Skip to contentSkip to conversation

drm/test: fix the gem shmem test to map the sg table.

PATCH Applied Dave Airlie 5 participants 6 messages Jul 15, 2024 – Aug 2, 2024

Subsystems (2)
lore
Share & tools
lore

Dave Airlie

author maintainer
#0 lore raw

PATCH 1/1 drm/test: fix the gem shmem test to map the sg table.

From: Dave Airlie <[email protected]>

The test here creates an sg table, but never maps it, when we get to drm_gem_shmem_free, the helper tries to unmap and this causes warnings on some platforms and debug kernels.

This also sets a 64-bit dma mask, as I see an swiotlb warning if I stick with the default 32-bit one.

Fixes 93032ae634d4 ("drm/test: add a test suite for GEM objects backed by shmem")
---
 drivers/gpu/drm/tests/drm_gem_shmem_test.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
drivers/gpu/drm/tests/drm_gem_shmem_test.c +11 -0
index 91202e40cde9..eb3a7a84be90 100644
@@ -102,6 +102,17 @@ static void drm_gem_shmem_test_obj_create_private(struct kunit *test)
 
 	sg_init_one(sgt->sgl, buf, TEST_SIZE);
 
+	/*
+	 * Set the DMA mask to 64-bits and map the sgtables
+	 * otherwise drm_gem_shmem_free will cause a warning
+	 * on debug kernels.
+	 * */
+	ret = dma_set_mask(drm_dev->dev, DMA_BIT_MASK(64));
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
+	ret = dma_map_sgtable(drm_dev->dev, sgt, DMA_BIDIRECTIONAL, 0);
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
 	/* Init a mock DMA-BUF */
 	buf_mock.size = TEST_SIZE;
 	attach_mock.dmabuf = &buf_mock;
Signature · 1 line
2.45.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 7 lines

-----Original Message-----
From: dri-devel <[email protected]> On Behalf Of Dave Airlie
Sent: Monday, July 15, 2024 4:36 AM
To: [email protected]
Subject: [PATCH] drm/test: fix the gem shmem test to map the sg table.

Quoted 5 lines

From: Dave Airlie <[email protected]>

The test here creates an sg table, but never maps it, when we get to drm_gem_shmem_free, the helper tries to unmap and this causes warnings on some platforms and debug kernels.

This looks pretty straightforward...

However, should drm_gem_shmem_free() really give an error if the mapping didn't happen?

I.e. just because you have an sgt pointer, should you also have a mapping?

If the errors are really just noise (form the specific platforms), and this patch is covering for that, then:

Thanks,

Mike

Quoted 37 lines

This also sets a 64-bit dma mask, as I see an swiotlb warning if I stick with the default 32-bit one.

Fixes 93032ae634d4 ("drm/test: add a test suite for GEM objects backed by

shmem")

---
drivers/gpu/drm/tests/drm_gem_shmem_test.c | 11 +++++++++++

1 file changed, 11 insertions(+)
drivers/gpu/drm/tests/drm_gem_shmem_test.c +11 -0
index 91202e40cde9..eb3a7a84be90 100644
@@ -102,6 +102,17 @@ static void
 drm_gem_shmem_test_obj_create_private(struct kunit *test)
 
 	sg_init_one(sgt->sgl, buf, TEST_SIZE);
 
+	/*
+	 * Set the DMA mask to 64-bits and map the sgtables
+	 * otherwise drm_gem_shmem_free will cause a warning
+	 * on debug kernels.
+	 * */
+	ret = dma_set_mask(drm_dev->dev, DMA_BIT_MASK(64));
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
+	ret = dma_map_sgtable(drm_dev->dev, sgt, DMA_BIDIRECTIONAL, 0);
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
 	/* Init a mock DMA-BUF */
 	buf_mock.size = TEST_SIZE;
	attach_mock.dmabuf = &buf_mock;

--
2.45.0

Acked-by
#2 lore raw
Quoted 13 lines · Michael J
Quoted 7 lines

-----Original Message-----
From: dri-devel <[email protected]> On Behalf Of Dave Airlie
Sent: Monday, July 15, 2024 4:36 AM
To: [email protected]
Subject: [PATCH] drm/test: fix the gem shmem test to map the sg table.

Quoted 5 lines

From: Dave Airlie <[email protected]>

The test here creates an sg table, but never maps it, when we get to drm_gem_shmem_free, the helper tries to unmap and this causes warnings on some platforms and debug kernels.

Quoted 6 lines

This looks pretty straightforward...

However, should drm_gem_shmem_free() really give an error if the mapping didn't happen?

I.e. just because you have an sgt pointer, should you also have a mapping?

Yes, I think only allocating an sgt but not setting it up is a bug. So the fix looks correct, and isn't just papering over noise.

Quoted 4 lines

If the errors are really just noise (form the specific platforms), and this patch is covering for that, then:

Cheers, Sima

Quoted 43 lines

Thanks,

Mike

Quoted 37 lines

This also sets a 64-bit dma mask, as I see an swiotlb warning if I stick with the default 32-bit one.

Fixes 93032ae634d4 ("drm/test: add a test suite for GEM objects backed by

shmem")

---
drivers/gpu/drm/tests/drm_gem_shmem_test.c | 11 +++++++++++

1 file changed, 11 insertions(+)
drivers/gpu/drm/tests/drm_gem_shmem_test.c +11 -0
index 91202e40cde9..eb3a7a84be90 100644
@@ -102,6 +102,17 @@ static void
 drm_gem_shmem_test_obj_create_private(struct kunit *test)
 
 	sg_init_one(sgt->sgl, buf, TEST_SIZE);
 
+	/*
+	 * Set the DMA mask to 64-bits and map the sgtables
+	 * otherwise drm_gem_shmem_free will cause a warning
+	 * on debug kernels.
+	 * */
+	ret = dma_set_mask(drm_dev->dev, DMA_BIT_MASK(64));
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
+	ret = dma_map_sgtable(drm_dev->dev, sgt, DMA_BIDIRECTIONAL, 0);
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
 	/* Init a mock DMA-BUF */
 	buf_mock.size = TEST_SIZE;
	attach_mock.dmabuf = &buf_mock;

--
2.45.0

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
#3 lore raw
Quoted 24 lines

-----Original Message-----
From: Daniel Vetter <[email protected]>
Sent: Tuesday, July 16, 2024 5:34 AM
To: Ruhl, Michael J <[email protected]>
Cc: Dave Airlie <[email protected]>; [email protected]
Subject: Re: [PATCH] drm/test: fix the gem shmem test to map the sg table.

Quoted 2 lines · Michael J
Quoted 2 lines

-----Original Message-----
From: dri-devel <[email protected]> On Behalf Of

Dave

Quoted 13 lines
Quoted 10 lines

Airlie
Sent: Monday, July 15, 2024 4:36 AM
To: [email protected]
Subject: [PATCH] drm/test: fix the gem shmem test to map the sg table.

From: Dave Airlie <[email protected]>

The test here creates an sg table, but never maps it, when we get to drm_gem_shmem_free, the helper tries to unmap and this causes warnings on some platforms and debug kernels.

This looks pretty straightforward...

Quoted 7 lines
Quoted 4 lines

However, should drm_gem_shmem_free() really give an error if the mapping didn't happen?

I.e. just because you have an sgt pointer, should you also have a mapping?

Yes, I think only allocating an sgt but not setting it up is a bug. So the fix looks correct, and isn't just papering over noise.

I guess my concern here is that the mapping could fail.

If that happens, what is the error path?

Can I call _shmem_free?

Mike

Quoted 57 lines
Quoted 1 line

If the errors are really just noise (form the specific platforms), and this patch is

covering

Quoted 3 lines

for that, then:

Cheers, Sima

Quoted 26 lines

Thanks,

Mike

Quoted 21 lines

This also sets a 64-bit dma mask, as I see an swiotlb warning if I stick with the default 32-bit one.

Fixes 93032ae634d4 ("drm/test: add a test suite for GEM objects backed by

shmem")

---
drivers/gpu/drm/tests/drm_gem_shmem_test.c | 11 +++++++++++

1 file changed, 11 insertions(+)
drivers/gpu/drm/tests/drm_gem_shmem_test.c +0 -0
index 91202e40cde9..eb3a7a84be90 100644
@@ -102,6 +102,17 @@ static void
 drm_gem_shmem_test_obj_create_private(struct kunit *test)
 
 	sg_init_one(sgt->sgl, buf, TEST_SIZE);
 
Quoted 17 lines
Quoted 14 lines
drivers/gpu/drm/tests/drm_gem_shmem_test.c +11 -0
+	/*
+	 * Set the DMA mask to 64-bits and map the sgtables
+	 * otherwise drm_gem_shmem_free will cause a warning
+	 * on debug kernels.
+	 * */
+	ret = dma_set_mask(drm_dev->dev, DMA_BIT_MASK(64));
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
+	ret = dma_map_sgtable(drm_dev->dev, sgt, DMA_BIDIRECTIONAL, 0);
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
 	/* Init a mock DMA-BUF */
 	buf_mock.size = TEST_SIZE;
	attach_mock.dmabuf = &buf_mock;
Quoted 2 lines

--
2.45.0

--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

Reviewed-by
#4 lore raw
Quoted 29 lines · Michael J
Quoted 29 lines

-----Original Message-----
From: Daniel Vetter <[email protected]>
Sent: Tuesday, July 16, 2024 5:34 AM
To: Ruhl, Michael J <[email protected]>
Cc: Dave Airlie <[email protected]>; [email protected]
Subject: Re: [PATCH] drm/test: fix the gem shmem test to map the sg table.

Quoted 2 lines · Michael J
Quoted 2 lines

-----Original Message-----
From: dri-devel <[email protected]> On Behalf Of

Dave

Quoted 17 lines
Quoted 10 lines

Airlie
Sent: Monday, July 15, 2024 4:36 AM
To: [email protected]
Subject: [PATCH] drm/test: fix the gem shmem test to map the sg table.

From: Dave Airlie <[email protected]>

The test here creates an sg table, but never maps it, when we get to drm_gem_shmem_free, the helper tries to unmap and this causes warnings on some platforms and debug kernels.

This looks pretty straightforward...

However, should drm_gem_shmem_free() really give an error if the mapping didn't happen?

I.e. just because you have an sgt pointer, should you also have a mapping?

Quoted 8 lines
Quoted 2 lines

Yes, I think only allocating an sgt but not setting it up is a bug. So the fix looks correct, and isn't just papering over noise.

I guess my concern here is that the mapping could fail.

If that happens, what is the error path?

Can I call _shmem_free?

In this case, if the mapping fails, the test case will be aborted, and the sg_table will be freed by the action that calls sg_free_table_wrapper(). However, I also think drm_gem_shmem_free() should behave well even if the sg_table is not mapped. Is there any advantage in issuing a warning when freeing the object if the sg_table is not mapped?

Thanks,
Marco

Quoted 61 lines

Mike

Quoted 57 lines
Quoted 1 line

If the errors are really just noise (form the specific platforms), and this patch is

covering

Quoted 3 lines

for that, then:

Cheers, Sima

Quoted 26 lines

Thanks,

Mike

Quoted 21 lines

This also sets a 64-bit dma mask, as I see an swiotlb warning if I stick with the default 32-bit one.

Fixes 93032ae634d4 ("drm/test: add a test suite for GEM objects backed by

shmem")

---
drivers/gpu/drm/tests/drm_gem_shmem_test.c | 11 +++++++++++

1 file changed, 11 insertions(+)
drivers/gpu/drm/tests/drm_gem_shmem_test.c +0 -0
index 91202e40cde9..eb3a7a84be90 100644
@@ -102,6 +102,17 @@ static void
 drm_gem_shmem_test_obj_create_private(struct kunit *test)
 
 	sg_init_one(sgt->sgl, buf, TEST_SIZE);
 
Quoted 17 lines
Quoted 14 lines
drivers/gpu/drm/tests/drm_gem_shmem_test.c +11 -0
+	/*
+	 * Set the DMA mask to 64-bits and map the sgtables
+	 * otherwise drm_gem_shmem_free will cause a warning
+	 * on debug kernels.
+	 * */
+	ret = dma_set_mask(drm_dev->dev, DMA_BIT_MASK(64));
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
+	ret = dma_map_sgtable(drm_dev->dev, sgt, DMA_BIDIRECTIONAL, 0);
+	KUNIT_ASSERT_EQ(test, ret, 0);
+
 	/* Init a mock DMA-BUF */
 	buf_mock.size = TEST_SIZE;
	attach_mock.dmabuf = &buf_mock;
Quoted 2 lines

--
2.45.0

--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

maintainer Applied
#5 lore raw
In reply to Dave Airlie
Quoted 8 lines · Dave Airlie

The test here creates an sg table, but never maps it, when we get to drm_gem_shmem_free, the helper tries to unmap and this causes warnings on some platforms and debug kernels.

This also sets a 64-bit dma mask, as I see an swiotlb warning if I stick with the default 32-bit one.

[...]

Applied to misc/kernel.git (drm-misc-fixes).

Thanks!
Maxime