[libvirt] [PATCH v2 2/8] storage_backend_iscsi_direct: Simplify vol zeroing

[Michal Privoznik]

So far we have two branches: either we zero BLOCK_PER_PACKET
(currently 128) block at one, or if we're close to the last block
then we zero out one block at the time. This is very suboptimal.
We know how many block are there left. Might as well just write
them all at once.

Also, SCSI protocol has WRITESAME command which allows us to
write a single buffer multiple times. This means, that we have to
transfer the buffer far less frequently and have the destination
write it over and over again. Ideally, we would allocate a buffer
with size equivalent to block size of the LUN and then write it
as many times as there are blocks on the LUN. Sadly, this works
well in theory but in practise I've found out that it's possible
to write only 4096 blocks at once. Any higher value leads to a
failure. But still better than Twili^Wwhat we have now.
Since 4096 seems like an arbitrary number, I'm also implementing
a mechanism that exponentially decreases the number of blocks
we're trying to write at once. It starts at 4096 blocks and if
that's too much the number is halved, and so on.

Signed-off-by: Michal Privoznik <mprivozn at redhat.com>
---
 src/storage/storage_backend_iscsi_direct.c | 55 ++++++++++++++++------
 1 file changed, 41 insertions(+), 14 deletions(-)

diff --git a/src/storage/storage_backend_iscsi_direct.c b/src/storage/storage_backend_iscsi_direct.c
index fc3b6550f7..0af7adf32c 100644
--- a/src/storage/storage_backend_iscsi_direct.c
+++ b/src/storage/storage_backend_iscsi_direct.c
@@ -39,9 +39,11 @@
 
 #define ISCSI_DEFAULT_TARGET_PORT 3260
 #define VIR_ISCSI_TEST_UNIT_TIMEOUT 30 * 1000
-#define BLOCK_PER_PACKET 128
 #define VOL_NAME_PREFIX "unit:0:0:"
 
+/* Empirically tested to be the highest value to work without any error. */
+#define WIPE_BLOCKS_AT_ONCE 4096
+
 VIR_LOG_INIT("storage.storage_backend_iscsi_direct");
 
 static struct iscsi_context *
@@ -624,6 +626,7 @@ virStorageBackendISCSIDirectVolWipeZero(virStorageVolDefPtr vol,
     uint64_t lba = 0;
     uint32_t block_size;
     uint64_t nb_block;
+    uint64_t wipe_at_once = WIPE_BLOCKS_AT_ONCE;
     struct scsi_task *task = NULL;
     int lun = 0;
     int ret = -1;
@@ -635,25 +638,49 @@ virStorageBackendISCSIDirectVolWipeZero(virStorageVolDefPtr vol,
         return ret;
     if (virISCSIDirectGetVolumeCapacity(iscsi, lun, &block_size, &nb_block))
         return ret;
-    if (VIR_ALLOC_N(data, block_size * BLOCK_PER_PACKET))
+    if (VIR_ALLOC_N(data, block_size))
         return ret;
 
+    VIR_DEBUG("Starting zeroing of lun=%d block_size=%ju nb_block=%ju wipe_at_once=%ju",
+              lun, (uintmax_t) block_size, (uintmax_t) nb_block, (uintmax_t) wipe_at_once);
+
     while (lba < nb_block) {
-        if (nb_block - lba > block_size * BLOCK_PER_PACKET) {
+        const uint64_t to_write = MIN(nb_block - lba + 1, wipe_at_once);
+        bool fail = false;
 
-            if (!(task = iscsi_write16_sync(iscsi, lun, lba, data,
-                                            block_size * BLOCK_PER_PACKET,
-                                            block_size, 0, 0, 0, 0, 0)))
-                return -1;
-            scsi_free_scsi_task(task);
-            lba += BLOCK_PER_PACKET;
-        } else {
-            if (!(task = iscsi_write16_sync(iscsi, lun, lba, data, block_size,
-                                        block_size, 0, 0, 0, 0, 0)))
-                return -1;
+        if (!(task = iscsi_writesame16_sync(iscsi, lun, lba,
+                                            data, block_size,
+                                            to_write,
+                                            0, 0, 0, 0)))
+            fail = true;
+
+        if (task &&
+            task->status == SCSI_STATUS_CHECK_CONDITION &&
+            task->sense.key == SCSI_SENSE_ILLEGAL_REQUEST &&
+            task->sense.ascq == SCSI_SENSE_ASCQ_INVALID_FIELD_IN_CDB) {
+            /* This means that we tried to write too much blocks at once.
+             * Halve it (if it's not small enough already) and retry. */
+            if (wipe_at_once > 1) {
+                wipe_at_once /= 2;
+                VIR_DEBUG("Halving wipe_at_once to %ju", (uintmax_t) wipe_at_once);
+                scsi_free_scsi_task(task);
+                continue;
+            }
+
+            fail = true;
+        }
+
+        if (fail) {
+            virReportError(VIR_ERR_INTERNAL_ERROR,
+                           _("failed to run writesame: %s"),
+                           iscsi_get_error(iscsi));
             scsi_free_scsi_task(task);
-            lba++;
+            return -1;
         }
+
+        scsi_free_scsi_task(task);
+
+        lba += to_write;
     }
 
     return 0;
-- 
2.19.2