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:   Wed, 28 Jun 2023 13:20:34 +0200
From:   Johan Hovold <johan@...nel.org>
To:     Maximilian Luz <luzmaximilian@...il.com>
Cc:     Bjorn Andersson <andersson@...nel.org>,
        Andy Gross <agross@...nel.org>,
        Konrad Dybcio <konrad.dybcio@...aro.org>,
        Ard Biesheuvel <ardb@...nel.org>,
        Ilias Apalodimas <ilias.apalodimas@...aro.org>,
        Srinivas Kandagatla <srinivas.kandagatla@...aro.org>,
        Sudeep Holla <sudeep.holla@....com>,
        Steev Klimaszewski <steev@...i.org>,
        linux-arm-msm@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v4 2/4] firmware: qcom_scm: Clear scm pointer on probe
 failure

On Mon, May 29, 2023 at 01:03:49AM +0200, Maximilian Luz wrote:
> When setting-up the IRQ goes wrong, the __scm pointer currently remains
> set even though we fail to probe the driver successfully. Due to this,
> access to __scm may go wrong since associated resources (clocks, ...)
> have been released. Therefore, clear the __scm pointer when setting-up
> the IRQ fails.
> 
> Fixes: 6bf325992236 ("firmware: qcom: scm: Add wait-queue handling logic")
> Signed-off-by: Maximilian Luz <luzmaximilian@...il.com>
> ---
> 
> Patch introduced in v4
> 
> ---
>  drivers/firmware/qcom_scm.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c
> index fde33acd46b7..d0070b833889 100644
> --- a/drivers/firmware/qcom_scm.c
> +++ b/drivers/firmware/qcom_scm.c
> @@ -1488,8 +1488,10 @@ static int qcom_scm_probe(struct platform_device *pdev)
>  	} else {
>  		ret = devm_request_threaded_irq(__scm->dev, irq, NULL, qcom_scm_irq_handler,
>  						IRQF_ONESHOT, "qcom-scm", __scm);
> -		if (ret < 0)
> +		if (ret < 0) {
> +			__scm = NULL;

This looks fragile at best. Clients use qcom_scm_is_available() to see
if __scm is available and do not expect it to go away once it is live.

It looks like you can hold off on initialising __scm until you've
requested the interrupt, either by using IRQ_NOAUTOEN or fixing
qcom_scm_waitq_wakeup() so that it doesn't use __scm directly.

That would also take care of the previous branch which may also leave
__scm set after the structure itself has been released on errors.

You'll have similar problems when registering qseecom which currently
depend on __scm being set, though. Clearing the pointer in that case is
clearly broken as you currently rely on devres for deregistering the aux
clients on errors (i.e. the clients using __scm are still registered
when you clear the pointer in patch 3/4).

>  			return dev_err_probe(scm->dev, ret, "Failed to request qcom-scm irq\n");
> +		}
>  	}
>  
>  	__get_convention();

Johan

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ