mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "Christophe Leroy (CS GROUP)" <chleroy@kernel.org>
To: jaidevshastri@vt.edu
Cc: linuxppc-dev@lists.ozlabs.org,
	linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH] soc: fsl: qbman: order the probed flags with release/acquire
Date: Tue, 22 Sep 2026 12:28:12 +0200	[thread overview]
Message-ID: <0a35cf00-32ba-4a21-b05c-edd44d3ca219@kernel.org> (raw)
In-Reply-To: <20260921-mb-qbman-v1-1-35f3321aa330@vt.edu>

Hi,

Le 22/09/2026 à 02:47, Jaidev Shastri via B4 Relay a écrit :
> [Vous ne recevez pas souvent de courriers de devnull+jaidevshastri.vt.edu@kernel.org. Découvrez pourquoi ceci est important à https://aka.ms/LearnAboutSenderIdentification ]
> 
> From: Jaidev Shastri <jaidevshastri@vt.edu>
> 
> dpaa_eth, caam and the portal drivers poll qman_is_probed(),
> bman_is_probed(), qman_portals_probed() and bman_portals_probed() before
> they touch the state the probe functions set up: qman_ip_rev, the CCSR
> and portal register maps, the FQD and PFDR bases, affine_portals and the
> pool allocators.
> 
> Each flag is set with a plain store at the end of its probe function and
> read with a plain load in the exported accessor. The stores that build
> the state are not ordered before the store to the flag, and the
> consumer's load of the flag is not ordered before its loads of the
> state. A consumer probing on another CPU can see the flag set and then
> read state that is still stale or zeroed, and program the hardware with
> it.
> 
> Set the flags with smp_store_release() and read them with
> smp_load_acquire(). The error paths storing -1 keep their plain stores:
> a consumer that sees -1 fails its own probe without touching the state.
> 
> Found with MBCheck, a static herd7-based memory consistency checker.

Sashiko made comments, can you have a look ?

https://sashiko.dev/#/patchset/20260921-mb-qbman-v1-1-35f3321aa330@vt.edu

Christophe

> 
> Signed-off-by: Jaidev Shastri <jaidevshastri@vt.edu>
> ---
>   drivers/soc/fsl/qbman/bman_ccsr.c   |  6 ++++--
>   drivers/soc/fsl/qbman/bman_portal.c |  6 ++++--
>   drivers/soc/fsl/qbman/qman_ccsr.c   | 11 +++++++++--
>   drivers/soc/fsl/qbman/qman_portal.c |  6 ++++--
>   4 files changed, 21 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/soc/fsl/qbman/bman_ccsr.c b/drivers/soc/fsl/qbman/bman_ccsr.c
> index b0f26f6f7..c5be35aba 100644
> --- a/drivers/soc/fsl/qbman/bman_ccsr.c
> +++ b/drivers/soc/fsl/qbman/bman_ccsr.c
> @@ -180,7 +180,8 @@ static irqreturn_t bman_isr(int irq, void *ptr)
> 
>   int bman_is_probed(void)
>   {
> -       return __bman_probed;
> +       /* Pairs with smp_store_release() in fsl_bman_probe(). */
> +       return smp_load_acquire(&__bman_probed);
>   }
>   EXPORT_SYMBOL_GPL(bman_is_probed);
> 
> @@ -279,7 +280,8 @@ static int fsl_bman_probe(struct platform_device *pdev)
>                  return ret;
>          }
> 
> -       __bman_probed = 1;
> +       /* Order bm_ccsr_start, bman_ip_rev and the pool allocator before the flag. */
> +       smp_store_release(&__bman_probed, 1);
> 
>          return 0;
>   };
> diff --git a/drivers/soc/fsl/qbman/bman_portal.c b/drivers/soc/fsl/qbman/bman_portal.c
> index 4d7b9caee..056fe3fbb 100644
> --- a/drivers/soc/fsl/qbman/bman_portal.c
> +++ b/drivers/soc/fsl/qbman/bman_portal.c
> @@ -90,7 +90,8 @@ static int bman_online_cpu(unsigned int cpu)
> 
>   int bman_portals_probed(void)
>   {
> -       return __bman_portals_probed;
> +       /* Pairs with smp_store_release() in bman_portal_probe(). */
> +       return smp_load_acquire(&__bman_portals_probed);
>   }
>   EXPORT_SYMBOL_GPL(bman_portals_probed);
> 
> @@ -157,7 +158,8 @@ static int bman_portal_probe(struct platform_device *pdev)
>          spin_lock(&bman_lock);
>          cpu = cpumask_first_zero(&portal_cpus);
>          if (cpu >= nr_cpu_ids) {
> -               __bman_portals_probed = 1;
> +               /* All CPU-bound portals are initialised and in affine_bportals. */
> +               smp_store_release(&__bman_portals_probed, 1);
>                  /* unassigned portal, skip init */
>                  spin_unlock(&bman_lock);
>                  goto check_cleanup;
> diff --git a/drivers/soc/fsl/qbman/qman_ccsr.c b/drivers/soc/fsl/qbman/qman_ccsr.c
> index aa5348f49..b46cfb964 100644
> --- a/drivers/soc/fsl/qbman/qman_ccsr.c
> +++ b/drivers/soc/fsl/qbman/qman_ccsr.c
> @@ -711,7 +711,8 @@ static int qman_resource_init(struct device *dev)
> 
>   int qman_is_probed(void)
>   {
> -       return __qman_probed;
> +       /* Pairs with smp_store_release() in fsl_qman_probe(). */
> +       return smp_load_acquire(&__qman_probed);
>   }
>   EXPORT_SYMBOL_GPL(qman_is_probed);
> 
> @@ -864,7 +865,13 @@ static int fsl_qman_probe(struct platform_device *pdev)
>          if (ret)
>                  return ret;
> 
> -       __qman_probed = 1;
> +       /*
> +        * Publish the flag only after every store made above (qman_ip_rev,
> +        * qm_ccsr_start, the FQD/PFDR bases, the work queue) is visible to
> +        * the consumers that poll qman_is_probed() and then call into
> +        * qman_set_sdest(), qman_liodn_fixup(), qman_alloc_*().
> +        */
> +       smp_store_release(&__qman_probed, 1);
> 
>          return 0;
>   }
> diff --git a/drivers/soc/fsl/qbman/qman_portal.c b/drivers/soc/fsl/qbman/qman_portal.c
> index 456ef5d5c..181c0f373 100644
> --- a/drivers/soc/fsl/qbman/qman_portal.c
> +++ b/drivers/soc/fsl/qbman/qman_portal.c
> @@ -175,7 +175,8 @@ static int qman_online_cpu(unsigned int cpu)
> 
>   int qman_portals_probed(void)
>   {
> -       return __qman_portals_probed;
> +       /* Pairs with smp_store_release() in qman_portal_probe(). */
> +       return smp_load_acquire(&__qman_portals_probed);
>   }
>   EXPORT_SYMBOL_GPL(qman_portals_probed);
> 
> @@ -251,7 +252,8 @@ static int qman_portal_probe(struct platform_device *pdev)
>          spin_lock(&qman_lock);
>          cpu = cpumask_first_zero(&portal_cpus);
>          if (cpu >= nr_cpu_ids) {
> -               __qman_portals_probed = 1;
> +               /* All CPU-bound portals are initialised and in affine_portals. */
> +               smp_store_release(&__qman_portals_probed, 1);
>                  /* unassigned portal, skip init */
>                  spin_unlock(&qman_lock);
>                  goto check_cleanup;
> 
> ---
> base-commit: 93f51579e7df248780214094418f205253383cc5
> change-id: 20260921-mb-qbman-84f56ef0f84c
> 
> Best regards,
> --
> Jaidev Shastri <jaidevshastri@vt.edu>
> 
> 


      reply	other threads:[~2026-09-22 10:28 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-09-22  0:47 Jaidev Shastri via B4 Relay
2026-09-22 10:28 ` Christophe Leroy (CS GROUP) [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=0a35cf00-32ba-4a21-b05c-edd44d3ca219@kernel.org \
    --to=chleroy@kernel.org \
    --cc=jaidevshastri@vt.edu \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linuxppc-dev@lists.ozlabs.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®