drm/i915: Only wait upon the execution timeline when unlocked

In order to walk the list of all timelines, we currently require the
struct_mutex. We are sometimes called prior to the struct_mutex being
taken by the caller (i.e !I915_WAIT_LOCKED) in which case we can only
trust the global execution timelines (as these are owned by the device).
This means in the unlocked phase we can only wait upon the currently
executing requests and not all queued.

[  175.743243] general protection fault: 0000 [#1] SMP
[  175.743263] Modules linked in: nls_iso8859_1 intel_rapl x86_pkg_temp_thermal coretemp kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul ghash_clmulni_intel iwlwifi aesni_intel aes_x86_64 lrw snd_soc_rt5640 gf128mul snd_soc_rl6231 snd_soc_core glue_helper snd_compress snd_pcm_dmaengine snd_hda_codec_hdmi ablk_helper snd_hda_codec_realtek cryptd snd_hda_codec_generic serio_raw cfg80211 snd_hda_intel snd_hda_codec ir_lirc_codec snd_hda_core lirc_dev snd_hwdep snd_pcm lpc_ich mei_me mei snd_seq_midi shpchp snd_seq_midi_event snd_rawmidi snd_seq snd_seq_device snd_timer rc_rc6_mce acpi_als nuvoton_cir kfifo_buf rc_core snd industrialio snd_soc_sst_acpi soundcore snd_soc_sst_match i2c_designware_platform 8250_dw i2c_designware_core dw_dmac spi_pxa2xx_platform mac_hid acpi_pad parport_pc ppdev lp parport
[  175.743509]  autofs4 i915 e1000e psmouse ptp pps_core xhci_pci ehci_pci ahci xhci_hcd ehci_hcd libahci video sdhci_acpi sdhci i2c_hid hid
[  175.743560] CPU: 2 PID: 2386 Comm: wtdg_monitor.sh Tainted: G     U          4.9.0-rc4-nightly+ #2
[  175.743581] Hardware name:                  /NUC5i7RYB, BIOS RYBDWi35.86A.0358.2016.0606.1423 06/06/2016
[  175.743603] task: ffff88024509ba80 task.stack: ffffc9007bd18000
[  175.743618] RIP: 0010:[<ffffffffa01af29b>]  [<ffffffffa01af29b>] i915_gem_wait_for_idle+0x3b/0x140 [i915]
[  175.743660] RSP: 0000:ffffc9007bd1b9b8  EFLAGS: 00010297
[  175.743674] RAX: ffff88024489d248 RBX: 0000000000000000 RCX: 0000000000000000
[  175.743691] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff880244898000
[  175.743708] RBP: ffffc9007bd1b9f0 R08: 0000000000000000 R09: 0000000000000001
[  175.743724] R10: 00000028eaf42792 R11: 0000000000000001 R12: dead000000000100
[  175.743741] R13: dead000000000148 R14: ffffc9007bd1ba5f R15: 0000000000000005
[  175.743758] FS:  00007f2638330700(0000) GS:ffff880256d00000(0000) knlGS:0000000000000000
[  175.743777] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  175.743791] CR2: 00007f885c8cea40 CR3: 00000002416b5000 CR4: 00000000003406e0
[  175.743808] Stack:
[  175.743816]  ffff88024489d248 000000004509ba80 ffff880244898000 ffff88024509ba80
[  175.743840]  00000000ffff8b69 ffffc9007bd1ba5f ffffc9007bd1ba5e ffffc9007bd1ba28
[  175.743863]  ffffffffa01b661d 00000000ffffffff 0000000000000000 ffff880244898000
[  175.743886] Call Trace:
[  175.743906]  [<ffffffffa01b661d>] i915_gem_shrinker_lock_uninterruptible.constprop.5+0x5d/0xc0 [i915]
[  175.743937]  [<ffffffffa01b6cd0>] i915_gem_shrinker_oom+0x30/0x1b0 [i915]
[  175.743955]  [<ffffffff8109ca79>] notifier_call_chain+0x49/0x70
[  175.743971]  [<ffffffff8109cd9d>] __blocking_notifier_call_chain+0x4d/0x70
[  175.743988]  [<ffffffff8109cdd6>] blocking_notifier_call_chain+0x16/0x20
[  175.744005]  [<ffffffff811885dc>] out_of_memory+0x22c/0x480
[  175.744020]  [<ffffffff81205542>] __alloc_pages_slowpath+0x851/0x8ec
[  175.744037]  [<ffffffff8118ca51>] __alloc_pages_nodemask+0x2c1/0x310
[  175.744054]  [<ffffffff811d8ea8>] alloc_pages_current+0x88/0x120
[  175.744070]  [<ffffffff811833a4>] __page_cache_alloc+0xb4/0xc0
[  175.744086]  [<ffffffff811865ca>] filemap_fault+0x29a/0x500
[  175.744101]  [<ffffffff81299aa6>] ext4_filemap_fault+0x36/0x50
[  175.744117]  [<ffffffff811b3d4a>] __do_fault+0x6a/0xe0
[  175.744131]  [<ffffffff811b97ee>] handle_mm_fault+0xd0e/0x1330
[  175.744147]  [<ffffffff8106738c>] __do_page_fault+0x23c/0x4d0
[  175.744162]  [<ffffffff81067650>] do_page_fault+0x30/0x80
[  175.744177]  [<ffffffff817ffbe8>] page_fault+0x28/0x30
[  175.744191] Code: 41 57 41 56 41 55 41 54 53 48 83 ec 10 4c 8b a7 48 52 00 00 89 75 d4 48 89 45 c8 49 39 c4 74 78 4d 8d 6c 24 48 41 bf 05 00 00 00 <49> 8b 5d 00 48 85 db 74 50 8b 83 20 01 00 00 85 c0 74 15 48 8b
[  175.744320] RIP  [<ffffffffa01af29b>] i915_gem_wait_for_idle+0x3b/0x140 [i915]
[  175.744351]  RSP <ffffc9007bd1b9b8>

Fixes: 80b204bce8f2 ("drm/i915: Enable multiple timelines")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20161111145809.9701-1-chris@chris-wilson.co.uk

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 97d164d0c04736254f018624c6c92908ef5e2bea..4ea3e537eceb9bdec72f6cdd3c6952d11567a1e3 100644 (file)
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2934,11 +2934,20 @@ static int wait_for_timeline(struct i915_gem_timeline *tl, unsigned int flags)
 
 int i915_gem_wait_for_idle(struct drm_i915_private *i915, unsigned int flags)
 {
-       struct i915_gem_timeline *tl;
        int ret;
 
-       list_for_each_entry(tl, &i915->gt.timelines, link) {
-               ret = wait_for_timeline(tl, flags);
+       if (flags & I915_WAIT_LOCKED) {
+               struct i915_gem_timeline *tl;
+
+               lockdep_assert_held(&i915->drm.struct_mutex);
+
+               list_for_each_entry(tl, &i915->gt.timelines, link) {
+                       ret = wait_for_timeline(tl, flags);
+                       if (ret)
+                               return ret;
+               }
+       } else {
+               ret = wait_for_timeline(&i915->gt.global_timeline, flags);
                if (ret)
                        return ret;
        }