linux-kernel - Re: [PATCH] usb: musb: Force-disable pullup on shutdown

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, 03 Apr 2019 17:52:20 +0200
From:   Paul Cercueil <paul@...pouillou.net>
To:     Bin Liu <b-liu@...com>
Cc:     linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org, od@...c.me
Subject: Re: [PATCH] usb: musb: Force-disable pullup on shutdown



Le mer. 3 avril 2019 à 17:46, Bin Liu <b-liu@...com> a écrit :
> On Wed, Apr 03, 2019 at 05:31:31PM +0200, Paul Cercueil wrote:
>> 
>> 
>>  Le mer. 3 avril 2019 à 15:26, Bin Liu <b-liu@...com> a écrit :
>>  >On Tue, Apr 02, 2019 at 09:58:42PM +0200, Paul Cercueil wrote:
>>  >> Hi,
>>  >>
>>  >> Le lun. 1 avril 2019 à 20:20, Bin Liu <b-liu@...com> a écrit :
>>  >> >On Mon, Apr 01, 2019 at 07:46:22PM +0200, Paul Cercueil wrote:
>>  >> >>
>>  >> >>
>>  >> >> Le lun. 1 avril 2019 à 19:17, Bin Liu <b-liu@...com> a 
>> écrit :
>>  >> >> >On Thu, Mar 21, 2019 at 03:42:46PM +0100, Paul Cercueil 
>> wrote:
>>  >> >> >> When the musb is shutdown, for instance when the driver is
>>  >> >>unloaded,
>>  >> >> >> force-disable the pullup. Otherwise, the host will still 
>> see
>>  >> >> >>the gadget
>>  >> >> >> device even after the shutdown.
>>  >> >> >
>>  >> >> >how would this happen?
>>  >> >> >
>>  >> >> >when musb-hdrc driver is unloaded, udc core removes the bound
>>  >> >>gadget
>>  >> >> >driver which calls musb_gadget_pullup() to disable the 
>> pullup.
>>  >> >>
>>  >> >> I'm testing with the jz4740-musb driver. I don't unload the
>>  >> >>module (it's
>>  >> >> built-in) but unbind it from sysfs.
>>  >> >
>>  >> >I did unbind too.
>>  >> >
>>  >> >root@...35x-evm:/sys/bus/platform/drivers/musb-hdrc# echo
>>  >> >musb-hdrc.0 > unbind
>>  >> >
>>  >> >or unbind the glue driver:
>>  >> >
>>  >> >root@...35x-evm:/sys/bus/platform/drivers/musb-dsps# echo
>>  >> >47401400.usb > unbind
>>  >> >
>>  >> >musb_gadget_pullup() is called in both cases.
>>  >> >
>>  >> >[ 3880.597014] [<bf444ab0>] (musb_gadget_pullup [musb_hdrc]) 
>> from
>>  >> >[<bf402cbc>] (usb_gadget_disconnect+0x3c/0xf4 [udc_core])
>>  >> >[ 3880.607959] [<bf402cbc>] (usb_gadget_disconnect [udc_core])
>>  >> >from [<bf403b28>] (usb_gadget_remove_driver+0x4c/0x90 
>> [udc_core])
>>  >> >[ 3880.619338] [<bf403b28>] (usb_gadget_remove_driver 
>> [udc_core])
>>  >> >from [<bf403d20>] (usb_del_gadget_udc+0x5c/0xc0 [udc_core])
>>  >>
>>  >> In my case this stops here, usb_del_gadget_udc() does not call
>>  >> usb_gadget_remove_driver(), that's why the pullup is never 
>> disabled.
>>  >>
>>  >> I guess that's because udc->driver is NULL; I'm testing with
>>  >
>>  >then the pullup should be disable by now.
>>  >
>>  >> CONFIG_USB_CONFIGFS,
>>  >> and I don't configure anything in sysfs before unbinding the 
>> driver.
>>  >
>>  >I didn't check on this, but I could imagine that
>>  >- when a configfs gadget is bound to the udc, .pullup() is called;
>>  >- when the configfs gadget is unbound from the udc, .pullup should 
>> be
>>  >  called again to disable the pullup.
>> 
>>  An important thing that I did not mention, is that the SoC boots
>>  from USB,
>>  so the pullup is active before the musb driver loads. Since in my 
>> case a
> 
> It sounds to me that the musb driver should disable the pullup during
> init.  Isn't it?

Yes. I can move it to musb_gadget_setup(). Should I still protect it 
with
the spinlock then?

>>  configfs gadget is never bound, then .pullup() is never called, and 
>> when
>>  I unbind the driver the pullup is still enabled.
> 
> Regards,
> -Bin.