From: Jason Gunthorpe <jgg@nvidia.com>
To: Alex Williamson <alex.williamson@redhat.com>
Cc: cohuck@redhat.com, kvm@vger.kernel.org,
linux-kernel@vger.kernel.org, peterx@redhat.com
Subject: Re: [PATCH v1 12/14] vfio/type1: Support batching of device mappings
Date: Mon, 8 Mar 2021 21:04:06 -0400 [thread overview]
Message-ID: <20210309010406.GE4247@nvidia.com> (raw)
In-Reply-To: <161524017090.3480.6508004360325488879.stgit@gimli.home>
On Mon, Mar 08, 2021 at 02:49:31PM -0700, Alex Williamson wrote:
> Populate the page array to the extent available to enable batching.
>
> Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
> drivers/vfio/vfio_iommu_type1.c | 10 +++++++++-
> 1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index e89f11141dee..d499bccfbe3f 100644
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -628,6 +628,8 @@ static int vaddr_get_pfns(struct vfio_iommu *iommu, struct vfio_dma *dma,
> vma = find_vma_intersection(mm, vaddr, vaddr + 1);
>
> if (vma && vma->vm_flags & VM_PFNMAP) {
> + unsigned long count, i;
> +
> if ((dma->prot & IOMMU_WRITE && !(vma->vm_flags & VM_WRITE)) ||
> (dma->prot & IOMMU_READ && !(vma->vm_flags & VM_READ))) {
> ret = -EFAULT;
> @@ -678,7 +680,13 @@ static int vaddr_get_pfns(struct vfio_iommu *iommu, struct vfio_dma *dma,
>
> *pfn = ((vaddr - vma->vm_start) >> PAGE_SHIFT) +
> dma->pfnmap->base_pfn;
> - ret = 1;
> + count = min_t(long,
> + (vma->vm_end - vaddr) >> PAGE_SHIFT, npages);
> +
> + for (i = 0; i < count; i++)
> + pages[i] = pfn_to_page(*pfn + i);
This isn't safe, we can't pass a VM_PFNMAP pfn into pfn_to_page(). The
whole api here with the batch should be using pfns not struct pages
Also.. this is not nice at all:
static int put_pfn(unsigned long pfn, int prot)
{
if (!is_invalid_reserved_pfn(pfn)) {
struct page *page = pfn_to_page(pfn);
unpin_user_pages_dirty_lock(&page, 1, prot & IOMMU_WRITE);
The manner in which the PFN was obtained should be tracked internally
to VFIO, not deduced externally by the pfn type. *only* pages returned
by pin_user_pages() should be used with unpin_user_pages() - the other
stuff must be kept distinct.
This is actually another bug with the way things are today, as if the
user gets a PFNMAP VMA that happens to point to a struct page (eg a
MIXEDMAP, these things exist in the kernel), the unpin will explode
when it gets here.
Something like what hmm_range_fault() does where the high bits of the
pfn encode information about it (there is always PAGE_SHIFT high bits
available for use) is much cleaner/safer.
Jason
next prev parent reply other threads:[~2021-03-09 1:05 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-03-08 21:47 [PATCH v1 00/14] vfio: Device memory DMA mapping improvements Alex Williamson
2021-03-08 21:47 ` [PATCH v1 01/14] vfio: Create vfio_fs_type with inode per device Alex Williamson
2021-03-09 8:36 ` Christoph Hellwig
2021-04-09 4:54 ` 答复: " Zengtao (B)
2021-04-09 14:24 ` Alex Williamson
2021-04-09 17:32 ` Jason Gunthorpe
2021-04-12 4:03 ` 答复: " Zengtao (B)
2021-04-12 4:09 ` Zengtao (B)
2021-03-08 21:47 ` [PATCH v1 02/14] vfio: Update vfio_add_group_dev() API Alex Williamson
2021-03-10 7:48 ` Christoph Hellwig
2021-03-10 12:19 ` Jason Gunthorpe
2021-03-10 15:28 ` Alex Williamson
2021-03-11 11:23 ` Christoph Hellwig
2021-03-08 21:47 ` [PATCH v1 03/14] vfio: Export unmap_mapping_range() wrapper Alex Williamson
2021-03-08 21:48 ` [PATCH v1 04/14] vfio/pci: Use vfio_device_unmap_mapping_range() Alex Williamson
2021-03-08 21:48 ` [PATCH v1 05/14] vfio: Create a vfio_device from vma lookup Alex Williamson
2021-03-08 21:48 ` [PATCH v1 06/14] vfio: Add vma to pfn callback Alex Williamson
2021-03-09 0:33 ` Jason Gunthorpe
2021-03-08 21:48 ` [PATCH v1 07/14] vfio: Add a device notifier interface Alex Williamson
2021-03-09 0:46 ` Jason Gunthorpe
2021-03-09 15:45 ` Alex Williamson
2021-03-09 16:47 ` Jason Gunthorpe
2021-03-10 7:56 ` Christoph Hellwig
2021-03-19 22:25 ` Alex Williamson
2021-03-22 15:16 ` Christoph Hellwig
2021-03-08 21:48 ` [PATCH v1 08/14] vfio/pci: Notify on device release Alex Williamson
2021-03-08 21:48 ` [PATCH v1 09/14] vfio/type1: Refactor pfn_list clearing Alex Williamson
2021-03-10 8:01 ` Christoph Hellwig
2021-03-08 21:49 ` [PATCH v1 10/14] vfio/type1: Pass iommu and dma objects through to vaddr_get_pfn Alex Williamson
2021-03-08 21:49 ` [PATCH v1 11/14] vfio/type1: Register device notifier Alex Williamson
2021-03-10 8:03 ` Christoph Hellwig
2021-03-08 21:49 ` [PATCH v1 12/14] vfio/type1: Support batching of device mappings Alex Williamson
2021-03-09 1:04 ` Jason Gunthorpe [this message]
2021-03-08 21:49 ` [PATCH v1 13/14] vfio: Remove extern from declarations across vfio Alex Williamson
2021-03-09 0:21 ` Halil Pasic
2021-03-09 1:07 ` Jason Gunthorpe
2021-03-08 21:49 ` [PATCH v1 14/14] vfio: Cleanup use of bare unsigned Alex Williamson
2021-03-09 1:07 ` Jason Gunthorpe
2021-03-09 8:31 ` Christoph Hellwig
2021-03-09 1:06 ` [PATCH v1 00/14] vfio: Device memory DMA mapping improvements Jason Gunthorpe
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210309010406.GE4247@nvidia.com \
--to=jgg@nvidia.com \
--cc=alex.williamson@redhat.com \
--cc=cohuck@redhat.com \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=peterx@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).