Dan reports that smatch thinks userspace can craft an out-of-bound bus family number. However, nd_cmd_clear_to_send() blocks all non-zero values of bus-family since only the kernel can initiate these commands. However, in the speculation path, family is a user controlled array index value so mask it for speculation safety. Also, since the nd_cmd_clear_to_send() safety is non-obvious and possibly may change in the future include input validation is if userspace could get past the nd_cmd_clear_to_send() gatekeeper.
Link: http://lore.kernel.org/r/20201111113000.GA1237157@mwanda Reported-by: Dan Carpenter dan.carpenter@oracle.com Fixes: 6450ddbd5d8e ("ACPI: NFIT: Define runtime firmware activation commands") Cc: stable@vger.kernel.org Signed-off-by: Dan Williams dan.j.williams@intel.com --- drivers/acpi/nfit/core.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c index cda7b6c52504..b11b08a60684 100644 --- a/drivers/acpi/nfit/core.c +++ b/drivers/acpi/nfit/core.c @@ -5,6 +5,7 @@ #include <linux/list_sort.h> #include <linux/libnvdimm.h> #include <linux/module.h> +#include <linux/nospec.h> #include <linux/mutex.h> #include <linux/ndctl.h> #include <linux/sysfs.h> @@ -479,8 +480,11 @@ int acpi_nfit_ctl(struct nvdimm_bus_descriptor *nd_desc, struct nvdimm *nvdimm, cmd_mask = nd_desc->cmd_mask; if (cmd == ND_CMD_CALL && call_pkg->nd_family) { family = call_pkg->nd_family; - if (!test_bit(family, &nd_desc->bus_family_mask)) + if (family > NVDIMM_BUS_FAMILY_MAX || + !test_bit(family, &nd_desc->bus_family_mask)) return -EINVAL; + family = array_index_nospec(family, + NVDIMM_BUS_FAMILY_MAX + 1); dsm_mask = acpi_desc->family_dsm_mask[family]; guid = to_nfit_bus_uuid(family); } else {