dup3 syscall - atomic set O_CLOEXEC with dup2
Davide Italiano
davide.italiano at gmail.com
Thu Jan 12 09:35:17 UTC 2012
On Thu, Jan 12, 2012 at 6:01 AM, Eitan Adler <lists at eitanadler.com> wrote:
> This is an implementation of dup3 for FreeBSD:
> man page here (with a FreeBSD patch coming soon):
> https://www.kernel.org/doc/man-pages/online/pages/man2/dup.2.html
>
> Is this implementation correct? If so any objection to adding this as
> a supported syscall?
>
>
> Index: sys/kern/kern_descrip.c
> ===================================================================
> --- sys/kern/kern_descrip.c (revision 229830)
> +++ sys/kern/kern_descrip.c (working copy)
> @@ -110,6 +110,7 @@
> /* Flags for do_dup() */
> #define DUP_FIXED 0x1 /* Force fixed allocation */
> #define DUP_FCNTL 0x2 /* fcntl()-style errors */
> +#define DUP_CLOEXEC 0x4 /* Enable O_CLOEXEC on the new fs */
>
> static int do_dup(struct thread *td, int flags, int old, int new,
> register_t *retval);
> @@ -307,7 +308,39 @@
> return (0);
> }
>
> +#ifndef _SYS_SYSPROTO_H_
> +struct dup3_args {
> + u_int from;
> + u_int to;
> + int flags;
> +};
> +#endif
> +
> /*
> + * Duplicate a file descriptor and allow for O_CLOEXEC
> + */
> +
> +/* ARGSUSED */
> +int
> +sys_dup3(struct thread * const td, struct dup3_args * const uap) {
> +
> + KASSERT(td != NULL, ("%s: td == NULL", __func__));
> + KASSERT(uap != NULL, ("%s: uap == NULL", __func__));
> +
> + if (uap->from == uap->to)
> + return EINVAL;
> +
> + if (uap->flags & ~O_CLOEXEC)
> + return EINVAL;
> +
> + const int dupflags = (uap->flags == O_CLOEXEC) ?
> DUP_FIXED|DUP_CLOEXEC : DUP_FIXED;
> +
> + return (do_dup(td, dupflags, (int)uap->from, (int)uap->to,
> + td->td_retval));
> + return (0);
> +}
> +
> +/*
> * Duplicate a file descriptor to a particular value.
> *
> * Note: keep in mind that a potential race condition exists when closing
> @@ -912,6 +945,9 @@
> fdp->fd_lastfile = new;
> *retval = new;
>
> + if (flags & DUP_CLOEXEC)
> + fdp->fd_ofileflags[new] |= UF_EXCLOSE;
> +
> /*
> * If we dup'd over a valid file, we now own the reference to it
> * and must dispose of it using closef() semantics (as if a
> Index: sys/kern/syscalls.master
> ===================================================================
> --- sys/kern/syscalls.master (revision 229830)
> +++ sys/kern/syscalls.master (working copy)
> @@ -951,5 +951,6 @@
> off_t offset, off_t len); }
> 531 AUE_NULL STD { int posix_fadvise(int fd, off_t offset, \
> off_t len, int advice); }
> +532 AUE_NULL STD { int dup3(u_int from, u_int to, int flags); }
> ; Please copy any additions and changes to the following compatability tables:
> ; sys/compat/freebsd32/syscalls.master
> Index: sys/compat/freebsd32/syscalls.master
> ===================================================================
> --- sys/compat/freebsd32/syscalls.master (revision 229830)
> +++ sys/compat/freebsd32/syscalls.master (working copy)
> @@ -997,3 +997,4 @@
> uint32_t offset1, uint32_t offset2,\
> uint32_t len1, uint32_t len2, \
> int advice); }
> +532 AUE_NULL STD { int dup3(u_int from, u_int to, int flags); }
>
>
> --
> Eitan Adler
> _______________________________________________
> freebsd-hackers at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-hackers
> To unsubscribe, send any mail to "freebsd-hackers-unsubscribe at freebsd.org"
>From what I can see it seems that dup3() is Linux specific and not
POSIX, so maybe there are some issues in adding it, but I may be
wrong.
Davide
More information about the freebsd-hackers
mailing list