This is what we're currently shipping in Debian. Enables the ability for
drivers to ship a text file listing PCI ID's they support, and have the
server read them on startup when no driver is specified. This works, but
isn't the final solution.
Removes "LookupKeyboardDevice" and "LookupPointerDevice" in favor of
inputInfo.keyboard and inputInfo.pointer, respectively; all use cases
are non-XI compliant anyway.
Matches linuxPci.c changes made in 8279444a54
Fixes compiler errors:
"ix86Pci.c", line 194: too many struct/union initializers
"ix86Pci.c", line 204: too many struct/union initializers
"ix86Pci.c", line 214: too many struct/union initializers
Even though they're defined to zero by the spec, we've seen an EDID block
where the (empty) ASCII strings were stuffed in a byte early, leading to the
descriptor being considered a detailed timing instead.
This was an attempt to avoid scratch gc creation and validation for paintwin
because that was expensive. This is not the case in current servers, and the
danger of failure to implement it correctly (as seen in all previous
implementations) is high enough to justify removing it. No performance
difference detected with x11perf -create -move -resize -circulate on Xvfb.
Leave the screen hooks for PaintWindow* in for now to avoid ABI change.
Add keyc->postdown, which represents the key state as of the last mieqEnqueue
call, and use it when we need to know the posted state, instead of the
processed state (keyc->down). Add small functions to getevents.c to query and
modify key state in postdown and use them all through, eliminating previously
broken uses.
In commit 41bb9fce47, the event delivery loop
for Xinput enabled keyboards was changed and accidentally used the wrong
index variable, causing random events to be delivered when returning from VT
switch.
In addition, in commit aeba855b07,
SIGIO was blocked during delivery of these events, but not for the entire
period the xf86Events array was being used. Block SIGIO for the whole loop
to avoid other event delivery from trashing the key release events.
(cherry picked from commit aa7ed1f5f3)
Previously, the server version reported by xdpyinfo and Xorg -version would
bear some vague resemblance to a X.Org katamari version, but in the presence
of modularization (and client-server relationships with different katamari
versions on each side) those numbers don't really make sense. Instead, just
report the package version.
When branching a stable branch, master's version should be immediately updated
to the endpoint of the stable branch plus a snapshot of 1 (for example,
1.4.0.1 after server-1.4-branch). The stable branch should then be changed to
RC0 at that time (1.3.99.0, for example).
This scheme was partially attempted for server 1.3, but lacked the appropriate
master updates, thus why it had to be revisited now. While here, we can also
remove a lot of versioning complexity since everything is based on the package
version.
One of these I introduced by listing dix and mi in the same library list to
simplify other servers. The other had been hacked around using libosandcommon,
which is now gone.
This cleans up server Makefile.ams a little bit, but also means that people
messing with configure.ac need to be careful with whether they put libraries
in the _LIBS or _SYS_LIBS targets. Hopefully the comment in configure.ac will
clarify the issues.
The author of the int10 code looked at the VBIOS POSTing code
in DOSEMU to get some initial idea on how to POST a VBIOS.
To give credit to the DOSEMU Team for this inspiration a comment
was added to the code which could suggest that code from the
GPLed DOSEMU was directly incorporated into this code.
This patch should clearify the situation.
Note that pciaccess doesn't yet have Net/OpenBSD support, but the relevant
code should go there instead of disconnected code in the X Server.
While here, remove the now-disabled INCLUDE_XF86_NO_DOMAIN from the headers,
and un-disable xf8StdAccResFromOS for those OSes without domain support which
will need it.
over to new system.
Need to update documentation and address some remaining vestiges of
old system such as CursorRec structure, fb "offman" structure, and
FontRec privates.
This was a bunch of poorly defined resource ranges per OS/platform combination
which were supposed to represent what regions could potentially have resources
allocated into them.
Composite's automatic redirection is a more general mechanism than the
ad-hoc BS machinery, so it's much prettier to implement the one in terms
of the other. Composite now wraps ChangeWindowAttributes and activates
automatic redirection for windows with backing store requested. The old
backing store infrastructure is completely gutted: ABI-visible structures
retain the function pointers, but they never get called, and all the
open-coded conditionals throughout the DIX layer to implement BS are gone.
Note that this is still not a strictly complete implementation of backing
store, since Composite will throw the bits away on unmap and therefore
WhenMapped and Always hints are equivalent.
xf86RandR12ScreenSetSize must protect calls to EnableDisableFBAccess with
suitable vtSema checks to avoid invoking driver code while the X server is
inactive.
The multi-crtc cursor code in hw/xfree86/modes holds a reference to the
current cursor. This reference must be correctly ref counted so the cursor
is not freed out from underneath this code.
This is where they should have been in the first place. All the rest of
the code in the server defines such things in the source files, not the
headers.
A bunch of CFLAGS had gone missing, so the build failed with errors like:
../../../../../hw/xfree86/os-support/linux/lnx_ev56.c:7:19: error: input.h: No such file or directory
../../../../../hw/xfree86/os-support/linux/lnx_ev56.c:8:24: error: scrnintstr.h: No such file or directory
As a result, we can remove the quirks that existed to flip the bits back around
for us. This is not confirmed in all cases due to lack of bugs containing EDID
blocks associated with the quirks, but is likely true.
The outport is most likely unnecessary on any currently used hardware,
the byte copy is necessary from what I know on IA64 and friends so leave it.
Add a new API entry point which lets a driver select the old behaviour if
such a needs is ever found.
This gives me ~20% speed up on startup on 945 hardware.
If NoAutoAddDevices is given as a server flag, then no devices will be added
from HAL events at all. If NoAutoEnableDevices is given, then the devices will
be added (and the DevicePresenceNotify sent), but not enabled, thus leaving
policy up to the client.
Make sure the font path is always 'built-ins' when we use built-in fonts,
rather than having it as a fixed path for a while, then clobbering it
halfway through startup.
Set the new randr crtc of the output before the output change notification is
delivered to the clients.
Remove RROutputSetCrtc as it is not really necessary. All we have to do is set
the output's crtc on RRCrtcNotify
When the PreferredMode option is selected in the config file, remove the
M_T_PREFERRED bit from all other preferred modes to force the config file
mode to be selected.
Code that disabled mode detection on disabled outputs would confuse
applications by listing said outputs as connected but without any modes.
This makes the disabled state in the config file affect only the initial
configuration and not subsequent modifications by RandR.
The DDX code was ignoring pending properties for computing when mode setting
was required. This meant that configurations differing only in property
values would not cause the mode to be set.
If your loader is as bad as elfloader, then it makes sense for the
server to have some stubs for you to assign to / break on. However it
is no longer 1996.
I made a mistake in some new code using MakeAtom, passing the size of the
string instead of the length of the string. Figuring there might be other
such mistakes, I reviewed the server code and found four bugs of the same
form.
The previous scheme didn't work when the client didn't create the core drawable,
e.g. the root or composite overlay window. Use refcounting via special client
resources to fix that.
Print debug messages only when the appropriate debug bit is set in the
8086 state vector, so you can focus in on the call you're actually
interested in.
This is to avoid issues with redirected windows which are located partly or
fully outside of a screen edge, resulting in unusual cliprects which the 3D
drivers generally can't handle. The symptoms in such cases would be incorrect
rendering or even crashes or hangs.
at server startup, and not against the virtual X/Y parameters
as they can change.
This fixes an issue when canGrow is TRUE and modes get dropped
when using the virtual X/Y parameters.
Major stylistic cleanups, greatly expanded cross-reference ("SEE ALSO")
section and some typo fixes.
This patch by Branden Robinson. Forward-ported by Fabio M. Di Nitto.
When the root window changed size, xf86XVFillKeyHelper would not revalidate
the GC, leaving the clip at the old size causing lossage (and possibly
memory corruption if the screen and frame buffer shrank).
Fixed by just using a scratch GC; saving memory, eliminating bugs and
shrinking the code.
To be used by AIGLX for GLX_EXT_texture_from_pixmap without several data copies.
The texOffsetStart hook must make sure that the given pixmap is accessible by
the GPU for texturing and return an 'offset' that can be used by the 3D
driver for that purpose.
The texOffsetFinish hook is called when the pixmap is no longer being used for
texturing.
DRI uses a non-screen block/wakeup handler which will be executed after the
screen block handler finishes. To ensure that the rotation block handler is
executed under the DRI lock, dynamically wrap the screen block handler for
rotation.
Leaving devices enabled during server startup can cause problems during the
initial mode setting in the server, especially when they are used for
different purposes by the X server than by the BIOS. Disabling all of them
before any mode setting is attempted provides a stable base upon which the
remaining mode setting operations can be built.
The code in hw/xfree86/os-support/bus/sparcPci.c:simbaCheckBus()
is trying to mimmick VGA routing by disabling I/O space responses
behind the Simba PCI-PCI controller.
Unfortunately, doing this also happens to disable access to the
IDE controller I/O space registers, thus crashing the system. The
granularity of the I/O disabling in the Simba controller is not
fine enough to disable VGA without also disabling the IDE controller
registers.
When the modules section is parsed, if a module is set to be loaded by
default, this will be logged. If it is redundantly specified in xorg.conf,
this will also be noted. None of this logging will happen if the xorg.conf
lacks a modules section.
This provides a new option, UseDefaultFontPath. This option is enabled by
default, and causes the X server to always append the default font path
(defined at compile time) to the font path for the server. This will allow
people to specify additional font paths if they want without breaking
their font path, thus hopefully avoiding ye olde "fixed front" problem.
Because this option is a ServerFlag option, the ServerFlags need to be
processed before the files section of the config file, so swap the order
that they are processed.
Provide default modules that may be overrided easily. Previously the
server would load a set of default modules, but only if none were
specified in the xorg.conf, or if you didn't have a xorg.conf at all. This
patch provides a default set and you can add only the "Load" instructions
to xorg.conf that you want without losing the defaults. Similarly, if you
don't want to load a module that's loaded by default, you can add "Disable
modulename" to your xorg.conf (see man xorg.conf in this release for
details). This allows for a minimal "Modules" section, where the user only
need specify what they want to be different. See bug #10541 for more.
The list of default modules is taken from the set loaded by default when
there was a xorg.conf containing no "Modules" section.
A potential problem for some users is that some users disable a module,
most notably DRI, by commenting out the "Load" line in their xorg.conf.
This needs to be changed to an uncommented "Disable" line, as DRI is
loaded by default.
Formerly we sized an array with a compile time constant, then initialized
its size to the same constant, but the Linux PCI init code would increase
that "constant". So if you happened to have more than 128 PCI devices,
you'd happily scribble into whatever variables happened to be in .bss
after that array.
Only really fixed for Linux atm. Other OSes will simply (still) fail to
work on video devices above the 128th PCI device.
SourceValidate is used exclusively by the software cursor code to pull the
cursor off of the screen before using the screen as a source operand. This
eliminates the software cursor from the frame buffer while painting the
rotated image though. Disabling this function by temporarily setting the
screen function pointer to NULL causes the cursor image to be captured.
(cherry picked from commit 05e1c45ade)
Setting a mode on an unrotated CRTC was causing all of the rotation updates
to be disabled; the loop looking for active rotation wasn't actually looking
at each crtc, it was looking at the modified crtc many times.
(cherry picked from commit 8b217dee3a)
I've managed to solve my own bug (#10545) by applying the following
patch to the xserver.
Please apply.
<Conspiracy mode on>
This monitor is "Vista Certified". I wonder if this is a pure coincidence...
<Conspiracy mode off>
With kind regards
Erik Andrén
(cherry picked from commit a63704f14a)
Option "Enable" "True" will force the server to enable an output at startup
time, even if the output is not connected. This also causes the default
modes to be added for this output, allowing even sync ranges to be used to
pick out standard modes.
(cherry picked from commit a3d73ba2cb)
By default, use the screen monitor section for output 0, however, a driver
can change which output gets the screen monitor by calling
xf86OutputUseScreenMonitor.
(cherry picked from commit f4a8e54caf)
The entity (device) has a locking SAREA and a master file descriptor
that optionally isn't closed between server generation.
The locking SAREA contains the device hardware lock.
Each DRI screen creates an new SAREA containing the drawable lock,
drawable-and private info, the drawable SAREA.
The first screen optionally shares its drawable SAREA with the
device SAREA.
Default is to close the master descriptor between server generations,
and to share the drawable SAREA of the first screen with the device locking
SAREA. Thus we should (hopefully) have full backwards compatibility.
Mesa changes to support single-device multiple screens are pending.
This Acer monitor reports support for 75hz refresh via EDID, and yet when
that rate is delivered, the monitor does not sync and reports out of range.
Use the existing 60hz quirk for this monitor.
(cherry picked from commit 1328a288e9)
By the time CloseScreen gets called, we can't call ProcessInputEvents, as
the event queue will get unhappy. So just unregister our hooks instantly,
and hope that they don't get called.
xf86SetSingleMode tries to resize all crtcs to match the selected mode. When
a CRTC has no matching mode, it now disables the CRTC (instead of crashing).
Also, poke the RandR extension when xf86SetSingleMode is done so that
appropriate events can be delivered, and so that future RandR queries return
correct information.
(cherry picked from commit dc6c4f6989)
/sys/devices reflects the bus topology, and we don't care that much.
Easier (and more reliable) to just look in /sys/bus/pci/devices, which
is a flat view.
When we see an evdev or vmmouse section, assume that it's a mouse, and
don't add a default mouse device. This will break users who have an
evdev keyboard section but no mouse, and want the mouse to get added
by default.
Now, fbcmap_mi.c contains the fb functions which just wrap mi functions.
Previously, these were in fbcmap.c and compiled when XFree86Server was defined.
Now, clients of fbcmap should either use fbcmap.c or fbcmap_mi.c and not worry
about setting the XFree86Server symbol.
Additionally, protect libcw setup behind checks for Render, to avoid
segfaulting if Render isn't available (xnest).
The previous setup was an ABI-preserving dance, which is better nuked now.
Now, anything that needs libcw must explicitly initialize it, and
miDisableCompositeWrapper (previously only called by EXA and presumably binary
drivers) is gone.
Yes, two changes in one commit. Sorry 'bout that.
The first change ensures that when pending property values have been
changed, a mode set to the current mode will actually do something, rather
than being identified as a no-op. In addition, the driver no longer needs to
manage the migration of pending to current values, that is handled both
within the xf86 mode setting code (to deal with non-RandR changes) as well
as within the RandR extension itself.
The second change eliminates the two-call Create/AttachScreen stuff that was
done in a failed attempt to create RandR resources before the screen
structures were allocated. Merging these back into the Create function is
cleaner.
(cherry picked from commit 57e87e0d00)
Conflicts:
randr/randrstr.h
randr/rrcrtc.c
I think master and server-1.3-branch are more in sync now.
desiredX and desiredY were not recorded during xf86InitialConfiguration.
desiredX, desiredY and desiredRotation were not recorded during
xf86SetSingleMode.
(cherry picked from commit 36e5227215)
Pending Properties take effect when the driver says they do, so provide an
API to tell DIX when a property effect is made. Also, allow driver
to reject property values in RRChangeOutputProperty.
(cherry picked from commit 8eb288fbd6)
The rotation state is stored in the xf86_config structure which is not
re-initialized at server reset time. Clean it up at CloseScreen time.
(cherry picked from commit f8db7665dc)
The former <X11/extensions/XKBsrv.h> has been pulled into the server now as
include/xkbsrv.h, and the world updated to look for it in the new place,
since it made no sense to define server API in an extension header. Any
further work along this line will need to do similar things with XKBgeom.h
and friends.
The DDC code sets the I2C timeouts to VESA standards, except that it had an
extra setting of the ByteTimeout value which was wrong (off by a factor of
50). Removing this should help DDC work on many more monitors. Note that the
Intel driver duplicated these settings, along with the error. Yay for cult
and paste coding.
The xf86 mode setting code was mis-using this field to try and store a
pointer to a DisplayModeRec, however, each output has its own copy of every
DisplayModeRec leaving the one in in the RRModeRec devPrivate field pointing
at a random DisplayModeRec.
Instead of attempting to rectify this, eliminating the devPrivate entirely
turned out to be very easy; the DDX code now accepts an arbitrary RRModeRec
structure and set that to the hardware, converting it on the fly to a
DisplayModeRec as needed.
(cherry picked from commit 3506b9376c)
The RandR protocol spec has several requests in support of user-defined
modes, but the implementation was stubbed out inside the X server. Fill out
the DIX portion and start on the xf86 DDX portion. It might be necessary to
add more code to the DDX to insert the user-defined modes into the output
mode list.
(cherry picked from commit 63cc2a51ef)
Conflicts:
randr/randrstr.h
Updated code to work in master with recent security API changes.
Rotation block handler was re-registering the rotation damage structure,
creating an infinite loop in the damage code. Track registration of the
damage structure to avoid this.
(cherry picked from commit b14f003b0e)
xf86_reload_cursors is supposed to be called from the crtc mode setting
commit hook; as that happens during server initialization, check for this
case.
(cherry picked from commit 5b77bf2d02)
This moves most of the cursor management code out of the intel driver and
into the general server code. Of course, the hope is that this code will be
useful for other driver writers as well.
Check out xf86Crtc.h for the usage information, making sure you add the
needed hooks to the crtc funcs structure for your driver.
(cherry picked from commit 4d81c99a46)
Previous version used monitor identifiers if present, otherwise output
names. That caused existing working configurations to break when additional
information was added to the configuration file.
(cherry picked from commit 3f5cedf00a)
Screen physical size is set to a random value before the RandR code gets
control, override that and reset it to a value based on the compat_output
physical size (if available). If that output has no physical size, just use
96dpi as the default resolution and set the physical size as appropriate.
(cherry picked from commit 843077f23a)
xf86SetDesiredModes applies the desired modes to each crtc (as selected by
xf86InitialConfiguration initially and modified by successful mode settings
afterwards). For crtcs without a desired mode, pScrn->currentMode is used to
select something workable.
(cherry picked from commit bcade98cca)
This function applies a single mode to the screen (as from RandR 1.1,
XFree86-VidModeExtension or XFree86-DGA) using a policy that selects one
output to reconfigure to the requested mode and then makes all other outputs
fit within that size.
(cherry picked from commit 5a595c1f76)
Box transformation from source to dest area was broken, leaving the wrong
areas painted when the crtc origin was non-zero.
When rotating from left to right, the pixmap doesn't get reallocated, and so
no damage was left in the pixmap from xf86RotatePrepare. Separately damage
the whole crtc area when this occurs to repaint the area.
(cherry picked from commit 2a50ca2160)
canGrow indicates to the DDX that the driver can enlarge the desktop via the
xf86_config->funcs->resize hook. If so, xf86InitialConfiguration will set
virtual[XY] to match the configuration it chooses and will leave the crtc config
size ranges alone. If FALSE, it will bloat the screen to fit the largest probed
mode and also set the crtc config max size to limit the desktop to the initial
virtual[XY] size.
This hook is called when the DDX needs to resize the screen. The driver is
responsible for changing virtualX and virtualY, along with any other related
screen properties (devPrivate.ptr, devKind, displayWidth, etc.).
Use the size range from the crtc config instead of randrp->virtual[XY] when
reporting the min and max screen sizes to the DDX.
The hook is called whenever the clipList of any DRI window changes, be it via
DRIClipNotify, DRICreateDrawable or DRIDrawablePrivDelete. This allows the
driver to keep track of which DRI windows are visible where.
in xf86CrtcSetMode, scrn->pScreen will be NULL during server startup time,
so don't try to set the subpixel order. subpixel order will be set in the
randr initialization anyways.
(cherry picked from commit 5f6f8616d8)
New modes header files required a few minor changes to be used by external
drivers, the most notable of which is the publication of the config file
parser header files.
Add monitor "Rotate" option taking one of "normal", "left", "inverted" or
"right". However, because initial mode selection is made before the screen
is completely initialized, we cannot create the shadow pixmap object at this
point. Pend the shadow pixmap creation until the block handler.
Note that this code is not completely functional yet.
Code added in hw/xfree86/modes came from the server-1.3-branch.
Portions of this code had previously been integrated into xf86Mode.c
and edid_modes.c.
To preserve hw/xfree86/modes as much as possible, the duplicate code from
the other files has been disabled; a more careful review would figure out
where that code actually belonged.
RandR 1.0 sizeID must be computed the same way every time, so when reporting
it in the ScreenChangeNotify event, just construct the usual 1.0 data block
and use that.
subpixel geometry information can be computed by looking at the connected
outputs and finding any with subpixel geometry and using one of those for
the global screen subpixel geometry. This might be improved by reporting
None if more than one screen has information and they conflict.
This code comes from the intel driver, so there's no history in this tree.
As the crtc/output-based mode selection code uses ddc, the ddc and i2c
modules have been merged into the server. Attempts to load them are safely
ignored now.
This instruction is being used in some debug VBIOSes. This implementation
doesn't even try to be accurate. Instead, it just increments the counter by a
fixed amount every time an rdtsc instruction in encountered, to avoid divides by
zero.
Our modes typically come from EDID or default modes, and when the monitor
asks for a specific mode, deciding to tweak it usually results in incorrect
display. And if the user is specifying a mode by hand, tweaking it then is
still pretty rude.
Reviewed by: ajax
Currently, the call to linuxPciOpenFile() is always made for read
only access which causes the subsequent mmap call to fail when the
memory is mapped read/write.
Xorg #9692
xf86ReadLegacyBIOS is only used by one function in int10/generic.c.
Move a generic implementation of that function there, rename it to
read_legcay_BIOS, and delete all remnants of it from all other places.
Convert xf86GetPciHostConfigFromTag to a new function called
get_parent_bridge. This name better represents what
xf86GetPciHostConfigFromTag is used for: walking up the lists of PCI
bridges from a device.
There's no need to store the slot information for a PCI device as its
ID. Instead, skip the middle man and just store a pointer to the
pci_device structure.
Rather than allocate a 9 byte buffer on each invocation, use a static
16 byte buffer. Use snprintf for safety. This commit should probably
be cherry-picked to the trunk.
Eliminate xf86GetPciDomain. The domain from libpciaccess is the
domain. Period. This means that 0 is a valid domain. Make sure that
INCLUDE_XF86_NO_DOMAIN is *not* set. Always run in "domain mode,"
even if the only domain possible is 0.
Convert all uses of PCITAG in int10 and vgaHW to 'struct pci_device'.
This allows the conversion of xf86ReadLegacyVideoBIOS and
xf86MapDomainMemory to 'struct pci_device' from PCITAG.
The fbdev API doesn't allow setting the pitch explicitly, so we have to set
the virtual width to the pitch we're using for drawing. This fixes corruption
after changing the virtual width with RandR.
The fbdev API allows the driver to 'accept' modes it doesn't really support by
modifying it to the nearest supported mode. Without this check, e.g. vesafb
would appear to accept all modes, even though it actually can't set any modes
other than the bootup mode at all.
This has been what has been used the most successfully post-damagetrack.
The current thinking is that:
1) We should be able to accelerate basically everything. So we don't need to
try to migrate trees of pixmaps permanently out of framebuffer to speed
CPU drawing up.
2) Migration is cheaper in the thrashing case, so we don't want to go to a lot
of effort to try (and fail badly) to find a working set.
- Added -extension & +extension to Xserver man page
- Changed Xorg synopsis from X11R6 to X11R7
- Clarified Xorg ancestry description
- Moved Solaris to free/Open Source OS list
- Removed references to MetroLink module loader & getconfig
- Converted (1) to (__appmansuffix__) in a few more places
- Replaced http://www.freedesktop.org/cvs/ with http://gitweb.freedesktop.org/