Re: linux-next: build failure after merge of the mm tree

[Stephen Rothwell]

Hi all,

On Mon, 2 May 2022 12:19:36 +0100 Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
>
> On Mon, May 02, 2022 at 08:49:03PM +1000, Stephen Rothwell wrote:
> > 
> > After merging the mm tree, today's linux-next build (arm
> > multi_v7_defconfig) failed like this:  
> [... i wish our BUILD_BUGs produced nicer output from the compiler ...]
> > Reverting the following commits makes the problem go away:
> > 
> >   2b58b3f33ba2 ("mm/shmem: convert shmem_swapin_page() to shmem_swapin_folio()")
> >   94cdf3e8c0bf ("mm/shmem: convert shmem_getpage_gfp to use a folio")
> >   3674fd6cadf5 ("mm/shmem: convert shmem_alloc_and_acct_page to use a folio")
> >   b0bb08b2d5f3 ("mm/shmem: turn shmem_alloc_page() into shmem_alloc_folio()")
> >   8d657a77c6fe ("mm/shmem: turn shmem_should_replace_page into shmem_should_replace_folio")
> >   9a44f3462edc ("mm/shmem: convert shmem_add_to_page_cache to take a folio")
> >   561fd8bee1dc ("mm/swap: add folio_throttle_swaprate")
> >   cb4e56ee240d ("mm/shmem: use a folio in shmem_unused_huge_shrink")
> >   22bf1b68e572 ("vmscan: remove remaining uses of page in shrink_page_list")
> >   7d15d41b7c4a ("vmscan: convert the activate_locked portion of shrink_page_list to folios")
> >   8a6aff9c51c7 ("vmscan: move initialisation of mapping down")
> >   b79338b3d217 ("vmscan: convert lazy freeing to folios")
> >   719426e40146 ("vmscan: convert page buffer handling to use folios")
> >   339ba7502e13 ("vmscan: convert dirty page handling to folios")  
> 
> Oops.  allnoconfig on x86 reproduces the problem.  This fixes it;
> happy to go back and produce a new set of patches for Andrew to
> preserve bisectability.
> 
> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> index 2999190adc22..e9e0d591061d 100644
> --- a/include/linux/huge_mm.h
> +++ b/include/linux/huge_mm.h
> @@ -347,7 +347,6 @@ static inline void prep_transhuge_page(struct page *page) {}
>  static inline bool
>  can_split_folio(struct folio *folio, int *pextra_pins)
>  {
> -	BUILD_BUG();
>  	return false;
>  }
>  static inline int
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 673a0e783496..d62936ffe74d 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -738,7 +738,7 @@ static int shmem_add_to_page_cache(struct folio *folio,
>  		xas_store(&xas, folio);
>  		if (xas_error(&xas))
>  			goto unlock;
> -		if (folio_test_large(folio)) {
> +		if (folio_test_pmd_mappable(folio)) {
>  			count_vm_event(THP_FILE_ALLOC);
>  			__lruvec_stat_mod_folio(folio, NR_SHMEM_THPS, nr);
>  		}
> @@ -1887,10 +1887,7 @@ static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
>  		goto unlock;
>  	}
>  
> -	if (folio_test_large(folio))
> -		hindex = round_down(index, HPAGE_PMD_NR);
> -	else
> -		hindex = index;
> +	hindex = round_down(index, folio_nr_pages(folio));
>  
>  	if (sgp == SGP_WRITE)
>  		__folio_set_referenced(folio);
> @@ -1909,7 +1906,7 @@ static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
>  	spin_unlock_irq(&info->lock);
>  	alloced = true;
>  
> -	if (folio_test_large(folio) &&
> +	if (folio_test_pmd_mappable(folio) &&
>  	    DIV_ROUND_UP(i_size_read(inode), PAGE_SIZE) <
>  			hindex + HPAGE_PMD_NR - 1) {
>  		/*
> 

I applied the above patch to the mm tree merge today (instead of all
the reverts) and it fixed the build problems for me.

-- 
Cheers,
Stephen Rothwell

Attachment: pgpk91Amp2SFL.pgp
Description: OpenPGP digital signature