Skip to contentSkip to conversation

kbuild: fix make -jN with multiple targets with make O=...

PATCH In review Sam Ravnborg 3 participants 5 messages Jan 15, 2006 – Jan 16, 2006 Kbuild

lore
Share & tools
lore
#0 lore raw

PATCH 2/2 kbuild: fix make -jN with multiple targets with make O=...

[It is pushed out at:
git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild.git]

The way multiple targets was handled with make O=... broke because for each high-level target make spawned a parallel make resulting in a broken build. Reported by Keith Owens <[email protected]>

---

 Makefile |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

296e0855b0f9a4ec9be17106ac541745a55b2ce1

Makefile +4 -3
index deedaf7..b3dd9db 100644
@@ -106,12 +106,13 @@ KBUILD_OUTPUT := $(shell cd $(KBUILD_OUT
 $(if $(KBUILD_OUTPUT),, \
      $(error output directory "$(saved-output)" does not exist))
 
-.PHONY: $(MAKECMDGOALS)
+.PHONY: $(MAKECMDGOALS) cdbuilddir
+$(MAKECMDGOALS) _all: cdbuilddir
 
-$(filter-out _all,$(MAKECMDGOALS)) _all:
+cdbuilddir:
 	$(if $(KBUILD_VERBOSE:1=),@)$(MAKE) -C $(KBUILD_OUTPUT) \
 	KBUILD_SRC=$(CURDIR) \
-	KBUILD_EXTMOD="$(KBUILD_EXTMOD)" -f $(CURDIR)/Makefile $@
+	KBUILD_EXTMOD="$(KBUILD_EXTMOD)" -f $(CURDIR)/Makefile $(MAKECMDGOALS)
 
 # Leave processing to above invocation of make
 skip-makefile := 1
Signature · 1 line
1.0.GIT
#1 lore raw
Quoted 7 lines · Sam Ravnborg

[It is pushed out at:
git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild.git]

The way multiple targets was handled with make O=... broke because for each high-level target make spawned a parallel make resulting in a broken build. Reported by Keith Owens <[email protected]>

When did it break? Are any of the released (not -git) kernels affected?

Quoted 19 lines

---

 Makefile |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

296e0855b0f9a4ec9be17106ac541745a55b2ce1

Makefile +2 -1
index deedaf7..b3dd9db 100644
@@ -106,12 +106,13 @@ KBUILD_OUTPUT := $(shell cd $(KBUILD_OUT
 $(if $(KBUILD_OUTPUT),, \
      $(error output directory "$(saved-output)" does not exist))
 
-.PHONY: $(MAKECMDGOALS)
+.PHONY: $(MAKECMDGOALS) cdbuilddir
+$(MAKECMDGOALS) _all: cdbuilddir

Makefile KBUILD_OUTPUT := $(shell cd $(KBUILD_OUT · line 118

Quoted 10 lines
Makefile +2 -2
 
-$(filter-out _all,$(MAKECMDGOALS)) _all:
+cdbuilddir:
 	$(if $(KBUILD_VERBOSE:1=),@)$(MAKE) -C $(KBUILD_OUTPUT) \
 	KBUILD_SRC=$(CURDIR) \
-	KBUILD_EXTMOD="$(KBUILD_EXTMOD)" -f $(CURDIR)/Makefile $@
+	KBUILD_EXTMOD="$(KBUILD_EXTMOD)" -f $(CURDIR)/Makefile $(MAKECMDGOALS)
 
 # Leave processing to above invocation of make
 skip-makefile := 1

Kalin.

Signature · 3 lines
|[ ~~~~~~~~~~~~~~~~~~~~~~ ]|
+-> http://ThinRope.net/ <-+
|[ ______________________ ]|
#2 lore raw
Quoted 10 lines · Kalin KOZHUHAROV (on Mon, 16 Jan 2006 11:09:51 +0900) wrote:
Quoted 7 lines · Sam Ravnborg

[It is pushed out at:
git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild.git]

The way multiple targets was handled with make O=... broke because for each high-level target make spawned a parallel make resulting in a broken build. Reported by Keith Owens <[email protected]>

When did it break? Are any of the released (not -git) kernels affected?

2.6.15 has the problem. It only triggers with the combination of a separate object directory _and_ multiple targets on the make command line _and_ running make in parallel (make -j).

#3 lore raw
Quoted 9 lines · Keith Owens
Quoted 8 lines · Kalin KOZHUHAROV (on Mon, 16 Jan 2006 11:09:51 +0900) wrote:
Quoted 7 lines · Sam Ravnborg

[It is pushed out at:
git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild.git]

The way multiple targets was handled with make O=... broke because for each high-level target make spawned a parallel make resulting in a broken build. Reported by Keith Owens <[email protected]>

Quoted 5 lines
Quoted 1 line

When did it break? Are any of the released (not -git) kernels affected?

2.6.15 has the problem. It only triggers with the combination of a separate object directory _and_ multiple targets on the make command line _and_ running make in parallel (make -j).

Thanks for the clarification!

I am safe as I don't usually use multiple targets, but I always compile with KBUILD_OUTPUT set and make -j5 (and distcc).

However, some of the out-of-tree modules might break, will stay on alert. A quick check through the relevant to my hardware ebuilds (/me on Gentoo) showed no show stoppers, they all do:

for T in $TARGETS; do make $T; done

(MAKEOPTS=-jN is handled internally on Gentoo if configured; and I use /var/kernels/out to compile my kernels)

Please get that patch into 2.6.15.2 if possible (seems many people have ppp problems, so I guess that will be released soon).

Kalin.

Signature · 3 lines
|[ ~~~~~~~~~~~~~~~~~~~~~~ ]|
+-> http://ThinRope.net/ <-+
|[ ______________________ ]|
#4 lore raw
Quoted 5 lines · Kalin KOZHUHAROV
Quoted 5 lines

When did it break? Are any of the released (not -git) kernels affected?

2.6.15 has the problem. It only triggers with the combination of a separate object directory _and_ multiple targets on the make command line _and_ running make in parallel (make -j).

I have not tested, but I beleive the bug has been present since day one of the make O=... feature inclusion.

Quoted 2 lines

Please get that patch into 2.6.15.2 if possible (seems many people have ppp problems, so I guess that will be released soon).

Since it has been present for long then we are not in a big hurry. One report in many months.

Sam