linux-kernel - Re: [PATCH] usb: gadget: fix possible regression introduced with ep->claimed

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20150821150050.GB527@saruman.tx.rr.com>
Date:	Fri, 21 Aug 2015 10:00:50 -0500
From:	Felipe Balbi <balbi@...com>
To:	Robert Baldyga <r.baldyga@...sung.com>
CC:	<balbi@...com>, <stern@...land.harvard.edu>,
	<r.baldyga@...kerion.com>, <linux-usb@...r.kernel.org>,
	<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] usb: gadget: fix possible regression introduced with
 ep->claimed

On Fri, Aug 21, 2015 at 11:01:29AM +0200, Robert Baldyga wrote:
> This patch fixes possible regression introduced by patch reworking endpoint
> claiming mechanism. It restores setring ep->driver_data to NULL in
> usb_ep_autoconfig_reset(), which was removed by patch [1].
> 
> [1] commit cc476b42a39d ("usb: gadget: encapsulate endpoint claiming
>     mechanism")
> 
> Reported-by: Felipe Balbi <balbi@...com>
> Signed-off-by: Robert Baldyga <r.baldyga@...sung.com>
> ---
> 
> Hi Felipe,
> 
> I was not able to reproduce the problem that you have described (functions
> trying to enable eps that are already enabled, and disable eps that are
> already disabled), but I have analysed the problem and I have concluded

it happens with all, you can't see it because dwc3 (only) has a WARN()
for when an endpoint is enabled and has previously been enabled
(drivers/usb/dwc3/gadget.c). If you track your endpoint's state
regarding being enabled or not, you can add a WARN() to see the error.

> that the only possible cause of regression can be that my patch removed
> setting ep->driver_data to NULL in usb_ep_autoconfig_reset() function.
> Because of this change initial state of ep->driver_data could be non-NULL
> value, which could change behavoiur of some functions. I can't imagine
> any other possible cause of regression introduced by such simple patch.

yeah, I'll test this later.

> 
> Thanks,
> Robert
> 
>  drivers/usb/gadget/epautoconf.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/usb/gadget/epautoconf.c b/drivers/usb/gadget/epautoconf.c
> index 978435a..6399c10 100644
> --- a/drivers/usb/gadget/epautoconf.c
> +++ b/drivers/usb/gadget/epautoconf.c
> @@ -186,6 +186,7 @@ void usb_ep_autoconfig_reset (struct usb_gadget *gadget)
>  
>  	list_for_each_entry (ep, &gadget->ep_list, ep_list) {
>  		ep->claimed = false;
> +		ep->driver_data = NULL;
>  	}
>  	gadget->in_epnum = 0;
>  	gadget->out_epnum = 0;
> -- 
> 1.9.1
> 

-- 
balbi

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)