From: Allison Henderson <achender@kernel.org>
To: Wentao Liang <vulab@iscas.ac.cn>,
"David S. Miller" <davem@davemloft.net>,
Eric Dumazet <edumazet@kernel.org>,
Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
Simon Horman <horms@kernel.org>
Cc: netdev@vger.kernel.org, linux-rdma@vger.kernel.org,
rds-devel@oss.oracle.com
Subject: Re: [PATCH] net/rds: drop the device reference for ODP memory regions
Date: Wed, 04 Nov 2026 15:15:46 -0700 [thread overview]
Message-ID: <b9b97bf371302d42319ce21fbd4532d39ee97a28.camel@kernel.org> (raw)
In-Reply-To: <20261104022810.1484182-1-vulab@iscas.ac.cn>
On Wed, 2026-11-04 at 02:28 +0000, Wentao Liang wrote:
> rds_ib_get_mr() takes a reference on the rds_ib_device with
> rds_ib_get_device(). The FRMR path stores it in ibmr->device and
> rds_ib_free_mr() drops it when the MR is freed.
>
> The ODP path allocates the ibmr with kzalloc_obj() and returns it
> without recording the device. rds_ib_odp_mr_worker() only deregisters
> the MR and frees the ibmr, so the reference taken for every ODP MR is
> never dropped and the rds_ib_device can no longer be released.
>
> Store the reference in ibmr->device and drop it in the worker after
> the MR has been deregistered, which also keeps the device and its PD
> alive for as long as the MR exists.
>
> Fixes: 2eafa1746f17 ("net/rds: Handle ODP mr registration/unregistration")
> Cc: stable@vger.kernel.org
> Assisted-by: DeepSeek-V4.1-Flash and OpenAI Codex
> Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
Hi Wentao,
Thanks for the patch. The fix itself looks reasonable, but few things to update: since
this is a bug fix, it should should target net with [PATCH net v2]. The commit message
should mention how this was found, and whether it was tested. Reproducing this needs an
ODP-capable device, so if it was only build-tested, please say so. It looks like there's
a few check patch nits to work through as well:
https://patchwork.kernel.org/project/netdevbpf/patch/20261104022810.1484182-1-vulab@iscas.ac.cn/
Other than that it looks straight forward to me.
Thanks,
Allison
> ---
> net/rds/ib_rdma.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/net/rds/ib_rdma.c b/net/rds/ib_rdma.c
> index db7e92e7bd29..8cd8295f66a3 100644
> --- a/net/rds/ib_rdma.c
> +++ b/net/rds/ib_rdma.c
> @@ -591,6 +591,7 @@ void *rds_ib_get_mr(struct scatterlist *sg, unsigned long nents,
> }
> ibmr->u.mr = ib_mr;
> ibmr->odp = 1;
> + ibmr->device = rds_ibdev;
>
> sge.addr = virt_addr;
> sge.length = length;
> @@ -697,5 +698,6 @@ static void rds_ib_odp_mr_worker(struct work_struct *work)
>
> ibmr = container_of(work, struct rds_ib_mr, work.work);
> ib_dereg_mr(ibmr->u.mr);
> + rds_ib_dev_put(ibmr->device);
> kfree(ibmr);
> }
next prev parent reply other threads:[~2026-11-04 22:15 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-04 2:28 [PATCH] net/rds: drop the device reference for ODP memory regions Wentao Liang
2026-11-04 2:33 ` netdev-bot+sinfo
2026-11-04 7:11 ` Krzysztof Kozlowski
2026-11-04 22:15 ` Allison Henderson [this message]
2026-11-08 3:04 ` netdev-bot+sashiko
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=b9b97bf371302d42319ce21fbd4532d39ee97a28.camel@kernel.org \
--to=achender@kernel.org \
--cc=davem@davemloft.net \
--cc=edumazet@kernel.org \
--cc=horms@kernel.org \
--cc=kuba@kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=rds-devel@oss.oracle.com \
--cc=vulab@iscas.ac.cn \
/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