Add FAILFAST support

[zfs.git] / module / zfs / vdev_disk.c

diff --git a/module/zfs/vdev_disk.c b/module/zfs/vdev_disk.c
index dbf9b08..8633925 100644 (file)
--- a/module/zfs/vdev_disk.c
+++ b/module/zfs/vdev_disk.c
@@ -91,6 +91,17 @@ bdev_capacity(struct block_device *bdev)
        return get_capacity(bdev->bd_disk);
 }
 
+static void
+vdev_disk_error(zio_t *zio)
+{
+#ifdef ZFS_DEBUG
+       printk("ZFS: zio error=%d type=%d offset=%llu "
+           "size=%llu flags=%x\n", zio->io_error, zio->io_type,
+           (u_longlong_t)zio->io_offset, (u_longlong_t)zio->io_size,
+           zio->io_flags);
+#endif
+}
+
 static int
 vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *ashift)
 {
@@ -133,9 +144,19 @@ vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *ashift)
        vd->vd_bdev = bdev;
        block_size =  vdev_bdev_block_size(bdev);
 
-       /* Check if this is a whole device.  When bdev->bd_contains ==
-        * bdev we have a whole device and not simply a partition. */
-       v->vdev_wholedisk = !!(bdev->bd_contains == bdev);
+       /* We think the wholedisk property should always be set when this
+        * function is called.  ASSERT here so if any legitimate cases exist
+        * where it's not set, we'll find them during debugging.  If we never
+        * hit the ASSERT, this and the following conditional statement can be
+        * removed. */
+       ASSERT3S(v->vdev_wholedisk, !=, -1ULL);
+
+       /* The wholedisk property was initialized to -1 in vdev_alloc() if it
+        * was unspecified.  In that case, check if this is a whole device.
+        * When bdev->bd_contains == bdev we have a whole device and not simply
+        * a partition. */
+       if (v->vdev_wholedisk == -1ULL)
+               v->vdev_wholedisk = (bdev->bd_contains == bdev);
 
        /* Clear the nowritecache bit, causes vdev_reopen() to try again. */
        v->vdev_nowritecache = B_FALSE;
@@ -222,6 +243,9 @@ vdev_disk_dio_put(dio_request_t *dr)
 
                if (zio) {
                        zio->io_error = error;
+                       ASSERT3S(zio->io_error, >=, 0);
+                       if (zio->io_error)
+                               vdev_disk_error(zio);
                        zio_interrupt(zio);
                }
        }
@@ -249,10 +273,10 @@ BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, size, error)
 #endif /* HAVE_2ARGS_BIO_END_IO_T */
 
        if (error == 0 && !test_bit(BIO_UPTODATE, &bio->bi_flags))
-               error = EIO;
+               error = -EIO;
 
        if (dr->dr_error == 0)
-               dr->dr_error = error;
+               dr->dr_error = -error;
 
        /* Drop reference aquired by __vdev_disk_physio */
        rc = vdev_disk_dio_put(dr);
@@ -318,15 +342,13 @@ retry:
        if (dr == NULL)
                return ENOMEM;
 
+       if (zio && !(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD)))
+                       bio_set_flags_failfast(bdev, &flags);
+
        dr->dr_zio = zio;
        dr->dr_rw = flags;
        block_size = vdev_bdev_block_size(bdev);
 
-#ifdef BIO_RW_FAILFAST
-       if (flags & (1 << BIO_RW_FAILFAST))
-               dr->dr_rw |= 1 << BIO_RW_FAILFAST;
-#endif /* BIO_RW_FAILFAST */
-
        /*
         * When the IO size exceeds the maximum bio size for the request
         * queue we are forced to break the IO in multiple bio's and wait
@@ -410,6 +432,7 @@ int
 vdev_disk_physio(struct block_device *bdev, caddr_t kbuf,
                 size_t size, uint64_t offset, int flags)
 {
+       bio_set_flags_failfast(bdev, &flags);
        return __vdev_disk_physio(bdev, NULL, kbuf, size, offset, flags);
 }
 
@@ -424,6 +447,9 @@ BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, size, rc)
                zio->io_vd->vdev_nowritecache = B_TRUE;
 
        bio_put(bio);
+       ASSERT3S(zio->io_error, >=, 0);
+       if (zio->io_error)
+               vdev_disk_error(zio);
        zio_interrupt(zio);
 
        BIO_END_IO_RETURN(0);
@@ -513,11 +539,6 @@ vdev_disk_io_start(zio_t *zio)
                return ZIO_PIPELINE_CONTINUE;
        }
 
-#ifdef BIO_RW_FAILFAST
-       if (zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD))
-               flags |= (1 << BIO_RW_FAILFAST);
-#endif /* BIO_RW_FAILFAST */
-
        error = __vdev_disk_physio(vd->vd_bdev, zio, zio->io_data,
                                   zio->io_size, zio->io_offset, flags);
        if (error) {