From 9a02c0bcafdbf681e76f816ad3f60dfb7dea13fb Mon Sep 17 00:00:00 2001
From: Maxim Levitsky <mlevitsk@redhat.com>
Date: Wed, 5 Jun 2019 13:57:03 +0200
Subject: [PATCH 15/23] file-posix: Support BDRV_REQ_NO_FALLBACK for zero
 writes

RH-Author: Maxim Levitsky <mlevitsk@redhat.com>
Message-id: <20190605135705.24526-8-mlevitsk@redhat.com>
Patchwork-id: 88560
O-Subject: [RHEL-7.7 qemu-kvm-rhev PATCH 7/9] file-posix: Support BDRV_REQ_NO_FALLBACK for zero writes
Bugzilla: 1648622
RH-Acked-by: Max Reitz <mreitz@redhat.com>
RH-Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
RH-Acked-by: John Snow <jsnow@redhat.com>

From: Kevin Wolf <kwolf@redhat.com>

We know that the kernel implements a slow fallback code path for
BLKZEROOUT, so if BDRV_REQ_NO_FALLBACK is given, we shouldn't call it.
The other operations we call in the context of .bdrv_co_pwrite_zeroes
should usually be quick, so no modification should be needed for them.
If we ever notice that there are additional problematic cases, we can
still make these conditional as well.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Eric Blake <eblake@redhat.com>

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1648622

Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
(Cherry picked from 738301e11758171defaa5a5237d584f8226af89f)

Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
---
 block/file-posix.c      | 24 ++++++++++++++++--------
 include/block/raw-aio.h |  1 +
 2 files changed, 17 insertions(+), 8 deletions(-)

diff --git a/block/file-posix.c b/block/file-posix.c
index 90c719f..d1926b3 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -632,7 +632,7 @@ static int raw_open_common(BlockDriverState *bs, QDict *options,
     }
 #endif
 
-    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
+    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK;
     ret = 0;
 fail:
     if (filename && (bdrv_flags & BDRV_O_TEMPORARY)) {
@@ -1449,14 +1449,19 @@ static ssize_t handle_aiocb_write_zeroes_block(RawPosixAIOData *aiocb)
     }
 
 #ifdef BLKZEROOUT
-    do {
-        uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes };
-        if (ioctl(aiocb->aio_fildes, BLKZEROOUT, range) == 0) {
-            return 0;
-        }
-    } while (errno == EINTR);
+    /* The BLKZEROOUT implementation in the kernel doesn't set
+     * BLKDEV_ZERO_NOFALLBACK, so we can't call this if we have to avoid slow
+     * fallbacks. */
+    if (!(aiocb->aio_type & QEMU_AIO_NO_FALLBACK)) {
+        do {
+            uint64_t range[2] = { aiocb->aio_offset, aiocb->aio_nbytes };
+            if (ioctl(aiocb->aio_fildes, BLKZEROOUT, range) == 0) {
+                return 0;
+            }
+        } while (errno == EINTR);
 
-    ret = translate_err(-errno);
+        ret = translate_err(-errno);
+    }
 #endif
 
     if (ret == -ENOTSUP) {
@@ -2535,6 +2540,9 @@ raw_do_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int bytes,
     if (blkdev) {
         acb.aio_type |= QEMU_AIO_BLKDEV;
     }
+    if (flags & BDRV_REQ_NO_FALLBACK) {
+        acb.aio_type |= QEMU_AIO_NO_FALLBACK;
+    }
 
     if (flags & BDRV_REQ_MAY_UNMAP) {
         acb.aio_type |= QEMU_AIO_DISCARD;
diff --git a/include/block/raw-aio.h b/include/block/raw-aio.h
index 2ffcd9d..5a926a3 100644
--- a/include/block/raw-aio.h
+++ b/include/block/raw-aio.h
@@ -40,6 +40,7 @@
 /* AIO flags */
 #define QEMU_AIO_MISALIGNED   0x1000
 #define QEMU_AIO_BLKDEV       0x2000
+#define QEMU_AIO_NO_FALLBACK  0x4000
 
 
 /* linux-aio.c - Linux native implementation */
-- 
1.8.3.1