On 2/18/26 9:22 PM, Warner Losh wrote:
From: Stacey Son <[email protected]>
Add implementation of __semctl(2) syscall for System V semaphore control
operations. Handles command translation, endianness conversion for GETVAL/
SETVAL, and array/structure conversions for GETALL/SETALL/IPC_STAT/IPC_SET.
Signed-off-by: Stacey Son <[email protected]>
Signed-off-by: Warner Losh <[email protected]>
---
bsd-user/bsd-misc.h | 116 ++++++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 116 insertions(+)
diff --git a/bsd-user/bsd-misc.h b/bsd-user/bsd-misc.h
index e1e552b58f..cb7fe1d6d0 100644
--- a/bsd-user/bsd-misc.h
+++ b/bsd-user/bsd-misc.h
...
+ case IPC_STAT:
+ case IPC_SET:
+ __get_user(target_buffer, (abi_ulong *)target_un);
+ err = target_to_host_semid_ds(&dsarg, target_buffer);
+ if (is_error(err)) {
+ return err;
+ }
+ arg.buf = &dsarg;
+ ret = get_errno(semctl(semid, semnum, host_cmd, arg));
+ err = host_to_target_semid_ds(target_buffer, &dsarg);
+ if (is_error(err)) {
+ return err;
+ }
+ break;
For all the cases using this pattern:
```
if (is_error(err)) {
return err;
}
```
+ case IPC_RMID:
+ case GETPID:
+ case GETNCNT:
+ case GETZCNT:
+ ret = get_errno(semctl(semid, semnum, host_cmd, NULL));
+ break;
+
+ default:
+ ret = -TARGET_EINVAL;
+ break;
+ }
Shouldn't we instead goto this epilogue to ensure we call unlock_user?
+ unlock_user(target_un, un_ptr, 1);
+ return ret;
+}
+
Regards,
Pierrick
