mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Vasant Hegde <vasant.hegde@amd.com>
To: lirongqing <lirongqing@baidu.com>, Joerg Roedel <joro@8bytes.org>,
	Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>,
	Will Deacon <will@kernel.org>,
	Robin Murphy <robin.murphy@arm.com>,
	iommu@lists.linux.dev, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] iommu/amd: RCU-protect iommu_ga_log_notifier
Date: Mon, 16 Nov 2026 20:18:56 +0530	[thread overview]
Message-ID: <301430f5-c96a-42e2-bd81-de564650f9a9@amd.com> (raw)
In-Reply-To: <20260924064709.2194-1-lirongqing@baidu.com>

Hi Li,


On 9/24/2026 12:17 PM, lirongqing wrote:
> From: Li RongQing <lirongqing@baidu.com>
> 
> iommu_poll_ga_log() accesses iommu_ga_log_notifier without an RCU
> read-side critical section, so synchronize_rcu() in the unregister
> path does not wait for an in-flight notifier callback.
> 
> Use rcu_assign_pointer() when updating the notifier and
> rcu_dereference() under scoped_guard(rcu) when invoking it. Load the
> pointer once to close the race with unregister.
> 
> Keep the RCU read-side critical section minimal: only the pointer load
> and the indirect notifier call stay inside scoped_guard(rcu). Move
> pr_debug and pr_err outside the section so that logging does not extend
> the window that synchronize_rcu() must wait for.

I didn't see your patch earlier and posted somewhat similar fix. But since you
posted earlier I will pick this. Can you please address below comments.



[1]
https://lore.kernel.org/linux-iommu/20261105041603.399016-4-vasant.hegde@amd.com/


> 
> Signed-off-by: Li RongQing <lirongqing@baidu.com>
> ---
>  drivers/iommu/amd/iommu.c | 30 +++++++++++++++++++-----------
>  1 file changed, 19 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/iommu/amd/iommu.c b/drivers/iommu/amd/iommu.c
> index a06cd61..a31464a 100644
> --- a/drivers/iommu/amd/iommu.c
> +++ b/drivers/iommu/amd/iommu.c
> @@ -1054,11 +1054,11 @@ static void iommu_poll_events(struct amd_iommu *iommu)
>  }
>  
>  #ifdef CONFIG_IRQ_REMAP
> -static int (*iommu_ga_log_notifier)(u32);
> +static int (__rcu *iommu_ga_log_notifier)(u32 ga_tag);
>  
>  int amd_iommu_register_ga_log_notifier(int (*notifier)(u32))
>  {
> -	iommu_ga_log_notifier = notifier;
> +	rcu_assign_pointer(iommu_ga_log_notifier, notifier);
>  
>  	/*
>  	 * Ensure all in-flight IRQ handlers run to completion before returning
> @@ -1094,17 +1094,25 @@ static void iommu_poll_ga_log(struct amd_iommu *iommu)
>  
>  		/* Handle GA entry */
>  		switch (GA_REQ_TYPE(log_entry)) {
> -		case GA_GUEST_NR:
> -			if (!iommu_ga_log_notifier)
> -				break;
> -
> -			pr_debug("%s: devid=%#x, ga_tag=%#x\n",
> -				 __func__, GA_DEVID(log_entry),
> -				 GA_TAG(log_entry));
> +		case GA_GUEST_NR: {
> +			int (*notifier)(u32 ga_tag);
> +			int ret = 0;
> +
> +			scoped_guard(rcu) {

Why we need scoped_guard()? we don't have multiple error condition to handle. I
think  simple rcu_read_lock() -> call notifier stuff -> rcu_read_unlock() is
good enough. That way we can avoid multiple if(notifier) check as well.

-Vasant


> +				notifier = rcu_dereference(iommu_ga_log_notifier);
> +				if (notifier)
> +					ret = notifier(GA_TAG(log_entry));
> +			}
>  
> -			if (iommu_ga_log_notifier(GA_TAG(log_entry)) != 0)
> -				pr_err("GA log notifier failed.\n");
> +			if (notifier) {
> +				pr_debug("%s: devid=%#x, ga_tag=%#x\n",
> +					 __func__, GA_DEVID(log_entry),
> +					 GA_TAG(log_entry));
> +				if (ret)
> +					pr_err("GA log notifier failed.\n");
> +			}
>  			break;
> +		}
>  		default:
>  			break;
>  		}


      parent reply	other threads:[~2026-11-16 14:49 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-24  6:47 lirongqing
2026-10-14  6:14 ` 答复: " Li,Rongqing
2026-11-16 14:48 ` Vasant Hegde [this message]

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=301430f5-c96a-42e2-bd81-de564650f9a9@amd.com \
    --to=vasant.hegde@amd.com \
    --cc=iommu@lists.linux.dev \
    --cc=joro@8bytes.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lirongqing@baidu.com \
    --cc=robin.murphy@arm.com \
    --cc=suravee.suthikulpanit@amd.com \
    --cc=will@kernel.org \
    /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

all inboxes | Powered by JetHome®