Hi Mulyadi,
struct embedded_fd_set {
unsigned long fds_bits[1];
};
http://lxr.linux.no/linux/include/linux/file.h#L26
[...]
so, I bet it's bitwise operation...should work atomically.
Yes, that's what I thought too, though the bit set used in fdtable
should be __kernel_fd_set and not embedded_fd_set, which is used in the
files_struct. Doesn't matter though, bit setting is bit setting.
FD_SET(fd, fdt->open_fds);
if (flags & O_CLOEXEC)
FD_SET(fd, fdt->close_on_exec);
else
FD_CLR(fd, fdt->close_on_exec);
However, between FD_SET(fd, fdt->open_fds) and FD_SET(fd,
fdt->close_on_exec)|FD_CLR(fd, fdt->close_on_exec) the state is
inconsistent.
I know that kernel preemption is disabled and that you shouldn't access
these entries until after the syscall has completed. However, interrupt
handling isn't disabled, so that gap can be enlarged "arbitrarily".
Whether this actually matters, I do not know as I do not know all legal
access paths/kernel control paths to this structure. For instance, if an
enumeration over all fds as in close_files() is always legal.
Cheers,
Martin
--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ
