| Commit message (Collapse) | Author | Files | Lines |
|---|
|
Upstream limits the number of processes invoked with the '-P 0'
option to 100.
On Windows the limit was set to MAXIMUM_WAIT_OBJECTS, as that's
the most WaitForMultipleObjects() can handle. Apply the same
limit to user-supplied '-P' values. This avoids the need to loop
over blocks of MAXIMUM_WAIT_OBJECTS.
"64 processes ought to be enough for anyone."
Saves 64-80 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
The principal change is to replace four separate arrays (one for
pids and three for time values) with a single array of structs.
There are some other minor adjustments.
Saves 96-144 bytes.
(GitHub PR #609)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
time is unknown.
|
|
Windows aggressively reuses PIDs of exited processes. When a new irrelevant process gets assigned the former PID of the parent of an orphaned process, `pkill -P`, `pgrep -P`, `ps -o ppid` sees the new process as the parent of the orphaned process, and PPID of the orphan is no longer 1. So in theory, all `pkill -P`, `pgrep -P` calls are potentially dangerous and possible to kill an innocent process which outlives its parent.
This issue is already discussed in various Internet communities:
- https://stackoverflow.com/questions/6593003/how-can-i-reliably-check-whether-one-windows-process-is-the-parent-of-another-in
- https://web.archive.org/web/20140701081232/https://www.sapphiresteel.com/Blog/Killing-Trees-the-Windows-way
Comparing process creation times seems a feasible mitigation that should almost always work (though not theoretically sound).
The executable is added 512 bytes after the patch.
|
|
Replace a call to 'FindNextFileA()' in 'get_reparse_tag()'.
This has no effect on the standard builds. Builds with the UTF-8
manifest included save 4-8 bytes as this removes the only remaining
call to 'FindNextFileA()'.
(GitHub issue #506)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
In certain circumstances 'FindNextFileA()' can fail because the
legacy DOS 8.3 short name doesn't fit in the buffer supplied.
This can be avoided by using 'FindFirstFileExA()' instead, but
with a setting that isn't available on Windows XP.
- The problem doesn't arise in the standard 32 and 64 bit builds,
so no change is required.
- Builds with full Unicode support target platforms which have the
required setting. They can use 'FindFirstFileExA()' unconditionally,
at the cost of 36-56 bytes.
- Builds which include the UTF-8 manifest but which are allowed
to run on systems prior to Windows 10 need to check whether they
need to call 'FindFirstFileExA()' at runtime. This adds 100 bytes.
(GitHub issue #506)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Since 2022 the 'xxd' which comes with 'vim' has allowed the
combination of options '-p -c 0' as a way to request output on
a single line. Details:
https://github.com/vim/vim/pull/9524
Add this feature to the BusyBox 'xxd' applet.
Adds 64 bytes.
(GitHub issue #610)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
function old new delta
readfields 442 437 -5
printfields 462 457 -5
join_main 843 760 -83
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/3 up/down: 0/-93) Total: -93 bytes
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
function old new delta
join_main - 843 +843
printfields - 462 +462
readfields - 442 +442
.rodata 107120 107332 +212
packed_usage 36078 36237 +159
applet_names 2870 2875 +5
applet_main 1652 1656 +4
------------------------------------------------------------------------------
(add/remove: 4/0 grow/shrink: 4/0 up/down: 2127/0) Total: 2127 bytes
text data bss dec hex filename
1080964 555 4992 1086511 10942f busybox_old
1083091 555 4992 1088638 109c7e busybox_unstripped
v2: Add a missing return statement which broke the '-t' option.
Thanks to Roberto A. Foglietta for pointing this out.
Signed-off-by: Morgan Bartlett <mjmouse9999@gmail.com>
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
The 'make' applet failed to build with FEATURE_CLEAN_UP enabled.
Add an early declaration of freerules() to avoid this.
This doesn't affect the default builds.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
ifsfree() does not only release allocated ifsregion nodes; it also clears
the global IFS region state used by ifsbreakup(). If argstr() raises an
error while expanding an argument, ash longjmps out of expandarg() before
that cleanup runs, leaving stale IFS split offsets behind.
A later expansion can reuse the stack for a shorter string. ifsbreakup()
then sees the stale IFS state, trusts the old offsets, and can walk past
the current stack block before dereferencing p.
Follow dash's root-cause fix: when an expansion-related handler catches
EXERROR and continues, restore the handler and call ifsfree(). Apply
the cleanup to redirectsafe(), expandstr(), and evaltree().
Upstream commit:
Date: Mon Dec 5 23:02:01 2022 +0800
expand: Add ifsfree to expand to fix a logic error that causes a buffer over-read
On Mon, Jun 20, 2022 at 02:27:10PM -0400, Alex Gorinson wrote:
> Due to a logic error in the ifsbreakup function in expand.c if a
> heredoc and normal command is run one after the other by means of a
> semi-colon, when the second command drops into ifsbreakup the command
> will be evaluated with the ifslastp/ifsfirst struct that was set when
> the here doc was evaluated. This results in a buffer over-read that
> can leak the program's heap, stack, and arena addresses which can be
> used to beat ASLR.
>
> Steps to Reproduce:
> First bug:
> cmd args: ~/exampleDir/example> dash
> $ M='AAAAAAAAAAAAAAAAA' <note: 17 A's>
> $ q00(){
> $ <<000;echo
> $ ${D?$M$M$M$M$M$M} <note: 6 $M's>
> $ 000
> $ }
> $ q00 <note: After the q00 is typed in, the leak
> should be echo'd out; this works with ash, busybox ash, and dash and
> with all option args.>
>
> Patch:
> Adding the following to expand.c will fix both bugs in one go.
> (Thank you to Harald van Dijk and Michael Greenberg for doing the
> heavy lifting for this patch!)
> ==========================
> --- a/src/expand.c
> +++ b/src/expand.c
> @@ -859,6 +859,7 @@
> if (discard)
> return -1;
>
> +ifsfree();
> sh_error("Bad substitution");
> }
>
> @@ -1739,6 +1740,7 @@
> } else
> msg = umsg;
> }
> +ifsfree();
> sh_error("%.*s: %s%s", end - var - 1, var, msg, tail);
> }
> ==========================
Thanks for the report!
I think it's better to add the ifsfree() call to the exception
handling path as other sh_error calls may trigger this too.
function old new delta
restore_handler_expandarg - 33 +33
evaltree 725 711 -14
static.redirectsafe 141 124 -17
expandstr 262 242 -20
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 0/3 up/down: 36/-45) Total: -18 bytes
Signed-off-by: Sanghyun Park <sanghyun.park.cnu@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
It broke single-character splitting
|
|
function old new delta
awk_printf 1381 1387 +6
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
Add a freelines() function to free the line data associated with
an FDAT struct. This has no effect in the default build but saves
48-64 bytes if FEATURE_CLEAN_UP is enabled.
Change the type of the 'first' flag in readfields() to be consistent
with that in printfields().
The 'lines' array should be freed for each FDAT struct during clean
up, as it was before I broke it.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Fix combinatorial
field_split() now takes in LINE as a parameter, rather than just the
fields parameter.
Saves 16 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
The user supplies field numbers with the -o, -1 and -2 options.
It feels more natural to convert these to 0-based field indices
in the code.
Saves 32-48 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
When splitting the fields of a line, keep the string allocated
for the line and make the 'fields' array point into it rather
than allocating a new string for each field.
Saves 48 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Reduce the amount of duplicated code in readfields() by handling
a pushed back line inside the main loop, with a flag to indicate
the first pass through the loop.
Replace some memcpy() calls with structure assignments.
Fix a bug in the code to increase the size of the 'lines' array.
It set the capacity to either 8 or 0.
Saves 112-128 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Patch by mjmouse9999@gmail.com
When running svlogd -tt with a long line of input, the code assumes a buffer
of at least linemax+26 length, but that is two bytes longer than the
actual buffer with the default of linemax = 1000.
fixing that.
No code size change.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
Change input to getopt32 from int to char * so that
max_count is parsed as string. Use xatol_sfx to convert
string to long.
function old new delta
cmp_main 595 631 +36
.rodata 107121 107120 -1
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/1 up/down: 36/-1) Total: 35 bytes
Signed-off-by: Anubhav Kokane <dev.anubhavk@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
Update the 'rpm' applet so it also respects the BB_SYSTEMROOT
environment variable.
Replace chdir_system_drive() with xchdir_system_drive(), which
exits on failure to change directory.
Saves 32-36 bytes.
(GitHub PR #603)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
* dpkg: update prefix
The prefix now works properly with BB_SYSTEMROOT as it no longer
writes everything with a prefixed /
For example, with BB_SYSTEMROOT=C:/somepath, the /var/lib/dpkg directory
is now underneath that, rather than the root of that drive.
Also added handling so that the extracted files are also there.
* dpkg: prefix updates
For clarity, the .list files now always include leading slashes,
matching what happens in upstream.
There is then special handling under remove_file_array to turn them
back into relative paths. The handling assumes that files can be both,
since it does get called with some relative paths.
Adds 14-54 bytes.
|
|
function old new delta
awk_printf 1349 1381 +32
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
function old new delta
awk_printf 1086 1349 +263
.rodata 107108 107121 +13
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 276/0) Total: 276 bytes
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
awk_printf() got bigger, and now variables need better names,
even if this makes them longer.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
Based on a patch by Anubhav Kokane <dev.anubhavk@gmail.com>
function old new delta
awk_printf 640 1086 +446
.rodata 107131 107108 -23
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/1 up/down: 446/-23) Total: 423 bytes
text data bss dec hex filename
1080210 555 4992 1085757 10913d busybox_old
1080633 555 4992 1086180 1092e4 busybox_unstripped
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
|
|
Background jobs were being reported in shells other than an
interactive, top-level shell.
Suppress this unnecessary report.
Adds 16 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
This script:
#!/bin/bash
(
( true ) &
( true ) &
( true ) &
( true ) &
( true ) &
)
resulted in a crash. The problem was introduced by commit
7b692ddf0c (ash: improved support for jobs built-in). This
commit copies the job table into child shells for use by the
'jobs' built-in.
The crash happens because when the job table is cleared in the
child it becomes available for reuse. If only four jobs are
required this is OK but going over four causes the table to be
reallocated. This doesn't work because in the child it's in
shared memory.
The fix is not to pass the job table to the child unless it's
required by the 'jobs' built-in. This is also more efficient.
Adds 0-16 bytes.
(GitHub issue #604)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Adopt some BusyBox idioms in join_main(). In particular:
- The '-a' and '-v' options are mutually exclusive. Enforce this
in getopt32() instead of by hand. Also, since they're mutually
exclusive we don't need a distinct list for each.
- Control some clean-up code with ENABLE_FEATURE_CLEAN_UP.
There are also a few cosmetic changes.
Saves 204-256 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
'make bloatcheck' didn't work in the ARM build because it didn't
handle the different output formats of llvm's 'objdump' and 'nm'.
Allow for 'objdump -h' printing 5 rather than 7 fields. This at
least gets us some output. 'nm' always seems to print symbol
sizes of zero, though, which isn't very useful.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Make the corners of the symbolic bitmap icon look more rounded.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Reduce the colour depth of the bitmap icons from 24-bits colour/
8-bit alpha to 8-bit colour/1-bit alpha for aterm.ico and 1-bit
colour/1-bit alpha for sterm.ico.
This reduces the size of the standard builds by 21KB. The builds
targeting Windows 10/11 already used smaller PNG icons.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Recent changes to our mntent routines added support for Windows
volume names. If such a name is available use it in our sync(2)
implementation to flush file buffers. Failing that, use the drive
letter as before.
This handles edge cases where only one or other of the volume name
or drive letter is available.
Adds 32 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
One Windows-specific line of code wasn't protected by #ifdef.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
* Revert "find: reset stat(2) flags before terminating"
This reverts commit 228ee18fbe8cf408440e7bbe91db4ef201312983.
* test: fix flags for stat with NOFORK
test is a NOFORK applet, so the flags under mingw_stat may not have been
reset before calling it. We have it so it also restores the flags after,
since it could be called in the middle of another applet.
This also needed a way to retrieve the flags, so they are just done
through the return value.
* Add comments about not having to restore flags
Just for future reference
* lineedit: use new flag restoring
* win32: simplify mingw_access calling
Since we now have a way to get the previous flags, mingw_access can be
changed to ignore any stat flags
* stat: comments from PR
Adds 32 bytes.
|
|
Commit a717c6844 (find: skip check for execute when not needed)
didn't reset stat(2) to its default behaviour (at my insistence).
This was incorrect. The '-exec cmd {} +' action calls a nofork
'cmd' directly without spawning a new process. If 'cmd' calls
'stat(2)' it won't be in its default state which can affect the
result.
Reintroduce the code to reset 'stat(2)'.
Adds 16 bytes.
(GitHub PR #600)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
See previous commits:
1fb1a65064748a54fba4bf4b5e8700f8e61f003f
6a7ccb6ca3bd9b661760879478ecf0d5f6112fc3
My testing shows a speedup of around 30%, but that will vary with
whether files are cached.
Test inside busybox repository with files cached:
Command | before | after
find >/dev/null | 0.83s | 0.56s
find -executable >/dev/null | 1.56s | 1.35s
|
|
Update 'lsattr' to distinguish between mount points and junctions.
Use the new 'mntent' functions in 'df' to display mount points as
well as drive letters. If mount point is associated with a drive
letter this is displayed as the filesystem. The Windows-specific
'-w' option forces the volume name to be displayed instead, if
avaialable.
Adds 87-111 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Add code to distinguish between volume mount points and junctions.
Volume mount points are to be treated like directries while
junctions are like symlinks.
Update the 'mntent' functions to scan drive letters first, then
volume names and their associated mount points. By default the
filesystem is returned as a drive letter, if possible, though
the volume name is also returned.
The function 'find_mount_point()' has been updated to use the
new 'mntent' routines.
Adds 772-803 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
The shell builtin 'pwd' takes a Windows-specific '-a' option to
display the current directory on each drive.
Use Windows APIs directly to scan for valid drives instead of
'getmntent()'. Future changes will result in 'getmntent()'
returning mounted volumes as well as drives.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Commit 62e4c5d29 (win32: special treatment for virtual hard disk)
added special treatment for mount points associated with virtual
hard disks. One statement made there was incorrect: a VHD without
a drive letter containing no files *does* have a '.' directory.
The error message for this case was misleading.
The changes made to handle VHDs were insufficiently general:
- lstat(2) only resolved symlinks for the explicit path '.'.
It should also do so for paths of the form 'path/to/.'.
- opendir(3) only generated entries for '.' and '..' directories
when FindFirstFileA() failed for the path '.'. It should also
do so for any path which resolves to a volume mount point.
Make the necessary changes.
Adds 64-80 bytes.
(GitHub issue #597)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Virtual hard disks (VHD) without a drive letter gave incorrect
results in 'ls' and 'stat'. There were two problems:
- If the disk is empty it doesn't have a '.' directory.
- If it contains files the '.' directory is actually the mount
point, which looks like a symbolic link and can't be opened
as a directory.
Modify our 'stat(2)' and 'opendir(3)' implementations to handle
these quirks.
Adds 78-80 bytes.
(GitHub issue #597)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
If busybox.exe was started from a Windows command line in a
directory which was the target of a symlink (or equivalent) 'ls'
and 'stat' were unable to report the correct details of the current
directory.
Call 'chdir()' from the BusyBox main program to update our notion
of our current directory.
Adds 16 bytes.
(GitHub issue #597)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Running 'C:/Windows/System32/cmd.exe' from the shell was found to
have the unxpected result of creating the directory '.exe'. This
is due to the weirdness of cmd.exe.
A similar problem with batch files was already being handled in
`spawnveq()` by converting forward slashes to backslashes in the
path to the executable file. Doing this unconditionally will fix
the above case too.
(GitHub issue #598)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
The 'lsattr' applet identified three specific types of reparse
points: junctions, symbolic links and app exec links. Any other
type was just shown as a reparse point.
A junction is actually a specific type of mount point; a volume
mount is another. The description 'junction' has been changed to
the more generic 'mount point'. 'lsattr' doesn't distinguish
between volume mounts and junctions.
Change the help text to say 'Unidentified reparse point' for reparse
point which aren't a mount point, symlink or app exec link.
Adds 29 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
It was reported that a mounted volume looked like a broken symlink
and that changing to the mount directory failed.
Preserve the format of the volume name when resolving the link.
Adds 84-96 bytes.
(GitHub issue #597)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Commit bda604a70 (ash: prevent leakage of process handles) added
job tracking for process substitutions. This had the unwanted
side-effect that if the user tried to exit from the shell after
a command involving process substitution the shell reported that
background jobs were present and refused to exit.
Use the flag introduced in commit e6c716317 (ash: don't report
completion of process substitution) to avoid this.
Adds 16 bytes.
(GitHub issue #587)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
The current implementation of 'stty' for Windows only uses a
limited subset of the flags in 'struct termios'. Remove some
unused features.
Saves 16-40 bytes.
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
If echo has been disabled by the command 'stty -echo' the shell
'read' built-in should match its behaviour on Linux and not echo
keyboard input.
Adds 32-48 bytes.
(GitHub issue #594)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
|
Commit a560fdf23 (win32: better handling of console state) changed
to using CONIN$/CONOUT$ to handle terminal modes.
In tcgetattr() CONOUT$ was used instead of CONIN$, leading to
incorrect mode bits being fetched.
Saves 8-16 bytes.
(GitHub issue #594)
Signed-off-by: Ron Yorston <rmy@pobox.com>
|
