Re: [PATCH] xen/xenbus: fix xenbus_setup_ring()

[Oleksandr Tyshchenko <Oleksandr_Tyshchenko@xxxxxxxx>]

On 15.09.22 17:31, Juergen Gross wrote:

Hello Juergen

> Commit 4573240f0764 ("xen/xenbus: eliminate xenbus_grant_ring()")
> introduced an error for initialization of multi-page rings.
>
> Cc: stable@xxxxxxxxxxxxxxx
> Fixes: 4573240f0764 ("xen/xenbus: eliminate xenbus_grant_ring()")
> Reported-by: Sander Eikelenboom <linux@xxxxxxxxxxxxxx>
> Signed-off-by: Juergen Gross <jgross@xxxxxxxx>
> ---
>   drivers/xen/xenbus/xenbus_client.c | 9 ++++++---
>   1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/xen/xenbus/xenbus_client.c 
> b/drivers/xen/xenbus/xenbus_client.c
> index d5f3f763717e..caa5c5c32f8e 100644
> --- a/drivers/xen/xenbus/xenbus_client.c
> +++ b/drivers/xen/xenbus/xenbus_client.c
> @@ -382,9 +382,10 @@ int xenbus_setup_ring(struct xenbus_device *dev, gfp_t 
> gfp, void **vaddr,
>       unsigned long ring_size = nr_pages * XEN_PAGE_SIZE;
>       grant_ref_t gref_head;
>       unsigned int i;
> +     void *addr;
>       int ret;
>   
> -     *vaddr = alloc_pages_exact(ring_size, gfp | __GFP_ZERO);
> +     addr = *vaddr = alloc_pages_exact(ring_size, gfp | __GFP_ZERO);
>       if (!*vaddr) {
>               ret = -ENOMEM;
>               goto err;
> @@ -401,13 +402,15 @@ int xenbus_setup_ring(struct xenbus_device *dev, gfp_t 
> gfp, void **vaddr,
>               unsigned long gfn;
>   
>               if (is_vmalloc_addr(*vaddr))
> -                     gfn = pfn_to_gfn(vmalloc_to_pfn(vaddr[i]));
> +                     gfn = pfn_to_gfn(vmalloc_to_pfn(addr));
>               else
> -                     gfn = virt_to_gfn(vaddr[i]);
> +                     gfn = virt_to_gfn(addr);
>   
>               grefs[i] = gnttab_claim_grant_reference(&gref_head);
>               gnttab_grant_foreign_access_ref(grefs[i], dev->otherend_id,
>                                               gfn, 0);
> +
> +             addr += PAGE_SIZE;

XEN_PAGE_SIZE?


Reviewed-by: Oleksandr Tyshchenko <oleksandr_tyshchenko@xxxxxxxx>


P.S.

I wondered why I didn't face the similar issue(s) as I used Linux 
v6.0.0-rc1 and several PV drivers including PV block device.

The answer is that the single-page ring is being used for all of them in 
my environment.


root@salvator-x-h3-4x2g-xt-domu:~# dmesg | grep xenbus_setup_ring
[    0.332449] vdispl vdispl-0: 0 xenbus_setup_ring: 1 pages
[    0.333464] vdispl vdispl-0: 0 xenbus_setup_ring: 1 pages
[    0.341350] vbd vbd-51713: 0 xenbus_setup_ring: 1 pages
[    0.342750] vbd vbd-51713: 0 xenbus_setup_ring: 1 pages
[    0.343261] vbd vbd-51713: 0 xenbus_setup_ring: 1 pages
[    0.343798] vbd vbd-51713: 0 xenbus_setup_ring: 1 pages
[    0.392969] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.518733] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.519198] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.519501] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.519948] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.520371] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.520805] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.521070] vif vif-0: 0 xenbus_setup_ring: 1 pages
[    0.575451] vsnd vsnd-0: 0 xenbus_setup_ring: 1 pages
[    0.576855] vsnd vsnd-0: 0 xenbus_setup_ring: 1 pages


>       }
>   
>       return 0;

-- 
Regards,

Oleksandr Tyshchenko