Re: [PATCH v5 8/9] mm/memory-failure: add hwpoison_boot_page() to flag an inherited frame

[Breno Leitao]

On Sat, Sep 19, 2026 at 03:58:25PM +0530, Shaikh Kamaluddin wrote:
> On Tue, Sep 15, 2026 at 05:53:42AM -0700, Breno Leitao wrote:
> > +
> > +	/*
> > +	 * The per memory block half of num_poisoned_pages_inc() has no block to
> > +	 * find at boot, and divides by zero looking for one. A hotplugged block
> > +	 * is already there.
> > +	 */
> > +	if (context == MEMINIT_HOTPLUG)
> > +		num_poisoned_pages_inc(pfn);
> > +	else
> > +		atomic_long_inc(&num_poisoned_pages);
> > +}
> 
> Hi Breno,
> 
> num_poisoned_pages_inc() currently assumes that the supplied PFN can
> be used for per-memory-block accounting and therefore calls
> memblk_nr_poison_inc() unconditionally. The early-boot path needs
> global-only accounting because the memory-block infrastructure is not
> initialized yet.
> 
> Could num_poisoned_pages_inc() treat -1UL as global-only accounting,
> matching num_poisoned_pages_sub()?
> 
> Example as below:
> 
> void num_poisoned_pages_inc(unsigned long pfn)
> {
> 	atomic_long_inc(&num_poisoned_pages);
> 
> 	if (pfn != -1UL)
> 		memblk_nr_poison_inc(pfn);
> }
> 
> 
> The caller could then use:
> 
> num_poisoned_pages_inc(context == MEMINIT_HOTPLUG ? pfn : -1UL);
> 
> This would keep updates to `num_poisoned_pages` encapsulated rather than
> manipulating the counter directly here, while also making the increment
> and decrement interfaces consistent.

Good point, and it is a better fit than what I have.

-1UL already means "global only" on the sub side, and it is not just a
convention on paper: remove_memory_block_devices() calls
num_poisoned_pages_sub(-1UL, memblk_nr_poison(mem)) when a block goes
away. Teaching inc() the same thing costs a branch and keeps
num_poisoned_pages behind the two helpers instead of growing a third
place that pokes the atomic directly.

I will fold this into v6:

      void num_poisoned_pages_inc(unsigned long pfn)
      {
              atomic_long_inc(&num_poisoned_pages);

              if (pfn != -1UL)
                      memblk_nr_poison_inc(pfn);
      }

Thanks for the review and suggestion,
--breno