Re: [PATCH 0/1] uvc gadget sg performance issues

[Dan Vacura]

Hi Michael, thanks for the prompt reply!

On Mon, Sep 26, 2022 at 10:15:41PM +0200, Michael Grzeschik wrote:
> Hi Dan!
> 
> On Mon, Sep 26, 2022 at 02:53:06PM -0500, Dan Vacura wrote:
> > 
> > Hello uvc gadget developers,
> > 
> > I'm working on a 5.15.41 based kernel on a qcom chipset with the dwc3
> > controller and I'm encountering two problems related to the recent performance
> > improvement changes:
> 
> What's about that odd kernel number. UVC is under heavy development, if
> you plan to work with this code, you should probably test top of tree.

Yes, it's a bit behind and it looks like some of the initial work you
did for scatter/gather got pulled into the 5.15 tree, but subsequent
changes didn't. I don't have much control over the kernel versioning as
we're part of the GKI Android initiative:
https://source.android.com/docs/core/architecture/kernel/generic-kernel-image
and we can only work off of what is provided, like this release line:
https://android.googlesource.com/kernel/common/+log/refs/heads/android13-5.15

Perhaps we can revert these changes for the 5.15 kernel (and other
versions) they were not intended for?

> 
> > https://patchwork.kernel.org/project/linux-usb/patch/20210628155311.16762-5-m.grzeschik@xxxxxxxxxxxxxx/  and
> > https://patchwork.kernel.org/project/linux-usb/patch/20210628155311.16762-6-m.grzeschik@xxxxxxxxxxxxxx/
> > 
> > If I revert these two changes, then I have much improved stability and a
> > transmission problem I'm seeing is gone. Has there been any success from
> > others on 5.15 with this uvc improvement and any recommendations for my
> > current problems?  Those being:
> > 
> > 1) a smmu panic, snippet here: 
> > 
> >    <3>[  718.314900][  T803] arm-smmu 15000000.apps-smmu: Unhandled arm-smmu context fault from a600000.dwc3!
> >    <3>[  718.314994][  T803] arm-smmu 15000000.apps-smmu: FAR    = 0x00000000efe60800
> >    <3>[  718.315023][  T803] arm-smmu 15000000.apps-smmu: PAR    = 0x0000000000000000
> >    <3>[  718.315048][  T803] arm-smmu 15000000.apps-smmu: FSR    = 0x40000402 [TF R SS ]
> >    <3>[  718.315074][  T803] arm-smmu 15000000.apps-smmu: FSYNR0    = 0x5f0003
> >    <3>[  718.315096][  T803] arm-smmu 15000000.apps-smmu: FSYNR1    = 0xaa02
> >    <3>[  718.315117][  T803] arm-smmu 15000000.apps-smmu: context bank#    = 0x1b
> >    <3>[  718.315141][  T803] arm-smmu 15000000.apps-smmu: TTBR0  = 0x001b0000c2a92000
> >    <3>[  718.315165][  T803] arm-smmu 15000000.apps-smmu: TTBR1  = 0x001b000000000000
> >    <3>[  718.315192][  T803] arm-smmu 15000000.apps-smmu: SCTLR  = 0x0a5f00e7 ACTLR  = 0x00000003
> >    <3>[  718.315245][  T803] arm-smmu 15000000.apps-smmu: CBAR  = 0x0001f300
> >    <3>[  718.315274][  T803] arm-smmu 15000000.apps-smmu: MAIR0   = 0xf404ff44 MAIR1   = 0x0000efe4
> >    <3>[  718.315297][  T803] arm-smmu 15000000.apps-smmu: SID = 0x40
> >    <3>[  718.315318][  T803] arm-smmu 15000000.apps-smmu: Client info: BID=0x5, PID=0xa, MID=0x2
> >    <3>[  718.315377][  T803] arm-smmu 15000000.apps-smmu: soft iova-to-phys=0x0000000000000000
> > 
> >    I can reduce this panic with the proposed patch, but it still happens until I
> >    disable the "req->no_interrupt = 1" logic.
> > 
> > 2) The frame is not fully transmitted in dwc3 with sg support enabled.
> > 
> >    There seems to be a mapping limit I'm seeing where only the roughly first
> >    70% of the total frame is sent. Interestingly, if I allocate a larger
> >    size for the buffer upfront, in uvc_queue_setup(), like sizes[0] =
> >    video->imagesize * 3. Then the issue rarely happens. For example, when I
> >    do YUYV I see green, uninitialized data, at the bottom part of the
> >    frame. If I do MJPG with smaller filled sizes, the transmission is fine.
> > 
> >    +-------------------------+
> >    |                         |
> >    |                         |
> >    |                         |
> >    |      Good data          |
> >    |                         |
> >    |                         |
> >    |                         |
> >    +-------------------------+
> >    |xxxxxxxxxxxxxxxxxxxxxxxxx|
> >    |xxxx  Bad data  xxxxxxxxx|
> >    |xxxxxxxxxxxxxxxxxxxxxxxxx|
> >    +-------------------------+
> > 
> > 
> > Appreciate any thoughts or feedback related to these issues.
> 
> Anyway, this is probably due to the frames being given back to early to
> the frameproducer. We have the following patches mainline now to fix this issue:
> 
> aef11279888c00e1841a3533a35d279285af3a51 usb: gadget: uvc: improve sg exit condition
> 9b969f93bcef9b3d9e92f1810e22bbd6c344a0e5 usb: gadget: uvc: giveback vb2 buffer on req complete

Yes, I did grab those in addition to some other necessary changes, noted
in my patch here:
https://patchwork.kernel.org/project/linux-usb/patch/20220926195307.110121-2-w36195@xxxxxxxxxxxx/ 
I also pulled a lot of patches from the dwc3 to be almost at parity with
top of tree for the core.c/h, ep0.c/h, and gadget.c/h files, but these
issues persisted.

Out of curiosity, have you tested these changes with dwc3 and if so,
have you tried "usb: gadget: uvc: decrease the interrupt load to a
quarter" with scatter/gather disabled? For me the crash occurs more
often.

Thanks,

Dan

> 
> You might need some patches beforehand to get them applied on your
> stack.
> 
> Regards,
> Michael
> 
> -- 
> Pengutronix e.K.                           |                             |
> Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
> 31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |