Recently we started seeing slight file corruption and random segmentation
faults with the 2.4.x MIPS kernel from CVS. The problems appeared after
upgrading from 2.4.20 (CVS 2003-01-13) to 2.4.21-pre4 (CVS 2003-05-06), which
introduced the new cache/tlb optimizations.
Most prominent indication of the file corruption is the corruption of
/etc/motd, of which the non-first lines are rewritten by the startup scripts on
every boot up.
The CPU contains a VR4120A core, running in big endian mode. It should be very
similar to the core in the VR4131, with the following exceptions:
- both the instruction and data cache are direct mapped instead of 2-way
- the data cache is only 8 KiB instead of 16 KiB
Perhaps this rings a bell?
Our cross-toolchain consists of gcc 3.2.2 and binutils 18.104.22.168. Userland is
Debian (mostly woody).
Relevant part of dmesg:
| CPU revision is: 00000c72
| Primary instruction cache 16kB, physically tagged, direct mapped, linesize 16
| Primary data cache 8kB direct mapped, linesize 16 bytes.
Relevant part of /proc/cpuinfo:
| cpu model : NEC VR4122 V7.2
| BogoMIPS : 165.88
| wait instruction : no
| microsecond timers : yes
| tlb_entries : 32
| extra interrupt vector : no
| hardware watchpoint : no
| VCED exceptions : not available
| VCEI exceptions : not available
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- firstname.lastname@example.org
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds