Skip to contentSkip to conversation

md: personality pushdown patches -- intro

PATCH In review Andre Noll 3 participants 6 messages May 29, 2009 – Jun 3, 2009 MD RAID

lore
Share & tools
lore

Andre Noll

author
#0 lore raw

cover md: personality pushdown patches -- intro

Here are the first two patches for pushing down raid level dependent code from md.c to the corresponding personalities.

The first patch is trivial as it moves only a printk() to the personalities. The second patch removes knowledge about bitmaps from md.c.

If the general approach used by the second patch is agreed with, I will follow up with patches that deal with data integrity registration and multipath handling which are the remaining two issues where md.c does what should be handled in personality code.

Thanks
Andre

Signature · 1 line
1.5.4.3
author
#1 lore raw

PATCH 1/2 md: Push down reconstruction log message to personality code.

Currently, the md layer checks in analyze_sbs() if the raid level supports reconstruction (mddev->level >= 1) and if reconstruction is in progress (mddev->recovery_cp != MaxSector).

Move that printk into the personality code of those raid levels that care (levels 1, 4, 5, 6, 10).

---
 drivers/md/md.c     |    9 ---------
 drivers/md/raid1.c  |    4 ++++
 drivers/md/raid10.c |    4 ++++
 drivers/md/raid5.c  |    4 ++++
 4 files changed, 12 insertions(+), 9 deletions(-)
drivers/md/md.c +0 -9
index a54ec91..1d71cc2 100644
@@ -2597,15 +2597,6 @@ static void analyze_sbs(mddev_t * mddev)
 			clear_bit(In_sync, &rdev->flags);
 		}
 	}
-
-
-
-	if (mddev->recovery_cp != MaxSector &&
-	    mddev->level >= 1)
-		printk(KERN_ERR "md: %s: raid array is not clean"
-		       " -- starting background reconstruction\n",
-		       mdname(mddev));
-
 }
 
 static void md_safemode_timeout(unsigned long data);
drivers/md/raid1.c +4 -0
index 4692a8a..d29842f 100644
@@ -2052,6 +2052,10 @@ static int run(mddev_t *mddev)
 		goto out_free_conf;
 	}
 
+	if (mddev->recovery_cp != MaxSector)
+		printk(KERN_NOTICE "raid1: %s is not clean"
+		       " -- starting background reconstruction\n",
+		       mdname(mddev));
 	printk(KERN_INFO 
 		"raid1: raid set %s active with %d out of %d mirrors\n",
 		mdname(mddev), mddev->raid_disks - mddev->degraded, 
drivers/md/raid10.c +4 -0
index d75cb96..c317543 100644
@@ -2186,6 +2186,10 @@ static int run(mddev_t *mddev)
 		goto out_free_conf;
 	}
 
+	if (mddev->recovery_cp != MaxSector)
+		printk(KERN_NOTICE "raid10: %s is not clean"
+		       " -- starting background reconstruction\n",
+		       mdname(mddev));
 	printk(KERN_INFO
 		"raid10: raid set %s active with %d out of %d devices\n",
 		mdname(mddev), mddev->raid_disks - mddev->degraded,
drivers/md/raid5.c +4 -0
index 9ec505b..cdd5b49 100644
@@ -4454,6 +4454,10 @@ static int run(mddev_t *mddev)
 	int working_disks = 0;
 	mdk_rdev_t *rdev;
 
+	if (mddev->recovery_cp != MaxSector)
+		printk(KERN_NOTICE "raid5: %s is not clean"
+		       " -- starting background reconstruction\n",
+		       mdname(mddev));
 	if (mddev->reshape_position != MaxSector) {
 		/* Check that we can continue the reshape.
 		 * Currently only disks can change, it must
Signature · 1 line
1.5.4.3
#2 lore raw

Re: [PATCH 1/2] md: Push down reconstruction log message to personality code.

Meil/Andre Hello
I am happy to see this patch. I have lots of "md" mess with the raid0 reshape because of some md "assumptions".
1. in md_do_sync , max_sectors is determined to be dev_sectors. Currently I had to patch to be
array_sectors in the case of level=0.
2. in procedure super_90_sync the minor version is determined by the reshape position ?
So now, when i reshape I see that the minor version simply changes .

thank you
raz

Quoted 83 lines · Andre Noll

Currently, the md layer checks in analyze_sbs() if the raid level supports reconstruction (mddev->level >= 1) and if reconstruction is in progress (mddev->recovery_cp != MaxSector).

Move that printk into the personality code of those raid levels that care (levels 1, 4, 5, 6, 10).

---
 drivers/md/md.c     |    9 ---------
 drivers/md/raid1.c  |    4 ++++
 drivers/md/raid10.c |    4 ++++
 drivers/md/raid5.c  |    4 ++++

 4 files changed, 12 insertions(+), 9 deletions(-)
drivers/md/md.c +0 -9
index a54ec91..1d71cc2 100644
@@ -2597,15 +2597,6 @@ static void analyze_sbs(mddev_t * mddev)
                        clear_bit(In_sync, &rdev->flags);
                }
        }
-
-
-
-       if (mddev->recovery_cp != MaxSector &&
-           mddev->level >= 1)
-               printk(KERN_ERR "md: %s: raid array is not clean"
-                      " -- starting background reconstruction\n",
-                      mdname(mddev));
-
  }
 
  static void md_safemode_timeout(unsigned long data);
drivers/md/raid1.c +4 -0
index 4692a8a..d29842f 100644
@@ -2052,6 +2052,10 @@ static int run(mddev_t *mddev)
                goto out_free_conf;
        }
 
+       if (mddev->recovery_cp != MaxSector)
+               printk(KERN_NOTICE "raid1: %s is not clean"
+                      " -- starting background reconstruction\n",
+                      mdname(mddev));

       printk(KERN_INFO
               "raid1: raid set %s active with %d out of %d mirrors\n", mdname(mddev), mddev->raid_disks - mddev->degraded,

drivers/md/raid10.c +4 -0
index d75cb96..c317543 100644
@@ -2186,6 +2186,10 @@ static int run(mddev_t *mddev)
                goto out_free_conf;
        }
 
+       if (mddev->recovery_cp != MaxSector)
+               printk(KERN_NOTICE "raid10: %s is not clean"
+                      " -- starting background reconstruction\n",
+                      mdname(mddev));

       printk(KERN_INFO
               "raid10: raid set %s active with %d out of %d devices\n", mdname(mddev), mddev->raid_disks - mddev->degraded,

drivers/md/raid5.c +4 -1
index 9ec505b..cdd5b49 100644
@@ -4454,6 +4454,10 @@ static int run(mddev_t *mddev)
        int working_disks = 0;
        mdk_rdev_t *rdev;
 
+       if (mddev->recovery_cp != MaxSector)
+               printk(KERN_NOTICE "raid5: %s is not clean"
+                      " -- starting background reconstruction\n",
+                      mdname(mddev));
        if (mddev->reshape_position != MaxSector) {
                /* Check that we can continue the reshape.
                 * Currently only disks can change, it must
--

1.5.4.3

Quoted 5 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

author
#3 lore raw

PATCH 2/2 md: Move check for bitmap presence to personality code.

In reply to Andre Noll

If the superblock of a component device indicates the presence of a bitmap but the corresponding raid personality does not support bitmaps (raid0, linear, multipath, faulty), then something is seriously wrong and we'd better refuse to run such an array.

Currently, this check is performed while the superblocks are examined, i.e. before entering personality code. Therefore the generic md layer must know which raid levels support bitmaps and which do not.

This patch avoids this layer violation without adding identical code to various personalities. This is accomplished by introducing a new public function to md.c, md_bitmap_present(), which replaces the hard-coded checks in the superblock loading functions.

A call to md_bitmap_present() is added to the ->run method of each personality which does not support bitmaps and assembly is aborted if at least one component device contains a bitmap.

---
 drivers/md/faulty.c    |    6 +++-
 drivers/md/linear.c    |    2 +
 drivers/md/md.c        |   64 +++++++++++++++++++++++++++++++----------------
 drivers/md/md.h        |    1 +
 drivers/md/multipath.c |    3 ++
 drivers/md/raid0.c     |    2 +
 6 files changed, 55 insertions(+), 23 deletions(-)
drivers/md/faulty.c +5 -1
index 8695809..cec5d1e 100644
@@ -298,8 +298,12 @@ static int run(mddev_t *mddev)
 {
 	mdk_rdev_t *rdev;
 	int i;
+	conf_t *conf;
 
-	conf_t *conf = kmalloc(sizeof(*conf), GFP_KERNEL);
+	if (md_bitmap_present(mddev))
+		return -EINVAL;
+
+	conf = kmalloc(sizeof(*conf), GFP_KERNEL);
 	if (!conf)
 		return -ENOMEM;
 
drivers/md/linear.c +2 -0
index ead1051..2ca5514 100644
@@ -183,6 +183,8 @@ static int linear_run (mddev_t *mddev)
 {
 	linear_conf_t *conf;
 
+	if (md_bitmap_present(mddev))
+		return -EINVAL;
 	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
 	conf = linear_conf(mddev, mddev->raid_disks);
 
drivers/md/md.c +42 -22

82 lines not on this page · raw message

drivers/md/md.h +1 -0
index 4274cdc..903479f 100644
@@ -438,5 +438,6 @@ extern void md_new_event(mddev_t *mddev);
 extern int md_allow_write(mddev_t *mddev);
 extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
 extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
+extern int md_bitmap_present(mddev_t *mddev);
 
 #endif /* _MD_MD_H */
drivers/md/multipath.c +3 -0
index ca387ba..d11285c 100644
@@ -421,6 +421,9 @@ static int multipath_run (mddev_t *mddev)
 	struct multipath_info *disk;
 	mdk_rdev_t *rdev;
 
+	if (md_bitmap_present(mddev))
+		return -EINVAL;
+
 	if (mddev->level != LEVEL_MULTIPATH) {
 		printk("multipath: %s: raid level not set to multipath IO (%d)\n",
 		       mdname(mddev), mddev->level);
drivers/md/raid0.c +2 -0
index 7d680bc..9d6f14e 100644
@@ -311,6 +311,8 @@ static int raid0_run(mddev_t *mddev)
 		printk(KERN_ERR "md/raid0: chunk size must be a power of 2.\n");
 		return -EINVAL;
 	}
+	if (md_bitmap_present(mddev))
+		return -EINVAL;
 	blk_queue_max_sectors(mddev->queue, mddev->chunk_sectors);
 	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
 
Signature · 1 line
1.5.4.3
#4 lore raw
In reply to Andre Noll
Quoted 6 lines · On Friday May 29, [email protected] wrote:

Here are the first two patches for pushing down raid level dependent code from md.c to the corresponding personalities.

The first patch is trivial as it moves only a printk() to the personalities. The second patch removes knowledge about bitmaps from md.c.

Thanks Andre.
This first is good and is now in my for-next branch.

The second is heading in the right direction.
However md_bitmap_present is as much of a layering violation in its own way as the previous code was - diving in to the metatype_type specific data.

You will notice in bitmap_create that an array has a bitmap if and only if (mddev->bitmap_file || mddev->bitmap_offset).

So that is the test which should be used in md_bitmap_present.

Also the function name "md_bitmap_present" give no hint that it will report an error - it reads like a simple test.
Maybe "md_check_no_bitmap" to indicate what the function is expecting to find?

Thanks,
NeilBrown

Quoted 10 lines

If the general approach used by the second patch is agreed with, I will follow up with patches that deal with data integrity registration and multipath handling which are the remaining two issues where md.c does what should be handled in personality code.

Thanks
Andre

Signature · 1 line
1.5.4.3
author
#5 lore raw
Quoted 8 lines · Neil Brown

The second is heading in the right direction.
However md_bitmap_present is as much of a layering violation in its own way as the previous code was - diving in to the metatype_type specific data.

You will notice in bitmap_create that an array has a bitmap if and only if (mddev->bitmap_file || mddev->bitmap_offset). So that is the test which should be used in md_bitmap_present.

I see. And this makes the code much simpler. In fact it's so simple that we might probably want to inline it.

Quoted 4 lines

Also the function name "md_bitmap_present" give no hint that it will report an error - it reads like a simple test.
Maybe "md_check_no_bitmap" to indicate what the function is expecting to find?

Yup, md_check_no_bitmap is more descriptive. The patch below addresses both issues.

Thanks
Andre
---

commit cb2e46af5393cfed1b46182e2ca19a4adb7a4e0d

Author: Andre Noll <[email protected]>
Date:   Tue Jun 2 21:49:29 2009 +0200

md: Move check for bitmap presence to personality code.

If the superblock of a component device indicates the presence of a bitmap but the corresponding raid personality does not support bitmaps (raid0, linear, multipath, faulty), then something is seriously wrong and we'd better refuse to run such an array.

Currently, this check is performed while the superblocks are examined, i.e. before entering personality code. Therefore the generic md layer must know which raid levels support bitmaps and which do not.

This patch avoids this layer violation without adding identical code to various personalities. This is accomplished by introducing a new public function to md.c, md_check_no_bitmap(), which replaces the hard-coded checks in the superblock loading functions.

A call to md_check_no_bitmap() is added to the ->run method of each personality which does not support bitmaps and assembly is aborted if at least one component device contains a bitmap.

drivers/md/faulty.c +5 -1
index 6e83b38..87d88db 100644
@@ -299,8 +299,12 @@ static int run(mddev_t *mddev)
 {
 	mdk_rdev_t *rdev;
 	int i;
+	conf_t *conf;
 
-	conf_t *conf = kmalloc(sizeof(*conf), GFP_KERNEL);
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
+
+	conf = kmalloc(sizeof(*conf), GFP_KERNEL);
 	if (!conf)
 		return -ENOMEM;
 
drivers/md/linear.c +2 -0
index 56c46e0..f9c713d 100644
@@ -183,6 +183,8 @@ static int linear_run (mddev_t *mddev)
 {
 	linear_conf_t *conf;
 
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
 	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
 	conf = linear_conf(mddev, mddev->raid_disks);
 
drivers/md/md.c +18 -22
index 812a1f7..aaeaa69 100644
@@ -748,6 +748,24 @@ struct super_type  {
 };
 
 /*
+ * Check that the given mddev has no bitmap.
+ *
+ * This function is called from the run method of all personalities that do not
+ * support bitmaps. It prints an error message and returns non-zero if mddev
+ * has a bitmap. Otherwise, it returns 0.
+ *
+ */
+int md_check_no_bitmap(mddev_t *mddev)
+{
+	if (!mddev->bitmap_file && !mddev->bitmap_offset)
+		return 0;
+	printk(KERN_ERR "%s: bitmaps are not supported for %s\n",
+		mdname(mddev), mddev->pers->name);
+	return 1;
+}
+EXPORT_SYMBOL(md_bitmap_present);
+
+/*
  * load_super for 0.90.0 
  */
 static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
@@ -800,17 +818,6 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
 	rdev->data_offset = 0;
 	rdev->sb_size = MD_SB_BYTES;
 
-	if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
-		if (sb->level != 1 && sb->level != 4
-		    && sb->level != 5 && sb->level != 6
-		    && sb->level != 10) {
-			/* FIXME use a better test */
-			printk(KERN_WARNING
-			       "md: bitmaps not supported for this level.\n");
-			goto abort;
-		}
-	}
-
 	if (sb->level == LEVEL_MULTIPATH)
 		rdev->desc_nr = -1;
 	else
@@ -1188,17 +1195,6 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
 		       bdevname(rdev->bdev,b));
 		return -EINVAL;
 	}
-	if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) {
-		if (sb->level != cpu_to_le32(1) &&
-		    sb->level != cpu_to_le32(4) &&
-		    sb->level != cpu_to_le32(5) &&
-		    sb->level != cpu_to_le32(6) &&
-		    sb->level != cpu_to_le32(10)) {
-			printk(KERN_WARNING
-			       "md: bitmaps not supported for this level.\n");
-			return -EINVAL;
-		}
-	}
 
 	rdev->preferred_minor = 0xffff;
 	rdev->data_offset = le64_to_cpu(sb->data_offset);
drivers/md/md.h +1 -0
index bac7c2b..ccc0ace 100644
@@ -437,5 +437,6 @@ extern void md_new_event(mddev_t *mddev);
 extern int md_allow_write(mddev_t *mddev);
 extern void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev);
 extern void md_set_array_sectors(mddev_t *mddev, sector_t array_sectors);
+extern int md_check_no_bitmap(mddev_t *mddev);
 
 #endif /* _MD_MD_H */
drivers/md/multipath.c +3 -0
index ca387ba..fb8a5e7 100644
@@ -421,6 +421,9 @@ static int multipath_run (mddev_t *mddev)
 	struct multipath_info *disk;
 	mdk_rdev_t *rdev;
 
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
+
 	if (mddev->level != LEVEL_MULTIPATH) {
 		printk("multipath: %s: raid level not set to multipath IO (%d)\n",
 		       mdname(mddev), mddev->level);
drivers/md/raid0.c +2 -0
index 0fcd9b5..3c78b42 100644
@@ -311,6 +311,8 @@ static int raid0_run(mddev_t *mddev)
 		printk(KERN_ERR "md/raid0: chunk size must be a power of 2.\n");
 		return -EINVAL;
 	}
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
 	blk_queue_max_sectors(mddev->queue, mddev->chunk_sectors);
 	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
 
Signature · 1 line
The only person who always got his work done by Friday was Robinson Crusoe