Re: [perfmon2] segmentation fault while running pfmon
Status: Beta
Brought to you by:
seranian
Menu
▾
▴
Re: [perfmon2] segmentation fault while running pfmon
|
From: Qingyuan D. <dd...@gm...> - 2009-09-24 21:27:40
|
Thank you! I will try your configuration.
On Thu, Sep 24, 2009 at 5:26 PM, stephane eranian <er...@go...>wrote:
> On Thu, Sep 24, 2009 at 11:24 PM, Qingyuan Deng <dd...@gm...> wrote:
> > Ok, I will try to replace the compiler with a different version and to
> see
> > what will happen. And could I have your software stack configuration on a
> > working Nehalem system?
>
> Ubuntu-8.10 x86_64 mode, gcc-4.3.2
>
> > Anyway, I appreciate your help!
> > Qingyuan
> >
> > On Thu, Sep 24, 2009 at 5:17 PM, stephane eranian <
> er...@go...>
> > wrote:
> >>
> >> On Thu, Sep 24, 2009 at 11:14 PM, Qingyuan Deng <dd...@gm...>
> wrote:
> >> > No, it doesn't work either. Exactly the same problem. Do you think it
> is
> >> > due
> >> > to the compiler? I am using gcc (Ubuntu 4.3.3-5ubuntu4) 4.3.3
> >>
> >> Nobody else reported this problem so far. Yet, as I said, I think I
> >> have encoutered
> >> it a while back.
> >> You see it could be something like the compiler is putting data in
> >> read-only sections
> >> but the code is trying to modify it. Or there is something corrupted
> >> in the event table
> >> for Nehalem.
> >>
> >> Do you have another type of machine installed with the same distro?
> >>
> >> Unfortunately, it is hard for me to help you because I cannot
> >> reproduce the problem
> >> on my systems.
> >>
> >> > Thank you!
> >> >
> >> > On Thu, Sep 24, 2009 at 5:08 PM, Qingyuan Deng <dd...@gm...>
> wrote:
> >> >>
> >> >> I thought I compile it as the default configuration file you put
> there
> >> >> which is "O2"... OK, I am trying to re-compile it with O0.
> >> >>
> >> >>
> >> >> On Thu, Sep 24, 2009 at 4:54 PM, stephane eranian
> >> >> <er...@go...>
> >> >> wrote:
> >> >>>
> >> >>> What is you try to compile libpfm with OPTIM=-O0?
> >> >>>
> >> >>>
> >> >>> On Thu, Sep 24, 2009 at 10:22 PM, Qingyuan Deng <dd...@gm...>
> >> >>> wrote:
> >> >>> > Starting program:
> >> >>> > /home/qdeng/pmu/perfmon2/libpfm-3.9/examples_v2.x/self
> >> >>> > [Thread debugging using libthread_db enabled]
> >> >>> > sycall base 333
> >> >>> > major version 2
> >> >>> > minor version 82
> >> >>> > [New Thread 0xb7eb96c0 (LWP 21252)]
> >> >>> >
> >> >>> > Program received signal SIGSEGV, Segmentation fault.
> >> >>> > [Switching to Thread 0xb7eb96c0 (LWP 21252)]
> >> >>> > 0xb7f31613 in strlen () from /lib/tls/i686/cmov/libc.so.6
> >> >>> > (gdb) backtrace
> >> >>> > #0 0xb7f31613 in strlen () from /lib/tls/i686/cmov/libc.so.6
> >> >>> > #1 0x0804a0a1 in pfm_get_max_event_name_len (len=0xbfc9491c)
> >> >>> > at pfmlib_common.c:916
> >> >>> > #2 0x08048d9b in main (argc=1, argv=0xbfc949d4) at self.c:106
> >> >>> > (gdb)
> >> >>> >
> >> >>> >
> >> >>> > On Thu, Sep 24, 2009 at 4:18 PM, stephane eranian
> >> >>> > <er...@go...>
> >> >>> > wrote:
> >> >>> >>
> >> >>> >> What about you try libpfm/examples_v2.x/self.
> >> >>> >> Does it crash too?
> >> >>> >>
> >> >>> >>
> >> >>> >> On Thu, Sep 24, 2009 at 10:13 PM, Qingyuan Deng <
> dd...@gm...>
> >> >>> >> wrote:
> >> >>> >> > Hi Stephane,
> >> >>> >> > Do you mean mean the i value in "for(i =0; i
> >> >>> >> > <pfm_current->pme_count;
> >> >>> >> > i++) "
> >> >>> >> > in pfmlib_common.c?
> >> >>> >> >
> >> >>> >> > Then it is 0 while the program stops.
> >> >>> >> >
> >> >>> >> > Thanks!
> >> >>> >> >
> >> >>> >> >
> --------------------------------------------------------------------
> >> >>> >> > (gdb) b 1899
> >> >>> >> > Breakpoint 1 at 0x804ead1: file pfmon.c, line 1899.
> >> >>> >> > (gdb) r --help
> >> >>> >> > Starting program: /usr/local/bin/pfmon --help
> >> >>> >> > [Thread debugging using libthread_db enabled]
> >> >>> >> > [New Thread 0xb7de48d0 (LWP 20951)]
> >> >>> >> > [Switching to Thread 0xb7de48d0 (LWP 20951)]
> >> >>> >> >
> >> >>> >> > Breakpoint 1, main (argc=2, argv=0xbf841dc4) at pfmon.c:1900
> >> >>> >> > 1900 pfmon_initialize(argv);
> >> >>> >> > (gdb) s
> >> >>> >> > pfmon_initialize (argv=0xbf841dc4) at pfmon.c:1387
> >> >>> >> > 1387 pfmon_get_version();
> >> >>> >> > (gdb) n
> >> >>> >> > 1389 ret = pfmon_api_probe();
> >> >>> >> > (gdb) n
> >> >>> >> > 1390 if (ret) {
> >> >>> >> > (gdb) n
> >> >>> >> > 1402 if (options.pfm_version != PERFMON_VERSION_20)
> >> >>> >> > (gdb) n
> >> >>> >> > 1403 options.opt_has_sets = 1;
> >> >>> >> > (gdb) n
> >> >>> >> > 1408 pfmon_arch_initialize();
> >> >>> >> > (gdb) n
> >> >>> >> > 1410 setup_common_signals();
> >> >>> >> > (gdb) n
> >> >>> >> > 1413 if (pfm_initialize() != PFMLIB_SUCCESS)
> >> >>> >> > (gdb) n
> >> >>> >> > 1416 pfm_get_version(&version);
> >> >>> >> > (gdb) n
> >> >>> >> > 1417 if (PFM_VERSION_MAJOR(version) < 3 ||
> >> >>> >> > PFM_VERSION_MINOR(version)
> >> >>> >> > < 2)
> >> >>> >> > (gdb) n
> >> >>> >> > 1421 options.arg_mem_max =
> >> >>> >> > pfmon_get_perfmon_arg_mem_max();
> >> >>> >> > (gdb) n
> >> >>> >> > 1423 ret = pfm_get_pmu_type(&options.pmu_type);
> >> >>> >> > (gdb) n
> >> >>> >> > 1424 if (ret != PFMLIB_SUCCESS) {
> >> >>> >> > (gdb) n
> >> >>> >> > 1428 ret = pfm_get_num_counters(&options.max_counters);
> >> >>> >> > (gdb) n
> >> >>> >> > 1429 if (ret != PFMLIB_SUCCESS) {
> >> >>> >> > (gdb) n
> >> >>> >> > 1433 if (check_forced_generic(argv[0])) {
> >> >>> >> > (gdb) n
> >> >>> >> > 1442 pfmon_detect();
> >> >>> >> > (gdb) n
> >> >>> >> > 1444 options.session_timeout = PFMON_NO_TIMEOUT;
> >> >>> >> > (gdb) n
> >> >>> >> > 1445 options.interval = PFMON_NO_TIMEOUT;
> >> >>> >> > (gdb) n
> >> >>> >> > 1450 long_val = sysconf(_SC_NPROCESSORS_ONLN);
> >> >>> >> > (gdb) n
> >> >>> >> > 1451 if (long_val == -1)
> >> >>> >> > (gdb) n
> >> >>> >> > 1454 options.online_cpus = long_val;
> >> >>> >> > (gdb) n
> >> >>> >> > 1456 long_val = sysconf(_SC_NPROCESSORS_CONF);
> >> >>> >> > (gdb) n
> >> >>> >> > 1457 if (long_val == -1)
> >> >>> >> > (gdb) n
> >> >>> >> > 1460 options.config_cpus = long_val;
> >> >>> >> > (gdb) n
> >> >>> >> > 1462 pfmon_get_phys_cpu_mask();
> >> >>> >> > (gdb) n
> >> >>> >> > 1464 clock_getres(CLOCK_MONOTONIC, &ts);
> >> >>> >> > (gdb) n
> >> >>> >> > 1465 options.clock_res = ts.tv_sec * 1000000000 +
> >> >>> >> > ts.tv_nsec;
> >> >>> >> > (gdb) n
> >> >>> >> > 1466 options.page_size = getpagesize();
> >> >>> >> > (gdb) n
> >> >>> >> > 1467 options.cpu_mhz = pfmon_find_cpu_speed();
> >> >>> >> > (gdb) n
> >> >>> >> > 1469 pfm_get_max_event_name_len(&len);
> >> >>> >> > (gdb) s
> >> >>> >> > pfm_get_max_event_name_len (len=0xbf841cc0) at
> >> >>> >> > pfmlib_common.c:907
> >> >>> >> > 907 if (PFMLIB_INITIALIZED() == 0)
> >> >>> >> > (gdb) n
> >> >>> >> > 909 if (len == NULL)
> >> >>> >> > (gdb) n
> >> >>> >> > 912 for(i=0; i < pfm_current->pme_count; i++) {
> >> >>> >> > (gdb) n
> >> >>> >> > 913 str = pfm_current->get_event_name(i);
> >> >>> >> > (gdb) p i
> >> >>> >> > $1 = 0
> >> >>> >> > (gdb) n
> >> >>> >> > 914 if (!str)
> >> >>> >> > (gdb) n
> >> >>> >> > 916 l = strlen(str);
> >> >>> >> > (gdb) n
> >> >>> >> >
> >> >>> >> > Program received signal SIGSEGV, Segmentation fault.
> >> >>> >> > 0xb7e5c613 in strlen () from /lib/tls/i686/cmov/libc.so.6
> >> >>> >> > (gdb)
> >> >>> >> >
> >> >>> >> >
> >> >>> >> > On Thu, Sep 24, 2009 at 4:20 AM, stephane eranian
> >> >>> >> > <er...@go...>
> >> >>> >> > wrote:
> >> >>> >> >>
> >> >>> >> >> On Thu, Sep 24, 2009 at 9:29 AM, Qingyuan Deng
> >> >>> >> >> <dd...@gm...>
> >> >>> >> >> wrote:
> >> >>> >> >> > I tracked into the program and found the segmentation fault
> >> >>> >> >> > happened
> >> >>> >> >> > at:
> >> >>> >> >> >
> >> >>> >> >> > in pfmon.c
> >> >>> >> >> > pfmon_initialize()->pfm_get_max_event_name_len(&len)
> >> >>> >> >> >
> >> >>> >> >> > then it jumped into pfmlib_common.c
> >> >>> >> >> >
> >> >>> >> >> > pfm_get_max_event_name_len (len=0xbffd9460) at
> >> >>> >> >> > pfmlib_common.c:907
> >> >>> >> >> > 907 if (PFMLIB_INITIALIZED() == 0)
> >> >>> >> >> > (gdb) n
> >> >>> >> >> > 909 if (len == NULL)
> >> >>> >> >> > (gdb) n
> >> >>> >> >> > 912 for(i=0; i < pfm_current->pme_count; i++) {
> >> >>> >> >> > (gdb) n
> >> >>> >> >> > 913 str = pfm_current->get_event_name(i);
> >> >>> >> >> > (gdb) n
> >> >>> >> >> > 914 if (!str)
> >> >>> >> >> > (gdb) n
> >> >>> >> >> > 916 l = strlen(str);
> >> >>> >> >> > (gdb) n
> >> >>> >> >> >
> >> >>> >> >> > Program received signal SIGSEGV, Segmentation fault.
> >> >>> >> >> > 0xb7ef5613 in strlen () from /lib/tls/i686/cmov/libc.so.6
> >> >>> >> >> >
> >> >>> >> >> > Here is the backtrace:
> >> >>> >> >> >
> >> >>> >> >> > Program received signal SIGSEGV, Segmentation fault.
> >> >>> >> >> > [Switching to Thread 0xb7e7a8d0 (LWP 5257)]
> >> >>> >> >> > 0xb7ef2613 in strlen () from /lib/tls/i686/cmov/libc.so.6
> >> >>> >> >> > (gdb) backtrace
> >> >>> >> >> > #0 0xb7ef2613 in strlen () from
> /lib/tls/i686/cmov/libc.so.6
> >> >>> >> >> > #1 0xb7fed411 in pfm_get_max_event_name_len
> (len=0xbf8d5d60)
> >> >>> >> >> > at pfmlib_common.c:916
> >> >>> >> >> > #2 0x0804dd45 in pfmon_initialize (argv=0xbf8d5e64) at
> >> >>> >> >> > pfmon.c:1469
> >> >>> >> >> > #3 0x0804eadf in main (argc=1, argv=0xbf8d5e64) at
> >> >>> >> >> > pfmon.c:1900
> >> >>> >> >> > (gdb)
> >> >>> >> >>
> >> >>> >> >>
> >> >>> >> >> I have seen this problem on an other machine as well. But I
> >> >>> >> >> don't
> >> >>> >> >> remember
> >> >>> >> >> what
> >> >>> >> >> the cause was. It has to do with the event table.
> >> >>> >> >> Could you print the value of i (the event index) so we know
> >> >>> >> >> where
> >> >>> >> >> it
> >> >>> >> >> stops?
> >> >>> >> >>
> >> >>> >> >> >
> >> >>> >> >> > On Thu, Sep 24, 2009 at 3:25 AM, stephane eranian
> >> >>> >> >> > <er...@go...>
> >> >>> >> >> > wrote:
> >> >>> >> >> >>
> >> >>> >> >> >> What does ldd pfmon say?
> >> >>> >> >> >>
> >> >>> >> >> >>
> >> >>> >> >> >> On Thu, Sep 24, 2009 at 9:01 AM, Qingyuan Deng
> >> >>> >> >> >> <dd...@gm...>
> >> >>> >> >> >> wrote:
> >> >>> >> >> >> > Hello,
> >> >>> >> >> >> >
> >> >>> >> >> >> > I patched the kernel and installed both libpfm and
> perfmon,
> >> >>> >> >> >> > however I
> >> >>> >> >> >> > got
> >> >>> >> >> >> > segmentation fault while I tried running command: "pfmon
> >> >>> >> >> >> > --help'.
> >> >>> >> >> >> > Could
> >> >>> >> >> >> > anyone help me to figure it out please?
> >> >>> >> >> >> >
> >> >>> >> >> >> > I am running Ubuntu 9.04 on kernel 2.6.29. I downloaded
> the
> >> >>> >> >> >> > latest
> >> >>> >> >> >> > version
> >> >>> >> >> >> > of libpfm, perfmon, and the base system patch from the
> >> >>> >> >> >> > sourceforge
> >> >>> >> >> >> > page.
> >> >>> >> >> >> > My
> >> >>> >> >> >> > processor is Intel core i7 920 and I adjusted related
> >> >>> >> >> >> > configuration
> >> >>> >> >> >> > when
> >> >>> >> >> >> > I
> >> >>> >> >> >> > configured the kernel. Following is the debug message
> from
> >> >>> >> >> >> > gdb:
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> >> >>> >> >> >> > qdeng@qdeng-desktop:~/pmu/perfmon2/pfmon-3.9$ sudo make
> >> >>> >> >> >> > install
> >> >>> >> >> >> > [sudo] password for qdeng:
> >> >>> >> >> >> > make[1]: Entering directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/pfmon'
> >> >>> >> >> >> > make[2]: Entering directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/pfmon/smpl_mod'
> >> >>> >> >> >> > make[2]: Nothing to be done for `all'.
> >> >>> >> >> >> > make[2]: Leaving directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/pfmon/smpl_mod'
> >> >>> >> >> >> > cc -o pfmon pfmon.o pfmon_smpl.o pfmon_util.o
> >> >>> >> >> >> > pfmon_system.o
> >> >>> >> >> >> > pfmon_task.o
> >> >>> >> >> >> > pfmon_symbols.o pfmon_results.o pfmon_hash.o
> >> >>> >> >> >> > pfmon_smpl_dfl.o
> >> >>> >> >> >> > pfmon_os.o
> >> >>> >> >> >> > pfmon_os_v2x.o pfmon_util_x86.o pfmon_i386_p6.o
> >> >>> >> >> >> > pfmon_gen_ia32.o
> >> >>> >> >> >> > pfmon_amd64.o pfmon_core.o pfmon_intel_atom.o
> >> >>> >> >> >> > pfmon_intel_nhm.o
> >> >>> >> >> >> > pfmon_pentium4.o -g -ggdb -Wall -Werror -D_REENTRANT
> >> >>> >> >> >> > -I/usr/local/include
> >> >>> >> >> >> > -DCONFIG_PFMON_I386 -DPFMON_DEBUG
> >> >>> >> >> >> > -DDATADIR=\"/usr/local/share/pfmon\"
> >> >>> >> >> >> > -I.
> >> >>> >> >> >> > -I/usr/include/libelf -D_GNU_SOURCE -DPFMON_DEBUG -g
> >> >>> >> >> >> > smpl_mod/libsmplfmt.a
> >> >>> >> >> >> > -lpthread -lm -lrt -Wl,-Bstatic -lelf -Wl,-Bdynamic
> >> >>> >> >> >> > -L/usr/local/lib
> >> >>> >> >> >> > -lpfm
> >> >>> >> >> >> > mkdir -p /usr/local/bin
> >> >>> >> >> >> > install -m 755 pfmon /usr/local/bin/pfmon
> >> >>> >> >> >> > ln -sf /usr/local/bin/pfmon /usr/local/bin/pfmon_gen
> >> >>> >> >> >> > make[1]: Leaving directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/pfmon'
> >> >>> >> >> >> > make[1]: Entering directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/tests'
> >> >>> >> >> >> > mkdir -p /usr/local/bin
> >> >>> >> >> >> > install -m 755 pfdbg /usr/local/bin/pfdbg
> >> >>> >> >> >> > make[1]: Leaving directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/tests'
> >> >>> >> >> >> > make[1]: Entering directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/docs'
> >> >>> >> >> >> > mkdir -p /usr/local/share/man/man1
> >> >>> >> >> >> > install -m 644 pfmon.1 /usr/local/share/man/man1
> >> >>> >> >> >> > make[1]: Leaving directory
> >> >>> >> >> >> > `/home/qdeng/pmu/perfmon2/pfmon-3.9/docs'
> >> >>> >> >> >> > qdeng@qdeng-desktop:~/pmu/perfmon2/pfmon-3.9$ pfmon
> --help
> >> >>> >> >> >> > Segmentation fault
> >> >>> >> >> >> > qdeng@qdeng-desktop:~/pmu/perfmon2/pfmon-3.9$ gdb pfmon
> >> >>> >> >> >> > GNU gdb 6.8-debian
> >> >>> >> >> >> > Copyright (C) 2008 Free Software Foundation, Inc.
> >> >>> >> >> >> > License GPLv3+: GNU GPL version 3 or later
> >> >>> >> >> >> > <http://gnu.org/licenses/gpl.html>
> >> >>> >> >> >> > This is free software: you are free to change and
> >> >>> >> >> >> > redistribute
> >> >>> >> >> >> > it.
> >> >>> >> >> >> > There is NO WARRANTY, to the extent permitted by law.
> Type
> >> >>> >> >> >> > "show
> >> >>> >> >> >> > copying"
> >> >>> >> >> >> > and "show warranty" for details.
> >> >>> >> >> >> > This GDB was configured as "i486-linux-gnu"...
> >> >>> >> >> >> > (gdb) run --help
> >> >>> >> >> >> > Starting program: /usr/local/bin/pfmon --help
> >> >>> >> >> >> > [Thread debugging using libthread_db enabled]
> >> >>> >> >> >> > [New Thread 0xb7e598d0 (LWP 4437)]
> >> >>> >> >> >> >
> >> >>> >> >> >> > Program received signal SIGSEGV, Segmentation fault.
> >> >>> >> >> >> > [Switching to Thread 0xb7e598d0 (LWP 4437)]
> >> >>> >> >> >> > 0xb7ed1613 in strlen () from /lib/tls/i686/cmov/libc.so.6
> >> >>> >> >> >> > (gdb)
> >> >>> >> >> >> >
> >> >>> >> >> >> > Thanks!
> >> >>> >> >> >> > Qingyuan
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> ------------------------------------------------------------------------------
> >> >>> >> >> >> > Come build with us! The BlackBerry® Developer
> >> >>> >> >> >> > Conference
> >> >>> >> >> >> > in
> >> >>> >> >> >> > SF,
> >> >>> >> >> >> > CA
> >> >>> >> >> >> > is the only developer event you need to attend this year.
> >> >>> >> >> >> > Jumpstart
> >> >>> >> >> >> > your
> >> >>> >> >> >> > developing skills, take BlackBerry mobile applications to
> >> >>> >> >> >> > market
> >> >>> >> >> >> > and
> >> >>> >> >> >> > stay
> >> >>> >> >> >> > ahead of the curve. Join us from November 9-12, 2009.
> >> >>> >> >> >> > Register
> >> >>> >> >> >> > now!
> >> >>> >> >> >> > http://p.sf.net/sfu/devconf
> >> >>> >> >> >> > _______________________________________________
> >> >>> >> >> >> > perfmon2-devel mailing list
> >> >>> >> >> >> > per...@li...
> >> >>> >> >> >> >
> https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
> >> >>> >> >> >> >
> >> >>> >> >> >> >
> >> >>> >> >> >
> >> >>> >> >> >
> >> >>> >> >
> >> >>> >> >
> >> >>> >
> >> >>> >
> >> >>
> >> >
> >> >
> >
> >
>
|