RTEMS | SPARC profiling is brokem (#5616)

[Matteo Concas (@matteo.concas)]

Matteo Concas created an issue: https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5616

Assignee: Matteo Concas

## Summary
Enabling `RTEMS_PROFILING` when building a SPARC BSP will lead to some undefined behavior. I have seen unaligned accesses:
```
grmon4> load /home/matteo/dev/gsoc/rtems-net-services/build/sparc-rtems7-gr765/iperf3test01.exe; run
                 0 .text              2.4MB /   2.4MB   [===============>] 100%
            25ce40 .tdata              20B              [===============>] 100%
            25ce58 .rtemsroset        3.3kB /   3.3kB   [===============>] 100%
            26a340 .data             93.6kB /  93.6kB   [===============>] 100%
            2819c0 .rtemsrwset        1.1MB /   1.1MB   [===============>] 100%
  Total size: 3.55MB (519.77kbit/s)
  Entry point 0x00000000
  Image /home/matteo/dev/gsoc/rtems-net-services/build/sparc-rtems7-gr740/iperf3test01.exe loaded


*** BEGIN OF TEST libBSD IPERF3 1 ***
*** TEST VERSION: 7.0.0
*** TEST STATE: EXPECTED_PASS
*** TEST BUILD: RTEMS_POSIX_API RTEMS_PROFILING
*** TEST TOOLS: 16.1.0 20260430 (RTEMS 7, RSB 348a70cf182d9e3452504757f53a78719190dc0b, Newlib a0ccec7e)
nexus0: <RTEMS Nexus device>
grethtwo0: <Ethernet Gbit MAC> on nexus0
info: grethtwo0: Ethernet address: 0e:b0:ba:5e:ba:11
info: lo0: link state changed to UP
greth: driver attached
**** PHY ****
Vendor: 885   Device: 22   Revision: 2
Current Operating Mode: 1000 Mbit Full Duplex
Autonegotiation Time: 7528ms
Connecting to host 172.16.103.1, port 5201
[ 28] local 172.16.103.2 port 13962 connected to 172.16.103.1 port 5201
  Error mode  (tt = 0x07, mem address not aligned)
  Hint: Before trap (tt = 0x89, trap instruction)
  0x001e1a48: 81c44000  jmp  %l1
```
or privileged instruction errors:
```
[...]
[ 28] local 172.16.103.2 port 36999 connected to 172.16.103.1 port 5201
  Error mode  (tt = 0x03, privileged instruction)
  Hint: Before trap (tt = 0x89, trap instruction)
  0x001e19d8: a9500000  mov  %wim, %l4
```

The common element here is that the local registers are not restored properly or overwritten somehow.

I will try to run the RTEMS testsuite with profiling enabled to see if any test becomes a fail with it.



<!-- Pre-set options
- milestone
-->

-- 
View it on GitLab: https://gitlab.rtems.org/rtems/rtos/rtems/-/issues/5616
You're receiving this email because of your account on gitlab.rtems.org.


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/bugs/attachments/20260608/681bf366/attachment.htm>