Re: [PATCH 2/2] thp: support split page table lock

[Naoya Horiguchi]

On Fri, Sep 06, 2013 at 07:46:25PM +0300, Kirill A. Shutemov wrote:
> Naoya Horiguchi wrote:
> > Hi Kirill,
> > 
> > On Fri, Sep 06, 2013 at 01:48:03PM +0300, Kirill A. Shutemov wrote:
> > > Naoya Horiguchi wrote:
> > > > Thp related code also uses per process mm->page_table_lock now.
> > > > So making it fine-grained can provide better performance.
> > > > 
> > > > This patch makes thp support split page table lock by using page->ptl
> > > > of the pages storing "pmd_trans_huge" pmds.
> > > > 
> > > > Some functions like pmd_trans_huge_lock() and page_check_address_pmd()
> > > > are expected by their caller to pass back the pointer of ptl, so this
> > > > patch adds to those functions new arguments for that. Rather than that,
> > > > this patch gives only straightforward replacement.
> > > > 
> > > > ChangeLog v3:
> > > >  - fixed argument of huge_pmd_lockptr() in copy_huge_pmd()
> > > >  - added missing declaration of ptl in do_huge_pmd_anonymous_page()
> > > > 
> > > > Signed-off-by: Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx>
> > > 
> > > Generally, looks good. Few notes:
> > > 
> > > I believe you need to convert __pte_alloc() to new locking. Not sure about
> > > __pte_alloc_kernel().
> > > Have you check all rest mm->page_table_lock, that they shouldn't be
> > > converted to new locking?
> > 
> > I thought that keeping __pte_alloc() using mm->page_table_lock was safe
> > because it uses bare mm->page_table_lock instead of pte_lockptr() even
> > before this patchset, but not 100% sure.
> > __pte_alloc() (and its family) are used in normal page path, so if it's
> > not safe, we've lived with unsafe code for very long (maybe since 2005).
> > Anyway, converting __pte_alloc() into split ptl could improve performance
> > (though we need testing to know what amount), so I'll try that.
> 
> No, before the patch mm->page_table_lock is what we need: it serializes
> setting up pmd, not adding pte to pmd and it's subject to change with new
> locking model.

OK, I see. This is a todo for the next post.
Thank you for explanation.

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