public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Lizhi Hou <lizhi.hou@amd.com>
To: Vineet Agarwal <agarwal.vineet2006@gmail.com>,
	<mamin506@gmail.com>, <ogabbay@kernel.org>
Cc: <dri-devel@lists.freedesktop.org>, <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v3] drm/amdxdna: fix pinned_vm accounting and error handling in user buffer pinning
Date: Mon, 4 May 2026 09:05:05 -0700	[thread overview]
Message-ID: <5a669bab-bceb-529c-351b-20fbb53c9ddd@amd.com> (raw)
In-Reply-To: <20260503055258.643546-1-agarwal.vineet2006@gmail.com>


On 5/2/26 22:50, Vineet Agarwal wrote:
> [You don't often get email from agarwal.vineet2006@gmail.com. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]
>
> amdxdna_get_ubuf() incorrectly accounted mm->pinned_vm using the
> requested number of pages before pin_user_pages_fast() completed.
>
> Since pin_user_pages_fast() can return partial success, this could
> lead to incorrect accounting and inconsistent cleanup on failure.

If pin_user_pages_fast() returns partial success, amdxdna_get_ubuf 
returns and pinned_vm will be restored. I do not think it is an issue here.

>
> Additionally, the RLIMIT_MEMLOCK check was performed after pinning,
> allowing excessive pin attempts before validation.

RLIMIT_MEMLOCK check is on line

    if (new_pinned > lock_limit && !capable(CAP_IPC_LOCK))

which is before pin_user_pages_fast. I do not see anything to fix.

>
> Fix this by:
>    - checking RLIMIT_MEMLOCK before attempting to pin pages
>    - handling partial pinning correctly and ensuring proper cleanup
>    - updating mm->pinned_vm only after all pages are successfully pinned
>    - removing incorrect error-path accounting and double-subtraction
>
> Also fix missing rollback when dma_buf_export() fails, which could
> leave mm->pinned_vm incremented without a corresponding release.

when dma_buf_export() fails, it jumps to destroy_pages which cleanup 
everything including restore pinned_vm.


Lizhi

>
> This ensures correct pinned memory accounting and consistent error
> handling aligned with other subsystems using GUP.
>
> Signed-off-by: Vineet Agarwal <agarwal.vineet2006@gmail.com>
>
> Changes in v3:
> - Fix missing pinned_vm rollback when dma_buf_export() fails
> ---
>   drivers/accel/amdxdna/amdxdna_ubuf.c | 26 +++++++++++++-------------
>   1 file changed, 13 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/accel/amdxdna/amdxdna_ubuf.c b/drivers/accel/amdxdna/amdxdna_ubuf.c
> index fb999aa25318..efce6b94fb0c 100644
> --- a/drivers/accel/amdxdna/amdxdna_ubuf.c
> +++ b/drivers/accel/amdxdna/amdxdna_ubuf.c
> @@ -129,7 +129,7 @@ struct dma_buf *amdxdna_get_ubuf(struct drm_device *dev,
>                                   u32 num_entries, void __user *va_entries)
>   {
>          struct amdxdna_dev *xdna = to_xdna_dev(dev);
> -       unsigned long lock_limit, new_pinned;
> +       unsigned long lock_limit;
>          struct amdxdna_drm_va_entry *va_ent;
>          struct amdxdna_ubuf_priv *ubuf;
>          u32 npages, start = 0;
> @@ -176,18 +176,17 @@ struct dma_buf *amdxdna_get_ubuf(struct drm_device *dev,
>
>          ubuf->nr_pages = exp_info.size >> PAGE_SHIFT;
>          lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
> -       new_pinned = atomic64_add_return(ubuf->nr_pages, &ubuf->mm->pinned_vm);
> -       if (new_pinned > lock_limit && !capable(CAP_IPC_LOCK)) {
> -               XDNA_DBG(xdna, "New pin %ld, limit %ld, cap %d",
> -                        new_pinned, lock_limit, capable(CAP_IPC_LOCK));
> +
> +       if (ubuf->nr_pages + atomic64_read(&ubuf->mm->pinned_vm) > lock_limit &&
> +               !capable(CAP_IPC_LOCK)) {
>                  ret = -ENOMEM;
> -               goto sub_pin_cnt;
> +               goto free_ent;
>          }
>
>          ubuf->pages = kvmalloc_objs(*ubuf->pages, ubuf->nr_pages);
>          if (!ubuf->pages) {
>                  ret = -ENOMEM;
> -               goto sub_pin_cnt;
> +               goto free_ent;
>          }
>
>          for (i = 0; i < num_entries; i++) {
> @@ -196,15 +195,17 @@ struct dma_buf *amdxdna_get_ubuf(struct drm_device *dev,
>                  ret = pin_user_pages_fast(va_ent[i].vaddr, npages,
>                                            FOLL_WRITE | FOLL_LONGTERM,
>                                            &ubuf->pages[start]);
> -               if (ret < 0 || ret != npages) {
> +               if (ret < 0)
> +                       goto destroy_pages;
> +               start += ret;
> +               if (ret != npages) {
>                          ret = -ENOMEM;
> -                       XDNA_ERR(xdna, "Failed to pin pages ret %d", ret);
>                          goto destroy_pages;
>                  }
> -
> -               start += ret;
>          }
>
> +       atomic64_add(ubuf->nr_pages, &ubuf->mm->pinned_vm);
> +
>          exp_info.ops = &amdxdna_ubuf_dmabuf_ops;
>          exp_info.priv = ubuf;
>          exp_info.flags = O_RDWR | O_CLOEXEC;
> @@ -212,6 +213,7 @@ struct dma_buf *amdxdna_get_ubuf(struct drm_device *dev,
>          dbuf = dma_buf_export(&exp_info);
>          if (IS_ERR(dbuf)) {
>                  ret = PTR_ERR(dbuf);
> +               atomic64_sub(ubuf->nr_pages, &ubuf->mm->pinned_vm);
>                  goto destroy_pages;
>          }
>          kvfree(va_ent);
> @@ -222,8 +224,6 @@ struct dma_buf *amdxdna_get_ubuf(struct drm_device *dev,
>          if (start)
>                  unpin_user_pages(ubuf->pages, start);
>          kvfree(ubuf->pages);
> -sub_pin_cnt:
> -       atomic64_sub(ubuf->nr_pages, &ubuf->mm->pinned_vm);
>   free_ent:
>          kvfree(va_ent);
>   free_ubuf:
> --
> 2.54.0
>

      reply	other threads:[~2026-05-04 16:05 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-02  3:17 [PATCH] drm/amdxdna: fix pinned_vm accounting and rlimit rollback Vineet Agarwal
2026-05-03  4:49 ` [PATCH] drm/amdxdna: fix pinned_vm accounting and error handling in user buffer pinning Vineet Agarwal
2026-05-03  5:50 ` [PATCH v3] " Vineet Agarwal
2026-05-04 16:05   ` Lizhi Hou [this message]

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=5a669bab-bceb-529c-351b-20fbb53c9ddd@amd.com \
    --to=lizhi.hou@amd.com \
    --cc=agarwal.vineet2006@gmail.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mamin506@gmail.com \
    --cc=ogabbay@kernel.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