Re: [PATCH 06/18] arm/altp2m: Add a(p2m) table flushing routines.

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

Hello Sergej,

On 04/07/16 13:12, Sergej Proskurin wrote:
>> +/* Reset this p2m table to be empty */
>> +static void p2m_flush_table(struct p2m_domain *p2m)
>> +{
>> +    struct page_info *top, *pg;
>> +    mfn_t mfn;
>> +    unsigned int i;
>> +
>> +    /* Check whether the p2m table has already been flushed before. */
>> +    if ( p2m->root == NULL)
>> +        return;
>> +
>> +    spin_lock(&p2m->lock);
>> +
>> +    /*
>> +     * "Host" p2m tables can have shared entries &c that need a bit more care
>> +     * when discarding them
>> +     */
>> +    ASSERT(!p2m_is_hostp2m(p2m));
>> +
>> +    /* Zap the top level of the trie */
>> +    top = p2m->root;
>> +
>> +    /* Clear all concatenated first level pages */
>> +    for ( i = 0; i < P2M_ROOT_PAGES; i++ )
>> +    {
>> +        mfn = _mfn(page_to_mfn(top + i));
>> +        clear_domain_page(mfn);
>> +    }
>> +
>> +    /* Free the rest of the trie pages back to the paging pool */
>> +    while ( (pg = page_list_remove_head(&p2m->pages)) )
>> +        if ( pg != top  )
>> +        {
>> +            /*
>> +             * Before freeing the individual pages, we clear them to prevent
>> +             * reusing old table entries in future p2m allocations.
>> +             */
>> +            mfn = _mfn(page_to_mfn(pg));
>> +            clear_domain_page(mfn);
>> +            free_domheap_page(pg);
>> +        }
>
> At this point, we prevent only the first root level page from being
> freed. In case there are multiple consecutive first level pages, one of
> them will be freed in the upper loop (and potentially crash the guest if
> the table is reused at a later point in time). However, testing for
> every concatenated page in the if clause of the while loop would further
> decrease the flushing performance. Thus, my question is, whether there
> is a good way to solve this issue?

The root pages are not part of p2m->pages, so there is no issue.

Regards,

-- 
Julien Grall

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