Re: [PATCH] nvme: introduce panic_on_double_cqe param

[Chaitanya Kulkarni <chaitanyak@nvidia.com>]

On 10/22/25 6:54 AM, Guixin Liu wrote:
> Add a new debug switch to control whether to trigger a kernel crash
> when duplicate CQEs are detected, in order to preserve the kernel
> context, such as sq, cq, and so on, for subsequent debugging and
> analysis.
>
> Signed-off-by: Guixin Liu <kanie@linux.alibaba.com>
> ---
>   drivers/nvme/host/core.c | 5 +++++
>   drivers/nvme/host/nvme.h | 3 +++
>   2 files changed, 8 insertions(+)
>
> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
> index fa4181d7de73..7a3f9129a39c 100644
> --- a/drivers/nvme/host/core.c
> +++ b/drivers/nvme/host/core.c
> @@ -95,6 +95,11 @@ module_param(apst_secondary_latency_tol_us, ulong, 0644);
>   MODULE_PARM_DESC(apst_secondary_latency_tol_us,
>   	"secondary APST latency tolerance in us");
>   
> +bool panic_on_double_cqe;
> +EXPORT_SYMBOL_GPL(panic_on_double_cqe);
> +module_param(panic_on_double_cqe, bool, 0644);
> +MODULE_PARM_DESC(panic_on_double_cqe, "crash the kernel to save the scene");
> +
>   /*
>    * Older kernels didn't enable protection information if it was at an offset.
>    * Newer kernels do, so it breaks reads on the upgrade if such formats were
> diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
> index 102fae6a231c..24010d5d15ce 100644
> --- a/drivers/nvme/host/nvme.h
> +++ b/drivers/nvme/host/nvme.h
> @@ -595,6 +595,8 @@ static inline u16 nvme_cid(struct request *rq)
>   	return nvme_cid_install_genctr(nvme_req(rq)->genctr) | rq->tag;
>   }
>   
> +extern bool panic_on_double_cqe;
> +
>   static inline struct request *nvme_find_rq(struct blk_mq_tags *tags,
>   		u16 command_id)
>   {
> @@ -612,6 +614,7 @@ static inline struct request *nvme_find_rq(struct blk_mq_tags *tags,
>   		dev_err(nvme_req(rq)->ctrl->device,
>   			"request %#x genctr mismatch (got %#x expected %#x)\n",
>   			tag, genctr, nvme_genctr_mask(nvme_req(rq)->genctr));
> +		BUG_ON(panic_on_double_cqe);
>   		return NULL;
>   	}
>   	return rq;


I'm really not sure this is a good idea, I'll leave to others.


-ck