Commit 2c83a726 authored by Lucas Stach's avatar Lucas Stach
Browse files

drm/etnaviv: bring back progress check in job timeout handler

When the hangcheck handler was replaced by the DRM scheduler timeout
handling we dropped the forward progress check, as this might allow
clients to hog the GPU for a long time with a big job.

It turns out that even reasonably well behaved clients like the
Armada Xorg driver occasionally trip over the 500ms timeout. Bring
back the forward progress check to get rid of the userspace regression.

We would still like to fix userspace to submit smaller batches
if possible, but that is for another day.

Cc: <>
Fixes: 6d7a20c0

 (drm/etnaviv: replace hangcheck with scheduler timeout)
Reported-by: default avatarRussell King <>
Signed-off-by: default avatarLucas Stach <>
Reviewed-by: default avatarEric Anholt <>
parent bf6ba3ae
...@@ -131,6 +131,9 @@ struct etnaviv_gpu { ...@@ -131,6 +131,9 @@ struct etnaviv_gpu {
struct work_struct sync_point_work; struct work_struct sync_point_work;
int sync_point_event; int sync_point_event;
/* hang detection */
u32 hangcheck_dma_addr;
void __iomem *mmio; void __iomem *mmio;
int irq; int irq;
...@@ -10,6 +10,7 @@ ...@@ -10,6 +10,7 @@
#include "etnaviv_gem.h" #include "etnaviv_gem.h"
#include "etnaviv_gpu.h" #include "etnaviv_gpu.h"
#include "etnaviv_sched.h" #include "etnaviv_sched.h"
#include "state.xml.h"
static int etnaviv_job_hang_limit = 0; static int etnaviv_job_hang_limit = 0;
module_param_named(job_hang_limit, etnaviv_job_hang_limit, int , 0444); module_param_named(job_hang_limit, etnaviv_job_hang_limit, int , 0444);
...@@ -85,6 +86,29 @@ static void etnaviv_sched_timedout_job(struct drm_sched_job *sched_job) ...@@ -85,6 +86,29 @@ static void etnaviv_sched_timedout_job(struct drm_sched_job *sched_job)
{ {
struct etnaviv_gem_submit *submit = to_etnaviv_submit(sched_job); struct etnaviv_gem_submit *submit = to_etnaviv_submit(sched_job);
struct etnaviv_gpu *gpu = submit->gpu; struct etnaviv_gpu *gpu = submit->gpu;
u32 dma_addr;
int change;
* If the GPU managed to complete this jobs fence, the timout is
* spurious. Bail out.
if (fence_completed(gpu, submit->out_fence->seqno))
* If the GPU is still making forward progress on the front-end (which
* should never loop) we shift out the timeout to give it a chance to
* finish the job.
dma_addr = gpu_read(gpu, VIVS_FE_DMA_ADDRESS);
change = dma_addr - gpu->hangcheck_dma_addr;
if (change < 0 || change > 16) {
gpu->hangcheck_dma_addr = dma_addr;
/* block scheduler */ /* block scheduler */
kthread_park(gpu->sched.thread); kthread_park(gpu->sched.thread);
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