linux-kernel - Re: [PATCH v3] remoteproc: stm32: fix probe error case

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20191115063228.GS3108315@builder>
Date:   Thu, 14 Nov 2019 22:32:28 -0800
From:   Bjorn Andersson <bjorn.andersson@...aro.org>
To:     Fabien Dessenne <fabien.dessenne@...com>
Cc:     Maxime Coquelin <mcoquelin.stm32@...il.com>,
        Alexandre Torgue <alexandre.torgue@...com>,
        Ohad Ben-Cohen <ohad@...ery.com>,
        linux-stm32@...md-mailman.stormreply.com,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        linux-remoteproc@...r.kernel.org,
        Loic Pallardy <loic.pallardy@...com>,
        Arnaud Pouliquen <arnaud.pouliquen@...com>
Subject: Re: [PATCH v3] remoteproc: stm32: fix probe error case

On Wed 13 Nov 00:52 PST 2019, Fabien Dessenne wrote:

> If the rproc driver is probed before the mailbox driver and if the rproc
> Device Tree node has some mailbox properties, the rproc driver probe
> shall be deferred instead of being probed without mailbox support.
> 
> Signed-off-by: Fabien Dessenne <fabien.dessenne@...com>
> ---
> Changes since v2: free other requested mailboxes after one request fails
> Changes since v1: test IS_ERR() before checking PTR_ERR()
> ---
>  drivers/remoteproc/stm32_rproc.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/remoteproc/stm32_rproc.c b/drivers/remoteproc/stm32_rproc.c
> index 2cf4b29..4b67480 100644
> --- a/drivers/remoteproc/stm32_rproc.c
> +++ b/drivers/remoteproc/stm32_rproc.c
> @@ -310,7 +310,7 @@ static const struct stm32_mbox stm32_rproc_mbox[MBOX_NB_MBX] = {
>  	}
>  };
>  
> -static void stm32_rproc_request_mbox(struct rproc *rproc)
> +static int stm32_rproc_request_mbox(struct rproc *rproc)
>  {
>  	struct stm32_rproc *ddata = rproc->priv;
>  	struct device *dev = &rproc->dev;
> @@ -329,10 +329,16 @@ static void stm32_rproc_request_mbox(struct rproc *rproc)
>  
>  		ddata->mb[i].chan = mbox_request_channel_byname(cl, name);
>  		if (IS_ERR(ddata->mb[i].chan)) {
> +			if (PTR_ERR(ddata->mb[i].chan) == -EPROBE_DEFER) {
> +				ddata->mb[i].chan = NULL;

So this relies on the caller jumping to stm32_rproc_free_mbox() to
release a subset of ddata->mb[x].chan. While this works I find it error
prone and would prefer the idiomatic solution of cleaning things up, in
this function, before returning.

So, could you please goto a snippet that loops backwards from i-- to 0
calling mbox_free_channel() and then return -EPROBE_DEFER instead?

Thanks,
Bjorn

> +				return -EPROBE_DEFER;
> +			}
>  			dev_warn(dev, "cannot get %s mbox\n", name);
>  			ddata->mb[i].chan = NULL;
>  		}
>  	}
> +
> +	return 0;
>  }
>  
>  static int stm32_rproc_set_hold_boot(struct rproc *rproc, bool hold)
> @@ -596,7 +602,9 @@ static int stm32_rproc_probe(struct platform_device *pdev)
>  	if (ret)
>  		goto free_rproc;
>  
> -	stm32_rproc_request_mbox(rproc);
> +	ret = stm32_rproc_request_mbox(rproc);
> +	if (ret)
> +		goto free_mb;
>  
>  	ret = rproc_add(rproc);
>  	if (ret)
> -- 
> 2.7.4
>