[Intel-gfx] [PATCH] drm/i915: Explicitly reset the seqno upon idling

[Chris Wilson]

During execbuffer emission we assert that we do not wrap around the
seqno used for semaphore breadcrumbs. However:

:kernel BUG at drivers/gpu/drm/i915/i915_gem_execbuffer.c:1239!
:invalid opcode: 0000 [#1] SMP
:CPU 0
:Modules linked in: usb_storage usblp bnep lockd sunrpc bluetooth rfkill vboxpci(O) vboxnetadp(O) vboxnetflt(O) vboxdrv(O) snd_hda_codec_hdmi snd_hda_codec_realtek coretemp mei lpc_ich snd_hda_intel mfd_core i2c_i801 kvm snd_hda_codec snd_hwdep e1000e microcode snd_pcm snd_page_alloc snd_timer snd soundcore serio_raw uinput binfmt_misc crc32c_intel ghash_clmulni_intel wmi i915 video i2c_algo_bit
drm_kms_helper drm i2c_core [last unloaded: scsi_wait_scan]
:Pid: 962, comm: X Tainted: G         C O 3.5.4-1.fc17.x86_64 #1 LENOVO 5032AJ3/
:RIP: 0010:[<ffffffffa008842c>]  [<ffffffffa008842c>] i915_gem_do_execbuffer.isra.10+0xcbc/0x1390 [i915]
:RSP: 0018:ffff8801214a3c08  EFLAGS: 00010286
:RAX: 0000000000000000 RBX: 0000000000000000 RCX: dead000000200200
:RDX: ffff880133561988 RSI: 00000000fffffe0c RDI: ffff88000248f8b0
:RBP: ffff8801214a3d28 R08: ffff88000248f8b0 R09: 0000000180400032
:R10: 000000003288eb01 R11: ffff8801214a3fd8 R12: ffff8801335618b0
:R13: ffff880133ebd800 R14: 0000000000000001 R15: ffff8801214a3bf8
:FS:  00007f5f0647a8c0(0000) GS:ffff88013e200000(0000) knlGS:0000000000000000
:CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
:CR2: 0000000003481000 CR3: 0000000134170000 CR4: 00000000000407f0
:DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
:DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
:Process X (pid: 962, threadinfo ffff8801214a2000, task ffff880123afae20)
:Stack:
: 0000000000000000 0000000000000000 0000000000000000 0000000000000000
: ffff880123851cc0 0000000000000001 ffff8801214a3c88 ffffffffa0082eb4
: ffff880100000002 000000c033560000 ffff8801214a3c88 ffff880123851ce0
:Call Trace:
: [<ffffffffa0082eb4>] ? i915_gem_object_set_to_gtt_domain+0x104/0x1a0 [i915]
: [<ffffffffa0089031>] i915_gem_execbuffer2+0xb1/0x290 [i915]
: [<ffffffffa00154f3>] drm_ioctl+0x4d3/0x580 [drm]
: [<ffffffffa0088f80>] ? i915_gem_execbuffer+0x480/0x480 [i915]
: [<ffffffff81188d5e>] ? do_readv_writev+0x18e/0x1e0
: [<ffffffff81199919>] do_vfs_ioctl+0x99/0x580
: [<ffffffff8127973a>] ? inode_has_perm.isra.31.constprop.61+0x2a/0x30
: [<ffffffff8127ad17>] ? file_has_perm+0x97/0xb0
: [<ffffffff81199e99>] sys_ioctl+0x99/0xa0
: [<ffffffff81614e29>] system_call_fastpath+0x16/0x1b
:Code: 59 fd ff ff 4c 89 ef e8 23 a4 ff ff 85 c0 90 0f 85 ad fc ff ff 4c 89 ef e8 82 9d ff ff 45 8b 4c 24 6c 45 85 c9 0f 84 02 fd ff ff <0f> 0b 4c 89 ef e8 fa a3 ff ff 85 c0 0f 85 85 fc ff ff 4c 89 ef
:RIP  [<ffffffffa008842c>] i915_gem_do_execbuffer.isra.10+0xcbc/0x1390 [i915]
: RSP <ffff8801214a3c08>

clearly shows us hitting this supposedly impossible wraparound. The
cause here is that after idling, retire-requests only resets the
breadcrumbs if there was a request on the ring. To avoid this after
idling, we can simply clear the breadcrumbs.

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=863861
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: stable at vger.kernel.org
---
 drivers/gpu/drm/i915/i915_gem.c |   11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 288d7b8..95c0cd0 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2163,6 +2163,15 @@ static int i915_ring_idle(struct intel_ring_buffer *ring)
 	return i915_wait_request(ring, i915_gem_next_request_seqno(ring));
 }
 
+static void i915_ring_reset_seqno(struct intel_ring_buffer *ring)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(ring->sync_seqno); i++)
+		if (seqno >= ring->sync_seqno[i])
+			ring->sync_seqno[i] = 0;
+}
+
 int i915_gpu_idle(struct drm_device *dev)
 {
 	drm_i915_private_t *dev_priv = dev->dev_private;
@@ -2178,6 +2187,8 @@ int i915_gpu_idle(struct drm_device *dev)
 		/* Is the device fubar? */
 		if (WARN_ON(!list_empty(&ring->gpu_write_list)))
 			return -EBUSY;
+
+		i915_ring_reset_seqno(ring);
 	}
 
 	return 0;
-- 
1.7.10.4