Re: [PATCH v6 08/18] drm/i915/alpm: table based establishment period

["Hogander, Jouni" <jouni.hogander@intel.com>]

On Tue, 2026-09-15 at 23:34 +0530, Animesh Manna wrote:
> Only use the table when the link rate is listed (idx valid)
> and the LTTPR count is within the table bounds. A zero entry
> means the rate has no tabulated value, so fall back to the
> computed establishment period below.
> 
> Note: The table and its values are taken from DP2.1 vesa spec.
> 
> v2: Add check for negative error codes from drm_dp_lttpr_count().
> [sashiko]
> 
> Cc: Jouni Högander <jouni.hogander@intel.com>
> Signed-off-by: Animesh Manna <animesh.manna@intel.com>
> ---
>  drivers/gpu/drm/i915/display/intel_alpm.c | 95
> +++++++++++++++++++++++
>  1 file changed, 95 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/display/intel_alpm.c
> b/drivers/gpu/drm/i915/display/intel_alpm.c
> index dc025058a626..cb24bba801a1 100644
> --- a/drivers/gpu/drm/i915/display/intel_alpm.c
> +++ b/drivers/gpu/drm/i915/display/intel_alpm.c
> @@ -122,6 +122,92 @@ static int get_dp_link_symbol_size(const struct
> intel_crtc_state *crtc_state)
>  	return intel_dp_is_uhbr(crtc_state) ? 32 : 10;
>  }
>  
> +#define DEFAULT_MAX_LTTPR_COUNT 7
> +#define MAX_SUPPORTED_DP_LINK_RATES 12
> +
> +static const u32
> establishment_periods[DEFAULT_MAX_LTTPR_COUNT][MAX_SUPPORTED_DP_LINK_
> RATES] = {
> +	/* 1620, 2160, 2430, 2700, 3240, 4320, 5400, 6750, 8100,
> 10000, 13500, 20000 */
> +	/* No Retimers */
> +	{ 61070, 0, 0, 56710, 0, 0, 53450, 0, 52360, 59050, 56750,
> 54620 },
> +	/*  1 Retimer  */
> +	{ 155670, 0, 0, 151310, 0, 0, 148050, 0, 146960, 153650,
> 151350, 149220 },
> +	/*  2 Retimers */
> +	{ 181160, 0, 0, 172450, 0, 0, 165910, 0, 163740, 177120,
> 172520, 168220 },
> +	/*  3 Retimers */
> +	{ 206650, 0, 0, 193580, 0, 0, 183780, 0, 180510, 200590,
> 193690, 187290 },
> +	/*  4 Retimers */
> +	{ 232140, 0, 0, 214710, 0, 0, 201650, 0, 197290, 224060,
> 214860, 206320 },
> +	/*  5 Retimers */
> +	{ 257620, 0, 0, 235850, 0, 0, 219510, 0, 214070, 247530,
> 236030, 225360 },
> +	/*  6 Retimers */
> +	{ 283110, 0, 0, 256980, 0, 0, 237380, 0, 230850, 271000,
> 257200, 244390 },
> +};
> +
> +enum link_rate_idx {
> +	LINK_RATE_INVALID = -1,
> +	LINK_RATE_1,
> +	LINK_RATE_2,
> +	LINK_RATE_3,
> +	LINK_RATE_4,
> +	LINK_RATE_5,
> +	LINK_RATE_6,
> +	LINK_RATE_7,
> +	LINK_RATE_8,
> +	LINK_RATE_9,
> +	LINK_RATE_10,
> +	LINK_RATE_11,
> +	LINK_RATE_12
> +};
> +
> +static int get_link_rate_index(int port_clock)
> +{
> +	int link_rate_index;
> +
> +	switch (port_clock) {
> +	case 162000:
> +		link_rate_index = LINK_RATE_1;	// Rate_1 (RBR) -
> 1.62 Gbps/Lane
> +		break;
> +	case 216000:
> +		link_rate_index = LINK_RATE_2;	// Rate_2       -
> 2.16 Gbps/Lane
> +		break;
> +	case 243000:
> +		link_rate_index = LINK_RATE_3;	// Rate_3       -
> 2.43 Gbps/Lane
> +		break;
> +	case 270000:
> +		link_rate_index = LINK_RATE_4;	// Rate_4 (HBR) -
> 2.70 Gbps/Lane
> +		break;
> +	case 324000:
> +		link_rate_index = LINK_RATE_5;	// Rate_5 (RBR2)-
> 3.24 Gbps/Lane
> +		break;
> +	case 432000:
> +		link_rate_index = LINK_RATE_6;	// Rate_6       -
> 4.32 Gbps/Lane
> +		break;
> +	case 540000:
> +		link_rate_index = LINK_RATE_7;	// Rate_7 (HBR2)-
> 5.40 Gbps/Lane
> +		break;
> +	case 675000:
> +		link_rate_index = LINK_RATE_8;	// Rate_8       -
> 6.75 Gbps/Lane
> +		break;
> +	case 810000:
> +		link_rate_index = LINK_RATE_9;	// Rate_9 (HBR3)-
> 8.10 Gbps/Lane
> +		break;
> +	case 1000000:
> +		link_rate_index = LINK_RATE_10;	// Rate_10
> (UHBR10) - 10.0 Gbps/Lane
> +		break;
> +	case 1350000:
> +		link_rate_index = LINK_RATE_11;	// Rate_11
> (UHBR13.5) - 13.5 Gbps/Lane
> +		break;
> +	case 2000000:
> +		link_rate_index = LINK_RATE_12;	// Rate_12
> (UHBR20) - 20.0 Gbps/Lane
> +		break;
> +	default:
> +		link_rate_index = LINK_RATE_INVALID;
> +		break;
> +	}
> +
> +	return link_rate_index;
> +}

Difference between what this table is giving us and what we are getting
from the formula is coming from SILENCE_PERIOD_TIME and tlfps_period.
In DP spec they are using max values while our calculation is using
midpoint.

I think we could just switch to max values on these and drop adding
tables completely.

BR,
Jouni Högander

> +
>  static int get_establishment_period(struct intel_dp *intel_dp,
>  				    const struct intel_crtc_state
> *crtc_state)
>  {
> @@ -138,8 +224,17 @@ static int get_establishment_period(struct
> intel_dp *intel_dp,
>  	if (intel_crtc_has_type(crtc_state, INTEL_OUTPUT_EDP)) {
>  		tcds = (7 + DIV_ROUND_UP(6500, tml_phy_lock) + 1) *
> tml_phy_lock;
>  	} else {
> +		int idx = get_link_rate_index(crtc_state-
> >port_clock);
> +
>  		tcds = 7 * tml_phy_lock;
>  		lttpr_count = drm_dp_lttpr_count(intel_dp-
> >lttpr_common_caps);
> +
> +		if (idx != LINK_RATE_INVALID && lttpr_count >= 0 &&
> +		    lttpr_count < DEFAULT_MAX_LTTPR_COUNT &&
> +		    establishment_periods[lttpr_count][idx]) {
> +			establishment_period =
> establishment_periods[lttpr_count][idx];
> +			return establishment_period;
> +		}
>  	}
>  
>  	if (lttpr_count > 0) {