On Sat, Sep 28, 2013 at 01:24:51AM +0300, Kirill A. Shutemov wrote:
> Cody P Schafer wrote:
> > On 09/27/2013 06:16 AM, Kirill A. Shutemov wrote:
> > > With split page table lock for PMD level we can't hold
> > > mm->page_table_lock while updating nr_ptes.
> > >
> > > Let's convert it to atomic_t to avoid races.
> > >
> >
> > > ---
> >
> > > diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> > > index 84e0c56e1e..99f19e850d 100644
> > > --- a/include/linux/mm_types.h
> > > +++ b/include/linux/mm_types.h
> > > @@ -339,6 +339,7 @@ struct mm_struct {
> > > pgd_t * pgd;
> > > atomic_t mm_users; /* How many users with
> > > user space? */
> > > atomic_t mm_count; /* How many references
> > > to "struct mm_struct" (users count as 1) */
> > > + atomic_t nr_ptes; /* Page table pages */
> > > int map_count; /* number of VMAs */
> > >
> > > spinlock_t page_table_lock; /* Protects page tables
> > > and some counters */
> > > @@ -360,7 +361,6 @@ struct mm_struct {
> > > unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE */
> > > unsigned long stack_vm; /* VM_GROWSUP/DOWN */
> > > unsigned long def_flags;
> > > - unsigned long nr_ptes; /* Page table pages */
> > > unsigned long start_code, end_code, start_data, end_data;
> > > unsigned long start_brk, brk, start_stack;
> > > unsigned long arg_start, arg_end, env_start, env_end;
> >
> > Will 32bits always be enough here? Should atomic_long_t be used instead?
>
> Good question!
>
> On x86_64 we need one table to cover 2M (512 entries by 4k, 21 bits) of
> virtual address space. Total size of virtual memory which can be covered
> by 31-bit (32 - sign) nr_ptes is 52 bits (31 + 21).
>
> Currently, on x86_64 with 4-level page tables we can use at most 48 bit of
> virtual address space (only half of it available for userspace), so we
> pretty safe here.
>
> Although, it can be a potential problem, if (when) x86_64 will implement
> 5-level page tables -- 57-bits of virtual address space.
>
> Any thoughts?
I'd just go with atomic_long_t to avoid having to worry about this in
the first place. It's been ulong forever and I'm not aware of struct
mm_struct size being an urgent issue. Cutting this type in half and
adding overflow checks adds more problems than it solves.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
