From: Yishai Hadas <yishaih@nvidia.com>
To: <jgg@ziepe.ca>, <leon@kernel.org>
Cc: <linux-rdma@vger.kernel.org>, <netdev@vger.kernel.org>,
<saeedm@nvidia.com>, <tariqt@nvidia.com>, <mbloch@nvidia.com>,
<kuba@kernel.org>, <pabeni@redhat.com>, <edumazet@kernel.org>,
<davem@davemloft.net>, <andrew+netdev@lunn.ch>,
<dtatulea@nvidia.com>, <moshe@nvidia.com>, <yishaih@nvidia.com>,
<maorg@nvidia.com>
Subject: [PATCH V2 rdma-next 15/21] RDMA/mlx5: Ignore completion events for a kernel CQ that is not ready
Date: Tue, 17 Nov 2026 19:09:08 +0200 [thread overview]
Message-ID: <20261117170914.27410-16-yishaih@nvidia.com> (raw)
In-Reply-To: <20261117170914.27410-1-yishaih@nvidia.com>
A kernel CQ can be reached from the completion interrupt before it is
set up: __ib_alloc_cq() sets its poll context and handler only after
ops.create_cq(), and devr->c0 has no handler, as it is never armed.
A trusted NIC raises no event for a CQ that is not armed, but a hostile
NIC in adversarial mode can. For devr->c0 that ends in a NULL call. For
a CQ from __ib_alloc_cq(), on a weakly ordered CPU, the interrupt can
see the handler but not the poll context, and crash in irq_poll.
In adversarial mode, give kernel CQs their own event callback, which
ignores events until the first arm, the point where the core has
finished setting up the CQ. mlx5_ib_arm_cq() marks the CQ ready with a
release before the doorbell, and the callback checks it with an
acquire.
Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
---
drivers/infiniband/hw/mlx5/cq.c | 31 +++++++++++++++++++++++++++-
drivers/infiniband/hw/mlx5/mlx5_ib.h | 2 ++
2 files changed, 32 insertions(+), 1 deletion(-)
diff --git a/drivers/infiniband/hw/mlx5/cq.c b/drivers/infiniband/hw/mlx5/cq.c
index a5197318d96f..831bdf238319 100644
--- a/drivers/infiniband/hw/mlx5/cq.c
+++ b/drivers/infiniband/hw/mlx5/cq.c
@@ -49,6 +49,24 @@ static void mlx5_ib_cq_comp(struct mlx5_core_cq *cq, struct mlx5_eqe *eqe)
ibcq->comp_handler(ibcq, ibcq->cq_context);
}
+/*
+ * A hostile NIC can raise an event for a kernel CQ before the core arms it,
+ * when its poll context and handler may not be visible yet. Ignore events
+ * until the core finished setting up the CQ.
+ */
+static void mlx5_ib_kernel_cq_comp(struct mlx5_core_cq *cq,
+ struct mlx5_eqe *eqe)
+{
+ /*
+ * Only the first events can find it unset. Pairs with the release in
+ * mlx5_ib_arm_cq().
+ */
+ if (!smp_load_acquire(&to_mibcq(cq)->ready))
+ return;
+
+ mlx5_ib_cq_comp(cq, eqe);
+}
+
static void mlx5_ib_cq_event(struct mlx5_core_cq *mcq, enum mlx5_event type)
{
struct mlx5_ib_cq *cq = container_of(mcq, struct mlx5_ib_cq, mcq);
@@ -853,6 +871,16 @@ int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
ret = 1;
spin_unlock_irqrestore(&cq->lock, irq_flags);
+ /*
+ * The core arms a kernel CQ only after its poll context and handler
+ * are set. Mark it ready before the doorbell, so an event raised
+ * right after it is not dropped.
+ */
+ if (ibcq->device->adversarial_mode && !cq->ready) {
+ /* Pairs with the acquire in mlx5_ib_kernel_cq_comp() */
+ smp_store_release(&cq->ready, true);
+ }
+
mlx5_cq_arm(&cq->mcq,
(flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ?
MLX5_CQ_DB_REQ_NOT_SOL : MLX5_CQ_DB_REQ_NOT,
@@ -1284,7 +1312,8 @@ int mlx5_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
MLX5_SET(cqc, cqc, c_eqn_or_apu_element, eqn);
MLX5_SET64(cqc, cqc, dbr_addr, cq->db.dma);
- cq->mcq.comp = mlx5_ib_cq_comp;
+ cq->mcq.comp = dev->ib_dev.adversarial_mode ? mlx5_ib_kernel_cq_comp :
+ mlx5_ib_cq_comp;
err = mlx5_core_create_cq(dev->mdev, &cq->mcq, cqb, inlen, out,
sizeof(out));
diff --git a/drivers/infiniband/hw/mlx5/mlx5_ib.h b/drivers/infiniband/hw/mlx5/mlx5_ib.h
index 5b7a3e3b456d..84aad754435e 100644
--- a/drivers/infiniband/hw/mlx5/mlx5_ib.h
+++ b/drivers/infiniband/hw/mlx5/mlx5_ib.h
@@ -592,6 +592,8 @@ struct mlx5_ib_cq {
enum ib_cq_notify_flags notify_flags;
struct work_struct notify_work;
u16 private_flags; /* Use mlx5_ib_cq_pr_flags */
+ /* adversarial_mode: kernel CQ events are handled only once set */
+ bool ready;
/* adversarial_mode: polled by the verifying UMR poller */
bool umr_verify;
};
--
2.18.1
next prev parent reply other threads:[~2026-11-17 17:11 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-17 17:08 [PATCH V2 rdma-next 00/21] mlx5: Adversarial mode for confidential computing guests Yishai Hadas
2026-11-17 17:08 ` [PATCH V2 rdma-next 01/21] net/mlx5: Add mlx5_trigger_health_work_force() Yishai Hadas
2026-11-17 17:08 ` [PATCH V2 rdma-next 02/21] net/mlx5: Derive adversarial mode from CC guest memory encryption Yishai Hadas
2026-11-17 17:08 ` [PATCH V2 rdma-next 03/21] net/mlx5: Add capability filtering in adversarial mode Yishai Hadas
2026-11-17 17:08 ` [PATCH V2 rdma-next 04/21] net/mlx5: Catch too low max number of EQs Yishai Hadas
2026-11-17 17:08 ` [PATCH V2 rdma-next 05/21] net/mlx5: Add macros for reading log-based caps Yishai Hadas
2026-11-17 17:08 ` [PATCH V2 rdma-next 06/21] RDMA/mlx5: Reject QP transport types not enabled by FW capability Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 07/21] RDMA/mlx5: Require IB_QPT_DRIVER for the DC vendor QP types Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 08/21] RDMA/core: Suppress kverbs_provider for adversarial_mode devices Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 09/21] RDMA/mlx5: Set ib_device::adversarial_mode from mlx5_core Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 10/21] RDMA/mlx5: Remove duplicate mlx5_ib_umr_context struct Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 11/21] RDMA/mlx5: Add shadow FIFO infrastructure for UMR completions Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 12/21] RDMA/mlx5: Verify UMR CQ polling against a hostile NIC Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 13/21] RDMA/mlx5: Keep the UMR QP failed once a forged CQE is detected Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 14/21] RDMA/core: Set the CQ comp_handler after its poll context is set up Yishai Hadas
2026-11-17 17:09 ` Yishai Hadas [this message]
2026-11-17 17:09 ` [PATCH V2 rdma-next 16/21] RDMA/mlx5: Validate EQE port and GSI port before indexing devr.ports[] Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 17/21] RDMA/mlx5: Validate firmware RQ/SQ state before sqrq_trans[] lookup Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 18/21] RDMA/mlx5: Clamp log_max_cq_sz shift in CQ size checks Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 19/21] RDMA/mlx5: Clamp log_max_* shifts in QP and WQ limit checks Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 20/21] RDMA/mlx5: Clamp log_max_* shifts in query_device and ucontext paths Yishai Hadas
2026-11-17 17:09 ` [PATCH V2 rdma-next 21/21] RDMA/core: Report and enforce adversarial mode via GET_CONTEXT Yishai Hadas
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=20261117170914.27410-16-yishaih@nvidia.com \
--to=yishaih@nvidia.com \
--cc=andrew+netdev@lunn.ch \
--cc=davem@davemloft.net \
--cc=dtatulea@nvidia.com \
--cc=edumazet@kernel.org \
--cc=jgg@ziepe.ca \
--cc=kuba@kernel.org \
--cc=leon@kernel.org \
--cc=linux-rdma@vger.kernel.org \
--cc=maorg@nvidia.com \
--cc=mbloch@nvidia.com \
--cc=moshe@nvidia.com \
--cc=netdev@vger.kernel.org \
--cc=pabeni@redhat.com \
--cc=saeedm@nvidia.com \
--cc=tariqt@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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox