[lwip-devel] [bugs #9076] Trimming of leading edge of received packets

[Kieran Mansley]

This mail is an automated notification from the bugs tracker
 of the project: lwIP - A Lightweight TCP/IP stack.




/**************************************************************************/
[bugs #9076] Full Item Snapshot:

URL: <http://savannah.nongnu.org/bugs/?func=detailitem&item_id=9076>
Project: lwIP - A Lightweight TCP/IP stack
Submitted by: Kieran Mansley
On: Mon 05/24/04 at 14:19

Category:  TCP
Severity:  3 - Ordinary
Item Group:  Faulty Behaviour
Resolution:  None
Assigned to:  kieranm
Status:  Open


Summary:  Trimming of leading edge of received packets

Original Submission:  
Bug in:
  Release:   Stable 0.7.1
  Module:    tcp_in.c
  Function   tcp_receive().

Hello and thanks to all contributors to this project!

The pbuf "tot_len" value is not adjusted correctly when "trimming
the first edge" of the received segment.

Here is the original code:

  off = pcb->rcv_nxt - seqno;
  if (inseg.p->len < off) {
    p = inseg.p;
    while (p->len < off) {
      off -= p->len;
      inseg.p->tot_len -= p->len;  /* <-----this doesn't work... */
      p->len = 0;
      p = p->next;
    }
    pbuf_header(p, -off);
  } else {

\\\\\\\\
The original code only works if the "trimming" is contained to the
first pbuf.  If the "trim" extends into the second pbuf two problems
result:

1.  Only the first pbuf has it's tot_len (partially) modified, the
other pbufs are left un-modified (this may or may not cause
problems later on...).

2.  The main issue is that the code doesn't take into account the
call to pbuf_header(p, -off).  The code drops out of the loop
with "inseg.p->tot_len" too big (by the amount remaining in "off").
This is handled fine by the call to pbuf_header() - if it's the
first pbuf in the chain.  Otherwise, the first several totals
are wrong...

In my case the call to lwip_recv() (which called netbuf_copy_partial())
returned without copying the requested number of bytes.  (The return
byte count however indicated all the requested bytes had been delivered.)

\\\\\\\\
So I believe this is what was intended:

  off = pcb->rcv_nxt - seqno;
  if (inseg.p->len < off) {
    u16_t new_tot;                    /* temp for new tot_len  */
    p = inseg.p;
    new_tot = inseg.p->tot_len - off; /* calc new tot_len */
    while (p->len < off) {
      off -= p->len;
      p->tot_len = new_tot;             /* each with len=0 has new_tot! */
      p->len = 0;
      p = p->next;
    }
    pbuf_header(p, -off);             /* now this works as intended */
  } else {

\\\\\\
Here is a printout of the pbuf chain before and after the "trim left"
adjustment (from the fixed code):

tcp_receive: before adjust left: pbuf=1541C, tot_len=512, len=74
tcp_receive: before adjust left: pbuf=154AA, tot_len=438, len=128
tcp_receive: before adjust left: pbuf=15538, tot_len=310, len=128
tcp_receive: before adjust left: pbuf=155C6, tot_len=182, len=128
tcp_receive: before adjust left: pbuf=15654, tot_len=54, len=54

tcp_receive: after adjust left: pbuf=1541C, tot_len=260, len=0
tcp_receive: after adjust left: pbuf=154AA, tot_len=260, len=0
tcp_receive: after adjust left: pbuf=15538, tot_len=260, len=78
tcp_receive: after adjust left: pbuf=155C6, tot_len=182, len=128
tcp_receive: after adjust left: pbuf=15654, tot_len=54, len=54

If there are any questions - feel free to email me...

Thanks again for all who have contributed.

Ed Kerekes











For detailed info, follow this link:
<http://savannah.nongnu.org/bugs/?func=detailitem&item_id=9076>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.nongnu.org/