[Qemu-devel] [PATCH 22/27] block/parallels: improve image reading performance

["Denis V. Lunev" <den@openvz.org>]

Try to perform IO for the biggest continuous block possible.
The performance for sequential read is increased from 220 Gb/sec to
360 Gb/sec for continous image on my SSD HDD.

Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Roman Kagan <rkagan@parallels.com>
CC: Kevin Wolf <kwolf@redhat.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
---
 block/parallels.c | 37 ++++++++++++++++++++++++++++++++-----
 1 file changed, 32 insertions(+), 5 deletions(-)

diff --git a/block/parallels.c b/block/parallels.c
index e8b3d09..d0128c9 100644
--- a/block/parallels.c
+++ b/block/parallels.c
@@ -96,6 +96,35 @@ static int cluster_remainder(BDRVParallelsState *s, int64_t sector_num,
     return MIN(nb_sectors, ret);
 }
 
+static int64_t block_status(BDRVParallelsState *s, int64_t sector_num,
+                            int nb_sectors, int *pnum)
+{
+    int64_t start_off = -2, prev_end_off = -2;
+
+    *pnum = 0;
+    while (nb_sectors > 0 || start_off == -2) {
+        int64_t offset = seek_to_sector(s, sector_num);
+        int to_end;
+
+        if (start_off == -2) {
+            start_off = offset;
+            prev_end_off = offset;
+        } else if (offset != prev_end_off) {
+            break;
+        }
+
+        to_end = cluster_remainder(s, sector_num, nb_sectors);
+        nb_sectors -= to_end;
+        sector_num += to_end;
+        *pnum += to_end;
+
+        if (offset > 0) {
+            prev_end_off += to_end;
+        }
+    }
+    return start_off;
+}
+
 static int64_t allocate_cluster(BlockDriverState *bs, int64_t sector_num)
 {
     BDRVParallelsState *s = bs->opaque;
@@ -133,11 +162,9 @@ static int64_t coroutine_fn parallels_co_get_block_status(BlockDriverState *bs,
     int64_t offset;
 
     qemu_co_mutex_lock(&s->lock);
-    offset = seek_to_sector(s, sector_num);
+    offset = block_status(s, sector_num, nb_sectors, pnum);
     qemu_co_mutex_unlock(&s->lock);
 
-    *pnum = cluster_remainder(s, sector_num, nb_sectors);
-
     if (offset < 0) {
         return 0;
     }
@@ -201,8 +228,8 @@ static coroutine_fn int parallels_co_readv(BlockDriverState *bs,
 
     qemu_co_mutex_lock(&s->lock);
     while (nb_sectors > 0) {
-        int64_t position = seek_to_sector(s, sector_num);
-        int n = cluster_remainder(s, sector_num, nb_sectors);
+        int n;
+        int64_t position = block_status(s, sector_num, nb_sectors, &n);
         int nbytes = n << BDRV_SECTOR_BITS;
 
         if (position < 0) {
-- 
1.9.1