[PATCH] dma-buf: Expose buffer size to userspace

[PATCH] dma-buf: Expose buffer size to userspace

[Christopher James Halse Rogers]

Each dma-buf has an associated size and it's reasonable for userspace
to want to know what it is.

Since userspace already has an fd, expose the size using the
size = lseek(fd, SEEK_END, 0); lseek(fd, SEEK_CUR, 0);
idiom.

Signed-off-by: Christopher James Halse Rogers <christopher.halse.rogers@canonical.com>
---

I've run into a point in the radeon DRM userspace where I need the
size of a dma-buf. I could add a radeon-specific mechanism to get that,
but this seems like something that would be more generally useful.

I'm not entirely sure about supporting both SEEK_END and SEEK_CUR; this
is somewhat of an abuse of lseek, as seeking obviously doesn't make sense.
It's the obivous idiom for getting the size of what's on the other end of a
file descriptor, though.

I didn't notice anywhere to document this; Documentation/dma-buf-api didn't
seem like the right place. Is there somewhere I've overlooked?

 drivers/base/dma-buf.c | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/drivers/base/dma-buf.c b/drivers/base/dma-buf.c
index 6687ba7..c33a857 100644
--- a/drivers/base/dma-buf.c
+++ b/drivers/base/dma-buf.c
@@ -77,9 +77,36 @@ static int dma_buf_mmap_internal(struct file *file, struct vm_area_struct *vma)
 	return dmabuf->ops->mmap(dmabuf, vma);
 }
 
+static loff_t dma_buf_llseek(struct file *file, loff_t offset, int whence)
+{
+	struct dma_buf *dmabuf;
+	loff_t base;
+
+	if (!is_dma_buf_file(file))
+		return -EBADF;
+
+	dmabuf = file->private_data;
+
+	/* only support discovering the end of the buffer,
+	   but also allow SEEK_SET to maintain the idiomatic
+	   SEEK_END(0), SEEK_CUR(0) pattern */
+	if (whence == SEEK_END)
+		base = dmabuf->size;
+	else if (whence == SEEK_SET)
+		base = 0;
+	else
+		return -EINVAL;
+
+	if (offset != 0)
+		return -EINVAL;
+
+	return base + offset;
+}
+
 static const struct file_operations dma_buf_fops = {
 	.release	= dma_buf_release,
 	.mmap		= dma_buf_mmap_internal,
+	.llseek		= dma_buf_llseek,
 };
 
 /*
@@ -133,6 +160,7 @@ struct dma_buf *dma_buf_export_named(void *priv, const struct dma_buf_ops *ops,
 	dmabuf->exp_name = exp_name;
 
 	file = anon_inode_getfile("dmabuf", &dma_buf_fops, dmabuf, flags);
+	file->f_mode |= FMODE_LSEEK;
 
 	dmabuf->file = file;
 
-- 
1.8.3.2

Re: [PATCH] dma-buf: Expose buffer size to userspace

[Daniel Vetter]

On Mon, Aug 05, 2013 at 04:22:00PM +1000, Christopher James Halse Rogers wrote:
> Each dma-buf has an associated size and it's reasonable for userspace
> to want to know what it is.
> 
> Since userspace already has an fd, expose the size using the
> size = lseek(fd, SEEK_END, 0); lseek(fd, SEEK_CUR, 0);
> idiom.
> 
> Signed-off-by: Christopher James Halse Rogers <christopher.halse.rogers@canonical.com>

Yeah, loosk good to me and rather useful, so (with the dma-buf docs
improved as suggested below):

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>

I've also written some small prime tests in igt, so also:

Tested-by: Daniel Vetter <daniel.vetter@ffwll.ch>

> ---
> 
> I've run into a point in the radeon DRM userspace where I need the
> size of a dma-buf. I could add a radeon-specific mechanism to get that,
> but this seems like something that would be more generally useful.
> 
> I'm not entirely sure about supporting both SEEK_END and SEEK_CUR; this
> is somewhat of an abuse of lseek, as seeking obviously doesn't make sense.
> It's the obivous idiom for getting the size of what's on the other end of a
> file descriptor, though.
> 
> I didn't notice anywhere to document this; Documentation/dma-buf-api didn't
> seem like the right place. Is there somewhere I've overlooked?

I think adding a section about various other userspace interfaces exposed
below the mmap support section would be good. Feel free to squash in the
belo diff for v2.

Cheers, Daniel

diff --git a/Documentation/dma-buf-sharing.txt b/Documentation/dma-buf-sharing.txt
index 0b23261..b3a8aa2 100644
--- a/Documentation/dma-buf-sharing.txt
+++ b/Documentation/dma-buf-sharing.txt
@@ -407,6 +407,18 @@ Being able to mmap an export dma-buf buffer object has 2 main use-cases:
    interesting ways depending upong the exporter (if userspace starts depending
    upon this implicit synchronization).
 
+Other Interfaces Exposed to Userspace on the dma-buf FD
+------------------------------------------------------
+
+- Since kernel 3.12 the dma-buf FD supports the llseek system call, but only
+  with offset=0 and whence=SEEK_END|SEEK_SET. SEEK_SET is supported to allowe
+  the usual size discover pattern size = SEEK_END(0); SEEK_SET(0). Every other
+  llseek operation will report -EINVAL.
+
+  If llseek on dma-buf FDs isn't support the kernel will report -ESPIPE for all
+  cases. Userspace can use this to detect support for discovering the dma-buf
+  size using llsee.
+
 Miscellaneous notes
 -------------------
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch