* [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait
@ 2016-06-09 15:05 Gustavo Padovan
2016-06-09 15:05 ` [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence() Gustavo Padovan
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Gustavo Padovan @ 2016-06-09 15:05 UTC (permalink / raw)
To: dri-devel
Cc: linux-kernel, Daniel Stone, Daniel Vetter, Rob Clark,
Greg Hackmann, John Harrison, laurent.pinchart, seanpaul,
marcheu, m.chehab, Sumit Semwal, Maarten Lankhorst,
Gustavo Padovan
From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Signalling doesn't need to be enabled at sync_file creation, it is only
required if userspace waiting the fence to signal through poll().
Thus we delay fence_add_callback() until poll is called. It only adds the
callback the first time poll() is called. This avoid re-adding the same
callback multiple times.
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
drivers/dma-buf/sync_file.c | 36 +++++++++++++++++++++++-------------
include/linux/fence.h | 12 ++++++++++++
2 files changed, 35 insertions(+), 13 deletions(-)
diff --git a/drivers/dma-buf/sync_file.c b/drivers/dma-buf/sync_file.c
index 00e9186..59f3445 100644
--- a/drivers/dma-buf/sync_file.c
+++ b/drivers/dma-buf/sync_file.c
@@ -91,9 +91,7 @@ struct sync_file *sync_file_create(struct fence *fence)
sync_file->cbs[0].fence = fence;
sync_file->cbs[0].sync_file = sync_file;
- if (fence_add_callback(fence, &sync_file->cbs[0].cb,
- fence_check_cb_func))
- atomic_dec(&sync_file->status);
+ INIT_LIST_HEAD(&sync_file->cbs[0].cb.node);
sync_file_debug_add(sync_file);
return sync_file;
@@ -130,8 +128,7 @@ static void sync_file_add_pt(struct sync_file *sync_file, int *i,
sync_file->cbs[*i].fence = fence;
sync_file->cbs[*i].sync_file = sync_file;
- if (!fence_add_callback(fence, &sync_file->cbs[*i].cb,
- fence_check_cb_func)) {
+ if (!fence_is_signaled(fence)) {
fence_get(fence);
(*i)++;
}
@@ -212,11 +209,8 @@ static void sync_file_free(struct kref *kref)
kref);
int i;
- for (i = 0; i < sync_file->num_fences; ++i) {
- fence_remove_callback(sync_file->cbs[i].fence,
- &sync_file->cbs[i].cb);
+ for (i = 0; i < sync_file->num_fences; ++i)
fence_put(sync_file->cbs[i].fence);
- }
sync_file_debug_remove(sync_file);
kfree(sync_file);
@@ -233,17 +227,33 @@ static int sync_file_release(struct inode *inode, struct file *file)
static unsigned int sync_file_poll(struct file *file, poll_table *wait)
{
struct sync_file *sync_file = file->private_data;
- int status;
+ int status, i;
poll_wait(file, &sync_file->wq, wait);
+ for (i = 0 ; i < sync_file->num_fences ; i++) {
+ if (fence_is_enabled(sync_file->cbs[i].fence))
+ continue;
+
+ if (fence_add_callback(sync_file->cbs[i].fence,
+ &sync_file->cbs[i].cb,
+ fence_check_cb_func))
+ atomic_dec(&sync_file->status);
+ }
+
status = atomic_read(&sync_file->status);
- if (!status)
- return POLLIN;
+ if (status > 0)
+ return 0;
+
+ for (i = 0 ; i < sync_file->num_fences ; i++)
+ fence_remove_callback(sync_file->cbs[i].fence,
+ &sync_file->cbs[i].cb);
+
if (status < 0)
return POLLERR;
- return 0;
+
+ return POLLIN;
}
static long sync_file_ioctl_merge(struct sync_file *sync_file,
diff --git a/include/linux/fence.h b/include/linux/fence.h
index 523ea3f..95a70b2 100644
--- a/include/linux/fence.h
+++ b/include/linux/fence.h
@@ -283,6 +283,18 @@ fence_is_signaled(struct fence *fence)
}
/**
+ * fence_is_enabled - return an indication if signalling is enabled
+ * @fence: [in] the fence to check
+ *
+ * Returns true if signalling was already enabled, false if not.
+ */
+static inline bool
+fence_is_enabled(struct fence *fence)
+{
+ return test_bit(FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags);
+}
+
+/**
* fence_is_later - return if f1 is chronologically later than f2
* @f1: [in] the first fence from the same context
* @f2: [in] the second fence from the same context
--
2.5.5
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence()
2016-06-09 15:05 [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait Gustavo Padovan
@ 2016-06-09 15:05 ` Gustavo Padovan
2016-06-10 12:23 ` Chris Wilson
2016-06-09 15:05 ` [PATCH 3/3] Documentation: add doc for sync_file_get_fence() Gustavo Padovan
2016-06-09 22:12 ` [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait Daniel Vetter
2 siblings, 1 reply; 6+ messages in thread
From: Gustavo Padovan @ 2016-06-09 15:05 UTC (permalink / raw)
To: dri-devel
Cc: linux-kernel, Daniel Stone, Daniel Vetter, Rob Clark,
Greg Hackmann, John Harrison, laurent.pinchart, seanpaul,
marcheu, m.chehab, Sumit Semwal, Maarten Lankhorst,
Gustavo Padovan
From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Creates a function that given an sync file descriptor returns a
fence_collection containing all fences in the sync_file.
If there is only one fence in the sync_file this fence itself is returned,
however if there is more than one, a fence_collection fence is returned.
v2: Comments by Daniel Vetter
- Adapt to new version of fence_collection_init()
- Hold a reference for the fence we return
v3: - Adapt to use fput() directly
- rename to sync_file_get_fence() as we always return one fence
v4: Adapt to use fence_array
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
drivers/dma-buf/sync_file.c | 61 +++++++++++++++++++++++++++++++++++++++++++++
include/linux/sync_file.h | 4 ++-
2 files changed, 64 insertions(+), 1 deletion(-)
diff --git a/drivers/dma-buf/sync_file.c b/drivers/dma-buf/sync_file.c
index 59f3445..daaf411 100644
--- a/drivers/dma-buf/sync_file.c
+++ b/drivers/dma-buf/sync_file.c
@@ -23,6 +23,7 @@
#include <linux/slab.h>
#include <linux/uaccess.h>
#include <linux/anon_inodes.h>
+#include <linux/fence-array.h>
#include <linux/sync_file.h>
#include <uapi/linux/sync_file.h>
@@ -122,6 +123,63 @@ err:
return NULL;
}
+/**
+ * sync_file_get_fence - get the fence related to the sync_file fd
+ * @fd: sync_file fd to get the fence from
+ *
+ * Ensures @fd references a valid sync_file and returns a fence that
+ * represents all fence in the sync_file.
+ *
+ * If there is only one fence in the sync_file, this fence is returned.
+ * If there is more than one, a fence_array containing all fences
+ * is created and its base fence object is returned.
+ * On both cases a reference to the returned fence is held. On error
+ * NULL is returned.
+ */
+struct fence *sync_file_get_fence(int fd)
+{
+ struct sync_file *sync_file;
+ struct fence_array *fence_array;
+ struct fence **fences;
+ int i;
+
+ sync_file = sync_file_fdget(fd);
+ if (!sync_file)
+ return NULL;
+
+ if (sync_file->num_fences == 1) {
+ struct fence *fence = sync_file->cbs[0].fence;
+
+ fence_get(fence);
+ fput(sync_file->file);
+ return fence;
+ }
+
+ fences = kcalloc(sync_file->num_fences, sizeof(**fences), GFP_KERNEL);
+ if (!fences) {
+ fput(sync_file->file);
+ return NULL;
+ }
+
+ for (i = 0 ; i < sync_file->num_fences ; i++)
+ fences[i] = sync_file->cbs[i].fence;
+
+ fence_array = fence_array_create(sync_file->num_fences, fences,
+ fence_context_alloc(1), 1, false);
+ if (!fence_array) {
+ kfree(fences);
+ fput(sync_file->file);
+ return NULL;
+ }
+
+ sync_file->fence_array = fence_array;
+ fence_get(&fence_array->base);
+ fput(sync_file->file);
+
+ return &fence_array->base;
+}
+EXPORT_SYMBOL(sync_file_get_fence);
+
static void sync_file_add_pt(struct sync_file *sync_file, int *i,
struct fence *fence)
{
@@ -209,6 +267,9 @@ static void sync_file_free(struct kref *kref)
kref);
int i;
+ if (sync_file->fence_array)
+ fence_put(&sync_file->fence_array->base);
+
for (i = 0; i < sync_file->num_fences; ++i)
fence_put(sync_file->cbs[i].fence);
diff --git a/include/linux/sync_file.h b/include/linux/sync_file.h
index c6ffe8b..33cb3d6 100644
--- a/include/linux/sync_file.h
+++ b/include/linux/sync_file.h
@@ -35,6 +35,7 @@ struct sync_file_cb {
* @num_fences: number of sync_pts in the fence
* @wq: wait queue for fence signaling
* @status: 0: signaled, >0:active, <0: error
+ * @fence_array: array with the fences in the sync_file
* @cbs: sync_pts callback information
*/
struct sync_file {
@@ -49,9 +50,10 @@ struct sync_file {
wait_queue_head_t wq;
atomic_t status;
+ struct fence_array *fence_array;
struct sync_file_cb cbs[];
};
struct sync_file *sync_file_create(struct fence *fence);
-
+struct fence *sync_file_get_fence(int fd);
#endif /* _LINUX_SYNC_H */
--
2.5.5
^ permalink raw reply related [flat|nested] 6+ messages in thread
* [PATCH 3/3] Documentation: add doc for sync_file_get_fence()
2016-06-09 15:05 [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait Gustavo Padovan
2016-06-09 15:05 ` [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence() Gustavo Padovan
@ 2016-06-09 15:05 ` Gustavo Padovan
2016-06-09 22:12 ` [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait Daniel Vetter
2 siblings, 0 replies; 6+ messages in thread
From: Gustavo Padovan @ 2016-06-09 15:05 UTC (permalink / raw)
To: dri-devel
Cc: linux-kernel, Daniel Stone, Daniel Vetter, Rob Clark,
Greg Hackmann, John Harrison, laurent.pinchart, seanpaul,
marcheu, m.chehab, Sumit Semwal, Maarten Lankhorst,
Gustavo Padovan
From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
Document the new function added to sync_file.c
v2: Adapt to fence_array
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
---
Documentation/sync_file.txt | 19 +++++++++++++++++++
1 file changed, 19 insertions(+)
diff --git a/Documentation/sync_file.txt b/Documentation/sync_file.txt
index eaf8297..46f5769 100644
--- a/Documentation/sync_file.txt
+++ b/Documentation/sync_file.txt
@@ -64,6 +64,25 @@ The sync_file fd now can be sent to userspace.
If the creation process fail, or the sync_file needs to be released by any
other reason fput(sync_file->file) should be used.
+Receiving Sync Files from Userspace
+-----------------------------------
+
+When userspace needs to send an in-fence to the driver it pass file descriptor
+of the Sync File to the kernel. The kernel then can retrieve the sync_file
+from it.
+
+Interface:
+ struct fence *sync_file_get_fence(int fd);
+
+
+The function return a struct fence pointer referencing the fence(s) in the Sync
+File. If the Sync File contains only one fence, this fence is returned. But if
+has more than one fence a fence_array[3] is returned. However the callers only
+see the arrays' base class which is struct fence. In both cases a reference to
+the returned fence is held.
+
+
References:
[1] struct sync_file in include/linux/sync_file.h
[2] All interfaces mentioned above defined in include/linux/sync_file.h
+[1] struct fence_array in include/linux/fence-array.h
--
2.5.5
^ permalink raw reply related [flat|nested] 6+ messages in thread
* Re: [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait
2016-06-09 15:05 [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait Gustavo Padovan
2016-06-09 15:05 ` [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence() Gustavo Padovan
2016-06-09 15:05 ` [PATCH 3/3] Documentation: add doc for sync_file_get_fence() Gustavo Padovan
@ 2016-06-09 22:12 ` Daniel Vetter
2 siblings, 0 replies; 6+ messages in thread
From: Daniel Vetter @ 2016-06-09 22:12 UTC (permalink / raw)
To: Gustavo Padovan
Cc: dri-devel, linux-kernel, Daniel Stone, Daniel Vetter, Rob Clark,
Greg Hackmann, John Harrison, laurent.pinchart, seanpaul,
marcheu, m.chehab, Sumit Semwal, Maarten Lankhorst,
Gustavo Padovan
On Thu, Jun 09, 2016 at 12:05:28PM -0300, Gustavo Padovan wrote:
> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>
> Signalling doesn't need to be enabled at sync_file creation, it is only
> required if userspace waiting the fence to signal through poll().
>
> Thus we delay fence_add_callback() until poll is called. It only adds the
> callback the first time poll() is called. This avoid re-adding the same
> callback multiple times.
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> ---
> drivers/dma-buf/sync_file.c | 36 +++++++++++++++++++++++-------------
> include/linux/fence.h | 12 ++++++++++++
> 2 files changed, 35 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/dma-buf/sync_file.c b/drivers/dma-buf/sync_file.c
> index 00e9186..59f3445 100644
> --- a/drivers/dma-buf/sync_file.c
> +++ b/drivers/dma-buf/sync_file.c
> @@ -91,9 +91,7 @@ struct sync_file *sync_file_create(struct fence *fence)
>
> sync_file->cbs[0].fence = fence;
> sync_file->cbs[0].sync_file = sync_file;
> - if (fence_add_callback(fence, &sync_file->cbs[0].cb,
> - fence_check_cb_func))
> - atomic_dec(&sync_file->status);
> + INIT_LIST_HEAD(&sync_file->cbs[0].cb.node);
>
> sync_file_debug_add(sync_file);
> return sync_file;
> @@ -130,8 +128,7 @@ static void sync_file_add_pt(struct sync_file *sync_file, int *i,
> sync_file->cbs[*i].fence = fence;
> sync_file->cbs[*i].sync_file = sync_file;
>
> - if (!fence_add_callback(fence, &sync_file->cbs[*i].cb,
> - fence_check_cb_func)) {
> + if (!fence_is_signaled(fence)) {
> fence_get(fence);
> (*i)++;
> }
> @@ -212,11 +209,8 @@ static void sync_file_free(struct kref *kref)
> kref);
> int i;
>
> - for (i = 0; i < sync_file->num_fences; ++i) {
> - fence_remove_callback(sync_file->cbs[i].fence,
> - &sync_file->cbs[i].cb);
> + for (i = 0; i < sync_file->num_fences; ++i)
> fence_put(sync_file->cbs[i].fence);
> - }
>
> sync_file_debug_remove(sync_file);
> kfree(sync_file);
> @@ -233,17 +227,33 @@ static int sync_file_release(struct inode *inode, struct file *file)
> static unsigned int sync_file_poll(struct file *file, poll_table *wait)
> {
> struct sync_file *sync_file = file->private_data;
> - int status;
> + int status, i;
>
> poll_wait(file, &sync_file->wq, wait);
>
> + for (i = 0 ; i < sync_file->num_fences ; i++) {
> + if (fence_is_enabled(sync_file->cbs[i].fence))
If some other sync_file (or anyone else really) added enabling, then
you'll miss adding a callback. I don't think this check here works. You
probably also need more testcases in the debugfs-based tests if this falls
through the cracks ;-)
-Daniel
> + continue;
> +
> + if (fence_add_callback(sync_file->cbs[i].fence,
> + &sync_file->cbs[i].cb,
> + fence_check_cb_func))
> + atomic_dec(&sync_file->status);
> + }
> +
> status = atomic_read(&sync_file->status);
>
> - if (!status)
> - return POLLIN;
> + if (status > 0)
> + return 0;
> +
> + for (i = 0 ; i < sync_file->num_fences ; i++)
> + fence_remove_callback(sync_file->cbs[i].fence,
> + &sync_file->cbs[i].cb);
> +
> if (status < 0)
> return POLLERR;
> - return 0;
> +
> + return POLLIN;
> }
>
> static long sync_file_ioctl_merge(struct sync_file *sync_file,
> diff --git a/include/linux/fence.h b/include/linux/fence.h
> index 523ea3f..95a70b2 100644
> --- a/include/linux/fence.h
> +++ b/include/linux/fence.h
> @@ -283,6 +283,18 @@ fence_is_signaled(struct fence *fence)
> }
>
> /**
> + * fence_is_enabled - return an indication if signalling is enabled
> + * @fence: [in] the fence to check
> + *
> + * Returns true if signalling was already enabled, false if not.
> + */
> +static inline bool
> +fence_is_enabled(struct fence *fence)
> +{
> + return test_bit(FENCE_FLAG_ENABLE_SIGNAL_BIT, &fence->flags);
> +}
> +
> +/**
> * fence_is_later - return if f1 is chronologically later than f2
> * @f1: [in] the first fence from the same context
> * @f2: [in] the second fence from the same context
> --
> 2.5.5
>
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence()
2016-06-09 15:05 ` [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence() Gustavo Padovan
@ 2016-06-10 12:23 ` Chris Wilson
2016-06-13 21:21 ` Gustavo Padovan
0 siblings, 1 reply; 6+ messages in thread
From: Chris Wilson @ 2016-06-10 12:23 UTC (permalink / raw)
To: Gustavo Padovan
Cc: dri-devel, marcheu, Daniel Stone, seanpaul, Daniel Vetter,
linux-kernel, laurent.pinchart, Gustavo Padovan, John Harrison,
m.chehab
On Thu, Jun 09, 2016 at 12:05:29PM -0300, Gustavo Padovan wrote:
> From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
>
> Creates a function that given an sync file descriptor returns a
> fence_collection containing all fences in the sync_file.
>
> If there is only one fence in the sync_file this fence itself is returned,
> however if there is more than one, a fence_collection fence is returned.
>
> v2: Comments by Daniel Vetter
> - Adapt to new version of fence_collection_init()
> - Hold a reference for the fence we return
>
> v3: - Adapt to use fput() directly
> - rename to sync_file_get_fence() as we always return one fence
>
> v4: Adapt to use fence_array
The sync_file looks just like a fence_array plus struct file integration.
Is there a good reason why we cannot just use fence_array here?
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence()
2016-06-10 12:23 ` Chris Wilson
@ 2016-06-13 21:21 ` Gustavo Padovan
0 siblings, 0 replies; 6+ messages in thread
From: Gustavo Padovan @ 2016-06-13 21:21 UTC (permalink / raw)
To: Chris Wilson, dri-devel, marcheu, Daniel Stone, seanpaul,
Daniel Vetter, linux-kernel, laurent.pinchart, Gustavo Padovan,
John Harrison, m.chehab
2016-06-10 Chris Wilson <chris@chris-wilson.co.uk>:
> On Thu, Jun 09, 2016 at 12:05:29PM -0300, Gustavo Padovan wrote:
> > From: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
> >
> > Creates a function that given an sync file descriptor returns a
> > fence_collection containing all fences in the sync_file.
> >
> > If there is only one fence in the sync_file this fence itself is returned,
> > however if there is more than one, a fence_collection fence is returned.
> >
> > v2: Comments by Daniel Vetter
> > - Adapt to new version of fence_collection_init()
> > - Hold a reference for the fence we return
> >
> > v3: - Adapt to use fput() directly
> > - rename to sync_file_get_fence() as we always return one fence
> >
> > v4: Adapt to use fence_array
>
> The sync_file looks just like a fence_array plus struct file integration.
> Is there a good reason why we cannot just use fence_array here?
I think the only reason is because I haven't thought about this before.
If we add proper support to fence_array we can make sync_files even
simpler.
Gustavo
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2016-06-13 21:21 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-06-09 15:05 [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait Gustavo Padovan
2016-06-09 15:05 ` [PATCH 2/3] dma-buf/sync_file: add sync_file_get_fence() Gustavo Padovan
2016-06-10 12:23 ` Chris Wilson
2016-06-13 21:21 ` Gustavo Padovan
2016-06-09 15:05 ` [PATCH 3/3] Documentation: add doc for sync_file_get_fence() Gustavo Padovan
2016-06-09 22:12 ` [PATCH 1/3] dma-buf/sync_file: only enable fence signalling during wait Daniel Vetter
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).