[PATCH] mm/hugetlb.c: Use __pa() instead of virt_to_phys() in early bootmem alloc code

[PATCH] mm/hugetlb.c: Use __pa() instead of virt_to_phys() in early bootmem alloc code

[Ritesh Harjani (IBM)]

Architecture like powerpc, checks for pfn_valid() in their
virt_to_phys() implementation (when CONFIG_DEBUG_VIRTUAL is enabled) [1].
Commit d49004c5f0c1 "arch, mm: consolidate initialization of nodes, zones and memory map"
changed the order of initialization between hugetlb_bootmem_alloc() and
free_area_init(). This means, pfn_valid() can now return false in
alloc_bootmem() path, since sparse_init() is not yet done.

Since, alloc_bootmem() uses memblock_alloc(.., MEMBLOCK_ALLOC_ACCESSIBLE), this
means these allocations are always going to happen below high_memory, where
__pa() should return valid physical addresses. Hence this patch converts
the two callers of virt_to_phys() in alloc_bootmem() path to __pa() to avoid
this bootup warning:

 ------------[ cut here ]------------
 WARNING: arch/powerpc/include/asm/io.h:879 at virt_to_phys+0x44/0x1b8, CPU#0: swapper/0
 Modules linked in:
 <...>
 NIP [c000000000601584] virt_to_phys+0x44/0x1b8
 LR [c000000004075de4] alloc_bootmem+0x144/0x1a8
 Call Trace:
 [c000000004d1fb50] [c000000004075dd4] alloc_bootmem+0x134/0x1a8
 [c000000004d1fba0] [c000000004075fac] __alloc_bootmem_huge_page+0x164/0x230
 [c000000004d1fbe0] [c000000004030bc4] alloc_bootmem_huge_page+0x44/0x138
 [c000000004d1fc10] [c000000004076e48] hugetlb_hstate_alloc_pages+0x350/0x5ac
 [c000000004d1fd30] [c0000000040782f0] hugetlb_bootmem_alloc+0x15c/0x19c
 [c000000004d1fd70] [c00000000406d7b4] mm_core_init_early+0x7c/0xdf4
 [c000000004d1ff30] [c000000004011d84] start_kernel+0xac/0xc58
 [c000000004d1ffe0] [c00000000000e99c] start_here_common+0x1c/0x20

[1]: https://lore.kernel.org/linuxppc-dev/87tsv5h544.ritesh.list@gmail.com/

Fixes: d49004c5f0c1 ("arch, mm: consolidate initialization of nodes, zones and memory map")
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
---
 mm/hugetlb.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 6e855a32de3d..43e0c95738a6 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -3101,7 +3101,7 @@ static __init void *alloc_bootmem(struct hstate *h, int nid, bool node_exact)
 			 * extract the actual node first.
 			 */
 			if (m)
-				listnode = early_pfn_to_nid(PHYS_PFN(virt_to_phys(m)));
+				listnode = early_pfn_to_nid(PHYS_PFN(__pa(m)));
 		}

 		if (m) {
@@ -3160,7 +3160,7 @@ int __alloc_bootmem_huge_page(struct hstate *h, int nid)
 	 * The head struct page is used to get folio information by the HugeTLB
 	 * subsystem like zone id and node id.
 	 */
-	memblock_reserved_mark_noinit(virt_to_phys((void *)m + PAGE_SIZE),
+	memblock_reserved_mark_noinit(__pa((void *)m + PAGE_SIZE),
 		huge_page_size(h) - PAGE_SIZE);

 	return 1;
--
2.53.0

Re: [PATCH] mm/hugetlb.c: Use __pa() instead of virt_to_phys() in early bootmem alloc code

[Mike Rapoport]

On Thu, Feb 26, 2026 at 05:56:30PM +0530, Ritesh Harjani (IBM) wrote:
> Architecture like powerpc, checks for pfn_valid() in their
> virt_to_phys() implementation (when CONFIG_DEBUG_VIRTUAL is enabled) [1].
> Commit d49004c5f0c1 "arch, mm: consolidate initialization of nodes, zones and memory map"
> changed the order of initialization between hugetlb_bootmem_alloc() and
> free_area_init(). This means, pfn_valid() can now return false in
> alloc_bootmem() path, since sparse_init() is not yet done.
> 
> Since, alloc_bootmem() uses memblock_alloc(.., MEMBLOCK_ALLOC_ACCESSIBLE), this
> means these allocations are always going to happen below high_memory, where
> __pa() should return valid physical addresses. Hence this patch converts
> the two callers of virt_to_phys() in alloc_bootmem() path to __pa() to avoid
> this bootup warning:
> 
>  ------------[ cut here ]------------
>  WARNING: arch/powerpc/include/asm/io.h:879 at virt_to_phys+0x44/0x1b8, CPU#0: swapper/0
>  Modules linked in:
>  <...>
>  NIP [c000000000601584] virt_to_phys+0x44/0x1b8
>  LR [c000000004075de4] alloc_bootmem+0x144/0x1a8
>  Call Trace:
>  [c000000004d1fb50] [c000000004075dd4] alloc_bootmem+0x134/0x1a8
>  [c000000004d1fba0] [c000000004075fac] __alloc_bootmem_huge_page+0x164/0x230
>  [c000000004d1fbe0] [c000000004030bc4] alloc_bootmem_huge_page+0x44/0x138
>  [c000000004d1fc10] [c000000004076e48] hugetlb_hstate_alloc_pages+0x350/0x5ac
>  [c000000004d1fd30] [c0000000040782f0] hugetlb_bootmem_alloc+0x15c/0x19c
>  [c000000004d1fd70] [c00000000406d7b4] mm_core_init_early+0x7c/0xdf4
>  [c000000004d1ff30] [c000000004011d84] start_kernel+0xac/0xc58
>  [c000000004d1ffe0] [c00000000000e99c] start_here_common+0x1c/0x20
> 
> [1]: https://lore.kernel.org/linuxppc-dev/87tsv5h544.ritesh.list@gmail.com/
> 
> Fixes: d49004c5f0c1 ("arch, mm: consolidate initialization of nodes, zones and memory map")
> Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>

Reviewed-by: Mike Rapoport (Microsoft) <rppt@kernel.org>

> ---
>  mm/hugetlb.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 6e855a32de3d..43e0c95738a6 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -3101,7 +3101,7 @@ static __init void *alloc_bootmem(struct hstate *h, int nid, bool node_exact)
>  			 * extract the actual node first.
>  			 */
>  			if (m)
> -				listnode = early_pfn_to_nid(PHYS_PFN(virt_to_phys(m)));
> +				listnode = early_pfn_to_nid(PHYS_PFN(__pa(m)));
>  		}
> 
>  		if (m) {
> @@ -3160,7 +3160,7 @@ int __alloc_bootmem_huge_page(struct hstate *h, int nid)
>  	 * The head struct page is used to get folio information by the HugeTLB
>  	 * subsystem like zone id and node id.
>  	 */
> -	memblock_reserved_mark_noinit(virt_to_phys((void *)m + PAGE_SIZE),
> +	memblock_reserved_mark_noinit(__pa((void *)m + PAGE_SIZE),
>  		huge_page_size(h) - PAGE_SIZE);
> 
>  	return 1;
> --
> 2.53.0
> 

-- 
Sincerely yours,
Mike.