Re: kernel locking for releasing incore security

From: Chris Wright (chrisw@private)
Date: Wed Dec 07 2005 - 17:56:44 PST


* Hawk Xu (h.xu@private) wrote:
> Serge E. Hallyn wrote:
> >Right, within reason.
> >
> >	1. In
> >		inode_free_security(struct inode *my_inode) {
> >		}
> >	you don't need to lock my_inode, because it won't be accessed by
> >	anyone else at that point.
> >
> > 
> >
> What about my_inode->i_security? Will it be accessed by anyone else at 
> this point? I'm sure that the data pointed by i_security is not shared 
> with anyone else.

my_inode is no longer accessible by anyone else at this point.

> And what about task->security at the point of task_free_security?

Same for task.  The only point is if you are doing some sort sharing of
->security (used in multiple tasks, kept on some list, etc), then you'd
have to do proper refcounting/locking.

thanks,
-chris



This archive was generated by hypermail 2.1.3 : Wed Dec 07 2005 - 17:57:45 PST