From: Leon Romanovsky <leon@kernel.org>
To: Hungyu Lin <dennylin0707@gmail.com>
Cc: jgg@ziepe.ca, yishaih@nvidia.com, linux-rdma@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH] RDMA/mlx5: Validate the DEVX command length before setting umem valid bits
Date: Mon, 16 Nov 2026 15:39:27 +0200 [thread overview]
Message-ID: <20261116133927.GO11319@unreal> (raw)
In-Reply-To: <20261115164658.88420-1-dennylin0707@gmail.com>
On Sun, Nov 15, 2026 at 04:46:58PM +0000, Hungyu Lin wrote:
> devx_set_umem_valid() sets the umem valid bits in the command supplied
> by user space, at offsets that depend on the opcode. The DEVX object
> create and modify methods only require the command to be as long as
> general_obj_in_cmd_hdr, which is 16 bytes, so when the command is
> shorter than the layout of its opcode the bits are written past the end
> of the command buffer. For MLX5_CMD_OP_CREATE_QP the modified words are
> 232 and 268 bytes into the command.
>
> The command buffer is allocated with uverbs_alloc(), so the write lands
> either past the internal buffer of the bundle, which normally lives on
> the stack of ib_uverbs_cmd_verbs(), or past a separate allocation.
>
> Fail with -EINVAL when the command does not cover the layout that is
> modified, as devx_handle_mkey_create() already does for
> MLX5_CMD_OP_CREATE_MKEY.
>
> Fixes: ba1a057da2f1 ("IB/mlx5: Set valid umem bit on DEVX")
> Cc: stable@vger.kernel.org
> Signed-off-by: Hungyu Lin <dennylin0707@gmail.com>
> ---
> drivers/infiniband/hw/mlx5/devx.c | 44 ++++++++++++++++++++++++++++++++-------
> 1 file changed, 37 insertions(+), 7 deletions(-)
Thanks for the report. We are aware of this issue and are working on a fix.
Unfortunately, the proposed solution would require DEVX userspace applications
to provide a size that is already known to the kernel, preventing us from
extending these structures in the future.
Thanks
next prev parent reply other threads:[~2026-11-16 13:39 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-15 16:46 [PATCH] RDMA/mlx5: Validate the DEVX command length before setting umem valid bits Hungyu Lin
2026-11-15 17:01 ` sashiko-bot
2026-11-16 13:39 ` Leon Romanovsky [this message]
2026-11-16 15:25 ` Hungyu Lin
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=20261116133927.GO11319@unreal \
--to=leon@kernel.org \
--cc=dennylin0707@gmail.com \
--cc=jgg@ziepe.ca \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=yishaih@nvidia.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.