From: Boris Ostrovsky <boris.ostrovsky@oracle.com>
To: Roman Skakun <rm.skakun@gmail.com>,
Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>,
Juergen Gross <jgross@suse.com>,
Stefano Stabellini <sstabellini@kernel.org>,
xen-devel@lists.xenproject.org, iommu@lists.linux-foundation.org,
linux-kernel@vger.kernel.org, Christoph Hellwig <hch@lst.de>
Cc: Oleksandr Tyshchenko <oleksandr_tyshchenko@epam.com>,
Oleksandr Andrushchenko <oleksandr_andrushchenko@epam.com>,
Volodymyr Babchuk <volodymyr_babchuk@epam.com>,
Roman Skakun <roman_skakun@epam.com>,
Andrii Anisov <andrii_anisov@epam.com>
Subject: Re: [PATCH 2/2] swiotlb-xen: override common mmap and get_sgtable dma ops
Date: Wed, 16 Jun 2021 10:12:55 -0400 [thread overview]
Message-ID: <2834cdc0-534c-4f07-1901-e468a7713c1f@oracle.com> (raw)
In-Reply-To: <20210616114205.38902-2-roman_skakun@epam.com>
On 6/16/21 7:42 AM, Roman Skakun wrote:
> This commit is dedicated to fix incorrect conversion from
> cpu_addr to page address in cases when we get virtual
> address which allocated through xen_swiotlb_alloc_coherent()
> and can be mapped in the vmalloc range.
> As the result, virt_to_page() cannot convert this address
> properly and return incorrect page address.
>
> Need to detect such cases and obtains the page address using
> vmalloc_to_page() instead.
>
> The reference code for mmap() and get_sgtable() was copied
> from kernel/dma/ops_helpers.c and modified to provide
> additional detections as described above.
>
> In order to simplify code there was added a new
> dma_cpu_addr_to_page() helper.
>
> Signed-off-by: Roman Skakun <roman_skakun@epam.com>
> Reviewed-by: Andrii Anisov <andrii_anisov@epam.com>
> ---
> drivers/xen/swiotlb-xen.c | 42 +++++++++++++++++++++++++++++++--------
> 1 file changed, 34 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c
> index 90bc5fc321bc..9331a8500547 100644
> --- a/drivers/xen/swiotlb-xen.c
> +++ b/drivers/xen/swiotlb-xen.c
> @@ -118,6 +118,14 @@ static int is_xen_swiotlb_buffer(struct device *dev, dma_addr_t dma_addr)
> return 0;
> }
>
> +static struct page *cpu_addr_to_page(void *cpu_addr)
> +{
> + if (is_vmalloc_addr(cpu_addr))
> + return vmalloc_to_page(cpu_addr);
> + else
> + return virt_to_page(cpu_addr);
> +}
> +
> static int
> xen_swiotlb_fixup(void *buf, size_t size, unsigned long nslabs)
> {
> @@ -337,7 +345,7 @@ xen_swiotlb_free_coherent(struct device *hwdev, size_t size, void *vaddr,
> int order = get_order(size);
> phys_addr_t phys;
> u64 dma_mask = DMA_BIT_MASK(32);
> - struct page *page;
> + struct page *page = cpu_addr_to_page(vaddr);
>
> if (hwdev && hwdev->coherent_dma_mask)
> dma_mask = hwdev->coherent_dma_mask;
> @@ -349,11 +357,6 @@ xen_swiotlb_free_coherent(struct device *hwdev, size_t size, void *vaddr,
> /* Convert the size to actually allocated. */
> size = 1UL << (order + XEN_PAGE_SHIFT);
>
> - if (is_vmalloc_addr(vaddr))
> - page = vmalloc_to_page(vaddr);
> - else
> - page = virt_to_page(vaddr);
> -
> if (!WARN_ON((dev_addr + size - 1 > dma_mask) ||
> range_straddles_page_boundary(phys, size)) &&
> TestClearPageXenRemapped(page))
> @@ -573,7 +576,23 @@ xen_swiotlb_dma_mmap(struct device *dev, struct vm_area_struct *vma,
> void *cpu_addr, dma_addr_t dma_addr, size_t size,
> unsigned long attrs)
> {
> - return dma_common_mmap(dev, vma, cpu_addr, dma_addr, size, attrs);
> + unsigned long user_count = vma_pages(vma);
> + unsigned long count = PAGE_ALIGN(size) >> PAGE_SHIFT;
> + unsigned long off = vma->vm_pgoff;
> + struct page *page = cpu_addr_to_page(cpu_addr);
> + int ret;
> +
> + vma->vm_page_prot = dma_pgprot(dev, vma->vm_page_prot, attrs);
> +
> + if (dma_mmap_from_dev_coherent(dev, vma, cpu_addr, size, &ret))
> + return ret;
> +
> + if (off >= count || user_count > count - off)
> + return -ENXIO;
> +
> + return remap_pfn_range(vma, vma->vm_start,
> + page_to_pfn(page) + vma->vm_pgoff,
> + user_count << PAGE_SHIFT, vma->vm_page_prot);
> }
I wonder now whether we could avoid code duplication between here and dma_common_mmap()/dma_common_get_sgtable() and use your helper there.
Christoph, would that work? I.e. something like
diff --git a/kernel/dma/ops_helpers.c b/kernel/dma/ops_helpers.c
index 910ae69cae77..43411c2fa47b 100644
--- a/kernel/dma/ops_helpers.c
+++ b/kernel/dma/ops_helpers.c
@@ -12,7 +12,7 @@ int dma_common_get_sgtable(struct device *dev, struct sg_table *sgt,
void *cpu_addr, dma_addr_t dma_addr, size_t size,
unsigned long attrs)
{
- struct page *page = virt_to_page(cpu_addr);
+ struct page *page = cpu_addr_to_page(cpu_addr);
int ret;
ret = sg_alloc_table(sgt, 1, GFP_KERNEL);
@@ -43,7 +43,7 @@ int dma_common_mmap(struct device *dev, struct vm_area_struct *vma,
return -ENXIO;
return remap_pfn_range(vma, vma->vm_start,
- page_to_pfn(virt_to_page(cpu_addr)) + vma->vm_pgoff,
+ page_to_pfn(cpu_addr_to_page(cpu_addr)) + vma->vm_pgoff,
user_count << PAGE_SHIFT, vma->vm_page_prot);
#else
return -ENXIO;
-boris
>
> /*
> @@ -585,7 +604,14 @@ xen_swiotlb_get_sgtable(struct device *dev, struct sg_table *sgt,
> void *cpu_addr, dma_addr_t handle, size_t size,
> unsigned long attrs)
> {
> - return dma_common_get_sgtable(dev, sgt, cpu_addr, handle, size, attrs);
> + struct page *page = cpu_addr_to_page(cpu_addr);
> + int ret;
> +
> + ret = sg_alloc_table(sgt, 1, GFP_KERNEL);
> + if (!ret)
> + sg_set_page(sgt->sgl, page, PAGE_ALIGN(size), 0);
> +
> + return ret;
> }
>
> const struct dma_map_ops xen_swiotlb_dma_ops = {
next prev parent reply other threads:[~2021-06-16 14:13 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-11 9:55 [PATCH] swiotlb-xen: override common mmap and get_sgtable dma ops Roman Skakun
2021-06-11 15:19 ` Boris Ostrovsky
2021-06-14 12:47 ` Roman Skakun
2021-06-14 15:45 ` Boris Ostrovsky
2021-06-16 11:45 ` Roman Skakun
2021-06-16 11:42 ` [PATCH 1/2] Revert "swiotlb-xen: remove xen_swiotlb_dma_mmap and xen_swiotlb_dma_get_sgtable" Roman Skakun
2021-06-16 11:42 ` [PATCH 2/2] swiotlb-xen: override common mmap and get_sgtable dma ops Roman Skakun
2021-06-16 14:12 ` Boris Ostrovsky [this message]
2021-06-16 14:21 ` Christoph Hellwig
2021-06-16 15:33 ` Boris Ostrovsky
2021-06-16 15:35 ` Christoph Hellwig
2021-06-16 15:39 ` Boris Ostrovsky
2021-06-16 15:44 ` Christoph Hellwig
2021-06-22 13:34 ` [PATCH v2] dma-mapping: use vmalloc_to_page for vmalloc addresses Roman Skakun
2021-07-14 0:15 ` Konrad Rzeszutek Wilk
2021-07-15 7:39 ` Roman Skakun
2021-07-15 16:58 ` Boris Ostrovsky
2021-07-15 17:00 ` Christoph Hellwig
2021-07-16 8:39 ` Roman Skakun
2021-07-16 9:35 ` Christoph Hellwig
2021-07-16 12:53 ` Roman Skakun
2021-07-16 15:29 ` Stefano Stabellini
2021-07-17 8:39 ` Roman Skakun
2021-07-19 9:22 ` Christoph Hellwig
2021-07-21 18:39 ` Roman Skakun
2021-07-14 1:23 ` Stefano Stabellini
2021-07-15 7:31 ` Roman Skakun
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=2834cdc0-534c-4f07-1901-e468a7713c1f@oracle.com \
--to=boris.ostrovsky@oracle.com \
--cc=andrii_anisov@epam.com \
--cc=hch@lst.de \
--cc=iommu@lists.linux-foundation.org \
--cc=jgross@suse.com \
--cc=konrad.wilk@oracle.com \
--cc=linux-kernel@vger.kernel.org \
--cc=oleksandr_andrushchenko@epam.com \
--cc=oleksandr_tyshchenko@epam.com \
--cc=rm.skakun@gmail.com \
--cc=roman_skakun@epam.com \
--cc=sstabellini@kernel.org \
--cc=volodymyr_babchuk@epam.com \
--cc=xen-devel@lists.xenproject.org \
/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).