| 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
| ||
|
Date: Mon, 4 Apr 2016 10:50:17 +0300 From: Roger Quadros <rogerq@...com> To: Felipe Balbi <balbi@...nel.org>, John Youn <John.Youn@...opsys.com> CC: "nsekhar@...com" <nsekhar@...com>, "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>, "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>, "Shankar, Abhishek" <a.shankar@...com>, "B, Ravi" <ravibabu@...com> Subject: Re: [PATCH 1/2] usb: dwc3: core: Introduce dwc3_device_reinit() +Abhishek, Ravi, Felipe, On 31/03/16 17:26, Felipe Balbi wrote: > > Hi, > > Roger Quadros <rogerq@...com> writes: >>>>>>>>>>> We will need this function for a workaround. >>>>>>>>>>> The function issues a softreset only to the device >>>>>>>>>>> controller and performs minimal re-initialization >>>>>>>>>>> so that the device controller can be usable. >>>>>>>>>>> >>>>>>>>>>> As some code is similar to dwc3_core_init() take out >>>>>>>>>>> common code into dwc3_get_gctl_quirks(). >>>>>>>>>>> >>>>>>>>>>> We add a new member (prtcap_mode) to struct dwc3 to >>>>>>>>>>> keep track of the current mode in the PRTCAPDIR register. >>>>>>>>>>> >>>>>>>>>>> Signed-off-by: Roger Quadros <rogerq@...com> >>>>>>>>>> >>>>>>>>>> I must say, I don't like this at all :-p There's ONE known silicon which >>>>>>>>>> needs this because of a poor silicon integration which took an IP with a >>>>>>>>>> known erratum where it can't be made to work on lower speeds and STILL >>>>>>>>>> was integrated without a superspeed PHY. >>>>>>>>>> >>>>>>>>>> There's a reason why I never tried to push this upstream myself ;-) >>>>>>>>>> >>>>>>>>>> I'm really thinking we might be better off adding a quirk flag to skip >>>>>>>>>> the metastability workaround and allow this ONE silicon to set the >>>>>>>>>> controller to lower speed. >>>>>>>>>> >>>>>>>>>> John, can you check with your colleagues if we would ever fall into >>>>>>>>>> STAR#9000525659 if we set maximum speed to high speed during driver >>>>>>>>>> probe and never touch it again ? I would assume we don't really fall >>>>>>>>>> into the metastability workaround, right ? We're not doing any sort of >>>>>>>>>> PM for dwc3... >>>>>>>>>> >>>>>>> >>>>>>> Hi Felipe, >>>>>>> >>>>>>> Do you mean to keep DCFG.speed to SS and set dwc->maximum_speed to HS? >>>>>>> I don't see an issue with that as long as we always ignore >>>>>>> dwc->maximum_speed when programming DCFG.speed for all affected >>>>>>> versions of the core. As long as the DCFG.speed = SS, you should not >>>>>>> hit the STAR. >>>>>> >>>>>> I actually mean changing DCFG.speed during driver probe and never >>>>>> touching it again. Would that still cause problems ? >>>>>> >>>>> >>>>> In that case I'm not sure. The engineer who would know is off until >>>>> next week so I'll get back to you as soon as I can talk to him about >>>>> it. >>>> >>> >>> So the engineers said that this issue can occur while set to HS and >>> the run/stop bit is changed so it seems that won't work. >> >> Thanks John. >> >> Felipe, >> >> any suggestion how we can fix this upstream? > > no idea, I don't have a lot of memory about this problem. I really don't > remember the details about this, is there an openly available errata > document which I could read ? /me goes search for it. > > I found [1] which tells me, the following: > > > | i819 | A Device Control Bit Meta-Stability for USB3.0 Controller in USB2.0 Mode | > |-------------+----------------------------------------------------------------------------| > | Criticality | Medium | > | | | > | Descritiion | When USB3.0 controller core is programmed to be a USB 2.0-only device | > | | hardware meta-stability on USB_DCTL[31]RUNSTOP bit causing the core to | > | | attempt high speed as well as SuperSpeed connection or completely miss | > | | the attach request. | > | | | > | Workaround | If the requirement is to always function in USB 2.0 mode, there is no | > | | workaround. | > | | Otherwise, you can always program the USB controller core to be SuperSpeed | > | | 3.0 capable (USB_DCFG[2:0]DEVSPD = 0x4) | > | | | > | Revisions | SR 1.1, 2.0 | > | Impacted | | > |-------------+----------------------------------------------------------------------------| > > So, TI's own documentation says that there is _no_ workaround. My We are working on updating that document. Apparently Synopsis has suggested this workaround. pasting verbatim " - As last step of device configuration we set the RUNSTOP bit in DCTL. - Once we set the RUNSTOP bit, we need to monitor GDBGLTSSM for 100 ms until one of the two below happens:. o We see the GDBGLTSSM.LTDB_LINK_STATE changing from 4 o We receive the USB 2.0 reset interrupt. If none of above happens then we can stop monitoring it. - If state change from 4 occurs issue a SoftReset thru DCTL.CSftRst and reconfigure Device. This time it is guaranteed that no metastability will occur so no need to do the 100ms timeout. " > question is, then: How are you sure that resetting the device actually > solves the issue ? Did you really hit the metastability problem and > noted that it works after a soft-reset ? How did you verify that I don't know if it solves the issue or not. It was suggested by Synopsis to TI's silicon team. I never hit the metastability problem detection condition in my overnight tests (i.e. LTDB_LINK_STATE != 4). I have verified that things work after a soft-reset by faking that the error happens. > Run/Stop was in a metastable state, considering that Run/Stop signal is > not visible outside the die ? LTDB_LINK_STATE != 4 within 100ms or RUNSTOP set is the condition to detect that the issue occurred. > > It seems to me that resetting the IP is just as "dangerous" as setting > the IP to High-speed in the first place. No ? The soft-reset is just a recovery mechanism if that error is ever hit. Putting the controller in reset state means it is in a known state. Why do you say it would be dangerous? The original workaround i.e. setting the High-speed instance to Super-speed to avoid this errata is causing another side effect. i.e. erratic interrupts happen and more than 2 seconds delay to enumerations. This problem is more serious and so we have to keep the controller in High-speed and tackle the meta-stability condition if it happens. cheers, -roger > > [1] http://www.ti.com/lit/er/sprz429h/sprz429h.pdf >
Powered by blists - more mailing lists