procps/proc/diskstats.c

1029 lines
32 KiB
C
Raw Normal View History

/*
* diskstat - Disk statistics - part of procps
*
* Copyright (C) 2003 Fabian Frederick
* Copyright (C) 2003 Albert Cahalan
2021-01-21 12:07:48 +05:30
* Copyright (C) 2015 Craig Small <csmall@dropbear.xyz>
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
* Copyright (C) 2016 Jim Warner <james.warner@comcast.net>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#include <errno.h>
#include <fcntl.h>
#include <dirent.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#include <time.h>
#include <unistd.h>
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#include <sys/stat.h>
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#include <sys/types.h>
#include <proc/procps-private.h>
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#include <proc/diskstats.h>
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/* The following define will cause the 'node_add' function to maintain our |
nodes list in ascending alphabetical order which could be used to avoid |
a sort on name. Without it, we default to a 'pull-up' stack at slightly |
more effort than a simple 'push-down' list to duplicate prior behavior. | */
//#define ALPHABETIC_NODES
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#define DISKSTATS_LINE_LEN 1024
#define DISKSTATS_NAME_LEN 34
#define DISKSTATS_FILE "/proc/diskstats"
#define SYSBLOCK_DIR "/sys/block"
#define STACKS_INCR 64 // amount reap stack allocations grow
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#define STR_COMPARE strverscmp
/* ----------------------------------------------------------------------- +
this provision can help ensure that our Item_table remains synchronized |
with the enumerators found in the associated header file. It's intended |
to only be used locally (& temporarily) at some point before a release! | */
// #define ITEMTABLE_DEBUG //--------------------------------------------- |
// ----------------------------------------------------------------------- +
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_data {
unsigned long reads;
unsigned long reads_merged;
unsigned long read_sectors;
unsigned long read_time;
unsigned long writes;
unsigned long writes_merged;
unsigned long write_sectors;
unsigned long write_time;
unsigned long io_inprogress;
unsigned long io_time;
unsigned long io_wtime;
};
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node {
char name[DISKSTATS_NAME_LEN+1];
int type;
int major;
int minor;
time_t stamped;
struct dev_data new;
struct dev_data old;
struct dev_node *next;
};
struct stacks_extent {
int ext_numstacks;
struct stacks_extent *next;
struct diskstats_stack **stacks;
};
struct ext_support {
int numitems; // includes 'logical_end' delimiter
enum diskstats_item *items; // includes 'logical_end' delimiter
struct stacks_extent *extents; // anchor for these extents
};
struct fetch_support {
struct diskstats_stack **anchor; // fetch consolidated extents
int n_alloc; // number of above pointers allocated
int n_inuse; // number of above pointers occupied
int n_alloc_save; // last known reap.stacks allocation
struct diskstats_reaped results; // count + stacks for return to caller
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
};
struct diskstats_info {
int refcount;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
FILE *diskstats_fp;
time_t old_stamp; // previous read seconds
time_t new_stamp; // current read seconds
struct dev_node *nodes; // dev nodes anchor
struct ext_support select_ext; // supports concurrent select/reap
struct ext_support fetch_ext; // supports concurrent select/reap
struct fetch_support fetch; // support for procps_diskstats_reap
struct diskstats_result get_this; // used by procps_diskstats_get
};
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
// ___ Results 'Set' Support ||||||||||||||||||||||||||||||||||||||||||||||||||
#define setNAME(e) set_diskstats_ ## e
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#define setDECL(e) static void setNAME(e) \
(struct diskstats_result *R, struct dev_node *N)
// regular assignment
#define DEV_set(e,t,x) setDECL(e) { R->result. t = N-> x; }
#define REG_set(e,t,x) setDECL(e) { R->result. t = N->new. x; }
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
// delta assignment
#define HST_set(e,t,x) setDECL(e) { R->result. t = ( N->new. x - N->old. x ); }
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
setDECL(noop) { (void)R; (void)N; }
library: minimize the use of 'cleanup_stacks' routines Some parts of our newlib implementation are the result of functions which have been propagated from module to module. In particular, those 'cleanup_stacks' routines are all similar & likely originated in the <pids> api. In that interface there was a need to free dynamically acquired memory before the result structure was reused to satisfy subsequent 'get', 'select' or 'reap' calls. This, in turn, led to a concept of 'dirty' stacks with the need to call one of two 'cleanup_stack' functions. None of the remaining interfaces deal with such memory yet they each had their own 'cleanup_stack' functions. Those functions were responsible for resetting each of the result unions to zero, excluding any 'noop' items. The bottom line is that for all interfaces, repetitive calls would require iterating through the stack(s) two separate times: once to 'cleanup' another to 'assign'. With this commit we will reduce iterations to just the 'assign' routine. A reset to zero will be accomplished in the 'extra' item set routine (which is the only one actually requiring any reset). All other items will be reinitialized automatically by a new current set value or upon reallocation when an items compliment changes. In the <pids> interface, any freeing of dynamic memory could have been accomplished by adding that 'freefunc' check to the 'assign' function. However, that requires an Item_table test with every item. Instead, we'll now satisfy such needs as the very first step in those set functions responsible for dynamically acquired memory. [ the <pids> api retains 2 'cleanup_stack' functions ] [ to accommodate stack(s) 'reset' & to serve 'unref' ] Lastly, all the 'itemize_stack' functions were tweaked by eliminating an unnecessary initialization of result unions. That objective was already accomplished by the calloc() in a 'stacks_alloc' function or the remaining 'cleanup_stack' routine found in the <pids> interface. Signed-off-by: Jim Warner <james.warner@comcast.net>
2019-05-22 10:30:00 +05:30
setDECL(extra) { (void)N; R->result.ul_int = 0; }
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
DEV_set(NAME, str, name)
DEV_set(TYPE, s_int, type)
DEV_set(MAJOR, s_int, major)
DEV_set(MINOR, s_int, minor)
REG_set(READS, ul_int, reads)
REG_set(READS_MERGED, ul_int, reads_merged)
REG_set(READ_SECTORS, ul_int, read_sectors)
REG_set(READ_TIME, ul_int, read_time)
REG_set(WRITES, ul_int, writes)
REG_set(WRITES_MERGED, ul_int, writes_merged)
REG_set(WRITE_SECTORS, ul_int, write_sectors)
REG_set(WRITE_TIME, ul_int, write_time)
REG_set(IO_TIME, ul_int, io_time)
REG_set(WEIGHTED_TIME, ul_int, io_wtime)
REG_set(IO_INPROGRESS, s_int, io_inprogress)
HST_set(DELTA_READS, s_int, reads)
HST_set(DELTA_READS_MERGED, s_int, reads_merged)
HST_set(DELTA_READ_SECTORS, s_int, read_sectors)
HST_set(DELTA_READ_TIME, s_int, read_time)
HST_set(DELTA_WRITES, s_int, writes)
HST_set(DELTA_WRITES_MERGED, s_int, writes_merged)
HST_set(DELTA_WRITE_SECTORS, s_int, write_sectors)
HST_set(DELTA_WRITE_TIME, s_int, write_time)
HST_set(DELTA_IO_TIME, s_int, io_time)
HST_set(DELTA_WEIGHTED_TIME, s_int, io_wtime)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#undef setDECL
#undef DEV_set
#undef REG_set
#undef HST_set
// ___ Sorting Support ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
struct sort_parms {
int offset;
enum diskstats_sort_order order;
};
#define srtNAME(t) sort_diskstats_ ## t
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#define srtDECL(t) static int srtNAME(t) \
(const struct diskstats_stack **A, const struct diskstats_stack **B, struct sort_parms *P)
srtDECL(s_int) {
const struct diskstats_result *a = (*A)->head + P->offset; \
const struct diskstats_result *b = (*B)->head + P->offset; \
return P->order * (a->result.s_int - b->result.s_int);
}
srtDECL(ul_int) {
const struct diskstats_result *a = (*A)->head + P->offset; \
const struct diskstats_result *b = (*B)->head + P->offset; \
if ( a->result.ul_int > b->result.ul_int ) return P->order > 0 ? 1 : -1; \
if ( a->result.ul_int < b->result.ul_int ) return P->order > 0 ? -1 : 1; \
return 0;
}
srtDECL(str) {
const struct diskstats_result *a = (*A)->head + P->offset;
const struct diskstats_result *b = (*B)->head + P->offset;
return P->order * STR_COMPARE(a->result.str, b->result.str);
}
srtDECL(noop) { \
(void)A; (void)B; (void)P; \
return 0;
}
#undef srtDECL
// ___ Controlling Table ||||||||||||||||||||||||||||||||||||||||||||||||||||||
typedef void (*SET_t)(struct diskstats_result *, struct dev_node *);
#ifdef ITEMTABLE_DEBUG
#define RS(e) (SET_t)setNAME(e), DISKSTATS_ ## e, STRINGIFY(DISKSTATS_ ## e)
#else
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#define RS(e) (SET_t)setNAME(e)
#endif
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
typedef int (*QSR_t)(const void *, const void *, void *);
#define QS(t) (QSR_t)srtNAME(t)
#define TS(t) STRINGIFY(t)
#define TS_noop ""
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/*
* Need it be said?
* This table must be kept in the exact same order as
* those *enum diskstats_item* guys ! */
static struct {
SET_t setsfunc; // the actual result setting routine
#ifdef ITEMTABLE_DEBUG
int enumnumb; // enumerator (must match position!)
char *enum2str; // enumerator name as a char* string
#endif
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
QSR_t sortfunc; // sort cmp func for a specific type
char *type2str; // the result type as a string value
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
} Item_table[] = {
/* setsfunc sortfunc type2str
------------------------ ----------- ---------- */
{ RS(noop), QS(noop), TS_noop },
{ RS(extra), QS(ul_int), TS_noop },
{ RS(NAME), QS(str), TS(str) },
{ RS(TYPE), QS(s_int), TS(s_int) },
{ RS(MAJOR), QS(s_int), TS(s_int) },
{ RS(MINOR), QS(s_int), TS(s_int) },
{ RS(READS), QS(ul_int), TS(ul_int) },
{ RS(READS_MERGED), QS(ul_int), TS(ul_int) },
{ RS(READ_SECTORS), QS(ul_int), TS(ul_int) },
{ RS(READ_TIME), QS(ul_int), TS(ul_int) },
{ RS(WRITES), QS(ul_int), TS(ul_int) },
{ RS(WRITES_MERGED), QS(ul_int), TS(ul_int) },
{ RS(WRITE_SECTORS), QS(ul_int), TS(ul_int) },
{ RS(WRITE_TIME), QS(ul_int), TS(ul_int) },
{ RS(IO_TIME), QS(ul_int), TS(ul_int) },
{ RS(WEIGHTED_TIME), QS(ul_int), TS(ul_int) },
{ RS(IO_INPROGRESS), QS(s_int), TS(s_int) },
{ RS(DELTA_READS), QS(s_int), TS(s_int) },
{ RS(DELTA_READS_MERGED), QS(s_int), TS(s_int) },
{ RS(DELTA_READ_SECTORS), QS(s_int), TS(s_int) },
{ RS(DELTA_READ_TIME), QS(s_int), TS(s_int) },
{ RS(DELTA_WRITES), QS(s_int), TS(s_int) },
{ RS(DELTA_WRITES_MERGED), QS(s_int), TS(s_int) },
{ RS(DELTA_WRITE_SECTORS), QS(s_int), TS(s_int) },
{ RS(DELTA_WRITE_TIME), QS(s_int), TS(s_int) },
{ RS(DELTA_IO_TIME), QS(s_int), TS(s_int) },
{ RS(DELTA_WEIGHTED_TIME), QS(s_int), TS(s_int) },
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
};
/* please note,
* this enum MUST be 1 greater than the highest value of any enum */
enum diskstats_item DISKSTATS_logical_end = MAXTABLE(Item_table);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#undef setNAME
#undef srtNAME
#undef RS
#undef QS
// ___ Private Functions ||||||||||||||||||||||||||||||||||||||||||||||||||||||
// --- dev_node specific support ----------------------------------------------
static struct dev_node *node_add (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node *this)
{
struct dev_node *prev, *walk;
#ifdef ALPHABETIC_NODES
if (!info->nodes
|| (STR_COMPARE(this->name, info->nodes->name) < 0)) {
this->next = info->nodes;
info->nodes = this;
return this;
}
prev = info->nodes;
walk = info->nodes->next;
while (walk) {
if (STR_COMPARE(this->name, walk->name) < 0)
break;
prev = walk;
walk = walk->next;
}
prev->next = this;
this->next = walk;
#else
if (!info->nodes)
info->nodes = this;
else {
walk = info->nodes;
do {
prev = walk;
walk = walk->next;
} while (walk);
prev->next = this;
}
#endif
return this;
} // end: node_add
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static void node_classify (
struct dev_node *this)
{
DIR *dirp;
struct dirent *dent;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/* all disks start off as partitions. this function
checks /sys/block and changes a device found there
into a disk. if /sys/block cannot have the directory
read, all devices are then treated as disks. */
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
this->type = DISKSTATS_TYPE_PARTITION;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (!(dirp = opendir(SYSBLOCK_DIR))) {
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
this->type = DISKSTATS_TYPE_DISK;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return;
}
while ((dent = readdir(dirp))) {
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (strcmp(this->name, dent->d_name) == 0) {
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
this->type = DISKSTATS_TYPE_DISK;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
break;
}
}
closedir(dirp);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
} // end: node_classify
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static struct dev_node *node_cut (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node *this)
{
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node *node = info->nodes;
if (this) {
if (this == node) {
info->nodes = node->next;
return this;
}
do {
if (this == node->next) {
node->next = node->next->next;
return this;
}
node = node->next;
} while (node);
}
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
} // end: node_cut
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static struct dev_node *node_get (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
const char *name)
{
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node *node = info->nodes;
while (node) {
if (strcmp(name, node->name) == 0)
break;
node = node->next;
}
if (node) {
/* if this disk or partition has somehow gotten stale, we'll lose
it and then pretend it was never actually found ...
[ we test against both stamps in case a 'read' was avoided ] */
if (node->stamped != info->old_stamp
&& (node->stamped != info->new_stamp)) {
free(node_cut(info, node));
node = NULL;
}
}
return node;
} // end: node_get
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static int node_update (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node *source)
{
struct dev_node *target = node_get(info, source->name);
if (!target) {
if (!(target = malloc(sizeof(struct dev_node))))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 0;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
memcpy(target, source, sizeof(struct dev_node));
// let's not distort the deltas when a new node is created ...
memcpy(&target->old, &target->new, sizeof(struct dev_data));
node_classify(target);
node_add(info, target);
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
}
// remember history from last time around ...
memcpy(&source->old, &target->new, sizeof(struct dev_data));
// preserve some stuff from the existing node struct ...
source->type = target->type;
source->next = target->next;
// finally 'update' the existing node struct ...
memcpy(target, source, sizeof(struct dev_node));
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
} // end: node_update
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
// ___ Private Functions ||||||||||||||||||||||||||||||||||||||||||||||||||||||
// --- generalized support ----------------------------------------------------
static inline void diskstats_assign_results (
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct diskstats_stack *stack,
struct dev_node *node)
{
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct diskstats_result *this = stack->head;
for (;;) {
enum diskstats_item item = this->item;
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
if (item >= DISKSTATS_logical_end)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
break;
Item_table[item].setsfunc(this, node);
++this;
}
return;
} // end: diskstats_assign_results
static void diskstats_extents_free_all (
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct ext_support *this)
{
while (this->extents) {
struct stacks_extent *p = this->extents;
this->extents = this->extents->next;
free(p);
};
} // end: diskstats_extents_free_all
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static inline struct diskstats_result *diskstats_itemize_stack (
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct diskstats_result *p,
int depth,
enum diskstats_item *items)
{
struct diskstats_result *p_sav = p;
int i;
for (i = 0; i < depth; i++) {
p->item = items[i];
++p;
}
return p_sav;
} // end: diskstats_itemize_stack
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static inline int diskstats_items_check_failed (
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
enum diskstats_item *items,
int numitems)
{
int i;
/* if an enum is passed instead of an address of one or more enums, ol' gcc
* will silently convert it to an address (possibly NULL). only clang will
* offer any sort of warning like the following:
*
* warning: incompatible integer to pointer conversion passing 'int' to parameter of type 'enum diskstats_item *'
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
* my_stack = procps_diskstats_select(info, DISKSTATS_noop, num);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
* ^~~~~~~~~~~~~~~~
*/
if (numitems < 1
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
|| (void *)items < (void *)(unsigned long)(2 * DISKSTATS_logical_end))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
for (i = 0; i < numitems; i++) {
// a diskstats_item is currently unsigned, but we'll protect our future
if (items[i] < 0)
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 1;
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
if (items[i] >= DISKSTATS_logical_end)
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
}
return 0;
} // end: diskstats_items_check_failed
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/*
* diskstats_read_failed:
*
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
* @info: info structure created at procps_diskstats_new
*
* Read the data out of /proc/diskstats putting the information
* into the supplied info structure
*
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
* Returns: 0 on success, 1 on error
*/
static int diskstats_read_failed (
struct diskstats_info *info)
{
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static const char *fmtstr = "%d %d %" STRINGIFY(DISKSTATS_NAME_LEN) \
"s %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu %lu";
char buf[DISKSTATS_LINE_LEN];
struct dev_node node;
int rc;
if (!info->diskstats_fp
&& (!(info->diskstats_fp = fopen(DISKSTATS_FILE, "r"))))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (fseek(info->diskstats_fp, 0L, SEEK_SET) == -1)
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return 1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
info->old_stamp = info->new_stamp;
info->new_stamp = time(NULL);
while (fgets(buf, DISKSTATS_LINE_LEN, info->diskstats_fp)) {
// clear out the soon to be 'current'values
memset(&node, 0, sizeof(struct dev_node));
rc = sscanf(buf, fmtstr
, &node.major
, &node.minor
, &node.name[0]
, &node.new.reads
, &node.new.reads_merged
, &node.new.read_sectors
, &node.new.read_time
, &node.new.writes
, &node.new.writes_merged
, &node.new.write_sectors
, &node.new.write_time
, &node.new.io_inprogress
, &node.new.io_time
, &node.new.io_wtime);
if (rc != 14) {
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = ERANGE;
return 1;
}
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
node.stamped = info->new_stamp;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
if (!node_update(info, &node))
return 1; // here, errno was set to ENOMEM
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
}
return 0;
} // end: diskstats_read_failed
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/*
* diskstats_stacks_alloc():
*
* Allocate and initialize one or more stacks each of which is anchored in an
* associated context structure.
*
* All such stacks will have their result structures properly primed with
* 'items', while the result itself will be zeroed.
*
* Returns a stacks_extent struct anchoring the 'heads' of each new stack.
*/
static struct stacks_extent *diskstats_stacks_alloc (
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct ext_support *this,
int maxstacks)
{
struct stacks_extent *p_blob;
struct diskstats_stack **p_vect;
struct diskstats_stack *p_head;
size_t vect_size, head_size, list_size, blob_size;
void *v_head, *v_list;
int i;
vect_size = sizeof(void *) * maxstacks; // size of the addr vectors |
vect_size += sizeof(void *); // plus NULL addr delimiter |
head_size = sizeof(struct diskstats_stack); // size of that head struct |
list_size = sizeof(struct diskstats_result) * this->numitems; // any single results stack |
blob_size = sizeof(struct stacks_extent); // the extent anchor itself |
blob_size += vect_size; // plus room for addr vects |
blob_size += head_size * maxstacks; // plus room for head thing |
blob_size += list_size * maxstacks; // plus room for our stacks |
/* note: all of our memory is allocated in one single blob, facilitating some later free(). |
as a minimum, it's important that all of those result structs themselves always be |
contiguous within every stack since they will be accessed via a relative position. | */
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (NULL == (p_blob = calloc(1, blob_size)))
return NULL;
p_blob->next = this->extents; // push this extent onto... |
this->extents = p_blob; // ...some existing extents |
p_vect = (void *)p_blob + sizeof(struct stacks_extent); // prime our vector pointer |
p_blob->stacks = p_vect; // set actual vectors start |
v_head = (void *)p_vect + vect_size; // prime head pointer start |
v_list = v_head + (head_size * maxstacks); // prime our stacks pointer |
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
for (i = 0; i < maxstacks; i++) {
p_head = (struct diskstats_stack *)v_head;
p_head->head = diskstats_itemize_stack((struct diskstats_result *)v_list, this->numitems, this->items);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
p_blob->stacks[i] = p_head;
v_list += list_size;
v_head += head_size;
}
p_blob->ext_numstacks = maxstacks;
return p_blob;
} // end: diskstats_stacks_alloc
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static int diskstats_stacks_fetch (
struct diskstats_info *info)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
{
#define n_alloc info->fetch.n_alloc
#define n_inuse info->fetch.n_inuse
#define n_saved info->fetch.n_alloc_save
struct stacks_extent *ext;
struct dev_node *node;
// initialize stuff -----------------------------------
if (!info->fetch.anchor) {
if (!(info->fetch.anchor = calloc(sizeof(void *), STACKS_INCR)))
return -ENOMEM;
n_alloc = STACKS_INCR;
}
if (!info->fetch_ext.extents) {
if (!(ext = diskstats_stacks_alloc(&info->fetch_ext, n_alloc)))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return -1; // here, errno was set to ENOMEM
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
memcpy(info->fetch.anchor, ext->stacks, sizeof(void *) * n_alloc);
}
// iterate stuff --------------------------------------
n_inuse = 0;
node = info->nodes;
while (node) {
if (!(n_inuse < n_alloc)) {
n_alloc += STACKS_INCR;
if ((!(info->fetch.anchor = realloc(info->fetch.anchor, sizeof(void *) * n_alloc)))
|| (!(ext = diskstats_stacks_alloc(&info->fetch_ext, STACKS_INCR))))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return -1; // here, errno was set to ENOMEM
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
memcpy(info->fetch.anchor + n_inuse, ext->stacks, sizeof(void *) * STACKS_INCR);
}
diskstats_assign_results(info->fetch.anchor[n_inuse], node);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
++n_inuse;
node = node->next;
}
// finalize stuff -------------------------------------
/* note: we go to this trouble of maintaining a duplicate of the consolidated |
extent stacks addresses represented as our 'anchor' since these ptrs |
are exposed to a user (um, not that we don't trust 'em or anything). |
plus, we can NULL delimit these ptrs which we couldn't do otherwise. | */
if (n_saved < n_inuse + 1) {
n_saved = n_inuse + 1;
if (!(info->fetch.results.stacks = realloc(info->fetch.results.stacks, sizeof(void *) * n_saved)))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return -1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
}
memcpy(info->fetch.results.stacks, info->fetch.anchor, sizeof(void *) * n_inuse);
info->fetch.results.stacks[n_inuse] = NULL;
info->fetch.results.total = n_inuse;
return n_inuse;
#undef n_alloc
#undef n_inuse
#undef n_saved
} // end: diskstats_stacks_fetch
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
static int diskstats_stacks_reconfig_maybe (
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct ext_support *this,
enum diskstats_item *items,
int numitems)
{
if (diskstats_items_check_failed(items, numitems))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return -1;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/* is this the first time or have things changed since we were last called?
if so, gotta' redo all of our stacks stuff ... */
if (this->numitems != numitems + 1
|| memcmp(this->items, items, sizeof(enum diskstats_item) * numitems)) {
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
// allow for our DISKSTATS_logical_end
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (!(this->items = realloc(this->items, sizeof(enum diskstats_item) * (numitems + 1))))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return -1; // here, errno was set to ENOMEM
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
memcpy(this->items, items, sizeof(enum diskstats_item) * numitems);
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
this->items[numitems] = DISKSTATS_logical_end;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
this->numitems = numitems + 1;
if (this->extents)
diskstats_extents_free_all(this);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return 1;
}
return 0;
} // end: diskstats_stacks_reconfig_maybe
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
// ___ Public Functions |||||||||||||||||||||||||||||||||||||||||||||||||||||||
// --- standard required functions --------------------------------------------
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/*
* procps_diskstats_new():
*
* @info: location of returned new structure
*
* Returns: < 0 on failure, 0 on success along with
* a pointer to a new context struct
*/
PROCPS_EXPORT int procps_diskstats_new (
struct diskstats_info **info)
{
struct diskstats_info *p;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
#ifdef ITEMTABLE_DEBUG
int i, failed = 0;
for (i = 0; i < MAXTABLE(Item_table); i++) {
if (i != Item_table[i].enumnumb) {
fprintf(stderr, "%s: enum/table error: Item_table[%d] was %s, but its value is %d\n"
, __FILE__, i, Item_table[i].enum2str, Item_table[i].enumnumb);
failed = 1;
}
}
if (failed) _Exit(EXIT_FAILURE);
#endif
if (info == NULL || *info != NULL)
return -EINVAL;
if (!(p = calloc(1, sizeof(struct diskstats_info))))
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return -ENOMEM;
p->refcount = 1;
/* do a priming read here for the following potential benefits: |
1) ensure there will be no problems with subsequent access |
2) make delta results potentially useful, even if 1st time |
3) elimnate need for history distortions 1st time 'switch' | */
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
if (diskstats_read_failed(p)) {
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
procps_diskstats_unref(&p);
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return -errno;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
}
*info = p;
return 0;
} // end: procps_diskstats_new
PROCPS_EXPORT int procps_diskstats_ref (
struct diskstats_info *info)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
{
if (info == NULL)
return -EINVAL;
info->refcount++;
return info->refcount;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
} // end: procps_diskstats_ref
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
PROCPS_EXPORT int procps_diskstats_unref (
struct diskstats_info **info)
{
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node *node;
if (info == NULL || *info == NULL)
return -EINVAL;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
(*info)->refcount--;
if ((*info)->refcount < 1) {
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
int errno_sav = errno;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if ((*info)->diskstats_fp) {
fclose((*info)->diskstats_fp);
(*info)->diskstats_fp = NULL;
}
node = (*info)->nodes;
while (node) {
struct dev_node *p = node;
node = p->next;
free(p);
}
if ((*info)->select_ext.extents)
diskstats_extents_free_all((&(*info)->select_ext));
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if ((*info)->select_ext.items)
free((*info)->select_ext.items);
if ((*info)->fetch.anchor)
free((*info)->fetch.anchor);
if ((*info)->fetch.results.stacks)
free((*info)->fetch.results.stacks);
if ((*info)->fetch_ext.extents)
diskstats_extents_free_all(&(*info)->fetch_ext);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if ((*info)->fetch_ext.items)
free((*info)->fetch_ext.items);
free(*info);
*info = NULL;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = errno_sav;
return 0;
}
return (*info)->refcount;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
} // end: procps_diskstats_unref
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
// --- variable interface functions -------------------------------------------
PROCPS_EXPORT struct diskstats_result *procps_diskstats_get (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
const char *name,
enum diskstats_item item)
{
struct dev_node *node;
time_t cur_secs;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = EINVAL;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (info == NULL)
return NULL;
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
if (item < 0 || item >= DISKSTATS_logical_end)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = 0;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/* we will NOT read the diskstat file with every call - rather, we'll offer
a granularity of 1 second between reads ... */
cur_secs = time(NULL);
if (1 <= cur_secs - info->new_stamp) {
if (diskstats_read_failed(info))
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
}
info->get_this.item = item;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
// with 'get', we must NOT honor the usual 'noop' guarantee
info->get_this.result.ul_int = 0;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
if (!(node = node_get(info, name))) {
errno = ENXIO;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
}
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
Item_table[item].setsfunc(&info->get_this, node);
return &info->get_this;
} // end: procps_diskstats_get
/* procps_diskstats_reap():
*
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
* Harvest all the requested disks information providing
* the result stacks along with the total number of harvested.
*
* Returns: pointer to a diskstats_reaped struct on success, NULL on error.
*/
PROCPS_EXPORT struct diskstats_reaped *procps_diskstats_reap (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
enum diskstats_item *items,
int numitems)
{
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = EINVAL;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (info == NULL || items == NULL)
return NULL;
if (0 > diskstats_stacks_reconfig_maybe(&info->fetch_ext, items, numitems))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return NULL; // here, errno may be overridden with ENOMEM
errno = 0;
if (diskstats_read_failed(info))
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
if (0 > diskstats_stacks_fetch(info))
return NULL;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return &info->fetch.results;
} // end: procps_diskstats_reap
/* procps_diskstats_select():
*
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
* Obtain all the requested disk/partition information then return
* it in a single library provided results stack.
*
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
* Returns: pointer to a diskstats_stack struct on success, NULL on error.
*/
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
PROCPS_EXPORT struct diskstats_stack *procps_diskstats_select (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
const char *name,
enum diskstats_item *items,
int numitems)
{
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct dev_node *node;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = EINVAL;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (info == NULL || items == NULL)
return NULL;
if (0 > diskstats_stacks_reconfig_maybe(&info->select_ext, items, numitems))
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
return NULL; // here, errno may be overridden with ENOMEM
errno = 0;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (!info->select_ext.extents
&& (!diskstats_stacks_alloc(&info->select_ext, 1)))
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
if (diskstats_read_failed(info))
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
if (!(node = node_get(info, name))) {
errno = ENXIO;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
}
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
diskstats_assign_results(info->select_ext.extents->stacks[0], node);
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return info->select_ext.extents->stacks[0];
} // end: procps_diskstats_select
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
/*
* procps_diskstats_sort():
*
* Sort stacks anchored in the passed stack pointers array
* based on the designated sort enumerator and specified order.
*
* Returns those same addresses sorted.
*
* Note: all of the stacks must be homogeneous (of equal length and content).
*/
PROCPS_EXPORT struct diskstats_stack **procps_diskstats_sort (
struct diskstats_info *info,
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct diskstats_stack *stacks[],
int numstacked,
enum diskstats_item sortitem,
enum diskstats_sort_order order)
{
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
struct diskstats_result *p;
struct sort_parms parms;
int offset;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = EINVAL;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
if (info == NULL || stacks == NULL)
return NULL;
// a diskstats_item is currently unsigned, but we'll protect our future
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
if (sortitem < 0 || sortitem >= DISKSTATS_logical_end)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
library: removed all the 'PROCPS_' enumerator prefixes Many of our item enumerator identifiers are very long, especially in that <VMSTAT> module. Additionally, they all contain the exact same universal 'PROCPS_' prefix. The origins for this are likely found in the desire to avoid name clashes with other potential include files. But with procps-ng newlib, we've probably gone way too far. Did 'PROCPS_PIDS_TICS_SYSTEM' actually offer more protection against clash than 'PIDS_TICS_SYSTEM' does? I don't think so. Besides, no matter how big that name becomes, one can never guarantee they'll never be some clash. And, conversely, extremely short names will not always create conflict. Of course, in either case when some clash occurs, one can always #undef that problem. Thus, this commit will eliminate that 'PROCPS_' prefix making all of those enum identifiers a little shorter. And, we'll still be well above some ridiculously short (criminally short) names found in some common headers: - - - - - - - - - - <term.h> - 'tab', 'TTY', etc - - - - - - - - - - - - - - - - <search.h> - 'ENTER', ENTRY', 'FIND', etc ------------------------------------------------------ Finally, with this as a last of the wholesale changes, we will have established the naming conventions below: . only functions will begin with that 'procps_' prefix . exposed structures begin with the module/header name . item enumerators begin like structs, but capitalized . other enumerators work exactly like item enumerators . macros and constants begin just like the enumerators ------------------------------------------------------ Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-21 10:30:00 +05:30
if (order != DISKSTATS_SORT_ASCEND && order != DISKSTATS_SORT_DESCEND)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
if (numstacked < 2)
return stacks;
offset = 0;
p = stacks[0]->head;
for (;;) {
if (p->item == sortitem)
break;
++offset;
if (p->item >= DISKSTATS_logical_end)
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
return NULL;
++p;
}
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
errno = 0;
parms.offset = offset;
library: improve and/or standardize 'errno' management With older library logic having been modified to avoid using those potentially deadly alloc.h routines, while improving 'errno' handling, we're ready to standardize and enhance newlib's approach to any potential errors. In so doing, we'll establish the following objectives: . . . . . . . . . . . . . functions returning an 'int' . an error will be indicated by a negative number that is always the inverse of some well known errno.h value . . . . . . . . . . . functions returning an 'address' . any error will be indicated by a NULL return pointer with the actual reason found in the formal errno value And, when errno is manipulated directly we will strive to do so whenever possible within those routines which have been declared with PROCPS_EXPORT. In other words, in the user callable functions defined in source last. [ But, that won't always be possible. In particular, ] [ all the 'read_failed' functions will sometimes set ] [ 'errno' so that they can serve callers returning a ] [ NULL or an int without duplicating a lot of logic. ] [ Also, that includes one subordinate function which ] [ was called by 'read_failed' in the <slabinfo> API. ] ------------------------------------------------------ Along the way, several additional miscellaneous issues were addressed. They're listed here now for posterity. . the '-1' return value passed outside the library was eliminated since it would erroneously equate to -EPERM . the stacks_fetch functions in <diskstats> and <stat> weren't checked for their possible minus return values . hash create was not checked in <meminfo> or <vmstat> . fixed 'new' function faulty parm check in <slabinfo> Signed-off-by: Jim Warner <james.warner@comcast.net>
2017-11-18 10:30:00 +05:30
parms.order = order;
library: normalize/standardize an i/f, <DISKSTATS> api This patch will bring this interface up to our 3rd gen standards. The following summarizes the major changes. * New delta provisions have been added to most fields. There are, of course, some fields for which a delta is inappropriate. They include the identifying items such as name, type, major and minor. Plus the io_inprogress field which already acts, in effect, as a delta value. * To provide delta support, dev_node historical values have become persistent. By the same token, the library must provide for future removal of disks/partitions. A timestamp is used to detect 'stale' data which will be deleted so as not to satisfy some get, select or reap. * Such persistent support is provided by a linked list which, by default, grows from the bottom down so as to maintain compatibility with the /proc/diskstats order. Initially, I was tempted to use the GNU tsearch (tree) provisions until I discovered the overhead of building that tree plus costs of a subsequent 'twalk'. Besides, walking such a tree means retrieval order would differ from an order required/expected by the vmstat program. * The '/sys/block' directory is no longer scanned with every refresh cycle. Rather, it's only accessed when a node is first encountered. Then, that node's 'type' is persistent for its lifetime like several other fields. * A sort provision was included, at virtually no cost, even though such a provision was not currently needed. Signed-off-by: Jim Warner <james.warner@comcast.net>
2016-07-20 10:30:00 +05:30
qsort_r(stacks, numstacked, sizeof(void *), (QSR_t)Item_table[p->item].sortfunc, &parms);
return stacks;
} // end: procps_diskstats_sort
// --- special debugging function(s) ------------------------------------------
/*
* The following isn't part of the normal programming interface. Rather,
* it exists to validate result types referenced in application programs.
*
* It's used only when:
* 1) the 'XTRA_PROCPS_DEBUG' has been defined, or
* 2) an #include of 'xtra-procps-debug.h' is used
*/
PROCPS_EXPORT struct diskstats_result *xtra_diskstats_get (
struct diskstats_info *info,
const char *name,
enum diskstats_item actual_enum,
const char *typestr,
const char *file,
int lineno)
{
struct diskstats_result *r = procps_diskstats_get(info, name, actual_enum);
if (actual_enum < 0 || actual_enum >= DISKSTATS_logical_end) {
fprintf(stderr, "%s line %d: invalid item = %d, type = %s\n"
, file, lineno, actual_enum, typestr);
}
if (r) {
char *str = Item_table[r->item].type2str;
if (str[0]
&& (strcmp(typestr, str)))
fprintf(stderr, "%s line %d: was %s, expected %s\n", file, lineno, typestr, str);
}
return r;
} // end: xtra_diskstats_get_
PROCPS_EXPORT struct diskstats_result *xtra_diskstats_val (
int relative_enum,
const char *typestr,
const struct diskstats_stack *stack,
struct diskstats_info *info,
const char *file,
int lineno)
{
char *str;
int i;
for (i = 0; stack->head[i].item < DISKSTATS_logical_end; i++)
;
if (relative_enum < 0 || relative_enum >= i) {
fprintf(stderr, "%s line %d: invalid relative_enum = %d, valid range = 0-%d\n"
, file, lineno, relative_enum, i-1);
return NULL;
}
str = Item_table[stack->head[relative_enum].item].type2str;
if (str[0]
&& (strcmp(typestr, str))) {
fprintf(stderr, "%s line %d: was %s, expected %s\n", file, lineno, typestr, str);
}
return &stack->head[relative_enum];
(void)info;
} // end: xtra_diskstats_val