dm thin: requeue bios to DM core if no_free_space and in read-only mode

Now that we switch the pool to read-only mode when the data device runs
out of space it causes active writers to get IO errors once we resume
after resizing the data device.

If no_free_space is set, save bios to the 'retry_on_resume_list' and
requeue them on resume (once the data or metadata device may have been
resized).

With this patch the resize_io test passes again (on slower storage):
 dmtest run --suite thin-provisioning -n /resize_io/

Later patches fix some subtle races associated with the pool mode
transitions done as part of the pool's -ENOSPC handling.  These races
are exposed on fast storage (e.g. PCIe SSD).

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Joe Thornber <ejt@redhat.com>

diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c
index 96ce36a1a764216f8768dacd828d802a50dc9d49..53252d2af2496aec75e15a6c6bf4a83bc4572f3c 100644 (file)
--- a/drivers/md/dm-thin.c
+++ b/drivers/md/dm-thin.c
@@ -164,7 +164,7 @@ struct pool {
 
        struct pool_features pf;
        bool low_water_triggered:1;     /* A dm event has been sent */
-       bool no_free_space:1;           /* A -ENOSPC warning has been issued */
+       bool no_free_space:1;           /* bios will be requeued if set */
 
        struct dm_bio_prison *prison;
        struct dm_kcopyd_client *copier;
@@ -982,6 +982,20 @@ static void retry_on_resume(struct bio *bio)
        spin_unlock_irqrestore(&pool->lock, flags);
 }
 
+static void handle_unserviceable_bio(struct pool *pool, struct bio *bio)
+{
+       /*
+        * When pool is read-only, no cell locking is needed because
+        * nothing is changing.
+        */
+       WARN_ON_ONCE(get_pool_mode(pool) != PM_READ_ONLY);
+
+       if (pool->no_free_space)
+               retry_on_resume(bio);
+       else
+               bio_io_error(bio);
+}
+
 static void retry_bios_on_resume(struct pool *pool, struct dm_bio_prison_cell *cell)
 {
        struct bio *bio;
@@ -991,7 +1005,7 @@ static void retry_bios_on_resume(struct pool *pool, struct dm_bio_prison_cell *c
        cell_release(pool, cell, &bios);
 
        while ((bio = bio_list_pop(&bios)))
-               retry_on_resume(bio);
+               handle_unserviceable_bio(pool, bio);
 }
 
 static void process_discard(struct thin_c *tc, struct bio *bio)
@@ -1245,7 +1259,7 @@ static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
        switch (r) {
        case 0:
                if (lookup_result.shared && (rw == WRITE) && bio->bi_size)
-                       bio_io_error(bio);
+                       handle_unserviceable_bio(tc->pool, bio);
                else {
                        inc_all_io_entry(tc->pool, bio);
                        remap_and_issue(tc, bio, lookup_result.block);
@@ -1254,7 +1268,7 @@ static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
 
        case -ENODATA:
                if (rw != READ) {
-                       bio_io_error(bio);
+                       handle_unserviceable_bio(tc->pool, bio);
                        break;
                }
 
@@ -1565,9 +1579,9 @@ static int thin_bio_map(struct dm_target *ti, struct bio *bio)
                if (get_pool_mode(tc->pool) == PM_READ_ONLY) {
                        /*
                         * This block isn't provisioned, and we have no way
-                        * of doing so.  Just error it.
+                        * of doing so.
                         */
-                       bio_io_error(bio);
+                       handle_unserviceable_bio(tc->pool, bio);
                        return DM_MAPIO_SUBMITTED;
                }
                /* fall through */