AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Lazar, Lijo" <lijo.lazar@amd.com>
To: "Xu, Feifei" <Feifei.Xu@amd.com>,
	"amd-gfx@lists.freedesktop.org" <amd-gfx@lists.freedesktop.org>
Cc: "Zhang, Hawking" <Hawking.Zhang@amd.com>,
	"Deucher, Alexander" <Alexander.Deucher@amd.com>,
	"Koenig, Christian" <Christian.Koenig@amd.com>,
	"Bhardwaj, Rajneesh" <Rajneesh.Bhardwaj@amd.com>,
	"Errabolu, Ramesh" <Ramesh.Errabolu@amd.com>
Subject: Re: [PATCH 6/7] drm/amdgpu: Check gmc requirement for reset on init
Date: Thu, 26 Sep 2024 15:27:58 +0530	[thread overview]
Message-ID: <ed560dc9-8b43-4db5-9a65-ee841d35eb8a@amd.com> (raw)
In-Reply-To: <CH2PR12MB4152398EF93FAEDC322ADBEEFE6A2@CH2PR12MB4152.namprd12.prod.outlook.com>



On 9/26/2024 2:31 PM, Xu, Feifei wrote:
> [AMD Official Use Only - AMD Internal Distribution Only]
> 
>>> +     refresh = (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) &&
>>> +               (adev->gmc.reset_flags & AMDGPU_GMC_INIT_RESET_NPS);
> 
> Is there a corner case that reloading with a different version tos and refreshing nps change co-exist?
> 

I guess you are referring to the below corner case
	1) Place NPS request
	2) Unload Driver
	3) Reinstall driver with a different TOS (possible but quite unlikely)
	4) Driver reload
	5) Driver checks TOS version first and goes for a reset
	6) reset_flag of GMC is not set, hence it doesn't refresh the NPS range.


I think changing the order in soc15_need_reset_on_init() to check for
NPS request before TOS version check will solve this.

Thanks,
Lijo

> Thanks,
> Feifei
> 
> -----Original Message-----
> From: amd-gfx <amd-gfx-bounces@lists.freedesktop.org> On Behalf Of Lijo Lazar
> Sent: Tuesday, September 24, 2024 1:57 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Zhang, Hawking <Hawking.Zhang@amd.com>; Deucher, Alexander <Alexander.Deucher@amd.com>; Koenig, Christian <Christian.Koenig@amd.com>; Bhardwaj, Rajneesh <Rajneesh.Bhardwaj@amd.com>; Errabolu, Ramesh <Ramesh.Errabolu@amd.com>
> Subject: [PATCH 6/7] drm/amdgpu: Check gmc requirement for reset on init
> 
> Add a callback to check if there is any condition detected by GMC block for reset on init. One case is if a pending NPS change request is detected. If reset is done because of NPS switch, refresh NPS info from discovery table.
> 
> Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 13 ++++++++++++-  drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h |  5 +++++
>  drivers/gpu/drm/amd/amdgpu/soc15.c      |  2 ++
>  3 files changed, 19 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
> index 21f1e65c9dc9..011fe3a847d0 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c
> @@ -1261,12 +1261,15 @@ int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev,
>         struct amdgpu_gmc_memrange *ranges;
>         int range_cnt, ret, i, j;
>         uint32_t nps_type;
> +       bool refresh;
> 
>         if (!mem_ranges)
>                 return -EINVAL;
> 
> +       refresh = (adev->init_lvl->level != AMDGPU_INIT_LEVEL_MINIMAL_XGMI) &&
> +                 (adev->gmc.reset_flags & AMDGPU_GMC_INIT_RESET_NPS);
>         ret = amdgpu_discovery_get_nps_info(adev, &nps_type, &ranges,
> -                                           &range_cnt, false);
> +                                           &range_cnt, refresh);
> 
>         if (ret)
>                 return ret;
> @@ -1392,3 +1395,11 @@ void amdgpu_gmc_prepare_nps_mode_change(struct amdgpu_device *adev)
>                         adev->dev,
>                         "NPS mode change request done, reload driver to complete the change\n");  }
> +
> +bool amdgpu_gmc_need_reset_on_init(struct amdgpu_device *adev) {
> +       if (adev->gmc.gmc_funcs->need_reset_on_init)
> +               return adev->gmc.gmc_funcs->need_reset_on_init(adev);
> +
> +       return false;
> +}
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
> index b13d6adb5efd..d4cd247fe574 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h
> @@ -78,6 +78,8 @@ enum amdgpu_memory_partition {
>          BIT(AMDGPU_NPS3_PARTITION_MODE) | BIT(AMDGPU_NPS4_PARTITION_MODE) | \
>          BIT(AMDGPU_NPS6_PARTITION_MODE) | BIT(AMDGPU_NPS8_PARTITION_MODE))
> 
> +#define AMDGPU_GMC_INIT_RESET_NPS  BIT(0)
> +
>  /*
>   * GMC page fault information
>   */
> @@ -169,6 +171,7 @@ struct amdgpu_gmc_funcs {
>         /* Request NPS mode */
>         int (*request_mem_partition_mode)(struct amdgpu_device *adev,
>                                           int nps_mode);
> +       bool (*need_reset_on_init)(struct amdgpu_device *adev);
>  };
> 
>  struct amdgpu_xgmi_ras {
> @@ -314,6 +317,7 @@ struct amdgpu_gmc {
>         const struct amdgpu_gmc_funcs   *gmc_funcs;
>         enum amdgpu_memory_partition    requested_nps_mode;
>         uint32_t supported_nps_modes;
> +       uint32_t reset_flags;
> 
>         struct amdgpu_xgmi xgmi;
>         struct amdgpu_irq_src   ecc_irq;
> @@ -468,5 +472,6 @@ int amdgpu_gmc_get_nps_memranges(struct amdgpu_device *adev,  int amdgpu_gmc_request_memory_partition(struct amdgpu_device *adev,
>                                         int nps_mode);
>  void amdgpu_gmc_prepare_nps_mode_change(struct amdgpu_device *adev);
> +bool amdgpu_gmc_need_reset_on_init(struct amdgpu_device *adev);
> 
>  #endif
> diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
> index 619933f252aa..97ca4931a7ef 100644
> --- a/drivers/gpu/drm/amd/amdgpu/soc15.c
> +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
> @@ -833,6 +833,8 @@ static bool soc15_need_reset_on_init(struct amdgpu_device *adev)
> 
>         if (amdgpu_psp_tos_reload_needed(adev))
>                 return true;
> +       if (amdgpu_gmc_need_reset_on_init(adev))
> +               return true;
>         /* Just return false for soc15 GPUs.  Reset does not seem to
>          * be necessary.
>          */
> --
> 2.25.1
> 

  reply	other threads:[~2024-09-26  9:58 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-09-24  5:56 [PATCH 0/7] Add support for dynamic NPS switch Lijo Lazar
2024-09-24  5:56 ` [PATCH 1/7] drm/amdgpu: Add option to refresh NPS data Lijo Lazar
2024-09-24 17:47   ` Bhardwaj, Rajneesh
2024-09-24  5:56 ` [PATCH 2/7] drm/amdgpu: Add PSP interface for NPS switch Lijo Lazar
2024-09-26  9:09   ` Xu, Feifei
2024-09-24  5:56 ` [PATCH 3/7] drm/amdgpu: Add gmc interface to request NPS mode Lijo Lazar
2024-09-26  9:08   ` Xu, Feifei
2024-09-24  5:56 ` [PATCH 4/7] drm/amdgpu: Add sysfs interfaces for " Lijo Lazar
2024-09-24 17:48   ` Bhardwaj, Rajneesh
2024-09-24  5:56 ` [PATCH 5/7] drm/amdgpu: Place NPS mode request on unload Lijo Lazar
2024-09-26  9:07   ` Xu, Feifei
2024-09-24  5:56 ` [PATCH 6/7] drm/amdgpu: Check gmc requirement for reset on init Lijo Lazar
2024-09-24 17:48   ` Bhardwaj, Rajneesh
2024-09-26  9:01   ` Xu, Feifei
2024-09-26  9:57     ` Lazar, Lijo [this message]
2024-09-27  3:56     ` Xu, Feifei
2024-09-27  4:39       ` Lazar, Lijo
2024-09-24  5:56 ` [PATCH 7/7] drm/amdgpu: Add NPS switch support for GC 9.4.3 Lijo Lazar
2024-09-26  9:08   ` Xu, Feifei
2024-09-26  5:09 ` [PATCH 0/7] Add support for dynamic NPS switch Lazar, Lijo

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ed560dc9-8b43-4db5-9a65-ee841d35eb8a@amd.com \
    --to=lijo.lazar@amd.com \
    --cc=Alexander.Deucher@amd.com \
    --cc=Christian.Koenig@amd.com \
    --cc=Feifei.Xu@amd.com \
    --cc=Hawking.Zhang@amd.com \
    --cc=Rajneesh.Bhardwaj@amd.com \
    --cc=Ramesh.Errabolu@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox