[Khadas WiP] VIM4 NVMe IO Errors

VIM4
102

You can check the source code. :wink:

103

@numbqq

Updated the kernel and it works with the WD 570, no serious testing, just functionality. Your test kernel is configured differently so it did break nftables/UFW so I went back to my custom kernel config. Where is the kernel config file in fenix?

104

Really? No change to che config.

1 Like
105

Willy Tarreau maintained 3.10 LTS kernel so he qualifies as ā€˜officialā€™ kernel hacker. Me not at all, in fact I learned just few weeks ago how shitty the situation with those Android vendor kernels is since I believed for many years Google would require those SoC vendors to rebase all their code on upstream kernel of the version they require for a certain Android version.

But thatā€™s not happening, they just forward port their crappy internal code base since forever, that way hiding every merge conflict that ever happened and as such even with a 5.4 version number the kernel running on any VIM4 is far away from official 5.4 LTS.

How far? You would need to take your time and efforts but itā€™s easily hundreds of thousands of differences. Anyone trusting into this crap is lost. But since people have no idea that neither SoC vendors nor board vendors give a shit they rely on this kernel even for security relevant stuff.

@foxsquirrel mentioned nftables. So I just checked out net/netfilter/nf_tables_api.c from official 5.4 LTS and from the Amlogic/Khadas mess:

https://www.diffchecker.com/frCrkIrU

Thatā€™s just a single file and that is something that has zero relationship to Amlogic adoptions (in such cases it would be understandable that official LTS and vendor kernel differ).

106

Yes, that is correct. I had to change the ā€œofficial releaseā€ kernel to get nftables and UFW to work.

Thank you for the link.

Tried to post the diff and the file is too large for the forum.

107

So far I have not had any IO issues under load. The VIM4 has been building RetroPie for about 12 hours now. Thank you and the dev team for all their efforts. It seems this is a very messy situation.

1 Like
108

The only thing that may be missing is the kernel headers, unless Iā€™ve missed them.

109

Iā€™m having problems with a Silicon Power SU001TBP34A60M28CA (1 TB Gen 3x4 NVMe). It appears as unclaimed under sudo lshw, it appears in sudo lspci, but not under sudo lsblk.

Iā€™m also getting system errors:

sudo dmesg | grep nvme
[ 0.686377] nvme nvme0: pci function 0000:01:00.0
[ 0.686618] nvme 0000:01:00.0: enabling device (0000 ā†’ 0002)
[ 128.773790] nvme nvme0: Device not ready; aborting initialisation
[ 128.773893] nvme nvme0: Removing after probe failure status: -19