mm: memblock: enforce overlap of memory.memblock and memory.reserved

Patch series "mm: fix initialization of struct page for holes in  memory layout", v2.

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 probelm was fixed, but a few
weeks ago Andrea Arcangeli hit the same bug [2] and after a long
discussion between us [3] I think these patches are the proper fix.

[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-foundation.org

This patch (of 2):

memblock does not require that the reserved memory ranges will be a subset
of memblock.memory.

As a result there may be reserved pages that are not in the range of any
zone or node because zone and node boundaries are detected based on
memblock.memory and pages that only present in memblock.reserved are not
taken into account during zone/node size detection.

Make sure that all ranges in memblock.reserved are added to
memblock.memory before calculating node and zone boundaries.

Link: https://lkml.kernel.org/r/20201209214304.6812-1-rppt@kernel.org
Link: https://lkml.kernel.org/r/20201209214304.6812-2-rppt@kernel.org
Fixes: 73a6e474cb37 ("mm: memmap_init: iterate over memblock regions rather that check each PFN")
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reported-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Qian Cai <cai@lca.pw>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

diff --git a/include/linux/memblock.h b/include/linux/memblock.h
index b93c44b9121ec050b19dd5a3e9b77976374fe4b9..9c5cc95c7cee906d83cbe8c57e1b7b43f7c63508 100644 (file)
--- a/include/linux/memblock.h
+++ b/include/linux/memblock.h
@@ -120,6 +120,7 @@ int memblock_clear_nomap(phys_addr_t base, phys_addr_t size);
 unsigned long memblock_free_all(void);
 void reset_node_managed_pages(pg_data_t *pgdat);
 void reset_all_zones_managed_pages(void);
+void memblock_enforce_memory_reserved_overlap(void);
 
 /* Low level functions */
 void __next_mem_range(u64 *idx, int nid, enum memblock_flags flags,

diff --git a/mm/memblock.c b/mm/memblock.c
index d24bcfa88d2f1bd42d96f4e52a6f3e8c2cd91396..b8b7be0561c462eebdcea8b59274df32bf4ddfe5 100644 (file)
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -1860,6 +1860,30 @@ void __init_memblock memblock_trim_memory(phys_addr_t align)
        }
 }
 
+/**
+ * memblock_enforce_memory_reserved_overlap - make sure every range in
+ * @memblock.reserved is covered by @memblock.memory
+ *
+ * The data in @memblock.memory is used to detect zone and node boundaries
+ * during initialization of the memory map and the page allocator. Make
+ * sure that every memory range present in @memblock.reserved is also added
+ * to @memblock.memory even if the architecture specific memory
+ * initialization failed to do so
+ */
+void __init memblock_enforce_memory_reserved_overlap(void)
+{
+       phys_addr_t start, end;
+       int nid;
+       u64 i;
+
+       __for_each_mem_range(i, &memblock.reserved, &memblock.memory,
+                            NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, &nid) {
+               pr_warn("memblock: reserved range [%pa-%pa] is not in memory\n",
+                       &start, &end);
+               memblock_add_node(start, (end - start), nid);
+       }
+}
+
 void __init_memblock memblock_set_current_limit(phys_addr_t limit)
 {
        memblock.current_limit = limit;

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 23ceaf8fdf553287590cc3feb2ccddf11f06e9a8..85f27bf2bbefcf2cabb3506a84415b543e4e784f 100644 (file)
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -7513,6 +7513,13 @@ void __init free_area_init(unsigned long *max_zone_pfn)
        memset(arch_zone_highest_possible_pfn, 0,
                                sizeof(arch_zone_highest_possible_pfn));
 
+       /*
+        * Some architectures (e.g. x86) have reserved pages outside of
+        * memblock.memory. Make sure these pages are taken into account
+        * when detecting zone and node boundaries
+        */
+       memblock_enforce_memory_reserved_overlap();
+
        start_pfn = find_min_pfn_with_active_regions();
        descending = arch_has_descending_max_zone_pfns();