Thomas Weißschuh [Wed, 11 Jan 2023 18:13:50 +0000 (18:13 +0000)]
lib/sysfs: add function blkdev_is_removable
This exactly represents the kernels notion of a "removable" blockdevice.
That is, the device itself is persistent but can contain some sort of
medium that can be changed at runtime.
Specifically it is *not* hotpluggable.
Karel Zak [Tue, 24 Jan 2023 15:45:42 +0000 (16:45 +0100)]
Merge branch 'signal_safety' of https://github.com/crrodriguez/util-linux
* 'signal_safety' of https://github.com/crrodriguez/util-linux:
sulogin: only assign to variables written by signal handlers
hardlink: calling putchar is off-limits on a signal handler
hardlink: last_signal should be a volatile sig_atomic_t
pg: calling exit on signal handler is not allowed
lib:pager: fix signal safety issues
write: signal_received should be volatile qualified
flock: timeout_expired must be volatile qualified
last: should not use errx/warnx on signal handlers
su-common: bool is a distinct type in c2x
fsck: only assign to cancel_requested
fsck: use sig_atomic_t type fot signal handler global vars
Karel Zak [Mon, 23 Jan 2023 13:09:27 +0000 (14:09 +0100)]
Merge branch 'bcachefs-fuzz' of https://github.com/t-8ch/util-linux
* 'bcachefs-fuzz' of https://github.com/t-8ch/util-linux:
libblkid: bcachefs: limit maximum size of read superblock
libblkid: iso9660: don't warn on isonum mismatch
libblkid: bcachefs: fix member_field_end
libblkid: bcachefs: validate size of member field
libblkid: bcachefs: fix field type
libblkid: bcachefs: add reproducer for oss-fuzz 55282
libblkid: bcachefs: fix endless loop
Samanta Navarro [Fri, 20 Jan 2023 11:57:15 +0000 (11:57 +0000)]
fsck.cramfs: print correct error on 32-bit systems
On 32-bit systems the file length check does not handle files correctly
which are larger than 4 GB. Use an unsigned long long which is already
in place for blkdev_get_size.
ThomasKaiser [Sun, 22 Jan 2023 09:36:01 +0000 (10:36 +0100)]
Add HiSilicon's 0x48/0xd40 Cortex-A76 variant.
Both Kirin 980 and Kirin 810 feature big cores with 0x48/0xd40 ID though they are marketed/documented as Cortex-A76 by HiSilicon themselves. Kirin 980 shows a r1p0 stepping while newer 810 has r3p0.
ThomasKaiser [Sat, 21 Jan 2023 17:03:11 +0000 (18:03 +0100)]
Add 53/002:Samsung Exynos-m3
ID 002 for the Exynos-m3 core seems weird since Exynos-m1 had 001 but Samsung's Mongoose line differs only by stepping between M1 and M2
* Mongoose 1, for example [Exynos 8890](https://github.com/ThomasKaiser/sbc-bench/blob/master/results/Exynos-8890-highperformance.cpuinfo): Exynos-m1 / r1p1
* Mongoose 2, for example [Exynos 8895](https://github.com/ThomasKaiser/sbc-bench/blob/master/results/Exynos-8895.cpuinfo): Exynos-m1 / r4p0
* Mongoose 3, for example [Exynos 9810](https://github.com/ThomasKaiser/sbc-bench/blob/master/results/Exynos-9810.cpuinfo): Exynos-m3 / r1p0
Karel Zak [Fri, 20 Jan 2023 14:20:55 +0000 (15:20 +0100)]
lsfd: remove strcpy(), keep things based on sizeof()
* use safe xstrncpy() rather than strcpy()
* use sizeof() for static buffers rather than macro with real size
(this solution is more robust for future code changes)
* use xcalloc() to zeroize allocated memory
Samanta Navarro [Fri, 20 Jan 2023 11:57:15 +0000 (11:57 +0000)]
fsck.cramfs: print correct error on 32-bit systems
On 32-bit systems the file length check does not handle files correctly
which are larger than 4 GB. Use an unsigned long long which is already
in place for blkdev_get_size.
It's probably typo in the original path (too many brackets:-)). Let's
be explicit about the types used in the calculation. '1 << n' means
the left shift is on 'int', but we need '(t)1 << n'.
Karel Zak [Tue, 17 Jan 2023 15:15:11 +0000 (16:15 +0100)]
Merge branch 'blkdev' of https://github.com/ferivoz/util-linux
* 'blkdev' of https://github.com/ferivoz/util-linux:
lib/blkdev: use off_t for max values
lib/blkdev: set errno in more cases
lib/blkdev: handle interrupted read call
Samanta Navarro [Fri, 13 Jan 2023 11:51:52 +0000 (11:51 +0000)]
lib/blkdev: set errno in more cases
Applications evaluate errno in case of a -1 return value. Examples
within util-linux are fsck.cramfs and mkfs.minix. Set errno more often
to offer reasonable error codes to the application and user.
Samanta Navarro [Fri, 13 Jan 2023 11:50:40 +0000 (11:50 +0000)]
lib/blkdev: handle interrupted read call
The read call may fail if it is interrupted by a signal. Do not assume
that we have reached the end of the underlying block device because of
this. Use read_all which retries in these cases.
hardlink: calling putchar is off-limits on a signal handler
Using putchar or any stdio function in signal handlers will
lead to funny results as they cannot be async signal safe.
One can write to stdout instead.
Samanta Navarro [Fri, 13 Jan 2023 11:53:41 +0000 (11:53 +0000)]
mkswap: do not use uninitialized stack value
If blkdev_get_size fails, then size is not set. Exit with an error code
and indicate what went wrong instead of continuing with random stack
content.
Proof of Concept:
```
$ mkswap /dev/null
mkswap: warning: truncating swap area to 17179869180 KiB
mkswap: /dev/null: insecure permissions 0666, fix with: chmod 0600 /dev/null
mkswap: unable to assign device to libblkid probe
```
The first output line depends on stack content and sometimes does not
show up at all. Abort operation if argument is neither regular file nor
block device.
Karel Zak [Thu, 12 Jan 2023 14:33:09 +0000 (15:33 +0100)]
include: add missing license lines
The header files are usually based on code from lib/. This commit
copies relevant license headers from lib/ to include/ to keep
things consistent. The very generic things (e.g. MBR definitions)
are always public domain.
Fixes: https://github.com/util-linux/util-linux/issues/2010 Signed-off-by: Karel Zak <kzak@redhat.com>
Samanta Navarro [Wed, 11 Jan 2023 11:57:21 +0000 (11:57 +0000)]
login: never send signals to init
If the signal handler is triggered after a failed fork attempt, then
child_pid will be -1. This in turn leads to a positive test and a
subsequent call of kill(1, signal). If signal was SIGTERM, then there
will be also another kill(1, SIGHUP) call.
Test explicitly for a positive child_pid value to prevent these cases.
Chris Webb [Tue, 10 Jan 2023 13:58:58 +0000 (13:58 +0000)]
unshare: support --map-users=inner:outer:count as well as outer,inner,count
The argument ordering in /proc/self/[gu]id_map and mount -o X-mount.idmap
is consistently inner:outer:count but unshare --map-users and --map-groups
were originally defined to expect outer,inner,count instead.
To reduce confusion caused by this mismatch, support a colon-separated
inner:outer:count ordering as well as the original comma-separated
outer,inner,count argument format. Although we accept both formats, only
document the new one in the usage text and deprecate the old format in
the man page.
Fixes: https://github.com/util-linux/util-linux/issues/1988 Signed-off-by: Chris Webb <chris@arachsys.com>
projects / thirdparty / util-linux.git / log
