[PATCH] zram: remove init_lock in zram_make_request

[Minchan Kim]

Admin could reset zram during I/O operation going on so we have
used zram->init_lock as read-side lock in I/O path to prevent
sudden zram meta freeing.

However, the init_lock is really troublesome.
We can't do call zram_meta_alloc under init_lock due to lockdep splat
because zram_rw_page is one of the function under reclaim path and
hold it as read_lock while other places in process context hold it
as write_lock. So, we have used allocation out of the lock to avoid
lockdep warn but it's not good for readability and finally, I met
another lockdep splat between init_lock and cpu_hotpulug from
kmem_cache_destroy during wokring zsmalloc compaction. :(

Yes, the ideal is to remove horrible init_lock of zram in rw path.
This patch removes it in rw path and instead, use kick_all_cpus_sync
and a bool init_done variable to check initialization done with
smp_[wmb|rmb].

Upon kick_all_cpus_sync returns, any CPU cannot access zram meta
any more due to init_done in zram_make_request so it's safe to
free meta. So, finally, we avoids init_lock in reclaim context
so we are free for deadlock.

Signed-off-by: Minchan Kim <minchan@xxxxxxxxxx>
---
 drivers/block/zram/zram_drv.c | 70 +++++++++++++++++++++++++------------------
 drivers/block/zram/zram_drv.h |  2 ++
 2 files changed, 43 insertions(+), 29 deletions(-)

diff --git a/drivers/block/zram/zram_drv.c b/drivers/block/zram/zram_drv.c
index a598ada817f0..404602b1932e 100644
--- a/drivers/block/zram/zram_drv.c
+++ b/drivers/block/zram/zram_drv.c
@@ -53,9 +53,16 @@ static ssize_t name##_show(struct device *d,		\
 }									\
 static DEVICE_ATTR_RO(name);
 
-static inline int init_done(struct zram *zram)
+static inline bool init_done(struct zram *zram)
 {
-	return zram->meta != NULL;
+	/*
+	 * init_done can be used without holding zram->init_lock in
+	 * read/write handler(ie, zram_make_request) but we should make sure
+	 * that zram->init_done should set up after meta initialization is
+	 * done. Look at disksize_store.
+	 */
+	smp_rmb();
+	return zram->init_done;
 }
 
 static inline struct zram *dev_to_zram(struct device *dev)
@@ -726,11 +733,8 @@ static void zram_reset_device(struct zram *zram, bool reset_capacity)
 		return;
 	}
 
-	zcomp_destroy(zram->comp);
 	zram->max_comp_streams = 1;
 
-	zram_meta_free(zram->meta);
-	zram->meta = NULL;
 	/* Reset stats */
 	memset(&zram->stats, 0, sizeof(zram->stats));
 
@@ -738,8 +742,16 @@ static void zram_reset_device(struct zram *zram, bool reset_capacity)
 	if (reset_capacity)
 		set_capacity(zram->disk, 0);
 
+	zram->init_done = false;
+	/* don't need smp_wmb because kick_all_cpus_sync does */
+	kick_all_cpus_sync();
+	/*
+	 * From now on, any read/write cannot access zram meta data
+	 * by init_done in the handler.
+	 */
+	zram_meta_free(zram->meta);
+	zcomp_destroy(zram->comp);
 	up_write(&zram->init_lock);
-
 	/*
 	 * Revalidate disk out of the init_lock to avoid lockdep splat.
 	 * It's okay because disk's capacity is protected by init_lock
@@ -762,10 +774,19 @@ static ssize_t disksize_store(struct device *dev,
 	if (!disksize)
 		return -EINVAL;
 
+	down_write(&zram->init_lock);
+	if (init_done(zram)) {
+		pr_info("Cannot change disksize for initialized device\n");
+		up_write(&zram->init_lock);
+		return -EBUSY;
+	}
+
 	disksize = PAGE_ALIGN(disksize);
 	meta = zram_meta_alloc(zram->disk->first_minor, disksize);
-	if (!meta)
+	if (!meta) {
+		up_write(&zram->init_lock);
 		return -ENOMEM;
+	}
 
 	comp = zcomp_create(zram->compressor, zram->max_comp_streams);
 	if (IS_ERR(comp)) {
@@ -775,17 +796,17 @@ static ssize_t disksize_store(struct device *dev,
 		goto out_free_meta;
 	}
 
-	down_write(&zram->init_lock);
-	if (init_done(zram)) {
-		pr_info("Cannot change disksize for initialized device\n");
-		err = -EBUSY;
-		goto out_destroy_comp;
-	}
-
 	zram->meta = meta;
 	zram->comp = comp;
 	zram->disksize = disksize;
 	set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT);
+	/*
+	 * Store operation of struct zram fields should complete
+	 * before init_done set up because zram_bvec_rw doesn't
+	 * hold an zram->init_lock.
+	 */
+	smp_wmb();
+	zram->init_done = true;
 	up_write(&zram->init_lock);
 
 	/*
@@ -797,10 +818,8 @@ static ssize_t disksize_store(struct device *dev,
 
 	return len;
 
-out_destroy_comp:
-	up_write(&zram->init_lock);
-	zcomp_destroy(comp);
 out_free_meta:
+	up_write(&zram->init_lock);
 	zram_meta_free(meta);
 	return err;
 }
@@ -907,7 +926,6 @@ static void zram_make_request(struct request_queue *queue, struct bio *bio)
 {
 	struct zram *zram = queue->queuedata;
 
-	down_read(&zram->init_lock);
 	if (unlikely(!init_done(zram)))
 		goto error;
 
@@ -918,12 +936,10 @@ static void zram_make_request(struct request_queue *queue, struct bio *bio)
 	}
 
 	__zram_make_request(zram, bio);
-	up_read(&zram->init_lock);
 
 	return;
 
 error:
-	up_read(&zram->init_lock);
 	bio_io_error(bio);
 }
 
@@ -951,17 +967,16 @@ static int zram_rw_page(struct block_device *bdev, sector_t sector,
 	struct bio_vec bv;
 
 	zram = bdev->bd_disk->private_data;
+
+	/* This should be another patch */
+	if (unlikely(!init_done(zram)))
+		return -EIO;
+
 	if (!valid_io_request(zram, sector, PAGE_SIZE)) {
 		atomic64_inc(&zram->stats.invalid_io);
 		return -EINVAL;
 	}
 
-	down_read(&zram->init_lock);
-	if (unlikely(!init_done(zram))) {
-		err = -EIO;
-		goto out_unlock;
-	}
-
 	index = sector >> SECTORS_PER_PAGE_SHIFT;
 	offset = sector & (SECTORS_PER_PAGE - 1) << SECTOR_SHIFT;
 
@@ -970,8 +985,6 @@ static int zram_rw_page(struct block_device *bdev, sector_t sector,
 	bv.bv_offset = 0;
 
 	err = zram_bvec_rw(zram, &bv, index, offset, rw);
-out_unlock:
-	up_read(&zram->init_lock);
 	/*
 	 * If I/O fails, just return error(ie, non-zero) without
 	 * calling page_endio.
@@ -1125,7 +1138,6 @@ static void destroy_device(struct zram *zram)
 
 	del_gendisk(zram->disk);
 	put_disk(zram->disk);
-
 	blk_cleanup_queue(zram->queue);
 }
 
diff --git a/drivers/block/zram/zram_drv.h b/drivers/block/zram/zram_drv.h
index e492f6bf11f1..dca265654285 100644
--- a/drivers/block/zram/zram_drv.h
+++ b/drivers/block/zram/zram_drv.h
@@ -107,6 +107,8 @@ struct zram {
 
 	/* Prevent concurrent execution of device init, reset and R/W request */
 	struct rw_semaphore init_lock;
+	bool init_done;
+
 	/*
 	 * This is the limit on amount of *uncompressed* worth of data
 	 * we can store in a disk.
-- 
1.9.1


-- 
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/