drbd: new on-disk activity log transaction format

Use a new on-disk transaction format for the activity log, which allows
for multiple changes to the active set per transaction.

Using 4k transaction blocks, we can now get rid of the work-around code
to deal with devices not supporting 512 byte logical block size.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>

drivers/block/drbd/drbd_int.h

@@ -1069,7 +1069,6 @@ struct drbd_conf {
 	atomic_t pp_in_use_by_net;	/* sendpage()d, still referenced by tcp */
 	wait_queue_head_t ee_wait;
 	struct page *md_io_page;	/* one page buffer for md_io */
-	struct page *md_io_tmpp;	/* for logical_block_size != 512 */
 	struct mutex md_io_mutex;	/* protects the md_io_buffer */
 	spinlock_t al_lock;
 	wait_queue_head_t al_wait;
@@ -1259,22 +1258,39 @@ extern void drbd_ldev_destroy(struct drbd_conf *mdev);
    * either at the end of the backing device
    * or on a separate meta data device. */
 
-#define MD_RESERVED_SECT (128LU << 11)  /* 128 MB, unit sectors */
 /* The following numbers are sectors */
+/* Allows up to about 3.8TB, so if you want more,
+ * you need to use the "flexible" meta data format. */
+#define MD_RESERVED_SECT (128LU << 11)  /* 128 MB, unit sectors */
 #define MD_AL_OFFSET	8    /* 8 Sectors after start of meta area */
-#define MD_AL_MAX_SIZE 64   /* = 32 kb LOG  ~ 3776 extents ~ 14 GB Storage */
-/* Allows up to about 3.8TB */
-#define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_MAX_SIZE)
+#define MD_AL_SECTORS	64   /* = 32 kB on disk activity log ring buffer */
+#define MD_BM_OFFSET (MD_AL_OFFSET + MD_AL_SECTORS)
 
-/* Since the smalles IO unit is usually 512 byte */
-#define MD_SECTOR_SHIFT	 9
-#define MD_SECTOR_SIZE	 (1<<MD_SECTOR_SHIFT)
+/* we do all meta data IO in 4k blocks */
+#define MD_BLOCK_SHIFT	12
+#define MD_BLOCK_SIZE	(1<<MD_BLOCK_SHIFT)
 
-/* activity log */
-#define AL_EXTENTS_PT ((MD_SECTOR_SIZE-12)/8-1) /* 61 ; Extents per 512B sector */
-#define AL_EXTENT_SHIFT 22		 /* One extent represents 4M Storage */
+/* One activity log extent represents 4M of storage */
+#define AL_EXTENT_SHIFT 22
 #define AL_EXTENT_SIZE (1<<AL_EXTENT_SHIFT)
 
+/* We could make these currently hardcoded constants configurable
+ * variables at create-md time (or even re-configurable at runtime?).
+ * Which will require some more changes to the DRBD "super block"
+ * and attach code.
+ *
+ * updates per transaction:
+ *   This many changes to the active set can be logged with one transaction.
+ *   This number is arbitrary.
+ * context per transaction:
+ *   This many context extent numbers are logged with each transaction.
+ *   This number is resulting from the transaction block size (4k), the layout
+ *   of the transaction header, and the number of updates per transaction.
+ *   See drbd_actlog.c:struct al_transaction_on_disk
+ * */
+#define AL_UPDATES_PER_TRANSACTION	 64	// arbitrary
+#define AL_CONTEXT_PER_TRANSACTION	919	// (4096 - 36 - 6*64)/4
+
 #if BITS_PER_LONG == 32
 #define LN2_BPL 5
 #define cpu_to_lel(A) cpu_to_le32(A)

drivers/block/drbd/drbd_main.c

@@ -2841,10 +2841,6 @@ void drbd_ldev_destroy(struct drbd_conf *mdev)
 		drbd_free_bc(mdev->ldev);
 		mdev->ldev = NULL;);
 
-	if (mdev->md_io_tmpp) {
-		__free_page(mdev->md_io_tmpp);
-		mdev->md_io_tmpp = NULL;
-	}
 	clear_bit(GO_DISKLESS, &mdev->flags);
 }
 

drivers/block/drbd/drbd_nl.c

@@ -527,7 +527,7 @@ static void drbd_md_set_sector_offsets(struct drbd_conf *mdev,
 	case DRBD_MD_INDEX_FLEX_INT:
 		bdev->md.md_offset = drbd_md_ss__(mdev, bdev);
 		/* al size is still fixed */
-		bdev->md.al_offset = -MD_AL_MAX_SIZE;
+		bdev->md.al_offset = -MD_AL_SECTORS;
 		/* we need (slightly less than) ~ this much bitmap sectors: */
 		md_size_sect = drbd_get_capacity(bdev->backing_bdev);
 		md_size_sect = ALIGN(md_size_sect, BM_SECT_PER_EXT);
@@ -751,8 +751,8 @@ static int drbd_check_al_size(struct drbd_conf *mdev)
 	unsigned int in_use;
 	int i;
 
-	if (!expect(mdev->sync_conf.al_extents >= 7))
-		mdev->sync_conf.al_extents = 127;
+	if (!expect(mdev->sync_conf.al_extents >= DRBD_AL_EXTENTS_MIN))
+		mdev->sync_conf.al_extents = DRBD_AL_EXTENTS_MIN;
 
 	if (mdev->act_log &&
 	    mdev->act_log->nr_elements == mdev->sync_conf.al_extents)
@@ -760,7 +760,7 @@ static int drbd_check_al_size(struct drbd_conf *mdev)
 
 	in_use = 0;
 	t = mdev->act_log;
-	n = lc_create("act_log", drbd_al_ext_cache, 1,
+	n = lc_create("act_log", drbd_al_ext_cache, AL_UPDATES_PER_TRANSACTION,
 		mdev->sync_conf.al_extents, sizeof(struct lc_element), 0);
 
 	if (n == NULL) {
@@ -932,7 +932,6 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
 	union drbd_state ns, os;
 	enum drbd_state_rv rv;
 	int cp_discovered = 0;
-	int logical_block_size;
 
 	drbd_reconfig_start(mdev);
 
@@ -1087,25 +1086,6 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
 
 	drbd_md_set_sector_offsets(mdev, nbc);
 
-	/* allocate a second IO page if logical_block_size != 512 */
-	logical_block_size = bdev_logical_block_size(nbc->md_bdev);
-	if (logical_block_size == 0)
-		logical_block_size = MD_SECTOR_SIZE;
-
-	if (logical_block_size != MD_SECTOR_SIZE) {
-		if (!mdev->md_io_tmpp) {
-			struct page *page = alloc_page(GFP_NOIO);
-			if (!page)
-				goto force_diskless_dec;
-
-			dev_warn(DEV, "Meta data's bdev logical_block_size = %d != %d\n",
-			     logical_block_size, MD_SECTOR_SIZE);
-			dev_warn(DEV, "Workaround engaged (has performance impact).\n");
-
-			mdev->md_io_tmpp = page;
-		}
-	}
-
 	if (!mdev->bitmap) {
 		if (drbd_bm_init(mdev)) {
 			retcode = ERR_NOMEM;
@@ -1804,14 +1784,12 @@ static int drbd_nl_syncer_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *n
 
 	if (!expect(sc.rate >= 1))
 		sc.rate = 1;
-	if (!expect(sc.al_extents >= 7))
-		sc.al_extents = 127; /* arbitrary minimum */
-#define AL_MAX ((MD_AL_MAX_SIZE-1) * AL_EXTENTS_PT)
-	if (sc.al_extents > AL_MAX) {
-		dev_err(DEV, "sc.al_extents > %d\n", AL_MAX);
-		sc.al_extents = AL_MAX;
-	}
-#undef AL_MAX
+
+	/* clip to allowed range */
+	if (!expect(sc.al_extents >= DRBD_AL_EXTENTS_MIN))
+		sc.al_extents = DRBD_AL_EXTENTS_MIN;
+	if (!expect(sc.al_extents <= DRBD_AL_EXTENTS_MAX))
+		sc.al_extents = DRBD_AL_EXTENTS_MAX;
 
 	/* to avoid spurious errors when configuring minors before configuring
 	 * the minors they depend on: if necessary, first create the minor we

include/linux/drbd.h

@@ -336,6 +336,10 @@ enum drbd_timeout_flag {
 #define DRBD_MAGIC 0x83740267
 #define DRBD_MAGIC_BIG 0x835a
 
+/* how I came up with this magic?
+ * base64 decode "actlog==" ;) */
+#define DRBD_AL_MAGIC 0x69cb65a2
+
 /* these are of type "int" */
 #define DRBD_MD_INDEX_INTERNAL -1
 #define DRBD_MD_INDEX_FLEX_EXT -2

include/linux/drbd_limits.h

@@ -102,10 +102,12 @@
 #define DRBD_RATE_DEF 250  /* kb/second */
 
   /* less than 7 would hit performance unnecessarily.
-   * 3833 is the largest prime that still does fit
-   * into 64 sectors of activity log */
+   * 919 slots context information per transaction,
+   * 32k activity log, 4k transaction size,
+   * one transaction in flight:
+   * 919 * 7 = 6433 */
 #define DRBD_AL_EXTENTS_MIN  7
-#define DRBD_AL_EXTENTS_MAX  3833
+#define DRBD_AL_EXTENTS_MAX  6433
 #define DRBD_AL_EXTENTS_DEF  127
 
 #define DRBD_AFTER_MIN  -1