Re: [PATCH v9 08/69] mm: start tracking VMAs with maple tree

[Liam Howlett]

* Liam R. Howlett <Liam.Howlett@xxxxxxxxxx> [220510 08:50]:
> * SeongJae Park <sj@xxxxxxxxxx> [220510 03:37]:
> > Hi Liam,
> > 
> > On Wed, 4 May 2022 01:07:52 +0000 Liam Howlett <liam.howlett@xxxxxxxxxx> wrote:
> > 
> > > From: "Liam R. Howlett" <Liam.Howlett@xxxxxxxxxx>
> > > 
> > > Start tracking the VMAs with the new maple tree structure in parallel with
> > > the rb_tree.  Add debug and trace events for maple tree operations and
> > > duplicate the rb_tree that is created on forks into the maple tree.
> > > 
> > > The maple tree is added to the mm_struct including the mm_init struct,
> > > added support in required mm/mmap functions, added tracking in kernel/fork
> > > for process forking, and used to find the unmapped_area and checked
> > > against what the rbtree finds.
> > > 
> > > This also moves the mmap_lock() in exit_mmap() since the oom reaper call
> > > does walk the VMAs.  Otherwise lockdep will be unhappy if oom happens.
> > > 
> > > Signed-off-by: Liam R. Howlett <Liam.Howlett@xxxxxxxxxx>
> > > Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx>
> > > ---
> > >  arch/x86/kernel/tboot.c     |   1 +
> > >  drivers/firmware/efi/efi.c  |   1 +
> > >  include/linux/mm.h          |   2 +
> > >  include/linux/mm_types.h    |   3 +
> > >  include/trace/events/mmap.h |  73 ++++++++
> > >  kernel/fork.c               |  20 ++-
> > >  mm/init-mm.c                |   2 +
> > >  mm/mmap.c                   | 323 +++++++++++++++++++++++++++++++++---
> > >  8 files changed, 396 insertions(+), 29 deletions(-)
> > > 
> > [...]
> > > diff --git a/mm/mmap.c b/mm/mmap.c
> > > index d7e120ad5825..e777da0132f6 100644
> > > --- a/mm/mmap.c
> > > +++ b/mm/mmap.c
> > [...]
> > > @@ -680,6 +744,56 @@ static void __vma_link_file(struct vm_area_struct *vma)
> > >  	}
> > >  }
> > >  
> > > +/*
> > > + * vma_mas_store() - Store a VMA in the maple tree.
> > > + * @vma: The vm_area_struct
> > > + * @mas: The maple state
> > > + *
> > > + * Efficient way to store a VMA in the maple tree when the @mas has already
> > > + * walked to the correct location.
> > > + *
> > > + * Note: the end address is inclusive in the maple tree.
> > > + */
> > > +inline void vma_mas_store(struct vm_area_struct *vma, struct ma_state *mas)
> > > +{
> > > +	trace_vma_store(mas->tree, vma);
> > > +	mas_set_range(mas, vma->vm_start, vma->vm_end - 1);
> > > +	mas_store_prealloc(mas, vma);
> > > +}
> > > +
> > > +/*
> > > + * vma_mas_remove() - Remove a VMA from the maple tree.
> > > + * @vma: The vm_area_struct
> > > + * @mas: The maple state
> > > + *
> > > + * Efficient way to remove a VMA from the maple tree when the @mas has already
> > > + * been established and points to the correct location.
> > > + * Note: the end address is inclusive in the maple tree.
> > > + */
> > > +static inline void vma_mas_remove(struct vm_area_struct *vma, struct ma_state *mas)
> > > +{
> > > +	trace_vma_mas_szero(mas->tree, vma->vm_start, vma->vm_end - 1);
> > > +	mas->index = vma->vm_start;
> > > +	mas->last = vma->vm_end - 1;
> > > +	mas_store_prealloc(mas, NULL);
> > > +}
> > 
> > Above two functions were defined in internal.h in v8[1], but moved to mmap.c in
> > this version.  As mmap.c is compiled when CONFIG_MMU, build fails when the
> > config is not set as below:
> > 
> >     .../mm/nommu.c: In function 'add_vma_to_mm':
> >     .../mm/nommu.c:575:2: error: implicit declaration of function 'vma_mas_store'; did you mean 'mas_store'? [-Werror=implicit-function-declaration]
> >       575 |  vma_mas_store(vma, &mas);
> >           |  ^~~~~~~~~~~~~
> >           |  mas_store
> >     .../mm/nommu.c: In function 'delete_vma_from_mm':
> >     .../mm/nommu.c:599:2: error: implicit declaration of function 'vma_mas_remove' [-Werror=implicit-function-declaration]
> >       599 |  vma_mas_remove(vma, &mas);
> >           |  ^~~~~~~~~~~~~~
> > 
> > Was there a reason to move these definitions?  I confirmed moving the
> > definition back to internal.h fixes this issue, but I'm unsure if that's the
> > right solution.
> > 
> > [1] https://lore.kernel.org/linux-mm/20220426150616.3937571-10-Liam.Howlett@xxxxxxxxxx/
> > 
> 
> I moved the functions as they were dropped elsewhere for cleanup.  I did
> check my m68k buildroot, but the buildroot seems to have not pulled the
> new code - it clones the repo externally.  I just changed the branch to
> use mm-unstable and that is indeed failing as you see above.  Your fix
> is correct - and I believe it also means I need to leave the damon
> change the way it is now.
> 

I recall now that I wanted to have the function traced.  Moving it to
the c file allowed me to add tracepoints pretty easily.  Having it in
the header does not work as it will redefine the tracepoints many times
and cause a bit of a mess.  I don't love putting the tracepoint as is in
another function but I think I will have to move to this method.  On the
plus side, the open-coded modifications to the maple tree may be able to
use the tracepoints then.