block: Fix incorrect alignment offset reporting and update documentation

queue_sector_alignment_offset returned the wrong value which caused partitions to report an incorrect alignment_offset. Since offset alignment calculation is needed several places it has been split into a separate helper function. The topology stacking function has been updated accordingly. Furthermore, comments have been added to clarify how the stacking function works. Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Tested-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

block: Fix incorrect alignment offset reporting and update documentation
queue_sector_alignment_offset returned the wrong value which caused partitions to report an incorrect alignment_offset. Since offset alignment calculation is needed several places it has been split into a separate helper function. The topology stacking function has been updated accordingly. Furthermore, comments have been added to clarify how the stacking function works. Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Tested-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Martin K. Petersen · Jens Axboe
1 parent 2f7a2d89a8
Showing 2 changed files with 42 additions and 13 deletions Side-by-side Diff
block/blk-settings.c
include/linux/blkdev.h
@@ -505,20 +505,30 @@
  
 /**
  * blk_stack_limits - adjust queue_limits for stacked devices
- * @t:	the stacking driver limits (top)
- * @b:  the underlying queue limits (bottom)
+ * @t:	the stacking driver limits (top device)
+ * @b:  the underlying queue limits (bottom, component device)
  * @offset:  offset to beginning of data within component device
  *
  * Description:
- *    Merges two queue_limit structs.  Returns 0 if alignment didn't
- *    change.  Returns -1 if adding the bottom device caused
- *    misalignment.
+ *    This function is used by stacking drivers like MD and DM to ensure
+ *    that all component devices have compatible block sizes and
+ *    alignments.  The stacking driver must provide a queue_limits
+ *    struct (top) and then iteratively call the stacking function for
+ *    all component (bottom) devices.  The stacking function will
+ *    attempt to combine the values and ensure proper alignment.
+ *
+ *    Returns 0 if the top and bottom queue_limits are compatible.  The
+ *    top device's block sizes and alignment offsets may be adjusted to
+ *    ensure alignment with the bottom device. If no compatible sizes
+ *    and alignments exist, -1 is returned and the resulting top
+ *    queue_limits will have the misaligned flag set to indicate that
+ *    the alignment_offset is undefined.
  */
 int blk_stack_limits(struct queue_limits *t, struct queue_limits *b,
 		     sector_t offset)
 {
 	sector_t alignment;
-	unsigned int top, bottom, granularity;
+	unsigned int top, bottom;
  
 	t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
 	t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
  
  
  
@@ -536,15 +546,18 @@
 	t->max_segment_size = min_not_zero(t->max_segment_size,
 					   b->max_segment_size);
  
-	granularity = max(b->physical_block_size, b->io_min);
-	alignment = b->alignment_offset - (offset & (granularity - 1));
+	alignment = queue_limit_alignment_offset(b, offset);
  
+	/* Bottom device has different alignment.  Check that it is
+	 * compatible with the current top alignment.
+	 */
 	if (t->alignment_offset != alignment) {
  
 		top = max(t->physical_block_size, t->io_min)
 			+ t->alignment_offset;
-		bottom = granularity + alignment;
+		bottom = max(b->physical_block_size, b->io_min) + alignment;
  
+		/* Verify that top and bottom intervals line up */
 		if (max(top, bottom) & (min(top, bottom) - 1))
 			t->misaligned = 1;
 	}
  
  
  
  
  
  
@@ -561,32 +574,39 @@
 	t->no_cluster |= b->no_cluster;
 	t->discard_zeroes_data &= b->discard_zeroes_data;
  
+	/* Physical block size a multiple of the logical block size? */
 	if (t->physical_block_size & (t->logical_block_size - 1)) {
 		t->physical_block_size = t->logical_block_size;
 		t->misaligned = 1;
 	}
  
+	/* Minimum I/O a multiple of the physical block size? */
 	if (t->io_min & (t->physical_block_size - 1)) {
 		t->io_min = t->physical_block_size;
 		t->misaligned = 1;
 	}
  
+	/* Optimal I/O a multiple of the physical block size? */
 	if (t->io_opt & (t->physical_block_size - 1)) {
 		t->io_opt = 0;
 		t->misaligned = 1;
 	}
  
+	/* Find lowest common alignment_offset */
 	t->alignment_offset = lcm(t->alignment_offset, alignment)
 		& (max(t->physical_block_size, t->io_min) - 1);
  
+	/* Verify that new alignment_offset is on a logical block boundary */
 	if (t->alignment_offset & (t->logical_block_size - 1))
 		t->misaligned = 1;
  
 	/* Discard alignment and granularity */
 	if (b->discard_granularity) {
+		unsigned int granularity = b->discard_granularity;
+		offset &= granularity - 1;
  
-		alignment = b->discard_alignment -
-			(offset & (b->discard_granularity - 1));
+		alignment = (granularity + b->discard_alignment - offset)
+			& (granularity - 1);
  
 		if (t->discard_granularity != 0 &&
 		    t->discard_alignment != alignment) {
@@ -598,6 +618,8 @@
 				t->discard_misaligned = 1;
 		}
  
+		t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
+						      b->max_discard_sectors);
 		t->discard_granularity = max(t->discard_granularity,
 					     b->discard_granularity);
 		t->discard_alignment = lcm(t->discard_alignment, alignment) &
@@ -1116,11 +1116,18 @@
 	return q->limits.alignment_offset;
 }
  
+static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t offset)
+{
+	unsigned int granularity = max(lim->physical_block_size, lim->io_min);
+
+	offset &= granularity - 1;
+	return (granularity + lim->alignment_offset - offset) & (granularity - 1);
+}
+
 static inline int queue_sector_alignment_offset(struct request_queue *q,
 						sector_t sector)
 {
-	return ((sector << 9) - q->limits.alignment_offset)
-		& (q->limits.io_min - 1);
+	return queue_limit_alignment_offset(&q->limits, sector << 9);
 }
  
 static inline int bdev_alignment_offset(struct block_device *bdev)
...	...	@@ -505,20 +505,30 @@
505	505
506	506	/**
507	507	* blk_stack_limits - adjust queue_limits for stacked devices
508		- * @t: the stacking driver limits (top)
509		- * @b: the underlying queue limits (bottom)
	508	+ * @t: the stacking driver limits (top device)
	509	+ * @b: the underlying queue limits (bottom, component device)
510	510	* @offset: offset to beginning of data within component device
511	511	*
512	512	* Description:
513		- * Merges two queue_limit structs. Returns 0 if alignment didn't
514		- * change. Returns -1 if adding the bottom device caused
515		- * misalignment.
	513	+ * This function is used by stacking drivers like MD and DM to ensure
	514	+ * that all component devices have compatible block sizes and
	515	+ * alignments. The stacking driver must provide a queue_limits
	516	+ * struct (top) and then iteratively call the stacking function for
	517	+ * all component (bottom) devices. The stacking function will
	518	+ * attempt to combine the values and ensure proper alignment.
	519	+ *
	520	+ * Returns 0 if the top and bottom queue_limits are compatible. The
	521	+ * top device's block sizes and alignment offsets may be adjusted to
	522	+ * ensure alignment with the bottom device. If no compatible sizes
	523	+ * and alignments exist, -1 is returned and the resulting top
	524	+ * queue_limits will have the misaligned flag set to indicate that
	525	+ * the alignment_offset is undefined.
516	526	*/
517	527	int blk_stack_limits(struct queue_limits t, struct queue_limits b,
518	528	sector_t offset)
519	529	{
520	530	sector_t alignment;
521		- unsigned int top, bottom, granularity;
	531	+ unsigned int top, bottom;
522	532
523	533	t->max_sectors = min_not_zero(t->max_sectors, b->max_sectors);
524	534	t->max_hw_sectors = min_not_zero(t->max_hw_sectors, b->max_hw_sectors);
525	535
526	536
527	537
...	...	@@ -536,15 +546,18 @@
536	546	t->max_segment_size = min_not_zero(t->max_segment_size,
537	547	b->max_segment_size);
538	548
539		- granularity = max(b->physical_block_size, b->io_min);
540		- alignment = b->alignment_offset - (offset & (granularity - 1));
	549	+ alignment = queue_limit_alignment_offset(b, offset);
541	550
	551	+ /* Bottom device has different alignment. Check that it is
	552	+ * compatible with the current top alignment.
	553	+ */
542	554	if (t->alignment_offset != alignment) {
543	555
544	556	top = max(t->physical_block_size, t->io_min)
545	557	+ t->alignment_offset;
546		- bottom = granularity + alignment;
	558	+ bottom = max(b->physical_block_size, b->io_min) + alignment;
547	559
	560	+ /* Verify that top and bottom intervals line up */
548	561	if (max(top, bottom) & (min(top, bottom) - 1))
549	562	t->misaligned = 1;
550	563	}
551	564
552	565
553	566
554	567
555	568
556	569
...	...	@@ -561,32 +574,39 @@
561	574	t->no_cluster \|= b->no_cluster;
562	575	t->discard_zeroes_data &= b->discard_zeroes_data;
563	576
	577	+ /* Physical block size a multiple of the logical block size? */
564	578	if (t->physical_block_size & (t->logical_block_size - 1)) {
565	579	t->physical_block_size = t->logical_block_size;
566	580	t->misaligned = 1;
567	581	}
568	582
	583	+ /* Minimum I/O a multiple of the physical block size? */
569	584	if (t->io_min & (t->physical_block_size - 1)) {
570	585	t->io_min = t->physical_block_size;
571	586	t->misaligned = 1;
572	587	}
573	588
	589	+ /* Optimal I/O a multiple of the physical block size? */
574	590	if (t->io_opt & (t->physical_block_size - 1)) {
575	591	t->io_opt = 0;
576	592	t->misaligned = 1;
577	593	}
578	594
	595	+ /* Find lowest common alignment_offset */
579	596	t->alignment_offset = lcm(t->alignment_offset, alignment)
580	597	& (max(t->physical_block_size, t->io_min) - 1);
581	598
	599	+ /* Verify that new alignment_offset is on a logical block boundary */
582	600	if (t->alignment_offset & (t->logical_block_size - 1))
583	601	t->misaligned = 1;
584	602
585	603	/* Discard alignment and granularity */
586	604	if (b->discard_granularity) {
	605	+ unsigned int granularity = b->discard_granularity;
	606	+ offset &= granularity - 1;
587	607
588		- alignment = b->discard_alignment -
589		- (offset & (b->discard_granularity - 1));
	608	+ alignment = (granularity + b->discard_alignment - offset)
	609	+ & (granularity - 1);
590	610
591	611	if (t->discard_granularity != 0 &&
592	612	t->discard_alignment != alignment) {
...	...	@@ -598,6 +618,8 @@
598	618	t->discard_misaligned = 1;
599	619	}
600	620
	621	+ t->max_discard_sectors = min_not_zero(t->max_discard_sectors,
	622	+ b->max_discard_sectors);
601	623	t->discard_granularity = max(t->discard_granularity,
602	624	b->discard_granularity);
603	625	t->discard_alignment = lcm(t->discard_alignment, alignment) &
...	...	@@ -1116,11 +1116,18 @@
1116	1116	return q->limits.alignment_offset;
1117	1117	}
1118	1118
	1119	+static inline int queue_limit_alignment_offset(struct queue_limits *lim, sector_t offset)
	1120	+{
	1121	+ unsigned int granularity = max(lim->physical_block_size, lim->io_min);
	1122	+
	1123	+ offset &= granularity - 1;
	1124	+ return (granularity + lim->alignment_offset - offset) & (granularity - 1);
	1125	+}
	1126	+
1119	1127	static inline int queue_sector_alignment_offset(struct request_queue *q,
1120	1128	sector_t sector)
1121	1129	{
1122		- return ((sector << 9) - q->limits.alignment_offset)
1123		- & (q->limits.io_min - 1);
	1130	+ return queue_limit_alignment_offset(&q->limits, sector << 9);
1124	1131	}
1125	1132
1126	1133	static inline int bdev_alignment_offset(struct block_device *bdev)