Mailing List Archive: [PATCH] ivtv-udma.c: handle get_user_pages() returning fewer pages than we asked for

From: Paul Cassella <fortytwo-ivtv@manetheren.bigw.org>

get_user_pages() may return fewer page than we ask for. As we give up and
return an error in this case, we need to put_page() each of the pages we
did get.

Signed-off-by: Paul Cassella <fortytwo-ivtv@manetheren.bigw.org>

---
I'm running with this on 2.6.37, though haven't triggered the condition.
This patch is against staging/for_v2.6.39, which compiles cleanly with it.

I'm not sure -EINVAL is the best return code vs -EFAULT or -ENOMEM, but
this mod doesn't change it.

drivers/media/video/ivtv/ivtv-udma.c | 8 +++++++-
1 files changed, 7 insertions(+), 1 deletions(-)

diff --git a/drivers/media/video/ivtv/ivtv-udma.c b/drivers/media/video/ivtv/ivtv-udma.c
index 1daf1dd..443ca35 100644
--- a/drivers/media/video/ivtv/ivtv-udma.c
+++ b/drivers/media/video/ivtv/ivtv-udma.c
@@ -132,7 +132,13 @@ int ivtv_udma_setup(struct ivtv *itv, unsigned long ivtv_dest_addr,
if (user_dma.page_count != err) {
IVTV_DEBUG_WARN("failed to map user pages, returned %d instead of %d\n",
err, user_dma.page_count);
- return -EINVAL;
+ if (err >= 0) {
+ for (i = 0; i < err; i++) {
+ put_page(dma->map[i]);
+ }
+ return -EINVAL;
+ }
+ return err;
}

dma->page_count = user_dma.page_count;
--
1.7.2.3

_______________________________________________
ivtv-devel mailing list
ivtv-devel@ivtvdriver.org
http://ivtvdriver.org/mailman/listinfo/ivtv-devel

On Sat, 2011-02-05 at 22:36 -0800, Paul Cassella wrote:
> From: Paul Cassella <fortytwo-ivtv@manetheren.bigw.org>
>
> get_user_pages() may return fewer page than we ask for. As we give up and
> return an error in this case, we need to put_page() each of the pages we
> did get.
>
> Signed-off-by: Paul Cassella <fortytwo-ivtv@manetheren.bigw.org>

Looks good.
I removed the braces in the 'for' statement to eliminate a checkpatch.pl
warning,

Signed-off-by: Andy Walls <awalls@md.metrocast.net>

Your first two patches are queued up here:

http://git.linuxtv.org/awalls/media_tree.git?a=shortlog;h=refs/heads/ivtv_39

> ---
> I'm running with this on 2.6.37, though haven't triggered the condition.
> This patch is against staging/for_v2.6.39, which compiles cleanly with it.
>
> I'm not sure -EINVAL is the best return code vs -EFAULT or -ENOMEM, but
> this mod doesn't change it.

It doesn't matter. The only caller of ivtv_udma_setup() doesn't care
about the exact value of an error return; it substitutes -EIO.

Regards,
Andy

> drivers/media/video/ivtv/ivtv-udma.c | 8 +++++++-
> 1 files changed, 7 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/media/video/ivtv/ivtv-udma.c b/drivers/media/video/ivtv/ivtv-udma.c
> index 1daf1dd..443ca35 100644
> --- a/drivers/media/video/ivtv/ivtv-udma.c
> +++ b/drivers/media/video/ivtv/ivtv-udma.c
> @@ -132,7 +132,13 @@ int ivtv_udma_setup(struct ivtv *itv, unsigned long ivtv_dest_addr,
> if (user_dma.page_count != err) {
> IVTV_DEBUG_WARN("failed to map user pages, returned %d instead of %d\n",
> err, user_dma.page_count);
> - return -EINVAL;
> + if (err >= 0) {
> + for (i = 0; i < err; i++) {
> + put_page(dma->map[i]);
> + }
> + return -EINVAL;
> + }
> + return err;
> }
>
> dma->page_count = user_dma.page_count;

_______________________________________________
ivtv-devel mailing list
ivtv-devel@ivtvdriver.org
http://ivtvdriver.org/mailman/listinfo/ivtv-devel