1
The following changes since commit 41feb5b955f0d7c9d071b2c5adbc404ae2895c7a:
1
The following changes since commit 887cba855bb6ff4775256f7968409281350b568c:
2
2
3
Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-3.0-pull-request' into staging (2018-06-05 10:38:33 +0100)
3
configure: Fix cross-building for RISCV host (v5) (2023-07-11 17:56:09 +0100)
4
4
5
are available in the git repository at:
5
are available in the Git repository at:
6
6
7
git://github.com/codyprime/qemu-kvm-jtc.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 68acc99f143b60ec4faa2903065b187d4d3c4bf3:
9
for you to fetch changes up to 75dcb4d790bbe5327169fd72b185960ca58e2fa6:
10
10
11
sheepdog: remove huge BSS object (2018-06-05 10:15:12 -0400)
11
virtio-blk: fix host notifier issues during dataplane start/stop (2023-07-12 15:20:32 -0400)
12
12
13
----------------------------------------------------------------
13
----------------------------------------------------------------
14
Sheepdog patches
14
Pull request
15
15
----------------------------------------------------------------
16
----------------------------------------------------------------
16
17
17
Paolo Bonzini (2):
18
Stefan Hajnoczi (1):
18
sheepdog: cleanup repeated expression
19
virtio-blk: fix host notifier issues during dataplane start/stop
19
sheepdog: remove huge BSS object
20
20
21
block/sheepdog.c | 28 +++++++++++++++-------------
21
hw/block/dataplane/virtio-blk.c | 67 +++++++++++++++++++--------------
22
1 file changed, 15 insertions(+), 13 deletions(-)
22
1 file changed, 38 insertions(+), 29 deletions(-)
23
23
24
--
24
--
25
2.13.6
25
2.40.1
26
27
diff view generated by jsdifflib
1
From: Paolo Bonzini <pbonzini@redhat.com>
1
The main loop thread can consume 100% CPU when using --device
2
virtio-blk-pci,iothread=<iothread>. ppoll() constantly returns but
3
reading virtqueue host notifiers fails with EAGAIN. The file descriptors
4
are stale and remain registered with the AioContext because of bugs in
5
the virtio-blk dataplane start/stop code.
2
6
3
The expression "SD_INODE_SIZE - sizeof(inode.data_vdi_id)" already has a macro
7
The problem is that the dataplane start/stop code involves drain
4
defined for the same value (though with a nicer definition using offsetof).
8
operations, which call virtio_blk_drained_begin() and
5
Replace it.
9
virtio_blk_drained_end() at points where the host notifier is not
10
operational:
11
- In virtio_blk_data_plane_start(), blk_set_aio_context() drains after
12
vblk->dataplane_started has been set to true but the host notifier has
13
not been attached yet.
14
- In virtio_blk_data_plane_stop(), blk_drain() and blk_set_aio_context()
15
drain after the host notifier has already been detached but with
16
vblk->dataplane_started still set to true.
6
17
7
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
18
I would like to simplify ->ioeventfd_start/stop() to avoid interactions
8
Reviewed-by: Fam Zheng <famz@redhat.com>
19
with drain entirely, but couldn't find a way to do that. Instead, this
9
Reviewed-by: Jeff Cody <jcody@redhat.com>
20
patch accepts the fragile nature of the code and reorders it so that
10
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
21
vblk->dataplane_started is false during drain operations. This way the
11
Message-Id: <20180523160721.14018-2-pbonzini@redhat.com>
22
virtio_blk_drained_begin() and virtio_blk_drained_end() calls don't
12
Signed-off-by: Jeff Cody <jcody@redhat.com>
23
touch the host notifier. The result is that
24
virtio_blk_data_plane_start() and virtio_blk_data_plane_stop() have
25
complete control over the host notifier and stale file descriptors are
26
no longer left in the AioContext.
27
28
This patch fixes the 100% CPU consumption in the main loop thread and
29
correctly moves host notifier processing to the IOThread.
30
31
Fixes: 1665d9326fd2 ("virtio-blk: implement BlockDevOps->drained_begin()")
32
Reported-by: Lukáš Doktor <ldoktor@redhat.com>
33
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
34
Tested-by: Lukas Doktor <ldoktor@redhat.com>
35
Message-id: 20230704151527.193586-1-stefanha@redhat.com
36
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
13
---
37
---
14
block/sheepdog.c | 6 +++---
38
hw/block/dataplane/virtio-blk.c | 67 +++++++++++++++++++--------------
15
1 file changed, 3 insertions(+), 3 deletions(-)
39
1 file changed, 38 insertions(+), 29 deletions(-)
16
40
17
diff --git a/block/sheepdog.c b/block/sheepdog.c
41
diff --git a/hw/block/dataplane/virtio-blk.c b/hw/block/dataplane/virtio-blk.c
18
index XXXXXXX..XXXXXXX 100644
42
index XXXXXXX..XXXXXXX 100644
19
--- a/block/sheepdog.c
43
--- a/hw/block/dataplane/virtio-blk.c
20
+++ b/block/sheepdog.c
44
+++ b/hw/block/dataplane/virtio-blk.c
21
@@ -XXX,XX +XXX,XX @@ static int sd_truncate(BlockDriverState *bs, int64_t offset,
45
@@ -XXX,XX +XXX,XX @@ int virtio_blk_data_plane_start(VirtIODevice *vdev)
46
47
memory_region_transaction_commit();
48
49
- /*
50
- * These fields are visible to the IOThread so we rely on implicit barriers
51
- * in aio_context_acquire() on the write side and aio_notify_accept() on
52
- * the read side.
53
- */
54
- s->starting = false;
55
- vblk->dataplane_started = true;
56
trace_virtio_blk_data_plane_start(s);
57
58
old_context = blk_get_aio_context(s->conf->conf.blk);
59
@@ -XXX,XX +XXX,XX @@ int virtio_blk_data_plane_start(VirtIODevice *vdev)
60
event_notifier_set(virtio_queue_get_host_notifier(vq));
22
}
61
}
23
62
24
/* we don't need to update entire object */
63
+ /*
25
- datalen = SD_INODE_SIZE - sizeof(s->inode.data_vdi_id);
64
+ * These fields must be visible to the IOThread when it processes the
26
+ datalen = SD_INODE_HEADER_SIZE;
65
+ * virtqueue, otherwise it will think dataplane has not started yet.
27
s->inode.vdi_size = offset;
66
+ *
28
ret = write_object(fd, s->bs, (char *)&s->inode,
67
+ * Make sure ->dataplane_started is false when blk_set_aio_context() is
29
vid_to_vdi_oid(s->inode.vdi_id), s->inode.nr_copies,
68
+ * called above so that draining does not cause the host notifier to be
30
@@ -XXX,XX +XXX,XX @@ static int sd_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info)
69
+ * detached/attached prematurely.
31
*/
70
+ */
32
strncpy(s->inode.tag, sn_info->name, sizeof(s->inode.tag));
71
+ s->starting = false;
33
/* we don't need to update entire object */
72
+ vblk->dataplane_started = true;
34
- datalen = SD_INODE_SIZE - sizeof(s->inode.data_vdi_id);
73
+ smp_wmb(); /* paired with aio_notify_accept() on the read side */
35
+ datalen = SD_INODE_HEADER_SIZE;
74
+
36
inode = g_malloc(datalen);
75
/* Get this show started by hooking up our callbacks */
37
76
if (!blk_in_drain(s->conf->conf.blk)) {
38
/* refresh inode. */
77
aio_context_acquire(s->ctx);
39
@@ -XXX,XX +XXX,XX @@ static int sd_snapshot_list(BlockDriverState *bs, QEMUSnapshotInfo **psn_tab)
78
@@ -XXX,XX +XXX,XX @@ int virtio_blk_data_plane_start(VirtIODevice *vdev)
40
/* we don't need to read entire object */
79
fail_guest_notifiers:
41
ret = read_object(fd, s->bs, (char *)&inode,
80
vblk->dataplane_disabled = true;
42
vid_to_vdi_oid(vid),
81
s->starting = false;
43
- 0, SD_INODE_SIZE - sizeof(inode.data_vdi_id), 0,
82
- vblk->dataplane_started = true;
44
+ 0, SD_INODE_HEADER_SIZE, 0,
83
return -ENOSYS;
45
s->cache_flags);
84
}
46
85
47
if (ret) {
86
@@ -XXX,XX +XXX,XX @@ void virtio_blk_data_plane_stop(VirtIODevice *vdev)
87
aio_wait_bh_oneshot(s->ctx, virtio_blk_data_plane_stop_bh, s);
88
}
89
90
+ /*
91
+ * Batch all the host notifiers in a single transaction to avoid
92
+ * quadratic time complexity in address_space_update_ioeventfds().
93
+ */
94
+ memory_region_transaction_begin();
95
+
96
+ for (i = 0; i < nvqs; i++) {
97
+ virtio_bus_set_host_notifier(VIRTIO_BUS(qbus), i, false);
98
+ }
99
+
100
+ /*
101
+ * The transaction expects the ioeventfds to be open when it
102
+ * commits. Do it now, before the cleanup loop.
103
+ */
104
+ memory_region_transaction_commit();
105
+
106
+ for (i = 0; i < nvqs; i++) {
107
+ virtio_bus_cleanup_host_notifier(VIRTIO_BUS(qbus), i);
108
+ }
109
+
110
+ /*
111
+ * Set ->dataplane_started to false before draining so that host notifiers
112
+ * are not detached/attached anymore.
113
+ */
114
+ vblk->dataplane_started = false;
115
+
116
aio_context_acquire(s->ctx);
117
118
/* Wait for virtio_blk_dma_restart_bh() and in flight I/O to complete */
119
@@ -XXX,XX +XXX,XX @@ void virtio_blk_data_plane_stop(VirtIODevice *vdev)
120
121
aio_context_release(s->ctx);
122
123
- /*
124
- * Batch all the host notifiers in a single transaction to avoid
125
- * quadratic time complexity in address_space_update_ioeventfds().
126
- */
127
- memory_region_transaction_begin();
128
-
129
- for (i = 0; i < nvqs; i++) {
130
- virtio_bus_set_host_notifier(VIRTIO_BUS(qbus), i, false);
131
- }
132
-
133
- /*
134
- * The transaction expects the ioeventfds to be open when it
135
- * commits. Do it now, before the cleanup loop.
136
- */
137
- memory_region_transaction_commit();
138
-
139
- for (i = 0; i < nvqs; i++) {
140
- virtio_bus_cleanup_host_notifier(VIRTIO_BUS(qbus), i);
141
- }
142
-
143
qemu_bh_cancel(s->bh);
144
notify_guest_bh(s); /* final chance to notify guest */
145
146
/* Clean up guest notifier (irq) */
147
k->set_guest_notifiers(qbus->parent, nvqs, false);
148
149
- vblk->dataplane_started = false;
150
s->stopping = false;
151
}
48
--
152
--
49
2.13.6
153
2.40.1
50
154
51
155
diff view generated by jsdifflib
Deleted patch
1
From: Paolo Bonzini <pbonzini@redhat.com>
2
1
3
block/sheepdog.o has a 4M static variable that is 90% of QEMU's whole .bss
4
section. Replace it with a heap-allocated block, and make it smaller too
5
since only the inode header is actually being used.
6
7
bss size goes down from 4464280 to 269976.
8
9
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
10
Reviewed-by: Jeff Cody <jcody@redhat.com>
11
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12
Message-Id: <20180523160721.14018-3-pbonzini@redhat.com>
13
Signed-off-by: Jeff Cody <jcody@redhat.com>
14
---
15
block/sheepdog.c | 22 ++++++++++++----------
16
1 file changed, 12 insertions(+), 10 deletions(-)
17
18
diff --git a/block/sheepdog.c b/block/sheepdog.c
19
index XXXXXXX..XXXXXXX 100644
20
--- a/block/sheepdog.c
21
+++ b/block/sheepdog.c
22
@@ -XXX,XX +XXX,XX @@ static int sd_snapshot_list(BlockDriverState *bs, QEMUSnapshotInfo **psn_tab)
23
QEMUSnapshotInfo *sn_tab = NULL;
24
unsigned wlen, rlen;
25
int found = 0;
26
- static SheepdogInode inode;
27
+ SheepdogInode *inode;
28
unsigned long *vdi_inuse;
29
unsigned int start_nr;
30
uint64_t hval;
31
uint32_t vid;
32
33
vdi_inuse = g_malloc(max);
34
+ inode = g_malloc(SD_INODE_HEADER_SIZE);
35
36
fd = connect_to_sdog(s, &local_err);
37
if (fd < 0) {
38
@@ -XXX,XX +XXX,XX @@ static int sd_snapshot_list(BlockDriverState *bs, QEMUSnapshotInfo **psn_tab)
39
}
40
41
/* we don't need to read entire object */
42
- ret = read_object(fd, s->bs, (char *)&inode,
43
+ ret = read_object(fd, s->bs, (char *)inode,
44
vid_to_vdi_oid(vid),
45
0, SD_INODE_HEADER_SIZE, 0,
46
s->cache_flags);
47
@@ -XXX,XX +XXX,XX @@ static int sd_snapshot_list(BlockDriverState *bs, QEMUSnapshotInfo **psn_tab)
48
continue;
49
}
50
51
- if (!strcmp(inode.name, s->name) && is_snapshot(&inode)) {
52
- sn_tab[found].date_sec = inode.snap_ctime >> 32;
53
- sn_tab[found].date_nsec = inode.snap_ctime & 0xffffffff;
54
- sn_tab[found].vm_state_size = inode.vm_state_size;
55
- sn_tab[found].vm_clock_nsec = inode.vm_clock_nsec;
56
+ if (!strcmp(inode->name, s->name) && is_snapshot(inode)) {
57
+ sn_tab[found].date_sec = inode->snap_ctime >> 32;
58
+ sn_tab[found].date_nsec = inode->snap_ctime & 0xffffffff;
59
+ sn_tab[found].vm_state_size = inode->vm_state_size;
60
+ sn_tab[found].vm_clock_nsec = inode->vm_clock_nsec;
61
62
snprintf(sn_tab[found].id_str, sizeof(sn_tab[found].id_str),
63
- "%" PRIu32, inode.snap_id);
64
+ "%" PRIu32, inode->snap_id);
65
pstrcpy(sn_tab[found].name,
66
- MIN(sizeof(sn_tab[found].name), sizeof(inode.tag)),
67
- inode.tag);
68
+ MIN(sizeof(sn_tab[found].name), sizeof(inode->tag)),
69
+ inode->tag);
70
found++;
71
}
72
}
73
@@ -XXX,XX +XXX,XX @@ out:
74
*psn_tab = sn_tab;
75
76
g_free(vdi_inuse);
77
+ g_free(inode);
78
79
if (ret < 0) {
80
return ret;
81
--
82
2.13.6
83
84
diff view generated by jsdifflib