Cross linking the device API doc #3
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Merged
Closed
rouming
added a commit
to rouming/eve
that referenced
this pull request
Nov 8, 2023
Patch drains possible events on qmp socket until true "qmp_capabilities"
response is received.
This targets a nasty and rare problem in "Connect(), Run()" sequence,
when asynchronous QEMU "event" just after the "qmp_capabilities" request
is accepted as response and further any QMP request gets completed with
response from the preceding "qmp_capabilities" response.
The following QMP protocol sequence is possible:
1 read {"QMP": {"version": {"qemu": {"micro": 0, "minor": 1, "major": 5}, ...}"
2 write {"execute":"qmp_capabilities"}
3 read {"timestamp": {"seconds": xxx, "microseconds": xxx}, "event": "..."}
4 read {"return": {}}
the lf-edge#3 is unexpected by the current Connect() implementation and "event" is
considered as a proper response on "qmp_capabilities", in other turn lf-edge#4 is
read in the go.mos.listen() and immediately pushed to the stream channel,
so any further QMP command (Run() call) will be immediately completed by
an empty response from line lf-edge#4.
The described problem of unexpected empty response line was observed
on this code qmp.SocketMonitor sequence:
Connect()
Run('{"execute":"query-status"}') <<< Returns empty response
Disconnect()
The problem is very rare and was observed ~5 times over a fairly long
period of time (several months), which corresponds to nature of the
described rare protocol race.
The current patch was tested on modified QEMU, where an aritifical
sleep() was introduced in the qmp_marshal_qmp_capabilities() call
just right after the qmp_qmp_capabilities() was invoked, so all
further events can be accepted by the QMP socket:
--- qapi/qapi-commands-control.c 2023-11-08 08:55:16.209007741 +0100
+++ qapi/qapi-commands-control.c.orig 2023-11-08 08:55:13.929005997 +0100
@@ -42,10 +42,6 @@
qmp_qmp_capabilities(arg.has_enable, arg.enable, &err);
+ printf(">>> BEFORE SLEEP 10\n");
+ sleep(10);
+ printf(">>> AFTER SLEEP 10\n");
+
error_propagate(errp, err);
out:
Any QMP event can be freely received by the QMP client, while the execution
flow of the qmp_marshal_qmp_capabilities() was interrupted or scheduled out.
The fix of the described is simple: read from the QMP socket until response
is received and drop all possible events.
Signed-off-by: Roman Penyaev <r.peniaev@gmail.com>
rouming
added a commit
to rouming/eve
that referenced
this pull request
Nov 8, 2023
Patch drains possible events on qmp socket until true "qmp_capabilities"
response is received.
This targets a nasty and rare problem in "Connect(), Run()" sequence,
when asynchronous QEMU "event" just after the "qmp_capabilities" request
is accepted as response and further any QMP request gets completed with
response from the preceding "qmp_capabilities" response.
The following QMP protocol sequence is possible:
1 read {"QMP": {"version": {"qemu": {"micro": 0, "minor": 1, "major": 5}, ...}"
2 write {"execute":"qmp_capabilities"}
3 read {"timestamp": {"seconds": xxx, "microseconds": xxx}, "event": "..."}
4 read {"return": {}}
the lf-edge#3 is unexpected by the current Connect() implementation and "event" is
considered as a proper response on "qmp_capabilities", in other turn lf-edge#4 is
read in the go.mos.listen() and immediately pushed to the stream channel,
so any further QMP command (Run() call) will be immediately completed by
an empty response from line lf-edge#4.
The described problem of unexpected empty response line was observed
on this code qmp.SocketMonitor sequence:
Connect()
Run('{"execute":"query-status"}') <<< Returns empty response
Disconnect()
The problem is very rare and was observed ~5 times over a fairly long
period of time (several months), which corresponds to nature of the
described rare protocol race.
The current patch was tested on modified QEMU, where an aritifical
sleep() was introduced in the qmp_marshal_qmp_capabilities() call
just right after the qmp_qmp_capabilities() was invoked, so all
further events can be accepted by the QMP socket:
--- qapi/qapi-commands-control.c 2023-11-08 08:55:16.209007741 +0100
+++ qapi/qapi-commands-control.c.orig 2023-11-08 08:55:13.929005997 +0100
@@ -42,10 +42,6 @@
qmp_qmp_capabilities(arg.has_enable, arg.enable, &err);
+ printf(">>> BEFORE SLEEP 10\n");
+ sleep(10);
+ printf(">>> AFTER SLEEP 10\n");
+
error_propagate(errp, err);
out:
Any QMP event can be freely received by the QMP client, while the execution
flow of the qmp_marshal_qmp_capabilities() was interrupted or scheduled out.
The fix of the described is simple: read from the QMP socket until response
is received and drop all possible events.
Signed-off-by: Roman Penyaev <r.peniaev@gmail.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adding a link to the controller API docs.