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>
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>
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>
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>
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>
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>
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>
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>
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 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>
In WaitForSomething(), the fd_set clientsWritable may be used
unitialized when the boolean AnyClientsWriteBlocked is set in the
WakeupHandler(). This leads to a crash in FlushAllOutput() after
x11proto's commit 2c94cdb453bc641246cc8b9a876da9799bee1ce7.
The problem did not manifest before because both the XFD_SIZE and the
maximum number of clients were set to 256. As the connectionTranslation
table was initalized for the 256 clients to 0, the test on the index not
being 0 was aborting before dereferencing the client #0.
As of commit 2c94cdb453bc641246cc8b9a876da9799bee1ce7 in x11proto, the
XFD_SIZE got bumped to 512. This lead the OutputPending fd_set to have
any fd above 256 to be uninitialized which in turns lead to reading an
index after the end of the ConnectionTranslation table. This index would
then be used to find the client corresponding to the fd marked as
pending writes and would also result to an out-of-bound access which
would usually be the fatal one.
Fix this by zeroing the clientsWritable fd_set at the beginning of
WaitForSomething(). In this case, the bottom part of the loop, which
would indirectly call FlushAllOutput, will not do any work but the next
call to select will result in the execution of the right codepath. This
is exactly what we want because we need to know the writable clients
before handling them. In the end, it also makes sure that the fds above
MaxClient are initialized, preventing the crash in FlushAllOutput().
Thanks to everyone involved in tracking this one down!
Reported-by: Karol Herbst <freedesktop@karolherbst.de>
Reported-by: Tobias Klausmann <tobias.klausmann@mni.thm.de>
Signed-off-by: Martin Peres <martin.peres@linux.intel.com>
Tested-by: Tobias Klausmann <tobias.klausmann@mni.thm.de>
Tested-by: Martin Peres <martin.peres@linux.intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91316
Cc: Ilia Mirkin <imirkin@alum.mit.edu>
Cc: Olivier Fourdan <ofourdan@redhat.com
Cc: Adam Jackson <ajax@redhat.com>
Cc: Alan Coopersmith <alan.coopersmith@oracle.com
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
This lets us stop using the 'pointer' typedef in Xdefs.h as 'pointer'
is used throughout the X server for other things, and having duplicate
names generates compiler warnings.
Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Eric Anholt <eric@anholt.net>
If a client passes a section of memory via file descriptor and then
subsequently truncates that file, the underlying pages will be freed
and the addresses invalidated. Subsequent accesses to the page will
fail with a SIGBUS error.
Trap that SIGBUS, figure out which segment was causing the error and
then allocate new pages to fill in for that region. Mark the offending
shared segment as invalid and free the resource ID so that the client
will be able to tell when subsequently attempting to use the segment.
Signed-off-by: Keith Packard <keithp@keithp.com>
v2: Use MAP_FIXED to simplify the recovery logic (Mark Kettenis)
v3: Also catch errors in ShmCreateSegment
Conflicts:
include/dix-config.h.in
include/xorg-config.h.in
And now that we have the accessors, localize it. No functional changes, just
preparing for a future change.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
If TimerSet() is called from a signal handler (synaptics tap handling code)
may result in list corruption if we're currently inside TimerSet().
See backtrace in
https://bugzilla.redhat.com/show_bug.cgi?id=814869
Block signals for all list manipulations in the timers.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Chase Douglas <chase.douglas@canonical.com>
Preparation work for per-device idle counters.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Jeremy Huddleston <jeremyhu@apple.com>
Reviewed-by: Jamey Sharp <jamey@minilop.net>
Reviewed-by: James Jones <jajones@nvidia.com>
This is strictly the application of the script 'x-indent-all.sh'
from util/modular. Compared to the patch that Daniel posted in
January, I've added a few indent flags:
-bap
-psl
-T PrivatePtr
-T pmWait
-T _XFUNCPROTOBEGIN
-T _XFUNCPROTOEND
-T _X_EXPORT
The typedefs were needed to make the output of sdksyms.sh match the
previous output, otherwise, the code is formatted badly enough that
sdksyms.sh generates incorrect output.
The generated code was compared with the previous version and found to
be essentially identical -- "assert" line numbers and BUILD_TIME were
the only differences found.
The comparison was done with this script:
dir1=$1
dir2=$2
for dir in $dir1 $dir2; do
(cd $dir && find . -name '*.o' | while read file; do
dir=`dirname $file`
base=`basename $file .o`
dump=$dir/$base.dump
objdump -d $file > $dump
done)
done
find $dir1 -name '*.dump' | while read dump; do
otherdump=`echo $dump | sed "s;$dir1;$dir2;"`
diff -u $dump $otherdump
done
Signed-off-by: Keith Packard <keithp@keithp.com>
Acked-by: Daniel Stone <daniel@fooishbar.org>
Acked-by: Alan Coopersmith <alan.coopersmith@oracle.com>
We can return from WaitForSomething with no clients ready for any number
of reasons. There's no reason to set up the scheduler timer when this
happens.
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Adam Jackson <ajax@redhat.com>
The only remaining X-functions used in server are XNF*, the rest is converted to
plain alloc/calloc/realloc/free/strdup.
X* functions are still exported from server and x* macros are still defined in
header file, so both ABI and API are not affected by this change.
Signed-off-by: Mikhail Gusarov <dottedmag@dottedmag.net>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
xextproto had Xlib client headers moved into libXext.
Protocol header files are named fooproto.h, header files with constants
foo.h or fooconst.h where foo.h was already in use for client-side headers.
Save in a few special cases, _X_EXPORT should not be used in C source
files. Instead, it should be used in headers, and the proper C source
include that header. Some special cases are symbols that need to be
shared between modules, but not expected to be used by external drivers,
and symbols that are accessible via LoaderSymbol/dlopen.
This patch also adds conditionally some new sdk header files, depending
on extensions enabled. These files were added to match pattern for
other extensions/modules, that is, have the headers "deciding" symbol
visibility in the sdk. These headers are:
o Xext/panoramiXsrv.h, Xext/panoramiX.h
o fbpict.h (unconditionally)
o vidmodeproc.h
o mioverlay.h (unconditionally, used only by xaa)
o xfixes.h (unconditionally, symbols required by dri2)
LoaderSymbol and similar functions now don't have different prototypes,
in loaderProcs.h and xf86Module.h, so that both headers can be included,
without the need of defining IN_LOADER.
xf86NewInputDevice() device prototype readded to xf86Xinput.h, but
not exported (and with a comment about it).
This is the biggest "visibility" patch. Instead of doing a "export"
symbol on demand, export everything in the sdk, so that if some module
fails due to an unresolved symbol, it is because it is using a symbol
not in the sdk.
Most exported symbols shouldn't really be made visible, neither
advertised in the sdk, as they are only used by a single shared object.
Symbols in the sdk (or referenced in sdk macros), but not defined
anywhere include:
XkbBuildCoreState()
XkbInitialMap
XkbXIUnsupported
XkbCheckActionVMods()
XkbSendCompatNotify()
XkbDDXFakePointerButton()
XkbDDXApplyConfig()
_XkbStrCaseCmp()
_XkbErrMessages[]
_XkbErrCode
_XkbErrLocation
_XkbErrData
XkbAccessXDetailText()
XkbNKNDetailMaskText()
XkbLookupGroupAndLevel()
XkbInitAtoms()
XkbGetOrderedDrawables()
XkbFreeOrderedDrawables()
XkbConvertXkbComponents()
XkbWriteXKBSemantics()
XkbWriteXKBLayout()
XkbWriteXKBKeymap()
XkbWriteXKBFile()
XkbWriteCFile()
XkbWriteXKMFile()
XkbWriteToServer()
XkbMergeFile()
XkmFindTOCEntry()
XkmReadFileSection()
XkmReadFileSectionName()
InitExtInput()
xf86CheckButton()
xf86SwitchCoreDevice()
RamDacSetGamma()
RamDacRestoreDACValues()
xf86Bpp
xf86ConfigPix24
xf86MouseCflags[]
xf86SupportedMouseTypes[]
xf86NumMouseTypes
xf86ChangeBusIndex()
xf86EntityEnter()
xf86EntityLeave()
xf86WrapperInit()
xf86RingBell()
xf86findOptionBoolean()
xf86debugListOptions()
LoadSubModuleLocal()
LoaderSymbolLocal()
getInt10Rec()
xf86CurrentScreen
xf86ReallocatePciResources()
xf86NewSerialNumber()
xf86RandRSetInitialMode()
fbCompositeSolidMask_nx1xn
fbCompositeSolidMask_nx8888x0565C
fbCompositeSolidMask_nx8888x8888C
fbCompositeSolidMask_nx8x0565
fbCompositeSolidMask_nx8x0888
fbCompositeSolidMask_nx8x8888
fbCompositeSrc_0565x0565
fbCompositeSrc_8888x0565
fbCompositeSrc_8888x0888
fbCompositeSrc_8888x8888
fbCompositeSrcAdd_1000x1000
fbCompositeSrcAdd_8000x8000
fbCompositeSrcAdd_8888x8888
fbGeneration
fbIn
fbOver
fbOver24
fbOverlayGeneration
fbRasterizeEdges
fbRestoreAreas
fbSaveAreas
composeFunctions
VBEBuildVbeModeList()
VBECalcVbeModeIndex()
TIramdac3030CalculateMNPForClock()
shadowBufPtr
shadowFindBuf()
miRRGetScreenInfo()
RRSetScreenConfig()
RRModePruneUnused()
PixmanImageFromPicture()
extern int miPointerGetMotionEvents()
miClipPicture()
miRasterizeTriangle()
fbPush1toN()
fbInitializeBackingStore()
ddxBeforeReset()
SetupSprite()
InitSprite()
DGADeliverEvent()
SPECIAL CASES
o defined as _X_INTERNAL
xf86NewInputDevice()
o defined as static
fbGCPrivateKey
fbOverlayScreenPrivateKey
fbScreenPrivateKey
fbWinPrivateKey
o defined in libXfont.so, but declared in xorg/dixfont.h
GetGlyphs()
QueryGlyphExtents()
QueryTextExtents()
ParseGlyphCachingMode()
InitGlyphCaching()
SetGlyphCachingMode()
The smart scheduler itimer currently always fires after each request
(which in turn causes the CPU to wake out of idle, burning precious
power). Rather than doing this, just stop the timer before going into
the select() portion of the WaitFor loop. It's a cheap system call, and
it will only get called if there's no more commands batched up from the
active fd.
This change also allows some of the functions to be simplified;
setitimer() will only fail if it's passed invalid data, and we don't do
that... so make it void and remove all the conditional code that deals
with failure.
The change also allows us to remove a few variables that were used for
housekeeping between the signal handler and the main loop.
Signed-off-by: Keith Packard <keithp@koto.keithp.com>