CVE-2024-40917

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> memblock: make memblock_set_node() also warn about use of MAX_NUMNODES<br /> <br /> On an (old) x86 system with SRAT just covering space above 4Gb:<br /> <br /> ACPI: SRAT: Node 0 PXM 0 [mem 0x100000000-0xfffffffff] hotplug<br /> <br /> the commit referenced below leads to this NUMA configuration no longer<br /> being refused by a CONFIG_NUMA=y kernel (previously<br /> <br /> NUMA: nodes only cover 6144MB of your 8185MB e820 RAM. Not used.<br /> No NUMA configuration found<br /> Faking a node at [mem 0x0000000000000000-0x000000027fffffff]<br /> <br /> was seen in the log directly after the message quoted above), because of<br /> memblock_validate_numa_coverage() checking for NUMA_NO_NODE (only). This<br /> in turn led to memblock_alloc_range_nid()&amp;#39;s warning about MAX_NUMNODES<br /> triggering, followed by a NULL deref in memmap_init() when trying to<br /> access node 64&amp;#39;s (NODE_SHIFT=6) node data.<br /> <br /> To compensate said change, make memblock_set_node() warn on and adjust<br /> a passed in value of MAX_NUMNODES, just like various other functions<br /> already do.