[Lustre-discuss] Building Lustre client on Linux kernel 3.14

Bartek bgrinning at gmail.com
Mon Jun 2 06:02:49 PDT 2014


Thanks Tao. I ended up fixing the "CONFIG_LNET_MAX_PAYLOAD redefined" error
by locally modifying Lustre configuration scripts because in an cloud
environment I'm working with rebuilding the kernel is not an option.

Also, in case anyone is interested in building the Lustre client on Debian
Wheezy with kernel 3.14-0.bpo.1-amd64, in addition to the above fix and the
patch I mentioned in my first message I also had to apply this:
http://review.whamcloud.com/#/c/10325 to fix the following error:
"lustre/llite/file.c:3545:24: error: invalid operands to binary == (have
‘kuid_t’ and ‘int’) current_fsuid() == 0 && .." More info can be found
here: https://jira.hpdd.intel.com/browse/LU-5058

Now moving on to functional testing the client...

Cheers,
Bartek


2014-05-31 18:56 GMT+02:00 Peng Tao <bergwolf at gmail.com>:

> On Thu, May 29, 2014 at 6:15 PM, Bartek <bgrinning at gmail.com> wrote:
> > Hello,
> >
> > I'm trying to build Lustre client on Debian Wheezy with kernel
> > 3.14-0.bpo.1-amd64. Here are my build steps are:
> > sh ./autogen.sh
> > sudo ./configure --disable-server
> > sudo dpkg-buildpackage
> > sudo dpkg -i lustre-source_2.5.59.0-32-gb265903_all.deb
> > sudo m-a -t build lustre
> > I also patched debian/rules to: ./configure --disable-server
> > --disable-ldfiskfs
> >
> > I'm aware of changes needed for latest kernels (3.12-14):
> > https://jira.hpdd.intel.com/browse/LU-4416
> > https://jira.hpdd.intel.com/browse/LU-4993
> >
> > So far, I applied the shrinker patch locally:
> > http://review.whamcloud.com/#/c/9300 and that allowed me to get past the
> > shrinker compilation error. But now I'm running into another error
> (error:
> > "CONFIG_LNET_MAX_PAYLOAD" redefined) for which I didn't find any
> mentions in
> > Jira:
> >
> > ...
> > echo "#define LUSTRE_RELEASE 3.14_0.bpo.1_amd64_gb265903" >> tmpver
> > cmp -s ../lustre/include/lustre/lustre_build_version.h tmpver > tmpdiff
> 2>
> > /dev/null &&   \
> >                 rm -f tmpver tmpdiff ||         \
> >     mv -f tmpver ../lustre/include/lustre/lustre_build_version.h
> > make[5]: Leaving directory `/usr/src/modules/lustre/lustre'
> > /usr/bin/make CC="gcc"  -C /lib/modules/3.14-0.bpo.1-amd64/build       \
> >   -f /usr/src/modules/lustre/build/Makefile
> > LUSTRE_LINUX_CONFIG=/lib/modules/3.14-0.bpo.1-amd64/build/.config \
> >   LINUXINCLUDE='-I$(srctree)/arch/$(SRCARCH)/include
> > -I$(srctree)/arch/$(SRCARCH)/include/generated -Iinclude $(if
> > $(KBUILD_SRC),-Iinclude2 -I$(srctree)/include)
> > -I$(srctree)/arch/$(SRCARCH)/include/uapi
> > -Iarch/$(SRCARCH)/include/generated/uapi -I$(srctree)/include/uapi
> > -Iinclude/generated/uapi -include
> > /lib/modules/3.14-0.bpo.1-amd64/source/include/linux/kconfig.h' \
> >   M=/usr/src/modules/lustre -o tmp_include_depends -o scripts -o \
> >   include/config/MARKER modules
> > make[5]: Entering directory `/usr/src/linux-headers-3.14-0.bpo.1-amd64'
> >   CC [M]  /usr/src/modules/lustre/libcfs/libcfs/linux/linux-tracefile.o
> > In file included from <command-line>:0:0:
> > /usr/src/modules/lustre/config.h:26:0: error: "CONFIG_LNET_MAX_PAYLOAD"
> > redefined [-Werror]
> > In file included from
> > /lib/modules/3.14-0.bpo.1-amd64/source/include/linux/kconfig.h:4:0,
> It seems that you have enabled CONFIG_LUSTRE_FS in your local kernel
> config which would then define CONFIG_LNET_MAX_PAYLOAD on its own. The
> easy workaround is to disable in-kernel client in your kernel and
> rebuild.
> I am not sure if it makes sense to support such use case as there
> would be two sets of client modules for the same kernel.
>
> Cheers,
> Tao
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20140602/51fe58ad/attachment.htm>


More information about the lustre-discuss mailing list