linux-kernel - Re: [PATCH V3 10/18] coresight: tmc: getting the right read_count on tmc

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

Date:	Mon, 25 Apr 2016 11:47:48 +0100
From:	Suzuki K Poulose <Suzuki.Poulose@....com>
To:	Mathieu Poirier <mathieu.poirier@...aro.org>,
	linux-arm-kernel@...ts.infradead.org
Cc:	linux-kernel@...r.kernel.org
Subject: Re: [PATCH V3 10/18] coresight: tmc: getting the right read_count on
 tmc_open()

On 22/04/16 18:14, Mathieu Poirier wrote:
> In function tmc_open(), if tmc_read_prepare() fails variable
> drvdata->read_count is not decremented, causing unwanted
> access to drvdata->buf and very likely, a crash dump.
>
> By moving the incrementation to a place where we know things
> are stable this kind of situation is avoided.
>
> Signed-off-by: Mathieu Poirier <mathieu.poirier@...aro.org>
> Reviewed-by: Suzuki K Poulose <Suzuki.Poulose@....com>
> ---
>   drivers/hwtracing/coresight/coresight-tmc.c | 3 ++-
>   1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/hwtracing/coresight/coresight-tmc.c b/drivers/hwtracing/coresight/coresight-tmc.c
> index e8e12a9b917a..55806352b1f1 100644
> --- a/drivers/hwtracing/coresight/coresight-tmc.c
> +++ b/drivers/hwtracing/coresight/coresight-tmc.c
> @@ -121,13 +121,14 @@ static int tmc_open(struct inode *inode, struct file *file)
>   						   struct tmc_drvdata, miscdev);
>   	int ret = 0;
>

On a second thought, I think there could be a race here.


> -	if (drvdata->read_count++)
> +	if (drvdata->read_count)
>   		goto out;
>
>   	ret = tmc_read_prepare(drvdata);
>   	if (ret)
>   		return ret;
>   out:

What prevents someone else doing a release() on the file when we get here, without
incrementing the read_count ? Also, read_count accesses are not protected. Either should
be covered by the drvdata->spinlock or convert it to atomic.



> +	drvdata->read_count++;
>   	nonseekable_open(inode, file);


Cheers
Suzuki