[Qemu-devel] [PATCH 19/19] block/parallels: optimize linear image expansion

[Denis V. Lunev]

Plain image expansion spends a lot of time to update image file size.
This seriously affects the performance. The following simple test
  qemu_img create -f parallels -o cluster_size=64k ./1.hds 64G
  qemu_io -n -c "write -P 0x11 0 1024M" ./1.hds
could be improved if the format driver will pre-allocate some space
in the image file with a reasonable chunk.

This patch preallocates 128 Mb using bdrv_write_zeroes, which should
normally use fallocate() call inside. Fallback to older truncate()
could be used as a fallback using image open options thanks to the
previous patch.

The benefit is around 15%.

This patch is final in this series. Block driver has near native
performance now.

Signed-off-by: Denis V. Lunev <address@hidden>
CC: Kevin Wolf <address@hidden>
CC: Stefan Hajnoczi <address@hidden>
---
 block/parallels.c | 35 +++++++++++++++++++++++++++++++++--
 1 file changed, 33 insertions(+), 2 deletions(-)

diff --git a/block/parallels.c b/block/parallels.c
index 12a9cea..5ec4a0d 100644
--- a/block/parallels.c
+++ b/block/parallels.c
@@ -82,6 +82,7 @@ typedef struct BDRVParallelsState {
     int bat_cache_off;
     int data_off;
 
+    int64_t  prealloc_off;
     uint64_t prealloc_size;
     ParallelsPreallocMode prealloc_mode;
 
@@ -216,9 +217,19 @@ static int parallels_open(BlockDriverState *bs, QDict 
*options, int flags,
         goto fail_options;
     }
 
-    for (i = 0; i < s->bat_size; i++)
+    for (i = 0; i < s->bat_size; i++) {
+        int64_t off;
         le32_to_cpus(&s->bat[i]);
 
+        if (s->bat[i] == 0) {
+            continue;
+        }
+        off = s->bat[i] * s->off_multiplier;
+        if (off >= s->prealloc_off) {
+            s->prealloc_off = off + s->tracks;
+        }
+    }
+
     qemu_co_mutex_init(&s->lock);
 
     s->bat_cache_off = -1;
@@ -230,6 +241,9 @@ static int parallels_open(BlockDriverState *bs, QDict 
*options, int flags,
     if (s->data_off == 0) {
         s->data_off = ROUND_UP(bat_offset(s->bat_size), BDRV_SECTOR_SIZE);
     }
+    if (s->prealloc_off == 0) {
+        s->prealloc_off = s->data_off >> BDRV_SECTOR_BITS;
+    }
 
     return 0;
 
@@ -338,7 +352,19 @@ static int64_t allocate_sector(BlockDriverState *bs, 
int64_t sector_num)
     }
 
     pos = bdrv_getlength(bs->file) >> BDRV_SECTOR_BITS;
-    bdrv_truncate(bs->file, (pos + s->tracks) << BDRV_SECTOR_BITS);
+    if (s->prealloc_off + s->tracks > pos) {
+        if (s->prealloc_mode == PRL_PREALLOC_MODE_FALLOCATE)
+            ret = bdrv_write_zeroes(bs->file, s->prealloc_off,
+                                    s->prealloc_size, 0);
+        else
+            ret = bdrv_truncate(bs->file,
+                    (s->prealloc_off + s->prealloc_size) << BDRV_SECTOR_BITS);
+        if (ret < 0) {
+            return ret;
+        }
+    }
+    pos = s->prealloc_off;
+    s->prealloc_off += s->tracks;
 
     ret = cache_bat(bs, idx, pos / s->off_multiplier);
     if (ret < 0) {
@@ -546,6 +572,11 @@ exit:
 static void parallels_close(BlockDriverState *bs)
 {
     BDRVParallelsState *s = bs->opaque;
+
+    if (bs->open_flags & BDRV_O_RDWR) {
+        bdrv_truncate(bs->file, s->prealloc_off << BDRV_SECTOR_BITS);
+    }
+
     qemu_vfree(s->bat_cache);
     g_free(s->bat);
 }
-- 
1.9.1