On Wed, Oct 14, 2020 at 04:07:10PM +0100, Burakov, Anatoly wrote:
> External Email
>
> ----------------------------------------------------------------------
> On 12-Oct-20 9:11 AM, Nithin Dabilpuram wrote:
> > Partial unmapping is not supported for VFIO IOMMU type1
> > by kernel. Though kernel gives return as zero, the unmapped size
> > returned will not be same as expected. So check for
> > returned unmap size and return error.
> >
> > For case of DMA map/unmap triggered by heap allocations,
> > maintain granularity of memseg page size so that heap
> > expansion and contraction does not have this issue.
>
> This is quite unfortunate, because there was a different bug that had to do
> with kernel having a very limited number of mappings available [1], as a
> result of which the page concatenation code was added.
>
> It should therefore be documented that the dma_entry_limit parameter should
> be adjusted should the user run out of the DMA entries.
>
> [1]
> https://urldefense.proofpoint.com/v2/url?u=https-3A__lore.kernel.org_lkml_155414977872.12780.13728555131525362206.stgit-40gimli.home_T_&d=DwICaQ&c=nKjWec2b6R0mOyPaz7xtfQ&r=FZ_tPCbgFOh18zwRPO9H0yDx8VW38vuapifdDfc8SFQ&m=3GMg-634_cdUCY4WpQPwjzZ_S4ckuMHOnt2FxyyjXMk&s=TJLzppkaDS95VGyRHX2hzflQfb9XLK0OiOszSXoeXKk&e=
Ack, I'll document it in guides/linux_gsg/linux_drivers.rst in vfio section.
>
> >
> > For user requested DMA map/unmap disallow partial unmapping
> > for VFIO type1.
> >
> > Fixes: 73a639085938 ("vfio: allow to map other memory regions")
> > Cc: anatoly.bura...@intel.com
> > Cc: sta...@dpdk.org
> >
> > Signed-off-by: Nithin Dabilpuram <ndabilpu...@marvell.com>
> > ---
> > lib/librte_eal/linux/eal_vfio.c | 34 ++++++++++++++++++++++++++++------
> > lib/librte_eal/linux/eal_vfio.h | 1 +
> > 2 files changed, 29 insertions(+), 6 deletions(-)
> >
> > diff --git a/lib/librte_eal/linux/eal_vfio.c
> > b/lib/librte_eal/linux/eal_vfio.c
> > index d26e164..ef95259 100644
> > --- a/lib/librte_eal/linux/eal_vfio.c
> > +++ b/lib/librte_eal/linux/eal_vfio.c
> > @@ -69,6 +69,7 @@ static const struct vfio_iommu_type iommu_types[] = {
> > {
> > .type_id = RTE_VFIO_TYPE1,
> > .name = "Type 1",
> > + .partial_unmap = false,
> > .dma_map_func = &vfio_type1_dma_map,
> > .dma_user_map_func = &vfio_type1_dma_mem_map
> > },
> > @@ -76,6 +77,7 @@ static const struct vfio_iommu_type iommu_types[] = {
> > {
> > .type_id = RTE_VFIO_SPAPR,
> > .name = "sPAPR",
> > + .partial_unmap = true,
> > .dma_map_func = &vfio_spapr_dma_map,
> > .dma_user_map_func = &vfio_spapr_dma_mem_map
> > },
> > @@ -83,6 +85,7 @@ static const struct vfio_iommu_type iommu_types[] = {
> > {
> > .type_id = RTE_VFIO_NOIOMMU,
> > .name = "No-IOMMU",
> > + .partial_unmap = true,
> > .dma_map_func = &vfio_noiommu_dma_map,
> > .dma_user_map_func = &vfio_noiommu_dma_mem_map
> > },
> > @@ -525,12 +528,19 @@ vfio_mem_event_callback(enum rte_mem_event type,
> > const void *addr, size_t len,
> > /* for IOVA as VA mode, no need to care for IOVA addresses */
> > if (rte_eal_iova_mode() == RTE_IOVA_VA && msl->external == 0) {
> > uint64_t vfio_va = (uint64_t)(uintptr_t)addr;
> > - if (type == RTE_MEM_EVENT_ALLOC)
> > - vfio_dma_mem_map(default_vfio_cfg, vfio_va, vfio_va,
> > - len, 1);
> > - else
> > - vfio_dma_mem_map(default_vfio_cfg, vfio_va, vfio_va,
> > - len, 0);
> > + uint64_t page_sz = msl->page_sz;
> > +
> > + /* Maintain granularity of DMA map/unmap to memseg size */
> > + for (; cur_len < len; cur_len += page_sz) {
> > + if (type == RTE_MEM_EVENT_ALLOC)
> > + vfio_dma_mem_map(default_vfio_cfg, vfio_va,
> > + vfio_va, page_sz, 1);
> > + else
> > + vfio_dma_mem_map(default_vfio_cfg, vfio_va,
> > + vfio_va, page_sz, 0);
> > + vfio_va += page_sz;
> > + }
> > +
>
> You'd also have to revert d1c7c0cdf7bac5eb40d3a2a690453aefeee5887b because
> currently the PA path will opportunistically concantenate contiguous
> segments into single mapping too.
Ack, I'll change it even for IOVA as PA mode. I missed that.
>
> > return;
> > }
> > @@ -1383,6 +1393,12 @@ vfio_type1_dma_mem_map(int vfio_container_fd,
> > uint64_t vaddr, uint64_t iova,
> > RTE_LOG(ERR, EAL, " cannot clear DMA remapping, error
> > %i (%s)\n",
> > errno, strerror(errno));
> > return -1;
> > + } else if (dma_unmap.size != len) {
> > + RTE_LOG(ERR, EAL, " unexpected size %"PRIu64" of DMA "
> > + "remapping cleared instead of %"PRIu64"\n",
> > + (uint64_t)dma_unmap.size, len);
> > + rte_errno = EIO;
> > + return -1;
> > }
> > }
> > @@ -1853,6 +1869,12 @@ container_dma_unmap(struct vfio_config *vfio_cfg,
> > uint64_t vaddr, uint64_t iova,
> > /* we're partially unmapping a previously mapped region, so we
> > * need to split entry into two.
> > */
> > + if (!vfio_cfg->vfio_iommu_type->partial_unmap) {
> > + RTE_LOG(DEBUG, EAL, "DMA partial unmap unsupported\n");
> > + rte_errno = ENOTSUP;
> > + ret = -1;
> > + goto out;
> > + }
>
> How would we ever arrive here if we never do more than 1 page worth of
> memory anyway? I don't think this is needed.
container_dma_unmap() is called by user via rte_vfio_container_dma_unmap()
and when he maps we don't split it as we don't about his memory.
So if he maps multiple pages and tries to unmap partially, then we should fail.
>
> > if (user_mem_maps->n_maps == VFIO_MAX_USER_MEM_MAPS) {
> > RTE_LOG(ERR, EAL, "Not enough space to store partial
> > mapping\n");
> > rte_errno = ENOMEM;
> > diff --git a/lib/librte_eal/linux/eal_vfio.h
> > b/lib/librte_eal/linux/eal_vfio.h
> > index cb2d35f..6ebaca6 100644
> > --- a/lib/librte_eal/linux/eal_vfio.h
> > +++ b/lib/librte_eal/linux/eal_vfio.h
> > @@ -113,6 +113,7 @@ typedef int (*vfio_dma_user_func_t)(int fd, uint64_t
> > vaddr, uint64_t iova,
> > struct vfio_iommu_type {
> > int type_id;
> > const char *name;
> > + bool partial_unmap;
> > vfio_dma_user_func_t dma_user_map_func;
> > vfio_dma_func_t dma_map_func;
> > };
> >
>
>
> --
> Thanks,
> Anatoly
