The xsocket_type() function had an optional "family" argument
that was enabled only if IPv6 is enabled. In the case of the
function was called with a valid AF_UNIX argument, and IPv6 is
disabled, this argument was silently ignored.
This patch makes the "family" argument mandatory, while keeping
the old behavior i.e., if AF_UNSPEC is passed, we try first IPv6
(if it's enabled) and fallback to IPv4.
Also I changed all callers of xsocket_type() to reflect its new
interface.
Signed-off-by: Jonh Wendell <jonh.wendell@vexcorp.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
If multiple consoles are found from the sysfs file, cttyhack fails:
cttyhack: can't open '/dev/tty0 ttyS0': No such file or directory
In such cases take the last one as the kernel will use that one for
/dev/console.
Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
When built with "sendfile" support, httpd was unable to send large files
(>2 GB) in one single connection, terminating it before the full file
has been sent.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
While at it, optimized :s/find/repl/ a bit
function old new delta
text_hole_make 120 150 +30
colon 2848 2844 -4
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
I am using quilt to manage patches internally. Quilt creates dir .pc
and stores unmodified files there. Since I made change in one of
Config.src quilt made a copy in .pc/xxx.patch/xxx/Config.src. When
I run make it calls gen_build_files.sh and it generates
.pc/xxx.patch/xxx/Config.in. Now when I want to pop patch quilt
thinks I have made changes to original xxx/Config.in.
IMO the best solution is just to ignore hidden directories in
gen_build_files.sh. This also results in shorter build time in case
busybox is under git/svn versioning, since it avoids searching
many directories for Config.src.
Signed-off-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This allows to find mount points of 'char' devices such as UBI volumes
which otherwise fail for example with 'df' command:
/ # df
Filesystem 1K-blocks Used Available Use% Mounted on
/dev/root 72091984 12360980 56068924 18% /
devtmpfs 115236 4 115232 0% /dev
tmpfs 28672 32 28640 0% /tmp
/dev/ubi0_0 360268 18348 341920 5% /tmp/mnt/userfs
/ # df /dev/ubi0_0
Filesystem 1K-blocks Used Available Use% Mounted on
devtmpfs 115236 4 115232 0% /dev
Signed-off-by: Javier Viguera <javier.viguera@digi.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
First change fixes "read -n NUM". Apparently poll() won't report
data availability if cc[VMIN] > 1 until there are at least cc[VMIN] bytes.
function old new delta
read_line_input 3885 3877 -8
shell_builtin_read 1097 1087 -10
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-18) Total: -18 bytes
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
function old new delta
udhcp_str2optset 415 532 +117
Signed-off-by: Michael McTernan <Michael.McTernan.2001@cs.bris.ac.uk>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
In original implementation /proc/net/vlan/config check only prints warning
and does not exit. Busybox version exits if this file is not found and this
prevents kernel module autoloading to work.
I think it is safe to remove this check since ioctl() call will fail
(with proper
error code) anyway if 802.1q module is not loaded (or autoloading is off).
Signed-off-by: Paulius Zaleckas <paulius.zaleckas@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
