Re: [Qemu-devel] [PATCH V2 WIP 0/2] vhost-scsi: new device supporting the tcm_vhost Linux kernel module

[Michael S. Tsirkin]

On Thu, Mar 14, 2013 at 05:25:42PM +0800, Asias He wrote:
> On Thu, Mar 14, 2013 at 12:25:14PM +0800, Asias He wrote:
> > On Tue, Mar 12, 2013 at 02:29:40PM +0800, Asias He wrote:
> > > This is on top of Paolo and Nick's work.
> > > 
> > > Current status:
> > > Works now (guest boots fine, no hang any more) with seabios's virtio-scsi 
> > > disabled.
> > > Rebased to latest qemu.org/master
> > > Change details are in commit log.
> > > 
> > > TODO:
> > > Make seabios happy.
> > 
> > Some updates:
> > 
> > In seabios, src/virtio-scsi.c:virtio_scsi_cmd()
> > 
> >     vring_kick(ioaddr, vq, 1);
> >     /* Wait for reply */
> >     while (!vring_more_used(vq))  ------------>>> we hang here
> >         usleep(5);
> > 
> > In tcm_vhost:
> > we got: vhost_get_vq_desc: head: -14, out: 0 in: 81216
> > 
> > The vring buffer provided by seabios is not correct?
> 
> More updates:
> 
> Basically, seabios + tcm_vhost works now, with the attached qemu and
> seabios patches.
> 
> We still have one more issue, vhost_verify_ring_mappings fails.
> 
> I am seeing this without the comment out hack.
> 
>    Unable to map ring buffer for ring 2
>    qemu-system-x86_64: /root/src/qemu/hw/vhost.c:426: vhost_set_memory:
>    Assertion `r >= 0' failed.
> 

Hmm it looks like a bug in cpu_physical_memory_map.
I think it can make length larger than what was passed in.
Could you try the following and see what happens?

diff --git a/hw/vhost.c b/hw/vhost.c
index 37777c2..a30e551 100644
--- a/hw/vhost.c
+++ b/hw/vhost.c
@@ -320,7 +320,9 @@ static int vhost_verify_ring_mappings(struct vhost_dev *dev,
         l = vq->ring_size;
         p = cpu_physical_memory_map(vq->ring_phys, &l, 1);
         if (!p || l != vq->ring_size) {
-            fprintf(stderr, "Unable to map ring buffer for ring %d\n", i);
+            fprintf(stderr, "Unable to map ring buffer for ring %d "
+                    "Debug data: %p %llx %llx\n", i, p,
+                    (unsigned long long)l, (unsigned long long)vq->ring_size);
             return -ENOMEM;
         }
         if (p != vq->ring) {


> > > Paolo Bonzini (2):
> > >   virtio-scsi: create VirtIOSCSICommon
> > >   vhost-scsi: new device supporting the tcm_vhost Linux kernel module
> > > 
> > >  hw/Makefile.objs           |   5 +-
> > >  hw/s390x/s390-virtio-bus.c |  35 +++++++
> > >  hw/vhost-scsi.c            | 242 
> > > +++++++++++++++++++++++++++++++++++++++++++++
> > >  hw/vhost-scsi.h            |  64 ++++++++++++
> > >  hw/virtio-pci.c            |  59 +++++++++++
> > >  hw/virtio-scsi.c           | 199 +++++++++----------------------------
> > >  hw/virtio-scsi.h           | 129 ++++++++++++++++++++++++
> > >  include/qemu/osdep.h       |   4 +
> > >  8 files changed, 586 insertions(+), 151 deletions(-)
> > >  create mode 100644 hw/vhost-scsi.c
> > >  create mode 100644 hw/vhost-scsi.h
> > > 
> > > -- 
> > > 1.8.1.4
> > > 
> > 
> > -- 
> > Asias
> 
> --------------- 8 ---------------> qemu patch:
> diff --git a/hw/vhost.c b/hw/vhost.c
> index 4d6aee3..0c52ec4 100644
> --- a/hw/vhost.c
> +++ b/hw/vhost.c
> @@ -421,10 +421,12 @@ static void vhost_set_memory(MemoryListener *listener,
>          return;
>      }
>  
> +#if 0
>      if (dev->started) {
>          r = vhost_verify_ring_mappings(dev, start_addr, size);
>          assert(r >= 0);
>      }
> +#endif
>  
>      if (!dev->log_enabled) {
>          r = ioctl(dev->control, VHOST_SET_MEM_TABLE, dev->mem);
> 
> --------------- 8 ---------------> seabios patch:
> diff --git a/src/virtio-scsi.c b/src/virtio-scsi.c
> index 879ddfb..4de1255 100644
> --- a/src/virtio-scsi.c
> +++ b/src/virtio-scsi.c
> @@ -147,6 +147,9 @@ init_virtio_scsi(struct pci_device *pci)
>          goto fail;
>      }
>  
> +    vp_set_status(ioaddr, VIRTIO_CONFIG_S_ACKNOWLEDGE |
> +                  VIRTIO_CONFIG_S_DRIVER | VIRTIO_CONFIG_S_DRIVER_OK);
> +
>      int i, tot;
>      for (tot = 0, i = 0; i < 256; i++)
>          tot += virtio_scsi_scan_target(pci, ioaddr, vq, i);
> @@ -154,8 +157,6 @@ init_virtio_scsi(struct pci_device *pci)
>      if (!tot)
>          goto fail;
>  
> -    vp_set_status(ioaddr, VIRTIO_CONFIG_S_ACKNOWLEDGE |
> -                  VIRTIO_CONFIG_S_DRIVER | VIRTIO_CONFIG_S_DRIVER_OK);
>      return;
>  
>  fail:
> diff --git a/src/virtio-scsi.h b/src/virtio-scsi.h
> index bbfbf30..96c3701 100644
> --- a/src/virtio-scsi.h
> +++ b/src/virtio-scsi.h
> @@ -26,7 +26,7 @@ struct virtio_scsi_req_cmd {
>      u8 prio;
>      u8 crn;
>      char cdb[VIRTIO_SCSI_CDB_SIZE];
> -};
> +} __attribute__((packed));
>  
>  /* This is the first element of the "in" scatter-gather list. */
>  struct virtio_scsi_resp_cmd {
> @@ -36,7 +36,7 @@ struct virtio_scsi_resp_cmd {
>      u8 status;
>      u8 response;
>      u8 sense[VIRTIO_SCSI_SENSE_SIZE];
> -};
> +} __attribute__((packed));
>  
>  #define VIRTIO_SCSI_S_OK            0

I see, the padding creates the problem?

> -- 
> Asias