Re: [PATCHv7 06/26] vfio/iommu_type1: implement the VFIO_DMA_MAP_FLAG_NOEXEC flag

[Antonios Motakis <a.motakis@virtualopensystems.com>]

On Fri, Sep 26, 2014 at 10:27 PM, Alex Williamson
<alex.williamson@redhat.com> wrote:
> On Fri, 2014-09-26 at 17:39 +0200, Antonios Motakis wrote:
>> On Wed, Sep 24, 2014 at 12:40 AM, Alex Williamson
>> <alex.williamson@redhat.com> wrote:
>> > On Tue, 2014-09-23 at 16:46 +0200, Antonios Motakis wrote:
>> >> Some IOMMU drivers, such as the ARM SMMU driver, make available the
>> >> IOMMU_NOEXEC flag, to set the page tables for a device as XN (execute never).
>> >> This affects devices such as the ARM PL330 DMA Controller, which respects
>> >> this flag and will refuse to fetch DMA instructions from memory where the
>> >> XN flag has been set.
>> >>
>> >> The flag can be used only if all IOMMU domains behind the container support
>> >> the IOMMU_NOEXEC flag. Also, if any mappings are created with the flag, any
>> >> new domains with devices will have to support it as well.
>> >>
>> >> Signed-off-by: Antonios Motakis <a.motakis@virtualopensystems.com>
>> >> ---
>> >>  drivers/vfio/vfio_iommu_type1.c | 38 +++++++++++++++++++++++++++++++++++++-
>> >>  1 file changed, 37 insertions(+), 1 deletion(-)
>> >>
>> >> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
>> >> index 0734fbe..09e5064 100644
>> >> --- a/drivers/vfio/vfio_iommu_type1.c
>> >> +++ b/drivers/vfio/vfio_iommu_type1.c
>> >> @@ -81,6 +81,26 @@ struct vfio_group {
>> >>  };
>> >>
>> >>  /*
>> >> + * This function returns true only if _all_ domains support the capability.
>> >> + */
>> >> +static int vfio_all_domains_have_iommu_noexec(struct vfio_iommu *iommu)
>> >
>> > Rename to vfio_domains_have_iommu_noexec() for consistency with the
>> > cache version.
>> >
>>
>> The logic here is a slightly different logic between the two. For
>> IOMMU_CACHE we generally check if any domain includes it,
>
> Not true, all the domains must support IOMMU_CACHE for the function to
> return 1.  In fact, the code is so identical that if we were to cache
> IOMMU_CAP_NOEXEC into domain->prot, we should probably only have one
> function:
>
> static int vfio_domains_have_iommu_flag(struct vfio_iommu *iommu, int flag);
>

You are absolutely correct, I managed to confuse myself when switching
from vfio_domains_have_iommu_exec to vfio_domains_have_iommu_noexec.

I will implement the shared function.

>>  for NOEXEC
>> in contract we need all domains to support it, otherwise we can't
>> expose the capability. Hence the _all_ addition in the name of the
>> function.
>>
>> >> +{
>> >> +     struct vfio_domain *d;
>> >> +     int ret = 1;
>> >> +
>> >> +     mutex_lock(&iommu->lock);
>> >> +     list_for_each_entry(d, &iommu->domain_list, next) {
>> >> +             if (!iommu_domain_has_cap(d->domain, IOMMU_CAP_NOEXEC)) {
>> >
>> > Should we cache this in domain->prot like we do for IOMMU_CACHE?
>> >
>> >> +                     ret = 0;
>> >> +                     break;
>> >> +             }
>> >> +     }
>> >> +     mutex_unlock(&iommu->lock);
>> >> +
>> >> +     return ret;
>> >> +}
>> >> +
>> >> +/*
>> >>   * This code handles mapping and unmapping of user data buffers
>> >>   * into DMA'ble space using the IOMMU
>> >>   */
>> >> @@ -546,6 +566,11 @@ static int vfio_dma_do_map(struct vfio_iommu *iommu,
>> >>               prot |= IOMMU_WRITE;
>> >>       if (map->flags & VFIO_DMA_MAP_FLAG_READ)
>> >>               prot |= IOMMU_READ;
>> >> +     if (map->flags & VFIO_DMA_MAP_FLAG_NOEXEC) {
>> >> +             if (!vfio_all_domains_have_iommu_noexec(iommu))
>> >> +                     return -EINVAL;
>> >> +             prot |= IOMMU_NOEXEC;
>> >> +     }
>> >>
>> >>       if (!prot || !size || (size | iova | vaddr) & mask)
>> >>               return -EINVAL;
>> >> @@ -636,6 +661,12 @@ static int vfio_iommu_replay(struct vfio_iommu *iommu,
>> >>               dma = rb_entry(n, struct vfio_dma, node);
>> >>               iova = dma->iova;
>> >>
>> >> +             /* if any of the mappings to be replayed has the NOEXEC flag
>> >> +              * set, then the new iommu domain must support it */
>> >
>> > nit, please fix the comment style to match the rest of the file.
>> >
>>
>> Ack
>>
>> >> +             if ((dma->prot | IOMMU_NOEXEC) &&
>> >> +                 !iommu_domain_has_cap(domain->domain, IOMMU_CAP_NOEXEC))
>> >> +                     return -EINVAL;
>> >> +
>> >>               while (iova < dma->iova + dma->size) {
>> >>                       phys_addr_t phys = iommu_iova_to_phys(d->domain, iova);
>> >>                       size_t size;
>> >> @@ -890,6 +921,10 @@ static long vfio_iommu_type1_ioctl(void *iommu_data,
>> >>                       if (!iommu)
>> >>                               return 0;
>> >>                       return vfio_domains_have_iommu_cache(iommu);
>> >> +             case VFIO_IOMMU_PROT_NOEXEC:
>> >> +                     if (!iommu)
>> >> +                             return 0;
>> >> +                     return vfio_all_domains_have_iommu_noexec(iommu);
>> >>               default:
>> >>                       return 0;
>> >>               }
>> >> @@ -913,7 +948,8 @@ static long vfio_iommu_type1_ioctl(void *iommu_data,
>> >>       } else if (cmd == VFIO_IOMMU_MAP_DMA) {
>> >>               struct vfio_iommu_type1_dma_map map;
>> >>               uint32_t mask = VFIO_DMA_MAP_FLAG_READ |
>> >> -                             VFIO_DMA_MAP_FLAG_WRITE;
>> >> +                             VFIO_DMA_MAP_FLAG_WRITE |
>> >> +                             VFIO_DMA_MAP_FLAG_NOEXEC;
>> >>
>> >>               minsz = offsetofend(struct vfio_iommu_type1_dma_map, size);
>> >>
>> >
>> >
>> >
>>
>>
>>
>
>
>



-- 
Antonios Motakis
Virtual Open Systems