Re: Linux 5.10-rc4; graphics alignment
Hi Am 24.11.20 um 17:27 schrieb David Laight: From: David Laight Sent: 20 November 2020 15:39 From: Thomas Zimmermann Sent: 20 November 2020 13:42 ... I did a diff from v5.10-rc4 to drm-tip to look for suspicious changes. Some candidates are 8e3784dfef8a ("drm/ast: Reload gamma LUT after changing primary plane's color format") Ok, that one fixes the screen colours (etc). So 8e3784dfef8a was good and then HEAD^ was bad. I might try to bisect the breakage. The stack splat is entirely different. I'll try to bisect that on Linus's tree. The good news is I'm not getting the stack splat on rc5. I'm not sure I can be bothered to find out when :-) Applying 8e3784dfef8a to rc5 by hand also fixes the display colours. I've added this commit to drm-misc-fixes and it should show up in the upstream kernel soonish. Best regards Thomas David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer OpenPGP_0x680DC11D530B7A23.asc Description: application/pgp-keys OpenPGP_signature Description: OpenPGP digital signature ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
RE: Linux 5.10-rc4; graphics alignment
From: David Laight > Sent: 20 November 2020 15:39 > > From: Thomas Zimmermann > > Sent: 20 November 2020 13:42 > ... > > I did a diff from v5.10-rc4 to drm-tip to look for suspicious changes. > > Some candidates are > > > >8e3784dfef8a ("drm/ast: Reload gamma LUT after changing primary > > plane's color format") > > Ok, that one fixes the screen colours (etc). > So 8e3784dfef8a was good and then HEAD^ was bad. > > I might try to bisect the breakage. > > The stack splat is entirely different. > I'll try to bisect that on Linus's tree. The good news is I'm not getting the stack splat on rc5. I'm not sure I can be bothered to find out when :-) Applying 8e3784dfef8a to rc5 by hand also fixes the display colours. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: Linux 5.10-rc4; graphics alignment
Hi Am 24.11.20 um 17:27 schrieb David Laight: From: David Laight Sent: 20 November 2020 15:39 From: Thomas Zimmermann Sent: 20 November 2020 13:42 ... I did a diff from v5.10-rc4 to drm-tip to look for suspicious changes. Some candidates are 8e3784dfef8a ("drm/ast: Reload gamma LUT after changing primary plane's color format") Ok, that one fixes the screen colours (etc). So 8e3784dfef8a was good and then HEAD^ was bad. I might try to bisect the breakage. The stack splat is entirely different. I'll try to bisect that on Linus's tree. The good news is I'm not getting the stack splat on rc5. I'm not sure I can be bothered to find out when :-) Applying 8e3784dfef8a to rc5 by hand also fixes the display colours. Sounds good. I'm counting this as solved then. Thanks for putting all this work into it. Best regards Thomas David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer OpenPGP_0x680DC11D530B7A23.asc Description: application/pgp-keys OpenPGP_signature Description: OpenPGP digital signature ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
RE: Linux 5.10-rc4; graphics alignment
From: Thomas Zimmermann > Sent: 20 November 2020 13:42 ... > I did a diff from v5.10-rc4 to drm-tip to look for suspicious changes. > Some candidates are > >8e3784dfef8a ("drm/ast: Reload gamma LUT after changing primary > plane's color format") Ok, that one fixes the screen colours (etc). So 8e3784dfef8a was good and then HEAD^ was bad. I might try to bisect the breakage. The stack splat is entirely different. I'll try to bisect that on Linus's tree. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
RE: Linux 5.10-rc4; graphics alignment
From: Thomas Zimmermann > Sent: 20 November 2020 10:14 ... > > Is there any way to bisect through the parts of the > > drm merge patch into v5.10-rc1 ? > > > > That ought to be quicker (and less error prone) than > > the bisect builds I was doing. > > > > Note that the stack 'splat' is due to a later change. > > It is separate from the broken pixel alignment. > > > > I actually saw the vga text go 'funny' while the boot > > was outputting all the [OK] messages (from systemd?) > > before the graphic login stole tty1 (bloody stupid > > to use tty1). > > > > I don't need to use the failing system today, I'll > > have another go at isolating the failure. > > You can use drm-tip for testing, where many of the DRM patches go through. > >https://cgit.freedesktop.org/drm/drm-tip/ > > It's fairly up-to-date. Any idea of tags either side of the 5.10 merge? > I have two systems with AST chips and neither shows any of the symptoms > you describe; nor do we have such reports about drivers that use a > similar stack (hibmc, bochs). Could you provide the output of > >dmesg | grep drm [2.112303] fb0: switching to astdrmfb from EFI VGA [2.120222] ast :02:00.0: [drm] Using P2A bridge for configuration [2.120233] ast :02:00.0: [drm] AST 2400 detected [2.120247] ast :02:00.0: [drm] Analog VGA only [2.120257] ast :02:00.0: [drm] dram MCLK=408 Mhz type=1 bus_width=16 [2.121121] [drm] Initialized ast 0.1.0 20120228 for :02:00.0 on minor 0 [2.125838] fbcon: astdrmfb (fb0) is primary device [2.152179] ast :02:00.0: [drm] fb0: astdrmfb frame buffer device [6.061034] systemd[1]: Condition check resulted in Load Kernel Module drm being skipped. The output is the same for both good and bad kernels. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
RE: Linux 5.10-rc4; graphics alignment
From: Thomas Zimmermann > Sent: 20 November 2020 11:27 ... > >> You can use drm-tip for testing, where many of the DRM patches go through. > >> > >> https://cgit.freedesktop.org/drm/drm-tip/ > >> > >> It's fairly up-to-date. > > > > Any idea of tags either side of the 5.10 merge? > > The final commit before v5.9 appears to be > >Fixes: 33c8256b3bcc ("drm/amd/display: Change ABM config init interface") > > I'd try this as a good commit. For the bad commit, just try HEAD. HEAD off that tree works. Colours ok and no stack backtrace. Ideas?? David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
RE: Linux 5.10-rc4; graphics alignment
> Hi David > > Am 18.11.20 um 23:01 schrieb David Laight: ... > Did you try Daniel's suggestion of testing with the direct parent commit? (I was on holiday yesterday and didn't want to spend a sunny afternoon doing bisects.) I've just done that and it is bad. Is there any way to bisect through the parts of the drm merge patch into v5.10-rc1 ? That ought to be quicker (and less error prone) than the bisect builds I was doing. Note that the stack 'splat' is due to a later change. It is separate from the broken pixel alignment. I actually saw the vga text go 'funny' while the boot was outputting all the [OK] messages (from systemd?) before the graphic login stole tty1 (bloody stupid to use tty1). I don't need to use the failing system today, I'll have another go at isolating the failure. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
RE: Linux 5.10-rc4; graphics alignment
From: Thomas Zimmermann > Sent: 20 November 2020 12:30 > > Am 20.11.20 um 12:45 schrieb David Laight: > > From: Thomas Zimmermann > >> Sent: 20 November 2020 11:27 > > ... > You can use drm-tip for testing, where many of the DRM patches go > through. > > https://cgit.freedesktop.org/drm/drm-tip/ > > It's fairly up-to-date. > >>> > >>> Any idea of tags either side of the 5.10 merge? > >> > >> The final commit before v5.9 appears to be > >> > >> Fixes: 33c8256b3bcc ("drm/amd/display: Change ABM config init > >> interface") > >> > >> I'd try this as a good commit. For the bad commit, just try HEAD. > > > > HEAD off that tree works. > > Colours ok and no stack backtrace. > > > > Ideas?? > > The good news is that it's been fixed. All you have to do is wait for > the fix to hit upstream. > > Did you try the patch that Dave linked? That patch makes no difference to my system. The condition is false so it doesn't corrupt the flags. (I printed the values to see.) David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: Linux 5.10-rc4; graphics alignment
Hi Am 20.11.20 um 13:53 schrieb David Laight: From: Thomas Zimmermann Sent: 20 November 2020 12:30 Am 20.11.20 um 12:45 schrieb David Laight: From: Thomas Zimmermann Sent: 20 November 2020 11:27 ... You can use drm-tip for testing, where many of the DRM patches go through. https://cgit.freedesktop.org/drm/drm-tip/ It's fairly up-to-date. Any idea of tags either side of the 5.10 merge? The final commit before v5.9 appears to be Fixes: 33c8256b3bcc ("drm/amd/display: Change ABM config init interface") I'd try this as a good commit. For the bad commit, just try HEAD. HEAD off that tree works. Colours ok and no stack backtrace. Ideas?? The good news is that it's been fixed. All you have to do is wait for the fix to hit upstream. Did you try the patch that Dave linked? That patch makes no difference to my system. The condition is false so it doesn't corrupt the flags. (I printed the values to see.) I see, so it's probably something else that has been fixed in drm-tip. It's likely that at least one commit in drm-tip is also broken. I'd try to find one of these. One idea is to go through the commits that went into the related files. Maybe go back to the HEAD of drm-tip and try something like git log --oneline -- drivers/gpu/drm/ast/ drivers/gpu/drm/drm_gem_vram_helper.c drivers/gpu/drm/ttm/ That lists recent ast and memory-management commits. One of them might be broken and can serve an a starting point for a bisect. I did a diff from v5.10-rc4 to drm-tip to look for suspicious changes. Some candidates are 8e3784dfef8a ("drm/ast: Reload gamma LUT after changing primary plane's color format") 2b8283ff1a60 ("drm/vram_helper: implement a ttm move callback.") 6a6e5988a265 ("drm/ttm: replace last move_notify with delete_mem_notify") I'd test them against their parent commits to see if anything changes. Thanks for all your efforts! Best regards Thomas David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer OpenPGP_0x680DC11D530B7A23.asc Description: application/pgp-keys OpenPGP_signature Description: OpenPGP digital signature ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: Linux 5.10-rc4; graphics alignment
Hi Am 20.11.20 um 12:45 schrieb David Laight: From: Thomas Zimmermann Sent: 20 November 2020 11:27 ... You can use drm-tip for testing, where many of the DRM patches go through. https://cgit.freedesktop.org/drm/drm-tip/ It's fairly up-to-date. Any idea of tags either side of the 5.10 merge? The final commit before v5.9 appears to be Fixes: 33c8256b3bcc ("drm/amd/display: Change ABM config init interface") I'd try this as a good commit. For the bad commit, just try HEAD. HEAD off that tree works. Colours ok and no stack backtrace. Ideas?? The good news is that it's been fixed. All you have to do is wait for the fix to hit upstream. Did you try the patch that Dave linked? If not, go back to v5.10-rc4 and do git am The patch is attached. Please report back if this fixes the issue. Best regards Thomas David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer --- Begin Message --- From: Dave Airlie In 7053e0eab473119503f6565b4e398f9a73122481 drm/vram-helper: stop using TTM placement flags it appears the flags got mixed up. This should fix a regression on ast [ 64.782340] WARNING: CPU: 51 PID: 1964 at drivers/gpu/drm/drm_gem_vram_helper.c:284 drm_gem_vram_offset+0x35/0x40 [drm_vram_helper] [ 64.782411] CPU: 51 PID: 1964 Comm: Xorg Not tainted 5.10.0-rc3 #12 [ 64.782413] Hardware name: To be filled. [ 64.782419] RIP: 0010:drm_gem_vram_offset+0x35/0x40 [drm_vram_helper] [ 64.782424] Code: 00 48 89 e5 85 c0 74 17 48 83 bf 78 01 00 00 00 74 18 48 8b 87 80 01 00 00 5d 48 c1 e0 0c c3 0f 0b 48 c7 c0 ed ff ff ff 5d c3 <0f> 0b 31 c0 5d c3 0f 1f 44 00 00 0f 1f 44 00 00 55 48 8b 87 18 06 [ 64.782427] RSP: 0018:a9128909fa68 EFLAGS: 00010246 [ 64.782431] RAX: 0002 RBX: 95a5c25e1ec0 RCX: c02b6600 [ 64.782433] RDX: 959e49824000 RSI: 95a5c25e0b40 RDI: 959e4b1c2c00 [ 64.782434] RBP: a9128909fa68 R08: 0040 R09: 95a9c5dcb688 [ 64.782436] R10: R11: 0001 R12: 959e49824000 [ 64.782437] R13: R14: R15: 95a5c5c56f00 [ 64.782440] FS: 7f485d466a80() GS:95a9afcc() knlGS: [ 64.782442] CS: 0010 DS: ES: CR0: 80050033 [ 64.782444] CR2: 7f485e202000 CR3: 000c82a0e000 CR4: 003506e0 [ 64.782446] Call Trace: [ 64.782455] ast_cursor_page_flip+0x22/0x100 [ast] [ 64.782460] ast_cursor_plane_helper_atomic_update+0x46/0x70 [ast] [ 64.782477] drm_atomic_helper_commit_planes+0xbd/0x220 [drm_kms_helper] [ 64.782493] drm_atomic_helper_commit_tail_rpm+0x3a/0x70 [drm_kms_helper] [ 64.782507] commit_tail+0x99/0x130 [drm_kms_helper] [ 64.782521] drm_atomic_helper_commit+0x123/0x150 [drm_kms_helper] [ 64.782551] drm_atomic_commit+0x4a/0x50 [drm] [ 64.782565] drm_atomic_helper_update_plane+0xe7/0x140 [drm_kms_helper] [ 64.782592] __setplane_atomic+0xcc/0x110 [drm] [ 64.782619] drm_mode_cursor_universal+0x13e/0x260 [drm] [ 64.782647] drm_mode_cursor_common+0xef/0x220 [drm] [ 64.782654] ? tomoyo_path_number_perm+0x6f/0x200 [ 64.782680] ? drm_mode_cursor_ioctl+0x60/0x60 [drm] [ 64.782706] drm_mode_cursor2_ioctl+0xe/0x10 [drm] [ 64.782727] drm_ioctl_kernel+0xae/0xf0 [drm] [ 64.782749] drm_ioctl+0x241/0x3f0 [drm] [ 64.782774] ? drm_mode_cursor_ioctl+0x60/0x60 [drm] [ 64.782781] ? tomoyo_file_ioctl+0x19/0x20 [ 64.782787] __x64_sys_ioctl+0x91/0xc0 [ 64.782792] do_syscall_64+0x38/0x90 [ 64.782797] entry_SYSCALL_64_after_hwframe+0x44/0xa9 Signed-off-by: Dave Airlie Cc: Wen Pu Cc: David Laight Cc: Christian König --- drivers/gpu/drm/drm_gem_vram_helper.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/drm_gem_vram_helper.c b/drivers/gpu/drm/drm_gem_vram_helper.c index 50cad0e4a92e..2896a057b771 100644 --- a/drivers/gpu/drm/drm_gem_vram_helper.c +++ b/drivers/gpu/drm/drm_gem_vram_helper.c @@ -140,7 +140,7 @@ static void drm_gem_vram_placement(struct drm_gem_vram_object *gbo, unsigned int c = 0; if (pl_flag & DRM_GEM_VRAM_PL_FLAG_TOPDOWN) - pl_flag = TTM_PL_FLAG_TOPDOWN; + invariant_flag = TTM_PL_FLAG_TOPDOWN; gbo->placement.placement = gbo->placements; gbo->placement.busy_placement = gbo->placements; -- 2.20.1 ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel --- End Message --- OpenPGP_0x680DC11D530B7A23.asc Description: application/pgp-keys OpenPGP_signature Description: OpenPGP digital signature ___ dri-devel mai
Re: Linux 5.10-rc4; graphics alignment
Hi Am 20.11.20 um 11:51 schrieb David Laight: From: Thomas Zimmermann Sent: 20 November 2020 10:14 ... Is there any way to bisect through the parts of the drm merge patch into v5.10-rc1 ? That ought to be quicker (and less error prone) than the bisect builds I was doing. Note that the stack 'splat' is due to a later change. It is separate from the broken pixel alignment. I actually saw the vga text go 'funny' while the boot was outputting all the [OK] messages (from systemd?) before the graphic login stole tty1 (bloody stupid to use tty1). I don't need to use the failing system today, I'll have another go at isolating the failure. You can use drm-tip for testing, where many of the DRM patches go through. https://cgit.freedesktop.org/drm/drm-tip/ It's fairly up-to-date. Any idea of tags either side of the 5.10 merge? The final commit before v5.9 appears to be Fixes: 33c8256b3bcc ("drm/amd/display: Change ABM config init interface") I'd try this as a good commit. For the bad commit, just try HEAD. Best regards Thomas I have two systems with AST chips and neither shows any of the symptoms you describe; nor do we have such reports about drivers that use a similar stack (hibmc, bochs). Could you provide the output of dmesg | grep drm [2.112303] fb0: switching to astdrmfb from EFI VGA [2.120222] ast :02:00.0: [drm] Using P2A bridge for configuration [2.120233] ast :02:00.0: [drm] AST 2400 detected [2.120247] ast :02:00.0: [drm] Analog VGA only [2.120257] ast :02:00.0: [drm] dram MCLK=408 Mhz type=1 bus_width=16 [2.121121] [drm] Initialized ast 0.1.0 20120228 for :02:00.0 on minor 0 [2.125838] fbcon: astdrmfb (fb0) is primary device [2.152179] ast :02:00.0: [drm] fb0: astdrmfb frame buffer device [6.061034] systemd[1]: Condition check resulted in Load Kernel Module drm being skipped. The output is the same for both good and bad kernels. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer OpenPGP_0x680DC11D530B7A23.asc Description: application/pgp-keys OpenPGP_signature Description: OpenPGP digital signature ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: Linux 5.10-rc4; graphics alignment
Hi Am 20.11.20 um 10:52 schrieb David Laight: Hi David Am 18.11.20 um 23:01 schrieb David Laight: ... Did you try Daniel's suggestion of testing with the direct parent commit? (I was on holiday yesterday and didn't want to spend a sunny afternoon doing bisects.) Makes sense :) I've just done that and it is bad. Is there any way to bisect through the parts of the drm merge patch into v5.10-rc1 ? That ought to be quicker (and less error prone) than the bisect builds I was doing. Note that the stack 'splat' is due to a later change. It is separate from the broken pixel alignment. I actually saw the vga text go 'funny' while the boot was outputting all the [OK] messages (from systemd?) before the graphic login stole tty1 (bloody stupid to use tty1). I don't need to use the failing system today, I'll have another go at isolating the failure. You can use drm-tip for testing, where many of the DRM patches go through. https://cgit.freedesktop.org/drm/drm-tip/ It's fairly up-to-date. I have two systems with AST chips and neither shows any of the symptoms you describe; nor do we have such reports about drivers that use a similar stack (hibmc, bochs). Could you provide the output of dmesg | grep drm Best regards Thomas David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales) -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer OpenPGP_0x680DC11D530B7A23.asc Description: application/pgp-keys OpenPGP_signature Description: OpenPGP digital signature ___ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel