[PATCH] drm/amdgpu: Add recovery_lock to save bad pages function

[Li, Candice]

[Public]

Thanks for the review, Tao. Updated the position for unlocking.

Fix race condition failure during UMC UE injection.

Signed-off-by: Candice Li <candice.li at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 08133de21fdd63..53b957a5b9a65c 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -1935,9 +1935,11 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev)
        if (!con || !con->eh_data)
                return 0;

+       mutex_lock(&con->recovery_lock);
        control = &con->eeprom_control;
        data = con->eh_data;
        save_count = data->count - control->ras_num_recs;
+       mutex_unlock(&con->recovery_lock);
        /* only new entries are saved */
        if (save_count > 0) {
                if (amdgpu_ras_eeprom_append(control,
--
2.17.1



Thanks,
Candice

-----Original Message-----
From: Zhou1, Tao <Tao.Zhou1 at amd.com> 
Sent: Tuesday, November 16, 2021 4:27 PM
To: Li, Candice <Candice.Li at amd.com>; amd-gfx at lists.freedesktop.org
Cc: Clements, John <John.Clements at amd.com>
Subject: RE: [PATCH] drm/amdgpu: Add recovery_lock to save bad pages function

[AMD Official Use Only]

> -----Original Message-----
> From: Li, Candice <Candice.Li at amd.com>
> Sent: Tuesday, November 16, 2021 4:02 PM
> To: amd-gfx at lists.freedesktop.org
> Cc: Clements, John <John.Clements at amd.com>; Zhou1, Tao
> <Tao.Zhou1 at amd.com>; Li, Candice <Candice.Li at amd.com>
> Subject: [PATCH] drm/amdgpu: Add recovery_lock to save bad pages function
>
> Fix race condition failure during UMC UE injection.
>
> Signed-off-by: Candice Li <candice.li at amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> index 08133de21fdd63..711b5fb26d47d4 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> @@ -1931,10 +1931,12 @@ int amdgpu_ras_save_bad_pages(struct
> amdgpu_device *adev)
>       struct ras_err_handler_data *data;
>       struct amdgpu_ras_eeprom_control *control;
>       int save_count;
> +     int ret = 0;
>
>       if (!con || !con->eh_data)
>               return 0;
>
> +     mutex_lock(&con->recovery_lock);
>       control = &con->eeprom_control;
>       data = con->eh_data;
>       save_count = data->count - control->ras_num_recs; @@ -1944,13

[Tao] Since recovery_lock is dedicated to protecting eh_data, can we unlock it here?

> +1946,16 @@ int amdgpu_ras_save_bad_pages(struct amdgpu_device *adev)
>                                            &data->bps[control->ras_num_recs],
>                                            save_count)) {
>                       dev_err(adev->dev, "Failed to save EEPROM table
> data!");
> -                     return -EIO;
> +                     ret = -EIO;
> +                     goto out;
>               }
>
>               dev_info(adev->dev, "Saved %d pages to EEPROM table.\n",
> save_count);
>       }
>
> -     return 0;
> +out:
> +     mutex_unlock(&con->recovery_lock);
> +     return ret;
>  }
>
>  /*
> --
> 2.17.1