Skip to contentSkip to conversation

drm/xe/kunit: fix link failure with built-in xe

PATCH Superseded Arnd Bergmann 5 participants 9 messages Feb 24, 2024 – Feb 26, 2024

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 drm/xe/kunit: fix link failure with built-in xe

From: Arnd Bergmann <[email protected]>

When the driver is built-in but the tests are in loadable modules, the helpers don't actually get put into the driver:

ERROR: modpost: "xe_kunit_helper_alloc_xe_device" [drivers/gpu/drm/xe/tests/xe_test.ko] undefined!

Change the Makefile to ensure they are always part of the driver even when the rest of the kunit tests are in loadable modules.

The tests/xe_kunit_helpers.c file depends on DRM_KUNIT_TEST_HELPERS, so this has to always be selected by the main XE module now, rather than the actual tests. In turn, the "depends on (m || (y && KUNIT=y))" doesn't really do what it tried and can just be removed.

Fixes 5095d13d758b ("drm/xe/kunit: Define helper functions to allocate fake xe device")
---
 drivers/gpu/drm/xe/Kconfig       | 3 ++-
 drivers/gpu/drm/xe/Kconfig.debug | 1 -
 drivers/gpu/drm/xe/Makefile      | 6 ++++--
 3 files changed, 6 insertions(+), 4 deletions(-)
drivers/gpu/drm/xe/Kconfig +2 -1
index 6d4428b19a4c..2948650680e1 100644
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only
 config DRM_XE
 	tristate "Intel Xe Graphics"
-	depends on DRM && PCI && MMU && (m || (y && KUNIT=y))
+	depends on DRM && PCI && MMU
 	depends on ACPI_VIDEO || !ACPI
 	select INTERVAL_TREE
 	# we need shmfs for the swappable backing store, and in particular
@@ -11,6 +11,7 @@ config DRM_XE
 	select DRM_BUDDY
 	select DRM_EXEC
 	select DRM_KMS_HELPER
+	select DRM_KUNIT_TEST_HELPERS if DRM_XE_KUNIT_TEST != n
 	select DRM_PANEL
 	select DRM_SUBALLOC_HELPER
 	select DRM_DISPLAY_DP_HELPER
drivers/gpu/drm/xe/Kconfig.debug +0 -1
index 549065f57a78..df02e5d17d26 100644
@@ -76,7 +76,6 @@ config DRM_XE_KUNIT_TEST
 	depends on DRM_XE && KUNIT && DEBUG_FS
 	default KUNIT_ALL_TESTS
 	select DRM_EXPORT_FOR_TESTS if m
-	select DRM_KUNIT_TEST_HELPERS
 	help
 	  Choose this option to allow the driver to perform selftests under
 	  the kunit framework
drivers/gpu/drm/xe/Makefile +4 -2
index 4c6ffe4b2172..b596e4482a9b 100644
@@ -158,8 +158,10 @@ xe-$(CONFIG_PCI_IOV) += \
 	xe_lmtt_2l.o \
 	xe_lmtt_ml.o
 
-xe-$(CONFIG_DRM_XE_KUNIT_TEST) += \
-	tests/xe_kunit_helpers.o
+# include helpers for tests even when XE is built-in
+ifdef CONFIG_DRM_XE_KUNIT_TEST
+xe-y += tests/xe_kunit_helpers.o
+endif
 
 # i915 Display compat #defines and #includes
 subdir-ccflags-$(CONFIG_DRM_XE_DISPLAY) += \
Signature · 1 line
2.39.2
#1 lore raw

PATCH 2/3 drm/xe/mmio: fix build warning for BAR resize on 32-bit

From: Arnd Bergmann <[email protected]>

clang complains about a nonsensical test on builds with a 32-bit phys_addr_t, which means resizing will always fail:

drivers/gpu/drm/xe/xe_mmio.c:109:23: error: result of comparison of constant 4294967296 with expression of type 'resource_size_t' (aka 'unsigned int') is always false [-Werror,-Wtautological-constant-out-of-range-compare]

  109 |                     root_res->start > 0x100000000ull)
      |                     ~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~

Previously, BAR resize was always disallowed on 32-bit kernels, but this apparently changed recently. Since 32-bit machines can in theory support PAE/LPAE for large address spaces, this may end up useful, so change the driver to shut up the warning but still work when phys_addr_t/resource_size_t is 64 bit wide.

Fixes 9a6e6c14bfde ("drm/xe/mmio: Use non-atomic writeq/readq variant for 32b")
Fixes ea97a66a2218 ("drm/xe: Disable 32bits build")
---
 drivers/gpu/drm/xe/xe_mmio.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_mmio.c +1 -1
index e3db3a178760..7ba2477452d7 100644
@@ -106,7 +106,7 @@ static void xe_resize_vram_bar(struct xe_device *xe)
 
 	pci_bus_for_each_resource(root, root_res, i) {
 		if (root_res && root_res->flags & (IORESOURCE_MEM | IORESOURCE_MEM_64) &&
-		    root_res->start > 0x100000000ull)
+		    (u64)root_res->start > 0x100000000ul)
 			break;
 	}
 
Signature · 1 line
2.39.2
Acked-by on v1 2/3
#2 lore raw

Re: [PATCH 2/3] drm/xe/mmio: fix build warning for BAR resize on 32-bit

Quoted 9 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

clang complains about a nonsensical test on builds with a 32-bit phys_addr_t, which means resizing will always fail:

drivers/gpu/drm/xe/xe_mmio.c:109:23: error: result of comparison of constant 4294967296 with expression of type 'resource_size_t' (aka 'unsigned int') is always false [-Werror,-Wtautological-constant-out-of-range-compare]

 109 |                     root_res->start > 0x100000000ull)
     |                     ~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~
Quoted 8 lines

Previously, BAR resize was always disallowed on 32-bit kernels, but this apparently changed recently. Since 32-bit machines can in theory support PAE/LPAE for large address spaces, this may end up useful, so change the driver to shut up the warning but still work when phys_addr_t/resource_size_t is 64 bit wide.

Fixes 9a6e6c14bfde ("drm/xe/mmio: Use non-atomic writeq/readq variant for 32b")
Fixes ea97a66a2218 ("drm/xe: Disable 32bits build")

this second Fixes should not be here? How would "disabling 32bits build" break 32bits build? I think just the first one is enough, otherwise 237412e45390 ("drm/xe: Enable 32bits build") is your next good candidate.

Lucas De Marchi

Quoted 21 lines

---
drivers/gpu/drm/xe/xe_mmio.c | 2 +-

1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_mmio.c +1 -1
index e3db3a178760..7ba2477452d7 100644
@@ -106,7 +106,7 @@ static void xe_resize_vram_bar(struct xe_device *xe)
 
 	pci_bus_for_each_resource(root, root_res, i) {
 		if (root_res && root_res->flags & (IORESOURCE_MEM | IORESOURCE_MEM_64) &&
-		    root_res->start > 0x100000000ull)
+		    (u64)root_res->start > 0x100000000ul)
 			break;
 	}
 
Signature · 1 line
2.39.2
#3 lore raw

PATCH 3/3 drm/xe/xe2: fix 64-bit division in pte_update_size

In reply to Arnd Bergmann

From: Arnd Bergmann <[email protected]>

This function does not build on 32-bit targets when the compiler fails to reduce DIV_ROUND_UP() into a shift:

ld.lld: error: undefined symbol: __aeabi_uldivmod

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

referenced by xe_migrate.c
drivers/gpu/drm/xe/xe_migrate.o:(pte_update_size) in archive vmlinux.a

There are two instances in this function. Change the first to use an open-coded shift with the same behavior, and the second one to a 32-bit calculation, which is sufficient here as the size is never more than 2^32 pages (16TB).

Fixes ea97a66a2218 ("drm/xe: Disable 32bits build")
---
 drivers/gpu/drm/xe/xe_migrate.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_migrate.c +1 -1
index a66fdf2d2991..ee1bb938c493 100644
@@ -462,7 +462,7 @@ static u32 pte_update_size(struct xe_migrate *m,
 	} else {
 		/* Clip L0 to available size */
 		u64 size = min(*L0, (u64)avail_pts * SZ_2M);
-		u64 num_4k_pages = DIV_ROUND_UP(size, XE_PAGE_SIZE);
+		u32 num_4k_pages = (size + XE_PAGE_SIZE - 1) >> XE_PTE_SHIFT;
 
 		*L0 = size;
 		*L0_ofs = xe_migrate_vm_addr(pt_ofs, 0);
Signature · 1 line
2.39.2
#4 lore raw

Re: [PATCH 3/3] drm/xe/xe2: fix 64-bit division in pte_update_size

Quoted 9 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

This function does not build on 32-bit targets when the compiler fails to reduce DIV_ROUND_UP() into a shift:

ld.lld: error: undefined symbol: __aeabi_uldivmod

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

referenced by xe_migrate.c
drivers/gpu/drm/xe/xe_migrate.o:(pte_update_size) in archive vmlinux.a

Quoted 6 lines

There are two instances in this function. Change the first to use an open-coded shift with the same behavior, and the second one to a 32-bit calculation, which is sufficient here as the size is never more than 2^32 pages (16TB).

Fixes ea97a66a2218 ("drm/xe: Disable 32bits build")

same comment as in patch 2... should rather be the commit *enabling* 32b builds?

Should this be something to be dealt with at the DIV_ROUND_UP() layer? Why did the compiler fail to reduce it here but didn't in the other 11 uses in the xe driver?

thanks
Lucas De Marchi

Quoted 21 lines

---
drivers/gpu/drm/xe/xe_migrate.c | 2 +-

1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/xe/xe_migrate.c +1 -1
index a66fdf2d2991..ee1bb938c493 100644
@@ -462,7 +462,7 @@ static u32 pte_update_size(struct xe_migrate *m,
 	} else {
 		/* Clip L0 to available size */
 		u64 size = min(*L0, (u64)avail_pts * SZ_2M);
-		u64 num_4k_pages = DIV_ROUND_UP(size, XE_PAGE_SIZE);
+		u32 num_4k_pages = (size + XE_PAGE_SIZE - 1) >> XE_PTE_SHIFT;
 
 		*L0 = size;
 		*L0_ofs = xe_migrate_vm_addr(pt_ofs, 0);
Signature · 1 line
2.39.2
#5 lore raw

✗ CI.Patch_applied: failure for series starting with [1/3] drm/xe/kunit: fix link failure with built-in xe

In reply to Arnd Bergmann
== Series Details ==

== Series Details ==

Series: series starting with [1/3] drm/xe/kunit: fix link failure with built-in xe
URL : https://patchwork.freedesktop.org/series/130344/
State : failure

== Summary ==

=== Applying kernel patches on branch 'drm-tip' with base: ===
Base commit: 7291e2e67dff drm-tip: 2024y-02m-23d-21h-08m-22s UTC integration manifest === git am output follows ===
error: patch failed: drivers/gpu/drm/xe/Kconfig:1
error: drivers/gpu/drm/xe/Kconfig: patch does not apply
hint: Use 'git am --show-current-patch=diff' to see the failed patch
Applying: drm/xe/kunit: fix link failure with built-in xe
Patch failed at 0001 drm/xe/kunit: fix link failure with built-in xe When you have resolved this problem, run "git am --continue".
If you prefer to skip this patch, run "git am --skip" instead.
To restore the original branch and stop patching, run "git am --abort".

#6 lore raw
In reply to Arnd Bergmann
Quoted 7 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

When the driver is built-in but the tests are in loadable modules, the helpers don't actually get put into the driver:

ERROR: modpost: "xe_kunit_helper_alloc_xe_device" [drivers/gpu/drm/xe/tests/xe_test.ko] undefined!

Quoted 7 lines

Change the Makefile to ensure they are always part of the driver even when the rest of the kunit tests are in loadable modules.

The tests/xe_kunit_helpers.c file depends on DRM_KUNIT_TEST_HELPERS, so this has to always be selected by the main XE module now, rather than the actual tests. In turn, the "depends on (m || (y && KUNIT=y))" doesn't really do what it tried and can just be removed.

it actually did, which was to workaround issues prior to the commit you are pointing out. What it did was to make sure xe.ko is m, or if it's built-in, kunit is also built-in. Apparently the problem here is that the xe_test.ko is missing the symbols.

See commit 08987a8b6820 ("drm/xe: Fix build with KUNIT=m").

I'm happy to remove it though if it's indeed not needed anymore.

Lucas De Marchi

Quoted 62 lines
Fixes 5095d13d758b ("drm/xe/kunit: Define helper functions to allocate fake xe device")

---

drivers/gpu/drm/xe/Kconfig       | 3 ++-
drivers/gpu/drm/xe/Kconfig.debug | 1 -
drivers/gpu/drm/xe/Makefile      | 6 ++++--
3 files changed, 6 insertions(+), 4 deletions(-)
drivers/gpu/drm/xe/Kconfig +0 -0
index 6d4428b19a4c..2948650680e1 100644
@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: GPL-2.0-only

config DRM_XE
tristate "Intel Xe Graphics"

drivers/gpu/drm/xe/Kconfig +1 -1
-	depends on DRM && PCI && MMU && (m || (y && KUNIT=y))
+	depends on DRM && PCI && MMU
	depends on ACPI_VIDEO || !ACPI
	select INTERVAL_TREE
	# we need shmfs for the swappable backing store, and in particular
drivers/gpu/drm/xe/Kconfig +1 -0
@@ -11,6 +11,7 @@ config DRM_XE
 	select DRM_BUDDY
 	select DRM_EXEC
 	select DRM_KMS_HELPER
+	select DRM_KUNIT_TEST_HELPERS if DRM_XE_KUNIT_TEST != n
 	select DRM_PANEL
 	select DRM_SUBALLOC_HELPER
 	select DRM_DISPLAY_DP_HELPER
drivers/gpu/drm/xe/Kconfig.debug +0 -1
index 549065f57a78..df02e5d17d26 100644
@@ -76,7 +76,6 @@ config DRM_XE_KUNIT_TEST
 	depends on DRM_XE && KUNIT && DEBUG_FS
 	default KUNIT_ALL_TESTS
 	select DRM_EXPORT_FOR_TESTS if m
-	select DRM_KUNIT_TEST_HELPERS
 	help
 	  Choose this option to allow the driver to perform selftests under
 	  the kunit framework
drivers/gpu/drm/xe/Makefile +4 -2
index 4c6ffe4b2172..b596e4482a9b 100644
@@ -158,8 +158,10 @@ xe-$(CONFIG_PCI_IOV) += \
 	xe_lmtt_2l.o \
 	xe_lmtt_ml.o
 
-xe-$(CONFIG_DRM_XE_KUNIT_TEST) += \
-	tests/xe_kunit_helpers.o
+# include helpers for tests even when XE is built-in
+ifdef CONFIG_DRM_XE_KUNIT_TEST
+xe-y += tests/xe_kunit_helpers.o
+endif
 
 # i915 Display compat #defines and #includes
 subdir-ccflags-$(CONFIG_DRM_XE_DISPLAY) += \
Signature · 1 line
2.39.2
#7 lore raw
Quoted 16 lines · Lucas De Marchi
Quoted 7 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

When the driver is built-in but the tests are in loadable modules, the helpers don't actually get put into the driver:

ERROR: modpost: "xe_kunit_helper_alloc_xe_device" [drivers/gpu/drm/xe/tests/xe_test.ko] undefined!

Quoted 7 lines

Change the Makefile to ensure they are always part of the driver even when the rest of the kunit tests are in loadable modules.

The tests/xe_kunit_helpers.c file depends on DRM_KUNIT_TEST_HELPERS, so this has to always be selected by the main XE module now, rather than the actual tests. In turn, the "depends on (m || (y && KUNIT=y))" doesn't really do what it tried and can just be removed.

Quoted 8 lines

it actually did, which was to workaround issues prior to the commit you are pointing out. What it did was to make sure xe.ko is m, or if it's built-in, kunit is also built-in. Apparently the problem here is that the xe_test.ko is missing the symbols.

See commit 08987a8b6820 ("drm/xe: Fix build with KUNIT=m").

I'm happy to remove it though if it's indeed not needed anymore.

I think the main issue with that original patch is that it keeps the direct include of the test files.

I assume it was done to allow kunit tests of static functions, but now that VISIBLE_IF_KUNIT and EXPORT_SYMBOL_IF_KUNIT are a thing, you don't have the need for that include, and you can build your tests in a module while the code is builtin.

And the conversion should be pretty minimal.

Maxime

#8 lore raw
In reply to Lucas De Marchi
Quoted 16 lines · Lucas De Marchi
Quoted 7 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

When the driver is built-in but the tests are in loadable modules, the helpers don't actually get put into the driver:

ERROR: modpost: "xe_kunit_helper_alloc_xe_device" [drivers/gpu/drm/xe/tests/xe_test.ko] undefined!

Quoted 7 lines

Change the Makefile to ensure they are always part of the driver even when the rest of the kunit tests are in loadable modules.

The tests/xe_kunit_helpers.c file depends on DRM_KUNIT_TEST_HELPERS, so this has to always be selected by the main XE module now, rather than the actual tests. In turn, the "depends on (m || (y && KUNIT=y))" doesn't really do what it tried and can just be removed.

Quoted 4 lines

it actually did, which was to workaround issues prior to the commit you are pointing out. What it did was to make sure xe.ko is m, or if it's built-in, kunit is also built-in. Apparently the problem here is that the xe_test.ko is missing the symbols.

Ah, I misunderstood the intention, as I was thrown off by the redundant 'y &&' which sounds like it was trying to force XE to be built-in rather than modular when Kunit is !=m.

The more common way to write this is 'depends on KUNIT || !KUNIT', with some drivers writing it as 'depends on m || KUNIT!=m'.

I double-checked now and found that the dependency is indeed still needed:

WARNING: unmet direct dependencies detected for DRM_KUNIT_TEST_HELPERS
Depends on [m]: HAS_IOMEM [=y] && DRM [=y] && KUNIT [=m]
Selected by [y]:
- DRM_XE [=y] && HAS_IOMEM [=y] && DRM [=y] && PCI [=y] && MMU [=y] && (ACPI_VIDEO [=y] || !ACPI [=y]) && DRM_XE_KUNIT_TEST [=m]!=n Selected by [m]:
- DRM_KUNIT_TEST [=m] && HAS_IOMEM [=y] && DRM [=y] && KUNIT [=m] && MMU [=y]

Quoted 3 lines

See commit 08987a8b6820 ("drm/xe: Fix build with KUNIT=m").

I'm happy to remove it though if it's indeed not needed anymore.

Ideally the xe.ko module should not depend on anything exported by lib/kunit, but for the moment, the tests/xe_kunit_helpers.o file is still included in xe.ko and in turn depends on kunit.

Changing this is probably a little more complicated than my patch, so I'll just send a v2 without the incorrect line.

Arnd