Eric Anholt eric@anholt.net hat am 8. Oktober 2017 um 19:09 geschrieben:
Stefan Wahren stefan.wahren@i2se.com writes:
Hi Eric,
Eric Anholt eric@anholt.net hat am 6. Oktober 2017 um 21:42 geschrieben:
Stefan Wahren stefan.wahren@i2se.com writes:
This fixes the race between vc4_overflow_mem_work and the init of the job lock. Otherwise we could trigger a NULL pointer dereference during VC4 binding.
Link: https://github.com/anholt/linux/issues/114 Signed-off-by: Stefan Wahren stefan.wahren@i2se.com Fixes: d5b1a78a772f ("drm/vc4: Add support for drawing 3D frames.")
drivers/gpu/drm/vc4/vc4_gem.c | 1 - drivers/gpu/drm/vc4/vc4_irq.c | 1 + 2 files changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/vc4/vc4_gem.c b/drivers/gpu/drm/vc4/vc4_gem.c index d0c6bfb..47d964f 100644 --- a/drivers/gpu/drm/vc4/vc4_gem.c +++ b/drivers/gpu/drm/vc4/vc4_gem.c @@ -1088,7 +1088,6 @@ vc4_gem_init(struct drm_device *dev) INIT_LIST_HEAD(&vc4->render_job_list); INIT_LIST_HEAD(&vc4->job_done_list); INIT_LIST_HEAD(&vc4->seqno_cb_list);
spin_lock_init(&vc4->job_lock);
INIT_WORK(&vc4->hangcheck.reset_work, vc4_reset_work); setup_timer(&vc4->hangcheck.timer,
diff --git a/drivers/gpu/drm/vc4/vc4_irq.c b/drivers/gpu/drm/vc4/vc4_irq.c index 7d7af3a..d508d13 100644 --- a/drivers/gpu/drm/vc4/vc4_irq.c +++ b/drivers/gpu/drm/vc4/vc4_irq.c @@ -195,6 +195,7 @@ vc4_irq_preinstall(struct drm_device *dev) struct vc4_dev *vc4 = to_vc4_dev(dev);
init_waitqueue_head(&vc4->job_wait_queue);
spin_lock_init(&vc4->job_lock); INIT_WORK(&vc4->overflow_mem_work, vc4_overflow_mem_work);
/* Clear any pending interrupts someone might have left around
Are you sure this is a fix? We don't attach the IRQ handler until V3D bind, and vc4_gem_init happens before component_bind_all(), right?
i agree i should have mark it as a RFC. But is it really impossible that vc4_overflow_mem_work is triggered before vc4_gem_init?
As far as I can see, it gets queued from the IRQ handler, the IRQ handler is added during V3D bind, and binding happens after GEM init.
Hmm. If we fail out of component binding and try again, we'll end up doing the job_wait_queue and overflow_mem_work initialization on the same pointer twice. Will that cause any trouble? We cancel any pending work during uninstall (V3D unbind path), but does drm_irq_uninstall() make sure that the irq handler has finished?
I cannot see an issue.