[Intel-gfx] [PATCH 2/3] drm/i915: Fix fbdev unload sequence
First thing we need to do is unregister the fbdev instance, but we can't just go ahead and kfree it. That must wait until the hotplug and polling work are stopped, since they can race with the with the teardown. That means we need to split up the fbdev teardown into the unregister part and the cleanup part. I originally suspected that this was broken in one of the unload shuffles, but on closer inspection the oldest sequence I've dug out also gets this wrong. Just not quite so badly. I've run drv_module_reload a few hundred times and it's rock solid compared to insta-death beforehand. This bug seems to have been uncovered by commit 88be58be886f1215cc73dc8c273c985eecd7385c Author: Daniel VetterDate: Thu Jul 6 15:00:19 2017 +0200 drm/i915/fbdev: Always forward hotplug events But the effect of that seems to only be to increase the race window enough to make it blow up easier. I'm not exactly clear on what's going on there ... v2: Fix whitespace and use fetch_and_zero (Chris). Testcase: igt/drv_module_reload Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101791 Cc: martin.pe...@free.fr Cc: ch...@chris-wilson.co.uk Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_drv.c | 3 +-- drivers/gpu/drm/i915/intel_display.c | 3 +++ drivers/gpu/drm/i915/intel_drv.h | 9 +++-- drivers/gpu/drm/i915/intel_fbdev.c | 22 +++--- 4 files changed, 26 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index d310d8245dca..2c2afb19975d 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1240,6 +1240,7 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) */ static void i915_driver_unregister(struct drm_i915_private *dev_priv) { + intel_fbdev_unregister(dev_priv); intel_audio_deinit(dev_priv); intel_gpu_ips_teardown(); @@ -1371,8 +1372,6 @@ void i915_driver_unload(struct drm_device *dev) struct drm_i915_private *dev_priv = to_i915(dev); struct pci_dev *pdev = dev_priv->drm.pdev; - intel_fbdev_fini(dev); - if (i915_gem_suspend(dev_priv)) DRM_ERROR("failed to idle hardware; continuing to unload!\n"); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ecb4e30673fc..e10fae2faa88 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15111,6 +15111,9 @@ void intel_modeset_cleanup(struct drm_device *dev) */ drm_kms_helper_poll_fini(dev); + /* poll work can call into fbdev, hence clean that up afterwards */ + intel_fbdev_fini(dev_priv); + intel_unregister_dsm_handler(); intel_fbc_global_disable(dev_priv); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 3132260f18ce..7361e983d953 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1597,7 +1597,8 @@ void intel_hpd_poll_init(struct drm_i915_private *dev_priv); #ifdef CONFIG_DRM_FBDEV_EMULATION extern int intel_fbdev_init(struct drm_device *dev); extern void intel_fbdev_initial_config_async(struct drm_device *dev); -extern void intel_fbdev_fini(struct drm_device *dev); +extern void intel_fbdev_unregister(struct drm_i915_private *dev_priv); +extern void intel_fbdev_fini(struct drm_i915_private *dev_priv); extern void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous); extern void intel_fbdev_output_poll_changed(struct drm_device *dev); extern void intel_fbdev_restore_mode(struct drm_device *dev); @@ -1611,7 +1612,11 @@ static inline void intel_fbdev_initial_config_async(struct drm_device *dev) { } -static inline void intel_fbdev_fini(struct drm_device *dev) +static inline void intel_fbdev_unregister(struct drm_i915_private *dev_priv) +{ +} + +static inline void intel_fbdev_fini(struct drm_i915_private *dev_priv) { } diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c index b953365a3eec..391992373d27 100644 --- a/drivers/gpu/drm/i915/intel_fbdev.c +++ b/drivers/gpu/drm/i915/intel_fbdev.c @@ -538,8 +538,6 @@ static void intel_fbdev_destroy(struct intel_fbdev *ifbdev) * trying to rectify all the possible error paths leading here. */ - drm_fb_helper_unregister_fbi(>helper); - drm_fb_helper_fini(>helper); if (ifbdev->vma) { @@ -727,8 +725,10 @@ static void intel_fbdev_initial_config(void *data, async_cookie_t cookie) /* Due to peculiar init order wrt to hpd handling this is separate. */ if (drm_fb_helper_initial_config(>helper, -ifbdev->preferred_bpp)) - intel_fbdev_fini(ifbdev->helper.dev); +
Re: [Intel-gfx] [PATCH 2/3] drm/i915: Fix fbdev unload sequence
Quoting Daniel Vetter (2017-07-14 20:14:38) Second nit: > +void intel_fbdev_fini(struct drm_i915_private *dev_priv) > +{ > + struct intel_fbdev *ifbdev = dev_priv->fbdev; > + > + if (!ifbdev) > + return; > + > intel_fbdev_destroy(ifbdev); > dev_priv->fbdev = NULL; Maybe struct intel_fbdev *ifbdev; ifbdev = fetch_and_zero(_priv->fbdev); if (!fbdev) return; intel_fbdev_destroy(ifbdev); -Chris ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] [PATCH 2/3] drm/i915: Fix fbdev unload sequence
Quoting Daniel Vetter (2017-07-14 20:14:38) > First thing we need to do is unregister the fbdev instance, but we > can't just go ahead and kfree it. That must wait until the hotplug and > polling work are stopped, since they can race with the with the > teardown. That means we need to split up the fbdev teardown into the > unregister part and the cleanup part. > > I originally suspected that this was broken in one of the unload > shuffles, but on closer inspection the oldest sequence I've dug out > also gets this wrong. Just not quite so badly. > > I've run drv_module_reload a few hundred times and it's rock solid > compared to insta-death beforehand. This bug seems to have been > uncovered by > > commit 88be58be886f1215cc73dc8c273c985eecd7385c > Author: Daniel Vetter> Date: Thu Jul 6 15:00:19 2017 +0200 > > drm/i915/fbdev: Always forward hotplug events > > But the effect of that seems to only be to increase the race window > enough to make it blow up easier. I'm not exactly clear on what's > going on there ... > > Testcase: igt/drv_module_reload > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101791 > Cc: martin.pe...@free.fr > Cc: ch...@chris-wilson.co.uk > Signed-off-by: Daniel Vetter Splitting out the unregister and running it early makes sense and is how we expect to unload to proceed. Reviewed-by: Chris Wilson Nit: > @@ -1383,7 +1383,6 @@ void i915_driver_unload(struct drm_device *dev) > intel_gvt_cleanup(dev_priv); > > i915_driver_unregister(dev_priv); > - > intel_modeset_cleanup(dev); I think it will useful to maintain the visual break here. (I know next patch moves it again...) -Chris ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx
[Intel-gfx] [PATCH 2/3] drm/i915: Fix fbdev unload sequence
First thing we need to do is unregister the fbdev instance, but we can't just go ahead and kfree it. That must wait until the hotplug and polling work are stopped, since they can race with the with the teardown. That means we need to split up the fbdev teardown into the unregister part and the cleanup part. I originally suspected that this was broken in one of the unload shuffles, but on closer inspection the oldest sequence I've dug out also gets this wrong. Just not quite so badly. I've run drv_module_reload a few hundred times and it's rock solid compared to insta-death beforehand. This bug seems to have been uncovered by commit 88be58be886f1215cc73dc8c273c985eecd7385c Author: Daniel VetterDate: Thu Jul 6 15:00:19 2017 +0200 drm/i915/fbdev: Always forward hotplug events But the effect of that seems to only be to increase the race window enough to make it blow up easier. I'm not exactly clear on what's going on there ... Testcase: igt/drv_module_reload Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=101791 Cc: martin.pe...@free.fr Cc: ch...@chris-wilson.co.uk Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_drv.c | 3 +-- drivers/gpu/drm/i915/intel_display.c | 3 +++ drivers/gpu/drm/i915/intel_drv.h | 9 +++-- drivers/gpu/drm/i915/intel_fbdev.c | 21 +++-- 4 files changed, 26 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index d310d8245dca..3a6dc04bd51e 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -1240,6 +1240,7 @@ static void i915_driver_register(struct drm_i915_private *dev_priv) */ static void i915_driver_unregister(struct drm_i915_private *dev_priv) { + intel_fbdev_unregister(dev_priv); intel_audio_deinit(dev_priv); intel_gpu_ips_teardown(); @@ -1371,7 +1372,6 @@ void i915_driver_unload(struct drm_device *dev) struct drm_i915_private *dev_priv = to_i915(dev); struct pci_dev *pdev = dev_priv->drm.pdev; - intel_fbdev_fini(dev); if (i915_gem_suspend(dev_priv)) DRM_ERROR("failed to idle hardware; continuing to unload!\n"); @@ -1383,7 +1383,6 @@ void i915_driver_unload(struct drm_device *dev) intel_gvt_cleanup(dev_priv); i915_driver_unregister(dev_priv); - intel_modeset_cleanup(dev); /* diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index ecb4e30673fc..e10fae2faa88 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -15111,6 +15111,9 @@ void intel_modeset_cleanup(struct drm_device *dev) */ drm_kms_helper_poll_fini(dev); + /* poll work can call into fbdev, hence clean that up afterwards */ + intel_fbdev_fini(dev_priv); + intel_unregister_dsm_handler(); intel_fbc_global_disable(dev_priv); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 3132260f18ce..7361e983d953 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -1597,7 +1597,8 @@ void intel_hpd_poll_init(struct drm_i915_private *dev_priv); #ifdef CONFIG_DRM_FBDEV_EMULATION extern int intel_fbdev_init(struct drm_device *dev); extern void intel_fbdev_initial_config_async(struct drm_device *dev); -extern void intel_fbdev_fini(struct drm_device *dev); +extern void intel_fbdev_unregister(struct drm_i915_private *dev_priv); +extern void intel_fbdev_fini(struct drm_i915_private *dev_priv); extern void intel_fbdev_set_suspend(struct drm_device *dev, int state, bool synchronous); extern void intel_fbdev_output_poll_changed(struct drm_device *dev); extern void intel_fbdev_restore_mode(struct drm_device *dev); @@ -1611,7 +1612,11 @@ static inline void intel_fbdev_initial_config_async(struct drm_device *dev) { } -static inline void intel_fbdev_fini(struct drm_device *dev) +static inline void intel_fbdev_unregister(struct drm_i915_private *dev_priv) +{ +} + +static inline void intel_fbdev_fini(struct drm_i915_private *dev_priv) { } diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c index b953365a3eec..82280e104692 100644 --- a/drivers/gpu/drm/i915/intel_fbdev.c +++ b/drivers/gpu/drm/i915/intel_fbdev.c @@ -538,8 +538,6 @@ static void intel_fbdev_destroy(struct intel_fbdev *ifbdev) * trying to rectify all the possible error paths leading here. */ - drm_fb_helper_unregister_fbi(>helper); - drm_fb_helper_fini(>helper); if (ifbdev->vma) { @@ -727,8 +725,10 @@ static void intel_fbdev_initial_config(void *data, async_cookie_t cookie) /* Due to peculiar init order wrt to hpd handling this is separate. */ if (drm_fb_helper_initial_config(>helper, -ifbdev->preferred_bpp)) -