Re: Wrong Perf Backtraces

ahmadkhorrami <ahmadkhorrami@xxxxxxxx> · Fri, 27 Mar 2020 15:34:20 +0430

I do the following:
If this line is in the perf script backtrace:

            7f21ffe256db g_main_context_iteration+0x2b 

(/usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.5600.4)

I run the following command:

gdb -batch -ex 'file /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.5600.4' 

-ex 'disass g_main_context_iteration'.

Regards.

On 2020-03-27 15:29, ahmadkhorrami wrote:

Hi,

Thanks! Could you tell me what should be changed in order to make the 

code runnable on your system, if it is possible?

Regards.

On 2020-03-27 13:50, Jiri Olsa wrote:

On Thu, Mar 26, 2020 at 10:49:12PM +0430, ahmadkhorrami wrote:

Hi,
Here is the link for the python script:
https://gofile.io/?c=1ZSLwe
It is written in python-3 and takes the perf script output as input.

It looks for consecutive repeated backtrace lines and checks if the 

function

in these lines calls itself at the offset in the line (i.e., checks if

recursion is possible). If not possible it reports an error. Could you 

check

to see if any error is detected in your outputs, please?
I'm getting tons of following output:

/usr/lib/x86_64-linux-gnu/libgtk-3.so.0.2200.30: No such file or 

directory.

No symbol table is loaded.  Use the "file" command.

7ffff71b9bc1 gtk_css_node_invalidate_timestamp+0x31 

(/usr/lib/x86_64-linux-gnu/libgtk-3.so.0.2200.30)

I assume it's because I have all the dso binaries stored
under .biuldid path, while you check the output name

jirka

Regards.
On 2020-03-26 20:09, Jiri Olsa wrote:

On Thu, Mar 26, 2020 at 05:50:27PM +0430, ahmadkhorrami wrote:

Hi,
First of all, many thanks for your time. Did you say that the first
file has
problems?

The first file (http://gofile.io/?c=qk6oXv) has repeated gmallocn()s
while
the second (https://gofile.io/?c=oGxgSM) also has problems with
unmatched
(not necessarily repeated) function calls. I am not sure if the
kernel for
the second one is 5.4.7 or the generic Ubuntu kernel. But the first
one is
certainly 5.4.7. Just to be clear, there were many instances of these
unmatched <caller, callees>.
I can se all the files, but I just can't see the issue yet
but it's probably because of issues with perf archive..

let's see if somebody else can chime in

I have a simple python script that checks for this situation. It
disassembles functions using GDB and checks the (directly called)
target of
each caller. I will put some comments in the script and upload it.
Could you
check to see if the python script detects any mismatches in your
backtraces?
It takes the perf script output file as input. I will upload the
script in
an hour.
ok

jirka