Re: [PATCH -rt] scheduling while atomic in remove_proc_entry()

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Fri, 2006-12-29 at 13:12 -0800, Daniel Walker wrote:

> 
> Signed-Off-By: Daniel Walker <dwalker@mvista.com>
> 
> ---
>  fs/proc/generic.c |    3 ++-
>  1 files changed, 2 insertions(+), 1 deletion(-)
> 
> Index: linux-2.6.19/fs/proc/generic.c
> ===================================================================
> --- linux-2.6.19.orig/fs/proc/generic.c
> +++ linux-2.6.19/fs/proc/generic.c
> @@ -555,7 +555,6 @@ static void proc_kill_inodes(struct proc
>  	/*
>  	 * Actually it's a partial revoke().
>  	 */
> -	filevec_add_drain_all();
>  	lock_list_for_each_entry(filp, &sb->s_files, f_u.fu_llist) {
>  		struct dentry * dentry = filp->f_path.dentry;
>  		struct inode * inode;
> @@ -738,6 +737,8 @@ void remove_proc_entry(const char *name,
>  		break;
>  	}
>  	spin_unlock(&proc_subdir_lock);
> +
> +	filevec_add_drain_all();
>  out:
>  	return;
>  }

Well, no. Draining after the inspect 'all' loop doesn't make sense, but
looking at 2.6.20-rc3-rt0 remove_proc_entry() looks sane.