md/lock: ensure updates to page_attrs are properly locked.

Page attributes are set using __set_bit rather than set_bit as
it normally called under a spinlock so the extra atomicity is not
needed.

However there are two places where we might set or clear page
attributes without holding the spinlock.
So add the spinlock in those cases.

This might be the cause of occasional reports that bits a aren't
getting clear properly - theory is that BITMAP_PAGE_PENDING gets lost
when BITMAP_PAGE_NEEDWRITE is set or cleared.  This is an
inconvenience, not a threat to data safety.

Signed-off-by: NeilBrown <neilb@suse.de>
This commit is contained in:
NeilBrown 2011-11-23 10:18:52 +11:00
parent 257a4b42af
commit 7c8f424798

View File

@ -1106,10 +1106,12 @@ void bitmap_write_all(struct bitmap *bitmap)
*/
int i;
spin_lock_irq(&bitmap->lock);
for (i = 0; i < bitmap->file_pages; i++)
set_page_attr(bitmap, bitmap->filemap[i],
BITMAP_PAGE_NEEDWRITE);
bitmap->allclean = 0;
spin_unlock_irq(&bitmap->lock);
}
static void bitmap_count_page(struct bitmap *bitmap, sector_t offset, int inc)
@ -1605,7 +1607,9 @@ void bitmap_dirty_bits(struct bitmap *bitmap, unsigned long s, unsigned long e)
for (chunk = s; chunk <= e; chunk++) {
sector_t sec = (sector_t)chunk << CHUNK_BLOCK_SHIFT(bitmap);
bitmap_set_memory_bits(bitmap, sec, 1);
spin_lock_irq(&bitmap->lock);
bitmap_file_set_bit(bitmap, sec);
spin_unlock_irq(&bitmap->lock);
if (sec < bitmap->mddev->recovery_cp)
/* We are asserting that the array is dirty,
* so move the recovery_cp address back so