On Tue, Dec 16, 2025 at 07:41:11AM +0100, Christophe Leroy (CS GROUP) wrote:

Le 14/12/2025 à 14:02, Mukesh Kumar Chaurasiya a écrit :

quoted

From: Mukesh Kumar Chaurasiya <redacted>

Convert the PowerPC syscall entry and exit paths to use the generic
entry/exit framework by selecting GENERIC_ENTRY and integrating with
the common syscall handling routines.

This change transitions PowerPC away from its custom syscall entry and
exit code to use the generic helpers such as:
  - syscall_enter_from_user_mode()
  - syscall_exit_to_user_mode()

As part of this migration:
  - The architecture now selects GENERIC_ENTRY in Kconfig.
  - Old tracing, seccomp, and audit handling in ptrace.c is removed in
    favor of generic entry infrastructure.
  - interrupt.c and syscall.c are simplified to delegate context
    management and user exit handling to the generic entry path.
  - The new pt_regs field `exit_flags` introduced earlier is now used
    to carry per-syscall exit state flags (e.g. _TIF_RESTOREALL).

This aligns PowerPC with the common entry code used by other
architectures and reduces duplicated logic around syscall tracing,
context tracking, and signal handling.

The performance benchmarks from perf bench basic syscall are below:

perf bench syscall usec/op

| Test            | With Patch | Without Patch | % Change |
| --------------- | ---------- | ------------- | -------- |
| getppid usec/op | 0.207795   | 0.210373      | -1.22%   |
| getpgid usec/op | 0.206282   | 0.211676      | -2.55%   |
| fork usec/op    | 833.986    | 814.809       | +2.35%   |
| execve usec/op  | 360.939    | 365.168       | -1.16%   |

perf bench syscall ops/sec

| Test            | With Patch | Without Patch | % Change |
| --------------- | ---------- | ------------- | -------- |
| getppid ops/sec | 48,12,433  | 47,53,459     | +1.24%   |
| getpgid ops/sec | 48,47,744  | 47,24,192     | +2.61%   |
| fork ops/sec    | 1,199      | 1,227         | -2.28%   |
| execve ops/sec  | 2,770      | 2,738         | +1.16%   |

I get about 2% degradation on powerpc 8xx, and it is quite stable over time
when repeating the test.

'perf bench syscall all' on powerpc 8xx (usec per op):

| Test            | With Patch | Without Patch | % Change |
| --------------- | ---------- | ------------- | -------- |
| getppid usec/op | 2.63       | 2.63          | ~ 0%     |
| getpgid usec/op | 2.26       | 2.22          | +2,80%   |
| fork usec/op    | 15300      | 15000         | +2,00%   |
| execve usec/op  | 45700      | 45200         | +1.10%   |

Ohk,
Do you have any idea where we might be loosing performance? I don't have
an 8xx chip so i am not sure where to look.

Thanks for testing it out.

Regards,
Mukesh

Christophe