[PATCH] selftests/mm: link with thp_settings when necessary

[PATCH] selftests/mm: link with thp_settings when necessary

[Wei Yang]

Currently all test cases are linked with thp_settings, while only 6
out of 50+ targets rely on it.

Instead of making thp_settings as a common dependency, link it only
when necessary.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
---
 tools/testing/selftests/mm/Makefile | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/mm/Makefile b/tools/testing/selftests/mm/Makefile
index d4f19f87053b..eea4881c918a 100644
--- a/tools/testing/selftests/mm/Makefile
+++ b/tools/testing/selftests/mm/Makefile
@@ -158,14 +158,19 @@ TEST_FILES += write_hugetlb_memory.sh
 
 include ../lib.mk
 
-$(TEST_GEN_PROGS): vm_util.c thp_settings.c
-$(TEST_GEN_FILES): vm_util.c thp_settings.c
+$(TEST_GEN_PROGS): vm_util.c
+$(TEST_GEN_FILES): vm_util.c
 
 $(OUTPUT)/uffd-stress: uffd-common.c
 $(OUTPUT)/uffd-unit-tests: uffd-common.c
-$(OUTPUT)/uffd-wp-mremap: uffd-common.c
+$(OUTPUT)/uffd-wp-mremap: uffd-common.c thp_settings.c
 $(OUTPUT)/protection_keys: pkey_util.c
 $(OUTPUT)/pkey_sighandler_tests: pkey_util.c
+$(OUTPUT)/cow: thp_settings.c
+$(OUTPUT)/migration: thp_settings.c
+$(OUTPUT)/khugepaged: thp_settings.c
+$(OUTPUT)/ksm_tests: thp_settings.c
+$(OUTPUT)/soft-dirty: thp_settings.c
 
 ifeq ($(ARCH),x86_64)
 BINARIES_32 := $(patsubst %,$(OUTPUT)/%,$(BINARIES_32))
-- 
2.34.1

Re: [PATCH] selftests/mm: link with thp_settings when necessary

[David Hildenbrand]

On 01.08.25 10:54, Wei Yang wrote:
> Currently all test cases are linked with thp_settings, while only 6
> out of 50+ targets rely on it.
> 
> Instead of making thp_settings as a common dependency, link it only
> when necessary.


You don't state why we should care about that? I don't see how binary 
size is a problem, why do you think it is?

-- 
Cheers,

David / dhildenb

Re: [PATCH] selftests/mm: link with thp_settings when necessary

[Lorenzo Stoakes]

On Fri, Aug 01, 2025 at 11:02:35AM +0200, David Hildenbrand wrote:
> On 01.08.25 10:54, Wei Yang wrote:
> > Currently all test cases are linked with thp_settings, while only 6
> > out of 50+ targets rely on it.
> >
> > Instead of making thp_settings as a common dependency, link it only
> > when necessary.
>
>
> You don't state why we should care about that? I don't see how binary size
> is a problem, why do you think it is?

Yeah overall I'm with David on this, I'd rather we keep things simple (so
you don't need to remember to link against this file) and just plonk all
the stuff you need to link to together in the Makefile.

It's not really a big deal when it comes to test stuff I don't think.

Cheers, Lorenzo

Re: [PATCH] selftests/mm: link with thp_settings when necessary

[Wei Yang]

On Fri, Aug 01, 2025 at 11:02:35AM +0200, David Hildenbrand wrote:
>On 01.08.25 10:54, Wei Yang wrote:
>> Currently all test cases are linked with thp_settings, while only 6
>> out of 50+ targets rely on it.
>> 
>> Instead of making thp_settings as a common dependency, link it only
>> when necessary.
>
>
>You don't state why we should care about that? I don't see how binary size is
>a problem, why do you think it is?
>

Yes, it doesn't help on the binary size perspective.

I come up with this when doing some test. Each time a change in thp_settings
would rebuild all the tests. My intent is to speed up the build during coding.

Not sure it worth a change.

>Cheers,
>
>David / dhildenb

-- 
Wei Yang
Help you, Help me

Re: [PATCH] selftests/mm: link with thp_settings when necessary

[Wei Yang]

On Fri, Aug 01, 2025 at 12:14:34PM +0100, Lorenzo Stoakes wrote:
>On Fri, Aug 01, 2025 at 11:02:35AM +0200, David Hildenbrand wrote:
>> On 01.08.25 10:54, Wei Yang wrote:
>> > Currently all test cases are linked with thp_settings, while only 6
>> > out of 50+ targets rely on it.
>> >
>> > Instead of making thp_settings as a common dependency, link it only
>> > when necessary.
>>
>>
>> You don't state why we should care about that? I don't see how binary size
>> is a problem, why do you think it is?
>
>Yeah overall I'm with David on this, I'd rather we keep things simple (so
>you don't need to remember to link against this file) and just plonk all
>the stuff you need to link to together in the Makefile.

I have thought about this, while current implementation do specify its
necessary dependency for some target like uffd-common.c/pkey_util.c.

So I thought we may keep the same style.

>
>It's not really a big deal when it comes to test stuff I don't think.
>

Yes, this is not a big deal in general. I come up with this idea because my test
machine is too old. This change could same me some time during
coding/debugging.

>Cheers, Lorenzo

-- 
Wei Yang
Help you, Help me