[Intel-gfx] [PATCH 9/9] drm/i915/dmc: protect against loading wrong firmware

[Lucas De Marchi]

On Fri, Jun 14, 2019 at 12:44:50PM -0700, Anusha Srivatsa wrote:
>
>
>>-----Original Message-----
>>From: De Marchi, Lucas
>>Sent: Friday, June 7, 2019 2:13 AM
>>To: intel-gfx at lists.freedesktop.org
>>Cc: Vivi, Rodrigo <rodrigo.vivi at intel.com>; Srivatsa, Anusha
>><anusha.srivatsa at intel.com>; De Marchi, Lucas <lucas.demarchi at intel.com>
>>Subject: [PATCH 9/9] drm/i915/dmc: protect against loading wrong firmware
>>
>>In intel_package_header version 2 there's a new field in the fw_info table that
>>must be 0, otherwise it's not the correct DMC firmware. Add a check for version 2
>>or later.
>>
>>Signed-off-by: Lucas De Marchi <lucas.demarchi at intel.com>
>>---
>> drivers/gpu/drm/i915/intel_csr.c | 14 +++++++++++---
>> 1 file changed, 11 insertions(+), 3 deletions(-)
>>
>>diff --git a/drivers/gpu/drm/i915/intel_csr.c b/drivers/gpu/drm/i915/intel_csr.c
>>index 7608e4e2950d..864531aae1a5 100644
>>--- a/drivers/gpu/drm/i915/intel_csr.c
>>+++ b/drivers/gpu/drm/i915/intel_csr.c
>>@@ -120,7 +120,10 @@ struct intel_css_header {  } __packed;
>>
>> struct intel_fw_info {
>>-	u16 reserved1;
>>+	u8 reserved1;
>>+
>>+	/* reserved on package_header version 1, must be 0 on version 2 */
>>+	u8 dmc_id;
>>
>> 	/* Stepping (A, B, C, ..., *). * is a wildcard */
>> 	char stepping;
>>@@ -325,12 +328,16 @@ void intel_csr_load_program(struct drm_i915_private
>>*dev_priv)
>>  */
>> static u32 find_dmc_fw_offset(const struct intel_fw_info *fw_info,
>> 			      unsigned int num_entries,
>>-			      const struct stepping_info *si)
>>+			      const struct stepping_info *si,
>>+			      u8 package_ver)
>> {
>> 	u32 dmc_offset = CSR_DEFAULT_FW_OFFSET;
>> 	unsigned int i;
>>
>> 	for (i = 0; i < num_entries; i++) {
>>+		if (package_ver > 1 && fw_info[i].dmc_id != 0)
>>+			continue;
>
>Wont we need a message here? "Wrong DMC version, not loading v_x.0x" or something?.."

I don't think so. It's normal to have other blobs inside the firmware
that we don't care about. At most I could put a debug, but I don't think
we really care.

As long as we find one we should be fine. And if we don't, then we will
complain loudly later in this function with a DRM_ERROR().
dmc_offset remaining as CSR_DEFAULT_FW_OFFSET has a double meaning here:
either we didn't find any entry or we found one that had the offset set
to this value.

Lucas De Marchi

>
>Anusha
>> 		if (fw_info[i].substepping == '*' &&
>> 		    si->stepping == fw_info[i].stepping) {
>> 			dmc_offset = fw_info[i].offset;
>>@@ -508,7 +515,8 @@ parse_csr_fw_package(struct intel_csr *csr,
>>
>> 	fw_info = (const struct intel_fw_info *)
>> 		((u8 *)package_header + sizeof(*package_header));
>>-	dmc_offset = find_dmc_fw_offset(fw_info, num_entries, si);
>>+	dmc_offset = find_dmc_fw_offset(fw_info, num_entries, si,
>>+					package_header->header_ver);
>> 	if (dmc_offset == CSR_DEFAULT_FW_OFFSET) {
>> 		DRM_ERROR("DMC firmware not supported for %c stepping\n",
>> 			  si->stepping);
>>--
>>2.21.0
>