Re: Can prune_icache safely discard inodes which have only clean pages? (2.4.18)

[Andrew Morton <akpm@digeo.com>]

Bill Davenport wrote:
> 
> I've got a system which has a fairly large amount of physical memory (2GB)
> that experiences
> performance problems after a large number of files have been accessed.
> 
> ...

Your analysis is 100% correct.  It's a problem.

There's a fix for this in Andrea's kernel.
 
> ...
> 
> and I'd like to change it to:
> 
>  void prune_icache(int goal)
>  {
>   ...
>   while (entry != &inode_unused)
>   {
>    ...
>    if (inode->i_state & (I_FREEING|I_CLEAR|I_LOCK))
>     continue;
>    if ((inode->i_state != 0) || inode_has_buffers(inode))
>     continue;
>    if (inode->i_data.nrpages != 0) {
>     if ((!list_empty(&inode->i_data.dirty_pages)) ||
>         (!list_empty(&inode->i_data.locked_pages))) {
>      /* skip if any dirty or locked pages */
>      continue;
>     }
>    }

locked_pages tends to hold clean, unlocked pages, alas.  Testing
->dirty_pages makes sense.

If there are no dirty pages then you can run invalidate_inode_pages();
chances are, that will bring ->nrpages to zero, and all is well.