On 22/09/26 6:56 PM, Anushree Mathur wrote:

On 03/09/26 5:20 PM, Kirill A. Korinsky wrote:
Greetings,

OpenBSD/macppc does not complete early boot on mac99: OpenPIC can
reassert an interrupt still masked by CTPR; the edge triggered
decrementer does not raise on zero; successful CUDA ADB replies omit the
original command; and the fixed CUDA shift register delay can stretch
packet delivery past the guest timeout on hosts with coarse timers.

This series corrects these four behaviours and adds a functional test
which boots the official OpenBSD 7.9 macppc CD to the installer prompt.

P.S. this is second third to send the same patch because previous one
had missed maintainers.


Hi Kirill,

Thanks for working on this patch series. I applied this patch set and ran verification tests on a Power (ppc64le) host; everything worked as expected.

Here is the detailed test summary:

1. Functional Test:
   Command:
   QEMU_TEST_QEMU_BINARY=./qemu-system-ppc64 ./run pyvenv/bin/python3 ../tests/functional/ppc/test_openbsd.py

   Result:
   TAP version 13
   ok 1 test_openbsd.OpenBSDMac99.test_openbsd_79_installer
   1..1

2. Manual OpenBSD Boot (TCG):
   Booted OpenBSD 7.9 RAMDISK image (cd79.iso) on mac99 with qemu-system-ppc64.    Verified that OpenPIC, CPU decrementer calibration, and CUDA/ADB keyboard
   probing attached cleanly and reached the installer prompt:

   "Bochs VGA" rev 0x02 at pci0 dev 13 function 0 not configured
   softraid0 at root
   scsibus1 at softraid0: 256 targets
   bootpath: /pci@f2000000/mac-io@c/ata-3@21000/cdrom@0:/7.9/macppc/bsd.rd
   root on rd0a swap on rd0b dump on rd0b
   WARNING: CHECK AND RESET THE DATE!
   erase ^?, werase ^W, kill ^U, intr ^C, status ^T

   Welcome to the OpenBSD/macppc 7.9 installation program.
   (I)nstall, (U)pgrade, (A)utoinstall or (S)hell?

3. Patch 2/5 (Decrementer Zero Write) A/B Check:
   - Positive check: Kernel boots through clock calibration into idle loops cleanly.    - Negative check: Reverting the patch 2 change caused the guest to reliably stall      during early boot right after CPU probe. Re-applying resolved the hang immediately.

4. Concurrency & Non-Regression:
   - Verified OpenBSD mac99 boot with both single-threaded (thread=single) and
     multi-threaded TCG.
   - Verified KVM/TCG guest bringup (pseries) on the ppc64le host with different thread
     configurations without any regressions.

Missed to add the space between Tested-by and name so resending tag in correct format for b4 tool to pick it

Please feel free to add:
Tested-by: Anushree Mathur <[email protected]>
Please feel free to add:
Tested-by: Anushree Mathur<[email protected]>

Thanks,
Anushree Mathur

Reply via email to