block: force noio scope in blk_mq_freeze_queue

When block drivers or the core block code perform allocations with a
frozen queue, this could try to recurse into the block device to
reclaim memory and deadlock.  Thus all allocations done by a process
that froze a queue need to be done without __GFP_IO and __GFP_FS.
Instead of tying to track all of them down, force a noio scope as
part of freezing the queue.

Note that nvme is a bit of a mess here due to the non-owner freezes,
and they will be addressed separately.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20250131120352.1315351-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

14 files changed, 59 insertions, 39 deletions

diff --git a/drivers/block/aoe/aoedev.c b/drivers/block/aoe/aoedev.c
index 3523dd82d7a0..4db7f6ce8ade 100644
--- a/drivers/block/aoe/aoedev.c
+++ b/drivers/block/aoe/aoedev.c
@@ -226,10 +226,11 @@ aoedev_downdev(struct aoedev *d)
 	/* fast fail all pending I/O */
 	if (d->blkq) {
 		/* UP is cleared, freeze+quiesce to insure all are errored */
-		blk_mq_freeze_queue(d->blkq);
+		unsigned int memflags = blk_mq_freeze_queue(d->blkq);
+
 		blk_mq_quiesce_queue(d->blkq);
 		blk_mq_unquiesce_queue(d->blkq);
-		blk_mq_unfreeze_queue(d->blkq);
+		blk_mq_unfreeze_queue(d->blkq, memflags);
 	}
 
 	if (d->gd)
diff --git a/drivers/block/ataflop.c b/drivers/block/ataflop.c
index 110f9aca2667..a81ade622a01 100644
--- a/drivers/block/ataflop.c
+++ b/drivers/block/ataflop.c
@@ -746,6 +746,7 @@ static int do_format(int drive, int type, struct atari_format_descr *desc)
 	unsigned char	*p;
 	int sect, nsect;
 	unsigned long	flags;
+	unsigned int memflags;
 	int ret;
 
 	if (type) {
@@ -758,7 +759,7 @@ static int do_format(int drive, int type, struct atari_format_descr *desc)
 	}
 
 	q = unit[drive].disk[type]->queue;
-	blk_mq_freeze_queue(q);
+	memflags = blk_mq_freeze_queue(q);
 	blk_mq_quiesce_queue(q);
 
 	local_irq_save(flags);
@@ -817,7 +818,7 @@ static int do_format(int drive, int type, struct atari_format_descr *desc)
 	ret = FormatError ? -EIO : 0;
 out:
 	blk_mq_unquiesce_queue(q);
-	blk_mq_unfreeze_queue(q);
+	blk_mq_unfreeze_queue(q, memflags);
 	return ret;
 }
 
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index d1f1d6bef2e6..c05fe27a96b6 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -586,6 +586,7 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
 {
 	struct file *file = fget(arg);
 	struct file *old_file;
+	unsigned int memflags;
 	int error;
 	bool partscan;
 	bool is_loop;
@@ -623,14 +624,14 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
 
 	/* and ... switch */
 	disk_force_media_change(lo->lo_disk);
-	blk_mq_freeze_queue(lo->lo_queue);
+	memflags = blk_mq_freeze_queue(lo->lo_queue);
 	mapping_set_gfp_mask(old_file->f_mapping, lo->old_gfp_mask);
 	lo->lo_backing_file = file;
 	lo->old_gfp_mask = mapping_gfp_mask(file->f_mapping);
 	mapping_set_gfp_mask(file->f_mapping,
 			     lo->old_gfp_mask & ~(__GFP_IO|__GFP_FS));
 	loop_update_dio(lo);
-	blk_mq_unfreeze_queue(lo->lo_queue);
+	blk_mq_unfreeze_queue(lo->lo_queue, memflags);
 	partscan = lo->lo_flags & LO_FLAGS_PARTSCAN;
 	loop_global_unlock(lo, is_loop);
 
@@ -1255,6 +1256,7 @@ loop_set_status(struct loop_device *lo, const struct loop_info64 *info)
 	int err;
 	bool partscan = false;
 	bool size_changed = false;
+	unsigned int memflags;
 
 	err = mutex_lock_killable(&lo->lo_mutex);
 	if (err)
@@ -1272,7 +1274,7 @@ loop_set_status(struct loop_device *lo, const struct loop_info64 *info)
 	}
 
 	/* I/O needs to be drained before changing lo_offset or lo_sizelimit */
-	blk_mq_freeze_queue(lo->lo_queue);
+	memflags = blk_mq_freeze_queue(lo->lo_queue);
 
 	err = loop_set_status_from_info(lo, info);
 	if (err)
@@ -1294,7 +1296,7 @@ loop_set_status(struct loop_device *lo, const struct loop_info64 *info)
 	loop_update_dio(lo);
 
 out_unfreeze:
-	blk_mq_unfreeze_queue(lo->lo_queue);
+	blk_mq_unfreeze_queue(lo->lo_queue, memflags);
 	if (partscan)
 		clear_bit(GD_SUPPRESS_PART_SCAN, &lo->lo_disk->state);
 out_unlock:
@@ -1446,6 +1448,7 @@ static int loop_set_capacity(struct loop_device *lo)
 static int loop_set_dio(struct loop_device *lo, unsigned long arg)
 {
 	bool use_dio = !!arg;
+	unsigned int memflags;
 
 	if (lo->lo_state != Lo_bound)
 		return -ENXIO;
@@ -1459,18 +1462,19 @@ static int loop_set_dio(struct loop_device *lo, unsigned long arg)
 		vfs_fsync(lo->lo_backing_file, 0);
 	}
 
-	blk_mq_freeze_queue(lo->lo_queue);
+	memflags = blk_mq_freeze_queue(lo->lo_queue);
 	if (use_dio)
 		lo->lo_flags |= LO_FLAGS_DIRECT_IO;
 	else
 		lo->lo_flags &= ~LO_FLAGS_DIRECT_IO;
-	blk_mq_unfreeze_queue(lo->lo_queue);
+	blk_mq_unfreeze_queue(lo->lo_queue, memflags);
 	return 0;
 }
 
 static int loop_set_block_size(struct loop_device *lo, unsigned long arg)
 {
 	struct queue_limits lim;
+	unsigned int memflags;
 	int err = 0;
 
 	if (lo->lo_state != Lo_bound)
@@ -1485,10 +1489,10 @@ static int loop_set_block_size(struct loop_device *lo, unsigned long arg)
 	lim = queue_limits_start_update(lo->lo_queue);
 	loop_update_limits(lo, &lim, arg);
 
-	blk_mq_freeze_queue(lo->lo_queue);
+	memflags = blk_mq_freeze_queue(lo->lo_queue);
 	err = queue_limits_commit_update(lo->lo_queue, &lim);
 	loop_update_dio(lo);
-	blk_mq_unfreeze_queue(lo->lo_queue);
+	blk_mq_unfreeze_queue(lo->lo_queue, memflags);
 
 	return err;
 }
diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
index b63a0f29a54a..7bdc7eb808ea 100644
--- a/drivers/block/nbd.c
+++ b/drivers/block/nbd.c
@@ -1234,6 +1234,7 @@ static int nbd_add_socket(struct nbd_device *nbd, unsigned long arg,
 	struct socket *sock;
 	struct nbd_sock **socks;
 	struct nbd_sock *nsock;
+	unsigned int memflags;
 	int err;
 
 	/* Arg will be cast to int, check it to avoid overflow */
@@ -1247,7 +1248,7 @@ static int nbd_add_socket(struct nbd_device *nbd, unsigned long arg,
 	 * We need to make sure we don't get any errant requests while we're
 	 * reallocating the ->socks array.
 	 */
-	blk_mq_freeze_queue(nbd->disk->queue);
+	memflags = blk_mq_freeze_queue(nbd->disk->queue);
 
 	if (!netlink && !nbd->task_setup &&
 	    !test_bit(NBD_RT_BOUND, &config->runtime_flags))
@@ -1288,12 +1289,12 @@ static int nbd_add_socket(struct nbd_device *nbd, unsigned long arg,
 	INIT_WORK(&nsock->work, nbd_pending_cmd_work);
 	socks[config->num_connections++] = nsock;
 	atomic_inc(&config->live_connections);
-	blk_mq_unfreeze_queue(nbd->disk->queue);
+	blk_mq_unfreeze_queue(nbd->disk->queue, memflags);
 
 	return 0;
 
 put_socket:
-	blk_mq_unfreeze_queue(nbd->disk->queue);
+	blk_mq_unfreeze_queue(nbd->disk->queue, memflags);
 	sockfd_put(sock);
 	return err;
 }
diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index 5b393e4a1ddf..faafd7ff43d6 100644
--- a/drivers/block/rbd.c
+++ b/drivers/block/rbd.c
@@ -7281,9 +7281,10 @@ static ssize_t do_rbd_remove(const char *buf, size_t count)
 		 * Prevent new IO from being queued and wait for existing
 		 * IO to complete/fail.
 		 */
-		blk_mq_freeze_queue(rbd_dev->disk->queue);
+		unsigned int memflags = blk_mq_freeze_queue(rbd_dev->disk->queue);
+
 		blk_mark_disk_dead(rbd_dev->disk);
-		blk_mq_unfreeze_queue(rbd_dev->disk->queue);
+		blk_mq_unfreeze_queue(rbd_dev->disk->queue, memflags);
 	}
 
 	del_gendisk(rbd_dev->disk);
diff --git a/drivers/block/sunvdc.c b/drivers/block/sunvdc.c
index 88dcae6ec575..05c4aee7f262 100644
--- a/drivers/block/sunvdc.c
+++ b/drivers/block/sunvdc.c
@@ -1113,6 +1113,7 @@ static void vdc_requeue_inflight(struct vdc_port *port)
 static void vdc_queue_drain(struct vdc_port *port)
 {
 	struct request_queue *q = port->disk->queue;
+	unsigned int memflags;
 
 	/*
 	 * Mark the queue as draining, then freeze/quiesce to ensure
@@ -1121,12 +1122,12 @@ static void vdc_queue_drain(struct vdc_port *port)
 	port->drain = 1;
 	spin_unlock_irq(&port->vio.lock);
 
-	blk_mq_freeze_queue(q);
+	memflags = blk_mq_freeze_queue(q);
 	blk_mq_quiesce_queue(q);
 
 	spin_lock_irq(&port->vio.lock);
 	port->drain = 0;
-	blk_mq_unquiesce_queue(q);
+	blk_mq_unquiesce_queue(q, memflags);
 	blk_mq_unfreeze_queue(q);
 }
 
diff --git a/drivers/block/swim3.c b/drivers/block/swim3.c
index 9914153b365b..3aedcb5add61 100644
--- a/drivers/block/swim3.c
+++ b/drivers/block/swim3.c
@@ -840,6 +840,7 @@ static int grab_drive(struct floppy_state *fs, enum swim_state state,
 static void release_drive(struct floppy_state *fs)
 {
 	struct request_queue *q = disks[fs->index]->queue;
+	unsigned int memflags;
 	unsigned long flags;
 
 	swim3_dbg("%s", "-> release drive\n");
@@ -848,10 +849,10 @@ static void release_drive(struct floppy_state *fs)
 	fs->state = idle;
 	spin_unlock_irqrestore(&swim3_lock, flags);
 
-	blk_mq_freeze_queue(q);
+	memflags = blk_mq_freeze_queue(q);
 	blk_mq_quiesce_queue(q);
 	blk_mq_unquiesce_queue(q);
-	blk_mq_unfreeze_queue(q);
+	blk_mq_unfreeze_queue(q, memflags);
 }
 
 static int fd_eject(struct floppy_state *fs)
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index bbaa26b523b8..a4af39fc7ea2 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -1584,11 +1584,12 @@ static int virtblk_freeze(struct virtio_device *vdev)
 {
 	struct virtio_blk *vblk = vdev->priv;
 	struct request_queue *q = vblk->disk->queue;
+	unsigned int memflags;
 
 	/* Ensure no requests in virtqueues before deleting vqs. */
-	blk_mq_freeze_queue(q);
+	memflags = blk_mq_freeze_queue(q);
 	blk_mq_quiesce_queue_nowait(q);
-	blk_mq_unfreeze_queue(q);
+	blk_mq_unfreeze_queue(q, memflags);
 
 	/* Ensure we don't receive any more interrupts */
 	virtio_reset_device(vdev);
diff --git a/drivers/mtd/mtd_blkdevs.c b/drivers/mtd/mtd_blkdevs.c
index ee7e1d908986..847c11542f02 100644
--- a/drivers/mtd/mtd_blkdevs.c
+++ b/drivers/mtd/mtd_blkdevs.c
@@ -404,6 +404,7 @@ out_list_del:
 int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 {
 	unsigned long flags;
+	unsigned int memflags;
 
 	lockdep_assert_held(&mtd_table_mutex);
 
@@ -420,10 +421,10 @@ int del_mtd_blktrans_dev(struct mtd_blktrans_dev *old)
 	spin_unlock_irqrestore(&old->queue_lock, flags);
 
 	/* freeze+quiesce queue to ensure all requests are flushed */
-	blk_mq_freeze_queue(old->rq);
+	memflags = blk_mq_freeze_queue(old->rq);
 	blk_mq_quiesce_queue(old->rq);
 	blk_mq_unquiesce_queue(old->rq);
-	blk_mq_unfreeze_queue(old->rq);
+	blk_mq_unfreeze_queue(old->rq, memflags);
 
 	/* If the device is currently open, tell trans driver to close it,
 		then put mtd device, and don't touch it again */
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 76b615d4d5b9..40046770f1bf 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -2132,15 +2132,16 @@ static int nvme_update_ns_info_generic(struct nvme_ns *ns,
 		struct nvme_ns_info *info)
 {
 	struct queue_limits lim;
+	unsigned int memflags;
 	int ret;
 
 	lim = queue_limits_start_update(ns->disk->queue);
 	nvme_set_ctrl_limits(ns->ctrl, &lim);
 
-	blk_mq_freeze_queue(ns->disk->queue);
+	memflags = blk_mq_freeze_queue(ns->disk->queue);
 	ret = queue_limits_commit_update(ns->disk->queue, &lim);
 	set_disk_ro(ns->disk, nvme_ns_is_readonly(ns, info));
-	blk_mq_unfreeze_queue(ns->disk->queue);
+	blk_mq_unfreeze_queue(ns->disk->queue, memflags);
 
 	/* Hide the block-interface for these devices */
 	if (!ret)
@@ -2155,6 +2156,7 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 	struct nvme_id_ns_nvm *nvm = NULL;
 	struct nvme_zone_info zi = {};
 	struct nvme_id_ns *id;
+	unsigned int memflags;
 	sector_t capacity;
 	unsigned lbaf;
 	int ret;
@@ -2186,7 +2188,7 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 
 	lim = queue_limits_start_update(ns->disk->queue);
 
-	blk_mq_freeze_queue(ns->disk->queue);
+	memflags = blk_mq_freeze_queue(ns->disk->queue);
 	ns->head->lba_shift = id->lbaf[lbaf].ds;
 	ns->head->nuse = le64_to_cpu(id->nuse);
 	capacity = nvme_lba_to_sect(ns->head, le64_to_cpu(id->nsze));
@@ -2219,7 +2221,7 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 
 	ret = queue_limits_commit_update(ns->disk->queue, &lim);
 	if (ret) {
-		blk_mq_unfreeze_queue(ns->disk->queue);
+		blk_mq_unfreeze_queue(ns->disk->queue, memflags);
 		goto out;
 	}
 
@@ -2235,7 +2237,7 @@ static int nvme_update_ns_info_block(struct nvme_ns *ns,
 		ns->head->features |= NVME_NS_DEAC;
 	set_disk_ro(ns->disk, nvme_ns_is_readonly(ns, info));
 	set_bit(NVME_NS_READY, &ns->flags);
-	blk_mq_unfreeze_queue(ns->disk->queue);
+	blk_mq_unfreeze_queue(ns->disk->queue, memflags);
 
 	if (blk_queue_is_zoned(ns->queue)) {
 		ret = blk_revalidate_disk_zones(ns->disk);
@@ -2291,9 +2293,10 @@ static int nvme_update_ns_info(struct nvme_ns *ns, struct nvme_ns_info *info)
 	if (!ret && nvme_ns_head_multipath(ns->head)) {
 		struct queue_limits *ns_lim = &ns->disk->queue->limits;
 		struct queue_limits lim;
+		unsigned int memflags;
 
 		lim = queue_limits_start_update(ns->head->disk->queue);
-		blk_mq_freeze_queue(ns->head->disk->queue);
+		memflags = blk_mq_freeze_queue(ns->head->disk->queue);
 		/*
 		 * queue_limits mixes values that are the hardware limitations
 		 * for bio splitting with what is the device configuration.
@@ -2325,7 +2328,7 @@ static int nvme_update_ns_info(struct nvme_ns *ns, struct nvme_ns_info *info)
 		set_disk_ro(ns->head->disk, nvme_ns_is_readonly(ns, info));
 		nvme_mpath_revalidate_paths(ns);
 
-		blk_mq_unfreeze_queue(ns->head->disk->queue);
+		blk_mq_unfreeze_queue(ns->head->disk->queue, memflags);
 	}
 
 	return ret;
diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
index a85d190942bd..2a7635565083 100644
--- a/drivers/nvme/host/multipath.c
+++ b/drivers/nvme/host/multipath.c
@@ -60,7 +60,7 @@ void nvme_mpath_unfreeze(struct nvme_subsystem *subsys)
 	lockdep_assert_held(&subsys->lock);
 	list_for_each_entry(h, &subsys->nsheads, entry)
 		if (h->disk)
-			blk_mq_unfreeze_queue(h->disk->queue);
+			blk_mq_unfreeze_queue_nomemrestore(h->disk->queue);
 }
 
 void nvme_mpath_wait_freeze(struct nvme_subsystem *subsys)
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 4411426a7894..b86e259516a7 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -2723,6 +2723,7 @@ int
 scsi_device_quiesce(struct scsi_device *sdev)
 {
 	struct request_queue *q = sdev->request_queue;
+	unsigned int memflags;
 	int err;
 
 	/*
@@ -2737,7 +2738,7 @@ scsi_device_quiesce(struct scsi_device *sdev)
 
 	blk_set_pm_only(q);
 
-	blk_mq_freeze_queue(q);
+	memflags = blk_mq_freeze_queue(q);
 	/*
 	 * Ensure that the effect of blk_set_pm_only() will be visible
 	 * for percpu_ref_tryget() callers that occur after the queue
@@ -2745,7 +2746,7 @@ scsi_device_quiesce(struct scsi_device *sdev)
 	 * was called. See also https://lwn.net/Articles/573497/.
 	 */
 	synchronize_rcu();
-	blk_mq_unfreeze_queue(q);
+	blk_mq_unfreeze_queue(q, memflags);
 
 	mutex_lock(&sdev->state_mutex);
 	err = scsi_device_set_state(sdev, SDEV_QUIESCE);
diff --git a/drivers/scsi/scsi_scan.c b/drivers/scsi/scsi_scan.c
index 042329b74c6e..312d78213954 100644
--- a/drivers/scsi/scsi_scan.c
+++ b/drivers/scsi/scsi_scan.c
@@ -220,6 +220,7 @@ static int scsi_realloc_sdev_budget_map(struct scsi_device *sdev,
 	int new_shift = sbitmap_calculate_shift(depth);
 	bool need_alloc = !sdev->budget_map.map;
 	bool need_free = false;
+	unsigned int memflags;
 	int ret;
 	struct sbitmap sb_backup;
 
@@ -240,7 +241,7 @@ static int scsi_realloc_sdev_budget_map(struct scsi_device *sdev,
 	 * and here disk isn't added yet, so freezing is pretty fast
 	 */
 	if (need_free) {
-		blk_mq_freeze_queue(sdev->request_queue);
+		memflags = blk_mq_freeze_queue(sdev->request_queue);
 		sb_backup = sdev->budget_map;
 	}
 	ret = sbitmap_init_node(&sdev->budget_map,
@@ -256,7 +257,7 @@ static int scsi_realloc_sdev_budget_map(struct scsi_device *sdev,
 		else
 			sbitmap_free(&sb_backup);
 		ret = 0;
-		blk_mq_unfreeze_queue(sdev->request_queue);
+		blk_mq_unfreeze_queue(sdev->request_queue, memflags);
 	}
 	return ret;
 }
diff --git a/drivers/ufs/core/ufs-sysfs.c b/drivers/ufs/core/ufs-sysfs.c
index 796e37a1d859..3438269a5440 100644
--- a/drivers/ufs/core/ufs-sysfs.c
+++ b/drivers/ufs/core/ufs-sysfs.c
@@ -1439,6 +1439,7 @@ static ssize_t max_number_of_rtt_store(struct device *dev,
 	struct ufs_hba *hba = dev_get_drvdata(dev);
 	struct ufs_dev_info *dev_info = &hba->dev_info;
 	struct scsi_device *sdev;
+	unsigned int memflags;
 	unsigned int rtt;
 	int ret;
 
@@ -1458,14 +1459,16 @@ static ssize_t max_number_of_rtt_store(struct device *dev,
 
 	ufshcd_rpm_get_sync(hba);
 
+	memflags = memalloc_noio_save();
 	shost_for_each_device(sdev, hba->host)
-		blk_mq_freeze_queue(sdev->request_queue);
+		blk_mq_freeze_queue_nomemsave(sdev->request_queue);
 
 	ret = ufshcd_query_attr(hba, UPIU_QUERY_OPCODE_WRITE_ATTR,
 		QUERY_ATTR_IDN_MAX_NUM_OF_RTT, 0, 0, &rtt);
 
 	shost_for_each_device(sdev, hba->host)
-		blk_mq_unfreeze_queue(sdev->request_queue);
+		blk_mq_unfreeze_queue_nomemrestore(sdev->request_queue);
+	memalloc_noio_restore(memflags);
 
 	ufshcd_rpm_put_sync(hba);