Sertonix authored 1 year ago and Sören Tempel committed 1 year ago

*.{pre,post}-{install,upgrade,deinstall} scripts are run at the root
directory.

main/busybox: run ntpd as ntp user and not root Inspired by https://filipsjanis.com/articles/linux-rootless-ntpd-capabilities.html

macmpi authored 1 year ago and Sören Tempel committed 1 year ago

ttyGS0 is created by USB gadget g_serial or g_cdc composite driver

Dominique Martinet authored 1 year ago and Sören Tempel committed 1 year ago

return status was not initialized properly and could sometimes incorrectly
report an error when skipping a mount point

Reported-by: Frederik Bosch
Link: !52406 (comment 348411)
Link: http://lists.busybox.net/pipermail/busybox/2023-October/090534.html

Rene de Souza Pinto authored 1 year ago and Sören Tempel committed 1 year ago

Add to securetty file the serial consoles related to the following devices:

- NVIDIA Jetson Xavier NX developer kit
- Lenovo ThinkEdge SE70
- phyBOARD-Pollux board (i.MX8M Plus SoC)
- Advantech EPC-R3720 (i.MX8M Plus SoC)
- Rockchip RK3399 developer kit

Signed-off-by: Renê de Souza Pinto <rene@renesp.com.br>

Sören Tempel authored 1 year ago and Sören Tempel committed 1 year ago

Follow up for 743f6cf1

See https://git.busybox.net/busybox/commit/?id=e265c8d4c039729f2a68f3b1fb589c13c38d86f8

Dominique Martinet authored 1 year ago and Sören Tempel committed 1 year ago

find with -depth (implied by -delete) ignores -xdev, which has been a
long standing bug.
This bite us in scripts' cleanup, so would appreciate getting it fixed
as there is no easy workaround... (for example, something like this,
ignoring errors:
  unshare -m sh -c 'mount --bind "$1" /mnt && rm -rf /mnt' -- "$dir" && rmdir "$dir"
)

Patch was sent upstream as well, but no reply yet.

Link: https://bugs.busybox.net/show_bug.cgi?id=5756
Link: http://lists.busybox.net/pipermail/busybox/2023-September/090479.html

Fixes #15292

The read_line function may otherwise access uninitialized memory,
the patch is very minor so we might as well backport this patch.

ref: !49802 (comment 328857)
ref: https://lists.buildroot.org/pipermail/buildroot/2023-March/665882.html

The ash fixes backported in 5edf6355
has been assigned a CVE. Retroactively update the secfixes to reflect
that it has been fixed.

Celeste authored 1 year ago and Kevin Daudt committed 1 year ago

with comments showing how to disable DHCP overwriting resolv.conf

Closes #15198

Sören Tempel authored 1 year ago and Sören Tempel committed 1 year ago

Handy feature (especially in combination with the already enabled
CONFIG_FEATURE_DD_SIGNAL_HANDLING=y) with minimal impact on binary size.

Fixes #15159

The builtin has several known issues and has therefore been
disabled upstream recently, let's do the same downstream. See:

* https://git.busybox.net/busybox/commit/?id=5e0411a7fb510b9aecda0a850c76bdd62c50efa4
* https://bugs.busybox.net/show_bug.cgi?id=15619

Instead, rely on the sleep BusyBox applet (CONFIG_SLEEP)

Sören Tempel authored 1 year ago and Sören Tempel committed 1 year ago

See:

* https://bugs.busybox.net/show_bug.cgi?id=15601
* https://bugs.busybox.net/show_bug.cgi?id=15216

Sören Tempel authored 2 years ago and Sören Tempel committed 1 year ago

While ed(1) is mandated by POSIX BusyBox's ed implementation
is nowhere near POSIX compatible. Furthermore, we found some critical
bugs in the past which indicate that nobody is really actively using
this (e.g. #13504). Therefore, I suggest removing Busybox ed for now.
If you need an ed(1) implementation on your system use main/ed or
testing/oed instead. I believe it to be preferable to ship no ed
implementation at all by default instead of shipping a broken one.
As the current state of BusyBox ed will just be deeply confusing
to folks familiar with the ed editor.

There is a possibility then some packages now require an explicit
dependency on the ed package if they use ed as part of their build
process.

See also: alpine/aports#13504 (comment 213557)

the conditional is wrong, i.e. it was 'fixes' to run on cross only, not when !cross as intended
oops

can't execute a cross-compiled busybox without binfmt

fixes ash segfault with stty -echo among others

This is helpful for various tools which build and analyze Alpine-based images
to understand what files should be related to busybox, since they are not
recorded in the apkdb.

This is a ipv4ll client which may be used by tiny-cloud

Sören Tempel authored 2 years ago and Sören Tempel committed 2 years ago

See: http://lists.busybox.net/pipermail/busybox/2022-June/089769.html

Fixes !35149
Fixes #13789
Fixes #9923

Dominique Martinet authored 2 years ago and Sören Tempel committed 2 years ago

klogd depends on a syslog implementation for its output, but it
does not have to be 'syslog': syslog, rsyslog and syslog-ng all
provide a 'logger' virtual target which klogd can depend on.

(regardless of the fact that more sophisticated syslog implementations
can read from kernel logs directly and probably do not need klogd...)

alice authored 2 years ago and alice committed 2 years ago

- sets command_background=true and foreground arg for most init.d's
- s/SVCNAME/RC_SVCNAME for modern varname
- s|/var/run|/run| as /var/run is symlinked to /run, which is the linux rundir

This adds a patch to fix ash treating a resize event (SIGWINCH) as an
input of an empty line, see [upstream issue][1] for details.

[1]: https://bugs.busybox.net/show_bug.cgi?id=15256

the flag check for this seems incorrect- it triggers sha_ni use when
the instructions aren't available, but avx512vl/bw is - Elly on irc

causes failures in github actions, which have avx512 but no sha_no

Sören Tempel authored 2 years ago and Sören Tempel committed 2 years ago

Notable new features:

* HW acceleration for SHA1 and SHA256
	* Supported upstream for x86 and x86_64
	* Broken on x86 musl: http://lists.busybox.net/pipermail/busybox/2023-January/090078.html
* sleep(1) is an ash builtin now
* Custom helper script can now be supplied for udhcpc6
	* We still use /usr/share/udhcpc/default.script for both udhcpc and udhcpc6
	* It seems our default.script is not designed for IPv6 support
	* It may also be worthwhile to move udhcpc6 to -extra in the future
* Various bug fixes

New upstream applets:

* seedrng disabled in Alpine as it is provided by OpenRC already
* tsort (topological sort) disabled in Alpine
* tree (directory listings) enabled in Alpine

letting openrc background the service is preferred. also set RC_SVCNAME
vars for modern naming, and correct symlinked rundir

Hugo Barrera authored 2 years ago and alice committed 2 years ago

The cargo vars are now set in abuild.conf by default.

Newlines done with:

    find . -name APKBUILD -exec sed -i '/^$/N;/^\n$/D' {} \;

See also: https://stackoverflow.com/a/4522043

Patrycja Rosa authored 2 years ago and alice committed 2 years ago

AL32: unnecessary usage of braces