[Intel-gfx] [PATCH v8 3/4] drm/i915: Manipulate DBuf slices properly
Matt Roper
matthew.d.roper at intel.com
Fri Dec 13 19:01:07 UTC 2019
On Fri, Dec 13, 2019 at 03:02:27PM +0200, Stanislav Lisovskiy wrote:
> Start manipulating DBuf slices as a mask,
> but not as a total number, as current approach
> doesn't give us full control on all combinations
> of slices, which we might need(like enabling S2
> only can't enabled by setting enabled_slices=1).
>
> Removed wrong code from intel_get_ddb_size as
> it doesn't match to BSpec. For now still just
> use DBuf slice until proper algorithm is implemented.
>
> Other minor code refactoring to get prepared
> for major DBuf assignment changes landed:
> - As now enabled slices contain a mask
> we still need some value which should
> reflect how much DBuf slices are supported
> by the platform, now device info contains
> num_supported_dbuf_slices.
> - Removed unneeded assertion as we are now
> manipulating slices in a more proper way.
>
> v2: Start using enabled_slices in dev_priv
>
> v3: "enabled_slices" is now "enabled_dbuf_slices_mask",
> as this now sits in dev_priv independently.
>
> Signed-off-by: Stanislav Lisovskiy <stanislav.lisovskiy at intel.com>
> ---
> drivers/gpu/drm/i915/display/intel_display.c | 23 ++--
> .../drm/i915/display/intel_display_power.c | 100 ++++++++----------
> .../drm/i915/display/intel_display_power.h | 5 +
> .../drm/i915/display/intel_display_types.h | 2 +-
> drivers/gpu/drm/i915/i915_drv.h | 2 +-
> drivers/gpu/drm/i915/i915_pci.c | 6 +-
> drivers/gpu/drm/i915/intel_device_info.h | 1 +
> drivers/gpu/drm/i915/intel_pm.c | 49 +++------
> drivers/gpu/drm/i915/intel_pm.h | 2 +-
> 9 files changed, 84 insertions(+), 106 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/display/intel_display.c b/drivers/gpu/drm/i915/display/intel_display.c
> index 0e09d0c23b1d..42a0ea540d4f 100644
> --- a/drivers/gpu/drm/i915/display/intel_display.c
> +++ b/drivers/gpu/drm/i915/display/intel_display.c
> @@ -13359,12 +13359,12 @@ static void verify_wm_state(struct intel_crtc *crtc,
>
> skl_pipe_ddb_get_hw_state(crtc, hw->ddb_y, hw->ddb_uv);
>
> - hw_enabled_slices = intel_enabled_dbuf_slices_num(dev_priv);
> + hw_enabled_slices = intel_enabled_dbuf_slices_mask(dev_priv);
>
> if (INTEL_GEN(dev_priv) >= 11 &&
> - hw_enabled_slices != dev_priv->enabled_dbuf_slices_num)
> - DRM_ERROR("mismatch in DBUF Slices (expected %u, got %u)\n",
> - dev_priv->enabled_dbuf_slices_num,
> + hw_enabled_slices != dev_priv->enabled_dbuf_slices_mask)
> + DRM_ERROR("mismatch in DBUF Slices (expected %x, got %x)\n",
Minor nitpick: I'd write these with "0x%x" to make it more obvious to
the person reading the message that the value is hex and thus a bitmask
rather than a count.
> + dev_priv->enabled_dbuf_slices_mask,
> hw_enabled_slices);
>
> /* planes */
> @@ -14549,22 +14549,23 @@ static void intel_update_trans_port_sync_crtcs(struct intel_crtc *crtc,
> static void icl_dbuf_slice_pre_update(struct intel_atomic_state *state)
> {
> struct drm_i915_private *dev_priv = to_i915(state->base.dev);
> - u8 hw_enabled_slices = dev_priv->enabled_dbuf_slices_num;
> - u8 required_slices = state->enabled_dbuf_slices_num;
> + u8 hw_enabled_slices = dev_priv->enabled_dbuf_slices_mask;
> + u8 required_slices = state->enabled_dbuf_slices_mask;
> + u8 slices_union = hw_enabled_slices | required_slices;
>
> /* If 2nd DBuf slice required, enable it here */
> - if (INTEL_GEN(dev_priv) >= 11 && required_slices > hw_enabled_slices)
> - icl_dbuf_slices_update(dev_priv, required_slices);
> + if (INTEL_GEN(dev_priv) >= 11 && required_slices != hw_enabled_slices)
Should this be hw_enabled_slices != slices_union so that we only process
it when we're adding new slices and not in cases where we're only going
to be taking slices away?
> + icl_dbuf_slices_update(dev_priv, slices_union);
> }
>
> static void icl_dbuf_slice_post_update(struct intel_atomic_state *state)
> {
> struct drm_i915_private *dev_priv = to_i915(state->base.dev);
> - u8 hw_enabled_slices = dev_priv->enabled_dbuf_slices_num;
> - u8 required_slices = state->enabled_dbuf_slices_num;
> + u8 hw_enabled_slices = dev_priv->enabled_dbuf_slices_mask;
> + u8 required_slices = state->enabled_dbuf_slices_mask;
>
> /* If 2nd DBuf slice is no more required disable it */
> - if (INTEL_GEN(dev_priv) >= 11 && required_slices < hw_enabled_slices)
> + if (INTEL_GEN(dev_priv) >= 11 && required_slices != hw_enabled_slices)
> icl_dbuf_slices_update(dev_priv, required_slices);
> }
>
> diff --git a/drivers/gpu/drm/i915/display/intel_display_power.c b/drivers/gpu/drm/i915/display/intel_display_power.c
> index b8983422a882..ba384a5315f8 100644
> --- a/drivers/gpu/drm/i915/display/intel_display_power.c
> +++ b/drivers/gpu/drm/i915/display/intel_display_power.c
> @@ -1031,15 +1031,6 @@ static bool gen9_dc_off_power_well_enabled(struct drm_i915_private *dev_priv,
> (I915_READ(DC_STATE_EN) & DC_STATE_EN_UPTO_DC5_DC6_MASK) == 0);
> }
>
> -static void gen9_assert_dbuf_enabled(struct drm_i915_private *dev_priv)
> -{
> - u32 tmp = I915_READ(DBUF_CTL);
> -
> - WARN((tmp & (DBUF_POWER_STATE | DBUF_POWER_REQUEST)) !=
> - (DBUF_POWER_STATE | DBUF_POWER_REQUEST),
> - "Unexpected DBuf power power state (0x%08x)\n", tmp);
> -}
> -
> static void gen9_disable_dc_states(struct drm_i915_private *dev_priv)
> {
> struct intel_cdclk_state cdclk_state = {};
> @@ -1055,8 +1046,6 @@ static void gen9_disable_dc_states(struct drm_i915_private *dev_priv)
> /* Can't read out voltage_level so can't use intel_cdclk_changed() */
> WARN_ON(intel_cdclk_needs_modeset(&dev_priv->cdclk.hw, &cdclk_state));
>
> - gen9_assert_dbuf_enabled(dev_priv);
> -
I'm undecided on whether dropping the assertion completely is a good
idea or not. If I understand correctly, it was never meant to check
that our own handling of the dbuf is correct (and it was added back when
we only ever had a single dbuf slice on gen9), but rather than the DMC
did the right thing when enabling/disabling the dbuf around DC state
transitions. As far as I know, the DMC has been pretty solid recently,
but extra paranoia around firmware we don't directly control still seems
like a good idea to me.
> if (IS_GEN9_LP(dev_priv))
> bxt_verify_ddi_phy_power_wells(dev_priv);
>
> @@ -4254,72 +4243,71 @@ static void gen9_dbuf_disable(struct drm_i915_private *dev_priv)
> intel_dbuf_slice_set(dev_priv, DBUF_CTL, false);
> }
>
> -static u8 intel_dbuf_max_slices(struct drm_i915_private *dev_priv)
> +int intel_dbuf_max_slices(struct drm_i915_private *dev_priv)
I don't see any new usage of this function outside this file. I think
we can keep this static?
> {
> - if (INTEL_GEN(dev_priv) < 11)
> - return 1;
> - return 2;
> + return INTEL_INFO(dev_priv)->num_supported_dbuf_slices;
> +}
> +
> +void icl_program_dbuf_slices(struct drm_i915_private *dev_priv)
It looks like this never gets used outside this file either, so it could
probably be static. Or it's simple enough that we could just drop it
completely and inline the icl_dbuf_slices_update() directly at the
callsites.
> +{
> + const u8 hw_enabled_slices = dev_priv->enabled_dbuf_slices_mask;
> +
> + icl_dbuf_slices_update(dev_priv, hw_enabled_slices);
> }
>
> void icl_dbuf_slices_update(struct drm_i915_private *dev_priv,
> u8 req_slices)
> {
> - const u8 hw_enabled_slices = dev_priv->enabled_dbuf_slices_num;
> - bool ret;
> + int i;
> + int max_slices = intel_dbuf_max_slices(dev_priv);
>
> - if (req_slices > intel_dbuf_max_slices(dev_priv)) {
> + if (hweight8(req_slices) > intel_dbuf_max_slices(dev_priv)) {
> DRM_ERROR("Invalid number of dbuf slices requested\n");
> return;
> }
>
> - if (req_slices == hw_enabled_slices || req_slices == 0)
> - return;
> + DRM_DEBUG_KMS("Updating dbuf slices to %x\n", req_slices);
As noted above, I prefer "0x%x" to make it more obvious the number is a
bitmask. Up to you though.
>
> - if (req_slices > hw_enabled_slices)
> - ret = intel_dbuf_slice_set(dev_priv, DBUF_CTL_S2, true);
> - else
> - ret = intel_dbuf_slice_set(dev_priv, DBUF_CTL_S2, false);
> + for (i = 0; i < max_slices; i++) {
> + int slice_bit = BIT(i);
> + bool slice_set = (slice_bit & req_slices) != 0;
Technically you could drop the "!= 0" part since anything non-zero will
be evaluated as 'true.' Doesn't really matter though.
> +
> + switch (slice_bit) {
> + case DBUF_S1_BIT:
> + intel_dbuf_slice_set(dev_priv,
> + DBUF_CTL_S1,
> + slice_set);
> + break;
> + case DBUF_S2_BIT:
> + intel_dbuf_slice_set(dev_priv,
> + DBUF_CTL_S2,
> + slice_set);
> + break;
> + default:
> + MISSING_CASE(slice_bit);
> + }
It might be worth parameterizing DBUF_CTL so that we can simplify code
like this to just
intel_dbuf_slice_set(dev_priv, DBUF_CTL(i), slice_set);
or even just
intel_dbuf_slice_set(dev_priv, DBUF_CTL(i), req_slices & BIT(i))
as we do with other types of registers. That will also future-proof a
bit in case we ever get platforms with more dbuf slices.
> + }
>
> - if (ret)
> - dev_priv->enabled_dbuf_slices_num = req_slices;
> + dev_priv->enabled_dbuf_slices_mask = req_slices;
> }
>
> static void icl_dbuf_enable(struct drm_i915_private *dev_priv)
> {
> - I915_WRITE(DBUF_CTL_S1, I915_READ(DBUF_CTL_S1) | DBUF_POWER_REQUEST);
> - I915_WRITE(DBUF_CTL_S2, I915_READ(DBUF_CTL_S2) | DBUF_POWER_REQUEST);
> - POSTING_READ(DBUF_CTL_S2);
> -
> - udelay(10);
> -
> - if (!(I915_READ(DBUF_CTL_S1) & DBUF_POWER_STATE) ||
> - !(I915_READ(DBUF_CTL_S2) & DBUF_POWER_STATE))
> - DRM_ERROR("DBuf power enable timeout\n");
> - else
> - /*
> - * FIXME: for now pretend that we only have 1 slice, see
> - * intel_enabled_dbuf_slices_num().
> - */
> - dev_priv->enabled_dbuf_slices_num = 1;
> + /*
> + * Just power up 1 slice, we will
> + * figure out later which slices we have and what we need.
> + */
> + dev_priv->enabled_dbuf_slices_mask = DBUF_S1_BIT;
> + icl_program_dbuf_slices(dev_priv);
> }
>
> static void icl_dbuf_disable(struct drm_i915_private *dev_priv)
> {
> - I915_WRITE(DBUF_CTL_S1, I915_READ(DBUF_CTL_S1) & ~DBUF_POWER_REQUEST);
> - I915_WRITE(DBUF_CTL_S2, I915_READ(DBUF_CTL_S2) & ~DBUF_POWER_REQUEST);
> - POSTING_READ(DBUF_CTL_S2);
> -
> - udelay(10);
> -
> - if ((I915_READ(DBUF_CTL_S1) & DBUF_POWER_STATE) ||
> - (I915_READ(DBUF_CTL_S2) & DBUF_POWER_STATE))
> - DRM_ERROR("DBuf power disable timeout!\n");
> - else
> - /*
> - * FIXME: for now pretend that the first slice is always
> - * enabled, see intel_enabled_dbuf_slices_num().
> - */
> - dev_priv->enabled_dbuf_slices_num = 1;
> + /*
> + * Disable all slices
> + */
> + dev_priv->enabled_dbuf_slices_mask = 0;
> + icl_program_dbuf_slices(dev_priv);
> }
>
> static void icl_mbus_init(struct drm_i915_private *dev_priv)
> diff --git a/drivers/gpu/drm/i915/display/intel_display_power.h b/drivers/gpu/drm/i915/display/intel_display_power.h
> index 1da04f3e0fb3..0d9f87607eac 100644
> --- a/drivers/gpu/drm/i915/display/intel_display_power.h
> +++ b/drivers/gpu/drm/i915/display/intel_display_power.h
> @@ -311,8 +311,13 @@ intel_display_power_put_async(struct drm_i915_private *i915,
> for ((wf) = intel_display_power_get((i915), (domain)); (wf); \
> intel_display_power_put_async((i915), (domain), (wf)), (wf) = 0)
>
> +#define DBUF_S1_BIT BIT(0)
> +#define DBUF_S2_BIT BIT(1)
> +
> void icl_dbuf_slices_update(struct drm_i915_private *dev_priv,
> u8 req_slices);
> +void icl_program_dbuf_slices(struct drm_i915_private *dev_priv);
> +int intel_dbuf_max_slices(struct drm_i915_private *dev_priv);
>
> void chv_phy_powergate_lanes(struct intel_encoder *encoder,
> bool override, unsigned int mask);
> diff --git a/drivers/gpu/drm/i915/display/intel_display_types.h b/drivers/gpu/drm/i915/display/intel_display_types.h
> index 70e65c2d525d..ba2e41a03051 100644
> --- a/drivers/gpu/drm/i915/display/intel_display_types.h
> +++ b/drivers/gpu/drm/i915/display/intel_display_types.h
> @@ -518,7 +518,7 @@ struct intel_atomic_state {
> struct skl_ddb_values wm_results;
>
> /* Number of enabled DBuf slices */
> - u8 enabled_dbuf_slices_num;
> + u8 enabled_dbuf_slices_mask;
>
> struct i915_sw_fence commit_ready;
>
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 7a2d9fa5a9a6..ec4b9e3cef79 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -1210,7 +1210,7 @@ struct drm_i915_private {
> bool distrust_bios_wm;
> } wm;
>
> - u8 enabled_dbuf_slices_num; /* GEN11 has configurable 2 slices */
> + u8 enabled_dbuf_slices_mask; /* GEN11 has configurable 2 slices */
>
> struct dram_info {
> bool valid;
> diff --git a/drivers/gpu/drm/i915/i915_pci.c b/drivers/gpu/drm/i915/i915_pci.c
> index 877560b1031e..2068aac5ab6a 100644
> --- a/drivers/gpu/drm/i915/i915_pci.c
> +++ b/drivers/gpu/drm/i915/i915_pci.c
> @@ -614,7 +614,8 @@ static const struct intel_device_info intel_cherryview_info = {
> .has_gt_uc = 1, \
> .display.has_hdcp = 1, \
> .display.has_ipc = 1, \
> - .ddb_size = 896
> + .ddb_size = 896, \
> + .num_supported_dbuf_slices = 1
>
> #define SKL_PLATFORM \
> GEN9_FEATURES, \
> @@ -649,6 +650,7 @@ static const struct intel_device_info intel_skylake_gt4_info = {
> #define GEN9_LP_FEATURES \
> GEN(9), \
> .is_lp = 1, \
> + .num_supported_dbuf_slices = 1, \
> .display.has_hotplug = 1, \
> .engine_mask = BIT(RCS0) | BIT(VCS0) | BIT(BCS0) | BIT(VECS0), \
> .pipe_mask = BIT(PIPE_A) | BIT(PIPE_B) | BIT(PIPE_C), \
> @@ -737,6 +739,7 @@ static const struct intel_device_info intel_coffeelake_gt3_info = {
> GEN9_FEATURES, \
> GEN(10), \
> .ddb_size = 1024, \
> + .num_supported_dbuf_slices = 1, \
GEN10_FEATURES inherits everything from GEN9_FEATURES, so I think we can
drop this one.
> .display.has_dsc = 1, \
> .has_coherent_ggtt = false, \
> GLK_COLORS
> @@ -773,6 +776,7 @@ static const struct intel_device_info intel_cannonlake_info = {
> }, \
> GEN(11), \
> .ddb_size = 2048, \
> + .num_supported_dbuf_slices = 2, \
> .has_logical_ring_elsq = 1, \
> .color = { .degamma_lut_size = 33, .gamma_lut_size = 262145 }
>
> diff --git a/drivers/gpu/drm/i915/intel_device_info.h b/drivers/gpu/drm/i915/intel_device_info.h
> index 2725cb7fc169..7d4d122d2182 100644
> --- a/drivers/gpu/drm/i915/intel_device_info.h
> +++ b/drivers/gpu/drm/i915/intel_device_info.h
> @@ -180,6 +180,7 @@ struct intel_device_info {
> } display;
>
> u16 ddb_size; /* in blocks */
> + u8 num_supported_dbuf_slices; /* number of DBuf slices */
>
> /* Register offsets for the various display pipes and transcoders */
> int pipe_offsets[I915_MAX_TRANSCODERS];
> diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c
> index c2510978ccdf..111bcafd6e4c 100644
> --- a/drivers/gpu/drm/i915/intel_pm.c
> +++ b/drivers/gpu/drm/i915/intel_pm.c
> @@ -3616,26 +3616,22 @@ bool ilk_disable_lp_wm(struct drm_i915_private *dev_priv)
> return _ilk_disable_lp_wm(dev_priv, WM_DIRTY_LP_ALL);
> }
>
> -u8 intel_enabled_dbuf_slices_num(struct drm_i915_private *dev_priv)
> +u8 intel_enabled_dbuf_slices_mask(struct drm_i915_private *dev_priv)
> {
> - u8 enabled_dbuf_slices_num;
> -
> - /* Slice 1 will always be enabled */
> - enabled_dbuf_slices_num = 1;
> + u8 enabled_slices_mask = 0;
>
> /* Gen prior to GEN11 have only one DBuf slice */
> if (INTEL_GEN(dev_priv) < 11)
> - return enabled_dbuf_slices_num;
> + return DBUF_S1_BIT;
Do we want to assume this or should we just read it anyway to make sure?
As noted before, we might get into a disagreement with the DMC or
something and it would be good to know how the hardware is actually set.
> - /*
> - * FIXME: for now we'll only ever use 1 slice; pretend that we have
> - * only that 1 slice enabled until we have a proper way for on-demand
> - * toggling of the second slice.
> - */
> - if (0 && I915_READ(DBUF_CTL_S2) & DBUF_POWER_STATE)
> - enabled_dbuf_slices_num++;
> + /* Check if second DBuf slice is enabled */
> + if (I915_READ(DBUF_CTL_S1) & DBUF_POWER_STATE)
> + enabled_slices_mask |= DBUF_S1_BIT;
You have the comment about the second slice before your check of the
first slice... I think you can just drop the comment in this case.
>
> - return enabled_dbuf_slices_num;
> + if (I915_READ(DBUF_CTL_S2) & DBUF_POWER_STATE)
> + enabled_slices_mask |= DBUF_S2_BIT;
> +
> + return enabled_slices_mask;
> }
>
> /*
> @@ -3843,8 +3839,6 @@ static u16 intel_get_ddb_size(struct drm_i915_private *dev_priv,
> {
> struct drm_atomic_state *state = crtc_state->uapi.state;
> struct intel_atomic_state *intel_state = to_intel_atomic_state(state);
> - const struct drm_display_mode *adjusted_mode;
> - u64 total_data_bw;
> u16 ddb_size = INTEL_INFO(dev_priv)->ddb_size;
>
> WARN_ON(ddb_size == 0);
> @@ -3852,23 +3846,8 @@ static u16 intel_get_ddb_size(struct drm_i915_private *dev_priv,
> if (INTEL_GEN(dev_priv) < 11)
> return ddb_size - 4; /* 4 blocks for bypass path allocation */
>
> - adjusted_mode = &crtc_state->hw.adjusted_mode;
> - total_data_bw = total_data_rate * drm_mode_vrefresh(adjusted_mode);
> -
> - /*
> - * 12GB/s is maximum BW supported by single DBuf slice.
> - *
> - * FIXME dbuf slice code is broken:
> - * - must wait for planes to stop using the slice before powering it off
> - * - plane straddling both slices is illegal in multi-pipe scenarios
> - * - should validate we stay within the hw bandwidth limits
> - */
> - if (0 && (num_active > 1 || total_data_bw >= GBps(12))) {
> - intel_state->enabled_dbuf_slices_num = 2;
> - } else {
> - intel_state->enabled_dbuf_slices_num = 1;
> - ddb_size /= 2;
> - }
> + intel_state->enabled_dbuf_slices_mask = DBUF_S1_BIT;
> + ddb_size /= 2;
>
> return ddb_size;
> }
> @@ -4065,7 +4044,7 @@ void skl_pipe_ddb_get_hw_state(struct intel_crtc *crtc,
>
> void skl_ddb_get_hw_state(struct drm_i915_private *dev_priv)
> {
> - dev_priv->enabled_dbuf_slices_num = intel_enabled_dbuf_slices_num(dev_priv);
> + dev_priv->enabled_dbuf_slices_mask = intel_enabled_dbuf_slices_mask(dev_priv);
> }
>
> /*
> @@ -5204,7 +5183,7 @@ skl_compute_ddb(struct intel_atomic_state *state)
> struct intel_crtc *crtc;
> int ret, i;
>
> - state->enabled_dbuf_slices_num = dev_priv->enabled_dbuf_slices_num;
> + state->enabled_dbuf_slices_mask = dev_priv->enabled_dbuf_slices_mask;
>
> for_each_oldnew_intel_crtc_in_state(state, crtc, old_crtc_state,
> new_crtc_state, i) {
> diff --git a/drivers/gpu/drm/i915/intel_pm.h b/drivers/gpu/drm/i915/intel_pm.h
> index a476f6c730e9..7a7494d1224f 100644
> --- a/drivers/gpu/drm/i915/intel_pm.h
> +++ b/drivers/gpu/drm/i915/intel_pm.h
> @@ -33,7 +33,7 @@ void g4x_wm_get_hw_state(struct drm_i915_private *dev_priv);
> void vlv_wm_get_hw_state(struct drm_i915_private *dev_priv);
> void ilk_wm_get_hw_state(struct drm_i915_private *dev_priv);
> void skl_wm_get_hw_state(struct drm_i915_private *dev_priv);
> -u8 intel_enabled_dbuf_slices_num(struct drm_i915_private *dev_priv);
> +u8 intel_enabled_dbuf_slices_mask(struct drm_i915_private *dev_priv);
> void skl_pipe_ddb_get_hw_state(struct intel_crtc *crtc,
> struct skl_ddb_entry *ddb_y,
> struct skl_ddb_entry *ddb_uv);
> --
> 2.17.1
>
--
Matt Roper
Graphics Software Engineer
VTT-OSGC Platform Enablement
Intel Corporation
(916) 356-2795
More information about the Intel-gfx
mailing list