lkml.org 
[lkml]   [2007]   [Nov]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: per-bdi-throttling: synchronous writepage doesn't work correctly
From
Date
On Thu, 2007-11-01 at 18:16 +0100, Peter Zijlstra wrote:
> On Thu, 2007-11-01 at 18:09 +0100, Peter Zijlstra wrote:
> > On Thu, 2007-11-01 at 18:00 +0100, Miklos Szeredi wrote:
> > > > > Hi,
> > > > >
> > > > > It looks like bdi_thresh will always be zero if filesystem does
> > > > > synchronous writepage, resulting in very poor write performance.
> > > > >
> > > > > Hostfs (UML) is one such example, but there might be others.
> > > > >
> > > > > The only solution I can think of is to add a set_page_writeback();
> > > > > end_page_writeback() pair (or some reduced variant, that only does
> > > > > the proportions magic). But that means auditing quite a few
> > > > > filesystems...
> > > >
> > > > Ouch...
> > > >
> > > > I take it there is no other function that is shared between all these
> > > > writeout paths which we could stick a bdi_writeout_inc(bdi) in?
> > >
> > > No, and you can't detect it from the callers either I think.
> >
> > The page not having PG_writeback set on return is a hint, but not fool
> > proof, it could be the device is just blazing fast.
> >
> > I guess there is nothing to it but for me to grep writepage and manually
> > look at all hits...
>
> writepage: called by the VM to write a dirty page to backing store.
> This may happen for data integrity reasons (i.e. 'sync'), or
> to free up memory (flush). The difference can be seen in
> wbc->sync_mode.
> The PG_Dirty flag has been cleared and PageLocked is true.
> writepage should start writeout, should set PG_Writeback,
> and should make sure the page is unlocked, either synchronously
> or asynchronously when the write operation completes.
>
> If wbc->sync_mode is WB_SYNC_NONE, ->writepage doesn't have to
> try too hard if there are problems, and may choose to write out
> other pages from the mapping if that is easier (e.g. due to
> internal dependencies). If it chooses not to start writeout, it
> should return AOP_WRITEPAGE_ACTIVATE so that the VM will not keep
> calling ->writepage on that page.
>
> See the file "Locking" for more details.
>
>
> The "should set PG_Writeback" bit threw me off I guess.

Hmm, set_page_writeback() is also the one clearing the radix tree dirty
tag. So if that is not called, we get in a bit of a mess, no?

Which makes me think hostfs is buggy.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2007-11-01 20:09    [W:2.077 / U:0.020 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site