-----Original Message----- From: De Marchi, Lucas lucas.demarchi@intel.com Sent: Friday, April 29, 2022 1:50 PM To: Srivatsa, Anusha anusha.srivatsa@intel.com Cc: intel-gfx@lists.freedesktop.org; stable@vger.kernel.org Subject: Re: [PATCH] drm/i915/dmc: Add MMIO range restrictions
On Fri, Apr 29, 2022 at 01:39:03PM -0700, Anusha Srivatsa wrote:
-----Original Message----- From: De Marchi, Lucas lucas.demarchi@intel.com Sent: Tuesday, April 26, 2022 10:42 PM To: Srivatsa, Anusha anusha.srivatsa@intel.com Cc: intel-gfx@lists.freedesktop.org; stable@vger.kernel.org Subject: Re: [PATCH] drm/i915/dmc: Add MMIO range restrictions
On Tue, Apr 26, 2022 at 05:35:09PM -0700, Anusha Srivatsa wrote:
Bspec has added some steps that check forDMC MMIO range before programming them
v2: Fix for CI v3: move register defines to .h (Anusha)
- Check MMIO restrictions per pipe
- Add MMIO restricton for v1 dmc header as well (Lucas)
BSpec: 49193
Cc: stable@vger.kernel.org Cc: Lucas De Marchi lucas.demarchi@intel.com Signed-off-by: Anusha Srivatsa anusha.srivatsa@intel.com
drivers/gpu/drm/i915/display/intel_dmc.c | 48 ++++++++++++++++--- drivers/gpu/drm/i915/display/intel_dmc_regs.h | 31 ++++++++++++ 2 files changed, 72 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/i915/display/intel_dmc.c b/drivers/gpu/drm/i915/display/intel_dmc.c index 257cf662f9f4..ac7896835bfa 100644 --- a/drivers/gpu/drm/i915/display/intel_dmc.c +++ b/drivers/gpu/drm/i915/display/intel_dmc.c @@ -97,13 +97,6 @@ MODULE_FIRMWARE(SKL_DMC_PATH); #define BXT_DMC_MAX_FW_SIZE 0x3000 MODULE_FIRMWARE(BXT_DMC_PATH);
-#define DMC_DEFAULT_FW_OFFSET 0xFFFFFFFF -#define PACKAGE_MAX_FW_INFO_ENTRIES 20 -#define PACKAGE_V2_MAX_FW_INFO_ENTRIES 32 -#define DMC_V1_MAX_MMIO_COUNT 8 -#define DMC_V3_MAX_MMIO_COUNT 20 -#define DMC_V1_MMIO_START_RANGE 0x80000
struct intel_css_header { /* 0x09 for DMC */ u32 module_type; @@ -374,6 +367,43 @@ static void dmc_set_fw_offset(struct intel_dmc
*dmc,
} }
+static bool dmc_mmio_addr_sanity_check(struct intel_dmc *dmc, const
u32 *mmioaddr,
u32 mmio_count, int header_ver, u8
dmc_id) {
- struct drm_i915_private *i915 = container_of(dmc, typeof(*i915),
dmc);
- int i;
- if (header_ver == 1) {
for (i = 0; i < mmio_count; i++) {
if (mmioaddr[i] < DMC_MMIO_START_RANGE ||
mmioaddr[i] > DMC_MMIO_END_RANGE)
return false;
}
return missing here
- }
- /* Main DMC MMIO check */
- if (dmc_id == DMC_FW_MAIN) {
for (i = 0; i < mmio_count; i++) {
if (mmioaddr[i] < TGL_DMC_MMIO_START(dmc_id)
|| mmioaddr[i] > TGL_DMC_MMIO_END(dmc_id))
return false;
}
- }
- /* Pipe DMC MMIO check */
- if (IS_DG2(i915) || IS_ALDERLAKE_P(i915)) {
for (i = 0; i < mmio_count; i++) {
if (mmioaddr[i] < ADLP_PIPE_MMIO_START &&
mmioaddr[i] > ADLP_PIPE_MMIO_END)
return false;
}
for DG2, main should use TGL_DMC_MMIO_START? and then fail here because of another missing return above?
- } else if (IS_TIGERLAKE(i915) || IS_DG1(i915) ||
IS_ALDERLAKE_S(i915)) {
for (i = 0; i < mmio_count; i++) {
if (mmioaddr[i] < TGL_DMC_MMIO_START(dmc_id)
|| mmioaddr[i] > TGL_DMC_MMIO_END(dmc_id))
return false;
this is handling DMC_FW_MAIN twice.
Maybe something like this:
u32 start, end;
if (header_ver == 1) { start = DMC_MMIO_START_RANGE; end = DMC_MMIO_END_RANGE; } else if (dmc_id == DMC_FW_MAIN || IS_TIGERLAKE(i915) || IS_DG1(i915) || IS_ALDERLAKE_S(i915)) { start = TGL_DMC_MMIO_START(dmc_id); end = TGL_DMC_MMIO_END(dmc_id); } else if (IS_DG2(i915) || IS_ALDERLAKE_P(i915)) { start = ADLP_PIPE_MMIO_START; end = ADLP_PIPE_MMIO_END; } else { drm_warn(&i915->drm, "Unknown mmio range for sanity
check");
return false;
}
for (i = 0; i < mmio_count; i++) if (mmioaddr[i] < start || mmioaddr[i] > end) return false;
return true;
... untested, and may need tweaks depending on the answer to the question above on what range to use for ADL-P/DG2 on main DMC.
The above approach is definitely neater. The main DMC offset is the same for all platforms.
}
- }
- return true;
+}
static u32 parse_dmc_fw_header(struct intel_dmc *dmc, const struct intel_dmc_header_base
*dmc_header,
size_t rem_size, u8 dmc_id) @@ -443,6 +473,10
@@ static
u32 parse_dmc_fw_header(struct intel_dmc
*dmc,
return 0;
}
- if (!dmc_mmio_addr_sanity_check(dmc, mmioaddr, mmio_count,
dmc_header->header_ver, dmc_id))
drm_err(&i915->drm, "DMC firmware has Wrong MMIO
Addresses\n");
return 0;
you don't like DMC and decided to fail it for all platforms?
<facepalm>
- for (i = 0; i < mmio_count; i++) { dmc_info->mmioaddr[i] = _MMIO(mmioaddr[i]); dmc_info->mmiodata[i] = mmiodata[i]; diff --git
a/drivers/gpu/drm/i915/display/intel_dmc_regs.h b/drivers/gpu/drm/i915/display/intel_dmc_regs.h index d65e698832eb..235d1b721334 100644 --- a/drivers/gpu/drm/i915/display/intel_dmc_regs.h +++ b/drivers/gpu/drm/i915/display/intel_dmc_regs.h @@ -11,12 +11,43 @@ #define DMC_PROGRAM(addr, i) _MMIO((addr) + (i) * 4) #define DMC_SSP_BASE_ADDR_GEN9 0x00002FC0 #define DMC_HTP_ADDR_SKL 0x00500034 +#define DMC_DEFAULT_FW_OFFSET 0xFFFFFFFF #define DMC_SSP_BASE _MMIO(0x8F074) #define DMC_HTP_SKL _MMIO(0x8F004) #define DMC_LAST_WRITE _MMIO(0x8F034) #define DMC_LAST_WRITE_VALUE 0xc003b400 #define DMC_MMIO_START_RANGE 0x80000 #define DMC_MMIO_END_RANGE 0x8FFFF +#define PACKAGE_MAX_FW_INFO_ENTRIES 20 +#define PACKAGE_V2_MAX_FW_INFO_ENTRIES 32 +#define DMC_V1_MAX_MMIO_COUNT 8 +#define DMC_V3_MAX_MMIO_COUNT 20
why are you moving these to _regs? these describe the DMC header/blob
Yeah my mistake. While making the change they seemed like the right thing
to do.
+#define DMC_V1_MMIO_START_RANGE 0x80000 +#define _TGL_MAIN_MMIO_START 0x8F000 +#define _TGL_MAIN_MMIO_END 0x8FFFF +#define _TGL_PIPEA_MMIO_START 0x92000 +#define _TGL_PIPEA_MMIO_END 0x93FFF +#define _TGL_PIPEB_MMIO_START 0x96000 +#define _TGL_PIPEB_MMIO_END 0x97FFF +#define _TGL_PIPEC_MMIO_START 0x9A000 +#define _TGL_PIPEC_MMIO_END 0x9BFFF +#define _TGL_PIPED_MMIO_START 0x9E000 +#define _TGL_PIPED_MMIO_END 0x9FFFF +#define ADLP_PIPE_MMIO_START 0x5F000 +#define ADLP_PIPE_MMIO_END 0x5FFFF
+#define TGL_DMC_MMIO_START(pipe) _PICK(pipe,
_TGL_MAIN_MMIO_START,\
_PICK? Did you miss my previous review?
No. the thing is Main DMC with the Pipe DMCs are not evenly spaced. So
using PICK_EVEN is not the right choice here. We also don't need to do _MMIO really.....
Unless I am missing something, this seems like the right approach.
Because the name you chose for your variable:
TGL_DMC_MMIO_START(pipe) _PICK(pipe,
I was expecting this to be used only with the pipe DMC address, which are evenly spaced. It seems the argument you're expecting here is a dmc_id.
Ah, yes. I see the confusion now. It is expecting the dmc_id ,yes. In the usage of the macro in the beginning of the patch though, I am using dmc_id. Thought that would make it clearer, still pipe was the wrong choice of var name.
But.... you said:
The main DMC offset is the same for all platforms.
So, maybe just handle that separately and keep using pipe here? Then you can switch to _PICK_EVEN()
While the Pipe DMC s are evenly spaced and _PICK_EVEN is the right choice. The dmc_id for PipeA, PipeB ....will be 1,2....and not 0,1,2 so the helper will return the wrong values. But you are suggesting to use PIPE_A, PIPE_B etc which will be 0 indexed. But here in the code we are parsing the dmc binary to see if it has Pipe DMC and if so what the MMIO offsets are they in and if it is a valid blob or not. The data we can use at this point is the dmc_id...... Unless we do a conversion from dmc-id to the pipe:
If (DMC_FW_PIPEA) TGL_DMC_MMIO_START(PIPE_A) _PICK(
That will lead to individual such conditions per pipe. Anusha
Lucas De Marchi