[smartmontools-support] nvme drive partial support

[Kwon-Young Choi]

Hello,

> Various Windows programs which are useful for "rescue" purposes could be
> run from the limited Windows PE environment. For example: 7-Zip GUI,
> Cygwin (bash, dd, ddrescue, fdisk, hexedit, ...), VeraCrypt, and ...
> smartctl.
> This may also work with the dell firmware update tool. Would require
> 32-bit Win10 install media, as the 64-bit WinPE has no 32-bit subsystem.

Thanks for the tip, I found an archlinux wiki page describing how to
make a WinPE bootable drive, which is awesome. I'll try this later.

> What's left is the drive firmware bug. I possibly will add the
> workaround already described.

I'll make sure to report if something changed after I updated the firmware.

Best regards,

Kwon-Young Choi

On 15/11/2020 18:32, Christian Franke wrote:
> Kwon-Young Choi wrote:
>>> I don't know. If there is no new firmware available which fixes this,
>>> you could possibly report the problem to SK hynix customer support.
>> I found that there is a dell firmware update for my ssd but I need
>> windows to do the update :(
> 
> Various Windows programs which are useful for "rescue" purposes could be
> run from the limited Windows PE environment. For example: 7-Zip GUI,
> Cygwin (bash, dd, ddrescue, fdisk, hexedit, ...), VeraCrypt, and ...
> smartctl.
> This may also work with the dell firmware update tool. Would require
> 32-bit Win10 install media, as the 64-bit WinPE has no 32-bit subsystem.
> 
> 
> I wrote 2020-11-13:
>> ...
>>
>> There is actually a bug in smartctl.
>>
>> The difference in the commands above is the size read from the log. In
>> the early days of smartctl NVMe support (2016), I decided to always
>> read the full log regardless of the desired print size - and
>> unfortunately introduced a bug: smartctl reads too many (64) bytes
>> from the log. Most drives behave polite and return extra 64 null bytes.
> 
> After reading NVMe spec more carefully (as I apparently did 4+ years
> ago), I take that back. Smartctl interprets the number of Error Log Page
> Entries (ELPE) correctly. It is defined as a "0's based value": number N
> represents value N+1. The command "nvme error-log ..." does the same.
> 
> What's left is the drive firmware bug. I possibly will add the
> workaround already described.
> 
> Thanks,
> Christian
>