[PATCH v2 0/6] Fix errors when try to build kvm selftests on specified output

[PATCH v2 0/6] Fix errors when try to build kvm selftests on specified output

[Xiaoyao Li]

I attempted to build KVM selftests on a specified dir, unfortunately
neither	"make O=/path/to/mydir TARGETS=kvm" in tools/testing/selftests, nor
"make OUTPUT=/path/to/mydir" in tools/testing/selftests/kvm work.

This series aims to fix them.

Patch 1 fixes the issue that output directory is not exist.

Patch 2 and 3 are the preparation for kvm to get the right path of
installed linux headers.

Patch 4 and 6 prepare the INSTALL_HDR_PATH to tell sub TARGET where the
linux headers are installed.

Patch 5 fixes the issue that with OUTPUT specified, it still make the
linux tree dirty.

I only test the sub TARGET of kvm.
In theory, it won't break other TARGET of selftests.

Changes in v2:
 - fix the no directory issue in lib.mk
 - make kvm fixes seperate patch
 - Add the patch to fix linux src tree not clean issue

v1:
https://lore.kernel.org/kvm/20200315093425.33600-1-xiaoyao.li@intel.com/

Xiaoyao Li (6):
  selftests: Create directory when OUTPUT specified
  selftests: kvm: Include lib.mk earlier
  selftests: kvm: Use the default linux header path only when
    INSTALL_HDR_PATH not defined
  selftests: Create variable INSTALL_HDR_PATH if need to install linux
    headers to $(OUTPUT)/usr
  selftests: Generate build output of linux headers to
    $(OUTPUT)/linux-header-build
  selftests: export INSTALL_HDR_PATH if using "O" to specify output dir

 tools/testing/selftests/Makefile     |  6 +++++-
 tools/testing/selftests/kvm/Makefile |  9 +++++----
 tools/testing/selftests/lib.mk       | 19 ++++++++++++++++++-
 3 files changed, 28 insertions(+), 6 deletions(-)

-- 
2.20.1

[PATCH v2 1/6] selftests: Create directory when OUTPUT specified

[Xiaoyao Li]

When OUTPUT specified, it may encounter no directory error if there is
subdir in $TARGET for any of TEST_GEN_PROGS, TEST_GEN_PROGS_EXTENDED, or
TEST_GEN_FILES.

This issue is found by running

	make OUTPUT=/path/to/output

in tools/testing/selftests/kvm

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
---
 tools/testing/selftests/lib.mk | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/tools/testing/selftests/lib.mk b/tools/testing/selftests/lib.mk
index 1c8a1963d03f..78fb00bdf5c4 100644
--- a/tools/testing/selftests/lib.mk
+++ b/tools/testing/selftests/lib.mk
@@ -20,6 +20,16 @@ TEST_GEN_PROGS := $(patsubst %,$(OUTPUT)/%,$(TEST_GEN_PROGS))
 TEST_GEN_PROGS_EXTENDED := $(patsubst %,$(OUTPUT)/%,$(TEST_GEN_PROGS_EXTENDED))
 TEST_GEN_FILES := $(patsubst %,$(OUTPUT)/%,$(TEST_GEN_FILES))
 
+ifneq ($(TEST_GEN_PROGS),)
+$(shell mkdir -p $(sort $(dir $(TEST_GEN_PROGS))))
+endif
+ifneq ($(TEST_GEN_PROGS_EXTENDED),)
+$(shell mkdir -p $(sort $(dir $(TEST_GEN_PROGS_EXTENDED))))
+endif
+ifneq ($(TEST_GEN_FILES),)
+$(shell mkdir -p $(sort $(dir $(TEST_GEN_FILES))))
+endif
+
 ifdef KSFT_KHDR_INSTALL
 top_srcdir ?= ../../../..
 include $(top_srcdir)/scripts/subarch.include
-- 
2.20.1

[PATCH v2 2/6] selftests: kvm: Include lib.mk earlier

[Xiaoyao Li]

Include lib.mk a little bit earlier so that common lib.mk can setup
environment earlier.

This is for the following patch to prepare INSTALL_HDR_PATH earlier.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
---
 tools/testing/selftests/kvm/Makefile | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
index 67abc1dd50ee..504a5fb0fa71 100644
--- a/tools/testing/selftests/kvm/Makefile
+++ b/tools/testing/selftests/kvm/Makefile
@@ -43,6 +43,10 @@ TEST_GEN_PROGS_s390x += kvm_create_max_vcpus
 TEST_GEN_PROGS += $(TEST_GEN_PROGS_$(UNAME_M))
 LIBKVM += $(LIBKVM_$(UNAME_M))
 
+# After inclusion, $(OUTPUT) is defined and
+# $(TEST_GEN_PROGS) starts with $(OUTPUT)/
+include ../lib.mk
+
 INSTALL_HDR_PATH = $(top_srcdir)/usr
 LINUX_HDR_PATH = $(INSTALL_HDR_PATH)/include/
 LINUX_TOOL_INCLUDE = $(top_srcdir)/tools/include
@@ -62,9 +66,6 @@ pgste-option = $(call try-run, echo 'int main() { return 0; }' | \
 
 LDFLAGS += -pthread $(no-pie-option) $(pgste-option)
 
-# After inclusion, $(OUTPUT) is defined and
-# $(TEST_GEN_PROGS) starts with $(OUTPUT)/
-include ../lib.mk
 
 STATIC_LIBS := $(OUTPUT)/libkvm.a
 LIBKVM_OBJ := $(patsubst %.c, $(OUTPUT)/%.o, $(LIBKVM))
-- 
2.20.1

[PATCH v2 3/6] selftests: kvm: Use the default linux header path only when INSTALL_HDR_PATH not defined

[Xiaoyao Li]

This provides a way for lib.mk or parent Makefile to tell where the
linux headers are installed.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
---
 tools/testing/selftests/kvm/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/testing/selftests/kvm/Makefile b/tools/testing/selftests/kvm/Makefile
index 504a5fb0fa71..6f08cc512e15 100644
--- a/tools/testing/selftests/kvm/Makefile
+++ b/tools/testing/selftests/kvm/Makefile
@@ -47,7 +47,7 @@ LIBKVM += $(LIBKVM_$(UNAME_M))
 # $(TEST_GEN_PROGS) starts with $(OUTPUT)/
 include ../lib.mk
 
-INSTALL_HDR_PATH = $(top_srcdir)/usr
+INSTALL_HDR_PATH ?= $(top_srcdir)/usr
 LINUX_HDR_PATH = $(INSTALL_HDR_PATH)/include/
 LINUX_TOOL_INCLUDE = $(top_srcdir)/tools/include
 LINUX_TOOL_ARCH_INCLUDE = $(top_srcdir)/tools/arch/x86/include
-- 
2.20.1

[PATCH v2 4/6] selftests: Create variable INSTALL_HDR_PATH if need to install linux headers to $(OUTPUT)/usr

[Xiaoyao Li]

So sub TARGETs of selftest can know where the linux headers installed by
checking INSTALL_HDR_PATH.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
---
 tools/testing/selftests/lib.mk | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/lib.mk b/tools/testing/selftests/lib.mk
index 78fb00bdf5c4..b51ff93da4de 100644
--- a/tools/testing/selftests/lib.mk
+++ b/tools/testing/selftests/lib.mk
@@ -35,6 +35,12 @@ top_srcdir ?= ../../../..
 include $(top_srcdir)/scripts/subarch.include
 ARCH		?= $(SUBARCH)
 
+ifndef KSFT_KHDR_INSTALL_DONE
+ifneq (1,$(DEFAULT_INSTALL_HDR_PATH))
+INSTALL_HDR_PATH := $(OUTPUT)/usr
+endif
+endif
+
 # set default goal to all, so make without a target runs all, even when
 # all isn't the first target in the file.
 .DEFAULT_GOAL := all
@@ -59,7 +65,7 @@ ifndef KSFT_KHDR_INSTALL_DONE
 ifeq (1,$(DEFAULT_INSTALL_HDR_PATH))
 	make --no-builtin-rules ARCH=$(ARCH) -C $(top_srcdir) headers_install
 else
-	make --no-builtin-rules INSTALL_HDR_PATH=$$OUTPUT/usr \
+	make --no-builtin-rules INSTALL_HDR_PATH=$(INSTALL_HDR_PATH) \
 		ARCH=$(ARCH) -C $(top_srcdir) headers_install
 endif
 endif
-- 
2.20.1

[PATCH v2 5/6] selftests: Generate build output of linux headers to $(OUTPUT)/linux-header-build

[Xiaoyao Li]

When build linux headers for selftests with OUTPUT specified, it makes
linux src tree not clean.

Fix it by generating the output file of "headers_install" to a fixed dir
in $(OUTPUT).

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
---
 tools/testing/selftests/lib.mk | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tools/testing/selftests/lib.mk b/tools/testing/selftests/lib.mk
index b51ff93da4de..bb830e4817fc 100644
--- a/tools/testing/selftests/lib.mk
+++ b/tools/testing/selftests/lib.mk
@@ -66,6 +66,7 @@ ifeq (1,$(DEFAULT_INSTALL_HDR_PATH))
 	make --no-builtin-rules ARCH=$(ARCH) -C $(top_srcdir) headers_install
 else
 	make --no-builtin-rules INSTALL_HDR_PATH=$(INSTALL_HDR_PATH) \
+		O=$(OUTPUT)/linux-header-build \
 		ARCH=$(ARCH) -C $(top_srcdir) headers_install
 endif
 endif
-- 
2.20.1

[PATCH v2 6/6] selftests: export INSTALL_HDR_PATH if using "O" to specify output dir

[Xiaoyao Li]

When build kvm selftests in tools/testing/selftests directory with

	make O=/path/to/kselftests TARGETS=kvm

it fails building some kvm test binaries due to lack of header files.

Export INSTALL_HDR_PATH when "O" is specified, so that sub TARGET can get
the right kernel headers with INSTALL_HDR_PATH.

Signed-off-by: Xiaoyao Li <xiaoyao.li@intel.com>
---
 tools/testing/selftests/Makefile | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile
index be22dbe94a4c..f36bc6fd8086 100644
--- a/tools/testing/selftests/Makefile
+++ b/tools/testing/selftests/Makefile
@@ -110,6 +110,10 @@ ARCH           ?= $(SUBARCH)
 export KSFT_KHDR_INSTALL_DONE := 1
 export BUILD
 
+ifneq (1,$(DEFAULT_INSTALL_HDR_PATH))
+export	INSTALL_HDR_PATH := $(BUILD)/usr
+endif
+
 # build and run gpio when output directory is the src dir.
 # gpio has dependency on tools/gpio and builds tools/gpio
 # objects in the src directory in all cases making the src
@@ -142,7 +146,7 @@ khdr:
 ifeq (1,$(DEFAULT_INSTALL_HDR_PATH))
 	$(MAKE) --no-builtin-rules ARCH=$(ARCH) -C $(top_srcdir) headers_install
 else
-	$(MAKE) --no-builtin-rules INSTALL_HDR_PATH=$$BUILD/usr \
+	$(MAKE) --no-builtin-rules INSTALL_HDR_PATH=$(INSTALL_HDR_PATH) \
 		ARCH=$(ARCH) -C $(top_srcdir) headers_install
 endif
 
-- 
2.20.1

Re: [PATCH v2 0/6] Fix errors when try to build kvm selftests on specified output

[shuah]

On 3/25/20 8:01 AM, Xiaoyao Li wrote:
> I attempted to build KVM selftests on a specified dir, unfortunately
> neither	"make O=/path/to/mydir TARGETS=kvm" in tools/testing/selftests, nor
> "make OUTPUT=/path/to/mydir" in tools/testing/selftests/kvm work.
> 

Please elaborate on the problems you are seeing. I would like you
to describe in detail the problems you are seeing and how you are
fixing them in this patch series.

The problem you are fixing here is subdir structure not being
supported for relocatable builds and the Makefile not being
able to locate headers files. These are issues, however, these
need to be fixed in the kvm Makefile.

Please look at arm64, android, futex tests as examples. lib.mk
and main selftests Makefile allow for overrides for make targets.
When a test has sub-dir structure and libraries, it is easier to
handle these in the individual Makefile.

Please fix the problems you are seeing in kvm Makefile.

 >I only test the sub TARGET of kvm.
 >In theory, it won't break other TARGET of selftests.

When you change lib.mk which is a common infrastructure, theory
doesn't help. Statements like this make me very reluctant to
accept patches. :)

This is one reason why I asked Paolo to drop these patches.

thanks,
-- Shuah

Re: [PATCH v2 0/6] Fix errors when try to build kvm selftests on specified output

[Xiaoyao Li]

On 3/26/2020 12:12 AM, shuah wrote:
> On 3/25/20 8:01 AM, Xiaoyao Li wrote:
>> I attempted to build KVM selftests on a specified dir, unfortunately
>> neither    "make O=/path/to/mydir TARGETS=kvm" in 
>> tools/testing/selftests, nor
>> "make OUTPUT=/path/to/mydir" in tools/testing/selftests/kvm work.
>>
> 
> Please elaborate on the problems you are seeing. I would like you
> to describe in detail the problems you are seeing and how you are
> fixing them in this patch series.
> 
> The problem you are fixing here is subdir structure not being
> supported for relocatable builds and the Makefile not being
> able to locate headers files. These are issues, however, these
> need to be fixed in the kvm Makefile
Maybe I should have sent it as RFC.

OK, then I'll just resend patch 1 in v1 to fix the subdir not created 
issue for relocatable build, in kvm Makefile. Obviously lib.mk doesn't 
create it, and you don't allow me to touch lib.mk

About headers issue, since I'm not familiar with Makefile, I don't want 
to waste any more time on it and I decide to drop them.

I can add a "make mrproper" in my script to do the cleanup as a workaround.

And never make kvm selftests in parent dir with
	make O=/somewhere TARGETS=kvm
to workaround the header not found issue.

> Please look at arm64, android, futex tests as examples. lib.mk
> and main selftests Makefile allow for overrides for make targets.
> When a test has sub-dir structure and libraries, it is easier to
> handle these in the individual Makefile.
> 
> Please fix the problems you are seeing in kvm Makefile.
> 
>  >I only test the sub TARGET of kvm.
>  >In theory, it won't break other TARGET of selftests.
> 
> When you change lib.mk which is a common infrastructure, theory
> doesn't help. Statements like this make me very reluctant to
> accept patches. :)
> 
> This is one reason why I asked Paolo to drop these patches.
> 
> thanks,
> -- Shuah