[PATCH] drm/amdgpu: avoid ras error injection for retired page

[Zhou1, Tao]

> -----Original Message-----
> From: Chen, Guchun <Guchun.Chen at amd.com>
> Sent: 2019年9月30日 15:14
> To: Zhou1, Tao <Tao.Zhou1 at amd.com>; amd-gfx at lists.freedesktop.org;
> Zhang, Hawking <Hawking.Zhang at amd.com>
> Subject: RE: [PATCH] drm/amdgpu: avoid ras error injection for retired page
> 
> 
> 
> 
> Regards,
> Guchun
> 
> -----Original Message-----
> From: Zhou1, Tao <Tao.Zhou1 at amd.com>
> Sent: Monday, September 30, 2019 2:58 PM
> To: amd-gfx at lists.freedesktop.org; Chen, Guchun
> <Guchun.Chen at amd.com>; Zhang, Hawking <Hawking.Zhang at amd.com>
> Cc: Zhou1, Tao <Tao.Zhou1 at amd.com>
> Subject: [PATCH] drm/amdgpu: avoid ras error injection for retired page
> 
> check whether a page is bad page before error injection
> 
> Signed-off-by: Tao Zhou <tao.zhou1 at amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c | 38
> +++++++++++++++++++++++++
>  1 file changed, 38 insertions(+)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> index fe3a57e567c8..d50e565b0b20 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
> @@ -71,6 +71,9 @@ const char *ras_block_string[] = {
> 
>  atomic_t amdgpu_ras_in_intr = ATOMIC_INIT(0);
> 
> +static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
> +				uint64_t addr);
> +
>  static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
>  					size_t size, loff_t *pos)
>  {
> @@ -290,6 +293,13 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct
> file *f, const char __user *
>  			break;
>  		}
> 
> +		/* ce/ue error injection for a bad page is not allowed */
> +		if (amdgpu_ras_check_bad_page(adev, data.inject.address))
> {
> +			DRM_WARN("DRM WARN: 0x%llx has been marked
> as bad before error injection!\n",
> +					data.inject.address);
> +			break;
> +		}
> +
>  		/* data.inject.address is offset instead of absolute gpu
> address */
>  		ret = amdgpu_ras_error_inject(adev, &data.inject);
>  		break;
> @@ -1430,6 +1440,34 @@ static int amdgpu_ras_load_bad_pages(struct
> amdgpu_device *adev)
>  	return ret;
>  }
> 
> +/* check if an address belongs to bad page */ static bool
> +amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
> +				uint64_t addr)
> +{
> +	struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
> +	struct ras_err_handler_data *data;
> +	int i, ret = false;
> [Guchun]It's better to use bool type for the ret variable, to keep consistent
> with function return type?
> Apart from that, this patch is: Reviewed-by: Guchun Chen
> <guchun.chen at amd.com>

[Tao] Thanks, I'll correct it before submit.

> 
> +
> +	if (!con || !con->eh_data)
> +		return ret;
> +
> +	mutex_lock(&con->recovery_lock);
> +	data = con->eh_data;
> +	if (!data)
> +		goto out;
> +
> +	addr >>= AMDGPU_GPU_PAGE_SHIFT;
> +	for (i = 0; i < data->count; i++)
> +		if (addr == data->bps[i].retired_page) {
> +			ret = true;
> +			goto out;
> +		}
> +
> +out:
> +	mutex_unlock(&con->recovery_lock);
> +	return ret;
> +}
> +
>  static void amdgpu_ras_create_bad_pages_bo(struct amdgpu_device *adev)
> {
>  	/* Note: the caller should guarantee con and data are not NULL */
> --
> 2.17.1