From: Alex Williamson <alex.williamson@redhat.com>
To: Steve Sistare <steven.sistare@oracle.com>
Cc: kvm@vger.kernel.org, Cornelia Huck <cohuck@redhat.com>,
Kirti Wankhede <kwankhede@nvidia.com>
Subject: Re: [PATCH V2 4/9] vfio/type1: implement unmap all
Date: Fri, 22 Jan 2021 14:22:47 -0700 [thread overview]
Message-ID: <20210122142247.0046a862@omen.home.shazbot.org> (raw)
In-Reply-To: <1611078509-181959-5-git-send-email-steven.sistare@oracle.com>
On Tue, 19 Jan 2021 09:48:24 -0800
Steve Sistare <steven.sistare@oracle.com> wrote:
> Implement VFIO_DMA_UNMAP_FLAG_ALL.
>
> Signed-off-by: Steve Sistare <steven.sistare@oracle.com>
> ---
> drivers/vfio/vfio_iommu_type1.c | 22 +++++++++++++++++-----
> 1 file changed, 17 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> index c687174..ef83018 100644
> --- a/drivers/vfio/vfio_iommu_type1.c
> +++ b/drivers/vfio/vfio_iommu_type1.c
> @@ -1100,6 +1100,7 @@ static int vfio_dma_do_unmap(struct vfio_iommu *iommu,
> unsigned long pgshift;
> dma_addr_t iova = unmap->iova;
> unsigned long size = unmap->size;
> + bool unmap_all = !!(unmap->flags & VFIO_DMA_UNMAP_FLAG_ALL);
>
> mutex_lock(&iommu->lock);
>
> @@ -1109,8 +1110,13 @@ static int vfio_dma_do_unmap(struct vfio_iommu *iommu,
> if (iova & (pgsize - 1))
> goto unlock;
>
> - if (!size || size & (pgsize - 1))
> + if (unmap_all) {
> + if (iova || size)
> + goto unlock;
> + size = SIZE_MAX;
> + } else if (!size || size & (pgsize - 1)) {
> goto unlock;
> + }
>
> if (iova + size - 1 < iova || size > SIZE_MAX)
> goto unlock;
> @@ -1154,7 +1160,7 @@ static int vfio_dma_do_unmap(struct vfio_iommu *iommu,
> * will only return success and a size of zero if there were no
> * mappings within the range.
> */
> - if (iommu->v2) {
> + if (iommu->v2 && !unmap_all) {
> dma = vfio_find_dma(iommu, iova, 1);
> if (dma && dma->iova != iova)
> goto unlock;
> @@ -1165,7 +1171,7 @@ static int vfio_dma_do_unmap(struct vfio_iommu *iommu,
> }
>
> ret = 0;
> - while ((dma = vfio_find_dma(iommu, iova, size))) {
> + while ((dma = vfio_find_dma_first(iommu, iova, size))) {
Why is this necessary? Isn't vfio_find_dma_first() O(logN) for this
operation while vfio_find_dma() is O(1)?
> if (!iommu->v2 && iova > dma->iova)
> break;
> /*
> @@ -2538,6 +2544,7 @@ static int vfio_iommu_type1_check_extension(struct vfio_iommu *iommu,
> case VFIO_TYPE1_IOMMU:
> case VFIO_TYPE1v2_IOMMU:
> case VFIO_TYPE1_NESTING_IOMMU:
> + case VFIO_UNMAP_ALL:
> return 1;
> case VFIO_DMA_CC_IOMMU:
> if (!iommu)
> @@ -2710,6 +2717,8 @@ static int vfio_iommu_type1_unmap_dma(struct vfio_iommu *iommu,
> {
> struct vfio_iommu_type1_dma_unmap unmap;
> struct vfio_bitmap bitmap = { 0 };
> + uint32_t mask = VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP |
> + VFIO_DMA_UNMAP_FLAG_ALL;
> unsigned long minsz;
> int ret;
>
> @@ -2718,8 +2727,11 @@ static int vfio_iommu_type1_unmap_dma(struct vfio_iommu *iommu,
> if (copy_from_user(&unmap, (void __user *)arg, minsz))
> return -EFAULT;
>
> - if (unmap.argsz < minsz ||
> - unmap.flags & ~VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP)
> + if (unmap.argsz < minsz || unmap.flags & ~mask)
> + return -EINVAL;
> +
> + if ((unmap.flags & VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP) &&
> + (unmap.flags & ~VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP))
Somehow we're jumping from unmap-all and dirty-bitmap being mutually
exclusive to dirty-bitmap is absolutely exclusive, which seems like a
future bug or maintenance issue. Let's just test specifically what
we're deciding is unsupported. Thanks,
Alex
> return -EINVAL;
>
> if (unmap.flags & VFIO_DMA_UNMAP_FLAG_GET_DIRTY_BITMAP) {
next prev parent reply other threads:[~2021-01-22 21:25 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-19 17:48 [PATCH V2 0/9] vfio virtual address update Steve Sistare
2021-01-19 17:48 ` [PATCH V2 1/9] vfio: option to unmap all Steve Sistare
2021-01-19 17:48 ` [PATCH V2 2/9] vfio/type1: find first dma Steve Sistare
2021-01-19 17:48 ` [PATCH V2 3/9] vfio/type1: unmap cleanup Steve Sistare
2021-01-19 17:48 ` [PATCH V2 4/9] vfio/type1: implement unmap all Steve Sistare
2021-01-22 21:22 ` Alex Williamson [this message]
2021-01-27 23:07 ` Steven Sistare
2021-01-19 17:48 ` [PATCH V2 5/9] vfio: interfaces to update vaddr Steve Sistare
2021-01-19 17:48 ` [PATCH V2 6/9] vfio/type1: implement " Steve Sistare
2021-01-22 21:48 ` Alex Williamson
2021-01-27 23:23 ` Steven Sistare
2021-01-19 17:48 ` [PATCH V2 7/9] vfio: iommu driver notify callback Steve Sistare
2021-01-19 17:48 ` [PATCH V2 8/9] vfio/type1: implement " Steve Sistare
2021-01-22 23:00 ` Alex Williamson
2021-01-27 23:24 ` Steven Sistare
2021-01-19 17:48 ` [PATCH V2 9/9] vfio/type1: block on invalid vaddr Steve Sistare
2021-01-22 22:59 ` Alex Williamson
2021-01-27 23:25 ` Steven Sistare
2021-01-28 0:03 ` Alex Williamson
2021-01-28 17:07 ` Alex Williamson
2021-01-28 17:18 ` Steven Sistare
[not found] ` <55725169-de0d-4019-f96c-ded20dfde0d7@oracle.com>
2021-01-29 17:05 ` [PATCH V2 0/9] vfio virtual address update Alex Williamson
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=20210122142247.0046a862@omen.home.shazbot.org \
--to=alex.williamson@redhat.com \
--cc=cohuck@redhat.com \
--cc=kvm@vger.kernel.org \
--cc=kwankhede@nvidia.com \
--cc=steven.sistare@oracle.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).