On Mon, 17 Jan 2011 22:00:27 -0500
David Dillow <dillowda@xxxxxxxx> wrote:
> When using devices that support max_segments > BIO_MAX_PAGES (256),
> direct IO tries to allocate a bio with more pages than allowed, which
> leads to an oops in dio_bio_alloc(). Clamp the request to the supported
> maximum, and change dio_bio_alloc() to reflect that bio_alloc() will
> always return a bio when called with __GFP_WAIT and a valid number of
> vectors.
Which device driver triggers this condition?
> Also added cc to stable, as this has been a longstanding item.
Well. -stable only needs the patch if the driver which triggers the
problem is also there. But we don't know what driver that is, yet???
> --- a/fs/direct-io.c
> +++ b/fs/direct-io.c
> @@ -325,13 +325,18 @@ void dio_end_io(struct bio *bio, int error)
> }
> EXPORT_SYMBOL_GPL(dio_end_io);
>
> -static int
> +static void
> dio_bio_alloc(struct dio *dio, struct block_device *bdev,
> sector_t first_sector, int nr_vecs)
> {
> struct bio *bio;
>
> + /*
> + * bio_alloc() is guaranteed to return a bio when called with
> + * __GFP_WAIT and we request a valid number of vectors.
> + */
> bio = bio_alloc(GFP_KERNEL, nr_vecs);
> + BUG_ON(!bio);
This BUG_ON() is pretty pointless,
> bio->bi_bdev = bdev;
because the next statement will reliably oops, providing us with the
same information.
> bio->bi_sector = first_sector;
> @@ -342,7 +347,6 @@ dio_bio_alloc(struct dio *dio, struct block_device *bdev,
>
> dio->bio = bio;
> dio->logical_offset_in_bio = dio->cur_page_fs_offset;
> - return 0;
> }
>
> /*
> @@ -583,8 +587,9 @@ static int dio_new_bio(struct dio *dio, sector_t start_sector)
> goto out;
> sector = start_sector << (dio->blkbits - 9);
> nr_pages = min(dio->pages_in_io, bio_get_nr_vecs(dio->map_bh.b_bdev));
> + nr_pages = min(nr_pages, BIO_MAX_PAGES);
> BUG_ON(nr_pages <= 0);
> - ret = dio_bio_alloc(dio, dio->map_bh.b_bdev, sector, nr_pages);
> + dio_bio_alloc(dio, dio->map_bh.b_bdev, sector, nr_pages);
> dio->boundary = 0;
> out:
> return ret;
>
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
