Re: [PATCH] mm: consolidate pgtable_cache_init() and pgd_cache_init()

[Mike Rapoport]

On Wed, Aug 21, 2019 at 04:49:42PM +0100, Will Deacon wrote:
> On Wed, Aug 21, 2019 at 06:06:58PM +0300, Mike Rapoport wrote:
> > Both pgtable_cache_init() and pgd_cache_init() are used to initialize kmem
> > cache for page table allocations on several architectures that do not use
> > PAGE_SIZE tables for one or more levels of the page table hierarchy.
> > 
> > Most architectures do not implement these functions and use __week default
> > NOP implementation of pgd_cache_init(). Since there is no such default for
> > pgtable_cache_init(), its empty stub is duplicated among most
> > architectures.
> > 
> > Rename the definitions of pgd_cache_init() to pgtable_cache_init() and drop
> > empty stubs of pgtable_cache_init().
> > 
> > Signed-off-by: Mike Rapoport <rppt@xxxxxxxxxxxxx>
> > ---
> 
> [...]
> 
> > diff --git a/arch/arm64/mm/pgd.c b/arch/arm64/mm/pgd.c
> > index 7548f9c..4a64089 100644
> > --- a/arch/arm64/mm/pgd.c
> > +++ b/arch/arm64/mm/pgd.c
> > @@ -35,7 +35,7 @@ void pgd_free(struct mm_struct *mm, pgd_t *pgd)
> >  		kmem_cache_free(pgd_cache, pgd);
> >  }
> >  
> > -void __init pgd_cache_init(void)
> > +void __init pgtable_cache_init(void)
> >  {
> >  	if (PGD_SIZE == PAGE_SIZE)
> >  		return;
> 
> [...]
> 
> > diff --git a/init/main.c b/init/main.c
> > index b90cb5f..2fa8038 100644
> > --- a/init/main.c
> > +++ b/init/main.c
> > @@ -507,7 +507,7 @@ void __init __weak mem_encrypt_init(void) { }
> >  
> >  void __init __weak poking_init(void) { }
> >  
> > -void __init __weak pgd_cache_init(void) { }
> > +void __init __weak pgtable_cache_init(void) { }
> >  
> >  bool initcall_debug;
> >  core_param(initcall_debug, initcall_debug, bool, 0644);
> > @@ -565,7 +565,6 @@ static void __init mm_init(void)
> >  	init_espfix_bsp();
> >  	/* Should be run after espfix64 is set up. */
> >  	pti_init();
> > -	pgd_cache_init();
> >  }
> 
> AFAICT, this change means we now initialise our pgd cache before
> debug_objects_mem_init() has run.

Right.

> Is that going to cause fireworks with CONFIG_DEBUG_OBJECTS when we later
> free a pgd?

We don't allocate a pgd at that time, we only create the kmem cache for the
future allocations. And that cache is never destroyed anyway.

> Will

-- 
Sincerely yours,
Mike.