Re: [PATCH v7 4/8] block: add block layer APIs resembling Linux ZonedBlockDevice ioctls

[Stefan Hajnoczi <stefanha@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1664 bytes --]

On Tue, Aug 23, 2022 at 12:12:44PM +0800, Sam Li wrote:
> Stefan Hajnoczi <stefanha@redhat.com> 于2022年8月23日周二 08:49写道：
> > On Tue, Aug 16, 2022 at 02:25:18PM +0800, Sam Li wrote:
> > > +static int handle_aiocb_zone_report(void *opaque) {
> > > +#if defined(CONFIG_BLKZONED)
> > > +    RawPosixAIOData *aiocb = opaque;
> > > +    int fd = aiocb->aio_fildes;
> > > +    unsigned int *nr_zones = aiocb->zone_report.nr_zones;
> > > +    BlockZoneDescriptor *zones = aiocb->zone_report.zones;
> > > +    int64_t sector = aiocb->aio_offset;
> > > +
> > > +    struct blk_zone *blkz;
> > > +    int64_t rep_size;
> > > +    unsigned int nrz;
> > > +    int ret, n = 0, i = 0;
> > > +
> > > +    nrz = *nr_zones;
> > > +    rep_size = sizeof(struct blk_zone_report) + nrz * sizeof(struct blk_zone);
> > > +    g_autofree struct blk_zone_report *rep = NULL;
> > > +    rep = g_malloc(rep_size);
> > > +
> > > +    blkz = (struct blk_zone *)(rep + 1);
> > > +    while (n < nrz) {
> > > +        memset(rep, 0, rep_size);
> > > +        rep->sector = sector;
> > > +        rep->nr_zones = nrz - n;
> > > +
> > > +        ret = ioctl(fd, BLKREPORTZONE, rep);
> >
> > Does this ioctl() need "do { ... } while (ret == -1 && errno == EINTR)"?
> 
> No? We discussed this before. I guess even EINTR should be propagated
> back to the guest. Maybe Damien can talk more about why.

No, EINTR is an internal error that must be handled by QEMU. It means
the QEMU process' syscall was interrupted by a signal and the syscall
must be retried. The guest shouldn't see EINTR (and there is no
virtio-blk error code defined for it).

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]