Re: [PATCH 0/6] vhost: Reset batched descriptors on SET_VRING_BASE call

From: Eugenio Perez Martin
Date: Mon Mar 30 2020 - 05:16:03 EST


On Mon, Mar 30, 2020 at 9:34 AM Christian Borntraeger
<borntraeger@xxxxxxxxxx> wrote:
>
>
>
> On 30.03.20 09:18, Eugenio Perez Martin wrote:
> > On Mon, Mar 30, 2020 at 9:14 AM Christian Borntraeger
> > <borntraeger@xxxxxxxxxx> wrote:
> >>
> >>
> >> On 29.03.20 13:33, Eugenio PÃrez wrote:
> >>> Vhost did not reset properly the batched descriptors on SET_VRING_BASE event. Because of that, is possible to return an invalid descriptor to the guest.
> >>
> >> I guess this could explain my problems that I have seen during reset?
> >>
> >
> > Yes, I think so. The series has a test that should reproduce more or
> > less what you are seeing. However, it would be useful to reproduce on
> > your system and to know what causes qemu to send the reset :).
>
> I do see SET_VRING_BASE in the debug output
> [228101.438630] [2113] vhost:vhost_vring_ioctl:1668: VHOST_GET_VRING_BASE [vq=00000000618905fc][s.index=1][s.num=42424][vq->avail_idx=42424][vq->last_avail_idx=42424][vq->ndescs=0][vq->first_desc=0]
> [228101.438631] CPU: 54 PID: 2113 Comm: qemu-system-s39 Not tainted 5.5.0+ #344
> [228101.438632] Hardware name: IBM 3906 M04 704 (LPAR)
> [228101.438633] Call Trace:
> [228101.438634] [<00000004fc71c132>] show_stack+0x8a/0xd0
> [228101.438636] [<00000004fd10e72a>] dump_stack+0x8a/0xb8
> [228101.438639] [<000003ff80377600>] vhost_vring_ioctl+0x668/0x848 [vhost]
> [228101.438640] [<000003ff80395fd4>] vhost_net_ioctl+0x4f4/0x570 [vhost_net]
> [228101.438642] [<00000004fc9ccdd8>] do_vfs_ioctl+0x430/0x6f8
> [228101.438643] [<00000004fc9cd124>] ksys_ioctl+0x84/0xb0
> [228101.438645] [<00000004fc9cd1ba>] __s390x_sys_ioctl+0x2a/0x38
> [228101.438646] [<00000004fd12ff72>] system_call+0x2a6/0x2c8
> [228103.682732] [2122] vhost:vhost_vring_ioctl:1653: VHOST_SET_VRING_BASE [vq=000000009e1ac3e7][s.index=0][s.num=0][vq->avail_idx=27875][vq->last_avail_idx=27709][vq->ndescs=65][vq->first_desc=22]
> [228103.682735] CPU: 44 PID: 2122 Comm: CPU 0/KVM Not tainted 5.5.0+ #344
> [228103.682739] Hardware name: IBM 3906 M04 704 (LPAR)
> [228103.682741] Call Trace:
> [228103.682748] [<00000004fc71c132>] show_stack+0x8a/0xd0
> [228103.682752] [<00000004fd10e72a>] dump_stack+0x8a/0xb8
> [228103.682761] [<000003ff80377422>] vhost_vring_ioctl+0x48a/0x848 [vhost]
> [228103.682764] [<000003ff80395fd4>] vhost_net_ioctl+0x4f4/0x570 [vhost_net]
> [228103.682767] [<00000004fc9ccdd8>] do_vfs_ioctl+0x430/0x6f8
> [228103.682769] [<00000004fc9cd124>] ksys_ioctl+0x84/0xb0
> [228103.682771] [<00000004fc9cd1ba>] __s390x_sys_ioctl+0x2a/0x38
> [228103.682773] [<00000004fd12ff72>] system_call+0x2a6/0x2c8
> [228103.682794] [2122] vhost:vhost_vring_ioctl:1653: VHOST_SET_VRING_BASE [vq=00000000618905fc][s.index=1][s.num=0][vq->avail_idx=42424][vq->last_avail_idx=42424][vq->ndescs=0][vq->first_desc=0]
> [228103.682795] CPU: 44 PID: 2122 Comm: CPU 0/KVM Not tainted 5.5.0+ #344
> [228103.682797] Hardware name: IBM 3906 M04 704 (LPAR)
> [228103.682797] Call Trace:
> [228103.682799] [<00000004fc71c132>] show_stack+0x8a/0xd0
> [228103.682801] [<00000004fd10e72a>] dump_stack+0x8a/0xb8
> [228103.682804] [<000003ff80377422>] vhost_vring_ioctl+0x48a/0x848 [vhost]
> [228103.682806] [<000003ff80395fd4>] vhost_net_ioctl+0x4f4/0x570 [vhost_net]
> [228103.682808] [<00000004fc9ccdd8>] do_vfs_ioctl+0x430/0x6f8
> [228103.682810] [<00000004fc9cd124>] ksys_ioctl+0x84/0xb0
> [228103.682812] [<00000004fc9cd1ba>] __s390x_sys_ioctl+0x2a/0x38
> [228103.682813] [<00000004fd12ff72>] system_call+0x2a6/0x2c8
>
>
> Isnt that triggered by resetting the virtio devices during system reboot?
>

Yes. I don't know exactly why qemu is sending them, but vhost should
be able to "protect/continue" the same way it used to be before
batching patches.

Did you lose connectivity or experienced rebooting with this patches applied?

Thanks!