Skip to content

Commit f2ebb8e

Browse files
committed
s390/virtio_ccw: Don't allocate/assign airqs for non-existing queues
jira LE-3587 Rebuild_History Non-Buildable kernel-4.18.0-553.62.1.el8_10 commit-author David Hildenbrand <[email protected]> commit 2ccd42b Empty-Commit: Cherry-Pick Conflicts during history rebuild. Will be included in final tarball splat. Ref for failed cherry-pick at: ciq/ciq_backports/kernel-4.18.0-553.62.1.el8_10/2ccd42b9.failed If we finds a vq without a name in our input array in virtio_ccw_find_vqs(), we treat it as "non-existing" and set the vq pointer to NULL; we will not call virtio_ccw_setup_vq() to allocate/setup a vq. Consequently, we create only a queue if it actually exists (name != NULL) and assign an incremental queue index to each such existing queue. However, in virtio_ccw_register_adapter_ind()->get_airq_indicator() we will not ignore these "non-existing queues", but instead assign an airq indicator to them. Besides never releasing them in virtio_ccw_drop_indicators() (because there is no virtqueue), the bigger issue seems to be that there will be a disagreement between the device and the Linux guest about the airq indicator to be used for notifying a queue, because the indicator bit for adapter I/O interrupt is derived from the queue index. The virtio spec states under "Setting Up Two-Stage Queue Indicators": ... indicator contains the guest address of an area wherein the indicators for the devices are contained, starting at bit_nr, one bit per virtqueue of the device. And further in "Notification via Adapter I/O Interrupts": For notifying the driver of virtqueue buffers, the device sets the bit in the guest-provided indicator area at the corresponding offset. For example, QEMU uses in virtio_ccw_notify() the queue index (passed as "vector") to select the relevant indicator bit. If a queue does not exist, it does not have a corresponding indicator bit assigned, because it effectively doesn't have a queue index. Using a virtio-balloon-ccw device under QEMU with free-page-hinting disabled ("free-page-hint=off") but free-page-reporting enabled ("free-page-reporting=on") will result in free page reporting not working as expected: in the virtio_balloon driver, we'll be stuck forever in virtballoon_free_page_report()->wait_event(), because the waitqueue will not be woken up as the notification from the device is lost: it would use the wrong indicator bit. Free page reporting stops working and we get splats (when configured to detect hung wqs) like: INFO: task kworker/1:3:463 blocked for more than 61 seconds. Not tainted 6.14.0 #4 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. task:kworker/1:3 [...] Workqueue: events page_reporting_process Call Trace: [<000002f404e6dfb2>] __schedule+0x402/0x1640 [<000002f404e6f22e>] schedule+0x3e/0xe0 [<000002f3846a88fa>] virtballoon_free_page_report+0xaa/0x110 [virtio_balloon] [<000002f40435c8a4>] page_reporting_process+0x2e4/0x740 [<000002f403fd3ee2>] process_one_work+0x1c2/0x400 [<000002f403fd4b96>] worker_thread+0x296/0x420 [<000002f403fe10b4>] kthread+0x124/0x290 [<000002f403f4e0dc>] __ret_from_fork+0x3c/0x60 [<000002f404e77272>] ret_from_fork+0xa/0x38 There was recently a discussion [1] whether the "holes" should be treated differently again, effectively assigning also non-existing queues a queue index: that should also fix the issue, but requires other workarounds to not break existing setups. Let's fix it without affecting existing setups for now by properly ignoring the non-existing queues, so the indicator bits will match the queue indexes. [1] https://lore.kernel.org/all/[email protected]/ Fixes: a229989 ("virtio: don't allocate vqs when names[i] = NULL") Reported-by: Chandra Merla <[email protected]> Cc: [email protected] Signed-off-by: David Hildenbrand <[email protected]> Tested-by: Thomas Huth <[email protected]> Reviewed-by: Thomas Huth <[email protected]> Reviewed-by: Cornelia Huck <[email protected]> Acked-by: Michael S. Tsirkin <[email protected]> Acked-by: Christian Borntraeger <[email protected]> Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Heiko Carstens <[email protected]> (cherry picked from commit 2ccd42b) Signed-off-by: Jonathan Maple <[email protected]> # Conflicts: # drivers/s390/virtio/virtio_ccw.c
1 parent 809952d commit f2ebb8e

File tree

1 file changed

+158
-0
lines changed

1 file changed

+158
-0
lines changed
Lines changed: 158 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,158 @@
1+
s390/virtio_ccw: Don't allocate/assign airqs for non-existing queues
2+
3+
jira LE-3587
4+
Rebuild_History Non-Buildable kernel-4.18.0-553.62.1.el8_10
5+
commit-author David Hildenbrand <[email protected]>
6+
commit 2ccd42b959aaf490333dbd3b9b102eaf295c036a
7+
Empty-Commit: Cherry-Pick Conflicts during history rebuild.
8+
Will be included in final tarball splat. Ref for failed cherry-pick at:
9+
ciq/ciq_backports/kernel-4.18.0-553.62.1.el8_10/2ccd42b9.failed
10+
11+
If we finds a vq without a name in our input array in
12+
virtio_ccw_find_vqs(), we treat it as "non-existing" and set the vq pointer
13+
to NULL; we will not call virtio_ccw_setup_vq() to allocate/setup a vq.
14+
15+
Consequently, we create only a queue if it actually exists (name != NULL)
16+
and assign an incremental queue index to each such existing queue.
17+
18+
However, in virtio_ccw_register_adapter_ind()->get_airq_indicator() we
19+
will not ignore these "non-existing queues", but instead assign an airq
20+
indicator to them.
21+
22+
Besides never releasing them in virtio_ccw_drop_indicators() (because
23+
there is no virtqueue), the bigger issue seems to be that there will be a
24+
disagreement between the device and the Linux guest about the airq
25+
indicator to be used for notifying a queue, because the indicator bit
26+
for adapter I/O interrupt is derived from the queue index.
27+
28+
The virtio spec states under "Setting Up Two-Stage Queue Indicators":
29+
30+
... indicator contains the guest address of an area wherein the
31+
indicators for the devices are contained, starting at bit_nr, one
32+
bit per virtqueue of the device.
33+
34+
And further in "Notification via Adapter I/O Interrupts":
35+
36+
For notifying the driver of virtqueue buffers, the device sets the
37+
bit in the guest-provided indicator area at the corresponding
38+
offset.
39+
40+
For example, QEMU uses in virtio_ccw_notify() the queue index (passed as
41+
"vector") to select the relevant indicator bit. If a queue does not exist,
42+
it does not have a corresponding indicator bit assigned, because it
43+
effectively doesn't have a queue index.
44+
45+
Using a virtio-balloon-ccw device under QEMU with free-page-hinting
46+
disabled ("free-page-hint=off") but free-page-reporting enabled
47+
("free-page-reporting=on") will result in free page reporting
48+
not working as expected: in the virtio_balloon driver, we'll be stuck
49+
forever in virtballoon_free_page_report()->wait_event(), because the
50+
waitqueue will not be woken up as the notification from the device is
51+
lost: it would use the wrong indicator bit.
52+
53+
Free page reporting stops working and we get splats (when configured to
54+
detect hung wqs) like:
55+
56+
INFO: task kworker/1:3:463 blocked for more than 61 seconds.
57+
Not tainted 6.14.0 #4
58+
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
59+
task:kworker/1:3 [...]
60+
Workqueue: events page_reporting_process
61+
Call Trace:
62+
[<000002f404e6dfb2>] __schedule+0x402/0x1640
63+
[<000002f404e6f22e>] schedule+0x3e/0xe0
64+
[<000002f3846a88fa>] virtballoon_free_page_report+0xaa/0x110 [virtio_balloon]
65+
[<000002f40435c8a4>] page_reporting_process+0x2e4/0x740
66+
[<000002f403fd3ee2>] process_one_work+0x1c2/0x400
67+
[<000002f403fd4b96>] worker_thread+0x296/0x420
68+
[<000002f403fe10b4>] kthread+0x124/0x290
69+
[<000002f403f4e0dc>] __ret_from_fork+0x3c/0x60
70+
[<000002f404e77272>] ret_from_fork+0xa/0x38
71+
72+
There was recently a discussion [1] whether the "holes" should be
73+
treated differently again, effectively assigning also non-existing
74+
queues a queue index: that should also fix the issue, but requires other
75+
workarounds to not break existing setups.
76+
77+
Let's fix it without affecting existing setups for now by properly ignoring
78+
the non-existing queues, so the indicator bits will match the queue
79+
indexes.
80+
81+
[1] https://lore.kernel.org/all/[email protected]/
82+
83+
Fixes: a229989d975e ("virtio: don't allocate vqs when names[i] = NULL")
84+
Reported-by: Chandra Merla <[email protected]>
85+
86+
Signed-off-by: David Hildenbrand <[email protected]>
87+
Tested-by: Thomas Huth <[email protected]>
88+
Reviewed-by: Thomas Huth <[email protected]>
89+
Reviewed-by: Cornelia Huck <[email protected]>
90+
Acked-by: Michael S. Tsirkin <[email protected]>
91+
Acked-by: Christian Borntraeger <[email protected]>
92+
Link: https://lore.kernel.org/r/[email protected]
93+
Signed-off-by: Heiko Carstens <[email protected]>
94+
(cherry picked from commit 2ccd42b959aaf490333dbd3b9b102eaf295c036a)
95+
Signed-off-by: Jonathan Maple <[email protected]>
96+
97+
# Conflicts:
98+
# drivers/s390/virtio/virtio_ccw.c
99+
diff --cc drivers/s390/virtio/virtio_ccw.c
100+
index eb609e021105,4904b831c0a7..000000000000
101+
--- a/drivers/s390/virtio/virtio_ccw.c
102+
+++ b/drivers/s390/virtio/virtio_ccw.c
103+
@@@ -260,14 -299,20 +260,20 @@@ static struct airq_info *new_airq_info(
104+
return info;
105+
}
106+
107+
-static unsigned long *get_airq_indicator(struct virtqueue *vqs[], int nvqs,
108+
- u64 *first, void **airq_info)
109+
+static unsigned long get_airq_indicator(struct virtqueue *vqs[], int nvqs,
110+
+ u64 *first, void **airq_info)
111+
{
112+
- int i, j;
113+
+ int i, j, queue_idx, highest_queue_idx = -1;
114+
struct airq_info *info;
115+
- unsigned long *indicator_addr = NULL;
116+
+ unsigned long indicator_addr = 0;
117+
unsigned long bit, flags;
118+
119+
+ /* Array entries without an actual queue pointer must be ignored. */
120+
+ for (i = 0; i < nvqs; i++) {
121+
+ if (vqs[i])
122+
+ highest_queue_idx++;
123+
+ }
124+
+
125+
for (i = 0; i < MAX_AIRQ_AREAS && !indicator_addr; i++) {
126+
mutex_lock(&airq_areas_lock);
127+
if (!airq_areas[i])
128+
@@@ -275,9 -320,9 +281,9 @@@
129+
info = airq_areas[i];
130+
mutex_unlock(&airq_areas_lock);
131+
if (!info)
132+
- return NULL;
133+
+ return 0;
134+
write_lock_irqsave(&info->lock, flags);
135+
- bit = airq_iv_alloc(info->aiv, nvqs);
136+
+ bit = airq_iv_alloc(info->aiv, highest_queue_idx + 1);
137+
if (bit == -1UL) {
138+
/* Not enough vacancies. */
139+
write_unlock_irqrestore(&info->lock, flags);
140+
@@@ -285,9 -330,11 +291,17 @@@
141+
}
142+
*first = bit;
143+
*airq_info = info;
144+
++<<<<<<< HEAD
145+
+ indicator_addr = (unsigned long)info->aiv->vector;
146+
+ for (j = 0; j < nvqs; j++) {
147+
+ airq_iv_set_ptr(info->aiv, bit + j,
148+
++=======
149+
+ indicator_addr = info->aiv->vector;
150+
+ for (j = 0, queue_idx = 0; j < nvqs; j++) {
151+
+ if (!vqs[j])
152+
+ continue;
153+
+ airq_iv_set_ptr(info->aiv, bit + queue_idx++,
154+
++>>>>>>> 2ccd42b959aa (s390/virtio_ccw: Don't allocate/assign airqs for non-existing queues)
155+
(unsigned long)vqs[j]);
156+
}
157+
write_unlock_irqrestore(&info->lock, flags);
158+
* Unmerged path drivers/s390/virtio/virtio_ccw.c

0 commit comments

Comments
 (0)