[PATCH v2 3/6] configs: xilinx_versal_virt: disable USB_DWC3

Mon Nov 25 19:44:15 CET 2024

On 11/25/24 19:08, Tom Rini wrote:
> On Mon, Nov 25, 2024 at 07:03:41PM +0100, Michal Simek wrote:
>>
>>
>> On 11/25/24 11:35, Heinrich Schuchardt wrote:
>>> On 25.11.24 11:02, Michal Simek wrote:
>>>>
>>>>
>>>> On 11/25/24 10:21, Heinrich Schuchardt wrote:
>>>>> On 11/25/24 09:32, Michal Simek wrote:
>>>>>>
>>>>>>
>>>>>> On 11/23/24 22:45, Heinrich Schuchardt wrote:
>>>>>>> The CI uses the following command to launch xilinx_versal_virt_defconfig:
>>>>>>>
>>>>>>>       qemu-system-aarch64 -M xlnx-versal-virt \
>>>>>>>       -display none -m 4G -serial mon:stdio \
>>>>>>>       -device loader,file=u-boot,cpu-num=0
>>>>>>>
>>>>>>> 'usb start' or invoking eth_bootdev_hunt leads to a crash when function
>>>>>>> dwc3_core_init() tries to access a register at offset 0xc704 (DWC3_DCTL)
>>>>>>> relative to the register start address 0xfe20c100.
>>>>>>>
>>>>>>> Disable CONFIG_USB_DWC3 until the driver problem is fixed.
>>>>>>>
>>>>>>> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt at canonical.com>
>>>>>>> ---
>>>>>>> v2:
>>>>>>>      new patch
>>>>>>> ---
>>>>>>>    configs/xilinx_versal_virt_defconfig | 2 --
>>>>>>>    1 file changed, 2 deletions(-)
>>>>>>>
>>>>>>> diff --git a/configs/xilinx_versal_virt_defconfig
>>>>>>> b/configs/ xilinx_versal_virt_defconfig
>>>>>>> index c8f166c1221..06a240173ba 100644
>>>>>>> --- a/configs/xilinx_versal_virt_defconfig
>>>>>>> +++ b/configs/xilinx_versal_virt_defconfig
>>>>>>> @@ -153,8 +153,6 @@ CONFIG_USB=y
>>>>>>>    CONFIG_DM_USB_GADGET=y
>>>>>>>    CONFIG_USB_XHCI_HCD=y
>>>>>>>    CONFIG_USB_XHCI_DWC3=y
>>>>>>> -CONFIG_USB_DWC3=y
>>>>>>> -CONFIG_USB_DWC3_GENERIC=y
>>>>>>>    CONFIG_USB_ULPI_VIEWPORT=y
>>>>>>>    CONFIG_USB_ULPI=y
>>>>>>>    CONFIG_USB_GADGET=y
>>>>>>
>>>>>> NACK. That's not the way to go. If one test is failing in CI
>>>>>> disable that test not really support which is for HW too.
>>>>>>
>>>>>> Thanks,
>>>>>> Michal
>>>>>
>>>>> Hello Michal,
>>>>>
>>>>> It is not that a single test fails. USB is completely unusable
>>>>> on the board. 'usb start' leads to an immediate crash.
>>>>
>>>> I am not applying any patches on SOM with HEAD and usb is partially working.
>>>
>>> Don't you see a crash when executing 'usb start'?
>>> What QEMU are you using?
>>>
>>>> I can agree that it is not super stable with upstream only patches
>>>> but that's different story.
>>>>
>>>>>
>>>>> The problem already existed in v2021.01.
>>>>
>>>> Any log?
>>>
>>> $ git reset --hard origin/master
>>> $ make xilinx_versal_virt_defconfig
>>> $ CROSS_COMPILE=aarch64-linux-gnu- make -j$(nproc)
>>> $ qemu-system-riscv64 --version
>>> QEMU emulator version 9.0.2 (Debian 1:9.0.2+ds-4ubuntu7)
>>> $ qemu-system-aarch64 -M xlnx-versal-virt \
>>>   > -display none -m 4G -serial mon:stdio \
>>>   > -device loader,file=u-boot,cpu-num=0
>>>
>>>
>>> U-Boot 2025.01-rc2-00172-g880fcc49eb40 (Nov 25 2024 - 11:18:35 +0100)
>>>
>>> CPU:   UNKNOWN
>>> Model: Xilinx Versal Virtual development board
>>> DRAM:  2 GiB (effective 4 GiB)
>>> EL Level:       EL3
>>> Core:  32 devices, 13 uclasses, devicetree: board
>>> Warning: Device tree includes old 'u-boot,dm-' tags: please fix by 2023.07!
>>> MMC:   sdhci at f1040000: 0, sdhci at f1050000: 1
>>> Loading Environment from nowhere... OK
>>> In:    uart at ff000000
>>> Out:   uart at ff000000
>>> Err:   uart at ff000000
>>> Bootmode: JTAG_MODE
>>> Net:
>>> ZYNQ GEM: ff0c0000, mdio bus ff0c0000, phyaddr 0, interface rgmii-id
>>>
>>> Warning: ethernet at ff0c0000 (eth0) using random MAC address - 5e:f6:37:a3:68:24
>>> eth0: ethernet at ff0c0000
>>> ZYNQ GEM: ff0d0000, mdio bus ff0d0000, phyaddr 0, interface rgmii-id
>>>
>>> Warning: ethernet at ff0d0000 (eth1) using random MAC address - 5e:8a:ec:a9:d7:e8
>>> , eth1: ethernet at ff0d0000
>>> MMC: no card present
>>> MMC: no card present
>>> Cannot persist EFI variables without system partition
>>> Missing TPMv2 device for EFI_TCG_PROTOCOL
>>> Missing RNG device for EFI_RNG_PROTOCOL
>>> Hit any key to stop autoboot:  0
>>> Versal> usb start
>>> starting USB...
>>> Bus dwc3 at fe200000: "Synchronous Abort" handler, esr 0x96000050
>>> elr: 0000000008068fd8 lr : 0000000008069940 (reloc)
>>> elr: 000000007ff17fd8 lr : 000000007ff18940
>>> x0 : 00000000fe20c100 x1 : 0000000040000000
>>> x2 : 0000000033310000 x3 : 000000000000003f
>>> x4 : 000000007ffa74b0 x5 : 000000007c112880
>>> x6 : 000000007ffa74c0 x7 : 000000007c300000
>>> x8 : 0000000000000684 x9 : 000000007bda668c
>>> x10: 0000000000000003 x11: 0000000000000188
>>> x12: 0000000000000000 x13: 000000007bda6be0
>>> x14: 0000000000000000 x15: 000000007bda6224
>>> x16: 000000007fee9244 x17: 0000000000000000
>>> x18: 000000007bea6be0 x19: 000000007c111b48
>>> x20: 000000007c111240 x21: 0000000047822004
>>> x22: 000000007ff998b2 x23: 000000007ff998e8
>>> x24: 000000007ffd4514 x25: 0000000000000000
>>> x26: 0000000000000000 x27: 000000007c111130
>>> x28: 000000007c1110d0 x29: 000000007bda67b0
>>>
>>> Code: b9030660 f9416e60 d5033fbf 52a80001 (b9060401)
>>> Resetting CPU ...
>>>
>>> resetting ...
>>>
>>> $ cat u-boot.map
>>>    .text.dwc3_core_init
>>>                   0x0000000008068f24      0x6a0 drivers/usb/dwc3/core.o
>>>    .text.dwc3_of_parse
>>>                   0x00000000080695c4      0x310 drivers/usb/dwc3/core.o
>>>                   0x00000000080695c4                dwc3_of_parse
>>>
>>>>
>>>>> Do you know of any QEMU/U-Boot combination where USB ever worked on the board?
>>>>
>>>> Why was this visible in CI in past and why this pops up now?
>>>> What has changed? I don't think there was any change in QEMu
>>>> regarding USB but I can double check.
>>>
>>> We never started USB in the CI.
>>> With patch 6/6 of the series usb_bootdev_hunter() is invoked which starts USB.
>>
>> Did you push this to gitlab CI? Which test/py is failing there?
> 
> Er, we don't do "usb start" in CI yet, is why it doesn't fail there. Or
> are you suggesting this is a QEMU problem ?

This is where it is failing from dwc3_core_init().
dwc3_writel(dwc->regs, DWC3_DCTL, DWC3_DCTL_CSFTRST);

It means it is likely qemu model limitation.

Thanks,
Michal