merge xorg-server 21.1.18.

initial import of xorg-server-21.1.18

merge xorg-server 21.1.16.

initial import of xorg-server-21.1.16

merge xorg-server 21.1.14.

initial import of xorg-server-21.1.14

merge xorg-server 21.1.13.

initial import of xorg-server-21.1.13

fix build on alpha.

some clean up upstream didn't get tested on alpha it seems, with both
missing CARD8 etc identifiers as well as dead/duplicate code.

make sure we fill in the name field in generated / converted modes
fixes PR 58321, tested by tsutsui@

initial import of xorg-server-21.1.9

initial import of xorg-server-21.1.6

fix a couple of merge botches (duplicated) delete #if 0'd sections.

reduces diffs vs upstream.

initial import of xorg-server-21.1.5

avoid variable stack arrays.

USE_SSP=yes builds don't work with this new xorg-server version, adjust
the drmmode_set_gamma_lut() call to use malloc() instead of the stack,
and pass down a higher level pointer to gain access to the screen index
so error messages with xf86DrvMsg() work.

tested on pinebookpro.

Pull an upstream (post-21.1) fix for xf86CompatOutput().

https://gitlab.freedesktop.org/xorg/xserver/-/commit/75d70612888f18339703315549db781a22c0cb23
This is an additional fix against the previous revision that is
avaiable in 21.1 branch:
https://gitlab.freedesktop.org/xorg/xserver/-/commit/80eeff3ebac772e25c9107199989e677457dbe06
and may avoid a possible out of bounds access at least on dumb wsfb servers.

ok'ed by mrg@ on current-users@:
https://mail-index.netbsd.org/current-users/2022/07/24/msg042731.html

Restore -flipPixels option removed in upstream server-21.1-branch.

https://gitlab.freedesktop.org/xorg/xserver/-/commit/d1c00c859c6676fbb540420c9055788bc19cb18f
Note ScrnInfoRec structure in hw/xfree86/common/xf86str.h isn't reverted
because it isn't used by -flipPixels option so that we can keep ABI
compatibility among xf86-video-foo drivers.

This -flipPixels is mandatory at least by Xorg server users with
xf86-video-wsfb driver on NetBSD/mac68k and SE/30 mono screen
that has "black=1 and white=0" VRAM.

"go for it" by mrg@ on current-users@:
https://mail-index.netbsd.org/current-users/2022/07/24/msg042731.html

use PRIu64 and hopefully fix some clang builds.

initial import of xorg-server-21.1.4

make this actually build again on arm/mips.

use "1ULL << 32" for a 64 bit member, not 1UL.

fix merge botch.

merge xorg-serer 21.1.3.

initial import of xorg-server-21.1.3

recognize Southland Media MGX

merge xorg-server 1.20.13.

merge xorg-server 1.20.12 and xkeyboard-config 2.33.

initial import of xorg-server-1.20.12

elide cast warning: error: cast from pointer to integer of different size

Add missing include of compiler.h and make all declarations of
IOPortBase consistent.

use PORT_SIZE from compiler.h instead of int.

Provide canonical declaration of IOPortBase for MIPS here, too.

Avoid multiple definitions of the same variable.

initial import of xorg-server-1.20.11

merge xorg-server 1.20.10.

initial import of xorg-server-1.20.10

merge xorg-server 1.20.6.

initial import of xorg-server-1.20.6

Revert previous two commits, will fix this another way.

Don't autoload vesa on x86 at all as it conflicts with wsfb.

Try wsfb before vesa on x86. Modern video cards may not support VBE at all.

Try modesetting on x86, too.

Not the best option without functional GLAMOR, but it does work.

Match wsfb on x86, too.

When uefi booting without driver-specific GPU support, wsfb is the only
choice for a graphical driver.

While here, combine the wsfb cases.

merge xorg-server 1.20.4.

initial import of xorg-server-1.20.4

Try the modesetting driver before wsfb on NetBSD arm and aarch64

pEnt isn't initialized at this point, don't free it
Undo diff to upstream.

There's a good chance we aren't using this code at all.

Undo unnecessary diff to vendor branch.

s/pointer/void */ in a few places. remove now (void *) casts.

use xf86AddMatchedDriver().

merge xorg-server 1.20.3.

initial import of xorg-server-1.20.3

Don't error if we are unprivileged

Fix compile error on evbarm-aarch64. (incompatible pointer types initializing 'present_get_ust_msc_ptr')

apply fixes for CVEs 2017-12176 to 2017-12187.

--
From 1b1d4c04695dced2463404174b50b3581dbd857b Mon Sep 17 00:00:00 2001
From: Nathan Kidd <nkidd@opentext.com>
Date: Sun, 21 Dec 2014 01:10:03 -0500
Subject: hw/xfree86: unvalidated lengths

This addresses:
CVE-2017-12180 in XFree86-VidModeExtension
CVE-2017-12181 in XFree86-DGA
CVE-2017-12182 in XFree86-DRI
--
From 211e05ac85a294ef361b9f80d689047fa52b9076 Mon Sep 17 00:00:00 2001
From: Michal Srb <msrb@suse.com>
Date: Fri, 7 Jul 2017 17:21:46 +0200
Subject: Xi: Test exact size of XIBarrierReleasePointer

Otherwise a client can send any value of num_barriers and cause reading or swapping of values on heap behind the receive buffer.
--
From 4ca68b878e851e2136c234f40a25008297d8d831 Mon Sep 17 00:00:00 2001
From: Nathan Kidd <nkidd@opentext.com>
Date: Fri, 9 Jan 2015 10:09:14 -0500
Subject: dbe: Unvalidated variable-length request in ProcDbeGetVisualInfo
(CVE-2017-12177)

v2: Protect against integer overflow (Alan Coopersmith)
--
From 55caa8b08c84af2b50fbc936cf334a5a93dd7db5 Mon Sep 17 00:00:00 2001
From: Nathan Kidd <nkidd@opentext.com>
Date: Fri, 9 Jan 2015 11:43:05 -0500
Subject: xfixes: unvalidated lengths (CVE-2017-12183)

v2: Use before swap (Jeremy Huddleston Sequoia)

v3: Fix wrong XFixesCopyRegion checks (Alan Coopersmith)
--
From 859b08d523307eebde7724fd1a0789c44813e821 Mon Sep 17 00:00:00 2001
From: Nathan Kidd <nkidd@opentext.com>
Date: Wed, 24 Dec 2014 16:22:18 -0500
Subject: Xi: fix wrong extra length check in ProcXIChangeHierarchy
(CVE-2017-12178)
--
From 9c23685009aa96f4b861dcc5d2e01dbee00c4dd9 Mon Sep 17 00:00:00 2001
From: Michal Srb <msrb@suse.com>
Date: Fri, 7 Jul 2017 17:04:03 +0200
Subject: os: Make sure big requests have sufficient length.

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.
--
From b747da5e25be944337a9cd1415506fc06b70aa81 Mon Sep 17 00:00:00 2001
From: Nathan Kidd <nkidd@opentext.com>
Date: Fri, 9 Jan 2015 10:15:46 -0500
Subject: Unvalidated extra length in ProcEstablishConnection (CVE-2017-12176)

make this build on netbsd.

Fix -O0 build; <machine/bwx.h> should be included after ".arch ev56" pseudo-op.

reduce diff to upstream. no functional change.

prune xrefs to xorgconfig

fix misplaced conditionalized } that broke pointer detection under
vmware when running X without a config file.

Add explicit cast before writing to newly allocated memory.

XXX Beside the strange allocation size, this could be XNFstrdup.

CID 1373525: Checking for NULL after deref is kind of late.

CID 1373520: Fix memory leak, don't allocate needlessly.
CID 1373521: Fix memory leak.

CID 1373518: Memory corruption (off-by-one)

CID 1373515: check return value.

don't use const char for these types, many drivers write to them.

ok mrg

Silently report success for enabling IO on platforms without it.

fix printf format.

use PCI_MAGIC_IO_RANGE instead of hardcoding the wrong ( for shark ) offset
now things actually have a chance of working

fix build issues for sparc64. mostly constification.

check for __arm__ as well as __arm32__.

NetBSD calls it x86_86_iopl().

disable iopl in xorg-config.h as it is set in the reachover makefile.

fix merge botches.

first merge of netbsd changes. not tested yet.

initial import of xorg-server-1.18.4

Avoid legacy VGA I/O port access on MIPS. Xorg converts the I/O port
address to a pointer and dereferences it, and then seg faults.

This fixes Xorg on my Lemote YeeLoong netbook.

Try xf86-video-nouveau for nvidia devices by default.

Fall back to nv (and then vesa and then wsfb) if it doesn't work.

Fix some uninitialized variable warnings with clang.

remove \n\n that confuses the Xorg log file.

we've been happily using bsdPciInit since xorg was first imported.
remove local changes that were only relevant in the xfree86 days.

don't look for _drv.o files. we haven't built any since the bad old days
when all the world was xfree86.

remove superfluous casts.

Remove some redundant function redeclarations, as noted in
http://mail-index.netbsd.org/tech-x11/2015/01/09/msg001452.html

The changes come from upstream in commits:

* xserver:
ecf62755 Define prototypes for hw/xfree86/modes/xf86Modes.c only in xf86Modes.h.
- didn't add xf86SaveModeContents which isn't defined in our version

0bc41d5f Remove redundant redeclarations of functions in the same header file

5595e7ce randr: Remove redundant declaration.

471e5373 Remove duplicate declaration of xf86ValidateModesFlags in xf86Modes.h

9ef53e22 include: GetClientResolutions is declared in font.h

8b6a7500 Fix gcc warnings about redundant declarations of fallback functions
- just took ffs() #ifdef defense, so most not applied


* xf86-video-nv:
49ee1c26 Include xf86Modes.h to use functions from hw/xfree86/modes/xf86Modes.c.

* xf86-video-savage:
2e9217bb Include xf86Modes.h to use functions from hw/xfree86/modes/xf86Modes.c.

apply fixes for:

X.Org Security Advisory: Dec. 9, 2014
Protocol handling issues in X Window System servers

backported to 1.10.x by myself.


included are fixes for:

denial of service due to unchecked malloc in client authentication
CVE-2014-8091
integer overflows calculating memory needs for requests
CVE-2014-8092
CVE-2014-8093
CVE-2014-8094
out of bounds access due to not validating length or offset values in requests
CVE-2014-8095
CVE-2014-8096
CVE-2014-8097
CVE-2014-8098
CVE-2014-8099
CVE-2014-8100
CVE-2014-8101
CVE-2014-8102
CVE-2014-8103

more clang appeasement

Fix format string.

for i386 and amd64, append vesa to the end of the possible video driver
list. makes "startx" without a config file work on systems where there is
a known native driver, but it is not installed (like "vboxvideo").

OFIOCNEXTPROP -> OFIOCGET
now this can actually work

Fix abs usage.

Lint does not know about pragmas

fix for lint

add a field to struct sbus_device which records the framebuffer size reported
by ioctl(FBIOGATTR) so drivers can use it

merge xorg-server 1.10.6.

initial import of xorg-server-1.10.6

remove debug spam

do ioctl(KDSETMODE) when switching in and out of the Xserver's VT
now this works on top of radeonfb

Add parentheses in code that attempts to convert 3 bytes to a 24-bit value.
"+" has higher precedence than "<<", so the old code would have given
incorrect results. Found using clang -Wshift-op-parentheses.

when using dixSetPrivate() and friends it helps to initialize the key before
using it. Now colours on suncg6 look right again and we don't spin or crash
on exit anymore.
probably needs to be pulled into -6

Add vax to the list of non inb() users

if the video driver is "vmware" or "vmwlegacy" and no mouse section exists,
use the "vmmouse" input driver instead of "mouse" -- now xorg works properly
in vmware without a config file

xsrc/45442: X -configure doesn't detect S3 Trio32/64

Backout the following commit from upstream:
http://cgit.freedesktop.org/xorg/xserver/commit/hw/xfree86/common/xf86Helper.c?id=0ceac6f64f5ad9bc2ac4b19be2dd245ffba78b05

Despite what the log message says, xf86MatchDevice can be called at
configure time by drivers who still implement the legacy probing API (and
there are a lot of them!).

When configuring a new PCI or SBus device, remember to stash the bus
specific information (in pVideo for PCI or sVideo for SBus), otherwise
more than one driver may try to attach to the device and 'X -configure'
will fail with "Number of created screens does not match number of
detected devices".

re-apply patch from revision 1.2:

GenerateDriverlist: Change driver matching pattern to be the same as the
one used by the module loader.

include dix-config.h since this depends up on symbols defined there

merge cornflakes with xorg-server 1.10.3

initial import of xorg-server 1.10.3. highlights include:
- fixes xorg bugs 29969, 30367, 30527, 30267, 30260, 29046,
30260, 31548, 28672, 24887, 32115, 21457, 32436, 21827,
28414, 24703, 32803, 16318, 33449, 33324, 33929, 35082,
35209, 36146, 36119
- many additional NULL checks
- adjustments related to input ABI 12
- *lots* of dead code removal
- vga arbitrator fixes
- GL fixes
- EDID fixes
- conversion to using asprintf() in many places
- many fixes from valgrind, compiler warnings

see these for more details:
http://lists.freedesktop.org/archives/xorg-announce/2011-February/001612.html
http://lists.freedesktop.org/archives/xorg-announce/2011-April/001646.html
http://lists.freedesktop.org/archives/xorg-announce/2011-May/001675.html
http://lists.freedesktop.org/archives/xorg-announce/2011-July/001710.html

merge xorg-server 1.9.5

initial import of xorg-server 1.9.5
fixes for FDO bugs 33929, 32803, 24703, 28414, 24887, 30260, 31093, 29046, 30267
memory leak fixes
EDID and modes fixes for multi-screen
dri2 fixes
randr vs invalid size fixes
fixes for Xext triggers
fix for a black-screen after VT switch back (redhat bug 533217)
fixes for x86emu

- get platform #ifdefs in a slightly more sane order
- look for SBus and UPA devices on sparc(64)
- add default drivers for shark and sgimips
- only try wsfb if we can't find any other usable device
With this Xorg without config file works fine on shark, sparc(64) and amd64,
should work on other archs as well.

fix sbusDeviceTable to actually include driver names

get this a little closer to working on non-x86 by:
* not poking around in IO space trying to disable / enable ancient VGAs when
NO_LEGACY_VGA is defined
* actually checking if what we get from read_legacy_video_BIOS() is actually
a valid PC ROM image every time we try

Avoid unaligned accesses.

Needed to make my CATS start X without an xorg.conf.

Reviewed and tested by macallan@. thanks.

Fix merge and ARM define handling.

make this compile on sparc64

put back a little bit of BUS_ISA, enough for shark and chips driver.

put BUS_ISA back, we need it for the chips driver on shark

re-merge our local change to add XAAFillScanlineImageWriteRects()

fix a merge botch.

merge xorg-server 1.9.2.

import xorg-server 1.9.2. important changes since 1.6.5 include:

- built-in symbol visibility, default not export
- fixes for various crashes
- many fixes for XI2, EXA, dix, render, glx, xkb, EDID, randr, Xv,
resource handling, Xext, vfb, xfree86 / vga and composite support.
- fixes several memory leaks
- many many documentation updates
- record extension enabled
- no longer looks for XF86Config files

various build fixes for modern xextproto/libXext and friends. some of
this came from pkgsrc, and some of it i had to figure out myself.

fix a segfault in xf86EnterServerState() which would fire whenever there is
more than one graphics device
This is a bandaid, we ( or the Xorg people ) need to fix it properly by
getting rid of all the now useless RAC-related code.

merge xorg-server 1.6.5

initial import of xorg-server-1.6.5

let xf86AddBusDeviceToConfigure() accept BUS_ISA and BUS_NONE

simplify PIXMAP_IS_SCREEN()

don't fall back to software rendering when drawing into the screen pixmap
this takes care of visible artifacts with non-mappable framebuffers like
newport and crime

check that dixLookupPrivate() returned non-NULL before dereferencing
it. fixes a SEGV when starting ioquake3 on an X server started with
the "-dpi 100 -depth 16" flags, as reported by wiz@.

initial import of xorg-server-1.6.3

when writing PCI BusIDs include the domain ID if it's non-zero
While there also add
Option "Protocol" "wskbd"
Option "Device" "/dev/wskbd"
to the keyboard section on non-x86
Now X -configure should dump something more usable.

merge xorg-server 1.6.2

initial import of xorg-server-1.6.2

ShareVTs is only visible #if defined (SYSCONS_SUPPORT) ||
defined (PCVT_SUPPORT).

fixes sparc64 build for me.

add sparcDriverName() - from Sbus.c

avoid a bunch of int/ptr warnings.

remove a wrong extra {}.

merge xorg-server 1.6.1.190.

XXX: our Pci.[ch] changes need to be re-looked at, i think.

initial import of xorg-server-1.6.1.901

PR# xsrc/40939: nvidia driver can not deal with COMPOSITE/DAMAGE extensions

Apply the following patches from Xorg head:

* http://cgit.freedesktop.org/xorg/xserver/patch/?id=0c6987df3b9b3a37d201d740d8248c326449835e
XAA: Disable offscreen pixmaps by default.
Say Option "XaaOffscreenPixmaps" to turn them back on.

* http://cgit.freedesktop.org/xorg/xserver/patch/?id=01879e583b0ce0d8fddc37e36d9b69fe4f1d1e72
Don't enable XAA offscreen pixmaps with Option "XaaOffscreenPixmaps" "no".
xf86IsOptionSet is rarely the right function to use for boolean options
because it returns TRUE whenever the option is present, even if its
value is "no", "off", "0", etc.

Actually use scanline image writes to upload images if available

add horizontal mouse-wheel functionality to USB and WSMOUSE drivers

add a SetupMouse proc to condition the fd. Convert NetBSD's SetupAuto to
SetupMouse.

The {read,write}Dense{8,16} functions, as implemented, always to 32bits access
and truncate the result to 8/16 bits, or do read/modify/write cycles.
This is bad if the read or write cycle to adjacent registers change the
adapter's state (e.g. if an adjacent register is a fifo, we'll discard
values, or add additionnal, unwanted values in the fifo).

Fix: use the BWX instructions for 8 and 16 bit access.

Also submitted to Xorg as bug #19722

NetBSD-specific code doesn't work on ev6 systems. This is because the
code assumes there is at last 2 mappings, one dense and one sparse.
On ev6 there's only one mapping, and the code access abw[] out of bound.

Fix: check abw_count and if it's 1, use the same mapping for
dense_base and memory_base.

Also submitted to Xorg as bug #19721

On NetBSD/alpha in{b,w,l} and out{b,w,l} are not functions but #define to a
function pointer. This doesn't work with the SYMFUNC() macro, the compiler
outputs and error.

Fix: don't SYMFUNC() these on NetBSD/alpha
Also submitted to Xorg as bug #19720

fix ioctl.

Set the mouse event protocol version. (untested, but head will be broken
unless I add this).

patch from jmcneill@:

add a log message to notify what VT is in use. helps gdm/consolekit.

add __mips__

add ARM support.
For some idiotic reason Xorg still uses different variable for the IO base
address on different architectures.

Add support for PCI_VIA_TTYE0 hack.
( that way we can access PCI config space belonging to the console via ttyE0.
This is useful on machines where the console fb may not be accessible via
/dev/pci0. Also, it hides other PCI devices from the Xserver so it's less
likely to mess with our BARs )
From xsrc/xfree

Make PCI bus address ranges match reality on sparc64 and powerpc

more bits'n'pieces for sparc support

add missing bits & pieces for sparc(64) and Sbus support
While there, use ppc_video.c on sparc(64) as well, sparc64_vodeo.c is
almost identical and only misses EnableIO() stuff which we'll want to add
at some point anyway and then they'd be identlical to the ones from
ppc_video.c anyway.

GenerateDriverlist: Change driver matching pattern to be the same as the
one used by the module loader.

videoPtrToDriverName: use the 'intel' driver instead of 'i810'.

make this compile on sparc64

add sparc64 goo, from xsrc/xfree

Downgrade a few more FatalError()s to X_WARNING, while there use
PCI_MAGIC_IO_RANGE to mmap the IO range instead of a macppc-specific magic
number. We'll need this on shark and the other powerpc-based ports.

Failure to map something is not a FatalError()

revert previous since it causes xorg to break when loading configuration

copy arm_video.c from xsrc/xfree and tweak it.
our version of this seems to be more up-to-date/used than that in x.org

-#elif defined(__powerpc__)
+#elif defined(__powerpc__) || defined(__arm__) || defined(__mips__) || \
+ defined(__sparc__)

from xsrc/xfree bsdResource.c

look for __arm__ as well.

also check for __arm__.

#include <string.h>

adapt this to a reasonable include path

attempt to support netbsd/arm

initial import of xorg-server-1.4.2