This commit shows a swaynag message when an unsupported GPU is detected
which must be acknowledged by users. It also adds an environment
variable (`SWAY_UNSUPPORTED_GPU`) which may be used instead of the
`--unsupported-gpu` argument.
The `static` storage class for flag variables in main has also been
removed, as this should have no effect on the program.
Resolves: #8999
A temporary SUID detection that would cause sway to exit was introduced
when SUID operation was deprecated, intended to avoid cases where a user
would not heed the deprecation notice, continued to use SUID and ended
up with sway accidentally running as root.
Remove the check, as the three years that have passed is sufficient time
for users to discover the deprecation and adapt. We did not care if
users intentionally want to run sway as root through SUID, we only
wanted to avoid surprise root.
This ensures these functions are always called (even when a library
such as wlroots or libc perform the fork) and removes the need to
manually call them.
In [1] we re-introduced a debug flag to enable wl_drm. Time has
passed and Xwayland + VA-API + amdvlk now all support
linux-dmabuf-v1.
[1]: https://github.com/swaywm/sway/pull/7916
The behavior of handlers registered with signal(3p) is not well-defined
for signals delivered more than once, as laid out in the man page.
We should replace our use of signal with sigaction, but for SIGCHLD
specifically we can also just skip the signals altogether by setting the
handler to SIG_IGN which causes child reaping to not be required.
Fixes: https://github.com/swaywm/sway/pull/8567
cmd_exec_process is used whenever sway is meant to execute a child
process on behalf of the user, and had a lot of complexity.
In order to avoid having to wait on the user's process, a double-fork
was used, which in turn required us to wait on the outer process. In
order to track the child PID for launcher purposes, a pipe was used to
transmit the PID back to sway.
This resulted in sway blocking for 5-6 ms per exec on my system, which
is quite significant. The error handling was also quite lacking - the
read loop did not handle errors at all for example.
Instead, teach sway to handle SIGCHLD and do away with the double-fork.
This in turn allows us to get rid of the pipe as we can record the
child's PID directly. This reduces the time we block to just 1.5 ms on
my system. We'd be able to get down to just 150 µs if we could use
posix_spawn(3), but posix_spawn(3) cannot reset NOFILE. clone(2) or
vfork(2) would be alternatives, but that presents portability issues.
This change is replicated for swaybar, swaybg and swaynag handling,
which had similar albeit less complicated implementations.
Some commands require outputs to be enabled. These commands are deferred
to allow outputs to be discovered, but the delayed modeset might only
run some time later.
Force a modeset to occur before running deferred commands.
Fixes: https://github.com/swaywm/sway/issues/8433
7e69a7076f ("Drop wl_drm") has dropped wl_drm, however a lot of
software wasn't quite ready for this (Xwayland, libva, amdvlk).
Keep wl_drm disabled by default to pressure the wl_drm phase-out,
but add a -Dlegacy-wl-drm flag for users to restore the previous
behavior in the meantime.
References: https://github.com/swaywm/sway/issues/7897
This semi-colon looks like a typo. Luckily, it has no effect on the code as it's treated as an empty statement leading the switch case.
Really straightforward nitpick change, was just something I was confused by when reading over the code.
This ensures that those surprised by the deprecation of SUID operation
receive an error rather than accidentally having sway run as root.
This detection will be removed in a future release.
Try to gain SCHED_RR (round-robin) realtime scheduling privileges before
starting the server. This requires CAP_SYS_NICE on Linux systems.
We additionally register a pthread_atfork callback which resets the
scheduling class back to SCHED_OTHER (the Linux system default).
Due to CAP_SYS_NICE, setting RLIMIT_RTPRIO has no effect on the process
as documented within man 7 sched (from Linux):
Privileged (CAP_SYS_NICE) threads ignore the RLIMIT_RTPRIO limit;
as with older kernels, they can make arbitrary changes to
scheduling policy and priority. See getrlimit(2) for further
information on RLIMIT_RTPRIO
Note that this requires the sway distribution packagers to set the
CAP_SYS_NICE capability on the sway binary.
Supersedes #6992
SUID privilege drop is needed for the "builtin"-backend of libseat,
which copied our old "direct" backend behavior for the sake of
compatibility and ease of transition.
libseat now has a better alternative in the form of seatd-launch. It
uses the normal seatd daemon and libseat backend and takes care of SUID
for us.
Add a soft deprecation warning to highlight our future intent of
removing this code. The deprecation cycle is needed to avoid surprises
when sway no longer drops privileges.
Proprietary drivers require --unsupported-gpu to be allowed, and IPCs
require no option to be passed.
The only way to satisfy both is to run IPCs before checking for
proprietary drivers.
Wayland compositors handle many file descriptors: client
connections, DMA-BUFs, sync_files, wl_data_device pipes, and so
on. Bump the limit to the max.
Closes: https://github.com/swaywm/sway/issues/6285
Nvidia has historically been a bad actor in the open-source graphics
ecosystem because they required a special EGLStreams code-path
instead of exposing the de-facto standard GBM API. However, with
their upcoming release they now support GBM as well.
This is a push in the right direction for Nvidia, so there's no
reason we should be more hostile to them than to any other proprietary
driver. Let's remove the --my-next-gpu-wont-be-nvidia flag, and advise
users to use --unsupported-gpu now.
Note, proprietary Nvidia drivers are still unsupported by the Sway
project (just like all other proprietary drivers).
Can be useful to make sure a bugfix is included. In the future maybe
the wlroots version string could include a commit hash when built
from source, too.
Instead of letting wlroots print messages to stdout, route debugging
messages into Sway's logging functions. This allows a more consistent
output (e.g. if Sway or wlroots changes its output style, they don't get
out-of-sync).
I also added a [wlr] prefix to wlroots messages, not yet sure it's a
good thing.
If the environment variable is not defined, getenv returns NULL.
Passing a NULL pointer to the "%s" format specifier is undefined
behavior. Even if some implementations output "(null)", an empty
string is nicer.
This makes it so invalid configs will return the exit code 1 when the
validation flag is given. This also reduces the log level to SWAY_ERROR,
which makes it so only the errors are shown. If someone wants more
verbose output, the can use the -V/--verbose or -d/--debug flags.
Additionally, this also makes it so swaybg will not be spawned when
validating the config.
It looks like the code to drop privileges may have been broken via
commit 37f0e1f. That commit reverted the correct order from #911, which
first drops the gid then the uid. If setuid is called first then the
target user may not have the ability to setgid.
This adds the logic to defer binding execution while sway is still
initializing. Without this, the binding command would be executed, but
the command handler would return CMD_DEFER, which was being treated as
a failure to run. To avoid partial executions, this will defer all
bindings while config->active is false.
Fixes memory leaks in the form of:
Direct leak of 24 byte(s) in 1 object(s) allocated from:
#0 0x7f5f7c2f4f30 in __interceptor_malloc (/usr/lib/x86_64-linux-gnu/libasan.so.5+0xedf30)
#1 0x563c799569f2 in ipc_recv_response ../common/ipc-client.c:94
#2 0x563c79957062 in ipc_single_command ../common/ipc-client.c:138
#3 0x563c798a56cc in run_as_ipc_client ../sway/main.c:127
#4 0x563c798a6a3a in main ../sway/main.c:349
#5 0x7f5f7b4d609a in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2409a)
Fixes memory leaks in the form of:
Direct leak of 20 byte(s) in 1 object(s) allocated from:
#0 0x7f5f7c2f4f30 in __interceptor_malloc (/usr/lib/x86_64-linux-gnu/libasan.so.5+0xedf30)
#1 0x563c7995b36a in join_args ../common/stringop.c:268
#2 0x563c798a6a1a in main ../sway/main.c:348
#3 0x7f5f7b4d609a in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2409a)
This aborts sway and displays an error message about XDG_RUNTIME_DIR
not being set without initializing the wl_display or logging any other
information.
