Re: UBIFS errors since Linux 5.15

[Zhihao Cheng]

在 2025/11/8 11:03, Zhihao Cheng 写道:
> 在 2025/11/8 1:38, Eddie James 写道:
> > Hello,
> >
> > We are updating some BMC (Aspeed AST2600, ARM32) systems that were on 
> > Linux 5.4 but observe consistent UBIFS errors with Linux 5.15 and 
> > later. The errors seem to vary; there is a sampling below.
> >
> >
> > We performed a git bisect and came up with the following commit, which 
> > of course is completely unrelated to UBIFS. Reverting it does prevent 
> > the issue... Does anyone have any suggestions?
> >
> > commit 33fc9e26b7cb39f0d4219c875a2451802249c225
> [...]
> > [    8.180486] UBIFS error (ubi0:2 pid 66): ubifs_check_node: bad 
> > magic 0x101831, expected 0x6101831
>
> Hi, look here, the magic number 0x101831 vs 0x6101831, the difference is 
> the highest bit(0 vs 6[0110])
> > [    8.189578] UBIFS error (ubi0:2 pid 66): ubifs_check_node: bad node 
> > at LEB 17:64664
>
> [...]
> > [ 3739.678968] UBIFS error (ubi0:2 pid 13): ubifs_check_node: bad CRC: 
> > calculated 0x9179af29, read 0xa760f4f5
>
> [...]
> > [10412.723111] UBIFS error (ubi0:2 pid 297): ubifs_check_node: bad 
> > magic 0x6100830, expected 0x6101831
>
> and there, magic number 0x6100830 vs 0x6101831, the difference is lowest 
> bit(0 vs 1).

Sorry, I need to make a correction, the differences are the lowest bit(0 
vs 1) and the middle bit (0 vs 1). The number of flipped bits is also two.
> So, I guess the crc error is caused by the similar reason, maybe there 
> is something wrong with the nand driver or flash? According to the 
> version information(5.15, and the unrelated commit 33fc9e26b7cb3), is it 
> possible that there is a compatibility issue between your nand(bsp) 
> driver with kernel nand common layer(which from different versions)?
>
> > [10412.732436] UBIFS error (ubi0:2 pid 297): ubifs_check_node: bad 
> > node at LEB 13:60248
>
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/