[Intel-gfx] [PATCH] drm/i915: Update write_domains on active list after flush.

[Adam Lantos]

Hi Chris,

On Sun, Jan 31, 2010 at 4:55 PM, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> Before changing the status of a buffer with a pending write we will await
> upon a new flush for that buffer. So we can take advantage of any flushes
> posted whilst the buffer is active and pending processing by the GPU, by
> clearing its write_domain and updating its last_rendering_seqno -- thus
> saving a potential flush in deep queues and improves flushing behaviour
> upon eviction for both GTT space and fences.
>
> In order to reduce the time spent searching the active list for matching
> write_domains, we move those to a separate list whose elements are
> the buffers belong to the active/flushing list with pending writes.
>
> Also as Daniel Vetter pointed out, currently we need to emit a flush and
> wait on that flush *twice* in order to properly evict everything. The
> first flush will just move dirty buffers onto the flushing list,
> requiring the second, redundant, flush to finally move those onto the
> inactive list.  The lack of the additional flush in
> i915_gem_evict_everything() and i915_gpu_idle() will trigger BUGs, such
> as:
>
> [1400539.466422] [drm:i915_gem_object_pin] *ERROR* Failure to install fence: -28
> [1400539.480213] [drm:i915_gem_object_pin] *ERROR* Failure to install fence: -28
> [1400539.500195] [drm:i915_gem_object_pin] *ERROR* Failure to install fence: -28
> [1400539.541957] ------------[ cut here ]------------
> [1400539.541971] kernel BUG at
> /tmp/buildd/linux-2.6-2.6.32/debian/build/source_i386_none/drivers/gpu/drm/i915/i915_gem.c:2123!
> [1400539.541977] invalid opcode: 0000 [#1] SMP
> [1400539.541982] last sysfs file:
> /sys/devices/pci0000:00/0000:00:1e.0/0000:02:04.0/class
> [1400539.541986] Modules linked in: i915 drm_kms_helper drm i2c_algo_bit
> i2c_core ufs qnx4 hfsplus hfs minix ntfs vfat msdos fat jfs xfs exportfs
> reiserfs ext3 jbd ext2 nfs lockd fscache nfs_acl auth_rpcgss sunrpc loop
> joydev snd_intel8x0m snd_intel8x0 snd_ac97_codec snd_pcsp ac97_bus
> snd_pcm yenta_socket snd_timer rsrc_nonstatic snd shpchp soundcore
> psmouse dcdbas rng_core serio_raw snd_page_alloc pci_hotplug battery ac
> evdev processor ext4 mbcache jbd2 crc16 ide_cd_mod ide_gd_mod cdrom
> ide_pci_generic ata_generic libata scsi_mod uhci_hcd b44 ssb mmc_core
> pcmcia piix pcmcia_core intel_agp mii ehci_hcd ide_core usbcore nls_base
> agpgart button video output thermal fan thermal_sys [last unloaded:
> scsi_wait_scan]
> [1400539.542074]
> [1400539.542080] Pid: 23623, comm: X Not tainted (2.6.32-trunk-686 #1) Inspiron 1100
> [1400539.542085] EIP: 0060:[<d0c29202>] EFLAGS: 00213246 CPU: 0
> [1400539.542106] EIP is at i915_gem_evict_everything+0xf1/0xfe [i915]
> [1400539.542111] EAX: 00000000 EBX: cf74e000 ECX: cfbbd388 EDX: 0000dede
> [1400539.542115] ESI: 00000000 EDI: cf9f5400 EBP: cf74ee0c ESP: cf9ffdf8
> [1400539.542119]  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
> [1400539.542125] Process X (pid: 23623, ti=cf9fe000 task=ce250cc0 task.ti=cf9fe000)
> [1400539.542128] Stack:
> [1400539.542130]  cf74ee0c cf74ee20 00000022 00000022 ffffffe4 cf9f3360 d0c2a3e3 ccfcfe00
> [1400539.542140] <0> c5a2bcc0 cf9ffe94 cf9f5400 cf74e000 cf9f5414 c36e7000 ced63a00 00000000
> [1400539.542150] <0> d2fc5000 00000022 00000000 00000000 01c90073 bf930134 00000000 cf74e000
> [1400539.542160] Call Trace:
> [1400539.542174]  [<d0c2a3e3>] ? i915_gem_execbuffer+0x677/0xcc3 [i915]
> [1400539.542187]  [<d08a05c3>] ? drm_ioctl+0x1d2/0x268 [drm]
> [1400539.542199]  [<d0c29d6c>] ? i915_gem_execbuffer+0x0/0xcc3 [i915]
> [1400539.542212]  [<c10bc24d>] ? vfs_ioctl+0x49/0x5f
> [1400539.542218]  [<c10bc7b4>] ? do_vfs_ioctl+0x4aa/0x4e5
> [1400539.542225]  [<c1259c36>] ? schedule+0x78f/0x7dc
> [1400539.542233]  [<c1002cf7>] ? restore_sigcontext+0xbe/0xd5
> [1400539.542239]  [<c10bc830>] ? sys_ioctl+0x41/0x58
> [1400539.542245]  [<c10030fb>] ? sysenter_do_call+0x12/0x28
> [1400539.542248] Code: 0e 00 00 39 83 18 0e 00 00 75 14 8d 83 10 0e 00
> 00 39 83 10 0e 00 00 0f 94 c0 0f b6 c0 eb 02 31 c0 fe 83 0c 0e 00 00 85
> c0 75 04 <0f> 0b eb fe 5d 89 f0 5a 5b 5e 5f 5d c3 55 83 3d 44 c6 c4 d0
> 00
> [1400539.542297] EIP: [<d0c29202>] i915_gem_evict_everything+0xf1/0xfe [i915] SS:ESP 0068:cf9ffdf8
> [1400539.542314] ---[ end trace 2e3c5b62fd2600b8 ]---
>
> References:
>  Bug 25911 - 2.10.0 causes kernel oops and system hangs
>  http://bugs.freedesktop.org/show_bug.cgi?id=25911
>
>  Bug 26101 - [i915] xf86-video-intel 2.10.0 (and git) triggers kernel oops
>              within seconds after login
>  http://bugs.freedesktop.org/show_bug.cgi?id=26101
>
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
> Cc: stable at kernel.org

...

I was bitten by this problem on -rc5 some weeks ago, and now it
happens again on rc7: only using kde4 'present windows' after login,
and voilá, I get some artifacts and a kernel bug_on.

Should I try this patch and report back, or are you guys currently
working on an updated version? What's missing, should I provide more
debug info? It seems that I can reliably trigger this one.

Btw I'm using xorg-server-1.7.4.901-1, libdrm-2.4.17-3, mesa-7.7-1,
xf86-video-intel-2.9.1-1 all from archlinux repos.


thanks,
 Adam