MinGW defines SIG_BLOCK, but doesn't have signal masks, so rather than
checking for SIG_BLOCK, add a configure check for sigprocmask.
v2:
Also add check to meson.build
It can take some time for Xorg to start. If Xorg runs as a systemd
service and other services are based on it, they have no way to
determine when Xorg is really ready to accept requests. Let's use
sd_notify() provided by libsystemd to signal systemd for readiness.
If Xorg has not been started as a systemd service, this won't do
anything.
Signed-off-by: Topi Miettinen <toiwoton@gmail.com>
Differences from autotools:
* Autotools defined NO_ALLOCA for OSX builds. I don't think we need
this anymore as Xalloc.h is no longer used anywhere in the xserver.
* X11.bin is linked with -u,miDCInitialize, and then libserver_mi
provided to satisfy (just) that. It's been that way since the commit
which added it. We can't write the equivalent in meson due to linker
argument ordering issues, but do we really need to?
* An explicit -Dsecure-rpc=false is required for OSX, since in meson we
don't do the checks that XTRANS_SECURE_RPC_FLAGS did for the existence
of the specific RPC functions required.
In the meson build, functions to make up for the shortcomings of libc
are compiled into a separate library. We don't bother making the pixman
headers available (reasonably enough) to this compilation, but they are
required indirectly by dix.h. Just remove this unneeded include.
GetTimeInMillis is called first, which sets clockid to
CLOCK_MONOTONIC_COARSE, which is typically much lower resolution than
the callers of GetTimeInMicros want.
Prior to a779fda224, GetTimeInMillis and
GetTimeInMicros did not share a clockid.
Restore the clockid split to fix the granularity of GetTimeInMicros.
Signed-off-by: Peter Harris <pharris@opentext.com>
This contortion made a bit more sense before we got SetNotifyFd and
friends, but now there's no need for it.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
CVE-2018-14665 also made it possible to exploit this to access
memory. With -logfile forbidden when running with elevated privileges
this is no longer an issue.
Signed-off-by: Matthieu Herrb <matthieu@herrb.eu>
Reviewed-by: Adam Jackson <ajax@redhat.com>
xdmcpSocket survives during the reset, there is no
need to create a new one.
This commit restores logic that was broken by
49c0f2413d in Xorg 1.19.
Signed-off-by: Alexander Volkov <a.volkov@rusbitech.ru>
These are so close to identical that most DDXes implement one in terms
of the other. All the relevant cases can be distinguished by the error
code, so merge the functions together to make things simpler.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>
This variable was no longer being read anywhere. MAXCLIENTS the macro is
the compile-time maximum limit, LIMITCLIENTS the macro is the default
limit, LimitClients the variable is the limit for the current server.
Signed-off-by: Adam Jackson <ajax@redhat.com>
The client ID is only needed for XRes, and autotools build ignores the
--clientids= arg if xres is disabled. We haven't made a meson option
for disabling tracking client ids (is it actually worth a build
option?), so just make this depend on xres.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Currently our meson.build just makes the assumption that the libc is
going to provide RPC functions. This doesn't actually seem to be the
case on Fedora, which causes compilation to fail unexpectedly:
../../Projects/xserver/os/rpcauth.c:47:10: fatal error: rpc/rpc.h: No such file or directory
#include <rpc/rpc.h>
^~~~~~~~~~~
compilation terminated.
So, in the event that we can't use libtirpc ensure that we actually
check whether or not the libc provides rpc/rpc.h. If it doesn't, raise
an error.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Lyude Paul <lyude@redhat.com>
If a driver calls AttendClient() from within a timer callback we
need to re-compute the local 'are_ready' to prevent the attended
client from waiting until WaitForSomething() times out.
This is a fix similar to commit 9ed5b263.
Signed-off-by: Damien Leone <dleone@nvidia.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
As we are not freeing elements while iterating the list of timers, we
can forgo using the safe variant, and reduce the number of pointer
dances required for the insertion sort.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Currently, we use xorg_list_add(new, head->prev) which is functionaly
equivalent to xorg_list_append(), but with more pointer chasing, so
reduce the strain on the reader and compiler by using the simpler
append().
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Currently we only check timer expiry if there are no client fd (or
other input) waiting to be serviced. This makes it very easy to starve
the timers with long request queues, and so miss critical timestamps.
The timer subsystem is just another input waiting to be serviced, so
evaluate it on every loop like all the others, at the cost of calling
GetTimeInMillis() slightly more frequently. (A more invasive and likely
OS specific alternative would be to move the timer wheel to the local
equivalent of timerfd, and treat it as an input fd to the event loop
exactly equivalent to all the others, and so also serviced on every
pass. The trade-off being that the kernel timer wheel is likely more
efficiently integrated with epoll, but individual updates to each timer
would then require syscalls.)
Reviewed-by: Peter Harris <pharris@opentext.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Querying a pkg-config variable using the --variable option produces the
value of the given variable as stored in the pkg-config file and should
not be used to add directories to the include search path.
The reason for this is that it breaks cross-compilation, because header
files are installed relative to the host sysroot. pkg-config supports a
PKG_CONFIG_SYSROOT_DIR environment variable that points to this sysroot
and will prepend that to the path of directories in -I or -L options in
pkg-config's Cflags, Libs or Libs.private keywords. However, because no
context can be inferred from variable names, as opposed to the keywords
with fixed meaning, the sysroot path will not be prepended to them. The
build system is responsible for doing so if necessary since it is aware
of the context in which the variable is used.
Adding the include directory returned by pkg-config to the include path
leaks build system information into the cross-build and break with very
confusing errors such as this:
In file included from include/misc.h:82:0,
from dix/atom.c:55:
/usr/include/pthread.h:682:6: warning: '__regparm__' attribute directive ignored [-Wattributes]
__cleanup_fct_attribute;
^~~~~~~~~~~~~~~~~~~~~~~
or this:
In file included from include/misc.h:139:0,
from dix/atom.c:55:
/usr/include/stdlib.h:133:8: error: '_Float128' is not supported on this target
extern _Float128 strtof128 (const char *__restrict __nptr,
^~~~~~~~~
Fix this by replacing the include directory with the appropriate xproto
dependency required to add the correct include directory to the compile
command for subdirectories that are missing the dependency. As detailed
above, this gives pkg-config the opportunity to prepend the sysroot for
all paths in -I compiler options.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
Trivial way to reproduce the bug:
$ Xorg -logfile /tmp/mylog -config /etc/xpra/xorg.conf -displayfd 2
The server then moans:
Failed to rename log file "/tmp/mylog" to "/tmp/mylog": No such file or directory
And the log file is created but immediately renamed to "/tmp/mylog.old".
This is caused by the changes to the log file handling introduced by
this commit:
https://cgit.freedesktop.org/xorg/xserver/commit/?id=edcb6426f20c3be5dd5f50b76a686754aef2f64e
To fix this, only rename the logfile if the log filename contains the
magic substitution string "%s".
Signed-off-by: Antoine Martin <antoine@nagafix.co.uk>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Ben Crocker <bcrocker@redhat.com>
Reviewed-by: Antoine Martin <antoine@nagafix.co.uk>
Tested-by: Ben Crocker <bcrocker@redhat.com>
Having different types of code all trying to check for elevated privileges
is a bad idea. This implementation is the most thorough one.
Signed-off-by: Nicolai Hähnle <nicolai.haehnle@amd.com>
Reviewed-by: Ben Crocker <bcrocker@redhat.com>
Reviewed-by: Antoine Martin <antoine@nagafix.co.uk>
Tested-by: Ben Crocker <bcrocker@redhat.com>
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>
Otherwise this is broken on cygwin:
rrlease.c: In function ‘ProcRRCreateLease’:
rrlease.c:305:9: error: implicit declaration of function ‘WriteFdToClient’ [-Werror=implicit-function-declaration]
if (WriteFdToClient(client, fd, TRUE) < 0) {
Signed-off-by: Adam Jackson <ajax@redhat.com>
When xorg_backtrace calls unw_get_proc_name and an error occurs, offset
might not be set for the current frame.
Initialize offset for each frame so that the offset from another frame
cannot be used inadvertently.
Signed-off-by: Jeff Smith <whydoubt@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
When a monotonic clock is not available, GetTimeInMicros() returns the
time in nanoseconds. Instead, return the time in microseconds, as the
name indicates.
Signed-off-by: Jeff Smith <whydoubt@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
The inputthread is kept locked all the time while X server's VT is not active.
If the X server is terminated while not active, it will be stuck forever in
InputThreadFini waiting for the thread to join, but it wouldn't because it is
locked.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=103782
Signed-off-by: Michal Srb <msrb@suse.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Building with strict-aliasing rightly chirps here:
../os/xdmcp.c: In function ‘XdmcpRegisterConnection’:
../os/xdmcp.c:489:31: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
&((struct sockaddr_in6 *) &address)->sin6_addr.s6_addr[12];
^~~~~~~~~~~~
We have "const char *address", so &address here is a char ** (i.e., it
points to the slot on the stack containing the pointer to the character
array passed in as an argument). Casting that to a struct sockaddr_in6 *
is wrong, because it means that area of the stack will be reinterpreted
as a struct sockaddr_in6.
Instead, cast address, not &address.
Signed-off-by: Adam Jackson <ajax@redhat.com>
x11perf -noop with 200 xlogos connected is slightly faster with ports:
before after Operation
---------- ----------------- --------------------
18400000.0 19200000.0 (1.04) X protocol NoOperation
Acked-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Peter Harris <pharris@opentext.com>
AIX's poll only allows FD_SETSIZE entries in the fd list, which is
insufficient for expanded MaxClients.
As a bonus, x11perf -noop with ~250 xlogos connected is slightly faster
with pollset:
before after Operation
--------- ---------------- --------------------
5750000.0 5990000.0 (1.04) X protocol NoOperation
Signed-off-by: Peter Harris <pharris@opentext.com>
Acked-by: Keith Packard <keithp@keithp.com>
SIGQUIT is a normal termination request, but any other signal we handle
here wants a core. This has the effect of making FatalError's call to
AbortServer trigger the
if (CoreDump)
OsAbort();
path. This will allow us to remove some DDX code that has the same net
effect.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Don't reuse cmd for strtok output to ensure the proper pointer is
freed afterwards.
The code incorrectly assumed the pointer returned by strtok(cmd, ":")
would always point to cmd. However, strtok(str, sep) != str if str
begins with sep. This caused an invalid-free crash when running
a program under X with a name beginning with a colon.
Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=104123
Signed-off-by: Tomasz Śniatowski <kailoran@gmail.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Terminate a dead session when -once was passed. Don't restart it.
Signed-off-by: Daniel Martin <consume.noise@gmail.com>
Reviewed-by: Walter Harms <wharms@bfs.de>
This was always wide enough to work on an fd_mask ("mask" ffs
presumably). We don't operate on fd_masks anymore, so this can go.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Roundhouse kick replacing the various (sizeof(foo)/sizeof(foo[0])) with
the ARRAY_SIZE macro from dix.h when possible. A semantic patch for
coccinelle has been used first. Additionally, a few macros have been
inlined as they had only one or two users.
Signed-off-by: Daniel Martin <consume.noise@gmail.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
A client can send a big request where the 32B "length" field has value
0. When the big request header is removed and the length corrected,
the value will underflow to 0xFFFFFFFF. Functions processing the
request later will think that the client sent much more data and may
touch memory beyond the receive buffer.
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
The meson build gives me:
../os/utils.c: In function ‘LockServer’:
../os/utils.c:310:40: warning: ‘snprintf’ output may be truncated before the last format character [-Wformat-truncation=]
snprintf(pid_str, sizeof(pid_str), "%10ld\n", (long) getpid());
^~~~~~~~~
../os/utils.c:310:5: note: ‘snprintf’ output between 12 and 13 bytes into a destination of size 12
snprintf(pid_str, sizeof(pid_str), "%10ld\n", (long) getpid());
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Which seems to be due to the %d part meaning that a negative number's -
sign would be one wider than we're expecting. Fine, just coerce it to
unsigned.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
The function itself does not depend on the macro. Move it outside
of the ifdef guard and remove the identical copy in XWIN.
This is step 1 towards removing the duplication in winauth.c and moving
the OS specifics to os/
Signed-off-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Jon Turney <jon.turney@dronecode.org.uk>
The epoll code depends on epoll_create1, not epoll_create.
Signed-off-by: Peter Harris <pharris@opentext.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Eric Anholt <eric@anholt.net>
This ensures that we don't use the now-closed file descriptor in the
future.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
oc->trans_conn is set to NULL when the connection is closed. At this
point, oc->fd is no longer valid and shouldn't be used. Move
dereference of oc->fd up into YieldControlNoInput where the state of
oc->trans_conn can be checked in a single place.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
In set_poll_client, check oc->trans_conn to make sure the connection
is still running before changing the ospoll configuration of the file
descriptor in case some other bit of the server is now using this file
descriptor.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
AbortClient performs most of the same operations as
CloseDownFileDescriptor except that it doesn't call ospoll_remove,
leaving that unaware that the file descriptor has been closed.
If the file descriptor is re-used before the server comes back around
to clean up, and that new file descriptor is passed to SetNotifyFd,
then that function will mistakenly re-interpret the stale ClientPtr
returned by ospoll_data as a struct notify * instead and mangle data
badly.
To fix this, the patch does:
1) Change CloseDownFileDescriptor so that it can be called multiple
times on the same OsCommPtr. The calls related to the file
descriptor are moved inside the check for trans_conn and
oc->trans_conn is set to NULL after cleaning up.
2) Move the XdmcpCloseDisplay call into CloseDownFileDescriptor. I
don't think the actually matters as we just need to know at some
point that the session client has exited. Moving it avoids the
possibility of having this accidentally trigger from another client
with the same fd which closes down at around the same time.
3) Change AbortClient to call CloseDownFileDescriptor. This makes sure
that all of the fd-related clean up happens in the same way
everywhere, in particular ensures that ospoll is notified about the
closed file descriptor at the time it is closed and not some time later.
Debian-bug: https://bugs.debian.org/862824
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
This infrastructure is no longer read, only written; the mapping
from fd to client is now handled by ospoll.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Whatever problem this is trying to fix, we don't care. Just include the
thing and stop worrying about whether _POSIX_SOURCE is defined.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
Tested-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Use conf_data outside of include/ to avoid re-running detection of the
same functions.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Peter Harris <pharris@opentext.com>
Include dix-config.h first to pick up _GNU_SOURCE so we get the
definition for sigset_t.
Reported-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Test to build xserver_poll.c was inverted compared to autoconf. Build
xserver_poll.c if poll is missing.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
This makes sure the server will go look at the client again, notice
that the FD is no longer valid and close the client down.
Bugzilla: https://bugs.freedesktop.org/100863
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-and-Tested-by: Michel Dänzer <michel.daenzer@amd.com>
There are three copies of the same short sequence of operations to
close down a client when a write error occurs. Create a new function,
AbortClient, which performs these operations and then call it from the
three places.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-and-Tested-by: Michel Dänzer <michel.daenzer@amd.com>
This is all just stuff we wish we had in libc, and some of this gets
used in eg. the dmx utilities build, so split it to its own library to
avoid pulling in xserver stuff.
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Adam Jackson <ajax@redhat.com>
This is a work in progress that builds Xvfb, Xephyr, Xwayland, Xnest,
and Xdmx so far. The outline of Xquartz/Xwin support is in tree, but
hasn't been built yet. The unit tests are also not done.
The intent is to build this as a complete replacement for the
autotools system, then eventually replace autotools. meson is faster
to generate the build, faster to run the bulid, shorter to write the
build files in, and less error-prone than autotools.
v2: Fix indentation nits, move version declaration to project(), use
existing meson_options for version-config.h's vendor name/web.
Signed-off-by: Eric Anholt <eric@anholt.net>
Acked-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
We mostly use #ifdef throughout the tree, and this lets the generated
config.h files just be #define TOKEN instead of #define TOKEN 1.
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
Without this, assertion failures can make life hard for users and those
trying to help them.
v2:
* Change commit log wording slightly to "can make life hard", since
apparently e.g. logind can alleviate that somewhat.
* Set default handler for SIGABRT in
hw/xfree86/common/xf86Init.c:InstallSignalHandlers() and
hw/xquartz/quartz.c:QuartzInitOutput() (Eric Anholt)
Reviewed-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
The clever pointer tricks were actually not working, and we were doing
the byte-by-byte moves in general. By just doing the memcpy and
obvious byte swap code, we end up generating actual byte swap
instructions, thanks to optimizing compilers.
text data bss dec hex filename
before: 2240807 51552 132016 2424375 24fe37 hw/xfree86/Xorg
after: 2215167 51552 132016 2398735 249a0f hw/xfree86/Xorg
Signed-off-by: Eric Anholt <eric@anholt.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
Most of this is a legacy of the old "extmod" design where you could load
_some_ extensions dynamically but only if the server had been built with
support for them in the first place.
Note that since we now only initialize the DPMS extension if at least
one screen supports it, we no longer need DPMSCapableFlag: if it would
be false, we would never read its value.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
timingsafe_memcmp.c:21:1: warning: no previous prototype for ‘timingsafe_memcmp’ [-Wmissing-prototypes]
timingsafe_memcmp(const void *b1, const void *b2, size_t len)
Signed-off-by: Adam Jackson <ajax@redhat.com>
- typo in #ifdef check
- also need to add AC_CHECK_FUNCS([arc4random_buf])
Reported-by Eric Engestrom. Thanks
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Matthieu Herrb <matthieu@herrb.eu>
And the current code for MitToId has a use-after-free() issue.
[Also remove the actual implementations - ajax]
Signed-off-by: Matthieu Herrb <matthieu@herrb.eu>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Provide the function definition for systems that don't have it.
Signed-off-by: Matthieu Herrb <matthieu@herrb.eu>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
The input thread should generate events, not send them. Make it easier to
find the instances where it's doing so.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Tested-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Olivier Fourdan <ofourdan@redhat.com>
Fixes a regression from
commit 41da295eb5
Author: Keith Packard <keithp@keithp.com>
Date: Sun Nov 3 13:12:40 2013 -0800
Trap SIGBUS to handle truncated shared memory segments
that causes the SIGBUS handler to fail to chain up correctly and
corrupts nearby memory instead.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
==8734== Thread 2 InputThread:
==8734== Conditional jump or move depends on uninitialised value(s)
==8734== at 0x2FDB05: InputThreadDoWork (inputthread.c:333)
==8734== by 0x6924423: start_thread (pthread_create.c:333)
==8734== by 0x6C229BE: clone (clone.S:105)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Remove leftover from commit e10ba9e, MAX_TIMES_PER is not used anymore.
Signed-off-by: Olivier Fourdan <ofourdan@redhat.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Solaris 11.3.5 introduced support for /proc/pid/cmdline, so try it
first, and if we can't open it, then fallback to /proc/pid/psinfo
as we did before.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
A client which is attended while a grab is blocking execution of its
requests needs to be placed in the saved_ready_clients list so that it
will get scheduled once the grab terminates. Otherwise, if the client
never sends another request, there is no way for it to be placed in
the ready_clients list.
v2: Wrap comment above mark_client_saved_ready.
Remove test for OS_COMM_IGNORED which will always be true.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99333
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Fixes a regression introduced in 0b2f30834b. If a driver posts input
events during a timer function (wacom and synaptics do this during tap
timeouts), ProcessInputEvents() is not called for these events. There are no
new events on any fds, so the events just sit in the queue waiting for
something else to happen.
Fix this by simply returning 0 from check_timers if we ran at least one of
them or reset them all. This way the callers ospoll_wait will exit and
continue with normal processing.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
The custom os/os.O library reuses *.o files of os/libos.la.
The current rule assumes automake puts all the objects into per-target
am__*_la_OBJECTS variable. At least with AC_REPLACE_FUNCS, this no
longer holds (as wanted objects are put into LTLIBOBJS instead).
Depend on automake's result, the *.la library instead, to express demand
of any its dependencies being built.
Should be fixing randomly occuring "undefined reference to `strlcpy'"
errors when linking Xvfb and other DDX-es that could use os.O.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Mihail Konev <k.mvc@ya.ru>
If a work proc wakes up a sleeping client and it is ready to execute,
we need to re-compute the local 'are_ready' value before deciding
what timeout value to use in WaitForSomething.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98030
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
On Linux, setting the main thread's name changes the program name
(/proc/self/comm). Setting it to MainThread breaks scripts that rely on
the command name, e.g. ps -C Xorg.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
When putting a device node into a poll-request list, do not overwrite a
"please-remove" element with the same fd, so that a closed device file
is ospoll_remove'd prior to being ospoll_add'ed.
Before, the opposite order was possible, resulting in ospoll_add
considering the newly opened file being already polled, should it have a
fd for which the "please-remove" has not been procesed yet. In this
case, no further events would be seen from the device.
Signed-off-by: Mihail Konev <k.mvc@ya.ru>
Regressed-in: 52d6a1e832
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97880
Patchwork: https://patchwork.freedesktop.org/patch/113763/
Hit-and-Reduced-by: Hans de Goede <hdegoede@redhat.com>
Reviewed-and-Reduced-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
When a client is marked as write blocked, clear any old 'write ready'
bit in the osfds structure so that a new indication of write ready
(which is marked as edge trigggered) will trigger the callback.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Tested-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Tested-by: Matthieu Herrb <matthieu@herrb.eu>
When a client with pending output is ready (has request data pending),
FlushAllOutput will skip it to get all of the requests processed
before sending any queued output. That means FlushAllOutput is going
to return with some output pending to a client which isn't known to be
write blocked. And that means NewOutputPending needs to be set so that
FlushAllOutput will get called again to actually go flush this client.
It might be interesting to try just flushing the client to send any
queued data along the way. This patch just restores the server
behavior to what it was before the ospoll changes.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Tested-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Tested-by: Matthieu Herrb <matthieu@herrb.eu>
Autoconf logic borrowed from glib
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Tested-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
If RTLD_DI_SETSIGNAL is set to let us turn runtime linker/loader errors
into catchable signals, then we should only show the errors when catching
that signal, instead of tossing out red herrings to distract people with
unrelated crashes long after their last failed symbol lookup (especially
when using drivers built to support multiple API's by checking which
symbols are available before calling them).
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Aaron Plattner <aplattner@nvidia.com>
O_CLOEXEC is not a file bit. It is not setable with F_SETFL. One must
use it when calling open(2). To set it cloexec on an existing fd,
F_SETFD and FD_CLOEXEC must be used.
This also fixes a build failure regression on configurations that don't
have O_CLOEXEC defined.
cf: http://pubs.opengroup.org/onlinepubs/9699919799/functions/fcntl.html
Regressed-in: 30ac756798
Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Reviewed-by: Julien Cristau <jcristau@debian.org>
This change has two effects:
1. Only calls FlushCallbacks when we're actually flushing data to a
client. The unnecessary FlushCallback calls could cause significant
performance degradation with compositing, which is significantly
reduced even without any driver changes.
2. By passing the ClientPtr to FlushCallbacks, drivers can completely
eliminate unnecessary flushing of GPU commands by keeping track of
whether we're flushing any XDamageNotify events to the client for
which the corresponding rendering commands haven't been flushed to
the GPU yet.
Reviewed-by: Adam Jackson <ajax@redha.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
If a file descriptor is added or removed from an ospoll callback, then
the arrays containing file descriptor information will have all of
their indices changed, so the loop state is no longer consistent. Just
bail out and let the caller come back around to try again.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Instead of freeing the struct ospollfd elements when the fd is
removed by the user, delay that until epoll is idle so that we are
sure no epoll_event structures could contain the stale pointer. This
handles cases where an fd is removed from the ospoll callback
interface, and also in case the OS keeps stale pointers around after
the call to epoll_ctl with EPOLL_CTL_DEL.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Calling InputThreadRegisterDev twice with the same fd should replace
the existing function and args instead of creating a new entry with
the same fd.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
xf86AddEnabledDevice() prepends the new fd to the list,
xf86RemoveEnabledDevice() then searches for a matching fd and removes that
entry. If this is done for the same fd (and since we lose all information but
the actual fd) we usually unregister virtual devices in reverse order, causing
a dereference of already released memory.
Case in point:
- the wacom driver calls xf86AddEnabledDevice() once for the physical device,
then multiple times for the virtual subdevices
- when the physical device is unplugged, the driver calls
xf86RemoveEnabledDevice() for the physical device
- all we have is the fd, so we end up removing the last virtual device from
the fd set
- xf86DeleteInput() frees the physical device's pInfo
- the fd goes crazy with ENODEV, but a read_input() now passes the already
freed pInfo for the physical device
- boom
Fix this by appending to the fd list to provide bug-for-bug compatibility with
the old SIGIO code. This needs to be fixed in the driver, but meanwhile not
crashing the server provides for better user experience.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Keith Packard <keithp@keithp.com>
Do all timer stuff before blocking, avoiding a bunch of duplicate code
and merge common code in WaitForSomething.
The WaitForSomething changes need a bit of explanation to show that
the new code is effectively equivalent to the old. Eliding error
checking and trivial bits we've got:
Before:
if (ready clients)
timeout = 0
else
compute timeout
i = poll
if (i <= 0) {
if (ready clients)
return TRUE;
if (input)
return FALSE;
if (any ready timers) {
run timers
return FALSE;
}
} else {
if (input)
return FALSE;
if (any ready timers) {
run timers
return FALSE;
}
if (ready clients)
return TRUE;
}
After:
if (ready clients)
timeout = 0;
else
compute timeout
run_timers
poll
if (input)
return FALSE;
if (ready clients)
return TRUE;
The old code would return TRUE if there were ready clients and input
pending. Dispatch would then schedule that ready client, but before
processing any requests, it would notice that there was input pending
and go process it. The new code just checks for input first, which is
effectively the same.
If the poll timed out and there weren't clients ready, then timers
would get run.
If the poll didn't time out, then timers would get run, even if there
were clients now ready. Now, if the timeout interval was zero, that
means that the timers must have been ready *before* poll was
invoked. In this case, we should simply run the timers before calling
poll -- no sense calling poll just to discard any data that it
generates.
If the timeout interval was non-zero, and poll didn't timeout, then
either there aren't any timers to run, or we got a surprise and hit a
timer exactly as a client became ready to run. This is the one case
where the new code is different from the old; the new code delays the
timer call until the next time WaitForSomething is called.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
No sense having an open-coded linked list here, plus the doubly linked
list is more efficient
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
poll provides per-fd notification of failure, so we don't need
CheckConnections anymore.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
There's no reason to close these now that we don't care what file
descriptors we use.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
There's no reason not to offer ridiculous numbers of clients; only a
few static data structures are arrays of this length.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
With no code depending on the range of file descriptors, checking
for that can be eliminated.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Replace use of select(2) to avoid fd limits. Note that
InputThreadFillPipe used select as well, but none of the files passed
were non-blocking, so there was no need for that code at all.
v2: Keep ospoll API usage single threaded to avoid re-entrancy issues
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Eliminates all of the fd_set mangling in the server main thread
v2: Listen for POLLOUT while writes are blocked.
v3: Only mark client not ready on EAGAIN return from read
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This converts the dispatch loop into using a list of ready clients
instead of an array. This changes the WaitForSomething API so that it
notifies DIX when a client becomes ready to read, instead of returning
the set of ready clients.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This provides a wrapper around poll or epoll providing a
callback-based interface for monitoring activity on a large set of
file descriptors.
v2: use xserver_poll API instead of poll. Don't use WSAPoll as
that is broken.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Jon Turney
Topi Miettinen
Alan Coopersmith
Peter Harris
Adam Jackson
Matthieu Herrb
Andreas Fett
Alexander Volkov
Eric Anholt
Lyude Paul
Damien Leone
Chris Wilson
Thierry Reding
Antoine Martin
Nicolai Hähnle
Jeffrey Smith
Michal Srb
Tomasz Śniatowski
Daniel Martin
Emil Velikov
Keith Packard
Michel Dänzer
Peter Hutterer
Olivier Fourdan
Mihail Konev
Jeremy Huddleston Sequoia