RTEMS | bsps/mips: The compiler emits rdhwr for a processor which lacks it (#5753)

[Sebastian Huber (@sebhub)]

Issue created by Sebastian Huber: https://gitlab.rtems.org/rtems/rtos/rtems/-/work_items/5753



The compiler reads the thread pointer with `rdhwr $3, $29` for every
access to a thread-local object. The instruction exists since MIPS32r2.
The R3900 of `mips/jmr3904` raises a reserved instruction exception
instead, so RTEMS emulates the instruction in `mips_emulate_rdhwr_ulr()`
of `bsps/mips/shared/irq/vectorexceptions.c`.

No compiler option avoids this. `mips-rtems7-gcc` 15.2.0 emits the
instruction under every value of `-ftls-model`. It emits it for
`-march=r3900` as well, wrapped in `.set push`, `.set mips32r2` and
`.set pop`. The MIPS backend has no `-mtp=` option.

The ARM port shows the shape of the fix. `-mtp=soft` makes the compiler
emit a call of `__aeabi_read_tp()`, and RTEMS supplies that function in
`cpukit/score/cpu/arm/__aeabi_read_tp.c`.

Give the MIPS backend of GCC the same option. Add the helper to
`cpukit/score/cpu/mips/`. It returns the thread pointer of the executing
thread, like the ARM and the MicroBlaze helpers do. Put the option in
the `ABI_FLAGS` of every MIPS BSP, so an application object carries it
too. `_CPU_Use_thread_local_storage()` then has nothing to do.
`mips_emulate_rdhwr_ulr()` and the reserved instruction case of
`mips_vector_exceptions()` go away.

The general dynamic model is the alternative. The compiler emits a call
of `__tls_get_addr()` for every access, which is what the MicroBlaze
port uses. It costs more. The change touches the TLS relocations, and
the linker relaxes a general dynamic sequence to local exec in a static
link. That relaxation emits `rdhwr` again.

Found in the review of merge request !1401 by Kinsey Moore. This
description was created with Claude Code assistance.

-- 
View it on GitLab: https://gitlab.rtems.org/rtems/rtos/rtems/-/work_items/5753
You're receiving this email because of your account on gitlab.rtems.org. Unsubscribe from this thread: https://gitlab.rtems.org/-/sent_notifications/5-8vjt45fk72e4wq1wgore65ynz-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/20260904/f86a2879/attachment.htm>