Re: [PATCH] mm: fix page_mkclean_one (was: 2.6.19 file content corruption on ext3)

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

On Thu, 2006-12-21 at 10:16 +0100, Martin Schwidefsky wrote:
> On Thu, 2006-12-21 at 00:03 +0100, Peter Zijlstra wrote:
> > current version
> 
> Nitpicking ..
> 
> > @@ -444,17 +444,18 @@ static int page_mkclean_one(struct page
> >  	if (!pte)
> >  		goto out;
> > 
> > -	if (!pte_dirty(*pte) && !pte_write(*pte))
> > -		goto unlock;
> > +	while (pte_dirty(*pte) || pte_write(*pte)) {
> > +		pte_t entry;
> > 
> > -	entry = ptep_get_and_clear(mm, address, pte);
> > -	entry = pte_mkclean(entry);
> > -	entry = pte_wrprotect(entry);
> > -	ptep_establish(vma, address, pte, entry);
> > -	lazy_mmu_prot_update(entry);
> > -	ret = 1;
> > +		flush_cache_page(vma, address, pte_pfn(*pte));
> > +		entry = ptep_clear_flush(vma, address, pte);
> > +		entry = pte_wrprotect(entry);
> > +		entry = pte_mkclean(entry);
> > +		ptep_establish(vma, address, pte, entry);
> 
> Now you are flushing the tlb twice. ptep_clear_flush clears the pte and
> flushes the tlb, ptep_establish sets the new pte and flushes the tlb.
> Not good. Use set_pte_at instead of the ptep_establish.

Yeah, sorry, I already noticed and corrected that :-|

Also, I'm dubious about the while thing and stuck a WARN_ON(ret) thing
at the beginning of the loop. flush_tlb_page() does IPI the other cpus
to flush their tlb too, so there should not be a SMP race, Arjan?

> > +		lazy_mmu_prot_update(entry);
> > +		ret = 1;
> > +	}
> > 
> > -unlock:
> >  	pte_unmap_unlock(pte, ptl);
> >  out:
> >  	return ret;
>