Re: [PATCH v3 1/2] mm: protect free_pgtables with mmap_lock write lock in exit_mmap

[Suren Baghdasaryan]

On Tue, Dec 7, 2021 at 2:47 PM Suren Baghdasaryan <surenb@xxxxxxxxxx> wrote:
>
> On Tue, Dec 7, 2021 at 2:07 PM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
> >
> > On Tue, Dec 07, 2021 at 01:50:30PM -0800, Suren Baghdasaryan wrote:
> > > - Added a comment for vm_operations_struct::close, documenting restriction for
> > > taking mmap_lock in the callback, per Michal Hocko and Matthew Wilcox
> >
> > This should be a separate patch because it stands alone, but ...
>
> I thought about it and since it was relevant to the change in
> remove_vma locking, I thought it would fit here. However, if you
> insist on splitting it, I'll post it as a separate patch. Please let
> me know.
>
> >
> > >  struct vm_operations_struct {
> > >       void (*open)(struct vm_area_struct * area);
> > > +    /*
> > > +     * Called with mmap_lock lock held for write from __split_vma and
> > > +     * remove_vma, therefore should never take that lock.
> > > +     */
> >
> > Your whitespace indentation is weird.  And it'd be nice to make this a
> > kernel-doc comment (I know none of the others are, but that should be
> > fixed too).  And naming the functions that call it is wrong too.
> >
> >         /**
> >          * @close: Called when the VMA is being removed from the MM.
> >          * Context: Caller holds mmap_lock.

BTW, is the caller always required to hold mmap_lock for write or it
*might* hold it?

> >          */
>
> Ack. Will change and include in the next respin once I hear from you
> on the preference for a separate patch.
> Thanks!