[LU-2631] build fails on 2.6.18-348.el5 with o2ib Created: 17/Jan/13  Updated: 18/Feb/13  Resolved: 18/Feb/13

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 1.8.x (1.8.0 - 1.8.5)
Fix Version/s: Lustre 1.8.9

Type: Bug Priority: Minor
Reporter: Shuichi Ihara (Inactive) Assignee: Yang Sheng
Resolution: Fixed Votes: 0
Labels: patch
Environment:

2.6.18-348.el5 kernel (CentOS5.9)


Severity: 3
Rank (Obsolete): 6157

 Description   

Client build fails for 2.6.18-348.el5 if o2ib is enabled.
Becouse, there is small changes in 2.6.18-348.el5 kernel that vlan_dev_vlan_id() is added in rdma_vlan_dev_vlan_id() in include/rdma/ib_addr.h, but vlan_dev_vlan_id() is defined in scsi/fc_compat.h that is no definition of lustre-lnet.m4 and o2iblnd.h.



 Comments   
Comment by Shuichi Ihara (Inactive) [ 17/Jan/13 ]

submitted patches. http://review.whamcloud.com/5051

Comment by Peter Jones [ 17/Jan/13 ]

Yangsheng

Is this a duplicate with the work you are doing under LU-2601 to add support for RHEL5.9 to b1_8?

Peter

Comment by Yang Sheng [ 17/Jan/13 ]

Hi, Peter, I think so. But since our build infrastructure still not ready for this kernel. I have not trying it with OFED. So looks like this patch is needed.

Comment by Yang Sheng [ 17/Jan/13 ]

From the http://build.whamcloud.com/job/lustre-reviews/11942/

The ofa build looks like success. But it obviously isn't a normal case. The ofa + patchless unit should also failed since kernel-devel rpms absence. Follow the link, We can see without any rpms been built, just README is in place. I don't know why it to do so. Is it a issue?

Comment by Shuichi Ihara (Inactive) [ 18/Jan/13 ]

updated patches. The first patch didn't work for external OFED.

This only allows adding "#include <scsi/fc_compat.h>" if o2ib is compiled with in-kernel OFED and scsi/fc_compat.h is exist. We skip when it's compiled on any RHEL6.x and external OFED with RHEL5.x and RHEL6.x.

Comment by Yang Sheng [ 26/Jan/13 ]

Hi, Shuichi,

As below link: http://build.whamcloud.com/job/lustre-reviews/12548/

The rhel6 patchless build is both successful on ofa & inkernel. The rhel5 is both successful on inkernel server & client build. Do you sure also same as your case?

BTW: Joshua, I found rhel6 server ofa & inkernel build indicted successful but doesn't product anythings. And rhel5 ofa server & client also build nothing. Just 'README' is present. Any problem in there? Or it is a normal case?

Comment by Shuichi Ihara (Inactive) [ 30/Jan/13 ]

Yang,
in-kernel ofed is no problem on rhel5.9's kenrel. when you build with external ofed, it should fail. this doesn't matter wether it's for server and client's module.

Comment by Peter Jones [ 01/Feb/13 ]

Ihara

Which version of OFED are you using? IIUC OFA OFED does not have a release that supports RHEL5.9...

Peter

Comment by Shuichi Ihara (Inactive) [ 02/Feb/13 ]

Peter,

Sorry confusion. I have tested MLNX_OFED_LINUX-1.5.3-3.1.0-rhel5.8 and it worked, but you are right. MLNX_OFED_LINUX-1.5.3-3.1.0-rhel5.8 doesn't support RHEL5.9 even it worked with RHEL5.9. I hope Mellanox will release same stack for rhel5.9 soon , but agree this is not offically supported at this morment.

Howerver, I think this patch http://review.whamcloud.com/5051 are still needed for RHEL5.9's in-kernel OFED.

Yangsheng mentioned below.

As below link: http://build.whamcloud.com/job/lustre-reviews/12548/

The rhel6 patchless build is both successful on ofa & inkernel. The rhel5 is both successful on inkernel server & client build. Do you sure also same as your case?

It looks like that build RPMs well, but OpenIB gen2 is disabled due to test by 'configure' fails. Please see the console log of the latest b1_8 build. http://build.whamcloud.com/job/lustre-b1_8/
(from rhel5's inkenrel server full console log, http://build.whamcloud.com/job/lustre-b1_8/252/arch=x86_64,build_type=server,distro=el5,ib_stack=inkernel/consoleFull)

checking whether to enable GM support... no
checking whether to enable OpenIB support... no
checking whether to enable Cisco/TopSpin IB support... no
checking whether to enable Voltaire IB support... no
checking whether to enable Infinicon support... no
checking whether to enable OpenIB gen2 support... no

It seems that o2ib is disabled on any latest b1_8 builds and can't enable it. This is an defect by LU-2601.

Please see, if we compile lustre modules with '--with-o2ib" for rhel5.9 kernels by hand, it still fails.

  1. ./configure --with-o2ib --disable-server --with-linux=/usr/src/kernels/2.6.18-348.1.1.el5-x86_64/
    - snip -
    checking whether to enable Cisco/TopSpin IB support... no
    checking whether to enable Voltaire IB support... no
    checking whether to enable Infinicon support... no
    checking whether to enable OpenIB gen2 support... no
    configure: error: can't compile with kernel OpenIB gen2 headers
    

according to config.log. this is issue of LU-2631.

/usr/src/kernels/2.6.18-348.1.1.el5-x86_64/include/rdma/ib_addr.h: In function 'rdma_vlan_dev_vlan_id':
/usr/src/kernels/2.6.18-348.1.1.el5-x86_64/include/rdma/ib_addr.h:154: error: implicit declaration of func
tion 'vlan_dev_vlan_id'
/lustre-release/build/conftest.c: In function 'main':
/lustre-release/build/conftest.c:53: warning: unused variable 'rej_reason'
/lustre-release/build/conftest.c:52: warning: unused variable 'pool_fmr'
/lustre-release/build/conftest.c:51: warning: unused variable 'qp_attr'
/lustre-release/build/conftest.c:50: warning: unused variable 'device_attr'
/lustre-release/build/conftest.c:49: warning: unused variable 'conn_param'
/lustre-release/build/conftest.c:48: warning: unused variable 'cm_id'
make[1]: *** [/lustre-release/build/conftest.o] Error 1
make: *** [_module_/lustre-release/build] Error 2
Comment by Yang Sheng [ 02/Feb/13 ]

Hi,ihara, Could you please re-submit http://review.whamcloud.com/#change,5051 to verify the inkernel ofed build is ok? TIA

Comment by Shuichi Ihara (Inactive) [ 02/Feb/13 ]

Hi YangSheng, I just rebased and pushed patches again. thanks!

Comment by Peter Jones [ 18/Feb/13 ]

Landed for 1.8.9. 2.1.x landing covered as part of RHEL5.x kernel update

Generated at Sat Feb 10 01:26:51 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.