Re: [Freedreno] [PATCH 3/3] drm/msm/dpu: Add interface support for CRC debugfs

[Jessica Zhang <quic_jesszhan@quicinc.com>]

On 6/2/2022 3:51 PM, Dmitry Baryshkov wrote:
> On 28/05/2022 01:23, Jessica Zhang wrote:
>>
>>
>> On 5/27/2022 12:46 PM, Dmitry Baryshkov wrote:
>>> On 27/05/2022 21:54, Jessica Zhang wrote:
>>>> Add support for writing CRC values for the interface block to
>>>> the debugfs by calling the necessary MISR setup/collect methods.
>>>>
>>>> Signed-off-by: Jessica Zhang <quic_jesszhan@quicinc.com>
>>>> ---
>>>>   drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c    | 43 ++++++++++++++-
>>>>   drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h    |  3 +
>>>>   drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 61 
>>>> +++++++++++++++++++++
>>>>   drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.h | 22 ++++++++
>>>>   4 files changed, 128 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c 
>>>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
>>>> index ae09466663cf..e830fb1e910d 100644
>>>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
>>>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c
>>>> @@ -79,6 +79,8 @@ static enum dpu_crtc_crc_source 
>>>> dpu_crtc_parse_crc_source(const char *src_name)
>>>>       if (!strcmp(src_name, "auto") ||
>>>>           !strcmp(src_name, "lm"))
>>>>           return DPU_CRTC_CRC_SOURCE_LAYER_MIXER;
>>>> +    if (!strcmp(src_name, "intf"))
>>>> +        return DPU_CRTC_CRC_SOURCE_INTF;
>>>>       return DPU_CRTC_CRC_SOURCE_INVALID;
>>>>   }
>>>> @@ -94,8 +96,18 @@ static int dpu_crtc_verify_crc_source(struct 
>>>> drm_crtc *crtc,
>>>>           return -EINVAL;
>>>>       }
>>>> -    if (source == DPU_CRTC_CRC_SOURCE_LAYER_MIXER)
>>>> +    if (source == DPU_CRTC_CRC_SOURCE_LAYER_MIXER) {
>>>>           *values_cnt = crtc_state->num_mixers;
>>>> +    } else if (source == DPU_CRTC_CRC_SOURCE_INTF) {
>>>> +        struct drm_encoder *drm_enc = get_encoder_from_crtc(crtc);
>>>
>>> Let's do this correctly from the beginning. The CRTC can drive 
>>> several encoders. Do we want to get CRC from all of them or just from 
>>> the first one?
>>
>> In the case of multiple encoders, it would be better to collect CRCs 
>> from all of them.
> 
> Then this should become a loop.

Ah, I see what you mean. Yea, I can do that.

> 
>>
>>>
>>>> +
>>>> +        if (!drm_enc) {
>>>> +            DRM_ERROR("no encoder found for crtc %d\n", crtc->index);
>>>> +            return -ENODATA;
>>>> +        }
>>>> +
>>>> +        *values_cnt = dpu_encoder_get_num_phys(drm_enc);
>>>> +    }
>>>>       return 0;
>>>>   }
>>>> @@ -116,6 +128,18 @@ static void dpu_crtc_setup_lm_misr(struct 
>>>> dpu_crtc_state *crtc_state)
>>>>       }
>>>>   }
>>>> +static void dpu_crtc_setup_encoder_misr(struct drm_crtc *crtc)
>>>> +{
>>>> +    struct drm_encoder *drm_enc = get_encoder_from_crtc(crtc);
>>>> +
>>>> +    if (!drm_enc) {
>>>> +        DRM_ERROR("no encoder found for crtc %d\n", crtc->index);
>>>> +        return;
>>>> +    }
>>>> +
>>>> +    dpu_encoder_setup_misr(drm_enc);
>>>> +}
>>>> +
>>>>   static int dpu_crtc_set_crc_source(struct drm_crtc *crtc, const 
>>>> char *src_name)
>>>>   {
>>>>       enum dpu_crtc_crc_source source = 
>>>> dpu_crtc_parse_crc_source(src_name);
>>>> @@ -164,6 +188,8 @@ static int dpu_crtc_set_crc_source(struct 
>>>> drm_crtc *crtc, const char *src_name)
>>>>       if (source == DPU_CRTC_CRC_SOURCE_LAYER_MIXER)
>>>>           dpu_crtc_setup_lm_misr(crtc_state);
>>>> +    else if (source == DPU_CRTC_CRC_SOURCE_INTF)
>>>> +        dpu_crtc_setup_encoder_misr(crtc);
>>>>   cleanup:
>>>>       drm_modeset_unlock(&crtc->mutex);
>>>> @@ -212,6 +238,18 @@ static int dpu_crtc_get_lm_crc(struct drm_crtc 
>>>> *crtc, struct dpu_crtc_state *crt
>>>>               drm_crtc_accurate_vblank_count(crtc), crcs);
>>>>   }
>>>> +static int dpu_crtc_get_encoder_crc(struct drm_crtc *crtc)
>>>> +{
>>>> +    struct drm_encoder *drm_enc =  get_encoder_from_crtc(crtc);
>>>> +
>>>> +    if (!drm_enc) {
>>>> +        DRM_ERROR("no encoder found for crtc %d\n", crtc->index);
>>>> +        return -EINVAL;
>>>> +    }
>>>> +
>>>> +    return dpu_encoder_get_crc(drm_enc);
>>>> +}
>>>> +
>>>>   static int dpu_crtc_get_crc(struct drm_crtc *crtc)
>>>>   {
>>>>       struct dpu_crtc_state *crtc_state;
>>>> @@ -227,6 +265,9 @@ static int dpu_crtc_get_crc(struct drm_crtc *crtc)
>>>>       if (crtc_state->crc_source == DPU_CRTC_CRC_SOURCE_LAYER_MIXER)
>>>>           return dpu_crtc_get_lm_crc(crtc, crtc_state);
>>>> +    if (crtc_state->crc_source == DPU_CRTC_CRC_SOURCE_INTF)
>>>> +        return dpu_crtc_get_encoder_crc(crtc);
>>>> +
>>>>       return 0;
>>>>   }
>>>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h 
>>>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h
>>>> index b8785c394fcc..a60af034905d 100644
>>>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h
>>>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.h
>>>> @@ -1,5 +1,6 @@
>>>>   /* SPDX-License-Identifier: GPL-2.0-only */
>>>>   /*
>>>> + * Copyright (c) 2022 Qualcomm Innovation Center, Inc. All rights 
>>>> reserved.
>>>>    * Copyright (c) 2015-2021 The Linux Foundation. All rights reserved.
>>>>    * Copyright (C) 2013 Red Hat
>>>>    * Author: Rob Clark <robdclark@gmail.com>
>>>> @@ -73,11 +74,13 @@ struct dpu_crtc_smmu_state_data {
>>>>    * enum dpu_crtc_crc_source: CRC source
>>>>    * @DPU_CRTC_CRC_SOURCE_NONE: no source set
>>>>    * @DPU_CRTC_CRC_SOURCE_LAYER_MIXER: CRC in layer mixer
>>>> + * @DPU_CRTC_CRC_SOURCE_INTF: CRC in phys interface
>>>>    * @DPU_CRTC_CRC_SOURCE_INVALID: Invalid source
>>>>    */
>>>>   enum dpu_crtc_crc_source {
>>>>       DPU_CRTC_CRC_SOURCE_NONE = 0,
>>>>       DPU_CRTC_CRC_SOURCE_LAYER_MIXER,
>>>> +    DPU_CRTC_CRC_SOURCE_INTF,
>>>>       DPU_CRTC_CRC_SOURCE_MAX,
>>>>       DPU_CRTC_CRC_SOURCE_INVALID = -1
>>>>   };
>>>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c 
>>>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>>> index 52516eb20cb8..7740515f462d 100644
>>>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>>> @@ -14,6 +14,7 @@
>>>>   #include <drm/drm_crtc.h>
>>>>   #include <drm/drm_file.h>
>>>> +#include <drm/drm_vblank.h>
>>>>   #include <drm/drm_probe_helper.h>
>>>>   #include "msm_drv.h"
>>>> @@ -225,6 +226,66 @@ bool dpu_encoder_is_widebus_enabled(const 
>>>> struct drm_encoder *drm_enc)
>>>>       return dpu_enc->wide_bus_en;
>>>>   }
>>>> +int dpu_encoder_get_num_phys(const struct drm_encoder *drm_enc)
>>>> +{
>>>> +    struct dpu_encoder_virt *dpu_enc;
>>>> +
>>>> +    dpu_enc = to_dpu_encoder_virt(drm_enc);
>>>> +
>>>> +    return dpu_enc->num_phys_encs;
>>>> +}
>>>> +
>>>> +void dpu_encoder_setup_misr(const struct drm_encoder *drm_enc)
>>>> +{
>>>> +    struct dpu_encoder_virt *dpu_enc;
>>>> +
>>>> +    int i;
>>>> +
>>>> +    dpu_enc = to_dpu_encoder_virt(drm_enc);
>>>> +
>>>> +    for (i = 0; i < dpu_enc->num_phys_encs; i++) {
>>>> +        struct dpu_encoder_phys *phys = dpu_enc->phys_encs[i];
>>>> +
>>>> +        if (!phys->hw_intf || !phys->hw_intf->ops.setup_misr)
>>>> +            continue;
>>>
>>> Does WB support CRC?
>>
>> AFAIK, no.
>>
>>>
>>>> +
>>>> +        phys->hw_intf->ops.setup_misr(phys->hw_intf, true, 1);
>>>> +    }
>>>> +}
>>>> +
>>>> +int dpu_encoder_get_crc(const struct drm_encoder *drm_enc)
>>>> +{
>>>> +    struct dpu_encoder_virt *dpu_enc;
>>>> +    u32 crcs[MAX_PHYS_ENCODERS_PER_VIRTUAL];
>>>> +
>>>> +    int i, rc;
>>>> +
>>>> +    if (!drm_enc->crtc) {
>>>> +        DRM_ERROR("no crtc found for encoder %d\n", drm_enc->index);
>>>> +        return -EINVAL;
>>>> +    }
>>>> +
>>>> +    dpu_enc = to_dpu_encoder_virt(drm_enc);
>>>> +
>>>> +    for (i = 0; i < dpu_enc->num_phys_encs; i++) {
>>>> +        struct dpu_encoder_phys *phys = dpu_enc->phys_encs[i];
>>>> +
>>>> +        if (!phys->hw_intf || !phys->hw_intf->ops.collect_misr)
>>>> +            continue;
>>>> +
>>>> +        rc = phys->hw_intf->ops.collect_misr(phys->hw_intf, &crcs[i]);
>>>
>>> This doesn't look fully correct. Do we need to skip the indices for 
>>> the phys without a backing hw_intf?
>>
>> Sorry if I'm misunderstanding your question, but don't we need to have 
>> a backing hw_intf (and skip if there isn't any) since the methods for 
>> collecting/setting MISR registers is within the hw_intf?
> 
> Yes. So the question if we should skip the phys and leave the crcs[i] 
> untouched, skip the phys and sset crcs[i] to 0 or change 
> dpu_crtc_parse_crc_source() to return the number of intf-backed 
> phys_enc's and do not skip any crcs[i].

Thanks for the clarification.

Is it possible to hit a case where a phys_encoder won't have a 
corresponding hw_intf?

AFAIK, it seems guaranteed that a phys_encoder will have an hw_intf 
since dpu_encoder_setup_display will skip incrementing num_phys_encs if 
dpu_encoder_get_intf fails [1].

[1] 
https://gitlab.freedesktop.org/drm/msm/-/blob/msm-next/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c#L2263

Thanks,

Jessica Zhang

> 
> 
>>
>>>
>>>> +
>>>
>>> Extra empty line.
>>
>> Noted
>>
>>>
>>>> +        if (rc) {
>>>> +            if (rc != -ENODATA)
>>>
>>> Do we need to handle ENODATA in any specific way (like zeroing the 
>>> crcs[i])? If not, I'd suggest to drop this return code. Let's make an 
>>> error always an error.
>>
>> This is a carry-over from this change [1]. We wanted to have the 
>> ENODATA check to avoid spamming the driver debug logs when CRC is 
>> disabled for this block.
>>
>> [1] 
>> https://gitlab.freedesktop.org/drm/msm/-/commit/3ce8bdca394fc606b55e7c5ed779d171aaae5d09 
> 
> 
> Thanks for the reminder. I commented this in the previous patch now.
> 
> 
> 
> -- 
> With best wishes
> Dmitry