Re: [dpdk-dev] [RFC] Add support for device dma mask

["Burakov, Anatoly" <anatoly.burakov@intel.com>]

On 27-Jun-18 5:52 PM, Alejandro Lucero wrote:
> 
> 
> On Wed, Jun 27, 2018 at 2:24 PM, Burakov, Anatoly 
> <anatoly.burakov@intel.com <mailto:anatoly.burakov@intel.com>> wrote:
> 
>     On 27-Jun-18 11:13 AM, Alejandro Lucero wrote:
> 
> 
> 
>         On Wed, Jun 27, 2018 at 9:17 AM, Burakov, Anatoly
>         <anatoly.burakov@intel.com <mailto:anatoly.burakov@intel.com>
>         <mailto:anatoly.burakov@intel.com
>         <mailto:anatoly.burakov@intel.com>>> wrote:
> 
>              On 26-Jun-18 6:37 PM, Alejandro Lucero wrote:
> 
>                  This RFC tries to handle devices with addressing
>         limitations.
>                  NFP devices
>                  4000/6000 can just handle addresses with 40 bits implying
>                  problems for handling
>                  physical address when machines have more than 1TB of
>         memory. But
>                  because how
>                  iovas are configured, which can be equivalent to physical
>                  addresses or based on
>                  virtual addresses, this can be a more likely problem.
> 
>                  I tried to solve this some time ago:
> 
>         https://www.mail-archive.com/dev@dpdk.org/msg45214.html
>         <https://www.mail-archive.com/dev@dpdk.org/msg45214.html>
>                 
>         <https://www.mail-archive.com/dev@dpdk.org/msg45214.html
>         <https://www.mail-archive.com/dev@dpdk.org/msg45214.html>>
> 
>                  It was delayed because there was some changes in
>         progress with
>                  EAL device
>                  handling, and, being honest, I completely forgot about this
>                  until now, when
>                  I have had to work on supporting NFP devices with DPDK and
>                  non-root users.
> 
>                  I was working on a patch for being applied on main DPDK
>         branch
>                  upstream, but
>                  because changes to memory initialization during the
>         last months,
>                  this can not
>                  be backported to stable versions, at least the part
>         where the
>                  hugepages iovas
>                  are checked.
> 
>                  I realize stable versions only allow bug fixing, and this
>                  patchset could
>                  arguably not be considered as so. But without this, it
>         could be,
>                  although
>                  unlikely, a DPDK used in a machine with more than 1TB,
>         and then
>                  NFP using
>                  the wrong DMA host addresses.
> 
>                  Although virtual addresses used as iovas are more
>         dangerous, for
>                  DPDK versions
>                  before 18.05 this is not worse than with physical
>         addresses,
>                  because iovas,
>                  when physical addresses are not available, are based on a
>                  starting address set
>                  to 0x0.
> 
> 
>              You might want to look at the following patch:
> 
>         http://patches.dpdk.org/patch/37149/
>         <http://patches.dpdk.org/patch/37149/>
>              <http://patches.dpdk.org/patch/37149/
>         <http://patches.dpdk.org/patch/37149/>>
> 
>              Since this patch, IOVA as VA mode uses VA addresses, and
>         that has
>              been backported to earlier releases. I don't think there's
>         any case
>              where we used zero-based addresses any more.
> 
> 
>         But memsegs get the iova based on hugepages physaddr, and for VA
>         mode that is based on 0x0 as starting point.
> 
>         And as far as I know, memsegs iovas are what end up being used
>         for IOMMU mappings and what devices will use.
> 
> 
>     For when physaddrs are available, IOVA as PA mode assigns IOVA
>     addresses to PA, while IOVA as VA mode assigns IOVA addresses to VA
>     (both 18.05+ and pre-18.05 as per above patch, which was applied to
>     pre-18.05 stable releases).
> 
>     When physaddrs aren't available, IOVA as VA mode assigns IOVA
>     addresses to VA, both 18.05+ and pre-18.05, as per above patch.
> 
> 
> This is right.
> 
>     If physaddrs aren't available and IOVA as PA mode is used, then i as
>     far as i can remember, even though technically memsegs get their
>     addresses set to 0x0 onwards, the actual addresses we get in
>     memzones etc. are RTE_BAD_IOVA.
> 
> 
> This is not right. Not sure if this was the intention, but if PA mode 
> and physaddrs not available, this code inside vfio_type1_dma_map:
> 
> if(rte_eal_iova_mode() == RTE_IOVA_VA)
> 
> dma_map.iova = dma_map.vaddr;
> 
> else
> 
> dma_map.iova = ms[i].iova;
> 
> 
> does the IOMMU mapping using the iovas and not the vaddr, with the iovas 
> starting at 0x0.

Yep, you're right, apologies. I confused this with no-huge option.

-- 
Thanks,
Anatoly