linux-kernel - Re: perf script : wrong symoff in callchain

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 3 Oct 2017 13:45:41 +0200
From:   Matthieu CASTET <matthieu.castet@...rot.com>
To:     Adrian Hunter <adrian.hunter@...el.com>
Cc:     <linux-kernel@...r.kernel.org>,
        Arnaldo Carvalho de Melo <acme@...hat.com>,
        Chris Phlipot <cphlipot0@...il.com>
Subject: Re: perf script : wrong symoff in callchain

Le Tue, 3 Oct 2017 13:34:37 +0300,
Adrian Hunter <adrian.hunter@...el.com> a écrit :

> On 03/10/17 13:19, Matthieu CASTET wrote:
> > Hi,
> > 
> > while using perf on x86_64, I saw strange output for symoff.
> > 
> > $ perf record -g -- sleep 1
> > $ perf script -F comm,tid,pid,time,ip,sym,dso,symoff
> > 
> > [...]
> > sleep 11656/11656 1045318.546436: 
> >             7fff9542e5b5 __d_lookup_rcu+0x80006ae02035 ([kernel.kallsyms])
> >             7fff9541e132 lookup_fast+0x80006ae02052 ([kernel.kallsyms])
> >             7fff9541eae8 walk_component+0x80006ae02048 ([kernel.kallsyms])
> >             7fff9541efa2 link_path_walk+0x80006ae021b2 ([kernel.kallsyms])
> >             7fff9541c92d path_init+0x80006ae021bd ([kernel.kallsyms])
> >             7fff9542100b path_openat+0x80006ae020fb ([kernel.kallsyms])
> >             7fff953c14fe handle_mm_fault+0x80006ae020ee ([kernel.kallsyms])
> >             7fff95423669 do_filp_open+0x80006ae02099 ([kernel.kallsyms])
> >             7fff95433414 __alloc_fd+0x80006ae02044 ([kernel.kallsyms])
> >             7fff9540ff6e do_sys_open+0x80006ae0212e ([kernel.kallsyms])
> >             7fff9540ff6e do_sys_open+0x80006ae0212e ([kernel.kallsyms])
> >             7fff95850abb system_call_fast_compare_end+0x80006ae0200c ([kernel.kallsyms])
> >                    2aecf _nl_load_locale_from_archive+0xffff014b5a92841f (/lib/x86_64-linux-gnu/libc-2.24.so)
> > 
> > 
> > I tried to revert a4eb24a49566db77ee999b46603f602a0302f481 and I got
> > good result :
> > [...]
> > sleep 11656/11656 1045318.546436: 
> >             7fff9542e5b5 __d_lookup_rcu+0x35 ([kernel.kallsyms])
> >             7fff9541e132 lookup_fast+0x52 ([kernel.kallsyms])
> >             7fff9541eae8 walk_component+0x48 ([kernel.kallsyms])
> >             7fff9541efa2 link_path_walk+0x1b2 ([kernel.kallsyms])
> >             7fff9541c92d path_init+0x1bd ([kernel.kallsyms])
> >             7fff9542100b path_openat+0xfb ([kernel.kallsyms])
> >             7fff953c14fe handle_mm_fault+0xee ([kernel.kallsyms])
> >             7fff95423669 do_filp_open+0x99 ([kernel.kallsyms])
> >             7fff95433414 __alloc_fd+0x44 ([kernel.kallsyms])
> >             7fff9540ff6e do_sys_open+0x12e ([kernel.kallsyms])
> >             7fff9540ff6e do_sys_open+0x12e ([kernel.kallsyms])
> >             7fff95850abb system_call_fast_compare_end+0xc ([kernel.kallsyms])
> >                    2aecf _nl_load_locale_from_archive+0x41f (/lib/x86_64-linux-gnu/libc-2.24.so)
> > 
> > 
> > Any idea ?  
> 
> It could be the problem with add_callchain_ip() described here:
> 
> 	https://www.spinics.net/lists/linux-perf-users/msg03172.html

The db-export code removed the remapping [1]

So there a mix of mapping in callchain code :
- add_callchain_ip() store al.addr (ip remapped)
- fill_callchain_info do the translation (map_ip)
- sample__fprintf_callchain do the translation
- call_path_from_sample (db-export) don't do anymore the translation


So what should we do : 
- assume we store al.addr and remove mapping from all consumers ?
- store ip and do the mapping in consumers ?



[1]
commit 7a2544c004a6c576b1e307f30925b165affe6a22
Author: Chris Phlipot <cphlipot0@...il.com>
Date:   Tue May 10 20:26:48 2016 -0700

    perf script: Fix callchain addresses in db-export
    
    Remove the call to map_ip() to adjust al.addr, because it has already
    been called when assembling the callchain, in:
    
      thread__resolve_callchain_sample(perf_sample)
          add_callchain_ip(ip = perf_sample->callchain->ips[j])
              thread__find_addr_location(addr = ip)
                  thread__find_addr_map(addr) {
                      al->addr = addr
                      if (al->map)
                          al->addr = al->map->map_ip(al->map, al->addr);
                  }
    
    Calling it a second time can result in incorrect addresses being used.
    This can have effects such as duplicate symbols being created and
    exported.