blob: 6faeafac6ca56ee1b109d68b57957a3539ee299e (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
|
device: 0.936
other: 0.770
network: 0.658
graphic: 0.598
socket: 0.533
files: 0.513
performance: 0.478
semantic: 0.434
vnc: 0.389
permissions: 0.357
boot: 0.353
PID: 0.308
debug: 0.295
KVM: 0.086
Ask to extend vhost-user protocol to carry implementation defined error contexts
Additional information:
I am working on the Google [crosvm](https://chromium.googlesource.com/crosvm/crosvm/) project, which implements some `vhost-user` clients/servers defined by [this QEMU doc](https://qemu-project.gitlab.io/qemu/interop/vhost-user.html). I am wondering if we could add a protocol feature/protocol header flag bit to allow the payload of the reply to carry detailed implementation defined error contexts?
Specifically, I am working on the `vhost-user-gpu` device, which needs to send some memory mapping request to the frontend(the main process where VCPU lives), so that we can map some GPU memory to the guest. We are trying to diagnose a bug where the frontend can sometimes fail to perform the operation. However, we don't have access to the logs on the main process, so we are left with only very limited information on the `vhost-user-gpu` process. It could be helpful if we could send detailed implementation defined error contexts in the payload of the reply.
I am wondering in order for the upstream QEMU to accept such "spec" change to the `vhost-user` protocol, what the process should be like? Thanks.
|