dio: zero struct dio with kzalloc instead of manually
Zach Brown [Tue, 21 Aug 2007 00:12:01 +0000 (17:12 -0700)]
This patch uses kzalloc to zero all of struct dio rather than manually
trying to track which fields we rely on being zero.  It passed aio+dio
stress testing and some bug regression testing on ext3.

This patch was introduced by Linus in the conversation that lead up to
Badari's minimal fix to manually zero .map_bh.b_state in commit:

  6a648fa72161d1f6468dabd96c5d3c0db04f598a

It makes the code a bit smaller.  Maybe a couple fewer cachelines to
load, if we're lucky:

   text    data     bss     dec     hex filename
3285925  568506 1304616 5159047  4eb887 vmlinux
3285797  568506 1304616 5158919  4eb807 vmlinux.patched

I was unable to measure a stable difference in the number of cpu cycles
spent in blockdev_direct_IO() when pushing aio+dio 256K reads at
~340MB/s.

So the resulting intent of the patch isn't a performance gain but to
avoid exposing ourselves to the risk of finding another field like
.map_bh.b_state where we rely on zeroing but don't enforce it in the
code.

Signed-off-by: Zach Brown <zach.brown@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/direct-io.c

index 6874785..901dc55 100644 (file)
@@ -958,36 +958,22 @@ direct_io_worker(int rw, struct kiocb *iocb, struct inode *inode,
        ssize_t ret2;
        size_t bytes;
 
-       dio->bio = NULL;
        dio->inode = inode;
        dio->rw = rw;
        dio->blkbits = blkbits;
        dio->blkfactor = inode->i_blkbits - blkbits;
-       dio->start_zero_done = 0;
-       dio->size = 0;
        dio->block_in_file = offset >> blkbits;
-       dio->blocks_available = 0;
-       dio->cur_page = NULL;
 
-       dio->boundary = 0;
-       dio->reap_counter = 0;
        dio->get_block = get_block;
        dio->end_io = end_io;
-       dio->map_bh.b_private = NULL;
-       dio->map_bh.b_state = 0;
        dio->final_block_in_bio = -1;
        dio->next_block_for_io = -1;
 
-       dio->page_errors = 0;
-       dio->io_error = 0;
-       dio->result = 0;
        dio->iocb = iocb;
        dio->i_size = i_size_read(inode);
 
        spin_lock_init(&dio->bio_lock);
        dio->refcount = 1;
-       dio->bio_list = NULL;
-       dio->waiter = NULL;
 
        /*
         * In case of non-aligned buffers, we may need 2 more
@@ -995,8 +981,6 @@ direct_io_worker(int rw, struct kiocb *iocb, struct inode *inode,
         */
        if (unlikely(dio->blkfactor))
                dio->pages_in_io = 2;
-       else
-               dio->pages_in_io = 0;
 
        for (seg = 0; seg < nr_segs; seg++) {
                user_addr = (unsigned long)iov[seg].iov_base;
@@ -1184,7 +1168,7 @@ __blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode,
                }
        }
 
-       dio = kmalloc(sizeof(*dio), GFP_KERNEL);
+       dio = kzalloc(sizeof(*dio), GFP_KERNEL);
        retval = -ENOMEM;
        if (!dio)
                goto out;