From: Bart V. A. <bar...@gm...> - 2009-10-31 19:27:29
|
On Sat, Oct 31, 2009 at 8:18 PM, Gennadiy Nerubayev <pa...@gm...>wrote: > On Sat, Oct 31, 2009 at 8:08 AM, Bart Van Assche <bar...@gm... > > wrote: > >> On Tue, Oct 27, 2009 at 2:42 PM, Gennadiy Nerubayev <pa...@gm...>wrote: >> >>> On Tue, Oct 27, 2009 at 2:07 AM, Bart Van Assche < >>> bar...@gm...> wrote: >>> >>>> On Tue, Oct 27, 2009 at 2:55 AM, Gennadiy Nerubayev <pa...@gm...>wrote: >>>> >>>>> Has anyone tried yet to get ib_srpt working under 2.6.30.x (.9 in my >>>>> case) and 1.5 OFED? It compiles fine after the prerequisite addition of >>>>> PRE_CFLAGS, but on modprobe I get a GPF (see attached). >>>>> >>>> >>>> SRPT works fine with the 2.6.30.x and 2.6.31.x kernels without OFED >>>> 1.5. OFED 1.5 has not yet been released. Have you been testing OFED 1.5 >>>> alpha3, alpha4, beta1 or rc1 ? >>>> >>> >>> I've been testing with rc1 and up; currently on the most recent daily as >>> of yesterday. I need to use OFED instead of the in-kernel bits because I >>> need the SDP functionality in OFED, amongst a few other things :( >>> >> >> Hello Gennadiy, >> >> I have updated README.ofed, the Makefile and the patches directory inside >> the srpt directory. Can you please verify whether SCST-SRPT does now work >> with OFED 1.5 ? >> > > Hi Bart, > > Module inserts with no problems and ibsrpdm shows the targets, so it looks > good :) > Thanks for the feedback. The cause of the crash was that the Linux kernel include path and the OFED 1.5 include path were specified in the wrong order. As a result, the function ib_dma_map_single(), invoked from srpt_add_one(), was reading ib_device::dma_ops from the wrong offset (the struct ib_device declared in the OFED 1.5 header include/rdma/ib_verbs.h differs from the struct ib_device declared in the 2.6.30 header include/rdma/ib_verbs.h). Bart. |