[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH v13 Kernel 4/7] vfio iommu: Implementation of ioctl to for di
|
From: |
Alex Williamson |
|
Subject: |
Re: [PATCH v13 Kernel 4/7] vfio iommu: Implementation of ioctl to for dirty pages tracking. |
|
Date: |
Mon, 16 Mar 2020 13:25:58 -0600 |
On Tue, 17 Mar 2020 00:19:05 +0530
Kirti Wankhede <address@hidden> wrote:
> On 3/13/2020 11:43 PM, Alex Williamson wrote:
> > On Thu, 12 Mar 2020 23:23:24 +0530
> > Kirti Wankhede <address@hidden> wrote:
> >
> > Subject: s/to //
> >
> >> VFIO_IOMMU_DIRTY_PAGES ioctl performs three operations:
> >> - Start pinned and unpinned pages tracking while migration is active
> >> - Stop pinned and unpinned dirty pages tracking. This is also used to
> >> stop dirty pages tracking if migration failed or cancelled.
> >
> > I'm not sure why we need to specify "pinned and unpinned" above, it
> > works for iommu mapped pages too, and we expect it to later work other
> > dirtying mechanisms, like Yan's dma_rw interface.
> >
>
> iommu mapped pages are also pinned.
> When there are CPU writes, like Yan's dma_rw, then KVM core takes care
> of marking those pages dirty, then we don't need to track those pages
> again here, right?
No, KVM will mark vCPU writes as dirty, Yan's DMA r/w is specifically
providing an interface for the *host* CPU to dirty pages as if it was
device DMA. We need to fill that gap.
Yes, IOMMU mapped pages are pinned, but as soon as we start talking
about pinned and unpinned, I drift towards explicit page pinning
through mdev.
> >> - Get dirty pages bitmap. This ioctl returns bitmap of dirty pages, its
> >> user space application responsibility to copy content of dirty pages
> >> from source to destination during migration.
> >>
> >> To prevent DoS attack, memory for bitmap is allocated per vfio_dma
> >> structure. Bitmap size is calculated considering smallest supported page
> >> size. Bitmap is allocated when dirty logging is enabled for those
> >> vfio_dmas whose vpfn list is not empty or whole range is mapped, in
> >> case of pass-through device.
> >>
> >> Bitmap is populated for already pinned pages when bitmap is allocated for
> >> a vfio_dma with the smallest supported page size. Update bitmap from
> >> pinning and unpinning functions. When user application queries bitmap,
> >> check if requested page size is same as page size used to populated
> >> bitmap. If it is equal, copy bitmap, but if not equal, return error.
> >>
> >> Signed-off-by: Kirti Wankhede <address@hidden>
> >> Reviewed-by: Neo Jia <address@hidden>
> >> ---
> >> drivers/vfio/vfio_iommu_type1.c | 243
> >> +++++++++++++++++++++++++++++++++++++++-
> >> 1 file changed, 237 insertions(+), 6 deletions(-)
> >>
> >> diff --git a/drivers/vfio/vfio_iommu_type1.c
> >> b/drivers/vfio/vfio_iommu_type1.c
> >> index d386461e5d11..435e84269a28 100644
> >> --- a/drivers/vfio/vfio_iommu_type1.c
> >> +++ b/drivers/vfio/vfio_iommu_type1.c
> >> @@ -70,6 +70,7 @@ struct vfio_iommu {
> >> unsigned int dma_avail;
> >> bool v2;
> >> bool nesting;
> >> + bool dirty_page_tracking;
> >> };
> >>
> >> struct vfio_domain {
> >> @@ -90,6 +91,7 @@ struct vfio_dma {
> >> bool lock_cap; /* capable(CAP_IPC_LOCK) */
> >> struct task_struct *task;
> >> struct rb_root pfn_list; /* Ex-user pinned pfn list */
> >> + unsigned long *bitmap;
> >> };
> >>
> >> struct vfio_group {
> >> @@ -125,6 +127,7 @@ struct vfio_regions {
> >> (!list_empty(&iommu->domain_list))
> >>
> >> static int put_pfn(unsigned long pfn, int prot);
> >> +static unsigned long vfio_pgsize_bitmap(struct vfio_iommu *iommu);
> >>
> >> /*
> >> * This code handles mapping and unmapping of user data buffers
> >> @@ -174,6 +177,76 @@ static void vfio_unlink_dma(struct vfio_iommu *iommu,
> >> struct vfio_dma *old)
> >> rb_erase(&old->node, &iommu->dma_list);
> >> }
> >>
> >> +static inline unsigned long dirty_bitmap_bytes(unsigned int npages)
> >> +{
> >> + if (!npages)
> >> + return 0;
> >> +
> >> + return ALIGN(npages, BITS_PER_LONG) / sizeof(unsigned long);
> >
> > Yes, but no. The math works out here on LP64 systems because
> > sizeof(unsigned long) == BITS_PER_BYTE, but sizeof(unsigned long) is
> > irrelevant, we wants BITS_PER_BYTE. Also, the UAPI defines the bitmap
> > as an array of __u64, so rather than BITS_PER_LONG I think we want
> > BITS_PER_TYPE(u64).
> >
>
> Changing it to macro:
>
> #define DIRTY_BITMAP_BYTES(n) (ALIGN(n, BITS_PER_TYPE(u64)) /
> BITS_PER_BYTE)
>
>
> >> +}
> >> +
> >> +static int vfio_dma_bitmap_alloc(struct vfio_dma *dma, unsigned long
> >> pgsize)
> >> +{
> >> + if (!RB_EMPTY_ROOT(&dma->pfn_list) || dma->iommu_mapped) {
> >
> > Yan's patch series allows vendor drivers to dirty pages regardless of
> > either of these. I think we should do away with this an
> > unconditionally allocate a bitmap per vfio_dma.
> >
>
> Same as above - CPU writes are tracked by KVM core.
As above, nope.
> >> + unsigned long npages = dma->size / pgsize;
> >> +
> >> + dma->bitmap = kvzalloc(dirty_bitmap_bytes(npages), GFP_KERNEL);
> >> + if (!dma->bitmap)
> >> + return -ENOMEM;
> >> + }
> >> + return 0;
> >> +}
> >> +
> >> +static int vfio_dma_all_bitmap_alloc(struct vfio_iommu *iommu,
> >> + unsigned long pgsize)
> >> +{
> >> + struct rb_node *n = rb_first(&iommu->dma_list);
> >> + int ret;
> >> +
> >> + for (; n; n = rb_next(n)) {
> >> + struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
> >> + struct rb_node *n;
> >> +
> >> + ret = vfio_dma_bitmap_alloc(dma, pgsize);
> >> + if (ret) {
> >> + struct rb_node *p = rb_prev(n);
> >> +
> >> + for (; p; p = rb_prev(p)) {
> >> + struct vfio_dma *dma = rb_entry(n,
> >> + struct vfio_dma, node);
> >> +
> >> + kfree(dma->bitmap);
> >> + dma->bitmap = NULL;
> >> + }
> >> + return ret;
> >> + }
> >> +
> >> + if (!dma->bitmap)
> >> + continue;
> >> +
> >> + for (n = rb_first(&dma->pfn_list); n; n = rb_next(n)) {
> >> + struct vfio_pfn *vpfn = rb_entry(n, struct vfio_pfn,
> >> + node);
> >> +
> >> + bitmap_set(dma->bitmap,
> >> + (vpfn->iova - dma->iova) / pgsize, 1);
> >> + }
> >> + }
> >> + return 0;
> >> +}
> >> +
> >> +static void vfio_dma_all_bitmap_free(struct vfio_iommu *iommu)
> >> +{
> >> + struct rb_node *n = rb_first(&iommu->dma_list);
> >> +
> >> + for (; n; n = rb_next(n)) {
> >> + struct vfio_dma *dma = rb_entry(n, struct vfio_dma, node);
> >> +
> >> + kfree(dma->bitmap);
> >> + dma->bitmap = NULL;
> >> + }
> >> +}
> >> +
> >> /*
> >> * Helper Functions for host iova-pfn list
> >> */
> >> @@ -254,12 +327,16 @@ static struct vfio_pfn
> >> *vfio_iova_get_vfio_pfn(struct vfio_dma *dma,
> >> return vpfn;
> >> }
> >>
> >> -static int vfio_iova_put_vfio_pfn(struct vfio_dma *dma, struct vfio_pfn
> >> *vpfn)
> >> +static int vfio_iova_put_vfio_pfn(struct vfio_dma *dma, struct vfio_pfn
> >> *vpfn,
> >> + bool do_tracking, unsigned long pgsize)
> >> {
> >> int ret = 0;
> >>
> >> vpfn->ref_count--;
> >> if (!vpfn->ref_count) {
> >> + if (do_tracking && dma->bitmap)
> >> + bitmap_set(dma->bitmap,
> >> + (vpfn->iova - dma->iova) / pgsize, 1);
> >
> > This seems wrong, or at best redundant. The dirty bitmap should always
> > reflect all outstanding pinned pages. When ref_count drops to zero we
> > can stop tracking it, but it should already be set in the bitmap and we
> > shouldn't need to do anything here.
> >
>
> Yes, this is redundant. Removing it.
>
> >> ret = put_pfn(vpfn->pfn, dma->prot);
> >> vfio_remove_from_pfn_list(dma, vpfn);
> >> }
> >> @@ -484,7 +561,8 @@ static int vfio_pin_page_external(struct vfio_dma
> >> *dma, unsigned long vaddr,
> >> }
> >>
> >> static int vfio_unpin_page_external(struct vfio_dma *dma, dma_addr_t
> >> iova,
> >> - bool do_accounting)
> >> + bool do_accounting, bool do_tracking,
> >> + unsigned long pgsize)
> >> {
> >> int unlocked;
> >> struct vfio_pfn *vpfn = vfio_find_vpfn(dma, iova);
> >> @@ -492,7 +570,7 @@ static int vfio_unpin_page_external(struct vfio_dma
> >> *dma, dma_addr_t iova,
> >> if (!vpfn)
> >> return 0;
> >>
> >> - unlocked = vfio_iova_put_vfio_pfn(dma, vpfn);
> >> + unlocked = vfio_iova_put_vfio_pfn(dma, vpfn, do_tracking, pgsize);
> >
> > This is the only use of do_tracking, which as discussed above shouldn't
> > need to be done in this function, thus we shouldn't need to pass pgsize
> > either.
> >
>
> Right changing all calls to vfio_iova_put_vfio_pfn()
>
> >>
> >> if (do_accounting)
> >> vfio_lock_acct(dma, -unlocked, true);
> >> @@ -563,9 +641,26 @@ static int vfio_iommu_type1_pin_pages(void
> >> *iommu_data,
> >>
> >> ret = vfio_add_to_pfn_list(dma, iova, phys_pfn[i]);
> >> if (ret) {
> >> - vfio_unpin_page_external(dma, iova, do_accounting);
> >> + vfio_unpin_page_external(dma, iova, do_accounting,
> >> + false, 0);
> >
> > I might use PAGE_SIZE here rather than 0 just for sanity, but as above
> > we don't need these extra args.
> >
> >> goto pin_unwind;
> >> }
> >> +
> >> + if (iommu->dirty_page_tracking) {
> >
> > If we unconditionally allocated the bitmap, we could just test
> > dma->bitmap here and get rid of the allocation branch below (we'll need
> > to allocate the bitmap with the vfio_dma if new mappings are created
> > while dirty tracking is enabled though).
> >
> >> + unsigned long pgshift =
> >> + __ffs(vfio_pgsize_bitmap(iommu));
> >
> > Maybe a follow-up patch could cache this, it can only potentially
> > change as each new group is added. We'd probably want to deny adding a
> > group if the minimum pagesize changed while dirty tracking is enabled.
> >
> >> +
> >> + if (!dma->bitmap) {
> >> + ret = vfio_dma_bitmap_alloc(dma, 1 << pgshift);
> >> + if (ret) {
> >> + vfio_unpin_page_external(dma, iova,
> >> + do_accounting, false, 0);
> >> + goto pin_unwind;
> >> + }
> >> + }
> >> + bitmap_set(dma->bitmap,
> >> + (vpfn->iova - dma->iova) >> pgshift, 1);
> >> + }
> >> }
> >>
> >> ret = i;
> >> @@ -578,7 +673,7 @@ static int vfio_iommu_type1_pin_pages(void *iommu_data,
> >>
> >> iova = user_pfn[j] << PAGE_SHIFT;
> >> dma = vfio_find_dma(iommu, iova, PAGE_SIZE);
> >> - vfio_unpin_page_external(dma, iova, do_accounting);
> >> + vfio_unpin_page_external(dma, iova, do_accounting, false, 0);
> >
> > Unneeded.
> >
> >> phys_pfn[j] = 0;
> >> }
> >> pin_done:
> >> @@ -612,7 +707,8 @@ static int vfio_iommu_type1_unpin_pages(void
> >> *iommu_data,
> >> dma = vfio_find_dma(iommu, iova, PAGE_SIZE);
> >> if (!dma)
> >> goto unpin_exit;
> >> - vfio_unpin_page_external(dma, iova, do_accounting);
> >> + vfio_unpin_page_external(dma, iova, do_accounting,
> >> + iommu->dirty_page_tracking,
> >> PAGE_SIZE);
> >
> > Same.
> >
> >> }
> >>
> >> unpin_exit:
> >> @@ -800,6 +896,7 @@ static void vfio_remove_dma(struct vfio_iommu *iommu,
> >> struct vfio_dma *dma)
> >> vfio_unmap_unpin(iommu, dma, true);
> >> vfio_unlink_dma(iommu, dma);
> >> put_task_struct(dma->task);
> >> + kfree(dma->bitmap);
> >> kfree(dma);
> >> iommu->dma_avail++;
> >> }
> >> @@ -830,6 +927,54 @@ static unsigned long vfio_pgsize_bitmap(struct
> >> vfio_iommu *iommu)
> >> return bitmap;
> >> }
> >>
> >> +static int vfio_iova_dirty_bitmap(struct vfio_iommu *iommu, dma_addr_t
> >> iova,
> >> + size_t size, uint64_t pgsize,
> >> + unsigned char __user *bitmap)
> >> +{
> >> + struct vfio_dma *dma;
> >> + unsigned long pgshift = __ffs(pgsize);
> >> + unsigned int npages, bitmap_size;
> >> +
> >> + dma = vfio_find_dma(iommu, iova, 1);
> >> +
> >> + if (!dma)
> >> + return -EINVAL;
> >> +
> >> + if (dma->iova != iova || dma->size != size)
> >> + return -EINVAL;
> >> +
> >> + npages = dma->size >> pgshift;
> >> + bitmap_size = dirty_bitmap_bytes(npages);
> >> +
> >> + /* mark all pages dirty if all pages are pinned and mapped. */
> >> + if (dma->iommu_mapped)
> >> + bitmap_set(dma->bitmap, 0, npages);
> >
> > Oops, we only test if dma->bitmap exists below. Always allocating the
> > bitmap would resolve this too.
> >
> >> +
> >> + if (dma->bitmap) {
> >> + if (copy_to_user((void __user *)bitmap, dma->bitmap,
> >> + bitmap_size))
> >> + return -EFAULT;
> >> +
> >> + memset(dma->bitmap, 0, bitmap_size);
> >
> > Nope, we need to reprocess the bitmap to set all pinned pages as dirty
> > again. Those need to be reported _every_ time the user asks for the
> > bitmap overlapping them. We consider them to be continuously dirtied.
> >
> >> + }
> >> + return 0;
> >> +}
> >> +
> >> +static int verify_bitmap_size(unsigned long npages, unsigned long
> >> bitmap_size)
> >> +{
> >> + long bsize;
> >> +
> >> + if (!bitmap_size || bitmap_size > SIZE_MAX)
> >> + return -EINVAL;
> >
> > bitmap_size should be a size_t if we're going to compare it to
> > SIZE_MAX. bsize and the return of dirty_bitmap_bytes() should probably
> > also be a size_t.
> >
>
> changing all parameters to uint64_t to be consistent with uapi and
> compare with UINT_MAX.
UINT_MAX seems arbitrary, is this specified in our API? The size of a
vfio_dma is limited to what the user is able to pin, and therefore
their locked memory limit, but do we have an explicit limit elsewhere
that results in this limit here. I think a 4GB bitmap would track
something like 2^47 bytes of memory, that's pretty excessive, but still
an arbitrary limit.
BTW, we probably do need to be careful about allocating dirty bitmaps
for non-page backed mappings, those aren't limited by the user's locked
memory. Being sure that we can allocate the bitmap is another reason
to unconditionally create the bitmap when logging is enabled. Thanks,
Alex
> >> +
> >> + bsize = dirty_bitmap_bytes(npages);
> >> +
> >> + if (bitmap_size < bsize)
> >> + return -EINVAL;
> >> +
> >> + return 0;
> >> +}
> >> +
> >> static int vfio_dma_do_unmap(struct vfio_iommu *iommu,
> >> struct vfio_iommu_type1_dma_unmap *unmap)
> >> {
> >> @@ -2277,6 +2422,92 @@ static long vfio_iommu_type1_ioctl(void *iommu_data,
> >>
> >> return copy_to_user((void __user *)arg, &unmap, minsz) ?
> >> -EFAULT : 0;
> >> + } else if (cmd == VFIO_IOMMU_DIRTY_PAGES) {
> >> + struct vfio_iommu_type1_dirty_bitmap dirty;
> >> + uint32_t mask = VFIO_IOMMU_DIRTY_PAGES_FLAG_START |
> >> + VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP |
> >> + VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP;
> >> + int ret;
> >> +
> >> + if (!iommu->v2)
> >> + return -EACCES;
> >> +
> >> + minsz = offsetofend(struct vfio_iommu_type1_dirty_bitmap,
> >> + flags);
> >> +
> >> + if (copy_from_user(&dirty, (void __user *)arg, minsz))
> >> + return -EFAULT;
> >> +
> >> + if (dirty.argsz < minsz || dirty.flags & ~mask)
> >> + return -EINVAL;
> >> +
> >> + /* only one flag should be set at a time */
> >> + if (__ffs(dirty.flags) != __fls(dirty.flags))
> >> + return -EINVAL;
> >> +
> >> + if (dirty.flags & VFIO_IOMMU_DIRTY_PAGES_FLAG_START) {
> >> + unsigned long iommu_pgsize =
> >> + 1 << __ffs(vfio_pgsize_bitmap(iommu));
> >> +
> >> + mutex_lock(&iommu->lock);
> >> + ret = vfio_dma_all_bitmap_alloc(iommu, iommu_pgsize);
> >> + if (!ret)
> >> + iommu->dirty_page_tracking = true;
> >> + mutex_unlock(&iommu->lock);
> >> + return ret;
> >
> > If a user called this more than once we'd re-allocate all the bitmaps
> > and leak the previous ones. Let's avoid that by testing
> > dirty_page_tracking before we actually take any action. Do we want to
> > return error or success though if dirty tracking is already enabled?
> > I'm inclined to say success because we've provided no means for the
> > user to check the status.
>
> Success seems better.
>
> Thanks,
> Kirti
>
> >
> >> + } else if (dirty.flags & VFIO_IOMMU_DIRTY_PAGES_FLAG_STOP) {
> >> + mutex_lock(&iommu->lock);
> >> + if (iommu->dirty_page_tracking) {
> >> + iommu->dirty_page_tracking = false;
> >> + vfio_dma_all_bitmap_free(iommu);
> >> + }
> >> + mutex_unlock(&iommu->lock);
> >> + return 0;
> >
> > This one essentially does what I'm suggesting above already. Repeated
> > calls take no additional action and don't return an error.
> >
> >> + } else if (dirty.flags &
> >> + VFIO_IOMMU_DIRTY_PAGES_FLAG_GET_BITMAP) {
> >> + struct vfio_iommu_type1_dirty_bitmap_get range;
> >> + unsigned long pgshift;
> >> + size_t data_size = dirty.argsz - minsz;
> >> + uint64_t iommu_pgsize =
> >> + 1 << __ffs(vfio_pgsize_bitmap(iommu));
> >> +
> >> + if (!data_size || data_size < sizeof(range))
> >> + return -EINVAL;
> >> +
> >> + if (copy_from_user(&range, (void __user *)(arg + minsz),
> >> + sizeof(range)))
> >> + return -EFAULT;
> >> +
> >> + // allow only min supported pgsize
> >
> > Proper comment style please.
> >
> >> + if (range.pgsize != iommu_pgsize)
> >> + return -EINVAL;
> >> + if (range.iova & (iommu_pgsize - 1))
> >> + return -EINVAL;
> >> + if (!range.size || range.size & (iommu_pgsize - 1))
> >> + return -EINVAL;
> >> + if (range.iova + range.size < range.iova)
> >> + return -EINVAL;
> >> + if (!access_ok((void __user *)range.bitmap,
> >> + range.bitmap_size))
> >> + return -EINVAL;
> >> +
> >> + pgshift = __ffs(range.pgsize);
> >> + ret = verify_bitmap_size(range.size >> pgshift,
> >> + range.bitmap_size);
> >> + if (ret)
> >> + return ret;
> >> +
> >> + mutex_lock(&iommu->lock);
> >> + if (iommu->dirty_page_tracking)
> >> + ret = vfio_iova_dirty_bitmap(iommu, range.iova,
> >> + range.size, range.pgsize,
> >> + (unsigned char __user *)range.bitmap);
> >> + else
> >> + ret = -EINVAL;
> >> + mutex_unlock(&iommu->lock);
> >> +
> >> + return ret;
> >> + }
> >> }
> >>
> >> return -ENOTTY;
> >
>
- [PATCH v13 Kernel 0/7] KABIs to support migration for VFIO devices, Kirti Wankhede, 2020/03/12
- [PATCH v13 Kernel 1/7] vfio: KABI for migration interface for device state, Kirti Wankhede, 2020/03/12
- [PATCH v13 Kernel 2/7] vfio iommu: Remove atomicity of ref_count of pinned pages, Kirti Wankhede, 2020/03/12
- [PATCH v13 Kernel 3/7] vfio iommu: Add ioctl definition for dirty pages tracking., Kirti Wankhede, 2020/03/12
- [PATCH v13 Kernel 4/7] vfio iommu: Implementation of ioctl to for dirty pages tracking., Kirti Wankhede, 2020/03/12
- Re: [PATCH v13 Kernel 4/7] vfio iommu: Implementation of ioctl to for dirty pages tracking., Dr. David Alan Gilbert, 2020/03/18
- [PATCH v13 Kernel 5/7] vfio iommu: Update UNMAP_DMA ioctl to get dirty bitmap before unmap, Kirti Wankhede, 2020/03/12
- [PATCH v13 Kernel 6/7] vfio iommu: Adds flag to indicate dirty pages tracking capability support, Kirti Wankhede, 2020/03/12
- [PATCH v13 Kernel 7/7] vfio: Selective dirty page tracking if IOMMU backed device pins pages, Kirti Wankhede, 2020/03/12