Kernel Kernel bug caused by (cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()) on Banana Pi R64 (MT7622)

[Matthias Brugger]

Hi Nick,

On 09/11/2022 14:35, Nick wrote:
> Hi,
> while trying to bump OpenWrt Kernel to 6.1rc2 I noticed that the kernel is 
> crashing while booting on a Banana Pi R64 (MT7622):
> 
>> [    1.055565] ------------[ cut here ]------------
>> [    1.060204] Kernel BUG at regulator_check_voltage+0xb0/0xf0 [verbose debug 
>> info unavailable]
>> [    1.062418] mtk-pcie 1a143000.pcie: host bridge /pcie at 1a143000 ranges:
>> [    1.068656] Internal error: Oops - BUG: 00000000f2000800 [#1] SMP
>> [    1.075248] mtk-pcie 1a143000.pcie: Parsing ranges property...
>> [    1.081257] Modules linked in:
>> [    1.081264] CPU: 1 PID: 328 Comm: kworker/1:7 Tainted: G S                 
>> 6.1-rc2 #0
>> [    1.087088] mtk-pcie 1a143000.pcie:      MEM 0x0020000000..0x0027ffffff -> 
>> 0x0020000000
>> [    1.090126] Hardware name: Bananapi BPI-R64 (DT)
>> [    1.110541] Workqueue: events dbs_work_handler
>> [    1.114988] pstate: 20000005 (nzCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
>> [    1.121944] pc : regulator_check_voltage+0xb0/0xf0
>> [    1.126728] lr : regulator_set_voltage_unlocked+0x88/0x110
>> [    1.129638] mmc1: host does not support reading read-only switch, assuming 
>> write-enable
>> [    1.132207] sp : ffffffc00956bb30
>> [    1.132209] x29: ffffffc00956bb30 x28: ffffff8000efb400 x27: 0000000000000024
>> [    1.132219] x26: 00000000001312d0 x25: 0000000000118c30 x24: 00000000001312d0
>> [    1.132227] x23: 0000000000149970
>> [    1.146036] mmc1: new high speed SDHC card at address e624
>> [    1.150642]  x22: ffffff800038f800
>> [    1.159192] mmcblk1: mmc1:e624 SL16G 14.8 GiB
>> [    1.161068]  x21: ffffff8000efb100
>> [    1.161072] x20: 00000000001312d0
>> [    1.175424] GPT:partition_entry_array_crc32 values don't match: 0xa0b5ce6d 
>> != 0xab54d286
>> [    1.177757]  x19: 0000000000000000 x18: 00000000799b2550
>> [    1.181067] GPT:Primary header thinks Alt. header is not at the end of the 
>> disk.
>> [    1.189143] x17: 0000000000000003 x16: 0000000000000001 x15: 0000000000000000
>> [    1.189151] x14: 0000000000000000 x13: 0000000000000146 x12: 00000000fa83b2da
>> [    1.189159] x11: 000000000000013d x10: 0000000000000850
>> [    1.194472] GPT:305184 != 31116287
>> [    1.201842]  x9 : ffffffc00956b910
>> [    1.201846] x8 : ffffff8000b9edf0 x7 : 0000000000000001
>> [    1.208970] GPT:Alternate GPT header not at the end of the disk.
>> [    1.216092]  x6 : 00000000001312d0
>> [    1.216095] x5 : 0000000000118c30 x4 : 0000000000000000 x3 : 0000000000000000
>> [    1.216103] x2 : ffffffc00956bb68 x1 : ffffffc00956bb6c
>> [    1.221321] GPT:305184 != 31116287
>> [    1.224706]  x0 : ffffff800038f800
>> [    1.228095] GPT: Use GNU Parted to correct GPT errors.
>> [    1.233307]
>> [    1.233309] Call trace:
>> [    1.233312]  regulator_check_voltage+0xb0/0xf0
>> [    1.242680] FIT: Selected configuration: 
>> "config-mt7622-bananapi-bpi-r64-pcie1" (OpenWrt bananapi_bpi-r64 with 
>> mt7622-bananapi-bpi-r64-pcie1)
>> [    1.242694]  regulator_set_voltage+0x3c/0x64
>> [    1.249831] FIT:           kernel sub-image 0x00001000..0x0052fd0a 
>> "kernel-1" (ARM64 OpenWrt Linux-6.1-rc2)
>> [    1.255030]  mtk_cpufreq_voltage_tracking+0x11c/0x26c
>> [    1.255039]  mtk_cpufreq_set_target+0x1c4/0x350
>> [    1.258444] FIT:          flat_dt sub-image 0x00530000..0x005380c5 "fdt-1" 
>> (ARM64 OpenWrt bananapi_bpi-r64 device tree blob)
>> [    1.261820]  __cpufreq_driver_target+0x2f4/0x674
>> [    1.261826]  od_dbs_update+0xb8/0x19c
>> [    1.266969] FIT:          flat_dt sub-image 0x00539000..0x0053911a 
>> "fdt-mt7622-bananapi-bpi-r64-pcie1" (ARM64 OpenWrt bananapi_bpi-r64 device 
>> tree overlay mt7622-bananapi-bpi-r64-pcie1)
>> [    1.268431]  dbs_work_handler+0x3c/0x7c
>> [    1.270883] FIT:          flat_dt sub-image 0x0053a000..0x0053a20f 
>> "fdt-mt7622-bananapi-bpi-r64-sata" (ARM64 OpenWrt bananapi_bpi-r64 device tree 
>> overlay mt7622-bananapi-bpi-r64-sata)
>> [    1.275297]  process_one_work+0x200/0x3a0
>> [    1.287998] FIT:       filesystem sub-image 0x0053b000..0x00859fff 
>> "rootfs-1" (ARM64 OpenWrt bananapi_bpi-r64 rootfs)
>> [    1.292237]  worker_thread+0x170/0x4c0
>> [    1.292244]  kthread+0xd4/0xe0
>> [    1.302066] FIT: selecting configured loadable "rootfs-1" to be root 
>> filesystem
>> [    1.307092]  ret_from_fork+0x10/0x20
>> [    1.311631]  mmcblk1: p1 p2 p3 p4 p5 p6 p65(rootfs-1) p66(rootfs_data) p128
>> [    1.322903] Code: 6b04001f 54fffe6b 2a0003e4 17fffff3 (d4210000)
>> [    1.413322] ---[ end trace 0000000000000000 ]---
> The complete log can be found here:
> https://gist.githubusercontent.com/PolynomialDivision/395d009c84b426d780549c5fa1f64ff1/raw/886d621d2bf6f03429586adf8a14a6c37c8d8a7d/mt7622-6-1.log
> 
> Reverting commit "cpufreq: mediatek: Refine mtk_cpufreq_voltage_tracking()" 
> fixes the kernel bug:
> https://github.com/torvalds/linux/commit/6a17b3876bc8303612d7ad59ecf7cbc0db418bcd
> 
> The revert commit can be found here:
> https://github.com/PolynomialDivision/openwrt/commit/1df941d0334000e3aced43b7d50cdac0da8bf427
> 
> The branch I use to build the 6.1rc2 on a Banana Pi R64 can be found here:
> https://github.com/PolynomialDivision/openwrt/commits/bump-mt7622-rebase
> 

Thanks for the report.
Could you test with a plain upstream kernel? That would help us to verify that 
this is a upstream problem and not introduced by some openwrt patches.

Regards,
Matthias