Re: [PATCH RFC 1/1] block/export: FUSE-over-io_uring Support for QEMU FUSE Exports

[Brian Song <hibriansong@gmail.com>]

On 12/10/25 5:10 PM, Bernd Schubert wrote:
> Hi Brian et al,
> 
>> +static void fuse_uring_cqe_handler(CqeHandler *cqe_handler)
>> +{
>> +    FuseRingEnt *ent = container_of(cqe_handler, FuseRingEnt, fuse_cqe_handler);
>> +    FuseQueue *q = ent->q;
>> +    Coroutine *co;
>> +    FuseExport *exp = ent->q->exp;
>> +
>> +    int err = cqe_handler->cqe.res;
>> +    if (err != 0) {
>> +        /* TODO end_conn support */
>> +
>> +        /* -ENOTCONN is ok on umount  */
>> +        if (err != -EINTR && err != -EOPNOTSUPP &&
>> +            err != -EAGAIN && err != -ENOTCONN) {
>> +            fuse_export_halt(exp);
> 
> 
> what actually happened to these patches? I don't see it merged in the
> qemu repo? There is an issue here. I know we had discussed -EAGAIN and
> -EINTR. Actually this needs to be specially handled, at least
> FUSE_IO_URING_CMD_REGISTER, because it can arrive before FUSE_INIT
> is processed and then kernel will return -EAGAIN.
> 
> https://github.com/libfuse/libfuse/pull/1387/commits/34f3dec97cce508642679cb1b874e99ed36d8158
> 
> I need to split this up it also includes some other refactoring.
> 
> 
> Thanks,
> Bernd

Thanks for reaching out. I'm graduating at the end of the year, so after 
GSoC I got really busy with finishing school and job hunting. I've got 
some time these couple of days, so I'll go through the feedback on patch 
v3 from the mailing list and wrap up the remaining work

Thanks,
Brian