[PATCH] gpu: host1x: Do not use mapping cache for job submissions

[PATCH] gpu: host1x: Do not use mapping cache for job submissions

[Thierry Reding]

From: Thierry Reding <treding@nvidia.com>

Buffer mappings used in job submissions are usually small and not
rapidly reused as opposed to framebuffers (which are usually large and
rapidly reused, for example when page-flipping between double-buffered
framebuffers). Avoid going through the mapping cache for these buffers
since the cache would also lead to leaks if nobody is ever releasing
the cache's last reference. For DRM/KMS these last references are
dropped when the framebuffers are removed and therefore no longer
needed.

While at it, also add a note about the need to explicitly remove the
final reference to the mapping in the cache.

Signed-off-by: Thierry Reding <treding@nvidia.com>
---
 drivers/gpu/host1x/job.c | 4 ++--
 include/linux/host1x.h   | 5 +++++
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/host1x/job.c b/drivers/gpu/host1x/job.c
index 9424b3c08ab2..f81f3b9f7510 100644
--- a/drivers/gpu/host1x/job.c
+++ b/drivers/gpu/host1x/job.c
@@ -176,7 +176,7 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 			goto unpin;
 		}
 
-		map = host1x_bo_pin(dev, bo, direction, &client->cache);
+		map = host1x_bo_pin(dev, bo, direction, NULL);
 		if (IS_ERR(map)) {
 			host1x_bo_put(bo);
 			err = PTR_ERR(map);
@@ -226,7 +226,7 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 			goto unpin;
 		}
 
-		map = host1x_bo_pin(host->dev, g->bo, DMA_TO_DEVICE, &host->cache);
+		map = host1x_bo_pin(host->dev, g->bo, DMA_TO_DEVICE, NULL);
 		if (IS_ERR(map)) {
 			host1x_bo_put(g->bo);
 			err = PTR_ERR(map);
diff --git a/include/linux/host1x.h b/include/linux/host1x.h
index e8dc5bc41f79..46081afd0ca0 100644
--- a/include/linux/host1x.h
+++ b/include/linux/host1x.h
@@ -31,6 +31,11 @@ u64 host1x_get_dma_mask(struct host1x *host1x);
  * struct host1x_bo_cache - host1x buffer object cache
  * @mappings: list of mappings
  * @lock: synchronizes accesses to the list of mappings
+ *
+ * Note that entries are not periodically evicted from this cache and instead need to be
+ * explicitly released. This is used primarily for DRM/KMS where the cache's reference is
+ * released when the last reference to a buffer object represented by a mapping in this
+ * cache is dropped.
  */
 struct host1x_bo_cache {
 	struct list_head mappings;
-- 
2.35.1

[PATCH] gpu: host1x: Do not use mapping cache for job submissions

[Thierry Reding]

From: Thierry Reding <treding@nvidia.com>

Buffer mappings used in job submissions are usually small and not
rapidly reused as opposed to framebuffers (which are usually large and
rapidly reused, for example when page-flipping between double-buffered
framebuffers). Avoid going through the mapping cache for these buffers
since the cache would also lead to leaks if nobody is ever releasing
the cache's last reference. For DRM/KMS these last references are
dropped when the framebuffers are removed and therefore no longer
needed.

While at it, also add a note about the need to explicitly remove the
final reference to the mapping in the cache.

Signed-off-by: Thierry Reding <treding@nvidia.com>
---
 drivers/gpu/host1x/job.c | 4 ++--
 include/linux/host1x.h   | 5 +++++
 2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/host1x/job.c b/drivers/gpu/host1x/job.c
index 9424b3c08ab2..f81f3b9f7510 100644
--- a/drivers/gpu/host1x/job.c
+++ b/drivers/gpu/host1x/job.c
@@ -176,7 +176,7 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 			goto unpin;
 		}
 
-		map = host1x_bo_pin(dev, bo, direction, &client->cache);
+		map = host1x_bo_pin(dev, bo, direction, NULL);
 		if (IS_ERR(map)) {
 			host1x_bo_put(bo);
 			err = PTR_ERR(map);
@@ -226,7 +226,7 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 			goto unpin;
 		}
 
-		map = host1x_bo_pin(host->dev, g->bo, DMA_TO_DEVICE, &host->cache);
+		map = host1x_bo_pin(host->dev, g->bo, DMA_TO_DEVICE, NULL);
 		if (IS_ERR(map)) {
 			host1x_bo_put(g->bo);
 			err = PTR_ERR(map);
diff --git a/include/linux/host1x.h b/include/linux/host1x.h
index e8dc5bc41f79..46081afd0ca0 100644
--- a/include/linux/host1x.h
+++ b/include/linux/host1x.h
@@ -31,6 +31,11 @@ u64 host1x_get_dma_mask(struct host1x *host1x);
  * struct host1x_bo_cache - host1x buffer object cache
  * @mappings: list of mappings
  * @lock: synchronizes accesses to the list of mappings
+ *
+ * Note that entries are not periodically evicted from this cache and instead need to be
+ * explicitly released. This is used primarily for DRM/KMS where the cache's reference is
+ * released when the last reference to a buffer object represented by a mapping in this
+ * cache is dropped.
  */
 struct host1x_bo_cache {
 	struct list_head mappings;
-- 
2.35.1

Re: [PATCH] gpu: host1x: Do not use mapping cache for job submissions

[Jon Hunter]

On 24/03/2022 10:30, Thierry Reding wrote:
> From: Thierry Reding <treding@nvidia.com>
> 
> Buffer mappings used in job submissions are usually small and not
> rapidly reused as opposed to framebuffers (which are usually large and
> rapidly reused, for example when page-flipping between double-buffered
> framebuffers). Avoid going through the mapping cache for these buffers
> since the cache would also lead to leaks if nobody is ever releasing
> the cache's last reference. For DRM/KMS these last references are
> dropped when the framebuffers are removed and therefore no longer
> needed.
> 
> While at it, also add a note about the need to explicitly remove the
> final reference to the mapping in the cache.
> 
> Signed-off-by: Thierry Reding <treding@nvidia.com>


I have tested this and verified that it is working well.

Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>

Thanks
Jon

-- 
nvpublic

Re: [PATCH] gpu: host1x: Do not use mapping cache for job submissions

[Jon Hunter]

On 24/03/2022 10:30, Thierry Reding wrote:
> From: Thierry Reding <treding@nvidia.com>
> 
> Buffer mappings used in job submissions are usually small and not
> rapidly reused as opposed to framebuffers (which are usually large and
> rapidly reused, for example when page-flipping between double-buffered
> framebuffers). Avoid going through the mapping cache for these buffers
> since the cache would also lead to leaks if nobody is ever releasing
> the cache's last reference. For DRM/KMS these last references are
> dropped when the framebuffers are removed and therefore no longer
> needed.
> 
> While at it, also add a note about the need to explicitly remove the
> final reference to the mapping in the cache.
> 
> Signed-off-by: Thierry Reding <treding@nvidia.com>


I have tested this and verified that it is working well.

Reviewed-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Jon Hunter <jonathanh@nvidia.com>

Thanks
Jon

-- 
nvpublic

Re: [PATCH] gpu: host1x: Do not use mapping cache for job submissions

[Thierry Reding]

[-- Attachment #1: Type: text/plain, Size: 968 bytes --]

On Thu, Mar 24, 2022 at 11:30:25AM +0100, Thierry Reding wrote:
> From: Thierry Reding <treding@nvidia.com>
> 
> Buffer mappings used in job submissions are usually small and not
> rapidly reused as opposed to framebuffers (which are usually large and
> rapidly reused, for example when page-flipping between double-buffered
> framebuffers). Avoid going through the mapping cache for these buffers
> since the cache would also lead to leaks if nobody is ever releasing
> the cache's last reference. For DRM/KMS these last references are
> dropped when the framebuffers are removed and therefore no longer
> needed.
> 
> While at it, also add a note about the need to explicitly remove the
> final reference to the mapping in the cache.
> 
> Signed-off-by: Thierry Reding <treding@nvidia.com>
> ---
>  drivers/gpu/host1x/job.c | 4 ++--
>  include/linux/host1x.h   | 5 +++++
>  2 files changed, 7 insertions(+), 2 deletions(-)

Applied.

Thierry

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]