xserver-xorg-video-nouveau: Changes to 'ubuntu-jaunty'
New branch 'ubuntu-jaunty' available with the following commits:
commit 50e253a9005e6df74af656e4165090ed62a0b6a4
Author: Christopher James Halse Rogers <chris@CowboyLaputopu.(none)>
Date: Tue Mar 24 20:42:49 2009 +1100
Build with --with-kms=no (required for Ubuntu DRM headers).
Add a patch to fix --with-kms=no, and quilt patchsystem to apply it
commit faa15918a7307adc4ba17e4747af9ed1b5dd1498
Author: Christopher James Halse Rogers <chris@CowboyLaputopu.(none)>
Date: Tue Mar 24 17:19:25 2009 +1100
Munge maintainer field as per MaintainerFieldSpec
commit 805070fc35d6cb6d5f2987fb7060898030eec5a7
Author: Christopher James Halse Rogers <chris@CowboyLaputopu.(none)>
Date: Sun Mar 22 20:26:39 2009 +1100
Fix package version in debian/changelog
commit 328f8a32b6e3b1faeef67111f252be87cddb286d
Author: Christopher James Halse Rogers <chris@CowboyLaputopu.(none)>
Date: Sun Mar 22 20:11:15 2009 +1100
New upstream snapshot
commit e6f07dfb7d6ca922cb747419fd60e3313bd60427
Merge: d87aa85a62f7221bf59ea567596e108370b822ab 3e7fa97a50e42c0255876c79137dc1a40a6052a6
Author: Christopher James Halse Rogers <chris@CowboyLaputopu.(none)>
Date: Sun Mar 22 17:08:45 2009 +1100
Merge branch 'upstream-ubuntu-jaunty' into ubuntu-jaunty
* upstream-ubuntu-jaunty: (102 commits)
dri: pass object handle instead of offset if using real mm
dri2: initial implementation, only with driver pixmaps
kms: support framebuffer resize if driver pixmaps enabled
kms: copy old fb to new fb on modeset
Call NVEnterVT from NVScreenInit rather than duplicating it.
exa: fix multiple prepare/finish_access on a pixmap
Wrap bios endian casting
Switch to using pll_vals struct for passing pll stuff around
randr12: cull old unused pll code
Don't pointlessly overallocate cursor storage
randr12: nv10 hw cursor fixes/changes
randr12: use nv0x cursor colour expansion funcs on later cards, unify argb paths
randr12: store nv0x hw cursor image in VRAM
Kill pre-randr12 hwcursor code, and tidy randr12 hwcursor code
Remove calls to load ddc and i2c modules -- they're built-in since server 1.3
Hack around missing display table in rdivacky's GeForce 7050 PV/nForce 630a
Fix xserver shutdown for randr12 off
Remove bios lvds parsing pxclk test; ddc_permitted is a sufficient condition
Name some regs/values and reduce crtc reg struct size
improve drm/dri initialisation code
...
commit 3e7fa97a50e42c0255876c79137dc1a40a6052a6
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Wed Mar 18 09:36:33 2009 +1000
dri: pass object handle instead of offset if using real mm
commit 88efe405213a3ea23edd9bed971da1d2492021eb
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Mar 16 17:45:49 2009 +1000
dri2: initial implementation, only with driver pixmaps
Also, no DRI1 when driver pixmaps enabled.
commit 675126e0f0c1c077bde4b25ccb64f87d1d08547b
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Mar 16 15:06:31 2009 +1000
kms: support framebuffer resize if driver pixmaps enabled
commit 8b354b92a859cf277b19e51fc6cb109f4ea44b7a
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Mar 16 13:13:07 2009 +1000
kms: copy old fb to new fb on modeset
Requires driver pixmaps enabled, which suck on G8x currently. Paves the
way for a nicer transition from console->X. There's still a flash where
the screen goes black before X starts here, but this is on the way to what
we want.
commit 6658403cfd91556d06afecdde03feec7384c5398
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Mar 16 12:34:49 2009 +1000
Call NVEnterVT from NVScreenInit rather than duplicating it.
commit f59fef15c94fc5c6f216d0f5dcdcb3d3188c0049
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Mar 16 09:17:29 2009 +1000
exa: fix multiple prepare/finish_access on a pixmap
commit 75a1c4158a19bf0857de49b736fb87a40c095048
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sat Mar 14 15:48:39 2009 +0000
Wrap bios endian casting
Also convert some functions to use individual bios pointers, rather than
bios->data[offset + ...
Plus some minor bmp structure documentation and tidying
commit 3666d73333ecbdd3a4941a76af2528113db427bb
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Feb 25 20:21:42 2009 +0000
Switch to using pll_vals struct for passing pll stuff around
commit a3dbc06d4f9dc340f5c167b94d8e10ac8386c5b4
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Apr 11 16:51:30 2008 +0100
randr12: cull old unused pll code
commit 71c821b8119cbac8dcd91b27e3b7556d00ceb546
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sat Mar 14 19:06:12 2009 +0000
Don't pointlessly overallocate cursor storage
commit 7284c07a0ad6551ea99c28454f4e7308699835e6
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Mar 13 21:23:48 2009 +0000
randr12: nv10 hw cursor fixes/changes
Remove access to PCRTC_GPIO, it's unused and on NV10 overwrites
PCRTC_CURSOR_CONFIG leading to all sorts of cursor weirdness
NV10 hw cursor is now set-up on the assumption it can do 64x64 (we've been
using 64x32 up to now), but if it can't do cursors 64px high it can easily
be fixed to behave like nv0x at 32x32 (64x32's pretty silly)
It's possible that with the PCRTC_GPIO thing resolved the nv10 cursor can
be more featureful: the code would also be simpler if it supported 32bpp
cursors (test by seeing if setting NV_CRTC_CURSOR_CONFIG_32BPP has any
effect). It may even do alpha -- change pNv->alphaCursor to be
(NVArch >= 0x10) to find out.
commit bcc3a14b15cdad3647cc7df0f540df1111843be5
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Mar 13 02:31:35 2009 +0000
randr12: use nv0x cursor colour expansion funcs on later cards, unify argb paths
commit b21c807d2ced35ac1b06300f6e68d661710b336a
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Mar 13 03:52:54 2009 +0000
randr12: store nv0x hw cursor image in VRAM
nv_driver.c is already doing the alloc, nvidia uses the same method, and it saves code :)
commit 3ea4b194e83735145d5e9d5c5357f92ae6c6a569
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Thu Mar 12 15:37:59 2009 +0000
Kill pre-randr12 hwcursor code, and tidy randr12 hwcursor code
If you're not using randr12 you should be.
Misc fixes included:
Ensure PNVM bit set on PCRTC_CURSOR_CONFIG for consistency
Fix NV11 alpha cursors
Add missing bo unref
commit b988160dfa5a535fd042062d7318ab99fc7efa5a
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Thu Mar 12 16:31:24 2009 +0000
Remove calls to load ddc and i2c modules -- they're built-in since server 1.3
commit d5893ecf3dd915fc6937c61e6eae7c221edeb6e2
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Thu Mar 12 21:46:02 2009 +0000
Hack around missing display table in rdivacky's GeForce 7050 PV/nForce 630a
commit 369e1efd23c0180e210dbe898318b6b9efa1faad
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Thu Mar 12 16:09:27 2009 +0000
Fix xserver shutdown for randr12 off
commit 8343d6b2c0ebddf1a1a7d146890ed90278559fea
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Mar 11 01:11:02 2009 +0000
Remove bios lvds parsing pxclk test; ddc_permitted is a sufficient condition
commit 8b56c699ad39c64326d76658c741801bb4118c5f
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Mar 11 00:43:33 2009 +0000
Name some regs/values and reduce crtc reg struct size
TVOUT_LATENCY taken from nv11 .scp file
CRTC[]: even nvidia only use up to 9f, and they know what the regs do
commit 79d23d87b489332af521fbda418a13b9ae032b7d
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Fri Mar 13 11:43:54 2009 +1000
improve drm/dri initialisation code
The main motivation here was to fix the scenario where X is started without
nouveau.ko loaded, checks if kms is enabled, is told no because nouveau.ko
isn't loaded, and then later nouveau.ko gets loaded with kms enabled by
default and starts fighting with the ddx for the display.
One cool side-effect is that we're not depending on DRIScreenInit for our
DRM fd now, which means that if DRIScreenInit fails for a non-fatal reason
(like, NVIDIA's libglx being around instead of xorg's) we can still run
with full 2D acceleration.
commit bd1448216715681beebabe82b88c8b4d0e62eced
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Wed Mar 11 16:07:41 2009 +1000
kms: cleanup set_property
commit 7a455929785e00beb2d1e65007a3407427ca693a
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Wed Mar 11 15:51:59 2009 +1000
kms: get current values for output properties on startup
commit 7095e8942c74953fcf1615900e2d7864024ece5c
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Wed Mar 11 11:30:08 2009 +1000
kms: work around some xserver stupidity with DPMS
The server calls the output DPMS functions with the intention of
turning off outputs before reprogramming them. It probably shouldn't
be doing this in the case of the randr driver using set_mode_major,
and leave it up to the driver.
The DRM modesetting will usually reenable the outputs after a modeset
anyway, which is the behaviour that the xserver currently relies on.
However, there's a shortcut used in the drm to allow far faster fb
swaps without doing a modeset. When this path is hit, the DRM will
rightly not touch any power states and the display will never get
turned back on.
commit 8f9a58070c2fc55c922e9d82d0796e32959a8431
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Mar 10 00:20:35 2009 +0000
Another DCB 1.5 entry (seen in rh#455194)
Plus a few minor PLL tidyings
commit 9a719900adcebf47442a30fe1989509f93e13e16
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Mon Mar 9 15:57:06 2009 +0000
randr12: add output properties for digital vibrance and image sharpening
commit 678ddc2e56aea3deb1e95b2309b47e1f2cce0635
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Mon Mar 9 15:34:03 2009 +0000
randr12: functionate output properties creation
notes:
* setting pending at change time is necessary for props created as pending,
so that PostPending doesn't overwrite the correct current state with the
incorrect pending one
* our call to PostPending can go; relic from when props were updated in
update_output_fields
commit ed9bd8848fcfe6a77894bfe520a5aa25a9b3f769
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Wed Mar 4 15:10:24 2009 +1000
kms: remove early-exit from gamma_set, the kernel side works now
commit 98c9e4edb58374f18249e5c9c53b392fb8b4a1d1
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Wed Mar 4 13:40:49 2009 +1000
kms: work around some bong hits with dpms
This'll need to get fixed elsewhere, but in the meantime, we have working
DPMS *and* can actually see what's happening when X starts!
commit 88ae2c92c0efc1919dbaa7e8db4c37b86987046c
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Wed Mar 4 13:36:36 2009 +1000
kms: implement dpms
commit 5269cc8060e0ce5160dfca408b0807a4cf2ff17c
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sun Mar 8 20:07:23 2009 +0000
randr12: make props code slightly less awful to behold
commit 0b87c4964076101a2d8373fd3addd9355fb7765e
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sat Mar 7 21:13:55 2009 +0000
More robust PLL upclocking for old cards
commit cb237a34f31f8b8c841974cb64141a2f2e36c53b
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sat Mar 7 04:38:12 2009 +0000
bios/randr12: mode validation fixes (includes fix for #20298)
Remove arbitrary 12000kHz mode clock lower bound (#20298)
Use DCB supplied max dac clock (and remove a memcpy)
commit 9cfccd7281a715e45a4296226d9d52854230f346
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sat Mar 7 04:34:20 2009 +0000
Handle failure to calculate pll values better
Ensure pll limit table get parsed on nv30 and nv35 too
commit 1e994400913bc656b34440df67aa105aa2b211c4
Author: Stuart Bennett <sb476@cam.ac.uk>
Date: Fri Jan 16 22:26:28 2009 +0000
xv: return BadAlloc if destination pixmap is not put into EXA offscreen (#15792)
Lets the player get X errors, instead of the card choking trying to do system
memory addresses
We could use exaDrawableIsOffscreen() here, but that's only in new xservers.
commit 77347da64f86822f128e0dae3e89a6b80282c4ba
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Feb 27 01:22:13 2009 +0000
Remove duplicated DCB parsing
Store stuff in DCB private instead
commit 1b6cabeee4c0160335c174bd4e9111dbd9535aaa
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Mar 3 01:03:45 2009 +0000
Remove VBIOS access in parse_dcb_entry
commit 72fd0ae537e4db208bbfa142dd70144901d7f267
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Mar 3 21:51:46 2009 +0000
randr12: no need to re-detect edid for lvds, it's not going away
Reuse of nv_connector->edid is safe as nv_lvds_output_get_modes doesn't call
nv_output_get_edid_modes after a mode has been found once
commit ae981a8075395af40cf530f2aa24ac68d2ca0e77
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Thu Mar 5 20:21:39 2009 +0100
nv50: some minor modesetting changes
commit 42f99e68940a526f804e2046676c1d06d1bbbb42
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Thu Mar 5 12:11:03 2009 +1100
nv30/xv: restore original viewport/clipping after putimage
Fixes two cases of corruption I seen while using the texture adaptors with
a composite manager.
- desktop getting clipped to video size sometimes
- missing fonts on some menu items after xv+compmgr
commit 56bb8f8e81e9d152217df7863e59781009a9bb77
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Mar 4 19:48:29 2009 +0000
Allow parse_fp_mode_table for x86 cards that find themselves on ppc
commit 1c2f33ed33791f391d059778b1490222e8cd1e1d
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Mar 4 19:40:37 2009 +0000
Don't rely on the bios mobile feature bit for BMP biosen (#19986)
Quadro bios builds also set the mobile bit (not on nv10gl, but for all
observed subsequent Quadros before nv4x (intro of BIT bios structure).
So for BMP cards, use bit 6 of cr4b on head A to detect mobile biosen (all
sampled mobile bios images set this in their init scripts, and non-mobile
don't)
commit e436c7ee98c2eda9d5ba45f7d897d68497b52799
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Mar 4 20:59:42 2009 +0000
randr12: make treatment of CR4B more closely resemble blob's behaviour
Possibly totally unnecessary anyway
commit c61102b326455f6c63a55c8c73ed629994281588
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Feb 18 01:17:29 2009 +0000
Parse bios fp mode during modesetting, removing extra alloc
Also, it seems that the mode parsing was wrong on big endian for years, so
fix that (though Apple laptops have DDC panels anyway)
commit 631e15a03b8a790d342aefe677f8c5eb33935de6
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Mar 3 22:03:00 2009 +0000
Get dual_link and dithering bios info through lvds table parsing function
Now that lvds table is only parsed during modesetting, return the parsed
data in out parameters to make the ordering requirements obvious
Since calling this on nv50 is pointless, that code was commented out
commit 2d4377193f5e2166af3df32ddcfb1235be82764e
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Mar 3 21:52:04 2009 +0000
Call parse_lvds_manufacturer_table during modesetting only (#19986 partial fix)
Also remove SetOwner call now that locking crtcs doesn't change the set head
Moving parse_lvds_manufacturer_table to modesetting means it only gets
called for lvds outputs, removing the dependency on checking for the mobile
bios build. Another benefit is making the strap and EDID FP paths common.
A side effect is that as the lvds dithering toggle depends on the results of
parse_lvds_manufacturer_table, setting that moves to after calling the
parsing function
commit ce8f5d71dacbe0d56cfcb17dfe60f1c4c6cf1773
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Mar 3 17:24:09 2009 +0000
Variety of LVDS script changes
Remove lvds invocation static
No LVDS_INIT for BIT biosen
Merge in one-time call to LVDS_INIT script to main script function
Make sel_clk preservation stuff use regular reg writing functions
commit 071d5373df744c0c9578134fa634bce1e202d4f3
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Feb 17 16:45:50 2009 +0000
Read embedded bios edid only on demand
commit f04689cb326da409448d54694e2660cf4b4b7950
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Tue Mar 3 10:14:26 2009 +1000
move memset of fb after NVSave(), hopefully fix lost VGA fonts
commit 6961efb84dd6894aaaf010a9f620231dfcbad004
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Mar 2 11:35:05 2009 +1000
kms: fix check for xf86drmMode.h
commit 76fa6569ffd1a7f61e295753383bd01b246b5348
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Mon Mar 2 16:43:06 2009 +0000
Detypedef bios_t, rename bios io functions, group various helper functions
commit 39c9d0231c0d24bf91dbe9ab262f54fcad1b3b2f
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Feb 27 18:57:35 2009 +0000
Move parsed DCB to bios structs, split into public and private parts
commit 275daaf7be46e54ca59067d3f2a6e6790e1ce7e0
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Feb 27 17:46:26 2009 +0000
Split bios struct into struct for nv_bios.c, and exported public info struct
commit 8b3e966e3230add04aa05a9ded3f8a647afe0d91
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sun Mar 1 20:50:15 2009 +0000
Remove never updated bios opcodes 0x31, 0x5A and 0x73
Notices of removal left in case any bios using them ever appears, so they
can be found in git history
Also nuke a few unused reg access defines
commit ce9eb05f516899f56e66cc290fc7c4f9d3085882
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Feb 27 23:14:24 2009 +0000
Trivial bios convenience variable changes
commit cfeef1144943e72d62a2c0d9701ddf4a639deeca
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Fri Feb 27 17:30:40 2009 +0100
exa: forgot to check if EXA_SUPPORTS_PREPARE_AUX was defined
commit 37a85dcbd6f6134a43b5c9bb883553fdaf7c0bd9
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Fri Feb 27 17:23:42 2009 +0100
exa: we support PREPARE_AUX
commit 1249fd25a529702a4c6fb4831ef92e801f0b6122
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Fri Feb 27 17:21:47 2009 +0100
Revert "exa: support major version 3"
This reverts commit b7f99aec7282ca7c81d19ff3ad7d9ff22275d476.
commit 1c4a284a80ebed9f9d1e01c47b929481801566b5
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Feb 20 20:02:38 2009 +0000
Rework PLL decoding in terms of pll_vals, save/restore pll_vals on vt switch
If your nv04-nv4x vt after running nouveau is more broken than it was with
the previous commit, file a bug
commit 81b4732162915ab1ac0593e0ef21b56b089dafae
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Feb 20 19:11:41 2009 +0000
New common struct for pll values
Makes the compiler do our dirty accessing work for us
commit adf2e350341b983d27123550041c3b1ddaac4526
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Feb 25 02:11:27 2009 +0000
Check whether NV_RAMDAC_580 has changed when setting PLLs
commit b7f99aec7282ca7c81d19ff3ad7d9ff22275d476
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Tue Feb 24 19:13:23 2009 +0100
exa: support major version 3
commit d91fc78859b11d00c822e5cae70f0fc511b26701
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Tue Feb 24 08:31:40 2009 +1000
kms: fix resize func
Don't change virtualX/virtualY when we're not resizing the actual framebuffer,
just change displayWidth to account for the adjusted scanout size.
With driver pixmaps, we'll be able to do "proper" framebuffer resize.
commit 535498bcfe79858e08ae1f01125025f56bcd504f
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Feb 23 08:20:34 2009 +1000
clear the scanout buffer on statup to avoid seeing garbage on-screen
commit 022767373474a0dacf50ee6fcd6401db15018aca
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Fri Feb 20 13:59:58 2009 +1000
nv50/exa: properly align offset for mem_tile ioctl
I really really hate this ioctl()..
commit e09f50e5f9126c682289b7ba790f16b93a80b876
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Feb 18 21:04:00 2009 +0000
bios/randr12: fix dual link TMDS (#20006)
TMDS04 needs setting, just as for LVDS, and bit 31 of fp_control should also
be set (although in this instance the scripts ensure it is set)
Furthermore, stop dpms clobbering any changes made to fp_control by scripts
Many thanks to bmaass for testing
commit b968db77037790dd5be91d6fafa0552eb7e9520f
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Mon Feb 16 21:38:31 2009 +0000
Return of "bios: fix create_i2c_device for g80"
commit 0a439ba05aa192e7433d430acd91b60719d20caf
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Feb 11 20:52:43 2009 +0000
Parse DCB before running init
commit 870b6f0d2df3e64af5ac921594b1742f82329c49
Author: Stuart Bennett <sb476@cam.ac.uk>
Date: Tue Jan 20 05:29:38 2009 +0000
Split parsing and init into separate functions
commit 0280a0c975830b5c4b2edf4fff8a79fb7d67b9df
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Feb 11 18:53:56 2009 +0000
Attempt to remove pNv and bios_t use in parsing the DCB
It's meant to parse a section of rom image into useful things, so use of pNv
ought to be unusual.
Access to things contained in pNv and bios_t makes it harder to prove parsing
ordering requirements
commit 9179c6667cf859248a525c8d01922b69ed45c17b
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Wed Feb 11 18:33:47 2009 +0000
Variety of DCB enhancements
* ensure parsing of dcb_table.i2c_default_indices happens
* only merge entries on dcb versions < 2.1, and note consequnces of merging
* correct dcbent->index on post-merge compacted entries
* drop pointless memcpy
* return an error if no outputs found
* define dcb structure in nouveau_bios.h (ought to rename pNv->dcb_table to pNv->dcb one day)
commit 317ca477cdbb936f29360984bcee25a0ecda3aa3
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Mon Feb 16 21:15:51 2009 +0000
Revert "bios: fix create_i2c_device for g80"
This reverts commit ca72ca24784b9da09b6e3e0b39f3aa32b8f7b502.
It'll be back in a minute.
commit 7b25a3019beba6342505fa16327a5bcd43669f90
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Mon Feb 16 21:23:56 2009 +0100
nv04: unbreak copy
commit 60c8bb1391a51188a4981b4936c190c149b79b7f
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Feb 16 13:09:39 2009 +1000
default to autodetecting whether to enable the driver's kms paths
commit a31b1d0118dc7b4d07a2c232fcd5aafbf94f69b9
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Feb 16 09:49:39 2009 +1000
exa: reimplement driver-controlled pixmaps
commit d97e993072dc773ecfeb80b5e138b5e4598a1c3d
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Mon Feb 16 08:50:02 2009 +1000
exa: preparation for reintroducing driver-controlled pixmaps
commit 1e00c7d1e6edb7f6b4e6d12ef227d27413948f8f
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Sun Feb 15 18:52:59 2009 +0100
shadowfb: clip copied regions to frontbuffer size.
- There have been reports of crashed when moving windows outside the visible screen.
commit efe27964fa25f71dcbdc7a751210860369ea0e11
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Sun Feb 15 16:54:18 2009 +0100
nv50: correct a poorly written comment
commit 2573c065423af47e14263479c66f2308259f29b8
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Fri Feb 13 10:46:13 2009 +1000
exa: remove exaWaitSync() call in nouveau_exa_pixmap_map()
unnecessary, nouveau_bo_map() will sync if required.
commit 500053517cb5fe058468c8965f43e6c3209ba7b5
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Thu Feb 12 09:04:43 2009 +1000
exa: don't need exaMarkSync on DFS, the operation will be done already
commit a7e7c557dbf9b2fe54db401793608128de08bdf5
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Thu Feb 12 09:01:49 2009 +1000
xv: remove notifier usage
commit 8ecb8d5331bdb910beea416cdb0b662c624c5c96
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Thu Feb 12 09:00:07 2009 +1000
exa: remove notifier waits in UTS/DFS
nouveau_bo will handle CPU/GPU sync if necessary when bo_map() is called.
commit cfb884b816eebbdaa2103dd233c6f975834e536d
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Fri Feb 13 22:29:33 2009 +0100
bios: add some more valid regs
commit ca72ca24784b9da09b6e3e0b39f3aa32b8f7b502
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Fri Feb 13 22:18:18 2009 +0100
bios: fix create_i2c_device for g80
commit bbd7041791eacd357afa775236d4123dd98ab9c9
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Fri Feb 13 20:30:24 2009 +0100
bios: implement opcode 0x76 (INIT_IO_CONDITION).
commit 5dd2cbe454ee8560c814f3fd96d7c28e0fec1dc1
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Thu Feb 12 17:13:42 2009 +0100
exa: add nouveau_exa_pixmap_is_tiled
commit 15ccf721791fa4f61784ad00449630c6d2496328
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sat Feb 7 16:40:46 2009 +0000
randr12: anti-crash measures for #19854
commit 17e9d4bb8c46d67e201b2a369eff7db805dff07b
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Fri Apr 4 19:01:20 2008 +0100
randr12: use bios PLL setting routines for modesetting
Only used for setting modes currently, old code remains for restore
commit e444e0e348b3aab59671ffdeb605f9a57c46bf70
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Jan 20 03:23:40 2009 +0000
Readability improvements for nv_hw.c and nvreg.h
commit f5ea66c2ef2c726b6a0aff72843c379ab25dbaff
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Feb 10 19:01:43 2009 +0000
Pull in extras from CRTC PLL code to bios routines
* g7x magic bits
* 0xc040 manipulation for vplls
commit 416636509152cf12c801ac537cb8c1bd20c19b5e
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Jan 20 02:11:08 2009 +0000
Pull out ramdac580 setting and powerctrl_1 shift code to separate functions
And only twiddle powerctrl_1 for vplls on nv3x single stage pll chips
commit 8b7041833c11f32feb7fa5edc21648cc9db1a333
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Tue Feb 10 19:01:05 2009 +0000
Miscellaneous fixes to bios PLL code
* bios variable was only used for chip_version. simplify.
* set plls faster on single pll cards after nv15/16
* allow early return for single stage plls when already set
* single stage pll mode is only for nv41+
* 80 cols fixes
commit 01bb4c1576a45217953632c9f5612aea9173f95f
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Mon Feb 9 21:14:43 2009 +0000
Make PROM score badly, out of available bios images, if checksum bad
Problem exposed on a Mac following 0387ac32
Having PROM first is good, if the image is fine. But if the read-only PROM
has problems, not a good idea to use it if there are other options available
Also avoid result contamination, if a read method has returned early and not
overwritten previous image in *data
commit 30848dd378d05917c646a24df404846e6f9b2352
Author: Stuart Bennett <sb476@cam.ac.uk>
Date: Tue Jan 6 16:23:33 2009 +0000
randr12: bodge-in continuous timings when the scaler is in use (#19203)
Similar to 3354e660b0744976871683ce226f17e873f26b50 in the intel driver, except
I can't see much benefit in allowing bigger-than-native resolutions, so no sync
range hacking
A pity the xserver has no idea of scalers and so each driver ends up with a hack
commit 0387ac32bef6c714f54917d5d36071ae1224458c
Author: Stuart Bennett <stuart@freedesktop.org>
Date: Sun Feb 8 09:19:53 2009 +0000
Allow reading of PROM and PCI rom on PPC (for people using x86 cards)
commit 6ad66c0b6165822dde167fd1700bd14727ce7087
Author: Stuart Bennett <sb476@cam.ac.uk>
Date: Sun Jan 25 20:09:36 2009 +0000
bios/randr12: be strict about when to use modes from bios vs modes from ddc
Again, the Inspiron 8100 in #14567 is an example of wackiness, having a
non-0xf i2c_index, but using straps
If the previous commit didn't break your DCB 2.0 EDID equipped laptop, this will
commit b5391eb909f8b477c3c438f1136d788acfe6bfd8
Author: Stuart Bennett <sb476@cam.ac.uk>
Date: Sun Jan 25 03:03:29 2009 +0000
Simplify FP strapping semantics and trust bios parsing of use_straps_for_mode
In many ways a spiritual revert of 11de8d
The compound fp.strapping member faithfully implemented what nvidia sticks in
cr58 (for cr57 == 0xf), but this was never directly useful and had differing
overloaded interpretations with bios version, causing confusion. Return to
just using the result of get_fp_strap (g80 code of which unused and so dropped)
As noted in the commit, there have been DCB 2.0 laptops (as in bug 14567) which
don't set use_straps_for_mode when they should, so we set it for all DCB 1.4-2.0
cards. If EDID-ful DCB 1.4-2.0 laptops exist they will be broken, and will need
fixing later (simplistic fix: use_straps_for_mode = (get_fp_strap() == 0xf) )
Will also break nv40 straps-based laptops whose tables used reserved values,
but that's pretty broken anyway, and I've no evidence they exist
commit 34c2b1539bb2d69dd3bd5911c4fac62e46385d20
Author: Stuart Bennett <sb476@cam.ac.uk>
Date: Sat Jan 24 18:41:31 2009 +0000
bios/randr12: misc (mainly lvds related) clarifications and cleanups
get_fp_strap() moved (& unused g80 bodge removed) in anticipation of next commit
commit 98b8cada6c355d437925a92ef0413e96751ed567
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Sat Feb 7 21:02:31 2009 +0100
Always allocate 2 hw cursors.
- This hopefully fixes the legacy codepath again.
commit f1099810bb3cfe451877667a0224eb3d664e442f
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Sat Feb 7 20:04:54 2009 +0100
Fix a few warnings.
commit 691589d8c05b8b0354e359b4942711f6fffa9903
Author: Maarten Maathuis <madman2003@gmail.com>
Date: Sat Feb 7 19:59:28 2009 +0100
Fix some cursor fallout, mostly in legacy code.
commit 945f0cb644b6e4f34cc04927177fe53529aeb5cf
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Fri Feb 6 00:44:14 2009 +1000
kms: update drmmode_display
commit ba1f897159d910a25560796d11c9d0357869e1c0
Author: Ben Skeggs <skeggsb@gmail.com>
Date: Thu Feb 5 23:15:24 2009 +1000
keep resources around after leavevt, far too problematic currently
Reply to: