On Mon, Feb 02, 2015 at 02:09:12PM +0900, Sergey Senozhatsky wrote: > > the patch mosly looks good, except for one place: > > On (02/02/15 13:28), Minchan Kim wrote: > > @@ -783,6 +812,8 @@ static ssize_t disksize_store(struct device *dev, > > goto out_destroy_comp; > > } > > > > + init_waitqueue_head(&zram->io_done); > > + zram_meta_get(zram); > > it was > + init_completion(&zram->io_done); > + atomic_set(&zram->refcount, 1); > > I think we need to replace zram_meta_get(zram) with atomic_set(&zram->refcount, 1). > > ->refcount is 0 by default and atomic_inc_not_zero(&zram->refcount) will not > increment it here, nor anywhere else. > > > > zram->meta = meta; > > zram->comp = comp; > > zram->disksize = disksize; > > @@ -838,8 +869,8 @@ static ssize_t reset_store(struct device *dev, > > /* Make sure all pending I/O is finished */ > > fsync_bdev(bdev); > > bdput(bdev); > > - > > [..] > > > @@ -1041,6 +1075,7 @@ static int create_device(struct zram *zram, int device_id) > > int ret = -ENOMEM; > > > > init_rwsem(&zram->init_lock); > > + atomic_set(&zram->refcount, 0); > > sorry, I forgot that zram is kzalloc()-ated. so we can drop > > atomic_set(&zram->refcount, 0) > Everything are fixed. Ready to send a patch. But before sending, hope we fix umount race issue first. Thanks a lot, Sergey! -- Kind regards, Minchan Kim -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>