Re: [Xenomai-core] P2020 support for ftrace with ipipe 2.12-01 and xeno 2.5.5.1
On 2011-09-23 16:21, Jean-Michel Hautbois wrote: > 2011/9/23 Jan Kiszka mailto:jan.kis...@siemens.com>> > On 2011-09-23 15:58, Jean-Michel Hautbois wrote: >> 2011/9/23 Gilles Chanteperdrix >> mailto:gilles.chanteperd...@xenomai.org>> >> >>> On 09/23/2011 11:49 AM, Jean-Michel Hautbois wrote: OK, I have more traces (a few :)) : >>> >>> I meant the I-pipe tracer alone. The I-pipe tracer intead of other >>> ftrace tracers. >>> >>> >> Well, I think it works, I can do a cat /proc/ipipe/trace/max without any >> error in a kernel which doesn't have any other ftrace function. >> Do you have one test in particular in mind ? > > Check > http://git.kiszka.org/?p=ipipe.git;a=shortlog;h=refs/heads/queues/2.6.35-x86-trace > to get an impression of what is required to get ftrace working on > x86_64. It is not working on x86_32 e.g. as it depends on the arch > providing a NMI-safe, ie. task-stack independent current_thread_info(). > That is also not yet the case on Power. > > Interesting, and impressive. Do you plan supporting it for PowerPC ? > Is there a git branch for that too ? > I can test on a P2020, if you are interested ;-). Sorry, this is currently not on any local roadmap. Jan -- Siemens AG, Corporate Technology, CT T DE IT 1 Corporate Competence Center Embedded Linux ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] P2020 support for ftrace with ipipe 2.12-01 and xeno 2.5.5.1
2011/9/23 Jan Kiszka > On 2011-09-23 15:58, Jean-Michel Hautbois wrote: > > 2011/9/23 Gilles Chanteperdrix > > > >> On 09/23/2011 11:49 AM, Jean-Michel Hautbois wrote: > >>> OK, I have more traces (a few :)) : > >> > >> I meant the I-pipe tracer alone. The I-pipe tracer intead of other > >> ftrace tracers. > >> > >> > > Well, I think it works, I can do a cat /proc/ipipe/trace/max without any > > error in a kernel which doesn't have any other ftrace function. > > Do you have one test in particular in mind ? > > Check > > http://git.kiszka.org/?p=ipipe.git;a=shortlog;h=refs/heads/queues/2.6.35-x86-trace > to get an impression of what is required to get ftrace working on > x86_64. It is not working on x86_32 e.g. as it depends on the arch > providing a NMI-safe, ie. task-stack independent current_thread_info(). > That is also not yet the case on Power. > Interesting, and impressive. Do you plan supporting it for PowerPC ? Is there a git branch for that too ? I can test on a P2020, if you are interested ;-). JM ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] P2020 support for ftrace with ipipe 2.12-01 and xeno 2.5.5.1
On 2011-09-23 15:58, Jean-Michel Hautbois wrote: > 2011/9/23 Gilles Chanteperdrix > >> On 09/23/2011 11:49 AM, Jean-Michel Hautbois wrote: >>> OK, I have more traces (a few :)) : >> >> I meant the I-pipe tracer alone. The I-pipe tracer intead of other >> ftrace tracers. >> >> > Well, I think it works, I can do a cat /proc/ipipe/trace/max without any > error in a kernel which doesn't have any other ftrace function. > Do you have one test in particular in mind ? Check http://git.kiszka.org/?p=ipipe.git;a=shortlog;h=refs/heads/queues/2.6.35-x86-trace to get an impression of what is required to get ftrace working on x86_64. It is not working on x86_32 e.g. as it depends on the arch providing a NMI-safe, ie. task-stack independent current_thread_info(). That is also not yet the case on Power. Jan -- Siemens AG, Corporate Technology, CT T DE IT 1 Corporate Competence Center Embedded Linux ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] P2020 support for ftrace with ipipe 2.12-01 and xeno 2.5.5.1
2011/9/23 Gilles Chanteperdrix > On 09/23/2011 11:49 AM, Jean-Michel Hautbois wrote: > > OK, I have more traces (a few :)) : > > I meant the I-pipe tracer alone. The I-pipe tracer intead of other > ftrace tracers. > > Well, I think it works, I can do a cat /proc/ipipe/trace/max without any error in a kernel which doesn't have any other ftrace function. Do you have one test in particular in mind ? JM ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] P2020 support for ftrace with ipipe 2.12-01 and xeno 2.5.5.1
On 09/23/2011 11:49 AM, Jean-Michel Hautbois wrote: > OK, I have more traces (a few :)) : I meant the I-pipe tracer alone. The I-pipe tracer intead of other ftrace tracers. -- Gilles. ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] P2020 support for ftrace with ipipe 2.12-01 and xeno 2.5.5.1
Yes, ipipe tracing works. When it is the case, BTW, Here is the result of echoing a new tracer : $>cat max I-pipe worst-case tracing service on 2.6.35.11/ipipe-2.12-01 -- -- CPU: 0, Begin: 2593605556 cycles, Trace Points: 12 (-10/+5), Length: 268166 us Calibrated minimum trace-point overhead: 0.023 us +- Hard IRQs ('|': locked) |+ ||+--- |||+-- Xenomai +- Linux ('*': domain stalled, '+': current, '#': current+stalled) |+-- Delay flag ('+': > 1 us, '!': > 10 us) ||+- NMI noise ('N') ||| TypeUser Val. TimeDelay Function (Parent) #func 00.039 ipipe_check_context+0x8 (tick_nohz_stop_sched_tick+0x31c) | #begin 0x8001 00.051 ipipe_check_context+0xf0 (tick_nohz_stop_sched_tick+0x31c) | #end 0x8001 00.039 ipipe_check_context+0xdc (tick_nohz_stop_sched_tick+0x31c) #func 00.039 __ipipe_unstall_root+0x8 (tick_nohz_stop_sched_tick+0x320) | #begin 0x8000 00.043 __ipipe_unstall_root+0xbc (tick_nohz_stop_sched_tick+0x320) | #func 00.058 ipipe_check_context+0x8 (__ipipe_unstall_root+0x3c) | +end 0x8000 00.046 __ipipe_unstall_root+0xa8 (tick_nohz_stop_sched_tick+0x320) +func 00.041 ipipe_suspend_domain+0x8 (cpu_idle+0xb0) | +begin 0x8001 00.060 ipipe_suspend_domain+0x1e0 (cpu_idle+0xb0) | +end 0x8001 00.043 ipipe_suspend_domain+0x220 (cpu_idle+0xb0) >| +begin 0x8000 0! 2146.524 cpu_idle+0xe8 (rest_init+0x94) :| +func 2681660.045 __ipipe_grab_timer+0x8 (__ipipe_ret_from_except+0x0) :| +begin 0x0200 2681660.046 __ipipe_grab_timer+0x84 (__ipipe_ret_from_except+0x0) :| +func 2681660.060 __ipipe_dispatch_wired_nocheck+0x8 (__ipipe_grab_timer+0x138) :| # func 2681660.054 xnintr_clock_handler+0x8 (__ipipe_dispatch_wired_nocheck+0x104) :| # func 2681660.050 xntimer_tick_aperiodic+0x8 (xnintr_clock_handler+0xd0) :| # func 2681660.060 xntimer_next_local_shot+0x8 (xntimer_tick_aperiodic+0x304) :| # func 2681660.049 xnintr_host_tick+0x8 (xnintr_clock_handler+0x1f4) :| # func 2681660.085 __ipipe_set_irq_pending+0x8 (xnintr_host_tick+0x94) :| +func 2681660.065 __ipipe_walk_pipeline+0x8 (__ipipe_dispatch_wired_nocheck+0x148) :| +func 2681660.061 __ipipe_sync_stage+0x8 (__ipipe_walk_pipeline+0x1a8) <| #end 0x8000 2681660.042 __ipipe_sync_stage+0x238 (__ipipe_walk_pipeline+0x1a8) #func 2681660.042 irq_enter+0x8 (__ipipe_sync_stage+0x260) #func 2681660.046 rcu_irq_enter+0x8 (irq_enter+0x38) #func 2681670.043 idle_cpu+0x8 (irq_enter+0x44) #func 2681670.039 ipipe_check_context+0x8 (irq_enter+0x84) | #begin 0x8001 2681670.000 ipipe_check_context+0xf0 (irq_enter+0x84) $>echo function_graph > current_tracer [ 275.698664] [ cut here ] [ 275.755250] [ cut here ] [ 275.810547] Kernel stack overflow in process dc6bf020, r1=dc838250 [ 275.896276] Kernel panic - not syncing: kernel stack overflow [ 275.965200] Rebooting in 30 seconds.. 2011/9/23 Gilles Chanteperdrix > On 09/23/2011 09:58 AM, Jean-Michel Hautbois wrote: > > Hi List, > > > > I am trying to use ftrace on a P2020 board, using the following : > > - Linux 2.6.35-11 > > - Xenomai 2.5.5.1 (or 2.5.6, same problem) > > - ipipe 2.12-01 > > > > I launch it using : > > $> mkdir /tmp/debug > > $> mount -t debugfs nodev /tmp/debug > > $> cd /tmp/debug/tracing > > $> echo function_graph > current_tracer > > > > The last one fails (it hangs the CPU, I cannot do anything than hard > reset). > > Do you know what could cause that ? > > As I can't have any trace, I don't really know how to go further, and > before > > using a BDI I would like to know if it is a known issue, as I could not > find > > anything related to this in the recent changes... > > > > Thanks in advance for your help, > > Regards, > > JM > > Does the I-pipe tracer work? > See: > http://www.xenomai.org/index.php/I-pipe:Tracer > > > -- >Gilles. > ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core
Re: [Xenomai-core] P2020 support for ftrace with ipipe 2.12-01 and xeno 2.5.5.1
On 09/23/2011 09:58 AM, Jean-Michel Hautbois wrote: > Hi List, > > I am trying to use ftrace on a P2020 board, using the following : > - Linux 2.6.35-11 > - Xenomai 2.5.5.1 (or 2.5.6, same problem) > - ipipe 2.12-01 > > I launch it using : > $> mkdir /tmp/debug > $> mount -t debugfs nodev /tmp/debug > $> cd /tmp/debug/tracing > $> echo function_graph > current_tracer > > The last one fails (it hangs the CPU, I cannot do anything than hard reset). > Do you know what could cause that ? > As I can't have any trace, I don't really know how to go further, and before > using a BDI I would like to know if it is a known issue, as I could not find > anything related to this in the recent changes... > > Thanks in advance for your help, > Regards, > JM Does the I-pipe tracer work? See: http://www.xenomai.org/index.php/I-pipe:Tracer -- Gilles. ___ Xenomai-core mailing list Xenomai-core@gna.org https://mail.gna.org/listinfo/xenomai-core