linux-kernel - Re: [PATCH] usb: dwc3: Fix race between dwc3_set_mode and __dwc3_set

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]

Date:   Mon, 28 Nov 2022 19:14:16 +0000
From:   Thinh Nguyen <Thinh.Nguyen@...opsys.com>
To:     Sven Peter <sven@...npeter.dev>
CC:     Thinh Nguyen <Thinh.Nguyen@...opsys.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Felipe Balbi <felipe.balbi@...ux.intel.com>,
        Janne Grunau <j@...nau.net>,
        "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
        "stable@...r.kernel.org" <stable@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "asahi@...ts.linux.dev" <asahi@...ts.linux.dev>
Subject: Re: [PATCH] usb: dwc3: Fix race between dwc3_set_mode and
 __dwc3_set_mode

Hi,

On Mon, Nov 28, 2022, Sven Peter wrote:
> dwc->desired_dr_role is changed by dwc3_set_mode inside a spinlock but
> then read by __dwc3_set_mode outside of that lock. This can lead to a
> race condition when very quick successive role switch events happen:
> 
> CPU A
> 	dwc3_set_mode(DWC3_GCTL_PRTCAP_HOST) // first role switch event
> 		spin_lock_irqsave(&dwc->lock, flags);
> 		dwc->desired_dr_role = mode; // DWC3_GCTL_PRTCAP_HOST
> 		spin_unlock_irqrestore(&dwc->lock, flags);
> 		queue_work(system_freezable_wq, &dwc->drd_work);
> 
> CPU B
> 	__dwc3_set_mode
> 		// ....
> 		spin_lock_irqsave(&dwc->lock, flags);
> 		// desired_dr_role is DWC3_GCTL_PRTCAP_HOST
> 		dwc3_set_prtcap(dwc, dwc->desired_dr_role);
> 		spin_unlock_irqrestore(&dwc->lock, flags);
> 
> CPU A
> 	dwc3_set_mode(DWC3_GCTL_PRTCAP_DEVICE) // second event
> 		spin_lock_irqsave(&dwc->lock, flags);
> 		dwc->desired_dr_role = mode; // DWC3_GCTL_PRTCAP_DEVICE
> 		spin_unlock_irqrestore(&dwc->lock, flags);
> 
> CPU B (continues running __dwc3_set_mode)
> 	switch (dwc->desired_dr_role) { // DWC3_GCTL_PRTCAP_DEVICE
> 	// ....
> 	case DWC3_GCTL_PRTCAP_DEVICE:
> 		// ....
> 		ret = dwc3_gadget_init(dwc);
> 
> We then have DWC3_GCTL.DWC3_GCTL_PRTCAPDIR = DWC3_GCTL_PRTCAP_HOST and
> dwc->current_dr_role = DWC3_GCTL_PRTCAP_HOST but initialized the
> controller in device mode. It's also possible to get into a state
> where both host and device are intialized at the same time.
> Fix this race by creating a local copy of desired_dr_role inside
> __dwc3_set_mode while holding dwc->lock.
> 
> Fixes: 41ce1456e1db ("usb: dwc3: core: make dwc3_set_mode() work properly")
> Signed-off-by: Sven Peter <sven@...npeter.dev>
> ---
>  drivers/usb/dwc3/core.c | 16 ++++++++++------
>  1 file changed, 10 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/usb/dwc3/core.c b/drivers/usb/dwc3/core.c
> index 1f348bc867c2..fc38a8b13efa 100644
> --- a/drivers/usb/dwc3/core.c
> +++ b/drivers/usb/dwc3/core.c
> @@ -122,21 +122,25 @@ static void __dwc3_set_mode(struct work_struct *work)
>  	unsigned long flags;
>  	int ret;
>  	u32 reg;
> +	u32 desired_dr_role;
>  
>  	mutex_lock(&dwc->mutex);
> +	spin_lock_irqsave(&dwc->lock, flags);
> +	desired_dr_role = dwc->desired_dr_role;
> +	spin_unlock_irqrestore(&dwc->lock, flags);
>  
>  	pm_runtime_get_sync(dwc->dev);
>  
>  	if (dwc->current_dr_role == DWC3_GCTL_PRTCAP_OTG)
>  		dwc3_otg_update(dwc, 0);
>  
> -	if (!dwc->desired_dr_role)
> +	if (!desired_dr_role)
>  		goto out;
>  
> -	if (dwc->desired_dr_role == dwc->current_dr_role)
> +	if (desired_dr_role == dwc->current_dr_role)
>  		goto out;
>  
> -	if (dwc->desired_dr_role == DWC3_GCTL_PRTCAP_OTG && dwc->edev)
> +	if (desired_dr_role == DWC3_GCTL_PRTCAP_OTG && dwc->edev)
>  		goto out;
>  
>  	switch (dwc->current_dr_role) {
> @@ -164,7 +168,7 @@ static void __dwc3_set_mode(struct work_struct *work)
>  	 */
>  	if (dwc->current_dr_role && ((DWC3_IP_IS(DWC3) ||
>  			DWC3_VER_IS_PRIOR(DWC31, 190A)) &&
> -			dwc->desired_dr_role != DWC3_GCTL_PRTCAP_OTG)) {
> +			desired_dr_role != DWC3_GCTL_PRTCAP_OTG)) {
>  		reg = dwc3_readl(dwc->regs, DWC3_GCTL);
>  		reg |= DWC3_GCTL_CORESOFTRESET;
>  		dwc3_writel(dwc->regs, DWC3_GCTL, reg);
> @@ -184,11 +188,11 @@ static void __dwc3_set_mode(struct work_struct *work)
>  
>  	spin_lock_irqsave(&dwc->lock, flags);
>  
> -	dwc3_set_prtcap(dwc, dwc->desired_dr_role);
> +	dwc3_set_prtcap(dwc, desired_dr_role);
>  
>  	spin_unlock_irqrestore(&dwc->lock, flags);
>  
> -	switch (dwc->desired_dr_role) {
> +	switch (desired_dr_role) {
>  	case DWC3_GCTL_PRTCAP_HOST:
>  		ret = dwc3_host_init(dwc);
>  		if (ret) {
> -- 
> 2.25.1
> 

This looks good to me. This should also go on stable.

Acked-by: Thinh Nguyen <Thinh.Nguyen@...opsys.com>

Thanks!
Thinh