Atsushi Nemoto wrote: > On Thu, 19 Oct 2006 09:29:48 +0200, Franck Bui-Huu <vagabon.xyz@xxxxxxxxx> wrote: >>> But we need sign extension for values comes from the initrd_header >>> anyway. The initrd_header is fixed-size and can not hold 64-bit >>> address. >>> >> initrd_header gives only 2 numbers: size of initrd, and a magic number, >> well it's what I understood when rewriting the code. >> >> the start address of initrd is given by: >> >> initrd_header = __va(PAGE_ALIGN(__pa_symbol(&_end) + 8)) - 8; >> >> >> I don't think we need any sign extension here, do we ? > > Oh I was confused. You are right. yeah I was confused at the first look too. I think rewriting it like this: initrd_start = (unsigned long)(initrd_header + 2); is less confusing... > > BTW, we can just ignore initrd instead of panic() if invalid values > given. This place around is too early to print panic message. > Continuing with printk(KERN_ERR) will give us a chance to see what's > wrong. > ok, I'm rebuilding a new patchset and sending it pretty soon, Franck
