Found 2 bugs in OpenBSD 3.3

From: Rahul Patel <rahul(at)transmeta.com>
Date: Tue Jul 01 2003 - 02:28:12 EDT

Hello,

Our compatibility lab has found 2 bugs while testing OpenBSD 3.3 on one of the next generation Transmeta processors. Both of these hangs occurs while booting the install CD. Our compatibility lab has identified that both issues will occur with an AMD Opteron platform as well.

Bug #1 : Spurious Interrupts

---

One bug that we found is that OpenBSD seems to get hung if a spurious interrupt occurs. During the installation process, we found that the system gets a spurious interrupt vector 0x27. After this, OpenBSD is stuck forever in the IRQ7 handler. The spurious interrupt happens during most boot attempts, but not every time. The same spurious interrupt and subsequent hang usually occurs on the Opteron platform as well.

In the occasional case that the spurious interrupt does not occur, OpenBSD boot will continue then hang with the problem described below.

Bug #2 : CPU Mis-recognized

---

Our engineer found that OpenBSD will incorrectly identify our CPU as a 386, and branch off to legacy code which causes an unexpected page fault and hangs. To workaround this, you can set physical memory location 0x50b1c0 to 0x3 (CPUCLASS_686) after OpenBSD has finished detecting the CPU. With this hack, OpenBSD can successfully boot up to the menu.

This same cpu recognition failure occurs on the Opteron system as well.

We hope that this information will help allow OpenBSD to work with Transmeta processors as well as AMD Opteron, and potentially others.

Best Regards,
Rahul

o----------------------------o

  Rahul Patel
  Manager, Project Engineering
  Transmeta Corporation

o----------------------------o

Received on Tue Jul 1 02:46:44 2003