BUG: pin_count exceeded in i915_gem_object_pin

BUG: pin_count exceeded in i915_gem_object_pin

[Arnd Bergmann]

While testing the regression I reported, I got the oops below, from

int i915_gem_object_pin(struct drm_gem_object *obj, uint32_t alignment)
{
        struct drm_device *dev = obj->dev;
        struct drm_i915_private *dev_priv = dev->dev_private;
        struct drm_i915_gem_object *obj_priv = to_intel_bo(obj);
        int ret;
        BUG_ON(obj_priv->pin_count == DRM_I915_GEM_OBJECT_MAX_PIN_COUNT);

This happened when I killed Xorg as root after plugging in a different screen.
kdm usually restarts immediately, so I don't know if this was the dying Xorg
or the new one.

The kernel version was 2.6.36-rc7-next-20101015 with my fix for the other problem
applied.

	Arnd

Oct 17 19:48:15 wuerfel kernel: [ 1172.767083] ------------[ cut here ]------------
Oct 17 19:48:15 wuerfel kernel: [ 1172.767113] kernel BUG at /home/arnd/linux-2.6/drivers/gpu/drm/i915/i915_gem.c:4159!
Oct 17 19:48:15 wuerfel kernel: [ 1172.767142] invalid opcode: 0000 [#1] SMP 
Oct 17 19:48:15 wuerfel kernel: [ 1172.767162] last sysfs file: /sys/devices/pci0000:00/0000:00:02.1/resource
Oct 17 19:48:15 wuerfel kernel: [ 1172.767187] CPU 2 
Oct 17 19:48:15 wuerfel kernel: [ 1172.767197] Modules linked in: rfcomm sco bnep l2cap crc16 kvm_intel kvm snd_hda_codec_hdmi snd_hda_cod
ec_realtek snd_hda_intel snd_hda_codec snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm 8250_pnp snd_seq_dummy snd_seq_oss snd_seq_midi snd_raw
midi snd_seq_midi_event snd_seq snd_timer snd_seq_device i915 usb_storage snd usbhid drm_kms_helper soundcore btusb drm usb_libusual hid 8
250 serial_core snd_page_alloc coretemp i2c_algo_bit sg firewire_ohci firewire_core r8169 uhci_hcd ehci_hcd crc_itu_t usbcore
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] 
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] Pid: 7035, comm: Xorg Not tainted 2.6.36-rc7-next-20101015+ #98 P5Q-EM/P5Q-EM
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] RIP: 0010:[<ffffffffa012f5f4>]  [<ffffffffa012f5f4>] i915_gem_object_pin+0x35/0x17f [i915]
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] RSP: 0018:ffff880188873c28  EFLAGS: 00010246
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] RAX: 000000000003c000 RBX: ffff8801ae4dee00 RCX: ffff8801ac923598
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] RDX: ffff8801863eed80 RSI: 0000000000000000 RDI: ffff8801ae4dee00
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] RBP: ffff880188873c48 R08: ffff8801789338c8 R09: 0000000000000000
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] R10: ffff880188873c48 R11: ffff880177baeb80 R12: ffff8801ac922000
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] R13: 0000000000000000 R14: ffff8801ae4dee00 R15: ffff8801ac922000
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] FS:  00007f10d7b85840(0000) GS:ffff8800cdb00000(0000) knlGS:0000000000000000
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] CR2: 000000000314c688 CR3: 0000000188a4d000 CR4: 00000000000406e0
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] Process Xorg (pid: 7035, threadinfo ffff880188872000, task ffff8801789331b0)
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  ffff880188873c48 0000000000000001 0000000000000000 ffff8801afd5db78
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] <0> ffff880188873d58 ffffffffa0132108 ffff8801afd5db40 0000000000000000
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006] <0> ffffffff810d9fa5 0000000000000000 ffff8801afd5dbb0 0000000000000002
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffffa0132108>] i915_gem_do_execbuffer+0x55b/0xeb3 [i915]
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff810d9fa5>] ? might_fault+0x5c/0xac
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff810d9fee>] ? might_fault+0xa5/0xac
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff810d9fa5>] ? might_fault+0x5c/0xac
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffffa0131b95>] ? drm_malloc_ab+0x38/0x50 [i915]
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffffa0132b2e>] i915_gem_execbuffer2+0xce/0x12c [i915]
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff810d9fee>] ? might_fault+0xa5/0xac
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffffa00891bf>] drm_ioctl+0x27c/0x349 [drm]
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff8120172b>] ? do_raw_spin_lock+0x6b/0x120
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffffa0132a60>] ? i915_gem_execbuffer2+0x0/0x12c [i915]
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff813fe6ec>] ? _raw_spin_unlock+0x2b/0x2f
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff810de36f>] ? handle_mm_fault+0x6fb/0x718
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff81401d86>] ? do_page_fault+0x16f/0x2c3
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff81106d61>] do_vfs_ioctl+0x4bf/0x50e
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff8105fbf4>] ? up_read+0x23/0x3b
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff81002caa>] ? sysret_check+0x2e/0x69
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff81106e06>] sys_ioctl+0x56/0x79
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  [<ffffffff81002c72>] system_call_fastpath+0x16/0x1b
Oct 17 19:48:15 wuerfel kernel: [ 1172.768006]  RSP <ffff880188873c28>
Oct 17 19:48:15 wuerfel kernel: [ 1172.830801] ---[ end trace 4186e27271a72eb5 ]---

Re: BUG: pin_count exceeded in i915_gem_object_pin

[Chris Wilson]

On Sun, 17 Oct 2010 20:06:37 +0200, Arnd Bergmann <arnd@arndb.de> wrote:
> This happened when I killed Xorg as root after plugging in a different screen.
> kdm usually restarts immediately, so I don't know if this was the dying Xorg
> or the new one.

The new BUG_ON() has spotted that we are leaking pinned fb. I haven't
gotten to the bottom of this entirely yet, in theory this should be fixed
with the unpin fb added when disabling the connector in -next. I am not
totally convinced that explains everything in this case though:

https://bugs.freedesktop.org/show_bug.cgi?id=29857
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

Re: BUG: pin_count exceeded in i915_gem_object_pin

[Arnd Bergmann]

On Monday 18 October 2010, Chris Wilson wrote:
> On Sun, 17 Oct 2010 20:06:37 +0200, Arnd Bergmann <arnd@arndb.de> wrote:
> > This happened when I killed Xorg as root after plugging in a different screen.
> > kdm usually restarts immediately, so I don't know if this was the dying Xorg
> > or the new one.
> 
> The new BUG_ON() has spotted that we are leaking pinned fb. I haven't
> gotten to the bottom of this entirely yet, in theory this should be fixed
> with the unpin fb added when disabling the connector in -next. I am not
> totally convinced that explains everything in this case though:
> 
> https://bugs.freedesktop.org/show_bug.cgi?id=29857

Well, this was using -next from friday, which should already include that
fix, right?

It sounds to me that there is another unpin missing in the close path
for the file descriptor.

	Arnd