]> Git Repo - linux.git/commit
memblock: make memblock_set_node() also warn about use of MAX_NUMNODES
authorJan Beulich <[email protected]>
Wed, 29 May 2024 07:39:10 +0000 (09:39 +0200)
committerMike Rapoport (IBM) <[email protected]>
Fri, 31 May 2024 09:36:28 +0000 (12:36 +0300)
commite0eec24e2e199873f43df99ec39773ad3af2bff7
tree5dacdd27399136cc033a23e039a46c1846cf8c9d
parent1613e604df0cd359cf2a7fbd9be7a0bcfacfabd0
memblock: make memblock_set_node() also warn about use of MAX_NUMNODES

On an (old) x86 system with SRAT just covering space above 4Gb:

    ACPI: SRAT: Node 0 PXM 0 [mem 0x100000000-0xfffffffff] hotplug

the commit referenced below leads to this NUMA configuration no longer
being refused by a CONFIG_NUMA=y kernel (previously

    NUMA: nodes only cover 6144MB of your 8185MB e820 RAM. Not used.
    No NUMA configuration found
    Faking a node at [mem 0x0000000000000000-0x000000027fffffff]

was seen in the log directly after the message quoted above), because of
memblock_validate_numa_coverage() checking for NUMA_NO_NODE (only). This
in turn led to memblock_alloc_range_nid()'s warning about MAX_NUMNODES
triggering, followed by a NULL deref in memmap_init() when trying to
access node 64's (NODE_SHIFT=6) node data.

To compensate said change, make memblock_set_node() warn on and adjust
a passed in value of MAX_NUMNODES, just like various other functions
already do.

Fixes: ff6c3d81f2e8 ("NUMA: optimize detection of memory with no node id assigned by firmware")
Signed-off-by: Jan Beulich <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Mike Rapoport (IBM) <[email protected]>
mm/memblock.c
This page took 0.051973 seconds and 4 git commands to generate.