]> git.kernelconcepts.de Git - karo-tx-linux.git/commitdiff
dm thin: defer whole cells rather than individual bios
authorJoe Thornber <ejt@redhat.com>
Fri, 10 Oct 2014 12:43:14 +0000 (13:43 +0100)
committerMike Snitzer <snitzer@redhat.com>
Mon, 10 Nov 2014 20:25:28 +0000 (15:25 -0500)
This avoids dropping the cell, so increases the probability that other
bios will collect within the cell, rather than being passed individually
to the worker.

Also add required process_cell and process_discard_cell error handling
wrappers and set associated pool-mode function pointers accordingly.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
drivers/md/dm-bio-prison.h
drivers/md/dm-thin.c

index 997a43960e77bca64520767a6b1275dd43bcb0f4..c0cddb118582ac4b522b15a25e6f724db5746b42 100644 (file)
@@ -35,6 +35,7 @@ struct dm_cell_key {
  * themselves.
  */
 struct dm_bio_prison_cell {
+       struct list_head user_list;     /* for client use */
        struct rb_node node;
 
        struct dm_cell_key key;
index 52562710f6a077180153337aed31f2a0eafb3a1b..912d7f4d89d1c922c32d08d6c882110644f7d7f8 100644 (file)
@@ -202,6 +202,7 @@ struct pool_features {
 
 struct thin_c;
 typedef void (*process_bio_fn)(struct thin_c *tc, struct bio *bio);
+typedef void (*process_cell_fn)(struct thin_c *tc, struct dm_bio_prison_cell *cell);
 typedef void (*process_mapping_fn)(struct dm_thin_new_mapping *m);
 
 struct pool {
@@ -246,6 +247,9 @@ struct pool {
        process_bio_fn process_bio;
        process_bio_fn process_discard;
 
+       process_cell_fn process_cell;
+       process_cell_fn process_discard_cell;
+
        process_mapping_fn process_prepared_mapping;
        process_mapping_fn process_prepared_discard;
 };
@@ -282,6 +286,7 @@ struct thin_c {
        struct dm_thin_device *td;
        bool requeue_mode:1;
        spinlock_t lock;
+       struct list_head deferred_cells;
        struct bio_list deferred_bio_list;
        struct bio_list retry_on_resume_list;
        struct rb_root sort_bio_list; /* sorted list of deferred bios */
@@ -346,19 +351,6 @@ static void cell_release_no_holder(struct pool *pool,
        dm_bio_prison_free_cell(pool->prison, cell);
 }
 
-static void cell_defer_no_holder_no_free(struct thin_c *tc,
-                                        struct dm_bio_prison_cell *cell)
-{
-       struct pool *pool = tc->pool;
-       unsigned long flags;
-
-       spin_lock_irqsave(&tc->lock, flags);
-       dm_cell_release_no_holder(pool->prison, cell, &tc->deferred_bio_list);
-       spin_unlock_irqrestore(&tc->lock, flags);
-
-       wake_worker(pool);
-}
-
 static void cell_error_with_code(struct pool *pool,
                                 struct dm_bio_prison_cell *cell, int error_code)
 {
@@ -371,6 +363,16 @@ static void cell_error(struct pool *pool, struct dm_bio_prison_cell *cell)
        cell_error_with_code(pool, cell, -EIO);
 }
 
+static void cell_success(struct pool *pool, struct dm_bio_prison_cell *cell)
+{
+       cell_error_with_code(pool, cell, 0);
+}
+
+static void cell_requeue(struct pool *pool, struct dm_bio_prison_cell *cell)
+{
+       cell_error_with_code(pool, cell, DM_ENDIO_REQUEUE);
+}
+
 /*----------------------------------------------------------------*/
 
 /*
@@ -458,10 +460,28 @@ static void requeue_bio_list(struct thin_c *tc, struct bio_list *master)
                bio_endio(bio, DM_ENDIO_REQUEUE);
 }
 
+static void requeue_deferred_cells(struct thin_c *tc)
+{
+       struct pool *pool = tc->pool;
+       unsigned long flags;
+       struct list_head cells;
+       struct dm_bio_prison_cell *cell, *tmp;
+
+       INIT_LIST_HEAD(&cells);
+
+       spin_lock_irqsave(&tc->lock, flags);
+       list_splice_init(&tc->deferred_cells, &cells);
+       spin_unlock_irqrestore(&tc->lock, flags);
+
+       list_for_each_entry_safe(cell, tmp, &cells, user_list)
+               cell_requeue(pool, cell);
+}
+
 static void requeue_io(struct thin_c *tc)
 {
        requeue_bio_list(tc, &tc->deferred_bio_list);
        requeue_bio_list(tc, &tc->retry_on_resume_list);
+       requeue_deferred_cells(tc);
 }
 
 static void error_thin_retry_list(struct thin_c *tc)
@@ -706,6 +726,28 @@ static void cell_defer_no_holder(struct thin_c *tc, struct dm_bio_prison_cell *c
        wake_worker(pool);
 }
 
+static void thin_defer_bio(struct thin_c *tc, struct bio *bio);
+
+static void inc_remap_and_issue_cell(struct thin_c *tc,
+                                    struct dm_bio_prison_cell *cell,
+                                    dm_block_t block)
+{
+       struct bio *bio;
+       struct bio_list bios;
+
+       bio_list_init(&bios);
+       cell_release_no_holder(tc->pool, cell, &bios);
+
+       while ((bio = bio_list_pop(&bios))) {
+               if (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA))
+                       thin_defer_bio(tc, bio);
+               else {
+                       inc_all_io_entry(tc->pool, bio);
+                       remap_and_issue(tc, bio, block);
+               }
+       }
+}
+
 static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
 {
        if (m->bio) {
@@ -1193,19 +1235,21 @@ static void retry_bios_on_resume(struct pool *pool, struct dm_bio_prison_cell *c
                        retry_on_resume(bio);
 }
 
-static void process_discard(struct thin_c *tc, struct bio *bio)
+static void process_discard_cell(struct thin_c *tc, struct dm_bio_prison_cell *cell)
 {
        int r;
+       struct bio *bio = cell->holder;
        struct pool *pool = tc->pool;
-       struct dm_bio_prison_cell *cell, *cell2;
-       struct dm_cell_key key, key2;
+       struct dm_bio_prison_cell *cell2;
+       struct dm_cell_key key2;
        dm_block_t block = get_bio_block(tc, bio);
        struct dm_thin_lookup_result lookup_result;
        struct dm_thin_new_mapping *m;
 
-       build_virtual_key(tc->td, block, &key);
-       if (bio_detain(tc->pool, &key, bio, &cell))
+       if (tc->requeue_mode) {
+               cell_requeue(pool, cell);
                return;
+       }
 
        r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
        switch (r) {
@@ -1273,6 +1317,19 @@ static void process_discard(struct thin_c *tc, struct bio *bio)
        }
 }
 
+static void process_discard_bio(struct thin_c *tc, struct bio *bio)
+{
+       struct dm_bio_prison_cell *cell;
+       struct dm_cell_key key;
+       dm_block_t block = get_bio_block(tc, bio);
+
+       build_virtual_key(tc->td, block, &key);
+       if (bio_detain(tc->pool, &key, bio, &cell))
+               return;
+
+       process_discard_cell(tc, cell);
+}
+
 static void break_sharing(struct thin_c *tc, struct bio *bio, dm_block_t block,
                          struct dm_cell_key *key,
                          struct dm_thin_lookup_result *lookup_result,
@@ -1379,34 +1436,30 @@ static void provision_block(struct thin_c *tc, struct bio *bio, dm_block_t block
        }
 }
 
-static void process_bio(struct thin_c *tc, struct bio *bio)
+static void process_cell(struct thin_c *tc, struct dm_bio_prison_cell *cell)
 {
        int r;
        struct pool *pool = tc->pool;
+       struct bio *bio = cell->holder;
        dm_block_t block = get_bio_block(tc, bio);
-       struct dm_bio_prison_cell *cell;
-       struct dm_cell_key key;
        struct dm_thin_lookup_result lookup_result;
 
-       /*
-        * If cell is already occupied, then the block is already
-        * being provisioned so we have nothing further to do here.
-        */
-       build_virtual_key(tc->td, block, &key);
-       if (bio_detain(pool, &key, bio, &cell))
+       if (tc->requeue_mode) {
+               cell_requeue(pool, cell);
                return;
+       }
 
        r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
        switch (r) {
        case 0:
                if (lookup_result.shared) {
                        process_shared_bio(tc, bio, block, &lookup_result);
+                       // FIXME: we can't remap because we're waiting on a commit
                        cell_defer_no_holder(tc, cell); /* FIXME: pass this cell into process_shared? */
                } else {
                        inc_all_io_entry(pool, bio);
-                       cell_defer_no_holder(tc, cell);
-
                        remap_and_issue(tc, bio, lookup_result.block);
+                       inc_remap_and_issue_cell(tc, cell, lookup_result.block);
                }
                break;
 
@@ -1440,7 +1493,26 @@ static void process_bio(struct thin_c *tc, struct bio *bio)
        }
 }
 
-static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
+static void process_bio(struct thin_c *tc, struct bio *bio)
+{
+       struct pool *pool = tc->pool;
+       dm_block_t block = get_bio_block(tc, bio);
+       struct dm_bio_prison_cell *cell;
+       struct dm_cell_key key;
+
+       /*
+        * If cell is already occupied, then the block is already
+        * being provisioned so we have nothing further to do here.
+        */
+       build_virtual_key(tc->td, block, &key);
+       if (bio_detain(pool, &key, bio, &cell))
+               return;
+
+       process_cell(tc, cell);
+}
+
+static void __process_bio_read_only(struct thin_c *tc, struct bio *bio,
+                                   struct dm_bio_prison_cell *cell)
 {
        int r;
        int rw = bio_data_dir(bio);
@@ -1450,15 +1522,21 @@ static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
        r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
        switch (r) {
        case 0:
-               if (lookup_result.shared && (rw == WRITE) && bio->bi_iter.bi_size)
+               if (lookup_result.shared && (rw == WRITE) && bio->bi_iter.bi_size) {
                        handle_unserviceable_bio(tc->pool, bio);
-               else {
+                       if (cell)
+                               cell_defer_no_holder(tc, cell);
+               } else {
                        inc_all_io_entry(tc->pool, bio);
                        remap_and_issue(tc, bio, lookup_result.block);
+                       if (cell)
+                               inc_remap_and_issue_cell(tc, cell, lookup_result.block);
                }
                break;
 
        case -ENODATA:
+               if (cell)
+                       cell_defer_no_holder(tc, cell);
                if (rw != READ) {
                        handle_unserviceable_bio(tc->pool, bio);
                        break;
@@ -1477,11 +1555,23 @@ static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
        default:
                DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
                            __func__, r);
+               if (cell)
+                       cell_defer_no_holder(tc, cell);
                bio_io_error(bio);
                break;
        }
 }
 
+static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
+{
+       __process_bio_read_only(tc, bio, NULL);
+}
+
+static void process_cell_read_only(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+       __process_bio_read_only(tc, cell->holder, cell);
+}
+
 static void process_bio_success(struct thin_c *tc, struct bio *bio)
 {
        bio_endio(bio, 0);
@@ -1492,6 +1582,16 @@ static void process_bio_fail(struct thin_c *tc, struct bio *bio)
        bio_io_error(bio);
 }
 
+static void process_cell_success(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+       cell_success(tc->pool, cell);
+}
+
+static void process_cell_fail(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+       cell_error(tc->pool, cell);
+}
+
 /*
  * FIXME: should we also commit due to size of transaction, measured in
  * metadata blocks?
@@ -1624,6 +1724,45 @@ static void process_thin_deferred_bios(struct thin_c *tc)
        blk_finish_plug(&plug);
 }
 
+static void process_thin_deferred_cells(struct thin_c *tc)
+{
+       struct pool *pool = tc->pool;
+       unsigned long flags;
+       struct list_head cells;
+       struct dm_bio_prison_cell *cell, *tmp;
+
+       INIT_LIST_HEAD(&cells);
+
+       spin_lock_irqsave(&tc->lock, flags);
+       list_splice_init(&tc->deferred_cells, &cells);
+       spin_unlock_irqrestore(&tc->lock, flags);
+
+       if (list_empty(&cells))
+               return;
+
+       list_for_each_entry_safe(cell, tmp, &cells, user_list) {
+               BUG_ON(!cell->holder);
+
+               /*
+                * If we've got no free new_mapping structs, and processing
+                * this bio might require one, we pause until there are some
+                * prepared mappings to process.
+                */
+               if (ensure_next_mapping(pool)) {
+                       spin_lock_irqsave(&tc->lock, flags);
+                       list_add(&cell->user_list, &tc->deferred_cells);
+                       list_splice(&cells, &tc->deferred_cells);
+                       spin_unlock_irqrestore(&tc->lock, flags);
+                       break;
+               }
+
+               if (cell->holder->bi_rw & REQ_DISCARD)
+                       pool->process_discard_cell(tc, cell);
+               else
+                       pool->process_cell(tc, cell);
+       }
+}
+
 static void thin_get(struct thin_c *tc);
 static void thin_put(struct thin_c *tc);
 
@@ -1672,6 +1811,7 @@ static void process_deferred_bios(struct pool *pool)
 
        tc = get_first_thin(pool);
        while (tc) {
+               process_thin_deferred_cells(tc);
                process_thin_deferred_bios(tc);
                tc = get_next_thin(pool, tc);
        }
@@ -1850,6 +1990,8 @@ static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)
                dm_pool_metadata_read_only(pool->pmd);
                pool->process_bio = process_bio_fail;
                pool->process_discard = process_bio_fail;
+               pool->process_cell = process_cell_fail;
+               pool->process_discard_cell = process_cell_fail;
                pool->process_prepared_mapping = process_prepared_mapping_fail;
                pool->process_prepared_discard = process_prepared_discard_fail;
 
@@ -1862,6 +2004,8 @@ static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)
                dm_pool_metadata_read_only(pool->pmd);
                pool->process_bio = process_bio_read_only;
                pool->process_discard = process_bio_success;
+               pool->process_cell = process_cell_read_only;
+               pool->process_discard_cell = process_cell_success;
                pool->process_prepared_mapping = process_prepared_mapping_fail;
                pool->process_prepared_discard = process_prepared_discard_passdown;
 
@@ -1880,7 +2024,9 @@ static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)
                if (old_mode != new_mode)
                        notify_of_pool_mode_change(pool, "out-of-data-space");
                pool->process_bio = process_bio_read_only;
-               pool->process_discard = process_discard;
+               pool->process_discard = process_discard_bio;
+               pool->process_cell = process_cell_read_only;
+               pool->process_discard_cell = process_discard_cell;
                pool->process_prepared_mapping = process_prepared_mapping;
                pool->process_prepared_discard = process_prepared_discard_passdown;
 
@@ -1893,7 +2039,9 @@ static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)
                        notify_of_pool_mode_change(pool, "write");
                dm_pool_metadata_read_write(pool->pmd);
                pool->process_bio = process_bio;
-               pool->process_discard = process_discard;
+               pool->process_discard = process_discard_bio;
+               pool->process_cell = process_cell;
+               pool->process_discard_cell = process_discard_cell;
                pool->process_prepared_mapping = process_prepared_mapping;
                pool->process_prepared_discard = process_prepared_discard;
                break;
@@ -1962,6 +2110,20 @@ static void thin_defer_bio_with_throttle(struct thin_c *tc, struct bio *bio)
        throttle_unlock(&pool->throttle);
 }
 
+static void thin_defer_cell(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+       unsigned long flags;
+       struct pool *pool = tc->pool;
+
+       throttle_lock(&pool->throttle);
+       spin_lock_irqsave(&tc->lock, flags);
+       list_add_tail(&cell->user_list, &tc->deferred_cells);
+       spin_unlock_irqrestore(&tc->lock, flags);
+       throttle_unlock(&pool->throttle);
+
+       wake_worker(pool);
+}
+
 static void thin_hook_bio(struct thin_c *tc, struct bio *bio)
 {
        struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
@@ -1982,8 +2144,7 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
        dm_block_t block = get_bio_block(tc, bio);
        struct dm_thin_device *td = tc->td;
        struct dm_thin_lookup_result result;
-       struct dm_bio_prison_cell cell1, cell2;
-       struct dm_bio_prison_cell *cell_result;
+       struct dm_bio_prison_cell *virt_cell, *data_cell;
        struct dm_cell_key key;
 
        thin_hook_bio(tc, bio);
@@ -2008,7 +2169,7 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
         * there's a race with discard.
         */
        build_virtual_key(tc->td, block, &key);
-       if (dm_bio_detain(tc->pool->prison, &key, bio, &cell1, &cell_result))
+       if (bio_detain(tc->pool, &key, bio, &virt_cell))
                return DM_MAPIO_SUBMITTED;
 
        r = dm_thin_find_block(td, block, 0, &result);
@@ -2033,20 +2194,19 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
                         * More distant ancestors are irrelevant. The
                         * shared flag will be set in their case.
                         */
-                       thin_defer_bio(tc, bio);
-                       cell_defer_no_holder_no_free(tc, &cell1);
+                       thin_defer_cell(tc, virt_cell);
                        return DM_MAPIO_SUBMITTED;
                }
 
                build_data_key(tc->td, result.block, &key);
-               if (dm_bio_detain(tc->pool->prison, &key, bio, &cell2, &cell_result)) {
-                       cell_defer_no_holder_no_free(tc, &cell1);
+               if (bio_detain(tc->pool, &key, bio, &data_cell)) {
+                       cell_defer_no_holder(tc, virt_cell);
                        return DM_MAPIO_SUBMITTED;
                }
 
                inc_all_io_entry(tc->pool, bio);
-               cell_defer_no_holder_no_free(tc, &cell2);
-               cell_defer_no_holder_no_free(tc, &cell1);
+               cell_defer_no_holder(tc, data_cell);
+               cell_defer_no_holder(tc, virt_cell);
 
                remap(tc, bio, result.block);
                return DM_MAPIO_REMAPPED;
@@ -2058,14 +2218,13 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
                         * of doing so.
                         */
                        handle_unserviceable_bio(tc->pool, bio);
-                       cell_defer_no_holder_no_free(tc, &cell1);
+                       cell_defer_no_holder(tc, virt_cell);
                        return DM_MAPIO_SUBMITTED;
                }
                /* fall through */
 
        case -EWOULDBLOCK:
-               thin_defer_bio(tc, bio);
-               cell_defer_no_holder_no_free(tc, &cell1);
+               thin_defer_cell(tc, virt_cell);
                return DM_MAPIO_SUBMITTED;
 
        default:
@@ -2075,7 +2234,7 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
                 * pool is switched to fail-io mode.
                 */
                bio_io_error(bio);
-               cell_defer_no_holder_no_free(tc, &cell1);
+               cell_defer_no_holder(tc, virt_cell);
                return DM_MAPIO_SUBMITTED;
        }
 }
@@ -3394,6 +3553,7 @@ static int thin_ctr(struct dm_target *ti, unsigned argc, char **argv)
                goto out_unlock;
        }
        spin_lock_init(&tc->lock);
+       INIT_LIST_HEAD(&tc->deferred_cells);
        bio_list_init(&tc->deferred_bio_list);
        bio_list_init(&tc->retry_on_resume_list);
        tc->sort_bio_list = RB_ROOT;