Udev workers consume typically 50-100MiB virtual memory.
On systems with lots of CPUs and relatively low memory, that may
easily cause workers to be OOM-killed.
This patch limits the number of workers to 8 per GiB memory.
But don't let the limit drop below the smallest value we had
without this patch (8 + 1 * 2 = 10); on small systems, udev's
memory footprint is likely lower.
Files which are installed as-is (any .service and other unit files, .conf
files, .policy files, etc), are left as is. My assumption is that SPDX
identifiers are not yet that well known, so it's better to retain the
extended header to avoid any doubt.
I also kept any copyright lines. We can probably remove them, but it'd nice to
obtain explicit acks from all involved authors before doing that.
This reworks the SELinux and SMACK label fixing calls in a number of
ways:
1. The two separate boolean arguments of these functions are converted
into a flags type LabelFixFlags.
2. The operations are now implemented based on O_PATH. This should
resolve TTOCTTOU races between determining the label for the file
system object and applying it, as it it allows to pin the object
while we are operating on it.
3. When changing a label fails we'll query the label previously set, and
if matches what we want to set anyway we'll suppress the error.
Also, all calls to label_fix() are now (void)ified, when we ignore the
return values.
Fixes: #8566
NPAR is a technology that allows a single network interface to
be divided into number of partitions. The partitions show up
as functions on the same PCI device... when there are more than
8 functions, ARI (alternative routing-ID interpretation) is
used. With ARI is enabled, the 8 bit field that normally has 5
bits for the PCI device and 3 bits for the PCI function is instead
interpreted as (implicit) device 0, with 8 bits for the function
number.
Because the linux kernel exposes the PCI device/function numbers
to userspace the same regardless of whether ARI is enabled,
systemd predictable device naming can generate unpredictable
names in this case, because network names using the PCI slot use
the function number, but not the device number, causing systemd
to generate the same name for mulitple network devices (so some
will revert to the "ethX" names).
With this patch, device naming code checks if ARI is enabled for
a PCI network device, and uses the full 8-bit function number
for naming to avoid this situation. This should improve
readability and predictability of device names.
Here is an example of how this change would affect naming:
before patch | after patch
-----------------------------
ens2f0 | ens2f0 NPAR partition 0 (in PCI slot 2)
ens2f1 | ens2f1 NPAR partition 1
...
ens2f7 | ens2f7 NPAR partition 7
eth1 | ens2f8 NPAR partition 8
eth2 | ens2f9 NPAR partition 9
With PCI SR-IOV, a number of virtual network devices can be enabled,
all of which share the same physical network device. Currently,
udev generates names for SR-IOV virtual functions as if they were
independent network devices.
With this change, the predictable network device naming code will
check if a network device is an SR-IOV virtual device, and will
generate a name based on the physical PCI device plus a "v%u"
suffix. This should improve readability and predictability of
device names.
Here is an example of how this change would affect naming:
before patch | after patch
-----------------------------
eno1 | eno1 onboard NIC, physical function
enp101s0f0 | eno1v0 onboard NIC, SR-IOV virtual func 0
enp101s0f1 | eno1v1 onboard NIC, SR-IOV virtual func 1
To generate predictable network device names, the code in
udev-builting-net_id.c tries to match the PCI device address
of the network device to the entries in /sys/bus/pci/slots.
However, sometimes the slot number is not associated the
network controller PCI device itself, but rather with one of
its parents.
This change will try to find a match in /sys/bus/pci/slots for
the parents of the PCI network device, if it doesn't find a
match for the device itself.
If a device exposes more than 16 mouse buttons, we run into the BTN_JOYSTICK
range, also labelling it as joystick. And since 774ff9b this results in only
ID_INPUT_JOYSTICK but no ID_INPUT_MOUSE.
Fixes#8460
This macro will read a pointer of any type, return it, and set the
pointer to NULL. This is useful as an explicit concept of passing
ownership of a memory area between pointers.
This takes inspiration from Rust:
https://doc.rust-lang.org/std/option/enum.Option.html#method.take
and was suggested by Alan Jenkins (@sourcejedi).
It drops ~160 lines of code from our codebase, which makes me like it.
Also, I think it clarifies passing of ownership, and thus helps
readability a bit (at least for the initiated who know the new macro)
Since 2016, Hyperv devices moved to using standard way to expose UUID to sysfs. Fix the parsing function to work with the newer format.
Change log:
v2: changed code to work with both old and new path format
v3: changed guid_str_len type to size_t, fixed length in char guid[] in handle_scsi_hyperv()
Support was killed in kernel 4.15 as well as ethtool 4.13.
Justification was lack of use by drivers and too much of a maintenance burden.
https://www.spinics.net/lists/netdev/msg443815.html
Also moved config_parse_warn_compat to conf-parser.[ch] to fix compile errors.
This was a bug inadvertently added by commit 73fc96c8ac.
The intent of the check is to "match slot address with device by
stripping the function" (as the comment above states it), for example
match network device PCI address 0000:05:00.0 (including a .0 for
function) to PCI slot address 0000:05:00, but changing that to a streq()
call prevented the match.
Change that to startswith(), which should both fix the bug and make the
intent of the check more clear and prevent unintentional bugs from being
introduced by future refactorings.
"noreturn" is reserved and can be used in other header files we include:
[ 16s] In file included from /usr/include/gcrypt.h:30:0,
[ 16s] from ../src/journal/journal-file.h:26,
[ 16s] from ../src/journal/journal-vacuum.c:31:
[ 16s] /usr/include/gpg-error.h:1544:46: error: expected ‘,’ or ‘;’ before ‘)’ token
[ 16s] void gpgrt_log_bug (const char *fmt, ...) GPGRT_ATTR_NR_PRINTF(1,2);
Here we include grcrypt.h (which in turns include gpg-error.h) *after* we
"noreturn" was defined in macro.h.
"--offset" takes an optional argument; if none is specified,
stroull() will attempt to parse a NULL pointer. For example:
$ udevadm test-builtin 'blkid --offset' /sys/dev/block/8:1
Update "--offset" to require an argument; also verify that the
offset is not negative.
gcc-8 throws an error if it knows snprintf might truncate output and the
return value is ignored:
../src/udev/udev-builtin-net_id.c: In function 'dev_pci_slot':
../src/udev/udev-builtin-net_id.c:297:47: error: '%s' directive output may be truncated writing up to 255 bytes into a region of size between 0 and 4095 [-Werror=format-truncation=]
snprintf(str, sizeof str, "%s/%s/address", slots, dent->d_name);
^~
../src/udev/udev-builtin-net_id.c:297:17: note: 'snprintf' output between 10 and 4360 bytes into a destination of size 4096
snprintf(str, sizeof str, "%s/%s/address", slots, dent->d_name);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: some warnings being treated as errors
Let's check all return values. This actually makes the code better, because there's
no point in trying to open a file when the name has been truncated, etc.
If a touchpad has MT axes only but not ABS_X/ABS_Y (DualShock 4 controller),
then we hit both the conditions is_touchpad and the later check for
!has_abs_axes here, assigning is_mouse and ID_INPUT_MOUSE later.
This is a bug, we historically only assigned either of of the pointing device
tags ID_INPUT_MOUSE/TOUCHPAD/JOYSTICK/TOUCHSCREEN, never multiple of them.
Note that we cannot just check for has_abs_axes and has_mt_coordinates because
the apple touch mouse has both. We really need to check if the device has
already been assigned something else.
https://bugs.freedesktop.org/show_bug.cgi?id=105050
This patch adds safe_atoux16 for parsing an unsigned hexadecimal 16bit int, and
uses that for parsing USB device and vendor IDs.
This fixes a compile error with gcc-8 because while we know that USB IDs are 2 bytes,
the compiler does not know that.
../src/udev/udev-builtin-hwdb.c:80:38: error: '%04X' directive output may be
truncated writing between 4 and 8 bytes into a region of size between 2 and 6
[-Werror=format-truncation=]
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Signed-off-by: Patrick Uiterwijk <puiterwijk@redhat.com>
Coverity now started warning about this ("Calling unlinkat without checking
return value (as is done elsewhere 12 out of 15 times).", and it is right:
most of the time we should at list print a log message so people can figure
out something is wrong when this happens.
v2:
- use warning level in journald too (this is unlikely to happen ever, so it
should be safe to something that is visible by default).
There are cases that we want to trigger and settle only specific
commands. For example, let's say at boot time we want to make sure all
the graphics devices are working correctly because it's critical for
booting, but not the USB subsystem (we'll trigger USB events later). So
we do:
udevadm trigger --action="add" --subsystem-match="graphics"
udevadm settle
However, we cannot block the kernel from emitting kernel events from
discovering USB devices. So if any of the USB kernel event was emitted
before the settle command, the settle command would still wait for the
entire queue to complete. And if the USB event takes a long time to be
processed, the system slows down.
The new `settle` option allows the `trigger` command to wait for only
the triggered events, and effectively solves this problem.
On Linux the former is a compat alias to the latter, and that's really
weird, as inside the kernel the two are distinct. Which means we really
should stay away from it.
log.h really should only include the bare minimum of other headers, as
it is really pulled into pretty much everything else and already in
itself one of the most basic pieces of code we have.
Let's hence drop inclusion of:
1. sd-id128.h because it's entirely unneeded in current log.h
2. errno.h, dito.
3. sys/signalfd.h which we can replace by a simple struct forward
declaration
4. process-util.h which was needed for getpid_cached() which we now hide
in a funciton log_emergency_level() instead, which nicely abstracts
the details away.
5. sys/socket.h which was needed for struct iovec, but a simple struct
forward declaration suffices for that too.
Ultimately this actually makes our source tree larger (since users of
the functionality above must now include it themselves, log.h won't do
that for them), but I think it helps to untangle our web of includes a
tiny bit.
(Background: I'd like to isolate the generic bits of src/basic/ enough
so that we can do a git submodule import into casync for it)
Including BitsPerSecond or Duplex values in .link files did not work when
set_slinksettings was called because the routine was not copying the base
parameters to the structure given to ioctl. As a result, EINVAL was always
reported, and no change occurred on the Ethernet device.
