Initialize the request queue lock earlier such that the following race can no longer occur:

blk_init_queue_node blkcg_print_blkgs blk_alloc_queue_node (1) q->queue_lock = &q->__queue_lock (2) blkcg_init_queue(q) (3) spin_lock_irq(blkg->q->queue_lock) (4) q->queue_lock = lock (5) spin_unlock_irq(blkg->q->queue_lock) (6)

(1) allocate an uninitialized queue; (2) initialize queue_lock to its default internal lock; (3) initialize blkcg part of request queue, which will create blkg and then insert it to blkg_list; (4) traverse blkg_list and find the created blkg, and then take its queue lock, here it is the default *internal lock*; (5) *race window*, now queue_lock is overridden with *driver specified lock*; (6) now unlock *driver specified lock*, not the locked *internal lock*, unlock balance breaks.

The changes in this patch are as follows: - Rename blk_alloc_queue_node() into blk_alloc_queue_node2() and add a new queue lock argument. - Introduce a wrapper function with the same name and behavior as the old blk_alloc_queue_node() function. - Move the .queue_lock initialization from blk_init_queue_node() into blk_alloc_queue_node2(). - For all all block drivers that initialize .queue_lock explicitly, change the blk_alloc_queue() call in the driver into a blk_alloc_queue_node2() call and remove the explicit .queue_lock initialization. Additionally, initialize the spin lock that will be used as queue lock earlier if necessary.

Reported-by: Joseph Qi joseph.qi@linux.alibaba.com Signed-off-by: Bart Van Assche bart.vanassche@wdc.com Cc: Christoph Hellwig hch@lst.de Cc: Joseph Qi joseph.qi@linux.alibaba.com Cc: Philipp Reisner philipp.reisner@linbit.com Cc: Ulf Hansson ulf.hansson@linaro.org Cc: Kees Cook keescook@chromium.org Cc: stable@vger.kernel.org --- block/blk-core.c | 33 ++++++++++++++++++++++++--------- drivers/block/drbd/drbd_main.c | 4 ++-- drivers/block/umem.c | 7 +++---- drivers/mmc/core/queue.c | 3 +-- include/linux/blkdev.h | 2 ++ 5 files changed, 32 insertions(+), 17 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c index bd43bc50740a..cb18f57e5b13 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -946,7 +946,22 @@ static void blk_rq_timed_out_timer(struct timer_list *t) kblockd_schedule_work(&q->timeout_work); }

-struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) +/** + * blk_alloc_queue_node2 - allocate a request queue + * @gfp_mask: memory allocation flags + * @node_id: NUMA node to allocate memory from + * @lock: Pointer to a spinlock that will be used to e.g. serialize calls to + * the legacy .request_fn(). Only set this pointer for queues that use + * legacy mode and not for queues that use blk-mq. + * + * Note: use this function instead of calling blk_alloc_queue_node() and + * setting the queue lock pointer explicitly to avoid triggering a crash in + * the blkcg throttling code. That code namely makes sysfs attributes visible + * in user space before this function returns and the show method of these + * attributes uses the queue lock. + */ +struct request_queue *blk_alloc_queue_node2(gfp_t gfp_mask, int node_id, + spinlock_t *lock) { struct request_queue *q;

@@ -997,11 +1012,7 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) mutex_init(&q->sysfs_lock); spin_lock_init(&q->__queue_lock);

- /* - * By default initialize queue_lock to internal lock and driver can - * override it later if need be. - */ - q->queue_lock = &q->__queue_lock; + q->queue_lock = lock ? : &q->__queue_lock;

/* * A queue starts its life with bypass turned on to avoid @@ -1042,6 +1053,12 @@ struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) kmem_cache_free(blk_requestq_cachep, q); return NULL; } +EXPORT_SYMBOL(blk_alloc_queue_node2); + +struct request_queue *blk_alloc_queue_node(gfp_t gfp_mask, int node_id) +{ + return blk_alloc_queue_node2(gfp_mask, node_id, NULL); +} EXPORT_SYMBOL(blk_alloc_queue_node);

/** @@ -1088,13 +1105,11 @@ blk_init_queue_node(request_fn_proc *rfn, spinlock_t *lock, int node_id) { struct request_queue *q;

- q = blk_alloc_queue_node(GFP_KERNEL, node_id); + q = blk_alloc_queue_node2(GFP_KERNEL, node_id, lock); if (!q) return NULL;

q->request_fn = rfn; - if (lock) - q->queue_lock = lock; if (blk_init_allocated_queue(q) < 0) { blk_cleanup_queue(q); return NULL; diff --git a/drivers/block/drbd/drbd_main.c b/drivers/block/drbd/drbd_main.c index 4b4697a1f963..965e80b13443 100644 --- a/drivers/block/drbd/drbd_main.c +++ b/drivers/block/drbd/drbd_main.c @@ -2822,7 +2822,8 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig

drbd_init_set_defaults(device);

- q = blk_alloc_queue(GFP_KERNEL); + q = blk_alloc_queue_node2(GFP_KERNEL, NUMA_NO_NODE, + &resource->req_lock); if (!q) goto out_no_q; device->rq_queue = q; @@ -2854,7 +2855,6 @@ enum drbd_ret_code drbd_create_device(struct drbd_config_context *adm_ctx, unsig /* Setting the max_hw_sectors to an odd value of 8kibyte here This triggers a max_bio_size message upon first attach or connect */ blk_queue_max_hw_sectors(q, DRBD_MAX_BIO_SIZE_SAFE >> 8); - q->queue_lock = &resource->req_lock;

device->md_io.page = alloc_page(GFP_KERNEL); if (!device->md_io.page) diff --git a/drivers/block/umem.c b/drivers/block/umem.c index 8077123678ad..f6bb78782afa 100644 --- a/drivers/block/umem.c +++ b/drivers/block/umem.c @@ -888,13 +888,14 @@ static int mm_pci_probe(struct pci_dev *dev, const struct pci_device_id *id) card->Active = -1; /* no page is active */ card->bio = NULL; card->biotail = &card->bio; + spin_lock_init(&card->lock);

- card->queue = blk_alloc_queue(GFP_KERNEL); + card->queue = blk_alloc_queue_node2(GFP_KERNEL, NUMA_NO_NODE, + &card->lock); if (!card->queue) goto failed_alloc;

blk_queue_make_request(card->queue, mm_make_request); - card->queue->queue_lock = &card->lock; card->queue->queuedata = card;

tasklet_init(&card->tasklet, process_page, (unsigned long)card); @@ -968,8 +969,6 @@ static int mm_pci_probe(struct pci_dev *dev, const struct pci_device_id *id) dev_printk(KERN_INFO, &card->dev->dev, "Window size %d bytes, IRQ %d\n", data, dev->irq);

- spin_lock_init(&card->lock); - pci_set_drvdata(dev, card);

if (pci_write_cmd != 0x0F) /* If not Memory Write & Invalidate */ diff --git a/drivers/mmc/core/queue.c b/drivers/mmc/core/queue.c index 5ecd54088988..274a8c6c8d64 100644 --- a/drivers/mmc/core/queue.c +++ b/drivers/mmc/core/queue.c @@ -216,10 +216,9 @@ int mmc_init_queue(struct mmc_queue *mq, struct mmc_card *card, int ret = -ENOMEM;

mq->card = card; - mq->queue = blk_alloc_queue(GFP_KERNEL); + mq->queue = blk_alloc_queue_node2(GFP_KERNEL, NUMA_NO_NODE, lock); if (!mq->queue) return -ENOMEM; - mq->queue->queue_lock = lock; mq->queue->request_fn = mmc_request_fn; mq->queue->init_rq_fn = mmc_init_request; mq->queue->exit_rq_fn = mmc_exit_request; diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h index 781992c4124e..37723a8b799c 100644 --- a/include/linux/blkdev.h +++ b/include/linux/blkdev.h @@ -1331,6 +1331,8 @@ extern long nr_blockdev_pages(void); bool __must_check blk_get_queue(struct request_queue *); struct request_queue *blk_alloc_queue(gfp_t); struct request_queue *blk_alloc_queue_node(gfp_t, int); +struct request_queue *blk_alloc_queue_node2(gfp_t gfp_mask, int node_id, + spinlock_t *lock); extern void blk_put_queue(struct request_queue *); extern void blk_set_queue_dying(struct request_queue *);