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>
v2: rename as 'xserver_poll' to avoid potential library name
collisions. Provide 'xserver_poll.h' which uses the system
poll where available and falls back to this emulation otherwise.
Autodetects when this is required, building the emulation only
then
Source: https://github.com/bmc/poll
Signed-off-by: Adam Jackson <ajax@redhat.com>
The file descriptors passed to InputThreadFillPipe are always
blocking, so there's no need to use Select (or poll).
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
The timeout resolution offered in the AdjustWaitForDelay call is
only milliseconds, so passing around the timeout as a pointer to a
struct timeval is not helpful. Doing everything in milliseconds up to
the point of the select call simplifies the code without affecting
functionality at all.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Instead of having scheduling done in two places (one in
WaitForSomething, and the other in SmartScheduleClient), just stick
all of the scheduling in SmartScheduleClient.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
All uses of these interfaces should instead be using the NotifyFd API
instead.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This removes the last uses of fd_set from the server interfaces
outside of the OS layer itself.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Instead of open-coding a single FD wait, use NotifyFd to wait for the
FD to become readable before returning the error message.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This new libXfont API eliminates exposing internal X server symbols to
the font library, replacing those with a struct full of the entire API
needed to use that library.
v2: Use libXfont2 instead of libXfont_2
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
For link local addresses the XDMCP server would need to either know the
interface thru a scope identifier or try all available interfaces. If
they don't this address will fail in which case the XDMCP server could
still try the other addresses passed - however some only try the first
address and then give up.
Even if this seems to be the wrong place to fix this it seems to be
easier than fixing all display servers.
[ajax: Cleaned up commit message]
Reviewed-by: Adam Jackson <ajax@redhat.com>
As the man page for the latter states:
The effects of signal() in a multithreaded process are unspecified.
We already have an interface to call sigaction() instead, use it.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Timer processing can happen on either the main thread or the input
thread. As a result, it must be done under the input lock.
Signals are unrelated to timers now that SIGIO isn't used for input
processing, so stop blocking signals while processing timers.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
If the server calls AbortServer during the first-time initialization
(which can happen if you start the server on an already using
DISPLAY), then the dbus code will shut down and call the notify fd
interface. If the notify fd list hasn't been initialized, the server
will crash.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
The list of input devices may be changed by hotplugging while the
server is active, and those changes may come from either the main
thread or the input thread. That means the list of input devices needs
to be protected by a mutex.
This prevents input drivers from receiving I/O ready callbacks after
removing a device.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
By default the X server will try CLOCK_MONOTONIC_COARSE before
CLOCK_MONOTONIC, while A Wayland compositor may only support getting
their timestamps from the CLOCK_MONOTONIC clock. This causes various
issues since it may happen that a timestamp from CLOCK_MONOTONIC
retrieved before a sending an X request will still be "later" than the
timestamp the X server than gets after receiving the request, due to the
fact that CLOCK_MONOTONIC_COARSE has a lower resolution.
To avoid these issues, make Xwayland always use CLOCK_MONOTONIC, so
that it becomes possible for Wayland compositor only supporting
CLOCK_MONOTONIC and X server to use the same clock.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
Acked-by: Daniel Stone <daniels@collabora.com>
Tested-by: Olivier Fourdan <ofourdan@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
input_force_unlock was mis-using input_mutex_lock and leaving it set
to -1. As this is executed from OsInit at each server generation, on
the second time through, the mutex would be left locked (!) due to the
trylock call. This caused input to fail after the first server reset.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
This matches a change made in xcb and improves performance for a small
increase in memory usage.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
The current SIGIO signal handler method, used at generation of input events,
has a bunch of oddities. This patch introduces an alternative way using a
thread, which is used to select() all input device file descriptors.
A mutex was used to control the access to input structures by the main and input
threads. Two pipes to emit alert events (such hotplug ones) and guarantee the
proper communication between them was also used.
Co-authored-by: Fernando Carrijo <fcarrijo@freedesktop.org>
Signed-off-by: Tiago Vignatti <tiago.vignatti@nokia.com>
v2: Fix non-Xorg link. Enable where supported by default.
This also splits out the actual enabling of input threads to
DDX-specific patches which follow
v3: Make the input lock recursive
v4: Use regular RECURSIVE_MUTEXes instead of rolling our own
Respect the --disable-input-thread configuration option by
providing stubs that expose the same API/ABI.
Respond to style comments from Peter Hutterer.
v5: use __func__ in inputthread debug and error mesages.
Respond to style comments from Peter Hutterer.
v6: use AX_PTHREAD instead of inlining pthread tests.
Suggested by Emil Velikov <emil.l.velikov@gmail.com>
v7: Use pthread_sigmask instead of sigprocmask when using threads
Suggested by Adam Jackson <ajax@redhat.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
This removes all of the SIGIO handling support used for input
throughout the X server, preparing the way for using threads for input
handling instead.
Places calling OsBlockSIGIO and OsReleaseSIGIO are marked with calls
to stub functions input_lock/input_unlock so that we don't lose this
information.
xfree86 SIGIO support is reworked to use internal versions of
OsBlockSIGIO and OsReleaseSIGIO.
v2: Don't change locking order (Peter Hutterer)
v3: Comment weird && FALSE in xf86Helper.c
Leave errno save/restore in xf86ReadInput
Squash with stub adding patch (Peter Hutterer)
v4: Leave UseSIGIO config parameter so that
existing config files don't break (Peter Hutterer)
v5: Split a couple of independent patch bits out
of kinput.c (Peter Hutterer)
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
By the time we get to ComputeLocalClient, we've already done
NextAvailableClient → ReserveClientIds → DetermineClientCmd (assuming
we're built with #define CLIENTIDS), so we can look up the name of the
client process and refuse to treat ssh's X forwarding as if it were
local.
v2: (Michel Dänzer)
* Only match "ssh" itself, not other executable names starting with
that prefix.
* Ignore executable path for the match.
v3: (Michel Dänzer)
* Use GetClientCmdName (Mark Kettenis)
* Perform check on Windows as well, but only ignore path on Cygwin
(Martin Peres, Emil Velikov, Jon Turney)
v4: (Michel Dänzer)
* Cut of any colon and whatever comes after it. (Adam Jackson)
* Add bugzilla reference.
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93261
Signed-off-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
The type of fd_mask was changed in Cygwin 2.4.0 headers from 'long' to
'unsigned long'. This exposes an existing problem with winauth.c, which
includes Xwindows.h (which includes windows.h, which defines WIN32),
before including osdep.h, which causes the now conflicting definition of
fd_mask in osdep.h to be exposed:
In file included from ../os/osdep.h:198:18: error: conflicting types for
‘fd_mask’ typedef long int fd_mask; /usr/include/sys/select.h:46:23:
note: previous declaration of ‘fd_mask’ was here typedef unsigned long
fd_mask;
Adjust the include guards in osdep.h to make sure we only use WIN32
guarded code when not compiling for Cygwin (i.e. WIN32 && !__CYGWIN__)
This isn't a very elegant, but unfortunately appears to be the best
solution, since it doesn't seem to be possible to write the test in a
positive form.
Future work: Should also audit of all the other uses of WIN32 in
xserver, and make sure they are correct.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>
Removed from xtrans in 2012, and never wired up in the modular build
anyway.
Signed-off-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
When -displayfd is looping through the possible display ids to use,
if it can't open all the listening sockets for one (say when :0 is
already in use), it calls CloseWellKnownConnections to close all
the ListenTransConns entries before the point that ListenTransFds
was allocated & initialized, so CloseWellKnownConnections would
segfault trying to read entries from a NULL ListenTransFds pointer.
Introduced by commit 7b02f0b8
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=93212
Previously all X servers started with -displayfd would overwrite
Xorg.0.log - now a temporary name of Xorg.pid-<pid>.log is used
until after -displayfd finds an open display - then it is renamed
to the traditional Xorg.<display>.log name.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Replace the custom path for dealing with new incoming connections with
the general-purpose NotifyFd API.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
This removes the block and wakeup handlers and replaces them with a
combination of a NotifyFd callback and timers.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
This adds the ability to be notified when a file descriptor is
available for writing.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
This provides a callback-based interface to monitor file
descriptors beyond the usual client and device interfaces.
Modules within the server using file descriptors for reading and/or
writing can call
Bool SetNotifyFd(int fd, NotifyFdProcPtr notify_fd, int mask, void *data);
mask can be any combination of X_NOTIFY_READ and X_NOTIFY_WRITE.
When 'fd' becomes readable or writable, the notify_fd function will be
called with the 'fd', the ready conditions and 'data' values as arguments,
When the module no longer needs to monitor the fd, it will call
void RemoveNotifyFd(int fd);
RemoveNotifyFd may be called from the notify function.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
This allows the server to call GetTimeInMillis() after each request is
processed to avoid needing setitimer. -dumbSched now turns off the
setitimer.
Reviewed-by: Adam Jackson <ajax@redhat.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Globally replace #ifdef and #if defined usage of 'sun' with '__sun'
such that strict ISO compiler modes such as -ansi or -std=c99 can be used.
Signed-off-by: Richard PALO <richard@NetBSD.org>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
