Re: invalidate_inode_pages in 2.5.32/3

[Daniel Phillips]

On Monday 23 September 2002 22:41, Trond Myklebust wrote:
> >>>>> " " == Andrew Morton <[e-mail]> writes:
> 
>      > Would it be possible to mark the inode as "needs invalidation",
>      > and make user processes check that flag once they have i_sem?
> 
> Not good enough unless you add those checks at the VFS/MM level.

Please see Rik's suggestion and my followups where we are talking about
handling this in the MM.

> Think
> for instance about mmap() where the filesystem is usually not involved
> once a pagein has occurred.
> 
>      > Do we really need to invalidate individual pages, or is it
>      > real-life acceptable to invalidate the whole mapping?
> 
> Invalidating the mapping is certainly a good alternative if it can be
> done cleanly.

But invalidate_inode_pages is (usually) just trying to do exactly that.
It doesn't work.  Anyway, what if you have a 2 gig file with 1 meg
mmaped/locked/whatever by a database?

> Note that in doing so, we do not want to invalidate any reads or
> writes that may have been already scheduled. The existing mapping
> still would need to hang around long enough to permit them to
> complete.

With the mechanism I described above, that would just work.  The fault
path would do lock_page, thus waiting for the IO to complete.

-- 
Daniel