[U-Boot] saveenv corrupts QSPI flash with latest commit U-Boot 2019.04-rc4-00047-gcfb3e102c4

[Vignesh Raghavendra]

On 26/03/19 7:11 PM, Ashish Kumar wrote:
>>
>> On 26/03/19 10:36 AM, Ashish Kumar wrote:
>>> Hello Maintainers,
>>>
>>>
>>>
>>> With latest commit I see that saveenv command corrupts QSPI-flash,
>>> meaning if I read QSPI-flash at 0x0 offset RCW(reset configuration
>>> word) is erased after saveenv command was executed.
>>>
>>> This is tested on LS1088ARDB, but it should not be limited to
>>> LS1088ARDB rather it will valid for all LS Freescale ARM boards.( like
>>> LS1088, LS2080/LS2085, LS1012, LS1043, LS1046).
>>>
>>
>> Which is the SPI controller driver? Does the controller driver support
>> reading/writing to flash using 4 byte addressing opcode?
> 
> fsl_qspi.c. As per the migration it was not migrated to 4 byte. Although it actually supports 4 byte for some SoC and 3 byte for older SoC that is the reason you see CONFIG_SPI_FLASH_BAR in code. My concerned SoC are all supporting 4 byte. I have not added any code modification in the current u-boot.
>>

OK, does normal read/write to offset 0 work fine? That is:

sf probe
sf erase 0x0 40000
sf write <add1r> 0x0 0x40000 (write some random data)
sf read <addr2> 0x0 0x40000 (read back)
md.b <addr2>

If this fails. Could you post full debug log (all debug prints enabled
in spi_mem_exec_op()) to pastebin.ubuntu.com and provide a link here?

>> Could you enable all the debug prints in spi_mem_exec_op() (especially those at
>> the end)?
> Logs are very verbose: I have commented two debug logs which are in for loop
> Logs here are from the end, 
> 0b 33 f3 43 | [59B in] [ret 0] 

I dont know the address from which saveenv is trying to read from. But
does the address bytes match (33 f3 43)?

[...]
> 0b 33 ff f0 | [16B in] [ret 0]
> Erasing SPI flash...06 | [0B -] [ret 0]
> d8 | [3B out] [ret 0] 


This is the sector erase command but you haven't enabled log for the
part where address of the sector is dumped. Could you provide that info?

Also, the commit ID in the subject is not present in upstream tree. So I
am not sure what exactly maybe broken. Does 2019.01 work fine?
U-Boot had major sf layer refactoring in 2019.04-rc1. Could you see if
2019.04-rc1 works fine?




-- 
Regards
Vignesh