Re: [PATCH v4 2/4] memblock: update initialization of reserved pages

[David Woodhouse <dwmw2@infradead.org>]

[-- Attachment #1: Type: text/plain, Size: 3896 bytes --]

On Tue, 2025-04-01 at 14:33 +0300, Mike Rapoport wrote:
> On Mon, Mar 31, 2025 at 04:13:56PM +0100, David Woodhouse wrote:
> > On Mon, 2025-03-31 at 17:50 +0300, Mike Rapoport wrote:
> > > On Mon, Mar 31, 2025 at 01:50:33PM +0100, David Woodhouse wrote:
> > > > On Tue, 2021-05-11 at 13:05 +0300, Mike Rapoport wrote:
> > > > 
> > > > On platforms with large NOMAP regions (e.g. which are actually reserved
> > > > for guest memory to keep it out of the Linux address map and allow for
> > > > kexec-based live update of the hypervisor), this pointless loop ends up
> > > > taking a significant amount of time which is visible as guest steal
> > > > time during the live update.
> > > > 
> > > > Can reserve_bootmem_region() skip the loop *completely* if no PFN in
> > > > the range from start to end is valid? Or tweak the loop itself to have
> > > > an 'else' case which skips to the next valid PFN? Something like
> > > > 
> > > >  for(...) {
> > > >     if (pfn_valid(start_pfn)) {
> > > >        ...
> > > >     } else {
> > > >        start_pfn = next_valid_pfn(start_pfn);
> > > >     }
> > > >  }
> > > 
> > > My understanding is that you have large reserved NOMAP ranges that don't
> > > appear as memory at all, so no memory map for them is created and so
> > > pfn_valid() is false for pfns in those ranges.
> > > 
> > > If this is the case one way indeed would be to make
> > > reserve_bootmem_region() skip ranges with no valid pfns.
> > > 
> > > Another way could be to memblock_reserved_mark_noinit() such ranges and
> > > then reserve_bootmem_region() won't even get called, but that would require
> > > firmware to pass that information somehow.
> > 
> > I was thinking along these lines (not even build tested)...
> > 
> > I don't much like the (unsigned long)-1 part. I might make the helper
> > 'static inline bool first_valid_pfn (unsigned long *pfn)' and return
> > success or failure. But that's an implementation detail.
> > 
> > index 6d1fb6162ac1..edd27ba3e908 100644
> > --- a/include/asm-generic/memory_model.h
> > +++ b/include/asm-generic/memory_model.h
> > @@ -29,8 +29,43 @@ static inline int pfn_valid(unsigned long pfn)
> >         return pfn >= pfn_offset && (pfn - pfn_offset) < max_mapnr;
> >  }
> >  #define pfn_valid pfn_valid
> > +
> > +static inline unsigned long first_valid_pfn(unsigned long pfn)
> > +{
> > +       /* avoid <linux/mm.h> include hell */
> > +       extern unsigned long max_mapnr;
> > +       unsigned long pfn_offset = ARCH_PFN_OFFSET;
> > +
> > +       if (pfn < pfn_offset)
> > +               return pfn_offset;
> > +
> > +       if ((pfn - pfn_offset) < max_mapnr)
> > +               return pfn;
> > +
> > +       return (unsigned long)(-1);
> > +}
> 
> This seems about right for FLATMEM. For SPARSEMEM it would be something
> along these lines (I kept dubious -1):

Thanks. Is that right even with CONFIG_SPARSEMEM_VMEMMAP? It seems that
it's possible for pfn_valid() to be false for a given *page*, but there
may still be valid pages in the remainder of the same section in that
case? 

I think it should only skip to the next section if the current section
doesn't exist at all, not just when pfn_section_valid() return false?

I also wasn't sure how to cope with the rcu_read_lock_sched() that
happens in pfn_valid(). What's that protecting against? Does it mean
that by the time pfn_valid() returns true, that might not be the
correct answer any more?



> static inline unsigned long first_valid_pfn(unsigned long pfn)
> {
> 	unsigned long nr = pfn_to_section_nr(pfn);
> 
> 	do {
> 		if (pfn_valid(pfn))
> 			return pfn;
> 		pfn = section_nr_to_pfn(nr++);
> 	} while (nr < NR_MEM_SECTIONS);
> 
> 	return (unsigned long)-1;
> }
> 


[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 5069 bytes --]