Re: [PATCH] 2.5.67-ac2 direct-IO for IDE taskfile ioctl (1/4)

Jens Axboe (axboe@suse.de)
Thu, 24 Apr 2003 10:23:31 +0200


On Wed, Apr 23 2003, Bartlomiej Zolnierkiewicz wrote:
> bio = bio_map_user(bdev, uaddr, hdr.dxfer_len, reading);
> - if (bio) {
> - if (writing)
> - bio->bi_rw |= (1 << BIO_RW);
> -
> - nr_sectors = (bio->bi_size + 511) >> 9;
> -
> - if (bio->bi_size < hdr.dxfer_len) {
> - bio_endio(bio, bio->bi_size, 0);
> - bio_unmap_user(bio, 0);
> - bio = NULL;
> - }
> - }
> + if (bio && writing)
> + bio->bi_rw |= (1 << BIO_RW);

I think it's cleaner to have bio_map_user() set the direction bit,
instead of having every user do it. It also uncovered an old bug where
blk_queue_bounce() is called without the direction bit set in the bio,
uh oh...

Here's my preferred version. You also had the incremental bio processing
as a prereq for this bio addition, I removed that for now as well.

# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher.
# This patch includes the following deltas:
# ChangeSet 1.1216 -> 1.1217
# drivers/block/ll_rw_blk.c 1.167 -> 1.168
# include/linux/blkdev.h 1.100 -> 1.101
# fs/bio.c 1.41 -> 1.42
# drivers/block/scsi_ioctl.c 1.22 -> 1.23
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 03/04/24 axboe@smithers.home.kernel.dk 1.1217
# - Abstract out bio request preparation
# - Have bio_map_user() set data direction (fixes bug where blk_queue_bounce()
# is called without it set)
# - Split bio_map_user() in two
# --------------------------------------------
#
diff -Nru a/drivers/block/ll_rw_blk.c b/drivers/block/ll_rw_blk.c
--- a/drivers/block/ll_rw_blk.c Thu Apr 24 10:23:09 2003
+++ b/drivers/block/ll_rw_blk.c Thu Apr 24 10:23:09 2003
@@ -2196,6 +2196,21 @@
}
}

+void blk_rq_bio_prep(request_queue_t *q, struct request *rq, struct bio *bio)
+{
+ /* first three bits are identical in rq->flags and bio->bi_rw */
+ rq->flags |= (bio->bi_rw & 7);
+
+ rq->nr_phys_segments = bio_phys_segments(q, bio);
+ rq->nr_hw_segments = bio_hw_segments(q, bio);
+ rq->current_nr_sectors = bio_cur_sectors(bio);
+ rq->hard_cur_sectors = rq->current_nr_sectors;
+ rq->hard_nr_sectors = rq->nr_sectors = bio_sectors(bio);
+ rq->buffer = bio_data(bio);
+
+ rq->bio = rq->biotail = bio;
+}
+
int __init blk_dev_init(void)
{
int total_ram = nr_free_pages() << (PAGE_SHIFT - 10);
@@ -2285,3 +2300,5 @@
EXPORT_SYMBOL(__blk_stop_queue);
EXPORT_SYMBOL(blk_run_queue);
EXPORT_SYMBOL(blk_run_queues);
+
+EXPORT_SYMBOL(blk_rq_bio_prep);
diff -Nru a/drivers/block/scsi_ioctl.c b/drivers/block/scsi_ioctl.c
--- a/drivers/block/scsi_ioctl.c Thu Apr 24 10:23:09 2003
+++ b/drivers/block/scsi_ioctl.c Thu Apr 24 10:23:09 2003
@@ -193,18 +193,6 @@
* be a write to vm.
*/
bio = bio_map_user(bdev, uaddr, hdr.dxfer_len, reading);
- if (bio) {
- if (writing)
- bio->bi_rw |= (1 << BIO_RW);
-
- nr_sectors = (bio->bi_size + 511) >> 9;
-
- if (bio->bi_size < hdr.dxfer_len) {
- bio_endio(bio, bio->bi_size, 0);
- bio_unmap_user(bio, 0);
- bio = NULL;
- }
- }

/*
* if bio setup failed, fall back to slow approach
@@ -243,21 +231,10 @@
rq->hard_nr_sectors = rq->nr_sectors = nr_sectors;
rq->hard_cur_sectors = rq->current_nr_sectors = nr_sectors;

- if (bio) {
- /*
- * subtle -- if bio_map_user() ended up bouncing a bio, it
- * would normally disappear when its bi_end_io is run.
- * however, we need it for the unmap, so grab an extra
- * reference to it
- */
- bio_get(bio);
+ rq->bio = rq->biotail = bio;

- rq->nr_phys_segments = bio_phys_segments(q, bio);
- rq->nr_hw_segments = bio_hw_segments(q, bio);
- rq->current_nr_sectors = bio_cur_sectors(bio);
- rq->hard_cur_sectors = rq->current_nr_sectors;
- rq->buffer = bio_data(bio);
- }
+ if (bio)
+ blk_rq_bio_prep(q, rq, bio);

rq->data_len = hdr.dxfer_len;
rq->data = buffer;
@@ -268,8 +245,6 @@
if (!rq->timeout)
rq->timeout = BLK_DEFAULT_TIMEOUT;

- rq->bio = rq->biotail = bio;
-
start_time = jiffies;

/* ignore return value. All information is passed back to caller
@@ -277,11 +252,9 @@
* N.B. a non-zero SCSI status is _not_ necessarily an error.
*/
blk_do_rq(q, bdev, rq);
-
- if (bio) {
+
+ if (bio)
bio_unmap_user(bio, reading);
- bio_put(bio);
- }

/* write to all output members */
hdr.status = rq->errors;
diff -Nru a/fs/bio.c b/fs/bio.c
--- a/fs/bio.c Thu Apr 24 10:23:09 2003
+++ b/fs/bio.c Thu Apr 24 10:23:09 2003
@@ -434,19 +434,9 @@
return len;
}

-/**
- * bio_map_user - map user address into bio
- * @bdev: destination block device
- * @uaddr: start of user address
- * @len: length in bytes
- * @write_to_vm: bool indicating writing to pages or not
- *
- * Map the user space address into a bio suitable for io to a block
- * device. Caller should check the size of the returned bio, we might
- * not have mapped the entire range specified.
- */
-struct bio *bio_map_user(struct block_device *bdev, unsigned long uaddr,
- unsigned int len, int write_to_vm)
+static struct bio *__bio_map_user(struct block_device *bdev,
+ unsigned long uaddr, unsigned int len,
+ int write_to_vm)
{
unsigned long end = (uaddr + len + PAGE_SIZE - 1) >> PAGE_SHIFT;
unsigned long start = uaddr >> PAGE_SHIFT;
@@ -510,8 +500,11 @@
kfree(pages);

/*
- * check if the mapped pages need bouncing for an isa host.
+ * set data direction, and check if mapped pages need bouncing
*/
+ if (!write_to_vm)
+ bio->bi_rw |= (1 << BIO_RW);
+
blk_queue_bounce(q, &bio);
return bio;
out:
@@ -521,17 +514,42 @@
}

/**
- * bio_unmap_user - unmap a bio
- * @bio: the bio being unmapped
- * @write_to_vm: bool indicating whether pages were written to
- *
- * Unmap a bio previously mapped by bio_map_user(). The @write_to_vm
- * must be the same as passed into bio_map_user(). Must be called with
- * a process context.
+ * bio_map_user - map user address into bio
+ * @bdev: destination block device
+ * @uaddr: start of user address
+ * @len: length in bytes
+ * @write_to_vm: bool indicating writing to pages or not
*
- * bio_unmap_user() may sleep.
+ * Map the user space address into a bio suitable for io to a block
+ * device.
*/
-void bio_unmap_user(struct bio *bio, int write_to_vm)
+struct bio *bio_map_user(struct block_device *bdev, unsigned long uaddr,
+ unsigned int len, int write_to_vm)
+{
+ struct bio *bio;
+
+ bio = __bio_map_user(bdev, uaddr, len, write_to_vm);
+
+ if (bio) {
+ if (bio->bi_size < len) {
+ bio_endio(bio, bio->bi_size, 0);
+ bio_unmap_user(bio, 0);
+ return NULL;
+ }
+
+ /*
+ * subtle -- if __bio_map_user() ended up bouncing a bio,
+ * it would normally disappear when its bi_end_io is run.
+ * however, we need it for the unmap, so grab an extra
+ * reference to it
+ */
+ bio_get(bio);
+ }
+
+ return bio;
+}
+
+static void __bio_unmap_user(struct bio *bio, int write_to_vm)
{
struct bio_vec *bvec;
int i;
@@ -558,6 +576,23 @@
page_cache_release(bvec->bv_page);
}

+ bio_put(bio);
+}
+
+/**
+ * bio_unmap_user - unmap a bio
+ * @bio: the bio being unmapped
+ * @write_to_vm: bool indicating whether pages were written to
+ *
+ * Unmap a bio previously mapped by bio_map_user(). The @write_to_vm
+ * must be the same as passed into bio_map_user(). Must be called with
+ * a process context.
+ *
+ * bio_unmap_user() may sleep.
+ */
+void bio_unmap_user(struct bio *bio, int write_to_vm)
+{
+ __bio_unmap_user(bio, write_to_vm);
bio_put(bio);
}

diff -Nru a/include/linux/blkdev.h b/include/linux/blkdev.h
--- a/include/linux/blkdev.h Thu Apr 24 10:23:09 2003
+++ b/include/linux/blkdev.h Thu Apr 24 10:23:09 2003
@@ -391,6 +391,8 @@
extern void blk_queue_invalidate_tags(request_queue_t *);
extern void blk_congestion_wait(int rw, long timeout);

+extern void blk_rq_bio_prep(request_queue_t *, struct request *, struct bio *);
+
#define MAX_PHYS_SEGMENTS 128
#define MAX_HW_SEGMENTS 128
#define MAX_SECTORS 255

-- 
Jens Axboe

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/