bio: always copy back data for copied kernel requests
authorTejun Heo <tj@kernel.org>
Tue, 19 May 2009 09:33:06 +0000 (18:33 +0900)
committerJens Axboe <jens.axboe@oracle.com>
Tue, 19 May 2009 09:36:08 +0000 (11:36 +0200)
When a read bio_copy_kern() request fails, the content of the bounce
buffer is not copied back.  However, as request failure doesn't
necessarily mean complete failure, the buffer state can be useful.
This behavior is also inconsistent with the user map counterpart and
causes the subtle difference between bounced and unbounced IO causes
confusion.

This patch makes bio_copy_kern_endio() ignore @err and always copy
back data on request completion.

Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Boaz Harrosh <bharrosh@panasas.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
fs/bio.c

index 7bbc98f0eda151d27e22a720998a5f8480ce78f6..ee3bc67833d2bd24ff2dff6daad4da3a0e9b2efb 100644 (file)
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -1198,7 +1198,7 @@ static void bio_copy_kern_endio(struct bio *bio, int err)
                char *addr = page_address(bvec->bv_page);
                int len = bmd->iovecs[i].bv_len;
 
-               if (read && !err)
+               if (read)
                        memcpy(p, addr, len);
 
                __free_page(bvec->bv_page);