[PATCH] md: raid1 support for bitmap intent logging
authorNeilBrown <neilb@cse.unsw.edu.au>
Wed, 22 Jun 2005 00:17:23 +0000 (17:17 -0700)
committerLinus Torvalds <torvalds@ppc970.osdl.org>
Wed, 22 Jun 2005 02:07:46 +0000 (19:07 -0700)
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
drivers/md/raid1.c
include/linux/raid/raid1.h

index 3c5c916cb09ed9af10833bfde0854b5fe4951669..0fd4c3bfc851e41f7e84f1546f98f15d756ce749 100644 (file)
  * Fixes to reconstruction by Jakob Ã˜stergaard" <jakob@ostenfeld.dk>
  * Various fixes by Neil Brown <neilb@cse.unsw.edu.au>
  *
+ * Changes by Peter T. Breuer <ptb@it.uc3m.es> 31/1/2003 to support
+ * bitmapped intelligence in resync:
+ *
+ *      - bitmap marked during normal i/o
+ *      - bitmap used to skip nondirty blocks during sync
+ *
+ * Additions to bitmap code, (C) 2003-2004 Paul Clements, SteelEye Technology:
+ * - persistent bitmap code
+ *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
  * the Free Software Foundation; either version 2, or (at your option)
  * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#include "dm-bio-list.h"
 #include <linux/raid/raid1.h>
+#include <linux/raid/bitmap.h>
+
+#define DEBUG 0
+#if DEBUG
+#define PRINTK(x...) printk(x)
+#else
+#define PRINTK(x...)
+#endif
 
 /*
  * Number of guaranteed r1bios in case of extreme VM load:
@@ -287,9 +305,11 @@ static int raid1_end_write_request(struct bio *bio, unsigned int bytes_done, int
        /*
         * this branch is our 'one mirror IO has finished' event handler:
         */
-       if (!uptodate)
+       if (!uptodate) {
                md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
-       else
+               /* an I/O failed, we can't clear the bitmap */
+               set_bit(R1BIO_Degraded, &r1_bio->state);
+       } else
                /*
                 * Set R1BIO_Uptodate in our master bio, so that
                 * we will return a good error code for to the higher
@@ -309,6 +329,10 @@ static int raid1_end_write_request(struct bio *bio, unsigned int bytes_done, int
         * already.
         */
        if (atomic_dec_and_test(&r1_bio->remaining)) {
+               /* clear the bitmap if all writes complete successfully */
+               bitmap_endwrite(r1_bio->mddev->bitmap, r1_bio->sector,
+                               r1_bio->sectors,
+                               !test_bit(R1BIO_Degraded, &r1_bio->state));
                md_write_end(r1_bio->mddev);
                raid_end_bio_io(r1_bio);
        }
@@ -458,7 +482,10 @@ static void unplug_slaves(mddev_t *mddev)
 
 static void raid1_unplug(request_queue_t *q)
 {
-       unplug_slaves(q->queuedata);
+       mddev_t *mddev = q->queuedata;
+
+       unplug_slaves(mddev);
+       md_wakeup_thread(mddev->thread);
 }
 
 static int raid1_issue_flush(request_queue_t *q, struct gendisk *disk,
@@ -501,16 +528,16 @@ static void device_barrier(conf_t *conf, sector_t sect)
 {
        spin_lock_irq(&conf->resync_lock);
        wait_event_lock_irq(conf->wait_idle, !waitqueue_active(&conf->wait_resume),
-                           conf->resync_lock, unplug_slaves(conf->mddev));
+                           conf->resync_lock, raid1_unplug(conf->mddev->queue));
        
        if (!conf->barrier++) {
                wait_event_lock_irq(conf->wait_idle, !conf->nr_pending,
-                                   conf->resync_lock, unplug_slaves(conf->mddev));
+                                   conf->resync_lock, raid1_unplug(conf->mddev->queue));
                if (conf->nr_pending)
                        BUG();
        }
        wait_event_lock_irq(conf->wait_resume, conf->barrier < RESYNC_DEPTH,
-                           conf->resync_lock, unplug_slaves(conf->mddev));
+                           conf->resync_lock, raid1_unplug(conf->mddev->queue));
        conf->next_resync = sect;
        spin_unlock_irq(&conf->resync_lock);
 }
@@ -522,8 +549,12 @@ static int make_request(request_queue_t *q, struct bio * bio)
        mirror_info_t *mirror;
        r1bio_t *r1_bio;
        struct bio *read_bio;
-       int i, disks;
+       int i, targets = 0, disks;
        mdk_rdev_t *rdev;
+       struct bitmap *bitmap = mddev->bitmap;
+       unsigned long flags;
+       struct bio_list bl;
+
 
        /*
         * Register the new request and wait if the reconstruction
@@ -554,7 +585,7 @@ static int make_request(request_queue_t *q, struct bio * bio)
 
        r1_bio->master_bio = bio;
        r1_bio->sectors = bio->bi_size >> 9;
-
+       r1_bio->state = 0;
        r1_bio->mddev = mddev;
        r1_bio->sector = bio->bi_sector;
 
@@ -597,6 +628,13 @@ static int make_request(request_queue_t *q, struct bio * bio)
         * bios[x] to bio
         */
        disks = conf->raid_disks;
+#if 0
+       { static int first=1;
+       if (first) printk("First Write sector %llu disks %d\n",
+                         (unsigned long long)r1_bio->sector, disks);
+       first = 0;
+       }
+#endif
        rcu_read_lock();
        for (i = 0;  i < disks; i++) {
                if ((rdev=conf->mirrors[i].rdev) != NULL &&
@@ -607,13 +645,21 @@ static int make_request(request_queue_t *q, struct bio * bio)
                                r1_bio->bios[i] = NULL;
                        } else
                                r1_bio->bios[i] = bio;
+                       targets++;
                } else
                        r1_bio->bios[i] = NULL;
        }
        rcu_read_unlock();
 
-       atomic_set(&r1_bio->remaining, 1);
+       if (targets < conf->raid_disks) {
+               /* array is degraded, we will not clear the bitmap
+                * on I/O completion (see raid1_end_write_request) */
+               set_bit(R1BIO_Degraded, &r1_bio->state);
+       }
+
+       atomic_set(&r1_bio->remaining, 0);
 
+       bio_list_init(&bl);
        for (i = 0; i < disks; i++) {
                struct bio *mbio;
                if (!r1_bio->bios[i])
@@ -629,14 +675,23 @@ static int make_request(request_queue_t *q, struct bio * bio)
                mbio->bi_private = r1_bio;
 
                atomic_inc(&r1_bio->remaining);
-               generic_make_request(mbio);
-       }
 
-       if (atomic_dec_and_test(&r1_bio->remaining)) {
-               md_write_end(mddev);
-               raid_end_bio_io(r1_bio);
+               bio_list_add(&bl, mbio);
        }
 
+       bitmap_startwrite(bitmap, bio->bi_sector, r1_bio->sectors);
+       spin_lock_irqsave(&conf->device_lock, flags);
+       bio_list_merge(&conf->pending_bio_list, &bl);
+       bio_list_init(&bl);
+
+       blk_plug_device(mddev->queue);
+       spin_unlock_irqrestore(&conf->device_lock, flags);
+
+#if 0
+       while ((bio = bio_list_pop(&bl)) != NULL)
+               generic_make_request(bio);
+#endif
+
        return 0;
 }
 
@@ -716,7 +771,7 @@ static void close_sync(conf_t *conf)
 {
        spin_lock_irq(&conf->resync_lock);
        wait_event_lock_irq(conf->wait_resume, !conf->barrier,
-                           conf->resync_lock,  unplug_slaves(conf->mddev));
+                           conf->resync_lock,  raid1_unplug(conf->mddev->queue));
        spin_unlock_irq(&conf->resync_lock);
 
        if (conf->barrier) BUG();
@@ -830,10 +885,11 @@ static int end_sync_read(struct bio *bio, unsigned int bytes_done, int error)
         * or re-read if the read failed.
         * We don't do much here, just schedule handling by raid1d
         */
-       if (!uptodate)
+       if (!uptodate) {
                md_error(r1_bio->mddev,
                         conf->mirrors[r1_bio->read_disk].rdev);
-       else
+               set_bit(R1BIO_Degraded, &r1_bio->state);
+       } else
                set_bit(R1BIO_Uptodate, &r1_bio->state);
        rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
        reschedule_retry(r1_bio);
@@ -857,8 +913,10 @@ static int end_sync_write(struct bio *bio, unsigned int bytes_done, int error)
                        mirror = i;
                        break;
                }
-       if (!uptodate)
+       if (!uptodate) {
                md_error(mddev, conf->mirrors[mirror].rdev);
+               set_bit(R1BIO_Degraded, &r1_bio->state);
+       }
        update_head_pos(mirror, r1_bio);
 
        if (atomic_dec_and_test(&r1_bio->remaining)) {
@@ -878,6 +936,9 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
 
        bio = r1_bio->bios[r1_bio->read_disk];
 
+/*
+       if (r1_bio->sector == 0) printk("First sync write startss\n");
+*/
        /*
         * schedule writes
         */
@@ -905,10 +966,12 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
                atomic_inc(&conf->mirrors[i].rdev->nr_pending);
                atomic_inc(&r1_bio->remaining);
                md_sync_acct(conf->mirrors[i].rdev->bdev, wbio->bi_size >> 9);
+
                generic_make_request(wbio);
        }
 
        if (atomic_dec_and_test(&r1_bio->remaining)) {
+               /* if we're here, all write(s) have completed, so clean up */
                md_done_sync(mddev, r1_bio->sectors, 1);
                put_buf(r1_bio);
        }
@@ -937,6 +1000,26 @@ static void raid1d(mddev_t *mddev)
        for (;;) {
                char b[BDEVNAME_SIZE];
                spin_lock_irqsave(&conf->device_lock, flags);
+
+               if (conf->pending_bio_list.head) {
+                       bio = bio_list_get(&conf->pending_bio_list);
+                       blk_remove_plug(mddev->queue);
+                       spin_unlock_irqrestore(&conf->device_lock, flags);
+                       /* flush any pending bitmap writes to disk before proceeding w/ I/O */
+                       if (bitmap_unplug(mddev->bitmap) != 0)
+                               printk("%s: bitmap file write failed!\n", mdname(mddev));
+
+                       while (bio) { /* submit pending writes */
+                               struct bio *next = bio->bi_next;
+                               bio->bi_next = NULL;
+                               generic_make_request(bio);
+                               bio = next;
+                       }
+                       unplug = 1;
+
+                       continue;
+               }
+
                if (list_empty(head))
                        break;
                r1_bio = list_entry(head->prev, r1bio_t, retry_list);
@@ -1020,17 +1103,43 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
        int disk;
        int i;
        int write_targets = 0;
+       int sync_blocks;
 
        if (!conf->r1buf_pool)
+       {
+/*
+               printk("sync start - bitmap %p\n", mddev->bitmap);
+*/
                if (init_resync(conf))
                        return 0;
+       }
 
        max_sector = mddev->size << 1;
        if (sector_nr >= max_sector) {
+               /* If we aborted, we need to abort the
+                * sync on the 'current' bitmap chunk (there will
+                * only be one in raid1 resync.
+                * We can find the current addess in mddev->curr_resync
+                */
+               if (!conf->fullsync) {
+                       if (mddev->curr_resync < max_sector)
+                               bitmap_end_sync(mddev->bitmap,
+                                               mddev->curr_resync,
+                                               &sync_blocks, 1);
+                       bitmap_close_sync(mddev->bitmap);
+               }
+               if (mddev->curr_resync >= max_sector)
+                       conf->fullsync = 0;
                close_sync(conf);
                return 0;
        }
 
+       if (!conf->fullsync &&
+           !bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks)) {
+               /* We can skip this block, and probably several more */
+               *skipped = 1;
+               return sync_blocks;
+       }
        /*
         * If there is non-resync activity waiting for us then
         * put in a delay to throttle resync.
@@ -1069,6 +1178,7 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 
        r1_bio->mddev = mddev;
        r1_bio->sector = sector_nr;
+       r1_bio->state = 0;
        set_bit(R1BIO_IsSync, &r1_bio->state);
        r1_bio->read_disk = disk;
 
@@ -1103,6 +1213,11 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
                bio->bi_bdev = conf->mirrors[i].rdev->bdev;
                bio->bi_private = r1_bio;
        }
+
+       if (write_targets + 1 < conf->raid_disks)
+               /* array degraded, can't clear bitmap */
+               set_bit(R1BIO_Degraded, &r1_bio->state);
+
        if (write_targets == 0) {
                /* There is nowhere to write, so all non-sync
                 * drives must be failed - so we are finished
@@ -1122,6 +1237,14 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
                        len = (max_sector - sector_nr) << 9;
                if (len == 0)
                        break;
+               if (!conf->fullsync && sync_blocks == 0)
+                       if (!bitmap_start_sync(mddev->bitmap,
+                                              sector_nr, &sync_blocks))
+                               break;
+               if (sync_blocks < (PAGE_SIZE>>9))
+                       BUG();
+               if (len > (sync_blocks<<9)) len = sync_blocks<<9;
+
                for (i=0 ; i < conf->raid_disks; i++) {
                        bio = r1_bio->bios[i];
                        if (bio->bi_end_io) {
@@ -1144,6 +1267,7 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
                }
                nr_sectors += len>>9;
                sector_nr += len>>9;
+               sync_blocks -= (len>>9);
        } while (r1_bio->bios[disk]->bi_vcnt < RESYNC_PAGES);
  bio_full:
        bio = r1_bio->bios[disk];
@@ -1232,6 +1356,9 @@ static int run(mddev_t *mddev)
        init_waitqueue_head(&conf->wait_idle);
        init_waitqueue_head(&conf->wait_resume);
 
+       bio_list_init(&conf->pending_bio_list);
+       bio_list_init(&conf->flushing_bio_list);
+
        if (!conf->working_disks) {
                printk(KERN_ERR "raid1: no operational mirrors for %s\n",
                        mdname(mddev));
@@ -1260,16 +1387,15 @@ static int run(mddev_t *mddev)
        conf->last_used = j;
 
 
-
-       {
-               mddev->thread = md_register_thread(raid1d, mddev, "%s_raid1");
-               if (!mddev->thread) {
-                       printk(KERN_ERR 
-                               "raid1: couldn't allocate thread for %s\n", 
-                               mdname(mddev));
-                       goto out_free_conf;
-               }
+       mddev->thread = md_register_thread(raid1d, mddev, "%s_raid1");
+       if (!mddev->thread) {
+               printk(KERN_ERR
+                      "raid1: couldn't allocate thread for %s\n",
+                      mdname(mddev));
+               goto out_free_conf;
        }
+       if (mddev->bitmap) mddev->thread->timeout = mddev->bitmap->daemon_sleep * HZ;
+
        printk(KERN_INFO 
                "raid1: raid set %s active with %d out of %d mirrors\n",
                mdname(mddev), mddev->raid_disks - mddev->degraded, 
@@ -1394,7 +1520,7 @@ static int raid1_reshape(mddev_t *mddev, int raid_disks)
        spin_lock_irq(&conf->resync_lock);
        conf->barrier++;
        wait_event_lock_irq(conf->wait_idle, !conf->nr_pending,
-                           conf->resync_lock, unplug_slaves(mddev));
+                           conf->resync_lock, raid1_unplug(mddev->queue));
        spin_unlock_irq(&conf->resync_lock);
 
        /* ok, everything is stopped */
index abbfdd9afe1ec615e288fccced1f5a41b0dafbed..9d93cf12e8901d7cb34f8013d290eb4e8e4d1f1f 100644 (file)
@@ -36,12 +36,21 @@ struct r1_private_data_s {
        spinlock_t              device_lock;
 
        struct list_head        retry_list;
+       /* queue pending writes and submit them on unplug */
+       struct bio_list         pending_bio_list;
+       /* queue of writes that have been unplugged */
+       struct bio_list         flushing_bio_list;
+
        /* for use when syncing mirrors: */
 
        spinlock_t              resync_lock;
-       int nr_pending;
-       int barrier;
+       int                     nr_pending;
+       int                     barrier;
        sector_t                next_resync;
+       int                     fullsync;  /* set to 1 if a full sync is needed,
+                                           * (fresh device added).
+                                           * Cleared when a sync completes.
+                                           */
 
        wait_queue_head_t       wait_idle;
        wait_queue_head_t       wait_resume;
@@ -85,14 +94,17 @@ struct r1bio_s {
        int                     read_disk;
 
        struct list_head        retry_list;
+       struct bitmap_update    *bitmap_update;
        /*
         * if the IO is in WRITE direction, then multiple bios are used.
         * We choose the number when they are allocated.
         */
        struct bio              *bios[0];
+       /* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
 };
 
 /* bits for r1bio.state */
 #define        R1BIO_Uptodate  0
 #define        R1BIO_IsSync    1
+#define        R1BIO_Degraded  2
 #endif