On 11/26/19 2:32 PM, Jaegeuk Kim wrote:
+static void
+loop_mq_freeze_queue(struct loop_device *lo)
+{
+ for (;;) {
+ sync_blockdev(lo->lo_device);
+
+ /* I/O need to be drained during transfer transition */
+ blk_mq_freeze_queue(lo->lo_queue);
+
+ if (is_dirty_bdev(lo->lo_device))
+ blk_mq_unfreeze_queue(lo->lo_queue);
+ else
+ break;
+ }
+ kill_bdev(lo->lo_device);
+}
Maybe I overlooked something, but how does the above code prevent that
pages get dirtied after is_dirty_bdev() returns false?
blk_mq_freeze_queue() freezes lo->lo_queue but not the filesystem that
uses that request queue. Did you perhaps want to call a function that
freezes filesystem activity from inside the loop (not sure if such a
function already exists)?
[ ... ]
/* I/O need to be drained during transfer transition */
- blk_mq_freeze_queue(lo->lo_queue);
+ if (drop_cache)
+ loop_mq_freeze_queue(lo);
+ else
+ blk_mq_freeze_queue(lo->lo_queue);
How about always calling loop_mq_freeze_queue(), independent of the
value of 'drop_cache'? I think that will make the code both easier to
read and easier to test.
+ /* I/O need to be drained during transfer transition */
+ if (drop_cache)
+ loop_mq_freeze_queue(lo);
+ else
+ blk_mq_freeze_queue(lo->lo_queue);
Same comment here.
diff --git a/fs/block_dev.c b/fs/block_dev.c
index 9c073dbdc1b0..81fe3beef92b 100644
--- a/fs/block_dev.c
+++ b/fs/block_dev.c
@@ -74,6 +74,18 @@ static void bdev_write_inode(struct block_device *bdev)
spin_unlock(&inode->i_lock);
}
+bool is_dirty_bdev(struct block_device *bdev)
+{
+ struct inode *inode = bdev->bd_inode;
+ bool dirty;
+
+ spin_lock(&inode->i_lock);
+ dirty = inode->i_state & I_DIRTY ? true: false;
+ spin_unlock(&inode->i_lock);
+ return dirty;
+}
Since the Linux kernel 'bool' datatype conforms to C99, I think that the
" ? true : false" part can be left out. From include/linux/types.h:
typedef _Bool bool;
From http://www.open-std.org/jtc1/sc22/WG14/www/docs/n1256.pdf: "When
any scalar value is converted to _Bool, the result is 0 if the value
compares equal to 0; otherwise, the result is 1."
Thanks,
Bart.
