From: Mike Rapoport rppt@linux.ibm.com
Hi,
@Andrew, this is based on v5.11-mmotm-2021-02-18-18-29 with the previous version reverted
Commit 73a6e474cb37 ("mm: memmap_init: iterate over memblock regions rather that check each PFN") exposed several issues with the memory map initialization and these patches fix those issues.
Initially there were crashes during compaction that Qian Cai reported back in April [1]. It seemed back then that the problem was fixed, but a few weeks ago Andrea Arcangeli hit the same bug [2] and there was an additional discussion at [3].
I didn't appreciate variety of ways BIOSes can report memory in the first megabyte, so previous versions of this set caused all kinds of troubles.
The last version that implicitly extended node/zone to cover the complete section might also have unexpected side effects, so this time I'm trying to move in forward in baby steps.
This is mostly a return to the fist version that simply merges init_unavailable_pages() into memmap_init() so that the only effective change would be more sensible zone/node links in unavailable struct pages.
For now, I've dropped the patch that tried to make ZONE_DMA to span pfn 0 because it didn't cause any issues for really long time and there are way to many hidden mines around this.
I have an ugly workaround for "pfn 0" issue that IMHO is the safest way to deal with it until it could be gradually fixed properly:
https://git.kernel.org/pub/scm/linux/kernel/git/rppt/linux.git/commit/?h=mem...
v7: * add handling of section end that span beyond the populated zones
v6: https://lore.kernel.org/lkml/20210222105728.28636-1-rppt@kernel.org * only interleave initialization of unavailable pages in memmap_init(), so that it is essentially includes init_unavailable_pages().
v5: https://lore.kernel.org/lkml/20210208110820.6269-1-rppt@kernel.org * extend node/zone spans to cover complete sections, this allows to interleave the initialization of unavailable pages with "normal" memory map init. * drop modifications to x86 early setup
v4: https://lore.kernel.org/lkml/20210130221035.4169-1-rppt@kernel.org/ * make sure pages in the range 0 - start_pfn_of_lowest_zone are initialized even if an architecture hides them from the generic mm * finally make pfn 0 on x86 to be a part of memory visible to the generic mm as reserved memory.
v3: https://lore.kernel.org/lkml/20210111194017.22696-1-rppt@kernel.org * use architectural zone constraints to set zone links for struct pages corresponding to the holes * drop implicit update of memblock.memory * add a patch that sets pfn 0 to E820_TYPE_RAM on x86
v2: https://lore.kernel.org/lkml/20201209214304.6812-1-rppt@kernel.org/): * added patch that adds all regions in memblock.reserved that do not overlap with memblock.memory to memblock.memory in the beginning of free_area_init()
[1] https://lore.kernel.org/lkml/8C537EB7-85EE-4DCF-943E-3CC0ED0DF56D@lca.pw [2] https://lore.kernel.org/lkml/20201121194506.13464-1-aarcange@redhat.com [3] https://lore.kernel.org/mm-commits/20201206005401.qKuAVgOXr%akpm@linux-found...
Mike Rapoport (1): mm/page_alloc.c: refactor initialization of struct page for holes in memory layout
mm/page_alloc.c | 147 +++++++++++++++++++++--------------------------- 1 file changed, 64 insertions(+), 83 deletions(-)