| Message ID | 1435655187-8769-1-git-send-email-tvrtko.ursulin@linux.intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Tested-By: Intel Graphics QA PRTS (Patch Regression Test System Contact: shuang.he@intel.com)
Task id: 6642
-------------------------------------Summary-------------------------------------
Platform Delta drm-intel-nightly Series Applied
ILK 302/302 302/302
SNB 312/316 312/316
IVB 343/343 343/343
BYT -1 287/287 286/287
HSW 380/380 380/380
-------------------------------------Detailed-------------------------------------
Platform Test drm-intel-nightly Series Applied
*BYT igt@gem_tiled_partial_pwrite_pread@reads PASS(1) FAIL(1)
Note: You need to pay more attention to line start with '*'
On Tue, 30 Jun 2015, Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> wrote: > From: Tvrtko Ursulin <tvrtko.ursulin@intel.com> > > We had three failure modes here: > > 1. > Deadlock in intelfb_alloc failure path where it calls > drm_framebuffer_remove, which grabs the struct mutex and intelfb_create > (caller of intelfb_alloc) was already holding it. > > 2. > Double unreference on the object if __intel_framebuffer_create fails > since both it and the caller (intelfb_alloc) do the unreference. > > 3. > Deadlock in intelfb_create failure path where it calls > drm_framebuffer_unreference, which grabs the struct mutex and > intelfb_create was already holding it. > > v2: > * Reformat commit msg to 72 chars. (Lukas Wunner) > * Added third failure mode. (Lukas Wunner) > > Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> > Fixes: 60a5ca015ffd2aacfe5674b5a401cd2a37159e07 The format is Fixes: 60a5ca015ffd ("drm/i915: Add locking around framebuffer_references--") See Documentation/SubmittingPatches on how to have git output that format for you. BR, Jani. > Reported-By: Lukas Wunner <lukas@wunner.de> > Tested-By: Lukas Wunner <lukas@wunner.de> > Cc: Chris Wilson <chris@chris-wilson.co.uk> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> > Cc: Lukas Wunner <lukas@wunner.de> > --- > Ville, you suggested some changes earlier; > > """ > I suggest removing the unref from __intel_framebuffer_create(). > """ > > Do you view that as an improvement in code clarity/organisation, > or you think my version is actually wrong for some reason? > --- > drivers/gpu/drm/i915/intel_fbdev.c | 21 +++++++++++++-------- > 1 file changed, 13 insertions(+), 8 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c > index 2a1724e..118420b 100644 > --- a/drivers/gpu/drm/i915/intel_fbdev.c > +++ b/drivers/gpu/drm/i915/intel_fbdev.c > @@ -141,7 +141,7 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > { > struct intel_fbdev *ifbdev = > container_of(helper, struct intel_fbdev, helper); > - struct drm_framebuffer *fb; > + struct drm_framebuffer *fb = NULL; > struct drm_device *dev = helper->dev; > struct drm_mode_fb_cmd2 mode_cmd = {}; > struct drm_i915_gem_object *obj; > @@ -159,6 +159,8 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > mode_cmd.pixel_format = drm_mode_legacy_fb_format(sizes->surface_bpp, > sizes->surface_depth); > > + mutex_lock(&dev->struct_mutex); > + > size = mode_cmd.pitches[0] * mode_cmd.height; > size = PAGE_ALIGN(size); > obj = i915_gem_object_create_stolen(dev, size); > @@ -172,8 +174,9 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > > fb = __intel_framebuffer_create(dev, &mode_cmd, obj); > if (IS_ERR(fb)) { > + /* Drops object reference on failure. */ > ret = PTR_ERR(fb); > - goto out_unref; > + goto out; > } > > /* Flush everything out, we'll be doing GTT only from now on */ > @@ -183,15 +186,18 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > goto out_fb; > } > > + mutex_unlock(&dev->struct_mutex); > + > ifbdev->fb = to_intel_framebuffer(fb); > > return 0; > > out_fb: > - drm_framebuffer_remove(fb); > -out_unref: > drm_gem_object_unreference(&obj->base); > out: > + mutex_unlock(&dev->struct_mutex); > + if (fb) > + drm_framebuffer_remove(fb); > return ret; > } > > @@ -209,8 +215,6 @@ static int intelfb_create(struct drm_fb_helper *helper, > int size, ret; > bool prealloc = false; > > - mutex_lock(&dev->struct_mutex); > - > if (intel_fb && > (sizes->fb_width > intel_fb->base.width || > sizes->fb_height > intel_fb->base.height)) { > @@ -225,7 +229,7 @@ static int intelfb_create(struct drm_fb_helper *helper, > DRM_DEBUG_KMS("no BIOS fb, allocating a new one\n"); > ret = intelfb_alloc(helper, sizes); > if (ret) > - goto out_unlock; > + return ret; > intel_fb = ifbdev->fb; > } else { > DRM_DEBUG_KMS("re-using BIOS fb\n"); > @@ -237,6 +241,8 @@ static int intelfb_create(struct drm_fb_helper *helper, > obj = intel_fb->obj; > size = obj->base.size; > > + mutex_lock(&dev->struct_mutex); > + > info = framebuffer_alloc(0, &dev->pdev->dev); > if (!info) { > ret = -ENOMEM; > @@ -307,7 +313,6 @@ static int intelfb_create(struct drm_fb_helper *helper, > out_unpin: > i915_gem_object_ggtt_unpin(obj); > drm_gem_object_unreference(&obj->base); > -out_unlock: > mutex_unlock(&dev->struct_mutex); > return ret; > } > -- > 2.4.2 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Hi, not sure if another reaction from me is required, but in case there is: The patch as quoted below LGTM now, so unless Ville vetoes because of his suggestion (see below) I think this is ready for inclusion in 4.2. Thanks, Lukas On Tue, Jun 30, 2015 at 10:06:27AM +0100, Tvrtko Ursulin wrote: > From: Tvrtko Ursulin <tvrtko.ursulin@intel.com> > > We had three failure modes here: > > 1. > Deadlock in intelfb_alloc failure path where it calls > drm_framebuffer_remove, which grabs the struct mutex and intelfb_create > (caller of intelfb_alloc) was already holding it. > > 2. > Double unreference on the object if __intel_framebuffer_create fails > since both it and the caller (intelfb_alloc) do the unreference. > > 3. > Deadlock in intelfb_create failure path where it calls > drm_framebuffer_unreference, which grabs the struct mutex and > intelfb_create was already holding it. > > v2: > * Reformat commit msg to 72 chars. (Lukas Wunner) > * Added third failure mode. (Lukas Wunner) > > Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> > Fixes: 60a5ca015ffd2aacfe5674b5a401cd2a37159e07 > Reported-By: Lukas Wunner <lukas@wunner.de> > Tested-By: Lukas Wunner <lukas@wunner.de> > Cc: Chris Wilson <chris@chris-wilson.co.uk> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> > Cc: Lukas Wunner <lukas@wunner.de> > --- > Ville, you suggested some changes earlier; > > """ > I suggest removing the unref from __intel_framebuffer_create(). > """ > > Do you view that as an improvement in code clarity/organisation, > or you think my version is actually wrong for some reason? > --- > drivers/gpu/drm/i915/intel_fbdev.c | 21 +++++++++++++-------- > 1 file changed, 13 insertions(+), 8 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c > index 2a1724e..118420b 100644 > --- a/drivers/gpu/drm/i915/intel_fbdev.c > +++ b/drivers/gpu/drm/i915/intel_fbdev.c > @@ -141,7 +141,7 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > { > struct intel_fbdev *ifbdev = > container_of(helper, struct intel_fbdev, helper); > - struct drm_framebuffer *fb; > + struct drm_framebuffer *fb = NULL; > struct drm_device *dev = helper->dev; > struct drm_mode_fb_cmd2 mode_cmd = {}; > struct drm_i915_gem_object *obj; > @@ -159,6 +159,8 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > mode_cmd.pixel_format = drm_mode_legacy_fb_format(sizes->surface_bpp, > sizes->surface_depth); > > + mutex_lock(&dev->struct_mutex); > + > size = mode_cmd.pitches[0] * mode_cmd.height; > size = PAGE_ALIGN(size); > obj = i915_gem_object_create_stolen(dev, size); > @@ -172,8 +174,9 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > > fb = __intel_framebuffer_create(dev, &mode_cmd, obj); > if (IS_ERR(fb)) { > + /* Drops object reference on failure. */ > ret = PTR_ERR(fb); > - goto out_unref; > + goto out; > } > > /* Flush everything out, we'll be doing GTT only from now on */ > @@ -183,15 +186,18 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > goto out_fb; > } > > + mutex_unlock(&dev->struct_mutex); > + > ifbdev->fb = to_intel_framebuffer(fb); > > return 0; > > out_fb: > - drm_framebuffer_remove(fb); > -out_unref: > drm_gem_object_unreference(&obj->base); > out: > + mutex_unlock(&dev->struct_mutex); > + if (fb) > + drm_framebuffer_remove(fb); > return ret; > } > > @@ -209,8 +215,6 @@ static int intelfb_create(struct drm_fb_helper *helper, > int size, ret; > bool prealloc = false; > > - mutex_lock(&dev->struct_mutex); > - > if (intel_fb && > (sizes->fb_width > intel_fb->base.width || > sizes->fb_height > intel_fb->base.height)) { > @@ -225,7 +229,7 @@ static int intelfb_create(struct drm_fb_helper *helper, > DRM_DEBUG_KMS("no BIOS fb, allocating a new one\n"); > ret = intelfb_alloc(helper, sizes); > if (ret) > - goto out_unlock; > + return ret; > intel_fb = ifbdev->fb; > } else { > DRM_DEBUG_KMS("re-using BIOS fb\n"); > @@ -237,6 +241,8 @@ static int intelfb_create(struct drm_fb_helper *helper, > obj = intel_fb->obj; > size = obj->base.size; > > + mutex_lock(&dev->struct_mutex); > + > info = framebuffer_alloc(0, &dev->pdev->dev); > if (!info) { > ret = -ENOMEM; > @@ -307,7 +313,6 @@ static int intelfb_create(struct drm_fb_helper *helper, > out_unpin: > i915_gem_object_ggtt_unpin(obj); > drm_gem_object_unreference(&obj->base); > -out_unlock: > mutex_unlock(&dev->struct_mutex); > return ret; > } > -- > 2.4.2 >
On Tue, Jun 30, 2015 at 10:06:27AM +0100, Tvrtko Ursulin wrote: > From: Tvrtko Ursulin <tvrtko.ursulin@intel.com> > > We had three failure modes here: > > 1. > Deadlock in intelfb_alloc failure path where it calls > drm_framebuffer_remove, which grabs the struct mutex and intelfb_create > (caller of intelfb_alloc) was already holding it. > > 2. > Double unreference on the object if __intel_framebuffer_create fails > since both it and the caller (intelfb_alloc) do the unreference. > > 3. > Deadlock in intelfb_create failure path where it calls > drm_framebuffer_unreference, which grabs the struct mutex and > intelfb_create was already holding it. > > v2: > * Reformat commit msg to 72 chars. (Lukas Wunner) > * Added third failure mode. (Lukas Wunner) > > Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> > Fixes: 60a5ca015ffd2aacfe5674b5a401cd2a37159e07 > Reported-By: Lukas Wunner <lukas@wunner.de> > Tested-By: Lukas Wunner <lukas@wunner.de> > Cc: Chris Wilson <chris@chris-wilson.co.uk> > Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> > Cc: Lukas Wunner <lukas@wunner.de> > --- > Ville, you suggested some changes earlier; > > """ > I suggest removing the unref from __intel_framebuffer_create(). > """ > > Do you view that as an improvement in code clarity/organisation, > or you think my version is actually wrong for some reason? I find it rather unexpected that the function drops the passed reference on error. My usual rule is: do nothing on error, if possible. > --- > drivers/gpu/drm/i915/intel_fbdev.c | 21 +++++++++++++-------- > 1 file changed, 13 insertions(+), 8 deletions(-) > > diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c > index 2a1724e..118420b 100644 > --- a/drivers/gpu/drm/i915/intel_fbdev.c > +++ b/drivers/gpu/drm/i915/intel_fbdev.c > @@ -141,7 +141,7 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > { > struct intel_fbdev *ifbdev = > container_of(helper, struct intel_fbdev, helper); > - struct drm_framebuffer *fb; > + struct drm_framebuffer *fb = NULL; > struct drm_device *dev = helper->dev; > struct drm_mode_fb_cmd2 mode_cmd = {}; > struct drm_i915_gem_object *obj; > @@ -159,6 +159,8 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > mode_cmd.pixel_format = drm_mode_legacy_fb_format(sizes->surface_bpp, > sizes->surface_depth); > > + mutex_lock(&dev->struct_mutex); > + > size = mode_cmd.pitches[0] * mode_cmd.height; > size = PAGE_ALIGN(size); > obj = i915_gem_object_create_stolen(dev, size); > @@ -172,8 +174,9 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > > fb = __intel_framebuffer_create(dev, &mode_cmd, obj); > if (IS_ERR(fb)) { > + /* Drops object reference on failure. */ > ret = PTR_ERR(fb); > - goto out_unref; > + goto out; > } > > /* Flush everything out, we'll be doing GTT only from now on */ > @@ -183,15 +186,18 @@ static int intelfb_alloc(struct drm_fb_helper *helper, > goto out_fb; > } > > + mutex_unlock(&dev->struct_mutex); > + > ifbdev->fb = to_intel_framebuffer(fb); > > return 0; > > out_fb: > - drm_framebuffer_remove(fb); > -out_unref: > drm_gem_object_unreference(&obj->base); > out: > + mutex_unlock(&dev->struct_mutex); > + if (fb) > + drm_framebuffer_remove(fb); > return ret; > } > > @@ -209,8 +215,6 @@ static int intelfb_create(struct drm_fb_helper *helper, > int size, ret; > bool prealloc = false; > > - mutex_lock(&dev->struct_mutex); > - > if (intel_fb && > (sizes->fb_width > intel_fb->base.width || > sizes->fb_height > intel_fb->base.height)) { > @@ -225,7 +229,7 @@ static int intelfb_create(struct drm_fb_helper *helper, > DRM_DEBUG_KMS("no BIOS fb, allocating a new one\n"); > ret = intelfb_alloc(helper, sizes); > if (ret) > - goto out_unlock; > + return ret; > intel_fb = ifbdev->fb; > } else { > DRM_DEBUG_KMS("re-using BIOS fb\n"); > @@ -237,6 +241,8 @@ static int intelfb_create(struct drm_fb_helper *helper, > obj = intel_fb->obj; > size = obj->base.size; > > + mutex_lock(&dev->struct_mutex); > + > info = framebuffer_alloc(0, &dev->pdev->dev); > if (!info) { > ret = -ENOMEM; > @@ -307,7 +313,6 @@ static int intelfb_create(struct drm_fb_helper *helper, > out_unpin: > i915_gem_object_ggtt_unpin(obj); > drm_gem_object_unreference(&obj->base); > -out_unlock: > mutex_unlock(&dev->struct_mutex); > return ret; > } > -- > 2.4.2
On 07/02/2015 02:37 PM, Ville Syrjälä wrote: > On Tue, Jun 30, 2015 at 10:06:27AM +0100, Tvrtko Ursulin wrote: >> From: Tvrtko Ursulin <tvrtko.ursulin@intel.com> >> >> We had three failure modes here: >> >> 1. >> Deadlock in intelfb_alloc failure path where it calls >> drm_framebuffer_remove, which grabs the struct mutex and intelfb_create >> (caller of intelfb_alloc) was already holding it. >> >> 2. >> Double unreference on the object if __intel_framebuffer_create fails >> since both it and the caller (intelfb_alloc) do the unreference. >> >> 3. >> Deadlock in intelfb_create failure path where it calls >> drm_framebuffer_unreference, which grabs the struct mutex and >> intelfb_create was already holding it. >> >> v2: >> * Reformat commit msg to 72 chars. (Lukas Wunner) >> * Added third failure mode. (Lukas Wunner) >> >> Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> >> Fixes: 60a5ca015ffd2aacfe5674b5a401cd2a37159e07 >> Reported-By: Lukas Wunner <lukas@wunner.de> >> Tested-By: Lukas Wunner <lukas@wunner.de> >> Cc: Chris Wilson <chris@chris-wilson.co.uk> >> Cc: Ville Syrjälä <ville.syrjala@linux.intel.com> >> Cc: Lukas Wunner <lukas@wunner.de> >> --- >> Ville, you suggested some changes earlier; >> >> """ >> I suggest removing the unref from __intel_framebuffer_create(). >> """ >> >> Do you view that as an improvement in code clarity/organisation, >> or you think my version is actually wrong for some reason? > > I find it rather unexpected that the function drops the passed > reference on error. My usual rule is: do nothing on error, if possible. I just don't have time at the moment to evaluate the other call sites etc. So from my point of view it boils down to whether this patch improves things without making anything worse. If it does R-B would be cool. If it doesn't then it will have to make for free time to appear. Or someone is free to take it over. Regards, Tvrtko
Hi, On Thu, Jul 02, 2015 at 02:59:24PM +0100, Tvrtko Ursulin wrote: > On 07/02/2015 02:37 PM, Ville Syrjälä wrote: > >On Tue, Jun 30, 2015 at 10:06:27AM +0100, Tvrtko Ursulin wrote: > >>Ville, you suggested some changes earlier; > >>""" > >>I suggest removing the unref from __intel_framebuffer_create(). > >>""" > >>Do you view that as an improvement in code clarity/organisation, > >>or you think my version is actually wrong for some reason? > >I find it rather unexpected that the function drops the passed > >reference on error. My usual rule is: do nothing on error, if possible. > I just don't have time at the moment to evaluate the other call sites etc. > So from my point of view it boils down to whether this patch improves things > without making anything worse. If it does R-B would be cool. If it doesn't > then it will have to make for free time to appear. Or someone is free to > take it over. I think Ville's suggestion is merited, right now the unref on failure happens at the bottom of the call chain in __intel_framebuffer_create while the ref happened further up in the call chain. This should really be consolidated in the same place. That a double unref managed to sneak into intelfb_alloc proves this is error prone. I've just submitted a v3 which is functionally equivalent to Tvrtko's v2 but takes up Ville's suggestion. This consists of two patches now, the first one being equivalent to Tvrtko's patch sans fix for the double unref, the second one fixing the double unref by moving the unref further up the call chain. Of course I've tested this, works for me. Now you can decide for yourself which version you like better. :-) Thanks, Lukas
diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c index 2a1724e..118420b 100644 --- a/drivers/gpu/drm/i915/intel_fbdev.c +++ b/drivers/gpu/drm/i915/intel_fbdev.c @@ -141,7 +141,7 @@ static int intelfb_alloc(struct drm_fb_helper *helper, { struct intel_fbdev *ifbdev = container_of(helper, struct intel_fbdev, helper); - struct drm_framebuffer *fb; + struct drm_framebuffer *fb = NULL; struct drm_device *dev = helper->dev; struct drm_mode_fb_cmd2 mode_cmd = {}; struct drm_i915_gem_object *obj; @@ -159,6 +159,8 @@ static int intelfb_alloc(struct drm_fb_helper *helper, mode_cmd.pixel_format = drm_mode_legacy_fb_format(sizes->surface_bpp, sizes->surface_depth); + mutex_lock(&dev->struct_mutex); + size = mode_cmd.pitches[0] * mode_cmd.height; size = PAGE_ALIGN(size); obj = i915_gem_object_create_stolen(dev, size); @@ -172,8 +174,9 @@ static int intelfb_alloc(struct drm_fb_helper *helper, fb = __intel_framebuffer_create(dev, &mode_cmd, obj); if (IS_ERR(fb)) { + /* Drops object reference on failure. */ ret = PTR_ERR(fb); - goto out_unref; + goto out; } /* Flush everything out, we'll be doing GTT only from now on */ @@ -183,15 +186,18 @@ static int intelfb_alloc(struct drm_fb_helper *helper, goto out_fb; } + mutex_unlock(&dev->struct_mutex); + ifbdev->fb = to_intel_framebuffer(fb); return 0; out_fb: - drm_framebuffer_remove(fb); -out_unref: drm_gem_object_unreference(&obj->base); out: + mutex_unlock(&dev->struct_mutex); + if (fb) + drm_framebuffer_remove(fb); return ret; } @@ -209,8 +215,6 @@ static int intelfb_create(struct drm_fb_helper *helper, int size, ret; bool prealloc = false; - mutex_lock(&dev->struct_mutex); - if (intel_fb && (sizes->fb_width > intel_fb->base.width || sizes->fb_height > intel_fb->base.height)) { @@ -225,7 +229,7 @@ static int intelfb_create(struct drm_fb_helper *helper, DRM_DEBUG_KMS("no BIOS fb, allocating a new one\n"); ret = intelfb_alloc(helper, sizes); if (ret) - goto out_unlock; + return ret; intel_fb = ifbdev->fb; } else { DRM_DEBUG_KMS("re-using BIOS fb\n"); @@ -237,6 +241,8 @@ static int intelfb_create(struct drm_fb_helper *helper, obj = intel_fb->obj; size = obj->base.size; + mutex_lock(&dev->struct_mutex); + info = framebuffer_alloc(0, &dev->pdev->dev); if (!info) { ret = -ENOMEM; @@ -307,7 +313,6 @@ static int intelfb_create(struct drm_fb_helper *helper, out_unpin: i915_gem_object_ggtt_unpin(obj); drm_gem_object_unreference(&obj->base); -out_unlock: mutex_unlock(&dev->struct_mutex); return ret; }