On Thu, Feb 5, 2026 at 8:34 PM Richard Henderson <
[email protected]> wrote:

> On 2/6/26 03:26, Warner Losh wrote:
> > From: Stacey Son <[email protected]>
> >
> > Wire up semget(2) and semop(2) syscalls to get System V semaphore
> > implementation, as well the undocumented __semctl used to implement the
> > bits of the interface in libc.
> >
> > Signed-off-by: Stacey Son <[email protected]>
> > Signed-off-by: Warner Losh <[email protected]>
> > ---
> >   bsd-user/freebsd/os-syscall.c | 23 +++++++++++++++++++++++
> >   1 file changed, 23 insertions(+)
> >
> > diff --git a/bsd-user/freebsd/os-syscall.c
> b/bsd-user/freebsd/os-syscall.c
> > index ca2f6fdb66..6e38007bdd 100644
> > --- a/bsd-user/freebsd/os-syscall.c
> > +++ b/bsd-user/freebsd/os-syscall.c
> > @@ -37,6 +37,7 @@
> >   #include "bsd-file.h"
> >   #include "bsd-mem.h"
> >   #include "bsd-proc.h"
> > +#include "bsd-misc.h"
> >
> >   /* BSD dependent syscall shims */
> >   #include "os-stat.h"
> > @@ -879,6 +880,28 @@ static abi_long freebsd_syscall(void *cpu_env, int
> num, abi_long arg1,
> >           ret = do_bsd_shmdt(arg1);
> >           break;
> >
> > +        /*
> > +         * System V Semaphores
> > +         */
> > +    case TARGET_FREEBSD_NR_semget: /* semget(2) */
> > +        ret = do_bsd_semget(arg1, arg2, arg3);
> > +        break;
> > +
> > +    case TARGET_FREEBSD_NR_semop: /* semop(2) */
> > +        ret = do_bsd_semop(arg1, arg2, arg3);
> > +        break;
> > +
> > +    case TARGET_FREEBSD_NR___semctl: { /* __semctl() undocumented */
> > +        /*
> > +         * The semun argument to semctl is passed by value, so
> dereference the
> > +         * ptr argument.
> > +         */
> > +        abi_ulong atptr;
> > +        get_user_ual(atptr, (abi_ulong)arg4);
> > +        ret = do_bsd___semctl(arg1, arg2, arg3,
> > +                (union target_semun)(abi_ulong) atptr);
>
> Eh?  Does this really compile, casting an integer to a union?
> Something seems very confused here.
>

It works, but it's kinda nutso code. I wouldn't have thought you could cast
a long to a union...

It's even more nutso inside do_bsd__semctl()! I think that we need to honor
that this is really a pointer to a union and pass that in as such and then
__get/put_user rather than doing the tswap dance twice.

Warner

Reply via email to