Re: [PATCH v4 02/14] memblock: add MEMBLOCK_RSRV_KERN flag

[Wei Yang]

On Tue, Mar 11, 2025 at 07:27:23AM +0200, Mike Rapoport wrote:
>Hi Wei,
>
>On Mon, Mar 10, 2025 at 09:51:24AM +0000, Wei Yang wrote:
>> On Sun, Feb 23, 2025 at 12:22:29AM +0000, Wei Yang wrote:
>> >On Wed, Feb 19, 2025 at 09:24:31AM +0200, Mike Rapoport wrote:
>> >>Hi,
>> >>
>> >>On Tue, Feb 18, 2025 at 03:50:04PM +0000, Wei Yang wrote:
>> >>> On Thu, Feb 06, 2025 at 03:27:42PM +0200, Mike Rapoport wrote:
>> >>> >From: "Mike Rapoport (Microsoft)" <rppt@xxxxxxxxxx>
>> >>> >
>> >>> >to denote areas that were reserved for kernel use either directly with
>> >>> >memblock_reserve_kern() or via memblock allocations.
>> >>> >
>> >>> >Signed-off-by: Mike Rapoport (Microsoft) <rppt@xxxxxxxxxx>
>> >>> >---
>> >>> > include/linux/memblock.h | 16 +++++++++++++++-
>> >>> > mm/memblock.c            | 32 ++++++++++++++++++++++++--------
>> >>> > 2 files changed, 39 insertions(+), 9 deletions(-)
>> >>> >
>> >>> >diff --git a/include/linux/memblock.h b/include/linux/memblock.h
>> >>> >index e79eb6ac516f..65e274550f5d 100644
>> >>> >--- a/include/linux/memblock.h
>> >>> >+++ b/include/linux/memblock.h
>> >>> >@@ -50,6 +50,7 @@ enum memblock_flags {
>> >>> > 	MEMBLOCK_NOMAP		= 0x4,	/* don't add to kernel direct mapping */
>> >>> > 	MEMBLOCK_DRIVER_MANAGED = 0x8,	/* always detected via a driver */
>> >>> > 	MEMBLOCK_RSRV_NOINIT	= 0x10,	/* don't initialize struct pages */
>> >>> >+	MEMBLOCK_RSRV_KERN	= 0x20,	/* memory reserved for kernel use */
>> >>> 
>> >>> Above memblock_flags, there are comments on explaining those flags.
>> >>> 
>> >>> Seems we miss it for MEMBLOCK_RSRV_KERN.
>> >>
>> >>Right, thanks!
>> >> 
>> >>> > 
>> >>> > #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
>> >>> >@@ -1459,14 +1460,14 @@ phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
>> >>> > again:
>> >>> > 	found = memblock_find_in_range_node(size, align, start, end, nid,
>> >>> > 					    flags);
>> >>> >-	if (found && !memblock_reserve(found, size))
>> >>> >+	if (found && !__memblock_reserve(found, size, nid, MEMBLOCK_RSRV_KERN))
>> >>> 
>> >>> Maybe we could use memblock_reserve_kern() directly. If my understanding is
>> >>> correct, the reserved region's nid is not used.
>> >>
>> >>We use nid of reserved regions in reserve_bootmem_region() (commit
>> >>61167ad5fecd ("mm: pass nid to reserve_bootmem_region()")) but KHO needs to
>> >>know the distribution of reserved memory among the nodes before
>> >>memmap_init_reserved_pages().
>> >> 
>> >
>> >I took another look into this commit. There maybe a very corner case in which
>> >will leave a reserved region with no nid set.
>> >
>> >memmap_init_reserved_pages()
>> >    for_each_mem_region() {
>> >        ...
>> >	memblock_set_node(start, end, &memblock.reserved, nid);
>> >    }
>> >
>> >We leverage the iteration here to set nid to all regions in memblock.reserved.
>> >But memblock_set_node() may call memblock_double_array() to expand the array,
>> >which may get a range before current start. So we would miss to set the
>> >correct nid to the new reserved region.
>> >
>> >I have tried to create a case in memblock test. This would happen when there
>> >are 126 memblock.reserved regions. And the last region is across the last two
>> >node.
>> >
>> >One way to fix this is compare type->max in memblock_set_node(). Then check
>> >this return value in memmap_init_reserved_pages(). If we found the size
>> >changes, repeat the iteration.
>> >
>> >But this is a very trivial one, not sure it worth fix.
>> >
>> 
>> Hi, Mike
>> 
>> I have done a user space test which shows we may have a chance to leave a
>> region with non-nid set.
>> 
>> Not sure you are ok with my approach of fixing.
>
>Wouldn't it be better to check for a change in reserved.max in
>memmap_init_reserved_pages()?
> 

Sounds better.

Previously I thought we need to hide detail from user, but actually it is
already in memblock.c :-)

If you agree, I would like to prepare a fix.

>> -- 
>> Wei Yang
>> Help you, Help me
>
>-- 
>Sincerely yours,
>Mike.

-- 
Wei Yang
Help you, Help me