Re: question about prep_ jitdump in opcontrol

[Myke S. <Myk...@pa...>]

Daniel,

Thanks now I understand that this is to cover the re-boot case, so my  
work-around for the lack of an -lname option in BusyBox has that  
possible hole. I will document that in my local patch.

thanks

-myke

On Jan 22, 2010, at 3:02 AM, Daniel Hansel wrote:

> Hi Myke,
>
> the option "-lname" checks the symbolic link content for a special  
> pattern (i.e. the linked file name).
> This is necessary to check if the dump file is in use by the process  
> with the same PID.
>
> For example, if you use Oprofile to profile a Java application  
> Oprofile would create a dump file using the PID of the Java app.  
> process (e.g. pid=1000).
> If you reboot your machine without resetting the Oprofile data the  
> dump files would not be deleted.
> After that reboot the PID counter is restarted.
> Now you could get a new process with the PID 1000 used by the last  
> Oprofile run. But this new process is not using the dump file.
> In that case the check should delete the dump file due to the fact  
> that it is not used by the process with the PID 1000.
>
> The find option "-lname" is used for that.
>
> Kind regards,
> Daniel
>
> p.s.: Sorry for my late answer. I was very busy during this week.
>
> Myke Smith wrote:
>>
>> Daniel,
>>
>> Any comments on this change? Am I missing something fundamental in  
>> why
>> you need to check for the file by -lname?
>>
>> To get around -lname in BusyBox I was going to simply do:
>>
>>    # VMs may already be running when profiling is started, so
>>    # remove only dump files that are not in use
>>    for I in $dumpdir/*; do
>>        test -f $I || continue;
>>        local pid=`basename $I .dump`;
>>        if  ! test -d /proc/$pid; then
>>            rm -f $I;
>>        fi
>>
>>
>> thank you,
>>
>> -myke
>>
>> On Jan 7, 2010, at 5:12 AM, Maynard Johnson wrote:
>>
>>> Myke Smith wrote:
>>>> Hello,
>>>>
>>>> I have encountered a problem in prep_jitdump which is due to the
>>>> limited implementation of "find" in busybox.
>>>>
>>>>>    # VMs may already be running when profiling is started, so
>>>>>    # remove only dump files that are not in use
>>>>>    for I in $dumpdir/*; do
>>>>>        test -f $I || continue;
>>>>>        local pid=`basename $I .dump`;
>>>>>        if test -d /proc/$pid; then
>>>>>            local files=`find /proc/$pid/fd -lname $I`;
>>>>>            test -n "$files" && continue;
>>>>>        fi
>>>>>        rm -f $I;
>>>>>
>>>>
>>>> In busybox, the option "-lname" is not supported. My question,  
>>>> after
>>>> looking at the code, is why is it necessary to do all of this?  
>>>> Isn't
>>>> it sufficient to see that the pid is still defined in /proc and  
>>>> if not
>>>> remove the file from $dumpdir?
>>> Daniel, you are probably the best person to answer this question  
>>> since
>>> you committed this change.
>>>
>>> -Maynard
>>>>
>>>> thanks,
>>>>
>>>> -myke
>>>>
>>>> ------------------------------------------------------------------------------
>>>>
>>>> This SF.Net email is sponsored by the Verizon Developer Community
>>>> Take advantage of Verizon's best-in-class app development support
>>>> A streamlined, 14 day to market process makes app distribution fast
>>>> and easy
>>>> Join now and get one step closer to millions of Verizon customers
>>>> http://p.sf.net/sfu/verizon-dev2dev
>>>> _______________________________________________
>>>> oprofile-list mailing list
>>>> opr...@li...
>>>> https://lists.sourceforge.net/lists/listinfo/oprofile-list
>>>
>>
>