[Qemu-devel] [PATCH 0/5] Block patches
[PULL 0/1] Block patches
1
The following changes since commit fb68096da3d35e64c88cd610c1fa42766c58e92a:
1
The following changes since commit a3607def89f9cd68c1b994e1030527df33aa91d0:
2
2
3
Revert "tests: use memfd in vhost-user-test" (2018-02-13 09:51:52 +0000)
3
Update version for v6.2.0-rc4 release (2021-12-07 17:51:38 -0800)
4
4
5
are available in the Git repository at:
5
are available in the Git repository at:
6
6
7
git://github.com/stefanha/qemu.git tags/block-pull-request
7
https://gitlab.com/stefanha/qemu.git tags/block-pull-request
8
8
9
for you to fetch changes up to 64b01feca991e5b19a5d750ef77cdca92b68bdbb:
9
for you to fetch changes up to cf4fbc3030c974fff726756a7ceef8386cdf500b:
10
10
11
misc: fix spelling (2018-02-13 15:38:17 +0000)
11
block/nvme: fix infinite loop in nvme_free_req_queue_cb() (2021-12-09 09:19:49 +0000)
12
13
----------------------------------------------------------------
14
Pull request
15
16
An infinite loop fix for the userspace NVMe driver.
12
17
13
----------------------------------------------------------------
18
----------------------------------------------------------------
14
19
15
----------------------------------------------------------------
20
Stefan Hajnoczi (1):
21
block/nvme: fix infinite loop in nvme_free_req_queue_cb()
16
22
17
Fam Zheng (2):
23
block/nvme.c | 5 +++--
18
Add a git-publish configuration file
24
1 file changed, 3 insertions(+), 2 deletions(-)
19
README: Document 'git-publish' workflow
20
21
Marc-André Lureau (1):
22
misc: fix spelling
23
24
Stefan Hajnoczi (1):
25
vl: pause vcpus before stopping iothreads
26
27
Wolfgang Bumiller (1):
28
ratelimit: don't align wait time with slices
29
30
include/qemu/ratelimit.h | 11 ++++-----
31
util/qemu-coroutine-lock.c | 2 +-
32
vl.c | 12 ++++++++--
33
.gitpublish | 58 ++++++++++++++++++++++++++++++++++++++++++++++
34
README | 31 ++++++++++++++++++++++++-
35
5 files changed, 104 insertions(+), 10 deletions(-)
36
create mode 100644 .gitpublish
37
25
38
--
26
--
39
2.14.3
27
2.33.1
40
28
41
29
30
diff view generated by jsdifflib
[Qemu-devel] [PATCH 1/5] Add a git-publish configuration file
Deleted patch
1
From: Fam Zheng <famz@redhat.com>
2
1
3
git-publish [1] is a convenient tool to send patches and has been
4
popular among QEMU developers. Recently it has been made available in
5
Fedora official repo thanks to Stefan's work.
6
7
One nice feature of the tool is a per-project configuration with
8
profiles, especially in which the cccmd option is a handy method to
9
create the Cc list.
10
11
[1]: https://github.com/stefanha/git-publish
12
13
Signed-off-by: Fam Zheng <famz@redhat.com>
14
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
15
Message-id: 20180205054725.25634-2-famz@redhat.com
16
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
17
---
18
.gitpublish | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
19
1 file changed, 58 insertions(+)
20
create mode 100644 .gitpublish
21
22
diff --git a/.gitpublish b/.gitpublish
23
new file mode 100644
24
index XXXXXXX..XXXXXXX
25
--- /dev/null
26
+++ b/.gitpublish
27
@@ -XXX,XX +XXX,XX @@
28
+#
29
+# Common git-publish profiles that can be used to send patches to QEMU upstream.
30
+#
31
+# See https://github.com/stefanha/git-publish for more information
32
+#
33
+[gitpublishprofile "default"]
34
+base = master
35
+prefix = PATCH
36
+to = qemu-devel@nongnu.org
37
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
38
+
39
+[gitpublishprofile "rfc"]
40
+base = master
41
+prefix = RFC PATCH
42
+to = qemu-devel@nongnu.org
43
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
44
+
45
+[gitpublishprofile "stable"]
46
+base = master
47
+prefix = PATCH
48
+to = qemu-devel@nongnu.org
49
+cc = qemu-stable@nongnu.org
50
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
51
+
52
+[gitpublishprofile "trivial"]
53
+base = master
54
+prefix = PATCH
55
+to = qemu-devel@nongnu.org
56
+cc = qemu-trivial@nongnu.org
57
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
58
+
59
+[gitpublishprofile "block"]
60
+base = master
61
+prefix = PATCH
62
+to = qemu-devel@nongnu.org
63
+cc = qemu-block@nongnu.org
64
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
65
+
66
+[gitpublishprofile "arm"]
67
+base = master
68
+prefix = PATCH
69
+to = qemu-devel@nongnu.org
70
+cc = qemu-arm@nongnu.org
71
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
72
+
73
+[gitpublishprofile "s390"]
74
+base = master
75
+prefix = PATCH
76
+to = qemu-devel@nongnu.org
77
+cc = qemu-s390@nongnu.org
78
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
79
+
80
+[gitpublishprofile "ppc"]
81
+base = master
82
+prefix = PATCH
83
+to = qemu-devel@nongnu.org
84
+cc = qemu-ppc@nongnu.org
85
+cccmd = scripts/get_maintainer.pl --noroles --norolestats --nogit --nogit-fallback 2>/dev/null
86
--
87
2.14.3
88
89
diff view generated by jsdifflib
[Qemu-devel] [PATCH 2/5] README: Document 'git-publish' workflow
Deleted patch
1
From: Fam Zheng <famz@redhat.com>
2
1
3
[Fixed s/git-profile/git-publish/ typo and s/for once/the first time/.
4
--Stefan]
5
6
Signed-off-by: Fam Zheng <famz@redhat.com>
7
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
8
Message-id: 20180205054725.25634-3-famz@redhat.com
9
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
10
---
11
README | 31 ++++++++++++++++++++++++++++++-
12
1 file changed, 30 insertions(+), 1 deletion(-)
13
14
diff --git a/README b/README
15
index XXXXXXX..XXXXXXX 100644
16
--- a/README
17
+++ b/README
18
@@ -XXX,XX +XXX,XX @@ The QEMU source code is maintained under the GIT version control system.
19
20
git clone git://git.qemu.org/qemu.git
21
22
-When submitting patches, the preferred approach is to use 'git
23
+When submitting patches, one common approach is to use 'git
24
format-patch' and/or 'git send-email' to format & send the mail to the
25
qemu-devel@nongnu.org mailing list. All patches submitted must contain
26
a 'Signed-off-by' line from the author. Patches should follow the
27
guidelines set out in the HACKING and CODING_STYLE files.
28
29
+A 'git-publish' utility was created to make above process less
30
+cumbersome, and is highly recommended for making regular contributions,
31
+or even just for sending consecutive patch series revisions. It also
32
+requires a working 'git send-email' setup, and by default doesn't
33
+automate everything, so you may want to go through the above steps
34
+manually the first time.
35
+
36
+For installation instructions, please go to
37
+
38
+ https://github.com/stefanha/git-publish
39
+
40
+The workflow with 'git-publish' is:
41
+
42
+ $ git checkout master -b my-feature
43
+ $ # work on new commits, add your 'Signed-off-by' lines to each
44
+ $ git publish
45
+
46
+Your patch series will be sent and tagged as my-feature-v1 if you need to refer
47
+back to it in the future.
48
+
49
+Sending v2:
50
+
51
+ $ git checkout my-feature # same topic branch
52
+ $ # making changes to the commits (using 'git rebase', for example)
53
+ $ git publish
54
+
55
+Your patch series will be sent with 'v2' tag in the subject and the git tip
56
+will be tagged as my-feature-v2.
57
+
58
Additional information on submitting patches can be found online via
59
the QEMU website
60
61
--
62
2.14.3
63
64
diff view generated by jsdifflib
[Qemu-devel] [PATCH 3/5] vl: pause vcpus before stopping iothreads
[PULL 1/1] block/nvme: fix infinite loop in nvme_free_req_queue_cb()
1
Commit dce8921b2baaf95974af8176406881872067adfa ("iothread: Stop threads
1
When the request free list is exhausted the coroutine waits on
2
before main() quits") introduced iothread_stop_all() to avoid the
2
q->free_req_queue for the next free request. Whenever a request is
3
following virtio-scsi assertion failure:
3
completed a BH is scheduled to invoke nvme_free_req_queue_cb() and wake
4
up waiting coroutines.
4
5
5
assert(blk_get_aio_context(d->conf.blk) == s->ctx);
6
1. nvme_get_free_req() waits for a free request:
6
7
7
Back then the assertion failed because when bdrv_close_all() made
8
while (q->free_req_head == -1) {
8
d->conf.blk NULL, blk_get_aio_context() returned the global AioContext
9
...
9
instead of s->ctx.
10
trace_nvme_free_req_queue_wait(q->s, q->index);
11
qemu_co_queue_wait(&q->free_req_queue, &q->lock);
12
...
13
}
10
14
11
The same assertion can still fail today when vcpus submit new I/O
15
2. nvme_free_req_queue_cb() wakes up the coroutine:
12
requests after iothread_stop_all() has moved the BDS to the global
13
AioContext.
14
16
15
This patch hardens the iothread_stop_all() approach by pausing vcpus
17
while (qemu_co_enter_next(&q->free_req_queue, &q->lock)) {
16
before calling iothread_stop_all().
18
^--- infinite loop when free_req_head == -1
19
}
17
20
18
Note that the assertion failure is a race condition. It is not possible
21
nvme_free_req_queue_cb() and the coroutine form an infinite loop when
19
to reproduce it reliably.
22
q->free_req_head == -1. Fix this by checking q->free_req_head in
23
nvme_free_req_queue_cb(). If the free request list is exhausted, don't
24
wake waiting coroutines. Eventually an in-flight request will complete
25
and the BH will be scheduled again, guaranteeing forward progress.
20
26
21
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
27
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
22
Message-id: 20180201110708.8080-1-stefanha@redhat.com
28
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
29
Message-id: 20211208152246.244585-1-stefanha@redhat.com
23
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
30
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
24
---
31
---
25
vl.c | 12 ++++++++++--
32
block/nvme.c | 5 +++--
26
1 file changed, 10 insertions(+), 2 deletions(-)
33
1 file changed, 3 insertions(+), 2 deletions(-)
27
34
28
diff --git a/vl.c b/vl.c
35
diff --git a/block/nvme.c b/block/nvme.c
29
index XXXXXXX..XXXXXXX 100644
36
index XXXXXXX..XXXXXXX 100644
30
--- a/vl.c
37
--- a/block/nvme.c
31
+++ b/vl.c
38
+++ b/block/nvme.c
32
@@ -XXX,XX +XXX,XX @@ int main(int argc, char **argv, char **envp)
39
@@ -XXX,XX +XXX,XX @@ static void nvme_free_req_queue_cb(void *opaque)
33
40
NVMeQueuePair *q = opaque;
34
main_loop();
41
35
replay_disable_events();
42
qemu_mutex_lock(&q->lock);
36
+
43
- while (qemu_co_enter_next(&q->free_req_queue, &q->lock)) {
37
+ /* The ordering of the following is delicate. Stop vcpus to prevent new
44
- /* Retry all pending requests */
38
+ * I/O requests being queued by the guest. Then stop IOThreads (this
45
+ while (q->free_req_head != -1 &&
39
+ * includes a drain operation and completes all request processing). At
46
+ qemu_co_enter_next(&q->free_req_queue, &q->lock)) {
40
+ * this point emulated devices are still associated with their IOThreads
47
+ /* Retry waiting requests */
41
+ * (if any) but no longer have any work to do. Only then can we close
48
}
42
+ * block devices safely because we know there is no more I/O coming.
49
qemu_mutex_unlock(&q->lock);
43
+ */
50
}
44
+ pause_all_vcpus();
45
iothread_stop_all();
46
-
47
- pause_all_vcpus();
48
bdrv_close_all();
49
+
50
res_free();
51
52
/* vhost-user must be cleaned up before chardevs. */
53
--
51
--
54
2.14.3
52
2.33.1
55
53
56
54
diff view generated by jsdifflib
[Qemu-devel] [PATCH 4/5] ratelimit: don't align wait time with slices
Deleted patch
1
From: Wolfgang Bumiller <w.bumiller@proxmox.com>
2
1
3
It is possible for rate limited writes to keep overshooting a slice's
4
quota by a tiny amount causing the slice-aligned waiting period to
5
effectively halve the rate.
6
7
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8
Reviewed-by: Alberto Garcia <berto@igalia.com>
9
Message-id: 20180207071758.6818-1-w.bumiller@proxmox.com
10
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
11
---
12
include/qemu/ratelimit.h | 11 +++++------
13
1 file changed, 5 insertions(+), 6 deletions(-)
14
15
diff --git a/include/qemu/ratelimit.h b/include/qemu/ratelimit.h
16
index XXXXXXX..XXXXXXX 100644
17
--- a/include/qemu/ratelimit.h
18
+++ b/include/qemu/ratelimit.h
19
@@ -XXX,XX +XXX,XX @@ typedef struct {
20
static inline int64_t ratelimit_calculate_delay(RateLimit *limit, uint64_t n)
21
{
22
int64_t now = qemu_clock_get_ns(QEMU_CLOCK_REALTIME);
23
- uint64_t delay_slices;
24
+ double delay_slices;
25
26
assert(limit->slice_quota && limit->slice_ns);
27
28
@@ -XXX,XX +XXX,XX @@ static inline int64_t ratelimit_calculate_delay(RateLimit *limit, uint64_t n)
29
return 0;
30
}
31
32
- /* Quota exceeded. Calculate the next time slice we may start
33
- * sending data again. */
34
- delay_slices = (limit->dispatched + limit->slice_quota - 1) /
35
- limit->slice_quota;
36
+ /* Quota exceeded. Wait based on the excess amount and then start a new
37
+ * slice. */
38
+ delay_slices = (double)limit->dispatched / limit->slice_quota;
39
limit->slice_end_time = limit->slice_start_time +
40
- delay_slices * limit->slice_ns;
41
+ (uint64_t)(delay_slices * limit->slice_ns);
42
return limit->slice_end_time - now;
43
}
44
45
--
46
2.14.3
47
48
diff view generated by jsdifflib
[Qemu-devel] [PATCH 5/5] misc: fix spelling
Deleted patch
1
From: Marc-André Lureau <marcandre.lureau@redhat.com>
2
1
3
s/pupulate/populate
4
5
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
6
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
7
Message-id: 20180208162447.10851-1-marcandre.lureau@redhat.com
8
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
9
---
10
util/qemu-coroutine-lock.c | 2 +-
11
1 file changed, 1 insertion(+), 1 deletion(-)
12
13
diff --git a/util/qemu-coroutine-lock.c b/util/qemu-coroutine-lock.c
14
index XXXXXXX..XXXXXXX 100644
15
--- a/util/qemu-coroutine-lock.c
16
+++ b/util/qemu-coroutine-lock.c
17
@@ -XXX,XX +XXX,XX @@ void qemu_co_queue_run_restart(Coroutine *co)
18
* invalid memory. Therefore, use a temporary queue and do not touch
19
* the "co" coroutine as soon as you enter another one.
20
*
21
- * In its turn resumed "co" can pupulate "co_queue_wakeup" queue with
22
+ * In its turn resumed "co" can populate "co_queue_wakeup" queue with
23
* new coroutines to be woken up. The caller, who has resumed "co",
24
* will be responsible for traversing the same queue, which may cause
25
* a different wakeup order but not any missing wakeups.
26
--
27
2.14.3
28
29
diff view generated by jsdifflib

Patchew 2.1-devel-b67e4c2

© 2016 - 2025 Red Hat, Inc.