[PATCH v2] drm/amdgpu: Need to set the baco cap before baco reset

Deng, Emily Emily.Deng at amd.com
Thu May 23 05:07:35 UTC 2019


Sorry, I have pushed the change as Evan gave the reviewed-by, I will send another patch to reference your review comments, do you think it is Ok?
>-----Original Message-----
>From: Alex Deucher <alexdeucher at gmail.com>
>Sent: Thursday, May 23, 2019 11:54 AM
>To: Deng, Emily <Emily.Deng at amd.com>
>Cc: amd-gfx list <amd-gfx at lists.freedesktop.org>
>Subject: Re: [PATCH v2] drm/amdgpu: Need to set the baco cap before baco
>reset
>
>[CAUTION: External Email]
>
>On Wed, May 22, 2019 at 11:48 PM Alex Deucher <alexdeucher at gmail.com>
>wrote:
>>
>> On Wed, May 22, 2019 at 11:27 PM Emily Deng <Emily.Deng at amd.com>
>wrote:
>> >
>>
>> Please include a patch description.
The comment is lost when using "git send-email", I will check why.
>>
>> > Signed-off-by: Emily Deng <Emily.Deng at amd.com>
>> > ---
>> >  drivers/gpu/drm/amd/amdgpu/amdgpu_device.c         | 10 +++++++++-
>> >  drivers/gpu/drm/amd/include/kgd_pp_interface.h     |  1 +
>> >  drivers/gpu/drm/amd/powerplay/amd_powerplay.c      | 16
>++++++++++++++++
>> >  drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c |  1 +
>> >  .../amd/powerplay/hwmgr/vega10_processpptables.c   | 22
>++++++++++++++++++++++
>> >  .../amd/powerplay/hwmgr/vega10_processpptables.h   |  1 +
>> >  drivers/gpu/drm/amd/powerplay/inc/hwmgr.h          |  1 +
>> >  7 files changed, 51 insertions(+), 1 deletion(-)
>> >
>> > diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> > b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> > index d6286ed..5288763 100644
>> > --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> > +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
>> > @@ -2605,7 +2605,15 @@ int amdgpu_device_init(struct amdgpu_device
>*adev,
>> >         /* check if we need to reset the asic
>> >          *  E.g., driver was not cleanly unloaded previously, etc.
>> >          */
>> > -       if (!amdgpu_sriov_vf(adev) &&
>amdgpu_asic_need_reset_on_init(adev)) {
>> > +       if (amdgpu_passthrough(adev) &&
>> > + amdgpu_asic_need_reset_on_init(adev)) {
>>
>> This will change the current behavior on baremetal.
Ok, I will put the passthrough check to "if(adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->set_asic_baco_cap)"
>>
>> > +               if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs-
>>set_asic_baco_cap) {
>> > +                       r = adev->powerplay.pp_funcs->set_asic_baco_cap(adev-
>>powerplay.pp_handle);
>> > +                       if (r) {
>> > +                               dev_err(adev->dev, "set baco capability failed\n");
>> > +                               goto failed;
>> > +                       }
>> > +               }
>> > +
>>
>> This will also change the current behavior on bare metal.
>>
>> I think you may want to rework this a bit otherwise this change won't
>> really make any difference due to this patch:
>> https://cgit.freedesktop.org/~agd5f/linux/commit/?h=amd-staging-drm-ne
>> xt&id=60ae2cd5aec94dc6459bdee5c610bb5c76a1d0ae
>> We need to avoid breaking module reload (modeprobe amdgpu; modprobe
>-r
>> amdgpu; modeprobe amdgpu).
>> I think it would be cleaner to call set_asic_baco_cap() in
>> hwmgr_early_init() and then return true in soc15_need_reset_on_init()
>> is it's passthrough mode.  then everything should just work as is.
Ok, will call set_asic_baco_cap in vega10_hwmgr_init
>Assuming you set the set_asic_baco_cap callbacks for vega20 and vega12 as
>well. Otherwise, you'll have to limit it to vega10 which starts to get messy
>IMHO.
I don't add the callbacks for vega20 and vega12, as no platform to test these. And if vega20 or vega12 need this, then they could add this and test this.
Currently only for vega10.
>Alex
>
>>
>> >                 r = amdgpu_asic_reset(adev);
>> >                 if (r) {
>> >                         dev_err(adev->dev, "asic reset on init
>> > failed\n"); diff --git
>> > a/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>> > b/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>> > index 2b579ba..0dcc18d 100644
>> > --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>> > +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h
>> > @@ -285,6 +285,7 @@ struct amd_pm_funcs {
>> >         int (*set_hard_min_fclk_by_freq)(void *handle, uint32_t clock);
>> >         int (*set_min_deep_sleep_dcefclk)(void *handle, uint32_t clock);
>> >         int (*get_asic_baco_capability)(void *handle, bool *cap);
>> > +       int (*set_asic_baco_cap)(void *handle);
>> >         int (*get_asic_baco_state)(void *handle, int *state);
>> >         int (*set_asic_baco_state)(void *handle, int state);
>> >         int (*get_ppfeature_status)(void *handle, char *buf); diff
>> > --git a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>> > b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>> > index bea1587..9856760 100644
>> > --- a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>> > +++ b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c
>> > @@ -1404,6 +1404,21 @@ static int pp_set_active_display_count(void
>*handle, uint32_t count)
>> >         return ret;
>> >  }
>> >
>> > +static int pp_set_asic_baco_cap(void *handle) {
>> > +       struct pp_hwmgr *hwmgr = handle;
>> > +
>> > +       if (!hwmgr)
>> > +               return -EINVAL;
>> > +
>> > +       if (!hwmgr->pm_en || !hwmgr->hwmgr_func->set_asic_baco_cap)
>> > +               return 0;
>> > +
>> > +       hwmgr->hwmgr_func->set_asic_baco_cap(hwmgr);
>> > +
>> > +       return 0;
>> > +}
>> > +
>> >  static int pp_get_asic_baco_capability(void *handle, bool *cap)  {
>> >         struct pp_hwmgr *hwmgr = handle; @@ -1546,6 +1561,7 @@
>> > static const struct amd_pm_funcs pp_dpm_funcs = {
>> >         .set_hard_min_dcefclk_by_freq = pp_set_hard_min_dcefclk_by_freq,
>> >         .set_hard_min_fclk_by_freq = pp_set_hard_min_fclk_by_freq,
>> >         .get_asic_baco_capability = pp_get_asic_baco_capability,
>> > +       .set_asic_baco_cap = pp_set_asic_baco_cap,
>> >         .get_asic_baco_state = pp_get_asic_baco_state,
>> >         .set_asic_baco_state = pp_set_asic_baco_state,
>> >         .get_ppfeature_status = pp_get_ppfeature_status, diff --git
>> > a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>> > b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>> > index ed6c638..8dc23eb 100644
>> > --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>> > +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c
>> > @@ -5171,6 +5171,7 @@ static const struct pp_hwmgr_func
>vega10_hwmgr_funcs = {
>> >         .odn_edit_dpm_table = vega10_odn_edit_dpm_table,
>> >         .get_performance_level = vega10_get_performance_level,
>> >         .get_asic_baco_capability = smu9_baco_get_capability,
>> > +       .set_asic_baco_cap = vega10_baco_set_cap,
>>
>> Any chance you can add callbacks for vega12 and 20 as well?
Sorry, don't have this.
>> >         .get_asic_baco_state = smu9_baco_get_state,
>> >         .set_asic_baco_state = vega10_baco_set_state,
>> >         .enable_mgpu_fan_boost = vega10_enable_mgpu_fan_boost, diff
>> > --git
>a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c
>> > b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c
>> > index b6767d7..8fdeb23 100644
>> > --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c
>> > +++
>b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c
>> > @@ -1371,3 +1371,25 @@ int vega10_get_powerplay_table_entry(struct
>> > pp_hwmgr *hwmgr,
>> >
>> >         return result;
>> >  }
>> > +
>> > +int vega10_baco_set_cap(struct pp_hwmgr *hwmgr) {
>> > +       int result = 0;
>> > +       const ATOM_Vega10_POWERPLAYTABLE *powerplay_table;
>> > +
>> > +       powerplay_table = get_powerplay_table(hwmgr);
>> > +
>> > +       PP_ASSERT_WITH_CODE((powerplay_table != NULL),
>> > +               "Missing PowerPlay Table!", return -1);
>> > +
>> > +       result = check_powerplay_tables(hwmgr, powerplay_table);
>> > +
>> > +       PP_ASSERT_WITH_CODE((result == 0),
>> > +                           "check_powerplay_tables failed", return
>> > + result);
>> > +
>> > +       set_hw_cap(
>> > +                       hwmgr,
>> > +                       0 != (le32_to_cpu(powerplay_table->ulPlatformCaps) &
>ATOM_VEGA10_PP_PLATFORM_CAP_BACO),
>> > +                       PHM_PlatformCaps_BACO);
>> > +       return result;
>> > +}
>> > \ No newline at end of file
>>
>> Missing newline.
Ok, will fix this.
>> Alex
>>
>> > diff --git
>> > a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.h
>> > b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.h
>> > index d83ed2a..da5fbec 100644
>> > ---
>a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.h
>> > +++
>b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.h
>> > @@ -59,4 +59,5 @@ extern int
>> > vega10_get_number_of_powerplay_table_entries(struct pp_hwmgr
>*hwmgr);  extern int vega10_get_powerplay_table_entry(struct pp_hwmgr
>*hwmgr, uint32_t entry_index,
>> >                 struct pp_power_state *power_state, int (*call_back_func)(struct
>pp_hwmgr *, void *,
>> >                                 struct pp_power_state *, void *,
>> > uint32_t));
>> > +extern int vega10_baco_set_cap(struct pp_hwmgr *hwmgr);
>> >  #endif
>> > diff --git a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>> > b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>> > index bac3d85..14480ae 100644
>> > --- a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>> > +++ b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h
>> > @@ -339,6 +339,7 @@ struct pp_hwmgr_func {
>> >         int (*set_hard_min_dcefclk_by_freq)(struct pp_hwmgr *hwmgr,
>uint32_t clock);
>> >         int (*set_hard_min_fclk_by_freq)(struct pp_hwmgr *hwmgr,
>uint32_t clock);
>> >         int (*get_asic_baco_capability)(struct pp_hwmgr *hwmgr, bool
>> > *cap);
>> > +       int (*set_asic_baco_cap)(struct pp_hwmgr *hwmgr);
>> >         int (*get_asic_baco_state)(struct pp_hwmgr *hwmgr, enum
>BACO_STATE *state);
>> >         int (*set_asic_baco_state)(struct pp_hwmgr *hwmgr, enum
>BACO_STATE state);
>> >         int (*get_ppfeature_status)(struct pp_hwmgr *hwmgr, char
>> > *buf);
>> > --
>> > 2.7.4
>> >
>> > _______________________________________________
>> > amd-gfx mailing list
>> > amd-gfx at lists.freedesktop.org
>> > https://lists.freedesktop.org/mailman/listinfo/amd-gfx


More information about the amd-gfx mailing list