[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <e3ae190e-0888-3d4a-e969-9604d4ab8695@acm.org>
Date: Mon, 18 Nov 2019 15:14:05 -0800
From: Bart Van Assche <bvanassche@....org>
To: Zhengyuan Liu <liuzhengyuan@...inos.cn>, mingo <mingo@...nel.org>,
"alexander.levin" <alexander.levin@...rosoft.com>
Cc: linux-kernel <linux-kernel@...r.kernel.org>,
Peter Zijlstra <peterz@...radead.org>
Subject: Re: Compilation error for target liblockdep
On 11/18/19 1:20 AM, Zhengyuan Liu wrote:
> I got a compilation error while building target liblockdep and I think I'd
> better report it to you. The error info showed as bellow:
>
> # cd SRC/tools
> # make liblockdep
> DESCEND lib/lockdep
> CC lockdep.o
> In file included from lockdep.c:33:0:
> ../../../kernel/locking/lockdep.c:53:28: fatal error: linux/rcupdate.h: No such file or directory
> compilation terminated.
> mv: cannot stat './.lockdep.o.tmp': No such file or directory
> /home/lzy/kernel-upstream/linux-linus-ubuntu/tools/build/Makefile.build:96: recipe for target 'lockdep.o' failed
> make[2]: *** [lockdep.o] Error 1
> Makefile:121: recipe for target 'liblockdep-in.o' failed
> make[1]: *** [liblockdep-in.o] Error 2
> Makefile:68: recipe for target 'liblockdep' failed
> make: *** [liblockdep] Error 2
>
> BTW, It was introduced by commit a0b0fd53e1e ("locking/lockdep: Free lock classes that are no longer in use").
(+Peter)
Hi Zhengyuan Liu,
The approach of liblockdep is fragile. Every time an additional kernel
header is included from the lockdep code or a change is made in one of
the kernel headers used by lockdep, that change has to be ported to the
include files in the tools/lib/lockdep/include/liblockdep/ directory. I
think there are two possible solutions:
- Making the changes necessary to make liblockdep build again.
- Removing the code under tools/lib/lockdep and porting this code to the
new KUnit framework. If I understood the KUnit framework correctly it
is based on UML and hence does not require kernel headers to be
duplicated.
I'm not sure what the best approach is.
Thanks,
Bart.
Powered by blists - more mailing lists