[Intel-gfx] [PATCH] drm/i915: Fix errors when there is no free DSM

[Ville Syrjälä]

On Fri, Jan 21, 2022 at 11:24:50AM -0800, José Roberto de Souza wrote:
> Part of DSM(data stolen memory) is reserved for HW functions and GUC
> and in some platform this reserved block takes the whole DSM leaving
> no memory to allocated by i915_gem_object_create_stolen() and friends.
> 
> In such cases i915_gem_init_stolen() was not calling drm_mm_init()
> and returning 0, causing errors later when testing memory
> (intel_memory_region_memtest()) and in intel_memory_regions_hw_probe()
> saying setup of memory region failed:

I think you might that a bit backwards. There are a lot of other cases
where don't initialize stolen and still return 0 here. But in those cases
we *don't* call drm_mm_init() and everyone just uses drm_mm_initialized()
to check whether stolen was actually initialized or not.

> 
> checking generic (91000000 300000) vs hw (23fe7e000000 2000000)
> checking generic (91000000 300000) vs hw (237800000000 800000000)
> i915 0000:8c:00.0: [drm:i915_gem_init_stolen [i915]] GEN6_STOLEN_RESERVED = 0x00000003ff800185
> i915 0000:8c:00.0: [drm:i915_gem_init_stolen [i915]] Memory reserved for graphics device: 8192K, usable: 0K
> i915 0000:8c:00.0: Failed to read back from memory region:[mem 0x3ff800000-0x3ff7fffff] at [0x0000237bff800000 + 0x0000000000000000] for i915_gem_stolen_lmem_setup [i915]; wrote 0, read (ff, ff, ff)
> i915 0000:8c:00.0: [drm] *ERROR* Failed to setup region 6 (type=3:0), error -22
> [drm:intel_gt_setup_lmem [i915]] LMEM: debug trace data region: [0x0-0x2000000]
> i915 0000:8c:00.0: [drm:intel_gt_setup_lmem [i915]] Local memory: [mem 0x00000000-0x3faffffff]
> 
> Signed-off-by: José Roberto de Souza <jose.souza at intel.com>
> ---
>  drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 9 ++++++---
>  drivers/gpu/drm/i915/intel_memory_region.c | 3 +++
>  2 files changed, 9 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> index 26975d8577760..6e90357b2d1fd 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c
> @@ -495,13 +495,16 @@ static int i915_gem_init_stolen(struct intel_memory_region *mem)
>  	 * memory, so just consider the start. */
>  	reserved_total = stolen_top - reserved_base;
>  
> +	i915->stolen_usable_size =
> +		resource_size(&i915->dsm) - reserved_total;
> +
>  	drm_dbg(&i915->drm,
>  		"Memory reserved for graphics device: %lluK, usable: %lluK\n",
>  		(u64)resource_size(&i915->dsm) >> 10,
> -		((u64)resource_size(&i915->dsm) - reserved_total) >> 10);
> +		(u64)i915->stolen_usable_size >> 10);
>  
> -	i915->stolen_usable_size =
> -		resource_size(&i915->dsm) - reserved_total;
> +	if (i915->stolen_usable_size == 0)
> +		return -ENOMEM;
>  
>  	/* Basic memrange allocator for stolen space. */
>  	drm_mm_init(&i915->mm.stolen, 0, i915->stolen_usable_size);
> diff --git a/drivers/gpu/drm/i915/intel_memory_region.c b/drivers/gpu/drm/i915/intel_memory_region.c
> index c70d7e286a512..317d67fa3a36e 100644
> --- a/drivers/gpu/drm/i915/intel_memory_region.c
> +++ b/drivers/gpu/drm/i915/intel_memory_region.c
> @@ -324,6 +324,9 @@ int intel_memory_regions_hw_probe(struct drm_i915_private *i915)
>  
>  		if (IS_ERR(mem)) {
>  			err = PTR_ERR(mem);
> +			if (err == -ENOMEM)
> +				continue;
> +
>  			drm_err(&i915->drm,
>  				"Failed to setup region(%d) type=%d\n",
>  				err, type);
> -- 
> 2.34.1

-- 
Ville Syrjälä
Intel