On July 13, mhtran@xxxxxxxxxx wrote: > On alpha systems, the kernel computed checksum, csum_partial(), and user > space computed checksum are different. As a consequence, the kernel MD > driver aborts detection of MD superblock. This means running mdadm > --assemble -f or -U summaries will not always work. Weird... arch/alpha/checksum.c seem to do a 16 bit checksum rather than the 32bit csum that i386 does. It's mostly used for network code and IP uses 16bit csum, but it is odd that it claims to return a 32 bit number, but really it will always be 16bit... > > I have been wondering... Should MD driver use csum_partial() at all? Well, as we seem to want a 32bit checksum, I guess we cannot. But for historical reasons, we probably have to. I guess a patch much like yours is needed. I'll add something to make sure version-1 superblock are a bit more predictable. NeilBrown - To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
