[PATCH v3] drm/msm: gpu: don't abuse dma_alloc for non-DMA allocations

[PATCH v3] drm/msm: gpu: don't abuse dma_alloc for non-DMA allocations

[Arnd Bergmann]

In zap_shader_load_mdt(), we pass a pointer to a phys_addr_t
into dmam_alloc_coherent, which the compiler warns about:

drivers/gpu/drm/msm/adreno/a5xx_gpu.c: In function 'zap_shader_load_mdt':
drivers/gpu/drm/msm/adreno/a5xx_gpu.c:54:50: error: passing argument 3 of 'dmam_alloc_coherent' from incompatible pointer type [-Werror=incompatible-pointer-types]

The returned DMA address is later passed on to a function that
takes a phys_addr_t, so it's clearly wrong to use the DMA
mapping interface here: the memory may be uncached, or the
address may be completely wrong if there is an IOMMU connected
to the device. What the code actually wants to do is to get
the physical address from the reserved-mem node. It goes through
the dma-mapping interfaces for obscure reasons, and this
apparently only works by chance, relying on specific bugs
in the error handling of the arm64 dma-mapping implementation.

The same problem existed in the "venus" media driver, which was
now fixed by Stanimir Varbanov after long discussions.

In order to make some progress here, I have now ported his
approach over to the adreno driver. The patch is currently
untested, and should get a good review, but it is now much
simpler than the original, and it should be obvious what
goes wrong if I made a mistake in the port.

See also: a6e2d36bf6b7 ("media: venus: don't abuse dma_alloc for non-DMA allocations")
Cc: Stanimir Varbanov <stanimir.varbanov@linaro.org>
Fixes: 7c65817e6d38 ("drm/msm: gpu: Enable zap shader for A5XX")
Acked-and-Tested-by: Jordan Crouse <jcrouse@codeaurora.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
I think we want this to be applied for 4.13, as the upstream
code that was added in the merge window is seriously broken
without it. Resending this separately now, please apply this
and "[PATCH 1/2] drm/msm: gpu: call qcom_mdt interfaces only for
ARCH_QCOM".

v3: fix typo and wrong comment found by Jordan
v2: rewrite based on Stanimir's venus patch
---
 drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 72 +++++++++++------------------------
 drivers/gpu/drm/msm/adreno/a5xx_gpu.h |  2 -
 2 files changed, 23 insertions(+), 51 deletions(-)

diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
index 1d54c76a7778..e1138f6c823b 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.c
@@ -15,7 +15,7 @@
 #include <linux/cpumask.h>
 #include <linux/qcom_scm.h>
 #include <linux/dma-mapping.h>
-#include <linux/of_reserved_mem.h>
+#include <linux/of_address.h>
 #include <linux/soc/qcom/mdt_loader.h>
 #include "msm_gem.h"
 #include "msm_mmu.h"
@@ -29,6 +29,8 @@ static void a5xx_dump(struct msm_gpu *gpu);
 static int zap_shader_load_mdt(struct device *dev, const char *fwname)
 {
 	const struct firmware *fw;
+	struct device_node *np;
+	struct resource r;
 	phys_addr_t mem_phys;
 	ssize_t mem_size;
 	void *mem_region = NULL;
@@ -37,6 +39,21 @@ static int zap_shader_load_mdt(struct device *dev, const char *fwname)
 	if (!IS_ENABLED(CONFIG_ARCH_QCOM))
 		return -EINVAL;
 
+	np = of_get_child_by_name(dev->of_node, "zap-shader");
+	if (!np)
+		return -ENODEV;
+
+	np = of_parse_phandle(np, "memory-region", 0);
+	if (!np)
+		return -EINVAL;
+
+	ret = of_address_to_resource(np, 0, &r);
+	if (ret)
+		return ret;
+
+	mem_phys = r.start;
+	mem_size = resource_size(&r);
+
 	/* Request the MDT file for the firmware */
 	ret = request_firmware(&fw, fwname, dev);
 	if (ret) {
@@ -52,7 +69,7 @@ static int zap_shader_load_mdt(struct device *dev, const char *fwname)
 	}
 
 	/* Allocate memory for the firmware image */
-	mem_region = dmam_alloc_coherent(dev, mem_size, &mem_phys, GFP_KERNEL);
+	mem_region = memremap(mem_phys, mem_size,  MEMREMAP_WC);
 	if (!mem_region) {
 		ret = -ENOMEM;
 		goto out;
@@ -70,6 +87,9 @@ static int zap_shader_load_mdt(struct device *dev, const char *fwname)
 		DRM_DEV_ERROR(dev, "Unable to authorize the image\n");
 
 out:
+	if (mem_region)
+		memunmap(mem_region);
+
 	release_firmware(fw);
 
 	return ret;
@@ -372,45 +392,6 @@ static int a5xx_zap_shader_resume(struct msm_gpu *gpu)
 	return ret;
 }
 
-/* Set up a child device to "own" the zap shader */
-static int a5xx_zap_shader_dev_init(struct device *parent, struct device *dev)
-{
-	struct device_node *node;
-	int ret;
-
-	if (dev->parent)
-		return 0;
-
-	/* Find the sub-node for the zap shader */
-	node = of_get_child_by_name(parent->of_node, "zap-shader");
-	if (!node) {
-		DRM_DEV_ERROR(parent, "zap-shader not found in device tree\n");
-		return -ENODEV;
-	}
-
-	dev->parent = parent;
-	dev->of_node = node;
-	dev_set_name(dev, "adreno_zap_shader");
-
-	ret = device_register(dev);
-	if (ret) {
-		DRM_DEV_ERROR(parent, "Couldn't register zap shader device\n");
-		goto out;
-	}
-
-	ret = of_reserved_mem_device_init(dev);
-	if (ret) {
-		DRM_DEV_ERROR(parent, "Unable to set up the reserved memory\n");
-		device_unregister(dev);
-	}
-
-out:
-	if (ret)
-		dev->parent = NULL;
-
-	return ret;
-}
-
 static int a5xx_zap_shader_init(struct msm_gpu *gpu)
 {
 	static bool loaded;
@@ -439,11 +420,7 @@ static int a5xx_zap_shader_init(struct msm_gpu *gpu)
 		return -ENODEV;
 	}
 
-	ret = a5xx_zap_shader_dev_init(&pdev->dev, &a5xx_gpu->zap_dev);
-
-	if (!ret)
-		ret = zap_shader_load_mdt(&a5xx_gpu->zap_dev,
-			adreno_gpu->info->zapfw);
+	ret = zap_shader_load_mdt(&pdev->dev, adreno_gpu->info->zapfw);
 
 	loaded = !ret;
 
@@ -686,9 +663,6 @@ static void a5xx_destroy(struct msm_gpu *gpu)
 
 	DBG("%s", gpu->name);
 
-	if (a5xx_gpu->zap_dev.parent)
-		device_unregister(&a5xx_gpu->zap_dev);
-
 	if (a5xx_gpu->pm4_bo) {
 		if (a5xx_gpu->pm4_iova)
 			msm_gem_put_iova(a5xx_gpu->pm4_bo, gpu->aspace);
diff --git a/drivers/gpu/drm/msm/adreno/a5xx_gpu.h b/drivers/gpu/drm/msm/adreno/a5xx_gpu.h
index 6638bc85645d..6b20f28c75a0 100644
--- a/drivers/gpu/drm/msm/adreno/a5xx_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/a5xx_gpu.h
@@ -36,8 +36,6 @@ struct a5xx_gpu {
 	uint32_t gpmu_dwords;
 
 	uint32_t lm_leakage;
-
-	struct device zap_dev;
 };
 
 #define to_a5xx_gpu(x) container_of(x, struct a5xx_gpu, base)
-- 
2.9.0

Re: [PATCH v3] drm/msm: gpu: don't abuse dma_alloc for non-DMA allocations

[Bjorn Andersson]

On Wed 26 Jul 12:59 PDT 2017, Arnd Bergmann wrote:

> In zap_shader_load_mdt(), we pass a pointer to a phys_addr_t
> into dmam_alloc_coherent, which the compiler warns about:
> 
> drivers/gpu/drm/msm/adreno/a5xx_gpu.c: In function 'zap_shader_load_mdt':
> drivers/gpu/drm/msm/adreno/a5xx_gpu.c:54:50: error: passing argument 3 of 'dmam_alloc_coherent' from incompatible pointer type [-Werror=incompatible-pointer-types]
> 
> The returned DMA address is later passed on to a function that
> takes a phys_addr_t, so it's clearly wrong to use the DMA
> mapping interface here: the memory may be uncached, or the
> address may be completely wrong if there is an IOMMU connected
> to the device. What the code actually wants to do is to get
> the physical address from the reserved-mem node. It goes through
> the dma-mapping interfaces for obscure reasons, and this
> apparently only works by chance, relying on specific bugs
> in the error handling of the arm64 dma-mapping implementation.
> 
> The same problem existed in the "venus" media driver, which was
> now fixed by Stanimir Varbanov after long discussions.
> 
> In order to make some progress here, I have now ported his
> approach over to the adreno driver. The patch is currently
> untested, and should get a good review, but it is now much
> simpler than the original, and it should be obvious what
> goes wrong if I made a mistake in the port.
> 
> See also: a6e2d36bf6b7 ("media: venus: don't abuse dma_alloc for non-DMA allocations")
> Cc: Stanimir Varbanov <stanimir.varbanov@linaro.org>
> Fixes: 7c65817e6d38 ("drm/msm: gpu: Enable zap shader for A5XX")
> Acked-and-Tested-by: Jordan Crouse <jcrouse@codeaurora.org>
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Acked-by: Bjorn Andersson <bjorn.andersson@linaro.org>

However...

[..]
> @@ -37,6 +39,21 @@ static int zap_shader_load_mdt(struct device *dev, const char *fwname)
>  	if (!IS_ENABLED(CONFIG_ARCH_QCOM))
>  		return -EINVAL;
>  
> +	np = of_get_child_by_name(dev->of_node, "zap-shader");
> +	if (!np)
> +		return -ENODEV;
> +
> +	np = of_parse_phandle(np, "memory-region", 0);
> +	if (!np)
> +		return -EINVAL;
> +
> +	ret = of_address_to_resource(np, 0, &r);
> +	if (ret)
> +		return ret;
> +
> +	mem_phys = r.start;
> +	mem_size = resource_size(&r);
> +

...this means that it's now required that the reserved-memory region has
a "reg", the prior solution supported that we just specify a "size".

I have another driver where I need to figure out a mechanism of getting
hold of the dynamically allocated "base" of struct reserved_mem.


So I think it's appropriate to apply this fix now and loosen the
restrictions on placement later.

Regards,
Bjorn