Re: [patch-2.3.43-pre5] optimized vhangup(2)

[Peter Benie]

Tigran Aivazian writes ("[patch-2.3.43-pre5] optimized vhangup(2)"):
> The patch below does this to fs/open.c:sys_vhangup()
> 
>   a) removes a goto
> 
>   b) removes a local variable
> 
>   c) reduces the length of holding big kernel lock

[New code is...]
>  asmlinkage long sys_vhangup(void)
>  {
>  	if (!capable(CAP_SYS_TTY_CONFIG))
> 		return -EPERM;
>  	if (current->tty)
>  		tty_vhangup(current->tty);
> 	return 0;
>  }

Is that actually correct? If the program disassociates itself from the
controlling terminal between here and the kernel lock in
do_tty_hangup, will tty still be valid, and will it be owned by this
process?

Please keep the lock here - this system call is very rare (typ. once
per tty allocation) so deferring taking the kernel lock for a few
microseconds will have negligable affect on system performance.

Peter