linux/drivers/gpu, branch v2.6.35.11

revert-drm-radeon-kms-properly-compute-group_size-on-6xx-7xx

2011-02-06T19:04:06Z

Revert drm/radeon/kms: properly compute group_size on 6xx/7xx From: Steve Conklin We discovered a regression for Radeon users in our latest proposed kernel for 2.6.35 (Maverick), and have isolated it to this patch: http://git.kernel.org/?p=linux/kernel/git/longterm/linux-2.6.35.y.git;a=commit;h=b8e9a4a45f8427837f4dba89 +bda4d4e3f3a5c726 We took that patch as part of 2.6.35.10, and one of our testers has reported that our build of that kernel also exhibits the problem. These are mainline kernels built with the Ubuntu configs. http://kernel.ubuntu.com/~kernel-ppa/mainline/v2.6.35.10-maverick/ Our bug report is here: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/703553 Upstream bug report: https://bugzilla.kernel.org/show_bug.cgi?id=24802 Signed-off-by: Andi Kleen

drm/i915/dp: Fix I2C/EDID handling with active DisplayPort to DVI converter

2011-02-06T19:03:50Z

commit 8316f33766a82907c694267ff911e45e256f09f9 upstream. The DisplayPort standard (1.1a) states that: The I2C-over-AUX Reply field is valid only when Native AUX CH Reply field is AUX_ACK (00). When Native AUX CH Reply field is not 00, then, I2C-over-AUX Reply field must be 00 and be ignored. This fixes broken EDID reading when using an active DisplayPort to duallink DVI converter. If the AUX CH replier chooses to defer the transaction, a short read occurs and erroneous data is returned as the i2c reply due to a lack of length checking and failure to check for AUX ACK. As a result, broken EDIDs can look like: 0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef 00: bc bc bc ff bc bc bc ff bc bc bc ac bc bc bc 45 ???.???.???????E 10: bc bc bc 10 bc bc bc 34 bc bc bc ee bc bc bc 4c ???????4???????L 20: bc bc bc 50 bc bc bc 00 bc bc bc 40 bc bc bc 00 ???P???.???@???. 30: bc bc bc 01 bc bc bc 01 bc bc bc a0 bc bc bc 40 ???????????????@ 40: bc bc bc 00 bc bc bc 00 bc bc bc 00 bc bc bc 55 ???.???.???.???U 50: bc bc bc 35 bc bc bc 31 bc bc bc 20 bc bc bc fc ???5???1??? ???? 60: bc bc bc 4c bc bc bc 34 bc bc bc 46 bc bc bc 00 ???L???4???F???. 70: bc bc bc 38 bc bc bc 11 bc bc bc 20 bc bc bc 20 ???8??????? ??? 80: bc bc bc ff bc bc bc ff bc bc bc ff bc bc bc ff ???.???.???.???. ... which can lead to: [drm:drm_edid_block_valid] *ERROR* EDID checksum is invalid, remainder [drm:drm_edid_block_valid] *ERROR* Raw EDID: <3>30 30 30 30 30 30 30 32 38 32 30 32 63 63 31 61 000000028202cc1a <3>28 00 02 8c 00 00 00 00 18 00 00 00 00 00 00 00 (............... <3>20 4c 61 73 74 20 62 65 61 63 6f 6e 3a 20 33 32 Last beacon: 32 <3>32 30 6d 73 20 61 67 6f 46 00 05 8c 00 00 00 00 20ms agoF....... <3>36 00 00 00 00 00 00 00 00 0c 57 69 2d 46 69 20 6.........Wi-Fi <3>52 6f 75 74 65 72 01 08 82 84 8b 96 24 30 48 6c Router......$0Hl <3>03 01 01 06 02 00 00 2a 01 00 2f 01 00 32 04 0c .......*../..2.. <3>12 18 60 dd 09 00 10 18 02 00 00 01 00 00 18 00 ..`............. Signed-off-by: David Flynn Signed-off-by: Andi Kleen [ickle: fix up some surrounding checkpatch warnings] Signed-off-by: Chris Wilson Signed-off-by: Greg Kroah-Hartman

drm/radeon/kms: reorder display resume to avoid problems

2011-02-06T19:03:50Z

commit a93f344d3c04e4b84490c65f2a574387c593be40 upstream. On resume, we were attemping to unblank the displays before the timing and plls had be reprogrammed which led to atom timeouts waiting for things that are not yet programmed. Re-program the mode first, then reset the dpms state. This fixes the infamous atombios timeouts on resume. Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie Signed-off-by: Greg Kroah-Hartman Signed-off-by: Andi Kleen

drm/radeon/kms: fix evergreen asic reset

2011-02-06T19:03:50Z

commit 9f0c4f9c2f835eee1bbb93f96bf9483d56f1892b upstream. Only reset the grbm blocks, srbm tends to lock the GPU if not done properly and in most cases is not necessary. Also, no need to call asic init after reset the grbm blocks. Signed-off-by: Alex Deucher Signed-off-by: Andi Kleen Reviewed-by: Jerome Glisse Signed-off-by: Dave Airlie Signed-off-by: Greg Kroah-Hartman

drm/radeon/kms/evergreen: reset the grbm blocks at resume and init

2011-02-06T19:03:49Z

commit 86f5c9edbb3bac37cc8cee6528a929005ba72aad upstream. This fixes module reloading and resume as the gfx block seems to be left in a bad state in some cases. Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie Signed-off-by: Greg Kroah-Hartman Signed-off-by: Andi Kleen

nouveau: Acknowledge HPD irq in handler, not bottom half

2011-02-06T19:03:40Z

commit ab838338a2a9e0cb8346eb0cab9977be13e8dce5 upstream. The old code generated an interrupt storm bad enough to completely take down my system. Signed-off-by: Andy Lutomirski Signed-off-by: Ben Skeggs Signed-off-by: Greg Kroah-Hartman Signed-off-by: Andi Kleen

drm/radeon/kms: fix vram base calculation on rs780/rs880

2011-02-06T19:03:26Z

commit 8961d52d4cc52edf5672f8f2712c57162b736793 upstream. Avoid overflowing a 32 bit value. Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie Signed-off-by: Greg Kroah-Hartman Signed-off-by: Andi Kleen

drm/kms: remove spaces from connector names (v2)

2011-02-06T19:03:25Z

commit e76116ca9671e2e5239054a40303b94feab585ad upstream. Grub doesn't parse spaces in parameters correctly, so this makes it impossible to force video= parameters for kms on the grub kernel command line. v2: shorten the names to make them easier to type. Reported-by: Sergej Pupykin Cc: Sergej Pupykin Signed-off-by: Alex Deucher Signed-off-by: Dave Airlie Signed-off-by: Greg Kroah-Hartman Signed-off-by: Andi Kleen

drm/radeon/kms: don't apply 7xx HDP flush workaround on AGP

2011-02-06T19:03:25Z

commit f3886f85cfde578f1d0ba6e40ac5f9d70043923b upstream. It should be required for all 7xx asics, but seems to cause problems on some AGP 7xx chips. Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=19002 Signed-off-by: Alex Deucher Signed-off-by: Andi Kleen Reported-and-Tested-by: Duncan <1i5t5.duncan@cox.net> Signed-off-by: Dave Airlie Signed-off-by: Greg Kroah-Hartman

drm/i915/sdvo: Always add a 30ms delay to make SDVO TV detection reliable

2010-12-14T22:40:11Z

commit ba84cd1f2b5dd49bda9300c5a11373f7e14c3c66 upstream. Commit d09c23de intended to add a 30ms delay to give the ADD time to detect any TVs connected. However, it used the sdvo->is_tv flag to do so which is dependent upon the previous detection result and not whether the output supports TVs. Signed-off-by: Chris Wilson Signed-off-by: Greg Kroah-Hartman Signed-off-by: Andi Kleen