Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
On 12/04/13 17:13, Jonathan Nieder wrote: tags 703468 - moreinfo quit Geoff Crompton wrote: I ran the 'cherry-pick' command, and redid the make deb-pkg and dpkg -i steps. When I rebooted onto that kernel it worked fine. So it seems this patch fixes the 3.4.4 kernel. Beautiful. Let's take this upstream. Jonathan reported this to sta...@vger.kernel.org on the same day as above, and Greg KH said the 9a0f938bde74 change is now applied (presumably to the 3.4.y tree), and Luis Henriques replied saying he's queued it for 3.5 as well. I'd still like to know if/how/when this will be fixed for the Wheezy release though. This bug report is not Release Critical, so I doubt it will get in before Wheezy releases (I would like Debian Wheezy to be installable on apple imacs, but I'm not sure it's feasible at this stage of the release process). But will it be queued up for the first point release of Wheezy? Cheers, Geoff -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
On 12/04/13 09:40, Jonathan Nieder wrote: Could you try the attached patch against a 3.4.y kernel? It works like this: 0. prerequisites: apt-get install git build-essential 1. get the kernel history, if you don't already have it: git clone \ https://kernel.googlesource.com/pub/scm/linux/kernel/git/torvalds/linux.git 2. grab point releases: cd linux git remote add stable \ https://kernel.googlesource.com/pub/scm/linux/kernel/git/stable/linux-stable git fetch stable 3. configure, build, test: git checkout stable/linux-3.4.y cp /boot/config-$(uname -r) .config; # current configuration make localmodconfig; # optional: minimize configuration scripts/config --disable DEBUG_INFO make deb-pkg; # optionally with -jnum for parallel build dpkg -i ../name of package; # as root reboot ... test test test ... Hopefully it reproduces the bug. So I followed the above instructions, though I did a 'git checkout v3.4.4' instead of 'git checkout stable/linux-3.4.y'. This produced a linux-image-3.4.4_3.4.4-1_amd64.deb file that I installed. This kernel failed in the same manner, as we hoped. 4. try the patch: cd linux git cherry-pick -x -s 9a0f938bde74bf9e50bd75c8de9e38c1787398cd make deb-pkg; # maybe with -j4 dpkg -i ../name of package; # as root reboot ... test test test ... I ran the 'cherry-pick' command, and redid the make deb-pkg and dpkg -i steps. When I rebooted onto that kernel it worked fine. So it seems this patch fixes the 3.4.4 kernel. What git invocation should I do to test a debian Wheezy kernel? (on Monday morning, as I'm leaving the office now). Cheers, Geoff -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
tags 703468 - moreinfo quit Geoff Crompton wrote: I ran the 'cherry-pick' command, and redid the make deb-pkg and dpkg -i steps. When I rebooted onto that kernel it worked fine. So it seems this patch fixes the 3.4.4 kernel. Beautiful. Let's take this upstream. -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
On 11/04/13 10:54, Geoff Crompton wrote: I've also built a 3.2.39 kernel with some patches stripped. This kernel still does exhibit the problem, suggesting the patches I stripped out are not responsible. Those patches are: bugfix/x86/drm-i915-kick-any-firmware-framebuffers-before-claim.patch bugfix/x86/drm-i915-Only-kick-out-vesafb-if-we-takeover-the-fbc.patch bugfix/x86/drm-i915-add-quirk_invert_brightness-for-ncr-machine.patch bugfix/x86/drm-i915-Close-race-between-processing-unpin-task-an.patch bugfix/all/i915-ensure-that-VGA-plane-is-disabled.patch bugfix/x86/drm-i915-Only-increment-the-user-pin-count-after-suc.patch bugfix/x86/drm-i915-Disable-AsyncFlip-performance-optimisations.patch bugfix/x86/drm-i915-GFX_MODE-Flush-TLB-Invalidate-Mode-must-be-.patch bugfix/x86/drm-i915-dump-UTS_RELEASE-into-the-error_state.patch Here is a quick summary of what kernel's I've tried so far: works 3.8.5 works 3.6.9 fails 3.4.4 fails 3.2.41-2 fails 3.2.39-2 works 3.2.35-2 Looking at snapshots.debian.org, I'll try 3.5.5-1, as it seems so far Jonathan has been trying to get me to bisect in on the bug. I've done some more testing, using kernels from snapshot.debian.org. I downloaded all the available versions between what we'd already tested. I've tried to bisect onto the range. I now know: works 3.8.5 works 3.6.9 works 3.6.6 works 3.6.4 fails 3.5.5-1 fails 3.4.4 fails 3.2.41-2 fails 3.2.39-2 works 3.2.35-2 This suggests the bug is introduced from 3.5.5 upto 3.6.3, and has been backported into a Debian kernel above 3.2.35-2. Hope that is helpful, what is the next step? Cheers, Geoff -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
found 703468 linux/3.5.5-1~experimental.1 fixed 703468 linux/3.6.4-1~experimental.1 tags 703468 + upstream patch moreinfo quit Geoff Crompton wrote: works 3.6.4 fails 3.5.5-1 [...] Hope that is helpful, what is the next step? Very much so! Thanks for this work. 00:02.0 Display controller [0380]: Intel Corporation 2nd Generation Core Processor Family Integrated Graphics Controller [8086:0102] (rev 09) Could you try the attached patch against a 3.4.y kernel? It works like this: 0. prerequisites: apt-get install git build-essential 1. get the kernel history, if you don't already have it: git clone \ https://kernel.googlesource.com/pub/scm/linux/kernel/git/torvalds/linux.git 2. grab point releases: cd linux git remote add stable \ https://kernel.googlesource.com/pub/scm/linux/kernel/git/stable/linux-stable git fetch stable 3. configure, build, test: git checkout stable/linux-3.4.y cp /boot/config-$(uname -r) .config; # current configuration make localmodconfig; # optional: minimize configuration scripts/config --disable DEBUG_INFO make deb-pkg; # optionally with -jnum for parallel build dpkg -i ../name of package; # as root reboot ... test test test ... Hopefully it reproduces the bug. So 4. try the patch: cd linux git cherry-pick -x -s 9a0f938bde74bf9e50bd75c8de9e38c1787398cd make deb-pkg; # maybe with -j4 dpkg -i ../name of package; # as root reboot ... test test test ... Regards, Jonathan From: Chris Wilson ch...@chris-wilson.co.uk Date: Fri, 24 Aug 2012 09:12:22 +0100 Subject: drm/i915: Use the correct size of the GTT for placing the per-process entries commit 9a0f938bde74bf9e50bd75c8de9e38c1787398cd upstream. The current layout is to place the per-process tables at the end of the GTT. However, this is currently using a hardcoded maximum size for the GTT and not taking in account limitations imposed by the BIOS. Use the value for the total number of entries allocated in the table as provided by the configuration registers. Reported-by: Matthew Garrett m...@redhat.com Signed-off-by: Chris Wilson ch...@chris-wilson.co.uk Cc: Ben Widawsky b...@bwidawsk.net Signed-off-by: Daniel Vetter daniel.vet...@ffwll.ch Signed-off-by: Jonathan Nieder jrnie...@gmail.com --- drivers/gpu/drm/i915/i915_gem_gtt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index a135c61..99a7855 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -72,7 +72,7 @@ int i915_gem_init_aliasing_ppgtt(struct drm_device *dev) /* ppgtt PDEs reside in the global gtt pagetable, which has 512*1024 * entries. For aliasing ppgtt support we just steal them at the end for * now. */ - first_pd_entry_in_global_pt = 512*1024 - I915_PPGTT_PD_ENTRIES; + first_pd_entry_in_global_pt = dev_priv-mm.gtt-gtt_total_entries - I915_PPGTT_PD_ENTRIES; ppgtt = kzalloc(sizeof(*ppgtt), GFP_KERNEL); if (!ppgtt)
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
Geoff Crompton wrote: I've tried that. The 3.4.4-1~eperimental.1 linux-image I built from those sources does not boot, it fails in the same manner as the other 3.2 kernels after 3.2.35. That's a comfort. It means that the problem wasn't introduced by the backport. What's next? 3.6.9-1~experimental.1, to narrow down the range to search for a fix. The binary from snapshot.debian.org (linux-image-3.6-trunk-amd64) should install fine, or if you prefer to build from source then there are instructions at [1]. Thanks, Jonathan [1] http://kernel-handbook.alioth.debian.org/ch-common-tasks.html (Note in particular sections 4.2.1.1 and 4.2.5.) -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
On 10/04/13 16:35, Jonathan Nieder wrote: What's next? 3.6.9-1~experimental.1, to narrow down the range to search for a fix. The binary from snapshot.debian.org (linux-image-3.6-trunk-amd64) should install fine, or if you prefer to build from source then there are instructions at [1]. Thanks, Jonathan [1] http://kernel-handbook.alioth.debian.org/ch-common-tasks.html (Note in particular sections 4.2.1.1 and 4.2.5.) I've downloaded the deb file from snapshot.debian.org for 3.6.9-1~experimental.1 and tried that. It booted fine, this version did not show the problem we're exploring in this ticket. I've also built a 3.2.39 kernel with some patches stripped. This kernel still does exhibit the problem, suggesting the patches I stripped out are not responsible. Those patches are: bugfix/x86/drm-i915-kick-any-firmware-framebuffers-before-claim.patch bugfix/x86/drm-i915-Only-kick-out-vesafb-if-we-takeover-the-fbc.patch bugfix/x86/drm-i915-add-quirk_invert_brightness-for-ncr-machine.patch bugfix/x86/drm-i915-Close-race-between-processing-unpin-task-an.patch bugfix/all/i915-ensure-that-VGA-plane-is-disabled.patch bugfix/x86/drm-i915-Only-increment-the-user-pin-count-after-suc.patch bugfix/x86/drm-i915-Disable-AsyncFlip-performance-optimisations.patch bugfix/x86/drm-i915-GFX_MODE-Flush-TLB-Invalidate-Mode-must-be-.patch bugfix/x86/drm-i915-dump-UTS_RELEASE-into-the-error_state.patch Here is a quick summary of what kernel's I've tried so far: works 3.8.5 works 3.6.9 fails 3.4.4 fails 3.2.41-2 fails 3.2.39-2 works 3.2.35-2 Looking at snapshots.debian.org, I'll try 3.5.5-1, as it seems so far Jonathan has been trying to get me to bisect in on the bug. Cheers, Geoff -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
On 08/04/13 16:28, Jonathan Nieder wrote: Thanks for testing. As a next test, can you try the 3.4.4 binary package from http://snapshot.debian.org/package/linux/? Hope that helps, Jonathan I've tried that. The 3.4.4-1~eperimental.1 linux-image I built from those sources does not boot, it fails in the same manner as the other 3.2 kernels after 3.2.35. What's next? Cheers, Geoff -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
On 03/04/13 09:30, Jonathan Nieder wrote: Geoff Crompton wrote: I had been considering building a kernel for each patch where the kernel only includes that one patch. But with 17 patches, that seems like a lot of kernels to test (for me anyway). So I've further whittled that down by looking for patches that mention 'backport': geoffc@hulk:~/debs/linux-3.2.39$ cat ../i915_bug_patch_shortlist | xargs grep -li backport debian/patches/bugfix/all/i915-ensure-that-VGA-plane-is-disabled.patch debian/patches/bugfix/x86/drm-i915-add-quirk_invert_brightness-for-ncr-machine.patch debian/patches/bugfix/x86/drm-i915-Close-race-between-processing-unpin-task-an.patch debian/patches/bugfix/x86/drm-i915-Disable-AsyncFlip-performance-optimisations.patch debian/patches/bugfix/x86/drm-i915-dump-UTS_RELEASE-into-the-error_state.patch debian/patches/bugfix/x86/drm-i915-GFX_MODE-Flush-TLB-Invalidate-Mode-must-be-.patch debian/patches/bugfix/x86/drm-i915-kick-any-firmware-framebuffers-before-claim.patch debian/patches/bugfix/x86/drm-i915-Only-increment-the-user-pin-count-after-suc.patch For these 8 patches I'll attempt to build packages that only include a single patch from this short list, and see if one of those kernels crashes. I don't have access to the machine that I've seen the bug on until next week. Unfortunately I don't think the patches are independent. If you can find a way to bisect among them to see which introduces the problem, that would be great, but please also do test the 3.8.y kernel from experimental so we can get upstream help. Thanks again and sorry for the trouble, Jonathan I can confirm the patches are not independant. I found bugfix/x86/drm-i915-Only-kick-out-vesafb-if-we-takeover-the-fbc.patch didn't apply if the above patches weren't included (because I'd commented them in the series file). But I didn't get any further along that line of development. Instead I downloaded linux_3.8.5-1~experimental.1.dsc (and related) and tried to build that on my wheezy box. The build failed after a long time, but well after it had successfully built linux-image-3.8-trunk-amd64_3.8.5-1~gc70.1_amd64.deb. So I tried to install that (and found I also had to download and compile the initramfs-tools 0.110 package). But now I'm running the 3.8.5 kernel successfully. So this bug is not present in a 3.8.5 kernel. I guess given Jonathans comments on 2013-04-02 I should keep trying to build a 3.2.39 package without some selection of the above patches (and whatever else) to see if that does work, and then try adding back in patches until it fails. Cheers, Geoff -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
fixed 703468 linux/3.8.5-1~experimental.1 quit Geoff Crompton wrote: I downloaded linux_3.8.5-1~experimental.1.dsc (and related) and tried to build that on my wheezy box. The build failed after a long time, but well after it had successfully built linux-image-3.8-trunk-amd64_3.8.5-1~gc70.1_amd64.deb. [...] But now I'm running the 3.8.5 kernel successfully. So this bug is not present in a 3.8.5 kernel. Thanks for testing. As a next test, can you try the 3.4.4 binary package from http://snapshot.debian.org/package/linux/? Hope that helps, Jonathan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
Geoff Crompton wrote: I had been considering building a kernel for each patch where the kernel only includes that one patch. But with 17 patches, that seems like a lot of kernels to test (for me anyway). So I've further whittled that down by looking for patches that mention 'backport': geoffc@hulk:~/debs/linux-3.2.39$ cat ../i915_bug_patch_shortlist | xargs grep -li backport debian/patches/bugfix/all/i915-ensure-that-VGA-plane-is-disabled.patch debian/patches/bugfix/x86/drm-i915-add-quirk_invert_brightness-for-ncr-machine.patch debian/patches/bugfix/x86/drm-i915-Close-race-between-processing-unpin-task-an.patch debian/patches/bugfix/x86/drm-i915-Disable-AsyncFlip-performance-optimisations.patch debian/patches/bugfix/x86/drm-i915-dump-UTS_RELEASE-into-the-error_state.patch debian/patches/bugfix/x86/drm-i915-GFX_MODE-Flush-TLB-Invalidate-Mode-must-be-.patch debian/patches/bugfix/x86/drm-i915-kick-any-firmware-framebuffers-before-claim.patch debian/patches/bugfix/x86/drm-i915-Only-increment-the-user-pin-count-after-suc.patch For these 8 patches I'll attempt to build packages that only include a single patch from this short list, and see if one of those kernels crashes. I don't have access to the machine that I've seen the bug on until next week. Unfortunately I don't think the patches are independent. If you can find a way to bisect among them to see which introduces the problem, that would be great, but please also do test the 3.8.y kernel from experimental so we can get upstream help. Thanks again and sorry for the trouble, Jonathan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
found 703468 linux/3.2.41-2 quit Jonathan Nieder wrote: I think a good place to start would be to try 3.2.41-2 from unstable, since it includes some i915 fixes. Ah, looking back over the log I see that's been tried. So: A good next step would be to try 3.8.y from experimental. If it works, we can try to find what is missing in the 3.4.y driver backport. If it doesn't work, we can get help from upstream. Jonathan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#703468: [3.2.35-2 - 3.2.39-1 regression] fails to boot on apple iMac
Hi, Michael Gilbert wrote: The following string is still recognizable: i915_gem_init_ppgtt+0x93/0x16c [i915] This is going to take some work on your end to get fixed. 3.2.39-1 included backported patches from aspects of the i915 driver. I think a good place to start would be to try 3.2.41-2 from unstable, since it includes some i915 fixes. A good next step would be to try 3.8.y from experimental. If it works, we can try to find what is missing in the 3.4.y driver backport. If it doesn't work, we can get help from upstream. [...] You can find what has changed between 3.2.35 and 3.2.39 by using debdiff from the devscripts package Unfortunately that range includes a pretty major change: * Backport drm and agp subsystems from Linux 3.4.29 (closes: #687442) Hope that helps, Jonathan -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org