The following changes since commit 417296c8d8588f782018d01a317f88957e9786d6:

tests/qtest/netdev-socket: Raise connection timeout to 60 seconds (2023-02-09 11:23:53 +0000)

for you to fetch changes up to acbc8aee5b09222dc6a5cb88306b67bcbe37e30b:

iotests/detect-zeroes-registered-buf: add new test (2023-02-09 10:22:30 -0500)

A few fixes that I've picked up.

Akihiko Odaki (1):

vhost-user-fs: Back up vqs before cleaning up vhost_dev

Emanuele Giuseppe Esposito (1):

virtio-blk: add missing AioContext lock

Stefan Hajnoczi (4):

block: fix detect-zeroes= with BDRV_REQ_REGISTERED_BUF

qemu-io: use BdrvRequestFlags instead of int

qemu-io: add -r option to register I/O buffer

iotests/detect-zeroes-registered-buf: add new test

block/io.c | 3 +

hw/block/virtio-blk.c | 5 +

hw/virtio/vhost-user-fs.c | 4 +-

qemu-io-cmds.c | 215 +++++++++++-------

.../tests/detect-zeroes-registered-buf | 58 +++++

.../tests/detect-zeroes-registered-buf.out | 7 +

6 files changed, 210 insertions(+), 82 deletions(-)

create mode 100755 tests/qemu-iotests/tests/detect-zeroes-registered-buf

create mode 100644 tests/qemu-iotests/tests/detect-zeroes-registered-buf.out

2.39.1

From: Emanuele Giuseppe Esposito <eesposit@redhat.com>

virtio_blk_update_config() calls blk_get_geometry and blk_getlength,

and both functions eventually end up calling bdrv_poll_co when not

running in a coroutine:

- blk_getlength is a co_wrapper_mixed function

- blk_get_geometry calls bdrv_get_geometry -> bdrv_nb_sectors, a

co_wrapper_mixed function too

Since we are not running in a coroutine, we need to take s->blk

AioContext lock, otherwise bdrv_poll_co will inevitably call

AIO_WAIT_WHILE and therefore try to un unlock() an AioContext lock

that was never acquired.

RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=2167838

Steps to reproduce the issue: simply boot a VM with

-object '{"qom-type":"iothread","id":"iothread1"}' \

-blockdev '{"driver":"file","filename":"$QCOW2","aio":"native","node-name":"libvirt-1-storage","cache":{"direct":true,"no-flush":false},"auto-read-only":true,"discard":"unmap"}' \

-blockdev '{"node-name":"libvirt-1-format","read-only":false,"cache":{"direct":true,"no-flush":false},"driver":"qcow2","file":"libvirt-1-storage"}' \

-device virtio-blk-pci,iothread=iothread1,drive=libvirt-1-format,id=virtio-disk0,bootindex=1,write-cache=on

and observe that it will fail not manage to boot with "qemu_mutex_unlock_impl: Operation not permitted"

Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>

Acked-by: Michael S. Tsirkin <mst@redhat.com>

Tested-by: Lukáš Doktor <ldoktor@redhat.com>

Message-Id: <20230208111148.1040083-1-eesposit@redhat.com>

hw/block/virtio-blk.c | 5 +++++

1 file changed, 5 insertions(+)

@@ -XXX,XX +XXX,XX @@ static void virtio_blk_update_config(VirtIODevice *vdev, uint8_t *config)

uint64_t capacity;

int64_t length;

int blk_size = conf->logical_block_size;

+ AioContext *ctx;

+

+ ctx = blk_get_aio_context(s->blk);

+ aio_context_acquire(ctx);

blk_get_geometry(s->blk, &capacity);

memset(&blkcfg, 0, sizeof(blkcfg));

@@ -XXX,XX +XXX,XX @@ static void virtio_blk_update_config(VirtIODevice *vdev, uint8_t *config)

* per track (cylinder).

*/

length = blk_getlength(s->blk);

+ aio_context_release(ctx);

if (length > 0 && length / conf->heads / conf->secs % blk_size) {

blkcfg.geometry.sectors = conf->secs & ~s->sector_mask;

} else {

2.39.1

From: Akihiko Odaki <akihiko.odaki@daynix.com>

vhost_dev_cleanup() clears vhost_dev so back up its vqs member to free

the memory pointed by the member.

Fixes: 98fc1ada4c ("virtio: add vhost-user-fs base device")

Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Message-Id: <20230130140225.77964-1-akihiko.odaki@daynix.com>

---

hw/virtio/vhost-user-fs.c | 4 ++--

1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/hw/virtio/vhost-user-fs.c b/hw/virtio/vhost-user-fs.c

--- a/hw/virtio/vhost-user-fs.c

+++ b/hw/virtio/vhost-user-fs.c

@@ -XXX,XX +XXX,XX @@ static void vuf_device_unrealize(DeviceState *dev)

VirtIODevice *vdev = VIRTIO_DEVICE(dev);

VHostUserFS *fs = VHOST_USER_FS(dev);

+ struct vhost_virtqueue *vhost_vqs = fs->vhost_dev.vqs;

int i;

/* This will stop vhost backend if appropriate. */

@@ -XXX,XX +XXX,XX @@ static void vuf_device_unrealize(DeviceState *dev)

static struct vhost_dev *vuf_get_vhost(VirtIODevice *vdev)

2.39.1

The block layer APIs use BdrvRequestFlags while qemu-io code uses int.

Although the code compiles and runs fine, BdrvRequestFlags is clearer

because it differentiates between other types of flags like bdrv_open()

flags.

This is purely refactoring.

Reviewed-by: Eric Blake <eblake@redhat.com>

Reviewed-by: Hanna Czenczek <hreitz@redhat.com>

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Message-Id: <20230207203719.242926-3-stefanha@redhat.com>

---

qemu-io-cmds.c | 13 +++++++------

1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/qemu-io-cmds.c b/qemu-io-cmds.c

--- a/qemu-io-cmds.c

+++ b/qemu-io-cmds.c

@@ -XXX,XX +XXX,XX @@ static int do_pread(BlockBackend *blk, char *buf, int64_t offset,

static int do_pwrite(BlockBackend *blk, char *buf, int64_t offset,

- int64_t bytes, int flags, int64_t *total)

+ int64_t bytes, BdrvRequestFlags flags, int64_t *total)

int ret;

@@ -XXX,XX +XXX,XX @@ static int do_pwrite(BlockBackend *blk, char *buf, int64_t offset,

static int do_pwrite_zeroes(BlockBackend *blk, int64_t offset,

- int64_t bytes, int flags, int64_t *total)

+ int64_t bytes, BdrvRequestFlags flags,

+ int64_t *total)

int ret = blk_pwrite_zeroes(blk, offset, bytes,

flags | BDRV_REQ_ZERO_WRITE);

@@ -XXX,XX +XXX,XX @@ static int do_aio_readv(BlockBackend *blk, QEMUIOVector *qiov,

static int do_aio_writev(BlockBackend *blk, QEMUIOVector *qiov,

- int64_t offset, int flags, int *total)

+ int64_t offset, BdrvRequestFlags flags, int *total)

int async_ret = NOT_DONE;

@@ -XXX,XX +XXX,XX @@ static int write_f(BlockBackend *blk, int argc, char **argv)

struct timespec t1, t2;

bool Cflag = false, qflag = false, bflag = false;

bool Pflag = false, zflag = false, cflag = false, sflag = false;

- int flags = 0;

+ BdrvRequestFlags flags = 0;

int c, cnt, ret;

char *buf = NULL;

int64_t offset;

@@ -XXX,XX +XXX,XX @@ static int writev_f(BlockBackend *blk, int argc, char **argv)

{

struct timespec t1, t2;

bool Cflag = false, qflag = false;

- int flags = 0;

+ BdrvRequestFlags flags = 0;

int c, cnt, ret;

char *buf;

int64_t offset;

@@ -XXX,XX +XXX,XX @@ static int aio_write_f(BlockBackend *blk, int argc, char **argv)

int nr_iov, c;

int pattern = 0xcd;

struct aio_ctx *ctx = g_new0(struct aio_ctx, 1);

- int flags = 0;

+ BdrvRequestFlags flags = 0;

ctx->blk = blk;

while ((c = getopt(argc, argv, "CfiqP:uz")) != -1) {

2.39.1

When a write request is converted into a write zeroes request by the

detect-zeroes= feature, it is no longer associated with an I/O buffer.

The BDRV_REQ_REGISTERED_BUF flag doesn't make sense without an I/O

buffer and must be cleared because bdrv_co_do_pwrite_zeroes() fails with

-EINVAL when it's set.

Fiona Ebner <f.ebner@proxmox.com> bisected and diagnosed this QEMU 7.2

regression where writes containing zeroes to a blockdev with

discard=unmap,detect-zeroes=unmap fail.

Buglink: https://gitlab.com/qemu-project/qemu/-/issues/1404

Fixes: e8b6535533be ("block: add BDRV_REQ_REGISTERED_BUF request flag")

Tested-by: Fiona Ebner <f.ebner@proxmox.com>

Cc: qemu-stable@nongnu.org

Reviewed-by: Eric Blake <eblake@redhat.com>

Reviewed-by: Hanna Czenczek <hreitz@redhat.com>

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Message-Id: <20230207203719.242926-2-stefanha@redhat.com>

---

block/io.c | 3 +++

1 file changed, 3 insertions(+)

@@ -XXX,XX +XXX,XX @@ static int coroutine_fn bdrv_aligned_pwritev(BdrvChild *child,

if (bs->detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP) {

flags |= BDRV_REQ_MAY_UNMAP;

}

+

+ /* Can't use optimization hint with bufferless zero write */

+ flags &= ~BDRV_REQ_REGISTERED_BUF;

}

if (ret < 0) {

2.39.1

The blk_register_buf() API is an optimization hint that allows some

block drivers to avoid I/O buffer housekeeping or bounce buffers.

Add an -r option to register the I/O buffer so that qemu-io can be used

to test the blk_register_buf() API. The next commit will add a test that

uses the new option.

Reviewed-by: Eric Blake <eblake@redhat.com>

Reviewed-by: Hanna Czenczek <hreitz@redhat.com>

qemu-io-cmds.c | 204 +++++++++++++++++++++++++++++++------------------

1 file changed, 129 insertions(+), 75 deletions(-)

diff --git a/qemu-io-cmds.c b/qemu-io-cmds.c

--- a/qemu-io-cmds.c

+++ b/qemu-io-cmds.c

@@ -XXX,XX +XXX,XX @@ static int parse_pattern(const char *arg)

*/

#define MISALIGN_OFFSET 16

-static void *qemu_io_alloc(BlockBackend *blk, size_t len, int pattern)

+static void *qemu_io_alloc(BlockBackend *blk, size_t len, int pattern,

+ bool register_buf)

void *buf;

@@ -XXX,XX +XXX,XX @@ static void *qemu_io_alloc(BlockBackend *blk, size_t len, int pattern)

buf = blk_blockalign(blk, len);

memset(buf, pattern, len);

+ if (register_buf) {