Commit c998e8a0 authored by Chris Wilson's avatar Chris Wilson

drm/i915: Hold rpm during GEM suspend in driver unload/suspend

i915_gem_suspend() tries to access the device to ensure it is idle and
all writes from the device are flushed to memory. It assumed is already
held the runtime pm wakeref, but we should explicitly acquire it for our
access to be safe.

[  619.926287] WARNING: CPU: 3 PID: 9353 at drivers/gpu/drm/i915/intel_drv.h:1750 gen6_write32+0x23e/0x2a0 [i915]
[  619.926300] RPM wakelock ref not held during HW access
[  619.926311] Modules linked in: vgem x86_pkg_temp_thermal intel_powerclamp snd_hda_codec_hdmi snd_hda_codec_generic snd_hda_codec coretemp snd_hwdep crct10dif_pclmul snd_hda_core crc32_pclmul snd_pcm mei_me mei lpc_ich ghash_clmulni_intel i915(-) sdhci_pci sdhci mmc_core e1000e ptp pps_core prime_numbers [last unloaded: snd_hda_intel]
[  619.926578] CPU: 3 PID: 9353 Comm: drv_module_relo Tainted: G     U          4.10.0-CI-Trybot_609+ #1
[  619.926585] Hardware name: LENOVO 42962WU/42962WU, BIOS 8DET56WW (1.26 ) 12/01/2011
[  619.926592] Call Trace:
[  619.926609]  dump_stack+0x67/0x92
[  619.926625]  __warn+0xc6/0xe0
[  619.926640]  warn_slowpath_fmt+0x4a/0x50
[  619.926726]  gen6_write32+0x23e/0x2a0 [i915]
[  619.926801]  gen6_mm_switch+0x38/0x70 [i915]
[  619.926871]  i915_switch_context+0xec/0xa10 [i915]
[  619.926942]  i915_gem_switch_to_kernel_context+0x13c/0x2b0 [i915]
[  619.927019]  i915_gem_suspend+0x2b/0x180 [i915]
[  619.927079]  i915_driver_unload+0x22/0x200 [i915]
[  619.927093]  ? __this_cpu_preempt_check+0x13/0x20
[  619.927105]  ? trace_hardirqs_on_caller+0xe7/0x200
[  619.927118]  ? trace_hardirqs_on+0xd/0x10
[  619.927128]  ? _raw_spin_unlock_irqrestore+0x3d/0x60
[  619.927192]  i915_pci_remove+0x14/0x20 [i915]
[  619.927205]  pci_device_remove+0x34/0xb0
[  619.927219]  device_release_driver_internal+0x158/0x210
[  619.927234]  driver_detach+0x3b/0x80
[  619.927245]  bus_remove_driver+0x53/0xd0
[  619.927256]  driver_unregister+0x27/0x50
[  619.927267]  pci_unregister_driver+0x25/0xa0
[  619.927351]  i915_exit+0x1a/0xb1a [i915]
[  619.927362]  SyS_delete_module+0x193/0x1e0
[  619.927378]  entry_SYSCALL_64_fastpath+0x1c/0xb1
[  619.927386] RIP: 0033:0x7f82b46c5d37
[  619.927393] RSP: 002b:00007ffdb6f610d8 EFLAGS: 00000246 ORIG_RAX: 00000000000000b0
[  619.927408] RAX: ffffffffffffffda RBX: ffffffff81481ff3 RCX: 00007f82b46c5d37
[  619.927415] RDX: 0000000000000001 RSI: 0000000000000800 RDI: 000000000224f558
[  619.927422] RBP: ffffc90001187f88 R08: 0000000000000000 R09: 00007ffdb6f61100
[  619.927428] R10: 000000000224f4e0 R11: 0000000000000246 R12: 0000000000000000
[  619.927435] R13: 00007ffdb6f612b0 R14: 0000000000000000 R15: 0000000000000000
[  619.927451]  ? __this_cpu_preempt_check+0x13/0x20

or

[  641.646590] WARNING: CPU: 1 PID: 8913 at drivers/gpu/drm/i915/intel_drv.h:1750 intel_runtime_pm_get_noresume+0x8b/0x90 [i915]
[  641.646595] RPM wakelock ref not held during HW access
[  641.646600] Modules linked in: vgem snd_hda_codec_hdmi snd_hda_codec_generic x86_pkg_temp_thermal intel_powerclamp coretemp snd_hda_codec snd_hwdep crct10dif_pclmul snd_hda_core crc32_pclmul ghash_clmulni_intel snd_pcm mei_me mei i915(-) r8169 mii prime_numbers i2c_hid [last unloaded: snd_hda_intel]
[  641.646825] CPU: 1 PID: 8913 Comm: drv_module_relo Tainted: G     U          4.10.0-CI-Trybot_609+ #1
[  641.646836] Hardware name: TOSHIBA SATELLITE P50-C/06F4                            , BIOS 1.20 10/08/2015
[  641.646843] Call Trace:
[  641.646857]  dump_stack+0x67/0x92
[  641.646869]  __warn+0xc6/0xe0
[  641.646880]  warn_slowpath_fmt+0x4a/0x50
[  641.646893]  ? __this_cpu_preempt_check+0x13/0x20
[  641.646904]  ? trace_hardirqs_on_caller+0xe7/0x200
[  641.646957]  intel_runtime_pm_get_noresume+0x8b/0x90 [i915]
[  641.647022]  __i915_add_request+0x423/0x540 [i915]
[  641.647080]  i915_gem_switch_to_kernel_context+0x148/0x2b0 [i915]
[  641.647145]  i915_gem_suspend+0x2b/0x180 [i915]
[  641.647189]  i915_driver_unload+0x22/0x200 [i915]
[  641.647200]  ? __this_cpu_preempt_check+0x13/0x20
[  641.647210]  ? trace_hardirqs_on_caller+0xe7/0x200
[  641.647220]  ? trace_hardirqs_on+0xd/0x10
[  641.647231]  ? _raw_spin_unlock_irqrestore+0x3d/0x60
[  641.647276]  i915_pci_remove+0x14/0x20 [i915]
[  641.647293]  pci_device_remove+0x34/0xb0
[  641.647307]  device_release_driver_internal+0x158/0x210
[  641.647321]  driver_detach+0x3b/0x80
[  641.647330]  bus_remove_driver+0x53/0xd0
[  641.647338]  driver_unregister+0x27/0x50
[  641.647348]  pci_unregister_driver+0x25/0xa0
[  641.647415]  i915_exit+0x1a/0xb1a [i915]
[  641.647429]  SyS_delete_module+0x193/0x1e0
[  641.647444]  entry_SYSCALL_64_fastpath+0x1c/0xb1
[  641.647453] RIP: 0033:0x7fc622bd2d37
[  641.647463] RSP: 002b:00007ffff8ffb5c8 EFLAGS: 00000246 ORIG_RAX: 00000000000000b0
[  641.647475] RAX: ffffffffffffffda RBX: ffffffff81481ff3 RCX: 00007fc622bd2d37
[  641.647480] RDX: 0000000000000001 RSI: 0000000000000800 RDI: 0000000000d49118
[  641.647485] RBP: ffffc90000997f88 R08: 0000000000000000 R09: 00007ffff8ffb5f0
[  641.647491] R10: 0000000000d490a0 R11: 0000000000000246 R12: 0000000000000000
[  641.647498] R13: 00007ffff8ffb7a0 R14: 0000000000000000 R15: 0000000000000000
[  641.647510]  ? __this_cpu_preempt_check+0x13/0x20

v2: Keep holding rpm until the end to cover i915_gem_sanitize() as well.

Fixes: 5ab57c70 ("drm/i915: Flush logical context image out to memory upon suspend")
Fixes: 1c777c5d ("drm/i915/hsw: Fix GPU hang during resume from S3-devices state")
Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
Link: http://patchwork.freedesktop.org/patch/msgid/20170302083029.19576-1-chris@chris-wilson.co.ukReviewed-by: default avatarImre Deak <imre.deak@intel.com>
Cc: <stable@vger.kernel.org> # v4.9+
parent a746095c
...@@ -4244,6 +4244,7 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) ...@@ -4244,6 +4244,7 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv)
struct drm_device *dev = &dev_priv->drm; struct drm_device *dev = &dev_priv->drm;
int ret; int ret;
intel_runtime_pm_get(dev_priv);
intel_suspend_gt_powersave(dev_priv); intel_suspend_gt_powersave(dev_priv);
mutex_lock(&dev->struct_mutex); mutex_lock(&dev->struct_mutex);
...@@ -4258,13 +4259,13 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) ...@@ -4258,13 +4259,13 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv)
*/ */
ret = i915_gem_switch_to_kernel_context(dev_priv); ret = i915_gem_switch_to_kernel_context(dev_priv);
if (ret) if (ret)
goto err; goto err_unlock;
ret = i915_gem_wait_for_idle(dev_priv, ret = i915_gem_wait_for_idle(dev_priv,
I915_WAIT_INTERRUPTIBLE | I915_WAIT_INTERRUPTIBLE |
I915_WAIT_LOCKED); I915_WAIT_LOCKED);
if (ret) if (ret)
goto err; goto err_unlock;
i915_gem_retire_requests(dev_priv); i915_gem_retire_requests(dev_priv);
GEM_BUG_ON(dev_priv->gt.active_requests); GEM_BUG_ON(dev_priv->gt.active_requests);
...@@ -4310,11 +4311,12 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv) ...@@ -4310,11 +4311,12 @@ int i915_gem_suspend(struct drm_i915_private *dev_priv)
* machine in an unusable condition. * machine in an unusable condition.
*/ */
i915_gem_sanitize(dev_priv); i915_gem_sanitize(dev_priv);
goto out_rpm_put;
return 0; err_unlock:
err:
mutex_unlock(&dev->struct_mutex); mutex_unlock(&dev->struct_mutex);
out_rpm_put:
intel_runtime_pm_put(dev_priv);
return ret; return ret;
} }
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment