OFED 1.3 w/ a Lustre kernel (was Re: [ofa-general] kernel ib build (OFED 1.3) fails on SLES 10)

[Chris Worley]

I'm building OFED 1.3 for an RHEL5.1 kernel for Lustre 1.6.4.3:
2.6.18-53.1.13.el5_lustre.1.6.4.3smp.

The install.pl script errors-out at the end of building the kernel
modules RPM saying some built modules don't exist, but those modules
do get built; the only other warning is some undefined symbols coming
out of the modpost command.  The differences between the .configs of
the two kernels are minimal, so I think it's the same problem as the
attached (not getting the proper patch files ).

What's the work-around?

Chris
P.S. Here's a sample of the output of the modpost command:

WARNING: "scst_register_target_template"
[/var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/drivers/infiniband/ulp/srpt/ib_srpt.ko]
undefined!
WARNING: "scst_unregister"
[/var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/drivers/infiniband/ulp/srpt/ib_srpt.ko]
undefined!
                                       WARNING: "scst_register"
[/var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/drivers/infiniband/ulp/srpt/ib_srpt.ko]
undefined!
                                         WARNING:
"scst_unregister_target_template"
[/var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/drivers/infiniband/ulp/srpt/ib_srpt.ko]
undefined!
                       WARNING: "scst_register_session"
[/var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/drivers/infiniband/ulp/srpt/ib_srpt.ko]
undefined!
                                 WARNING: "scst_rx_mgmt_fn"
[/var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/drivers/infiniband/ulp/srpt/ib_srpt.ko]
undefined!
                                       WARNING: "scst_cmd_init_done"
[/var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/drivers/infiniband/ulp/srpt/ib_srpt.ko]
undefined!

Here's the errors at the end:

RPM build errors:
...
    File not found:
/var/tmp/OFED/lib/modules/2.6.18-53.1.13.el5_lustre.1.6.4.3smp/updates/kernel/drivers/scsi/iscsi_tcp.ko
    File not found:
/var/tmp/OFED/lib/modules/2.6.18-53.1.13.el5_lustre.1.6.4.3smp/updates/kernel/drivers/scsi/libiscsi.ko
    File not found:
/var/tmp/OFED/lib/modules/2.6.18-53.1.13.el5_lustre.1.6.4.3smp/updates/kernel/drivers/scsi/scsi_transport_iscsi.ko
    File not found:
/var/tmp/OFED/lib/modules/2.6.18-53.1.13.el5_lustre.1.6.4.3smp/updates/kernel/net/rds
    File not found:
/var/tmp/OFED/lib/modules/2.6.18-53.1.13.el5_lustre.1.6.4.3smp/updates/kernel/drivers/net/cxgb3
    File not found:
/var/tmp/OFED/lib/modules/2.6.18-53.1.13.el5_lustre.1.6.4.3smp/updates/kernel/drivers/net/mlx4

On Tue, Apr 8, 2008 at 8:43 AM, Brian J. Murrell <Brian.Murrell at sun.com> wrote:
> On Tue, 2008-04-08 at 10:13 +0200, Thomas Großmann wrote:
>> Hi,
>
> Hi
>
>> kernel ib build (OFED 1.3) fails on SLES 10.
>
> To be fair, it fails on Sun's version of the SLES 10 kernel for Lustre,
> and here is why:
>
>> Executing(%build): /bin/sh -e /var/tmp/rpm-tmp.52212
>> + umask 022
>> + cd /var/tmp/OFED_topdir/BUILD
>> + /bin/rm -rf /var/tmp/OFED
>> ++ dirname /var/tmp/OFED
>> + /bin/mkdir -p /var/tmp
>> + /bin/mkdir /var/tmp/OFED
>> + cd ofa_kernel-1.3
>> + rm -rf /var/tmp/OFED
>> + cd /var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3
>> + mkdir -p /var/tmp/OFED//usr/local/ofed-1.3/src
>> + cp -a /var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3 /var/tmp/OFED//usr/local/ofed-1.3/src
>> + ./configure --prefix=/usr/local/ofed-1.3 --kernel-version 2.6.16-54-0.2.5_lustre.1.6.4.3smp --kernel-sources /lib/modules/2.6.16-54-0.2.5_lustre.1.6.4.3smp/build --modules-dir /lib/modules/2.6.16-54-0.2.5_lustre.1.6.4.3smp/updates --with-core-mod --with-user_mad-mod --with-user_access-mod --with-addr_trans-mod --with-mthca-mod --with-mlx4-mod --with-cxgb3-mod --with-nes-mod --with-ipoib-mod --with-sdp-mod --with-srp-mod --with-srp-target-mod --with-rds-mod --with-qlgc_vnic-mod
>> ofed_patch.mk does not exist. running ofed_patch.sh
>> /var/tmp/OFED_topdir/BUILD/ofa_kernel-1.3/ofed_scripts/ofed_patch.sh --kernel-version 2.6.16-54-0.2.5_lustre.1.6.4.3smp
> ----------------------------------------------------------------------------------------------------------------------^
> This kernel version does match what ofed_patch.sh thinks is a SLES 10
> kernel because it is not of the form "2.6.16.*-*-*".  Here's the code in
> ofed_patch.sh which detects SLES 10 kernels and assigns the right patch
> series for it:
>
>        2.6.16.*-*-*)
>                minor=$(echo $KVERSION | cut -d"." -f4 | cut -d"-" -f1)
>                if [ $minor -lt 37 ]; then
>                        echo 2.6.16_sles10
>                elif [ $minor -lt 60 ]; then
>                        echo 2.6.16_sles10_sp1
>                else
>                        echo 2.6.16_sles10_sp2
>                fi
>        ;;
>
> The lustre kernel version for SLES 10 is
> "2.6.16-54-0.2.5_lustre.1.6.4.3smp".  In order for it to match the above
> code it needs to have a "-" put before the "smp" at the end.  I am
> working on the Lustre build process to do exactly this right at this
> moment as well as build our released RPMs with OFED 1.3 support right in
> them.  My work is being done in Lustre bugzilla ticket 15316.  When I
> have something working, I will post an attachment there with a patch for
> our current b1_6 that should apply to 1.6.4.3.
>
> In theory you should be able use the "--with-backport*" configure
> options to override this detection when building the RPMs however see my
> message to this list (inconsistent use of --with-backport[-patches])
> last Saturday about how this seems to be broken currently.
>
> Cheers,
> b.
>
>
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
>