[PATCH] ll_rw_blk: allow more flexibility for read_ahead_kb store
authorJens Axboe <axboe@suse.de>
Fri, 21 Jul 2006 18:30:28 +0000 (20:30 +0200)
committerJens Axboe <axboe@nelson.home.kernel.dk>
Sat, 30 Sep 2006 18:29:41 +0000 (20:29 +0200)
It can make sense to set read-ahead larger than a single request.
We should not be enforcing such policy on the user. Additionally,
using the BLKRASET ioctl doesn't impose such a restriction. So
additionally we now expose identical behaviour through the two.

Issue also reported by Anton <cbou@mail.ru>

Signed-off-by: Jens Axboe <axboe@suse.de>
block/ll_rw_blk.c

index 346be9ae31f659174051b399ac1c47be147ad4e6..e3980ec747c13ba30ea90673dc1dc554716ab399 100644 (file)
@@ -3806,9 +3806,6 @@ queue_ra_store(struct request_queue *q, const char *page, size_t count)
        ssize_t ret = queue_var_store(&ra_kb, page, count);
 
        spin_lock_irq(q->queue_lock);
-       if (ra_kb > (q->max_sectors >> 1))
-               ra_kb = (q->max_sectors >> 1);
-
        q->backing_dev_info.ra_pages = ra_kb >> (PAGE_CACHE_SHIFT - 10);
        spin_unlock_irq(q->queue_lock);