Re: [PATCH v2 2/3] x86/hvm: Allow writes to registers on the same page as MSI-X table

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 25/03/2023 2:49 am, Marek Marczykowski-Górecki wrote:
> diff --git a/xen/arch/x86/hvm/vmsi.c b/xen/arch/x86/hvm/vmsi.c
> index 9c82bf9b4ec2..9293009a4075 100644
> --- a/xen/arch/x86/hvm/vmsi.c
> +++ b/xen/arch/x86/hvm/vmsi.c
> @@ -438,6 +438,152 @@ static const struct hvm_io_ops msixtbl_mmio_ops = {
>      .write = _msixtbl_write,
>  };
>  
> +static void __iomem *msixtbl_page_handler_get_hwaddr(
> +    const struct domain *d,
> +    uint64_t address,
> +    bool write)
> +{
> +    const struct pci_dev *pdev = NULL;
> +    const struct msixtbl_entry *entry;
> +    int adj_type;
> +
> +    rcu_read_lock(&msixtbl_rcu_lock);
> +    /*
> +     * Check if it's on the same page as the end of the MSI-X table, but
> +     * outside of the table itself.
> +     */
> +    list_for_each_entry( entry, &d->arch.hvm.msixtbl_list, list )
> +    {
> +        if ( PFN_DOWN(address) == PFN_DOWN(entry->gtable) &&
> +             address < entry->gtable )
> +        {
> +            adj_type = ADJ_IDX_FIRST;
> +            pdev = entry->pdev;
> +            break;
> +        }
> +        if ( PFN_DOWN(address) == PFN_DOWN(entry->gtable + entry->table_len) 
> &&
> +             address >= entry->gtable + entry->table_len )
> +        {
> +            adj_type = ADJ_IDX_LAST;
> +            pdev = entry->pdev;
> +            break;
> +        }
> +    }
> +    rcu_read_unlock(&msixtbl_rcu_lock);
> +
> +    if ( !pdev )
> +        return NULL;
> +
> +    ASSERT(pdev->msix);
> +
> +    if ( !pdev->msix->adj_access_table_idx[adj_type] )
> +    {
> +        gdprintk(XENLOG_WARNING,
> +                 "Page for adjacent MSI-X table access not initialized for 
> %pp\n",
> +                 pdev);

One minor observation.   &pdev->sbdf

Otherwise things will go wrong if sbdf ever moves from being the first
element in a pdev.

~Andrew