Re: [PATCH 4/6] xentrace: ARM platform DOMID_XEN mapping support

[Julien Grall <julien.grall@arm.com>]

Hello Benjamin,

Thank you for the patch.

On 16/03/16 20:51, Benjamin Sanda wrote:
> From: bensanda <ben.sanda@dornerworks.com>
>
> Modified xenmem_add_to_physmap_one() to provide support for xentrace on the ARM platform. Checks for DOMID_XEN added via new function, get_pg_owner, ported from x86 code base. This provides correct calls to rcu_lock_domain() when DOMID_XEN is requested. DOMID_XEN checks also adde to skip page to MFN translation (xentrace sends a MFN dirrectly and so does not need to be translated).

A line in the commit message should not be longer than 72 characters.

>
> Signed-off-by: Benjamin Sanda <ben.sanda@dornerworks.com>
> ---
>   xen/arch/arm/mm.c | 56 +++++++++++++++++++++++++++++++++++++++++++++++++++++--
>   1 file changed, 54 insertions(+), 2 deletions(-)
>
> diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
> index 81f9e2e..b1d834f 100644
> --- a/xen/arch/arm/mm.c
> +++ b/xen/arch/arm/mm.c
> @@ -41,6 +41,7 @@
>   #include <xen/pfn.h>
>
>   struct domain *dom_xen, *dom_io, *dom_cow;
> +static struct domain *get_pg_owner(domid_t domid);

I would rather avoid to forward declare a function if there is no strict 
dependency on other functions. Instead, I would add the function before 
the caller.

>
>   /* Static start-of-day pagetables that we use before the allocators
>    * are up. These are used by all CPUs during bringup before switching
> @@ -1099,7 +1100,8 @@ int xenmem_add_to_physmap_one(
>       {
>           struct domain *od;
>           p2m_type_t p2mt;
> -        od = rcu_lock_domain_by_any_id(foreign_domid);
> +        od = get_pg_owner(foreign_domid);
> +
>           if ( od == NULL )
>               return -ESRCH;
>
> @@ -1132,7 +1134,17 @@ int xenmem_add_to_physmap_one(
>               return -EINVAL;
>           }
>
> -        mfn = page_to_mfn(page);
> +        /* If DOMID_XEN then no page to MFN translation is
> +        needed as we already have the MFN directly */
> +        if(DOMID_XEN !=od->domain_id)
> +        {
> +            mfn = page_to_mfn(page);
> +        }
> +        else
> +        {
> +            mfn = idx;
> +        }
> +

Please avoid to spread the DOMID_ID specific case everywhere. The cost 
to calculate the MFN of a page is very limited.

>           t = p2m_map_foreign;
>
>           rcu_unlock_domain(od);
> @@ -1312,6 +1324,46 @@ void clear_and_clean_page(struct page_info *page)
>       unmap_domain_page(p);
>   }
>
> +/* Ported from x86 architecture: checks for special domain requests for
> +DOMID_XEN or DOMID_IO which must be handled differently then guest domain
> +requests */
> +static struct domain *get_pg_owner(domid_t domid)

This function is very similar to the x86 one. I think it would benefit 
to implement get_pg_owner in the common code and add arch specific 
helper when it's necessary.

Also, please introduce the helper put_pg_owner to stay consistent.

> +{
> +    struct domain *pg_owner = NULL, *curr = current->domain;
> +
> +    if ( likely(domid == DOMID_SELF) )
> +    {
> +        pg_owner = rcu_lock_current_domain();
> +        goto out;
> +    }
> +
> +    if ( unlikely(domid == curr->domain_id) )
> +    {
> +        goto out;
> +    }
> +
> +    /* check for special domain cases of DOMID_IO or DOMID_XEN which
> +    must use rcu_lock_domain() and dom_xen/dom_io as the domid_t */
> +    switch ( domid )
> +    {
> +    case DOMID_IO:
> +        pg_owner = rcu_lock_domain(dom_io);
> +        break;
> +    case DOMID_XEN:
> +        pg_owner = rcu_lock_domain(dom_xen);
> +        break;
> +    default:
> +        if ( (pg_owner = rcu_lock_domain_by_id(domid)) == NULL )
> +        {
> +            break;
> +        }
> +        break;
> +    }
> +
> + out:
> +    return pg_owner;
> +}
> +
>   /*
>    * Local variables:
>    * mode: C
>

Regards,

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel