RTEMS | bsps/mips/malta: The board never resets and the PCI probe prints at every boot (#5730)

[Sebastian Huber (@sebhub)]

Issue created by Sebastian Huber: https://gitlab.rtems.org/rtems/rtos/rtems/-/work_items/5730



`bsp_reset()` writes `0xFF` to the software reset register of the Malta FPGA.
Qemu resets the board on `0x42` and ignores every other value, so the board
never resets. The comment above the write records where the value came from:
a local change to Qemu which recognised `0xFF` and which never reached the
project.

The store carries two more defects, and either one alone stops it from
reaching the register. Its address, `0x9F000500`, is the cached alias of the
register rather than the uncached one at `0xBF000500`. Its pointer is not
`volatile` and `RTEMS_UNREACHABLE()` follows it, so a compiler may drop the
store as dead. The compiler of RTEMS 7 does drop it: the store is absent from
`bsp_reset()` in the object file until the pointer becomes `volatile`.

The board therefore runs on past `RTEMS_UNREACHABLE()`. Every executable of
this BSP runs its test to the end, prints the end of test line and then loops
in exception frames. One run of `hello` printed 16686 of them in 25 seconds.
The simulator never exits, so a test runner can only end the run on a timeout
and every test of the BSP reads as a failure.

The PCI probe is the second defect. `print_bars()` prints the six base address
registers of a device it configures, and it prints them twice for one device.
Every executable opens with fourteen lines of that output ahead of the first
line of the application. The file carries a `DEBUG_PCI` switch which its other
messages use and these calls do not.

The BSP states that it has only ever been tested on Qemu, so nothing else has
exercised either path.

Found while checking whether the Malta board can join the simulator test
programme. This description was created with Claude Code assistance.

-- 
View it on GitLab: https://gitlab.rtems.org/rtems/rtos/rtems/-/work_items/5730
You're receiving this email because of your account on gitlab.rtems.org. Unsubscribe from this thread: https://gitlab.rtems.org/-/sent_notifications/5-9rndgoudmnd9u6lboyfzsdibh-1d/unsubscribe | Manage all notifications: https://gitlab.rtems.org/-/profile/notifications | Help: https://gitlab.rtems.org/help


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/bugs/attachments/20260824/d34e684f/attachment-0001.htm>