[Click] strange bug

[Eddie Kohler]

Ah, yes.  You should *never* refer to a packet after calling output.push() on it.

E


Amita Ekbote wrote:
> Hey,
> 
> The bug is kinda solved. I should have posted it , but here are the 
> details: 
> 
> I was modifying the packet in the run_timer function of timed source, 
> the code was something like this
> {
> output.push(p)
> p->data_modify //add sequence number 
> 
> }
> 
> i got the error because the push occurred before the modification, looks 
> like the packet was accessible for some time and then got deleted. I am 
> not completely familiar with how the push function works and am not sure 
> when the packet gets deleted. Looked like a race condition. 
> 
> Thanks ! 
> 
> On Tue, Feb 16, 2010 at 5:54 PM, Eddie Kohler <kohler at cs.ucla.edu 
> <mailto:kohler at cs.ucla.edu>> wrote:
> 
>     Hi Amrita,
> 
>     Sorry, it's been a while.  That is quite a weird tailroom() number!
>      It looks very much like memory corruption.  I have some questions.
> 
>     - Are you running on a 32- or 64-bit architecture?
>     - Would you feel comfortable sharing the code of your modified
>     timedsource?
> 
>     E
> 
> 
> 
>     Amita Ekbote wrote:
> 
>         Hello,
> 
>         I am trying to add sequence numbers in the data portion of a
>         packet, so i
>         edited the timedsource.cc to just push the sequence number in
>         the packet.
>         The code converts the number to string, "pushes" the strings
>         length on the
>         packet and copies it. I am getting the following error
> 
>         click: ../lib/packet.cc:416: WritablePacket*
>         Packet::expensive_uniqueify(int32_t, int32_t, bool): Assertion
>         `(extra_headroom >= (int32_t)(-headroom())) && (extra_tailroom >=
>         (int32_t)(-tailroom()))' failed.
> 
>         Initially the packet has no headroom or tailroom. The headroom() and
>         tailroom() return 0. The sequence of function calls is
>         push->expensive_push->expensive_uniqueify. Right before
>         expensive_uniqueify
>         is called the headroom and tailroom are still 0. I printed
>         values before the
>         assertion and I get these:
> 
>         extra headroom 128 headroom 0 extra_tailroom 0 tailroom -157483016
>         end_buffer 137 end_data 157483153
> 
>         There is nothing else that is modifying the packet before the
>         function call
>         or during the function call. I am clueless as to why the
>         end_data and
>         end_buffer values are so weird. Any suggestions as to why this
>         would be
>         happening would be really helpful.
> 
>         Thanks!
> 
> 
> 
> 
> -- 
> Amita Ekbote