"value":"In the Linux kernel, the following vulnerability has been resolved:\n\nx86/microcode/AMD: Fix out-of-bounds on systems with CPU-less NUMA nodes\n\nCurrently, load_microcode_amd() iterates over all NUMA nodes, retrieves their\nCPU masks and unconditionally accesses per-CPU data for the first CPU of each\nmask.\n\nAccording to Documentation/admin-guide/mm/numaperf.rst:\n\n \"Some memory may share the same node as a CPU, and others are provided as\n memory only nodes.\"\n\nTherefore, some node CPU masks may be empty and wouldn't have a \"first CPU\".\n\nOn a machine with far memory (and therefore CPU-less NUMA nodes):\n- cpumask_of_node(nid) is 0\n- cpumask_first(0) is CONFIG_NR_CPUS\n- cpu_data(CONFIG_NR_CPUS) accesses the cpu_info per-CPU array at an\n index that is 1 out of bounds\n\nThis does not have any security implications since flashing microcode is\na privileged operation but I believe this has reliability implications by\npotentially corrupting memory while flashing a microcode update.\n\nWhen booting with CONFIG_UBSAN_BOUNDS=y on an AMD machine that flashes\na microcode update. I get the following splat:\n\n UBSAN: array-index-out-of-bounds in arch/x86/kernel/cpu/microcode/amd.c:X:Y\n index 512 is out of range for type 'unsigned long[512]'\n [...]\n Call Trace:\n dump_stack\n __ubsan_handle_out_of_bounds\n load_microcode_amd\n request_microcode_amd\n reload_store\n kernfs_fop_write_iter\n vfs_write\n ksys_write\n do_syscall_64\n entry_SYSCALL_64_after_hwframe\n\nChange the loop to go over only NUMA nodes which have CPUs before determining\nwhether the first CPU on the respective node needs microcode update.\n\n [ bp: Massage commit message, fix typo. ]"
