Please apply the following commit:
Commit subject: usb: dwc3: gadget: Don't clear flags before transfer ended
Commit ID: a114c4ca64bd522aec1790c7e5c60c882f699d8f
Apply to: at least 4.19 stable, and 5.4 stable if possible. Note that all kernels from v4.18-rc1 up to 5.7-rc1 are affected.
Why apply it:
<https://github.com/torvalds/linux/commit/a114c4ca64bd522aec1790c7e5c60c882f…> fixes <https://github.com/torvalds/linux/commit/6d8a019614f3a7630e0a2c1be4bf1cfc23…>. Without this fix the built-in USB function source/sink test module fails to work with isochronous endpoints [1]. A side-effect of setting dep->flags = DWC3_EP_ENABLED; in dwc3_gadget_ep_cleanup_completed_requests() as part of disabling an ep is that a subsequent attempt to enable the endpoint will skip __dwc3_gadget_ep_enable() effectively leaving the ep disabled.
[1] Our gadget driver on TI AM5729 fails to work exactly like the built-in USB function source/sink test module when switching to alternate interface 1 because of the issue described above.
TI is currently using 4.19 and 5.4 stable kernels as the basis for their processor SDK kernels for their AM57x SoC and may switch to 5.4 stable at a later time. Thank you.
Kind Regards
—
Delio Brignoli
AudioScience, Inc.
USA Sales Toll Free 1-855-AUDIOSC
<www.audioscience.com> - <http://www.facebook.com/AudioScienceInc>
Dear Prike, dear Alex, dear Linux folks,
Am 13.04.20 um 10:44 schrieb Paul Menzel:
> A regression between causes a system with the AMD board MSI B350M MORTAR
> (MS-7A37) with an AMD Ryzen 3 2200G not to power off any more but just
> to halt.
>
> The regression is introduced in 9ebe5422ad6c..b032227c6293. I am in the
> process to bisect this, but maybe somebody already has an idea.
I found the Easter egg:
> commit 487eca11a321ef33bcf4ca5adb3c0c4954db1b58
> Author: Prike Liang <Prike.Liang(a)amd.com>
> Date: Tue Apr 7 20:21:26 2020 +0800
>
> drm/amdgpu: fix gfx hang during suspend with video playback (v2)
>
> The system will be hang up during S3 suspend because of SMU is pending
> for GC not respose the register CP_HQD_ACTIVE access request.This issue
> root cause of accessing the GC register under enter GFX CGGPG and can
> be fixed by disable GFX CGPG before perform suspend.
>
> v2: Use disable the GFX CGPG instead of RLC safe mode guard.
>
> Signed-off-by: Prike Liang <Prike.Liang(a)amd.com>
> Tested-by: Mengbing Wang <Mengbing.Wang(a)amd.com>
> Reviewed-by: Huang Rui <ray.huang(a)amd.com>
> Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com>
> Cc: stable(a)vger.kernel.org
It reverts cleanly on top of 5.7-rc1, and this fixes the issue.
Greg, please do not apply this to the stable series. The commit message
doesn’t even reference a issue/bug report, and doesn’t give a detailed
problem description. What system is it?
Dave, Alex, how to proceed? Revert? I created issue 1094 [1].
Kind regards,
Paul
[1]: https://gitlab.freedesktop.org/drm/amd/-/issues/1094