Re: [PATCH v1 1/4] mm: handle poisoning of pfn without struct pages

[Jason Gunthorpe]

On Sat, Sep 23, 2023 at 11:20:19AM +0800, Miaohe Lin wrote:

> >  /**
> >   * memory_failure - Handle memory failure of a page.
> >   * @pfn: Page Number of the corrupted page
> > @@ -2183,6 +2271,11 @@ int memory_failure(unsigned long pfn, int flags)
> >  	if (!(flags & MF_SW_SIMULATED))
> >  		hw_memory_failure = true;
> >  
> > +	if (!pfn_valid(pfn) && !arch_is_platform_page(PFN_PHYS(pfn))) {
> 
> Could it be better to add a helper here to detect the pfns without
> struct page?

pfn_valid is supposed to do that.

This arch_is_platform_page stuff is actually detecting Intel SGX
memory and routing it to arch_memory_failure()

It would have been more accurately named
arch_is_arch_memory_failure_pfn() or something

Actually that SGX stuff could probably be changed over to use the
interval tree of this series. Modify sgx_setup_epc_section() to
register tree nodes per-section and remove all this arch stuff
entirely.

Jason