[PATCH] drm/amdgpu/mes: keep enforce isolation up to date

[Alex Deucher]

On Mon, Feb 17, 2025 at 9:18 AM SRINIVASAN SHANMUGAM
<srinivasan.shanmugam at amd.com> wrote:
>
>
> On 2/17/2025 7:44 PM, Alex Deucher wrote:
> > On Sat, Feb 15, 2025 at 3:02 AM SRINIVASAN SHANMUGAM
> > <srinivasan.shanmugam at amd.com> wrote:
> >>
> >> On 2/14/2025 11:05 PM, Alex Deucher wrote:
> >>
> >> Re-send the mes message on resume to make sure the
> >> mes state is up to date.
> >>
> >> Fixes: 8521e3c5f058 ("drm/amd/amdgpu: limit single process inside MES")
> >> Signed-off-by: Alex Deucher <alexander.deucher at amd.com>
> >> Cc: Shaoyun Liu <shaoyun.liu at amd.com>
> >> Cc: Srinivasan Shanmugam <srinivasan.shanmugam at amd.com>
> >> ---
> >>   drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 13 ++++---------
> >>   drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c | 20 +++++++++++++++++++-
> >>   drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h |  2 +-
> >>   drivers/gpu/drm/amd/amdgpu/mes_v11_0.c  |  4 ++++
> >>   drivers/gpu/drm/amd/amdgpu/mes_v12_0.c  |  4 ++++
> >>   5 files changed, 32 insertions(+), 11 deletions(-)
> >>
> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> >> index b9bd6654f3172..a194bf3347cbc 100644
> >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
> >> @@ -1665,24 +1665,19 @@ static ssize_t amdgpu_gfx_set_enforce_isolation(struct device *dev,
> >>    }
> >>
> >>    mutex_lock(&adev->enforce_isolation_mutex);
> >> -
> >>    for (i = 0; i < num_partitions; i++) {
> >> - if (adev->enforce_isolation[i] && !partition_values[i]) {
> >> + if (adev->enforce_isolation[i] && !partition_values[i])
> >>    /* Going from enabled to disabled */
> >>    amdgpu_vmid_free_reserved(adev, AMDGPU_GFXHUB(i));
> >> - if (adev->enable_mes && adev->gfx.enable_cleaner_shader)
> >> - amdgpu_mes_set_enforce_isolation(adev, i, false);
> >> - } else if (!adev->enforce_isolation[i] && partition_values[i]) {
> >> + else if (!adev->enforce_isolation[i] && partition_values[i])
> >>    /* Going from disabled to enabled */
> >>    amdgpu_vmid_alloc_reserved(adev, AMDGPU_GFXHUB(i));
> >> - if (adev->enable_mes && adev->gfx.enable_cleaner_shader)
> >> - amdgpu_mes_set_enforce_isolation(adev, i, true);
> >> - }
> >>    adev->enforce_isolation[i] = partition_values[i];
> >>    }
> >> -
> >>    mutex_unlock(&adev->enforce_isolation_mutex);
> >>
> >> + amdgpu_mes_update_enforce_isolation(adev);
> >> +
> >>    return count;
> >>   }
> >>
> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
> >> index cee38bb6cfaf2..ca076306adba4 100644
> >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
> >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
> >> @@ -1508,7 +1508,8 @@ bool amdgpu_mes_suspend_resume_all_supported(struct amdgpu_device *adev)
> >>   }
> >>
> >>   /* Fix me -- node_id is used to identify the correct MES instances in the future */
> >> -int amdgpu_mes_set_enforce_isolation(struct amdgpu_device *adev, uint32_t node_id, bool enable)
> >> +static int amdgpu_mes_set_enforce_isolation(struct amdgpu_device *adev,
> >> +    uint32_t node_id, bool enable)
> >>   {
> >>    struct mes_misc_op_input op_input = {0};
> >>    int r;
> >> @@ -1530,6 +1531,23 @@ int amdgpu_mes_set_enforce_isolation(struct amdgpu_device *adev, uint32_t node_i
> >>    return r;
> >>   }
> >>
> >> +int amdgpu_mes_update_enforce_isolation(struct amdgpu_device *adev)
> >> +{
> >> + int i, r = 0;
> >> +
> >> + if (adev->enable_mes && adev->gfx.enable_cleaner_shader) {
> >> + mutex_lock(&adev->enforce_isolation_mutex);
> >> + for (i = 0; i < (adev->xcp_mgr ? adev->xcp_mgr->num_xcps : 1); i++) {
> >> + if (adev->enforce_isolation[i])
> >> + r |= amdgpu_mes_set_enforce_isolation(adev, i, true);
> >> + else
> >> + r |= amdgpu_mes_set_enforce_isolation(adev, i, false);
> >> + }
> >> + mutex_unlock(&adev->enforce_isolation_mutex);
> >> + }
> >> + return r;
> >> +}
> >> +
> >>   #if defined(CONFIG_DEBUG_FS)
> >>
> >>   static int amdgpu_debugfs_mes_event_log_show(struct seq_file *m, void *unused)
> >> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
> >> index 6a792ffc81e33..3a65c3788956d 100644
> >> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
> >> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
> >> @@ -532,6 +532,6 @@ static inline void amdgpu_mes_unlock(struct amdgpu_mes *mes)
> >>
> >>   bool amdgpu_mes_suspend_resume_all_supported(struct amdgpu_device *adev);
> >>
> >> -int amdgpu_mes_set_enforce_isolation(struct amdgpu_device *adev, uint32_t node_id, bool enable);
> >> +int amdgpu_mes_update_enforce_isolation(struct amdgpu_device *adev);
> >>
> >>   #endif /* __AMDGPU_MES_H__ */
> >> diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
> >> index 530371e6a7aee..fc7b17463cb4d 100644
> >> --- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
> >> +++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c
> >> @@ -1660,6 +1660,10 @@ static int mes_v11_0_hw_init(struct amdgpu_ip_block *ip_block)
> >>    goto failure;
> >>    }
> >>
> >> + r = amdgpu_mes_update_enforce_isolation(adev);
> >> + if (r)
> >> + goto failure;
> >> +
> >>
> >> Hi Alex,
> >>
> >> Should this also be moved to mes_v11_0_hw_init. Please let me know your thoughts?
> > I'm not sure I follow.  This is in hw_init.
> >
> > Alex
>
> Sorry, my mistake mes_v11_0_sw_init pls?

There's no need to call it in sw_init, plus the hw is not set up in
sw_init so you can't call it there anyway.  The whole point of this is
to update the firmware with the current sw state after a suspend or
reset.

Alex

>
> Thanks!
>
> Srini
>
> >>   out:
> >>    /*
> >>    * Disable KIQ ring usage from the driver once MES is enabled.
> >> diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
> >> index 6db88584dd529..ec91c78468f30 100644
> >> --- a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
> >> +++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c
> >> @@ -1773,6 +1773,10 @@ static int mes_v12_0_hw_init(struct amdgpu_ip_block *ip_block)
> >>    goto failure;
> >>    }
> >>
> >> + r = amdgpu_mes_update_enforce_isolation(adev);
> >> + if (r)
> >> + goto failure;
> >> +
> >>
> >> And Similarly here also?
> >>
> >> Thanks!
> >>
> >> Srini
> >>
> >>   out:
> >>    /*
> >>    * Disable KIQ ring usage from the driver once MES is enabled.