Messages in this thread | | | Date | Wed, 28 May 2014 18:44:06 -0300 | From | Arnaldo Carvalho de Melo <> | Subject | [BUG] perf probe segfaulting when asked for variable it doesn't find |
| |
Masami-san,
While trying:
[root@zoo ~]# perf probe 'vfs_getname=getname_flags:65 pathname=result->name:string' Failed to find the location of result at this address. Perhaps, it has been optimized out. Failed to find 'result' in this function. Error: Failed to add events. (-2) [root@zoo ~]# perf probe 'vfs_getname=getname_flags:65 pathname=result->name:string' Added new event: Segmentation fault (core dumped)
I got segfaulted while in the past I would get the much nicer:
[root@zoo ~]# perf probe 'vfs_getname=getname_flags:65 pathname=result->name:string' Failed to find the location of result at this address. Perhaps, it has been optimized out. Failed to find 'result' in this function. Error: Failed to add events. (-2) [root@zoo ~]#
The "Error:" part can and should be trimmed, just the two middle ones should be enough, but I'm digressing, I bisected it down to the cset at the bottom of this message.
In the past there was a 'result' variable at getname_flags, but now 'probe' isn't finding any, as it seems to have been optimized away, still haven't checked thoroughly:
[root@zoo ~]# perf probe -V getname_flags Available variables at getname_flags @<getname_flags+0> char* filename int flags int* empty @<getname+18> (No matched variables) @<user_path_create+37> (No matched variables) @<user_path_parent+37> (No matched variables) @<user_path_at_empty+50> (No matched variables) @<user_path_mountpoint_at+37> (No matched variables) @<sys_symlinkat+38> (No matched variables) @<sys_symlink+34> (No matched variables) [root@zoo ~]
3d918a12a1b3088ac16ff37fa52760639d6e2403 is the first bad commit commit 3d918a12a1b3088ac16ff37fa52760639d6e2403 Author: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Date: Fri Oct 11 16:10:26 2013 +0900
perf probe: Find fentry mcount fuzzed parameter location
At this point, --fentry (mcount function entry) option for gcc fuzzes the debuginfo variable locations by skipping the mcount instruction offset (on x86, this is a 5 byte call instruction).
This makes variable searching fail at the entry of functions which are mcount'ed.
e.g.) Available variables at vfs_read @<vfs_read+0> (No matched variables)
This patch adds additional location search at the function entry point to solve this issue, which tries to find the earliest address for the variable location.
Note that this only works with function parameters (formal parameters) because any local variables should not exist on the function entry address (those are not initialized yet).
With this patch, perf probe shows correct parameters if possible; # perf probe --vars vfs_read Available variables at vfs_read @<vfs_read+0> char* buf loff_t* pos size_t count struct file* file
Signed-off-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Ingo Molnar <mingo@redhat.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/20131011071025.15557.13275.stgit@udc4-manage.rcp.hitachi.co.jp Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
:040000 040000 4c416906285aa1488ed2badbaf3b3feee86f9578 978c6b47f442845e6e93a79a24aeb36bbca0b7da M tools [acme@zoo linux]$
| |