Merge branch 'for-3.17/core' of git://git.kernel.dk/linux-block

Pull block core bits from Jens Axboe:
"Small round this time, after the massive blk-mq dump for 3.16. This
pull request contains:

- Fixes for max_sectors overflow in ioctls from Akinoby Mita.

- Partition off-by-one bug fix in aix partitions from Dan Carpenter.

- Various small partition cleanups from Fabian Frederick.

- Fix for the block integrity code sometimes returning the wrong
vector count from Gu Zheng.

- Cleanup an re-org of the blk-mq queue enter/exit percpu counters
from Tejun. Dependent on the percpu pull for 3.17 (which was in
the block tree too), that you have already pulled in.

- A blkcg oops fix, also from Tejun"

* 'for-3.17/core' of git://git.kernel.dk/linux-block:
partitions: aix.c: off by one bug
blkcg: don't call into policy draining if root_blkg is already gone
Revert "bio: modify __bio_add_page() to accept pages that don't start a new segment"
bio: modify __bio_add_page() to accept pages that don't start a new segment
block: fix SG_[GS]ET_RESERVED_SIZE ioctl when max_sectors is huge
block: fix BLKSECTGET ioctl when max_sectors is greater than USHRT_MAX
block/partitions/efi.c: kerneldoc fixing
block/partitions/msdos.c: code clean-up
block/partitions/amiga.c: replace nolevel printk by pr_err
block/partitions/aix.c: replace count*size kzalloc by kcalloc
bio-integrity: add "bip_max_vcnt" into struct bio_integrity_payload
blk-mq: use percpu_ref for mq usage count
blk-mq: collapse __blk_mq_drain_queue() into blk_mq_freeze_queue()
blk-mq: decouble blk-mq freezing from generic bypassing
block, blk-mq: draining can't be skipped even if bypass_depth was non-zero
blk-mq: fix a memory ordering bug in blk_mq_queue_enter()

block/bio-integrity.c

···
 					  bs->bvec_integrity_pool);
 		if (!bip->bip_vec)
 			goto err;
+		bip->bip_max_vcnt = bvec_nr_vecs(idx);
 	} else {
 		bip->bip_vec = bip->bip_inline_vecs;
+		bip->bip_max_vcnt = inline_vecs;
 	}
 
 	bip->bip_slab = idx;
···
 }
 EXPORT_SYMBOL(bio_integrity_free);
 
-static inline unsigned int bip_integrity_vecs(struct bio_integrity_payload *bip)
-{
-	if (bip->bip_slab == BIO_POOL_NONE)
-		return BIP_INLINE_VECS;
-
-	return bvec_nr_vecs(bip->bip_slab);
-}
-
 /**
  * bio_integrity_add_page - Attach integrity metadata
  * @bio:	bio to update
···
 	struct bio_integrity_payload *bip = bio->bi_integrity;
 	struct bio_vec *iv;
 
-	if (bip->bip_vcnt >= bip_integrity_vecs(bip)) {
+	if (bip->bip_vcnt >= bip->bip_max_vcnt) {
 		printk(KERN_ERR "%s: bip_vec full\n", __func__);
 		return 0;
 	}

block/blk-core.c

···
  */
 void blk_queue_bypass_start(struct request_queue *q)
 {
-	bool drain;
-
 	spin_lock_irq(q->queue_lock);
-	drain = !q->bypass_depth++;
+	q->bypass_depth++;
 	queue_flag_set(QUEUE_FLAG_BYPASS, q);
 	spin_unlock_irq(q->queue_lock);
 
-	if (drain) {
+	/*
+	 * Queues start drained.  Skip actual draining till init is
+	 * complete.  This avoids lenghty delays during queue init which
+	 * can happen many times during boot.
+	 */
+	if (blk_queue_init_done(q)) {
 		spin_lock_irq(q->queue_lock);
 		__blk_drain_queue(q, false);
 		spin_unlock_irq(q->queue_lock);
···
 	 * prevent that q->request_fn() gets invoked after draining finished.
 	 */
 	if (q->mq_ops) {
-		blk_mq_drain_queue(q);
+		blk_mq_freeze_queue(q);
 		spin_lock_irq(lock);
 	} else {
 		spin_lock_irq(lock);

block/blk-mq.c

···
 
 static int blk_mq_queue_enter(struct request_queue *q)
 {
-	int ret;
+	while (true) {
+		int ret;
 
-	__percpu_counter_add(&q->mq_usage_counter, 1, 1000000);
-	smp_wmb();
+		if (percpu_ref_tryget_live(&q->mq_usage_counter))
+			return 0;
 
-	/* we have problems freezing the queue if it's initializing */
-	if (!blk_queue_dying(q) &&
-	    (!blk_queue_bypass(q) || !blk_queue_init_done(q)))
-		return 0;
-
-	__percpu_counter_add(&q->mq_usage_counter, -1, 1000000);
-
-	spin_lock_irq(q->queue_lock);
-	ret = wait_event_interruptible_lock_irq(q->mq_freeze_wq,
-		!blk_queue_bypass(q) || blk_queue_dying(q),
-		*q->queue_lock);
-	/* inc usage with lock hold to avoid freeze_queue runs here */
-	if (!ret && !blk_queue_dying(q))
-		__percpu_counter_add(&q->mq_usage_counter, 1, 1000000);
-	else if (blk_queue_dying(q))
-		ret = -ENODEV;
-	spin_unlock_irq(q->queue_lock);
-
-	return ret;
+		ret = wait_event_interruptible(q->mq_freeze_wq,
+				!q->mq_freeze_depth || blk_queue_dying(q));
+		if (blk_queue_dying(q))
+			return -ENODEV;
+		if (ret)
+			return ret;
+	}
 }
 
 static void blk_mq_queue_exit(struct request_queue *q)
 {
-	__percpu_counter_add(&q->mq_usage_counter, -1, 1000000);
+	percpu_ref_put(&q->mq_usage_counter);
 }
 
-void blk_mq_drain_queue(struct request_queue *q)
+static void blk_mq_usage_counter_release(struct percpu_ref *ref)
 {
-	while (true) {
-		s64 count;
+	struct request_queue *q =
+		container_of(ref, struct request_queue, mq_usage_counter);
 
-		spin_lock_irq(q->queue_lock);
-		count = percpu_counter_sum(&q->mq_usage_counter);
-		spin_unlock_irq(q->queue_lock);
-
-		if (count == 0)
-			break;
-		blk_mq_start_hw_queues(q);
-		msleep(10);
-	}
+	wake_up_all(&q->mq_freeze_wq);
 }
 
 /*
  * Guarantee no request is in use, so we can change any data structure of
  * the queue afterward.
  */
-static void blk_mq_freeze_queue(struct request_queue *q)
+void blk_mq_freeze_queue(struct request_queue *q)
 {
-	bool drain;
-
 	spin_lock_irq(q->queue_lock);
-	drain = !q->bypass_depth++;
-	queue_flag_set(QUEUE_FLAG_BYPASS, q);
+	q->mq_freeze_depth++;
 	spin_unlock_irq(q->queue_lock);
 
-	if (drain)
-		blk_mq_drain_queue(q);
+	percpu_ref_kill(&q->mq_usage_counter);
+	blk_mq_run_queues(q, false);
+	wait_event(q->mq_freeze_wq, percpu_ref_is_zero(&q->mq_usage_counter));
 }
 
 static void blk_mq_unfreeze_queue(struct request_queue *q)
···
 	bool wake = false;
 
 	spin_lock_irq(q->queue_lock);
-	if (!--q->bypass_depth) {
-		queue_flag_clear(QUEUE_FLAG_BYPASS, q);
-		wake = true;
-	}
-	WARN_ON_ONCE(q->bypass_depth < 0);
+	wake = !--q->mq_freeze_depth;
+	WARN_ON_ONCE(q->mq_freeze_depth < 0);
 	spin_unlock_irq(q->queue_lock);
-	if (wake)
+	if (wake) {
+		percpu_ref_reinit(&q->mq_usage_counter);
 		wake_up_all(&q->mq_freeze_wq);
+	}
 }
 
 bool blk_mq_can_queue(struct blk_mq_hw_ctx *hctx)
···
 	if (!q)
 		goto err_hctxs;
 
-	if (percpu_counter_init(&q->mq_usage_counter, 0))
+	if (percpu_ref_init(&q->mq_usage_counter, blk_mq_usage_counter_release))
 		goto err_map;
 
 	setup_timer(&q->timeout, blk_mq_rq_timer, (unsigned long) q);
···
 	blk_mq_exit_hw_queues(q, set, set->nr_hw_queues);
 	blk_mq_free_hw_queues(q, set);
 
-	percpu_counter_destroy(&q->mq_usage_counter);
+	percpu_ref_exit(&q->mq_usage_counter);
 
 	free_percpu(q->queue_ctx);
 	kfree(q->queue_hw_ctx);
···
 {
 	blk_mq_cpu_init();
 
-	/* Must be called after percpu_counter_hotcpu_callback() */
-	hotcpu_notifier(blk_mq_queue_reinit_notify, -10);
+	hotcpu_notifier(blk_mq_queue_reinit_notify, 0);
 
 	return 0;
 }

block/blk-mq.h

···
 void __blk_mq_complete_request(struct request *rq);
 void blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async);
 void blk_mq_init_flush(struct request_queue *q);
-void blk_mq_drain_queue(struct request_queue *q);
+void blk_mq_freeze_queue(struct request_queue *q);
 void blk_mq_free_queue(struct request_queue *q);
 void blk_mq_clone_flush_request(struct request *flush_rq,
 		struct request *orig_rq);

block/blk-sysfs.c

···
 	 * Initialization must be complete by now.  Finish the initial
 	 * bypass from queue allocation.
 	 */
-	blk_queue_bypass_end(q);
 	queue_flag_set_unlocked(QUEUE_FLAG_INIT_DONE, q);
+	blk_queue_bypass_end(q);
 
 	ret = blk_trace_init_sysfs(dev);
 	if (ret)

block/compat_ioctl.c

···
 	fmode_t mode = file->f_mode;
 	struct backing_dev_info *bdi;
 	loff_t size;
+	unsigned int max_sectors;
 
 	/*
 	 * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have
···
 	case BLKSSZGET: /* get block device hardware sector size */
 		return compat_put_int(arg, bdev_logical_block_size(bdev));
 	case BLKSECTGET:
-		return compat_put_ushort(arg,
-					 queue_max_sectors(bdev_get_queue(bdev)));
+		max_sectors = min_t(unsigned int, USHRT_MAX,
+				    queue_max_sectors(bdev_get_queue(bdev)));
+		return compat_put_ushort(arg, max_sectors);
 	case BLKROTATIONAL:
 		return compat_put_ushort(arg,
 					 !blk_queue_nonrot(bdev_get_queue(bdev)));

block/ioctl.c

···
 	struct backing_dev_info *bdi;
 	loff_t size;
 	int ret, n;
+	unsigned int max_sectors;
 
 	switch(cmd) {
 	case BLKFLSBUF:
···
 	case BLKDISCARDZEROES:
 		return put_uint(arg, bdev_discard_zeroes_data(bdev));
 	case BLKSECTGET:
-		return put_ushort(arg, queue_max_sectors(bdev_get_queue(bdev)));
+		max_sectors = min_t(unsigned int, USHRT_MAX,
+				    queue_max_sectors(bdev_get_queue(bdev)));
+		return put_ushort(arg, max_sectors);
 	case BLKROTATIONAL:
 		return put_ushort(arg, !blk_queue_nonrot(bdev_get_queue(bdev)));
 	case BLKRASET:

block/partitions/aix.c

···
 		numlvs = be16_to_cpu(p->numlvs);
 		put_dev_sector(sect);
 	}
-	lvip = kzalloc(sizeof(struct lv_info) * state->limit, GFP_KERNEL);
+	lvip = kcalloc(state->limit, sizeof(struct lv_info), GFP_KERNEL);
 	if (!lvip)
 		return 0;
 	if (numlvs && (d = read_part_sector(state, vgda_sector + 1, &sect))) {
···
 				continue;
 			}
 			lv_ix = be16_to_cpu(p->lv_ix) - 1;
-			if (lv_ix > state->limit) {
+			if (lv_ix >= state->limit) {
 				cur_lv_ix = -1;
 				continue;
 			}

block/partitions/amiga.c

···
  *  Re-organised Feb 1998 Russell King
  */
 
+#define pr_fmt(fmt) fmt
+
 #include <linux/types.h>
 #include <linux/affs_hardblocks.h>
 
···
 		data = read_part_sector(state, blk, &sect);
 		if (!data) {
 			if (warn_no_part)
-				printk("Dev %s: unable to read RDB block %d\n",
+				pr_err("Dev %s: unable to read RDB block %d\n",
 				       bdevname(state->bdev, b), blk);
 			res = -1;
 			goto rdb_done;
···
 		*(__be32 *)(data+0xdc) = 0;
 		if (checksum_block((__be32 *)data,
 				be32_to_cpu(rdb->rdb_SummedLongs) & 0x7F)==0) {
-			printk("Warning: Trashed word at 0xd0 in block %d "
-				"ignored in checksum calculation\n",blk);
+			pr_err("Trashed word at 0xd0 in block %d ignored in checksum calculation\n",
+			       blk);
 			break;
 		}
 
-		printk("Dev %s: RDB in block %d has bad checksum\n",
+		pr_err("Dev %s: RDB in block %d has bad checksum\n",
 		       bdevname(state->bdev, b), blk);
 	}
 
···
 		data = read_part_sector(state, blk, &sect);
 		if (!data) {
 			if (warn_no_part)
-				printk("Dev %s: unable to read partition block %d\n",
+				pr_err("Dev %s: unable to read partition block %d\n",
 				       bdevname(state->bdev, b), blk);
 			res = -1;
 			goto rdb_done;

block/partitions/efi.c

···
 /**
  * efi_crc32() - EFI version of crc32 function
  * @buf: buffer to calculate crc32 of
- * @len - length of buf
+ * @len: length of buf
  *
  * Description: Returns EFI-style CRC32 value for @buf
  * 
···
 
 /**
  * read_lba(): Read bytes from disk, starting at given LBA
- * @state
- * @lba
- * @buffer
- * @size_t
+ * @state: disk parsed partitions
+ * @lba: the Logical Block Address of the partition table
+ * @buffer: destination buffer
+ * @count: bytes to read
  *
  * Description: Reads @count bytes from @state->bdev into @buffer.
  * Returns number of bytes read on success, 0 on error.
···
 
 /**
  * alloc_read_gpt_entries(): reads partition entries from disk
- * @state
- * @gpt - GPT header
+ * @state: disk parsed partitions
+ * @gpt: GPT header
  * 
  * Description: Returns ptes on success,  NULL on error.
  * Allocates space for PTEs based on information found in @gpt.
···
 
 /**
  * alloc_read_gpt_header(): Allocates GPT header, reads into it from disk
- * @state
- * @lba is the Logical Block Address of the partition table
+ * @state: disk parsed partitions
+ * @lba: the Logical Block Address of the partition table
  * 
  * Description: returns GPT header on success, NULL on error.   Allocates
  * and fills a GPT header starting at @ from @state->bdev.
···
 
 /**
  * is_gpt_valid() - tests one GPT header and PTEs for validity
- * @state
- * @lba is the logical block address of the GPT header to test
- * @gpt is a GPT header ptr, filled on return.
- * @ptes is a PTEs ptr, filled on return.
+ * @state: disk parsed partitions
+ * @lba: logical block address of the GPT header to test
+ * @gpt: GPT header ptr, filled on return.
+ * @ptes: PTEs ptr, filled on return.
  *
  * Description: returns 1 if valid,  0 on error.
  * If valid, returns pointers to newly allocated GPT header and PTEs.
···
 
 /**
  * is_pte_valid() - tests one PTE for validity
- * @pte is the pte to check
- * @lastlba is last lba of the disk
+ * @pte:pte to check
+ * @lastlba: last lba of the disk
  *
  * Description: returns 1 if valid,  0 on error.
  */
···
 
 /**
  * compare_gpts() - Search disk for valid GPT headers and PTEs
- * @pgpt is the primary GPT header
- * @agpt is the alternate GPT header
- * @lastlba is the last LBA number
+ * @pgpt: primary GPT header
+ * @agpt: alternate GPT header
+ * @lastlba: last LBA number
+ *
  * Description: Returns nothing.  Sanity checks pgpt and agpt fields
  * and prints warnings on discrepancies.
  * 
···
 
 /**
  * find_valid_gpt() - Search disk for valid GPT headers and PTEs
- * @state
- * @gpt is a GPT header ptr, filled on return.
- * @ptes is a PTEs ptr, filled on return.
+ * @state: disk parsed partitions
+ * @gpt: GPT header ptr, filled on return.
+ * @ptes: PTEs ptr, filled on return.
+ *
  * Description: Returns 1 if valid, 0 on error.
  * If valid, returns pointers to newly allocated GPT header and PTEs.
  * Validity depends on PMBR being valid (or being overridden by the
···
 
 /**
  * efi_partition(struct parsed_partitions *state)
- * @state
+ * @state: disk parsed partitions
  *
  * Description: called from check.c, if the disk contains GPT
  * partitions, sets up partition entries in the kernel.

block/partitions/msdos.c

···
 		/*
 		 * First process the data partition(s)
 		 */
-		for (i=0; i<4; i++, p++) {
+		for (i = 0; i < 4; i++, p++) {
 			sector_t offs, size, next;
+
 			if (!nr_sects(p) || is_extended_partition(p))
 				continue;
 
···
 		 * It should be a link to the next logical partition.
 		 */
 		p -= 4;
-		for (i=0; i<4; i++, p++)
+		for (i = 0; i < 4; i++, p++)
 			if (nr_sects(p) && is_extended_partition(p))
 				break;
 		if (i == 4)
···
 		return;
 	}
 	/* Ensure we can handle previous case of VTOC with 8 entries gracefully */
-	max_nparts = le16_to_cpu (v->v_nparts) > 8 ? SOLARIS_X86_NUMSLICE : 8;
-	for (i=0; i<max_nparts && state->next<state->limit; i++) {
+	max_nparts = le16_to_cpu(v->v_nparts) > 8 ? SOLARIS_X86_NUMSLICE : 8;
+	for (i = 0; i < max_nparts && state->next < state->limit; i++) {
 		struct solaris_x86_slice *s = &v->v_slice[i];
 		char tmp[3 + 10 + 1 + 1];
 
···
 	/* The first sector of a Minix partition can have either
 	 * a secondary MBR describing its subpartitions, or
 	 * the normal boot sector. */
-	if (msdos_magic_present (data + 510) &&
+	if (msdos_magic_present(data + 510) &&
 	    SYS_IND(p) == MINIX_PARTITION) { /* subpartition table present */
 		char tmp[1 + BDEVNAME_SIZE + 10 + 9 + 1];
 
···
 	for (slot = 1 ; slot <= 4 ; slot++, p++) {
 		sector_t start = start_sect(p)*sector_size;
 		sector_t size = nr_sects(p)*sector_size;
+
 		if (!size)
 			continue;
 		if (is_extended_partition(p)) {
···
 			 * sector, although it may not be enough/proper.
 			 */
 			sector_t n = 2;
+
 			n = min(size, max(sector_size, n));
 			put_partition(state, slot, start, n);
 

block/scsi_ioctl.c

···
 	return err;
 }
 
+static int max_sectors_bytes(struct request_queue *q)
+{
+	unsigned int max_sectors = queue_max_sectors(q);
+
+	max_sectors = min_t(unsigned int, max_sectors, INT_MAX >> 9);
+
+	return max_sectors << 9;
+}
+
 static int sg_get_reserved_size(struct request_queue *q, int __user *p)
 {
-	unsigned val = min(q->sg_reserved_size, queue_max_sectors(q) << 9);
+	int val = min_t(int, q->sg_reserved_size, max_sectors_bytes(q));
 
 	return put_user(val, p);
 }
···
 
 	if (size < 0)
 		return -EINVAL;
-	if (size > (queue_max_sectors(q) << 9))
-		size = queue_max_sectors(q) << 9;
 
-	q->sg_reserved_size = size;
+	q->sg_reserved_size = min(size, max_sectors_bytes(q));
 	return 0;
 }
 

include/linux/bio.h

···
 
 	unsigned short		bip_slab;	/* slab the bip came from */
 	unsigned short		bip_vcnt;	/* # of integrity bio_vecs */
+	unsigned short		bip_max_vcnt;	/* integrity bio_vec slots */
 	unsigned		bip_owns_buf:1;	/* should free bip_buf */
 
 	struct work_struct	bip_work;	/* I/O completion */

include/linux/blkdev.h

···
 #include <linux/bsg.h>
 #include <linux/smp.h>
 #include <linux/rcupdate.h>
+#include <linux/percpu-refcount.h>
 
 #include <asm/scatterlist.h>
 
···
 	struct mutex		sysfs_lock;
 
 	int			bypass_depth;
+	int			mq_freeze_depth;
 
 #if defined(CONFIG_BLK_DEV_BSG)
 	bsg_job_fn		*bsg_job_fn;
···
 #endif
 	struct rcu_head		rcu_head;
 	wait_queue_head_t	mq_freeze_wq;
-	struct percpu_counter	mq_usage_counter;
+	struct percpu_ref	mq_usage_counter;
 	struct list_head	all_q_node;
 
 	struct blk_mq_tag_set	*tag_set;