Some requests signal partial completion. We currently record this
by updating bi_idx, bv_len, and bv_offset.
This is bad if the bi_io_vec is to be shared.
So instead keep in "struct request" the amount of the first bio
that has completed. This is "first_offset" (i.e. offset in to
first bio). Update and use that instead.
Signed-off-by: Neil Brown <[EMAIL PROTECTED]>
### Diffstat output
./block/ll_rw_blk.c | 38 ++++++++++++++++++++++++++++++--------
./drivers/ide/ide-io.c | 2 +-
./include/linux/blkdev.h | 22 +++++++++++++++++-----
3 files changed, 48 insertions(+), 14 deletions(-)
diff .prev/block/ll_rw_blk.c ./block/ll_rw_blk.c
--- .prev/block/ll_rw_blk.c 2007-07-31 11:20:46.000000000 +1000
+++ ./block/ll_rw_blk.c 2007-07-31 11:20:46.000000000 +1000
@@ -243,6 +243,7 @@ static void rq_init(struct request_queue
rq->errors = 0;
rq->bio = rq->biotail = NULL;
+ rq->first_offset = 0;
INIT_HLIST_NODE(&rq->hash);
RB_CLEAR_NODE(&rq->rb_node);
rq->ioprio = 0;
@@ -447,6 +448,7 @@ static inline struct request *start_orde
rq->cmd_flags |= q->ordered & QUEUE_ORDERED_FUA ? REQ_FUA : 0;
rq->elevator_private = NULL;
rq->elevator_private2 = NULL;
+ BUG_ON(rq->first_offset);
init_request_from_bio(rq, q->orig_bar_rq->bio);
rq->end_io = bar_end_io;
@@ -1214,6 +1216,7 @@ void blk_recount_segments(struct request
struct bio *nxt = bio->bi_next;
rq.q = q;
rq.bio = rq.biotail = bio;
+ rq.first_offset = 0;
bio->bi_next = NULL;
blk_recalc_rq_segments(&rq);
bio->bi_next = nxt;
@@ -2926,6 +2929,7 @@ static void init_request_from_bio(struct
req->hard_sector = req->sector = bio->bi_sector;
req->hard_nr_sectors = req->nr_sectors = bio_sectors(bio);
req->bio = req->biotail = bio;
+ req->first_offset = 0;
req->current_nr_sectors = req->hard_cur_sectors =
blk_rq_cur_sectors(req);
req->nr_phys_segments = bio_phys_segments(req->q, bio);
@@ -3411,22 +3415,30 @@ static int __end_that_request_first(stru
nbytes = bio->bi_size;
if (!ordered_bio_endio(req, bio, nbytes, error))
bio_endio(bio, nbytes, error);
+ req->first_offset = 0;
next_idx = 0;
bio_nbytes = 0;
} else {
int idx = bio->bi_idx + next_idx;
- if (unlikely(bio->bi_idx >= bio->bi_vcnt)) {
+ if (unlikely(idx >= bio->bi_vcnt)) {
blk_dump_rq_flags(req, "__end_that");
printk("%s: bio idx %d >= vcnt %d\n",
__FUNCTION__,
- bio->bi_idx, bio->bi_vcnt);
+ idx, bio->bi_vcnt);
break;
}
nbytes = bio_iovec_idx(bio, idx)->bv_len;
BIO_BUG_ON(nbytes > bio->bi_size);
+ if (req->first_offset > bio_nbytes + nbytes) {
+ bio_nbytes += nbytes;
+ nbytes = 0;
+ } else if (req->first_offset > bio_nbytes) {
+ nbytes -= req->first_offset - bio_nbytes;
+ bio_nbytes = req->first_offset;
+ }
/*
* not a complete bvec done
*/
@@ -3467,9 +3479,7 @@ static int __end_that_request_first(stru
if (bio_nbytes) {
if (!ordered_bio_endio(req, bio, bio_nbytes, error))
bio_endio(bio, bio_nbytes, error);
- bio->bi_idx += next_idx;
- bio_iovec(bio)->bv_offset += nr_bytes;
- bio_iovec(bio)->bv_len -= nr_bytes;
+ req->first_offset = bio_nbytes;
}
blk_recalc_rq_sectors(req, total_bytes >> 9);
@@ -3658,6 +3668,7 @@ void blk_rq_bio_prep(struct request_queu
rq->hard_nr_sectors = rq->nr_sectors = bio_sectors(bio);
rq->data_len = bio->bi_size;
+ rq->first_offset = 0;
rq->bio = rq->biotail = bio;
rq->buffer = blk_rq_data(rq);
rq->current_nr_sectors = blk_rq_cur_sectors(rq);
@@ -3668,14 +3679,25 @@ EXPORT_SYMBOL(blk_rq_bio_prep);
void *blk_rq_data(struct request *rq)
{
- return page_address(bio_page(rq->bio)) +
- bio_offset(rq->bio);
+ struct bio_vec bvec;
+ struct req_iterator i;
+
+ rq_for_each_segment(rq, i, bvec)
+ return page_address(bvec.bv_page) + bvec.bv_offset;
+
+ return NULL;
}
EXPORT_SYMBOL(blk_rq_data);
int blk_rq_cur_bytes(struct request *rq)
{
- return bio_iovec(rq->bio)->bv_len;
+ struct bio_vec bvec;
+ struct req_iterator i;
+
+ rq_for_each_segment(rq, i, bvec)
+ return bvec.bv_len;
+
+ return 0;
}
EXPORT_SYMBOL(blk_rq_cur_bytes);
diff .prev/drivers/ide/ide-io.c ./drivers/ide/ide-io.c
--- .prev/drivers/ide/ide-io.c 2007-07-31 11:20:43.000000000 +1000
+++ ./drivers/ide/ide-io.c 2007-07-31 11:20:46.000000000 +1000
@@ -1415,7 +1415,7 @@ static ide_startstop_t ide_dma_timeout_r
if (!rq->bio)
goto out;
- rq->sector = rq->bio->bi_sector;
+ rq->sector = rq->bio->bi_sector + (rq->first_offset >> 9);
rq->current_nr_sectors = blk_rq_cur_sectors(rq);
rq->hard_cur_sectors = rq->current_nr_sectors;
rq->buffer = blk_rq_data(rq);
diff .prev/include/linux/blkdev.h ./include/linux/blkdev.h
--- .prev/include/linux/blkdev.h 2007-07-31 11:20:46.000000000 +1000
+++ ./include/linux/blkdev.h 2007-07-31 11:20:46.000000000 +1000
@@ -254,6 +254,7 @@ struct request {
struct bio *bio;
struct bio *biotail;
+ int first_offset; /* offset into first bio in list */
struct hlist_node hash; /* merge hash */
/*
@@ -640,14 +641,25 @@ static inline void blk_queue_bounce(stru
struct req_iterator {
int i;
struct bio *bio;
+ int offset;
};
#define rq_for_each_segment(rq, _iter, bvec) \
- for (_iter.bio = (rq)->bio; _iter.bio; _iter.bio = _iter.bio->bi_next) \
- for (_iter.i = _iter.bio->bi_idx, \
- bvec = *bio_iovec_idx(_iter.bio, _iter.i); \
+ for (_iter.bio = (rq)->bio, _iter.offset = (rq)->first_offset; \
+ _iter.bio; \
+ _iter.bio = _iter.bio->bi_next, _iter.offset = 0) \
+ for (_iter.i = _iter.bio->bi_idx; \
_iter.i < _iter.bio->bi_vcnt; \
- _iter.i++, bvec = *bio_iovec_idx(_iter.bio, _iter.i) \
- )
+ _iter.i++ \
+ ) \
+ if (bvec = *bio_iovec_idx(_iter.bio, _iter.i), \
+ bvec.bv_offset += _iter.offset, \
+ bvec.bv_len <= _iter.offset \
+ ? (_iter.offset -= bvec.bv_len, 0) \
+ : (bvec.bv_len -= _iter.offset, \
+ _iter.offset = 0, \
+ 1))
+
+
#define rq_iter_last(rq, _iter) (_iter.bio->bi_next == NULL && \
_iter.i == _iter.bio->bi_vcnt - 1)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
