linux-raid.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: NeilBrown <neilb@suse.de>
To: linux-raid@vger.kernel.org
Subject: [md PATCH 03/17] md/raid10: preferentially read from replacement device if possible.
Date: Wed, 02 Nov 2011 16:25:43 +1100	[thread overview]
Message-ID: <20111102052543.17566.52804.stgit@notabene.brown> (raw)
In-Reply-To: <20111102051851.17566.52748.stgit@notabene.brown>

When reading (for array reads, not for recovery etc) we read from the
replacement device if it has recovered far enough.
This requires storing the chosen rdev in the 'r10_bio' so we can make
sure to drop the ref on the right device when the read finishes.

Signed-off-by: NeilBrown <neilb@suse.de>
---

 drivers/md/raid10.c |   36 +++++++++++++++++++++++-------------
 1 files changed, 23 insertions(+), 13 deletions(-)

diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 254a6ce..fffdc43 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -323,11 +323,13 @@ static void raid10_end_read_request(struct bio *bio, int error)
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
 	struct r10bio *r10_bio = bio->bi_private;
 	int slot, dev;
+	struct md_rdev *rdev;
 	struct r10conf *conf = r10_bio->mddev->private;
 
 
 	slot = r10_bio->read_slot;
 	dev = r10_bio->devs[slot].devnum;
+	rdev = r10_bio->devs[slot].rdev;
 	/*
 	 * this branch is our 'one mirror IO has finished' event handler:
 	 */
@@ -345,7 +347,7 @@ static void raid10_end_read_request(struct bio *bio, int error)
 		 */
 		set_bit(R10BIO_Uptodate, &r10_bio->state);
 		raid_end_bio_io(r10_bio);
-		rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
+		rdev_dec_pending(rdev, conf->mddev);
 	} else {
 		/*
 		 * oops, read error - keep the refcount on the rdev
@@ -354,7 +356,7 @@ static void raid10_end_read_request(struct bio *bio, int error)
 		printk_ratelimited(KERN_ERR
 				   "md/raid10:%s: %s: rescheduling sector %llu\n",
 				   mdname(conf->mddev),
-				   bdevname(conf->mirrors[dev].rdev->bdev, b),
+				   bdevname(rdev->bdev, b),
 				   (unsigned long long)r10_bio->sector);
 		set_bit(R10BIO_ReadError, &r10_bio->state);
 		reschedule_retry(r10_bio);
@@ -598,7 +600,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
 	int sectors = r10_bio->sectors;
 	int best_good_sectors;
 	sector_t new_distance, best_dist;
-	struct md_rdev *rdev;
+	struct md_rdev *rdev, *best_rdev;
 	int do_balance;
 	int best_slot;
 
@@ -607,6 +609,7 @@ static struct md_rdev *read_balance(struct r10conf *conf,
 retry:
 	sectors = r10_bio->sectors;
 	best_slot = -1;
+	best_rdev = NULL;
 	best_dist = MaxSector;
 	best_good_sectors = 0;
 	do_balance = 1;
@@ -628,10 +631,16 @@ retry:
 		if (r10_bio->devs[slot].bio == IO_BLOCKED)
 			continue;
 		disk = r10_bio->devs[slot].devnum;
-		rdev = rcu_dereference(conf->mirrors[disk].rdev);
+		rdev = rcu_dereference(conf->mirrors[disk].replacement);
+		if (rdev == NULL || test_bit(Faulty, &rdev->flags) ||
+		    r10_bio->devs[slot].addr + sectors > rdev->recovery_offset)
+			rdev = rcu_dereference(conf->mirrors[disk].rdev);
 		if (rdev == NULL)
 			continue;
-		if (!test_bit(In_sync, &rdev->flags))
+		if (test_bit(Faulty, &rdev->flags))
+			continue;
+		if (!test_bit(In_sync, &rdev->flags) &&
+		    r10_bio->devs[slot].addr + sectors > rdev->recovery_offset)
 			continue;
 
 		dev_sector = r10_bio->devs[slot].addr;
@@ -656,6 +665,7 @@ retry:
 				if (good_sectors > best_good_sectors) {
 					best_good_sectors = good_sectors;
 					best_slot = slot;
+					best_rdev = rdev;
 				}
 				if (!do_balance)
 					/* Must read from here */
@@ -684,16 +694,15 @@ retry:
 		if (new_distance < best_dist) {
 			best_dist = new_distance;
 			best_slot = slot;
+			best_rdev = rdev;
 		}
 	}
-	if (slot == conf->copies)
+	if (slot >= conf->copies) {
 		slot = best_slot;
+		rdev = best_rdev;
+	}
 
 	if (slot >= 0) {
-		disk = r10_bio->devs[slot].devnum;
-		rdev = rcu_dereference(conf->mirrors[disk].rdev);
-		if (!rdev)
-			goto retry;
 		atomic_inc(&rdev->nr_pending);
 		if (test_bit(Faulty, &rdev->flags)) {
 			/* Cannot risk returning a device that failed
@@ -991,6 +1000,7 @@ read_again:
 			    max_sectors);
 
 		r10_bio->devs[slot].bio = read_bio;
+		r10_bio->devs[slot].rdev = rdev;
 
 		read_bio->bi_sector = r10_bio->devs[slot].addr +
 			rdev->data_offset;
@@ -2090,10 +2100,9 @@ static int narrow_write_error(struct r10bio *r10_bio, int i)
 static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 {
 	int slot = r10_bio->read_slot;
-	int mirror = r10_bio->devs[slot].devnum;
 	struct bio *bio;
 	struct r10conf *conf = mddev->private;
-	struct md_rdev *rdev;
+	struct md_rdev *rdev = r10_bio->devs[slot].rdev;
 	char b[BDEVNAME_SIZE];
 	unsigned long do_sync;
 	int max_sectors;
@@ -2111,7 +2120,7 @@ static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 		fix_read_error(conf, mddev, r10_bio);
 		unfreeze_array(conf);
 	}
-	rdev_dec_pending(conf->mirrors[mirror].rdev, mddev);
+	rdev_dec_pending(rdev, mddev);
 
 	bio = r10_bio->devs[slot].bio;
 	bdevname(bio->bi_bdev, b);
@@ -2146,6 +2155,7 @@ read_more:
 		    r10_bio->sector - bio->bi_sector,
 		    max_sectors);
 	r10_bio->devs[slot].bio = bio;
+	r10_bio->devs[slot].rdev = rdev;
 	bio->bi_sector = r10_bio->devs[slot].addr
 		+ rdev->data_offset;
 	bio->bi_bdev = rdev->bdev;



  parent reply	other threads:[~2011-11-02  5:25 UTC|newest]

Thread overview: 18+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-11-02  5:25 [md PATCH 00/17] hot-replace support for RAID1 and RAID10 NeilBrown
2011-11-02  5:25 ` [md PATCH 01/17] md/raid10: prepare data structures for handling replacement NeilBrown
2011-11-02  5:25 ` NeilBrown [this message]
2011-11-02  5:25 ` [md PATCH 02/17] md/raid10: change read_balance to return an rdev NeilBrown
2011-11-02  5:25 ` [md PATCH 04/17] md/raid10: allow removal of failed replacement devices NeilBrown
2011-11-02  5:25 ` [md PATCH 13/17] md/raid1: Allow a failed replacement device to be removed NeilBrown
2011-11-02  5:25 ` [md PATCH 10/17] md/raid10: If there is a spare and a replaceable device, start replacement NeilBrown
2011-11-02  5:25 ` [md PATCH 06/17] md/raid10: Handle replacement devices during resync NeilBrown
2011-11-02  5:25 ` [md PATCH 15/17] md/raid1: recognise replacements when assembling arrays NeilBrown
2011-11-02  5:25 ` [md PATCH 12/17] md/raid1: Allocate spare to store replacement devices and their bios NeilBrown
2011-11-02  5:25 ` [md PATCH 07/17] md/raid10: handle recovery of replacement devices NeilBrown
2011-11-02  5:25 ` [md PATCH 14/17] md/raid1: handle activation of replacement device when recovery completes NeilBrown
2011-11-02  5:25 ` [md PATCH 09/17] md/raid10: recognise replacements when assembling array NeilBrown
2011-11-02  5:25 ` [md PATCH 11/17] md/raid1: Replace use of mddev->raid_disks with conf->raid_disks NeilBrown
2011-11-02  5:25 ` [md PATCH 08/17] md/raid10: Allow replacement device to be replace old drive NeilBrown
2011-11-02  5:25 ` [md PATCH 05/17] md/raid10: writes should get directed to replacement as well as original NeilBrown
2011-11-02  5:25 ` [md PATCH 17/17] md/raid1: Mark device replaceable when we see a write error NeilBrown
2011-11-02  5:25 ` [md PATCH 16/17] md/raid1: If there is a spare and a replaceable device, start replacement NeilBrown

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20111102052543.17566.52804.stgit@notabene.brown \
    --to=neilb@suse.de \
    --cc=linux-raid@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).