From: Chaitanya Kulkarni <chaitanyak@nvidia.com>
To: Casey Chen <cachen@purestorage.com>,
"linux-nvme@lists.infradead.org" <linux-nvme@lists.infradead.org>,
"linux-block@vger.kernel.org" <linux-block@vger.kernel.org>
Cc: "yzhong@purestorage.com" <yzhong@purestorage.com>,
"sconnor@purestorage.com" <sconnor@purestorage.com>,
"axboe@kernel.dk" <axboe@kernel.dk>,
"mkhalfella@purestorage.com" <mkhalfella@purestorage.com>
Subject: Re: [PATCH 1/1] nvme: fix use-after-free of admin queue via stale pointer
Date: Wed, 29 Oct 2025 23:00:29 +0000 [thread overview]
Message-ID: <65062542-e87c-4026-a58a-d5a29d03b8c9@nvidia.com> (raw)
In-Reply-To: <20251029210853.20768-2-cachen@purestorage.com>
On 10/29/25 2:08 PM, Casey Chen wrote:
> From: Yuanyuan Zhong <yzhong@purestorage.com>
>
> When a controller is deleted (e.g., via sysfs delete_controller), the
> admin queue is freed while userspace may still have open fd to the
> namespace block device. Userspace can issue IOCTLs on the open fd
> that access the freed admin queue through the stale ns->ctrl->admin_q
> pointer, causing a use-after-free.
>
> Fix this by taking an additional reference on the admin queue during
> namespace allocation and releasing it during namespace cleanup.
>
> Signed-off-by: Casey Chen <cachen@purestorage.com>
> Signed-off-by: Seamus Connor <sconnor@purestorage.com>
> ---
> drivers/nvme/host/core.c | 10 +++++++++-
> 1 file changed, 9 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
> index 8d8af58e79d1..184a6096a2be 100644
> --- a/drivers/nvme/host/core.c
> +++ b/drivers/nvme/host/core.c
> @@ -687,6 +687,7 @@ static void nvme_free_ns(struct kref *kref)
> {
> struct nvme_ns *ns = container_of(kref, struct nvme_ns, kref);
>
> + blk_put_queue(ns->ctrl->admin_q);
> put_disk(ns->disk);
> nvme_put_ns_head(ns->head);
> nvme_put_ctrl(ns->ctrl);
> @@ -3903,9 +3904,14 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, struct nvme_ns_info *info)
> struct gendisk *disk;
> int node = ctrl->numa_node;
>
would be a good idea to add a comment at both places to explain why we
are taking this additional reference ? since this is specifically needed
for userspace.
> + if (!blk_get_queue(ctrl->admin_q)) {
> + dev_err(ctrl->device, "failed to get admin_q %p\n", ctrl->admin_q);
> + return;
> + }
> +
> ns = kzalloc_node(sizeof(*ns), GFP_KERNEL, node);
-ck
next prev parent reply other threads:[~2025-10-29 23:00 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-10-29 21:08 [PATCH 0/1] cover letter Casey Chen
2025-10-29 21:08 ` [PATCH 1/1] nvme: fix use-after-free of admin queue via stale pointer Casey Chen
2025-10-29 23:00 ` Chaitanya Kulkarni [this message]
2025-10-29 23:33 ` Ming Lei
2025-10-30 0:39 ` Keith Busch
2025-10-30 8:12 ` Chaitanya Kulkarni
2025-10-31 2:54 ` Keith Busch
2025-10-31 8:16 ` Chaitanya Kulkarni
2025-10-31 19:19 ` Casey Chen
2025-10-31 19:31 ` Chaitanya Kulkarni
2025-11-04 22:39 ` Casey Chen
2025-11-04 23:00 ` Keith Busch
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=65062542-e87c-4026-a58a-d5a29d03b8c9@nvidia.com \
--to=chaitanyak@nvidia.com \
--cc=axboe@kernel.dk \
--cc=cachen@purestorage.com \
--cc=linux-block@vger.kernel.org \
--cc=linux-nvme@lists.infradead.org \
--cc=mkhalfella@purestorage.com \
--cc=sconnor@purestorage.com \
--cc=yzhong@purestorage.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