Hello, First a little background for the problem I'm looking at: I picked up an SR71000 support patch (from Jason Gunthorpe <jgg@debian.org> sent to this list) and have applied it to our 2.4 MIPS kernel tree. This is a Sandcraft EV-SR71000-500 cpu module running on a Galileo EV-64240-BP base board in big endian mode. What happens is the kernel comes up to the point in init where /sbin/init is execve'd and hangs (or waits) forever. Since our QED RM7000 module (which we also ported) is able to bootup all the way we don't suspect file system problems. This board has the PMON Galileo Technology monitor ver. 1.4 in 8-bit flash which does some SR71K cache init by doing a secondary/tertiary flash initialize operation as outlined in the SR71K Cpu Spec in section 7.5.18. At the load_mmu stage of bring-up I noticed I cannot call the "clear_enable_caches" routine without suffering an exeception. This is in arch/mips/mm/c-sr71000.c, routine ld_mmu_sr71000 where the call is made to clear_enable_caches as a KSEG1 address. I noticed the config register K0 is already set (before load_mmu) to 0x3 (Cached, Write-back) and that L2, L3 are enabled, no doubt due to some init done by the PMON monitor. Temporarily commenting out the call to clear_enable_caches step, the kernel gets past the exception problem and progresses to the point where execve seems to hang. So my questions are: Is this the latest SR71K support patch? (I can't get access in my browser to http://oss.sgi.com/mips/archive to see prior postings on this...) Is anyone else using this patch? If so, is it working ok and what hardware flavor(s) are you using? What SR71K cpu init is your bootloader or monitor doing before your kernel runs? Any ideas why the tag init loop and secondary/tertiary flash invalidate step in clear_enable_caches crashes the boot-up? Tia, Louis Louis Hamilton Red Hat, Inc. hamilton@redhat.com
