ll_rw_block mark_buffer_clean without the big kernel lock

Andrea Arcangeli (andrea@suse.de)
Wed, 23 Jun 1999 18:44:09 +0200 (CEST)

ll_rw_block will refile the buffer, and currently it can do that without
the big kernel lock held (at least in write-partial). I think we can't put
the lock_kernel just around mark_buffer_clean (in a
async_mark_buffer_clean style) because we would end waiting for the big
kernel lock but with the io_request_lock just acquired while other part of
the kernel gets the io_request_lock with the lock_kernel acquired.

I am not sure if we can make the buffer clean outside the io_request_lock
(looks not to be sure that the req->bh won't change under us). Anyway the
below patch will fix the race. It also fixes a wrong param to
async_mark_buffer_dirty.

Index: linux/fs/buffer.c
===================================================================
RCS file: /var/cvs/linux/fs/buffer.c,v
retrieving revision 1.1.1.23
diff -u -r1.1.1.23 buffer.c
--- linux/fs/buffer.c 1999/06/22 22:33:23 1.1.1.23
+++ linux/fs/buffer.c 1999/06/23 16:35:44
@@ -1539,7 +1539,7 @@
* lots of dirty pages.
*/
if (!test_and_set_bit(BH_Dirty, &bh->b_state)) {
- __atomic_mark_buffer_dirty(bh, bdf_prm.b_un.age_buffer);
+ __atomic_mark_buffer_dirty(bh, 0);
if (too_many_dirty_buffers)
balance_dirty(bh->b_dev);
}
Index: linux/drivers/block/ll_rw_blk.c
===================================================================
RCS file: /var/cvs/linux/drivers/block/ll_rw_blk.c,v
retrieving revision 1.1.1.9
diff -u -r1.1.1.9 ll_rw_blk.c
--- linux/drivers/block/ll_rw_blk.c 1999/06/14 17:12:07 1.1.1.9
+++ linux/drivers/block/ll_rw_blk.c 1999/06/23 16:36:26
@@ -634,13 +634,15 @@
for (i = 0; i < nr; i++) {
if (bh[i]) {
set_bit(BH_Req, &bh[i]->b_state);
+ lock_kernel();
#ifdef CONFIG_BLK_DEV_MD
if (MAJOR(bh[i]->b_dev) == MD_MAJOR) {
md_make_request(MINOR (bh[i]->b_dev), rw, bh[i]);
- continue;
- }
+ } else
#endif
- make_request(MAJOR(bh[i]->b_rdev), rw, bh[i]);
+ make_request(MAJOR(bh[i]->b_rdev), rw, bh[i]);
+ unlock_kernel();
+
}
}
return;

Andrea Arcangeli

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