Current linux kernels output no more than 32 groups
in /proc/{pid}/status.
Plan is to increase this limit.
This patch allows ps to not core dump if the buffer used to read status
file was too small.
# ps aux
Signal 11 (SEGV) caught by ps (procps-ng version 3.3.3).
ps:display.c:59: please report this bug
Also increases the size of the buffer from 1024 to 4096, since even with
32 groups we are close to the limit.
cat /proc/12731/status | wc
39 128 961
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
pgrep used delimiter and delimeter, the correct spelling is delimiter.
Thanks to Alexander Kobel for the heads-up.
Signed-off-by: Craig Small <csmall@enc.com.au>
The 'refactor and enhance column width management'
recent redesign produced many subsequent benefits,
the latest of which is automatically sized fixed-width
non-scalable columns.
As expected, there was a cost associated with these
many enhancements. That cost has now been identified
as a 1-4% performance degradation, depending on which
fields are being displayed.
This increased cost arises principally from current
drawing related function calls, whereas top-3.3.3 did
most of its drawing via macros effectively inlining
those duties.
This commit inlines the equivalent drawing functions,
thus eliminating the function call penalty, and places
this top on a par with top-3.3.3. The trade off is a
modest additional 4k in executable size.
Signed-off-by: Jim Warner <james.warner@comcast.net>
The recent introduction of a column widths override
(the 'X' command) provided for a user input amount
to be added to default field size which ranged from
5 to 10 bytes.
While that approach could prevent truncated data, the
different default sizes would almost certainly mean
some precious screen real estate was waisted.
This commit introduces the concept of dynamic widths
where top will add only enough to a field default to
prevent truncation for that specific field.
Now users have a choice between their explicit width
override or a width chosen by top to exactly match
display needs. The former is immediate but likely
wastes some horizontal space while the latter is
iterative but will be sized precisely.
Original 'X' Command:
commit 384afa494a
commit 47e1d063ac
Extensions to 'X' Command:
commit bbf8e44fb4
commit 7557f3f754
Signed-off-by: Jim Warner <james.warner@comcast.net>
There were some gaps in the alternate navigation keys
top provided. Additionally, some inconsistencies
existed in the supporting key table.
This commit adds the following new key equivalents,
mirroring the standard vim navigation keys:
. ctrl+alt+ k = pgup, ctrl+alt+ j = pgdown
. ctrl+alt+ h = home, ctrl+alt+ l = end
Also, the supporting table entries now consistently
follow these "directions":
. up/pgup, down/pgdown, left/home, right/end
Signed-off-by: Jim Warner <james.warner@comcast.net>
To support the cursor navigation keys, after saving
the termios structure top issues 'smkx/keypad_xmit'
during startup. However, some terminals appear to
treat that directive as persistent which leaves a
corrupted tty state after top exit.
This commit reverses the above terminal directive
via 'rmkx/keypad_local' just prior to restoring the
saved termios structure at program end.
For discovering this bug, and providing the 'rmkx'
clue to its solution, thanks to:
Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Jim Warner <james.warner@comcast.net>
Bug-Debian: http://bugs.debian.org/688731
kill would not permit negative PIDs and thought they were options. kill
now explicitly checks for unknown options and if they are numeric
assumes they are negative PIDs. The first negative PID stops any
further option processing.
Signed-off-by: Craig Small <csmall@enc.com.au>
This 'Sleeping in function' field was made variable
width because the length of current kernel symbols
usually exceeded the former top's 9 character limit.
As a variable width field it would steal valuable
horizontal display positions from other, more likely,
displayed fields such as COMMAND or CGROUPS.
With the advent of the new 'X' toggle, no fixed-width
non-scalable field need suffer permanent truncation.
Thus, WCHAN is being made fixed width with a default
size of 10 characters.
Signed-off-by: Jim Warner <james.warner@comcast.net>
top/top.c | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
I have no idea what the maximum length of a terminal
name might be. However, the library provides for up
to 128 characters (ouch).
So just to be safe, this commit extends the ability
to widen columns to embrace this field.
Signed-off-by: Jim Warner <james.warner@comcast.net>
This commit accommodates those fields which may have
suffered truncation due to these default limits:
. 5 digits for uid/gid type fields
. 8 characters for user/group type fields
With a new interactive command, users can increase the
width of all such fields, or return to the defaults.
Note:
There are no restrictions on the amount added to
the defaults. The user is free to vastly exceed
screen limits which simply means such fields can
never be displayed.
Signed-off-by: Jim Warner <james.warner@comcast.net>
This commit affords user control over justification
for both column headings and the subordinate data.
Separate toggles are provided for control of numeric
data and string data.
Signed-off-by: Jim Warner <james.warner@comcast.net>
Now that column headings are independent of column
data format and require no carefully managed padding
bytes they are candidates for nls translation.
This commit migrates all column headings to the .pot
file with additional translator guidance in the form
of maximum sizes to avoid truncation.
It also places these new additions adjacent to their
associated descriptions, which were already present.
Signed-off-by: Jim Warner <james.warner@comcast.net>
This commit accomplishes the following objectives:
* remove extra task_show parm added with 'Locate'
* avoid column overflow with subsequent misalignment
* eliminate spaces for column heading padding
* decouple column headings from column data formats
* eliminate all hardcoded column format specifiers
* generalize the inter-column spacing management
* remove Fieldstab.desc in favor of direct nls access
* set the stage for nls support of column headings
* set the stage for dynamic changes to justification
Signed-off-by: Jim Warner <james.warner@comcast.net>
This flaw was revealed under 'man2htm' and dates back
to the first Gitorious revised top submission.
Reference:
commit fd62123562
Date: Thu Mar 31 22:15:12 2011 +1100
Signed-off-by: Jim Warner <james.warner@comcast.net>
When the maj_delta and min_delta fields were added to
the proc_t, they necessitated some compiler generated
padding bytes.
With this slight reordering, those padding bytes are
no longer generated. And since the original commit
already broke the library ABI, now is an opportune
time to correct that misalignment.
Reference:
commit 7753bd1004
Signed-off-by: Jim Warner <james.warner@comcast.net>
The UNIX and POSIX standards require that user and
group names be printed as decimal integers when there
is insufficient room. This has led to a constant
stream of bug reports.
With this commit, long names will be truncated and
displayed with a trailing visual clue.
To avoid truncation. the UNIX and POSIX way to change
column width is to rename the column:
ps -o pid,user=CumbersomeUserNames -o comm
The easy way is to directly specify the desired width:
ps -o pid,user:19,comm
Reference:
http://www.freelists.org/post/procps/rhbz737215-ps-does-not-resolve-some-user-names
Signed-off-by: Jim Warner <james.warner@comcast.net>
Both options provide more information about a process using -X and -XX
flags. The data comes from /proc/PID/smaps so it may vary.
Signed-off-by: Craig Small <csmall@enc.com.au>
We got a bug report, that our project doesn't spell "SELinux"
consistently/correctly. I've fixed that and the patch is attached.
Signed-off-by: Craig Small <csmall@enc.com.au>
The TREE_RESCANS #define (formerly TREE_ONEPASS) has
been eliminated and the approach to forest view mode
redesigned. The chance of dangling children has been
eliminated and overhead reduced.
We now order processes on start_time (non-display)
and are therefore immune to any pid, ppid or tgid
anomalies when pid values wrap.
The new algorithm also accommodates any distortions
caused by the 3.3 kernel 'hidepid' provisions --
something guaranteed to produce dangling children
under the former approach.
Related References:
commit a2086dfdf6
commit cd608f462e
commit 41ed28aa5d
Signed-off-by: Jim Warner <james.warner@comcast.net>
The recent introduction of scrollable variable width
columns makes a process 'environment' a potentially
useful addition to top's displayable fields.
This commit exploits the following new library flag:
PROC_EDITENVRCVT
Signed-off-by: Jim Warner <james.warner@comcast.net>
In preparation for top scrollable environment display,
the new flag PROC_EDITENVRCVT was added to mirror the
existing single vector string handling for cgroup and
cmdline.
Signed-off-by: Jim Warner <james.warner@comcast.net>
In an effort to avoid dangling children when in forest
view mode, top defaulted to a complete rescan of every
proc_t for each child encountered.
That expense was never really cost justified and now
with the 3.3 kernel 'hidepid' provisions it no longer
can offer such protection.
With this commit, the TREE_ONEPASS define is changed
to TREE_RESCANS so as to reverse the default scan
behavior.
Signed-off-by: Jim Warner <james.warner@comcast.net>
This commit represents mostly spelling corrections
in comments. It also includes a few very minor logic
changes/relocations.
Signed-off-by: Jim Warner <james.warner@comcast.net>
This section purported to list fields in alphabetical
order, but this was not always true.
With this commit, strict ascii collating sequence is
now observed.
Signed-off-by: Jim Warner <james.warner@comcast.net>
With the introduction of intra-column scrolling, the
scroll coordinates message was enhanced to give some
hint of positioning within a scrolled column.
Rather than rebuild this somewhat costly string from
scratch with each frame, we'll now do the bulk of the
work only when column headers are constructed.
The only remaining per frame costs will then be the
addition of a few terminfo escapes and the current
Frame_maxtask count.
Signed-off-by: Jim Warner <james.warner@comcast.net>
This commit introduces horizontal scrolling within any
variable width column. Thus, an entire command line,
complete list of control groups, etc. can now be
viewed -- not just a screen width's portion.
It is activated when any variable width column:
. is (via field selection) or
. has become (via the right arrow key)
the only displayed field.
Then, the right and left arrow keys can be used in the
normal way to continue scrolling within that column.
The amount scrolled with each key press is currently
set as the normal tab stop increment of 8 characters.
Signed-off-by: Jim Warner <james.warner@comcast.net>
The control group hierarchies for any particular task
could conceivably grow quite large. However, the
library might impose an arbitrary limit of 1024 bytes
via fill_cgroup_cvt.
Two utility buffers of 128 KiB each were already
available for command line use. This commit simply
trades the smaller 1024 byte stack based buffers for
those much larger existing ones. Thus, truncation
can be avoided with no additional run-time costs.
Signed-off-by: Jim Warner <james.warner@comcast.net>
Some inconsistencies have emerged during development
of support for these relatively new proc_t fields.
For example, a PROC_FILLCGROUP flag (via file2strvec)
could return NULL in cgroup whereas PROC_EDITCGRPCVT
(via fill_cgroup_cvt) *almost* guaranteed a return
address (as is true for PROC_EDITCMDLCVT and cmdline).
But even PROC_EDITCGRPCVT could return NULL if the
kernel version was less than 2.6.24. Then with NULL
ps would display a "-" while top would show "n/a".
And while unlikely, with the PROC_FILLSTATUS flag (via
status2proc) a NULL supgid address was theoretically
possible and both ps and top would then show "n/a".
This commit standardizes the following usage:
. PROC_FILLSTATUS (via status2proc)
guarantees a valid supgid address
representing either a true comma
delimited list or "-"
. PROC_FILLCGROUP plus
PROC_EDITCGRPCVT (via fill_cgroup_cvt)
guarantees a cgroup single vector
representing either a true control
group hierarchy or "-"
And as was true before, the following remains true:
PROC_FILLCOM or
PROC_FILLARG (via file2strvec)
may return a NULL cmdline pointer
. PROC_FILLCGROUP (via file2strvec)
may return a NULL cgroup pointer
. PROC_FILLCOM or
PROC_FILLARG plus
PROC_EDITCMDLCVT (via fill_cmdline_cvt)
guarantees a cmdline single vector
representing either a true command
line or a bracketed program name
. PROC_FILLSTATUS plus
PROC_FILLSUPGRP (via supgrps_from_supgids)
guarantees a valid supgrp address
representing either a true comma
delimited list or "-"
Signed-off-by: Jim Warner <james.warner@comcast.net>
This commit restores the missing space between command
lines and the environment when the later is being
displayed. Below is a brief history of that elusive
character.
commit bb4f08ba29
Date: Thu Aug 11 07:42:14 2011 +1000
The ps program was altered for improved args/comm
compliance. At this time, the needed space was
present due to a buglet in the new library
read_unvectored function used by fill_cmdline_cvt.
commit a5881b5a4e
Date: Thu Dec 8 10:19:38 2011 -0600
The trailing space was eliminated so that the
file2strvec and fill_cmdline_cvt returned
command lines contained no trailing space.
However, this created a buglet when control group
hierarchies were displayed and the final cgroup
was empty.
This is also where the undetected ps buglet was
created.
commit c3a1239efe
Date: Sun Dec 11 12:00:50 2011 -0600
The control group anomaly was fixed but the impact
on ps args/environ was still not detected.
Signed-off-by: Jim Warner <james.warner@comcast.net>
'si' and 'so' values depend on the result of the unitConvert
function where the output is a fixed-point size of kb_per_page
after the conversion. It gives 4 for kB units and 0 for MB units.
This also causes problems when switching between 'K' and 'k'
since the output value is 4 in both cases and the result for
'k' and 'K' then doesn't differ ... I swapped the conversion with
multiplication in order to make the number higher so it doesn't
lose precision. Since the unitConvert now accepts long instead
of int, I had to change the input type from int to long.
Signed-off-by: Craig Small <csmall@enc.com.au>
