Hi Daisuke, The trace.c extension module is no longer functional as of Linux 4.2 due to a huge rework of the ftrace code which, most notably, changed symbol and structure names from "ftrace..." to "trace...". Can you take a look at addressing those changes, while maintaining backwards compatibility? Thanks, Dave

Hello Dave, The patch looks good to me, and has been tested on several x86/x64 system. It Works well and I think I should add an Acked-by: Qiao Nuohan <qiaonuohan(a)cn.fujitsu.com&gt; -- Regards Qiao Nuohan

On 09/24/2015 04:43 AM, Dave Anderson wrote: > > > ----- Original Message ----- >> >> Hi Daisuke, >> >> The trace.c extension module is no longer functional as of Linux 4.2 >> due to a huge rework of the ftrace code which, most notably, changed >> symbol and structure names from "ftrace..." to "trace...". >> >> Can you take a look at addressing those changes, while maintaining >> backwards compatibility? >> >> Thanks, >> Dave > > > Hi Daisuke, > > I *think* that the attached patch may be enough to handle the kernel > changes. > It uses this construct in the relevant locations to pick the correct > version: > > MAX(MEMBER_OFFSET("ftrace_event_call", ...), > MEMBER_OFFSET("trace_event_call", ...)) > > Note that I also added a gdb_set_crash_scope() call to init_offsets() > because there are multiple, different, "struct ring_buffer" declarations in > the kernel, and for whatever reason, crash was initially picking up the > wrong > one from "kernel/events/internal.h". > > Can you verify that this patch is sufficient and test its functionality? > Or if you prefer, can you come up with your own solution? > > Thanks, > Dave > > > > -- > Crash-utility mailing list > Crash-utility(a)redhat.com > https://www.redhat.com/mailman/listinfo/crash-utility >