On Friday 13 July 2007 12:35:07 pm David Gay wrote:
> On 7/3/07, Steve McKown <[EMAIL PROTECTED]> wrote:
> > In TinyOS 2.0.x, the C SDK's serialsource.c contains a logic error in
> > write_serial_packet() that can cause a deadlock. This situation occurs
> > when write_serial_packet() is called by the host and the attached mote
> > fails to see the ending SYNC delimiter sent (due to a communication
> > error). The host waits in read_and_process() for an ack from the mote,
> > and the mote waits in RXSTATE_INFO for more data from the host. The
> > deadlock is broken if the mote forwards a radio packet to the host.
>
> I've checked in a fix for this - I believe the problem is that
> write_serial_packet should always be checking for acks in non-blocking
> mode even when the serial_source isn't non-blocking (as far as I can
> tell, the old code would never deadlock with a non-blocking serial
> source). So the fix is slightly different from the one below.
>
> Let me know if the modified version (in CVS on the tinyos 2.x tree)
> still doesn't work...
Some more information. With DEBUG enabled, the latest code never shows
receipt of ack frames:
----------------------
resp [1250]: frame=93 guid=9100000027c1aa24 frame=93 sample=0 metadata=invalid
time=Fri Jul 13 15:46:21 2007
writing 44 26 00 ff ff 00 00 1b 00 96 00 00 00 5d 24 aa c1 27 00 00 00 91 00
00 00 5d 00 00 00 00 04 e2 46 97 f2 ad 00
encoded 7e 44 26 00 ff ff 00 00 1b 00 96 00 00 00 5d 24 aa c1 27 00 00 00 91
00 00 00 5d 00 00 00 00 04 e2 46 97 f2 ad 00 4c 2f 7e
*** write_serial_packet 1 ***
raw 7e 45 00 00 00 04 e2 0c 04 96 00 00 00 5e 24 aa c1 27 00 00 00 91 c6 3f 7e
received 45 00 00 00 04 e2 0c 04 96 00 00 00 5e 24 aa c1 27 00 00 00 91 c6 3f
crc 3fc6 comp 3fc6
time [1250]: frame=94 guid=9100000027c1aa24 exists
resp [1250]: frame=94 guid=9100000027c1aa24 frame=94 sample=0 metadata=invalid
time=Fri Jul 13 15:46:29 2007
writing 44 27 00 ff ff 00 00 1b 00 96 00 00 00 5e 24 aa c1 27 00 00 00 91 00
00 00 5e 00 00 00 00 04 e2 46 97 f2 b5 00
encoded 7e 44 27 00 ff ff 00 00 1b 00 96 00 00 00 5e 24 aa c1 27 00 00 00 91
00 00 00 5e 00 00 00 00 04 e2 46 97 f2 b5 00 4d d0 7e
*** write_serial_packet 1 ***
----------------------
With the older code, the ACK is received, and the dialog continues more
normally:
----------------------
resp [1185]: frame=536 guid=8200000027bc8524 frame=536 sample=0
metadata=invalid
time=Fri Jul 13 16:05:26 2007
writing 44 26 00 ff ff 00 00 1b 00 96 00 00 02 18 24 85 bc 27 00 00 00 82 00
00 02 18 00 00 00 00 04 a1 46 97 f7 26 00
encoded 7e 44 26 00 ff ff 00 00 1b 00 96 00 00 02 18 24 85 bc 27 00 00 00 82
00 00 02 18 00 00 00 00 04 a1 46 97 f7 26 00 58 84 7e
raw 7e 43 26 3b 1c 7e
received 43 26 3b 1c
crc 1c3b comp 1c3b
raw 7e 45 00 00 00 04 a1 20 04 90 00 00 02 19 24 85 bc 27 00 00 00 82 00 07 00
10 6d 56 20 20 20 20 20 20 20 20 20 20 20 20 20 20 c9 96 7e
received 45 00 00 00 04 a1 20 04 90 00 00 02 19 24 85 bc 27 00 00 00 82 00 07
00 10 6d 56 20 20 20 20 20 20 20 20 20 20 20 20 20 20 c9 96
crc 96c9 comp 96c9
dtyp [1185]: frame=537 guid=8200000027bc8524
dtype_guid=7 name=mV type=int16 (0)
raw 7e 45 00 00 00 04 a1 20 04 90 00 00 02 1a 24 85 bc 27 00 00 00 82 00 04 01
10 25 2a 31 30 30 20 20 20 20 20 20 20 20 20 20 20 4c 51 7e 7e 45 00 00 00 04
a1 20 04 90 00 00 02 1b
received 45 00 00 00 04 a1 20 04 90 00 00 02 1a 24 85 bc 27 00 00 00 82 00 04
01 10 25 2a 31 30 30 20 20 20 20 20 20 20 20 20 20 20 4c 51
crc 514c comp 514c
----------------------
Thanks,
Steve
>
> David Gay
>
> > The patch below may be a suitable resolution.
> >
> > Cheers,
> > Steve
> >
> > Index: serialsource.c
> > ===================================================================
> > --- serialsource.c (revision 1024)
> > +++ serialsource.c (working copy)
> > @@ -768,6 +768,9 @@
> > {
> > struct packet_list *entry;
> >
> > + if (source_wait(src, &deadline) < 0)
> > + return 1;
> > +
> > read_and_process(src);
> > entry = pop_protocol_packet(src, P_ACK);
> > if (entry)
> > @@ -779,8 +782,6 @@
> > if (acked == src->send.seqno)
> > return 0;
> > }
> > - else if (source_wait(src, &deadline) < 0)
> > - return 1;
> > }
> > }
> >
> > _______________________________________________
> > Tinyos-help mailing list
> > [email protected]
> > https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
>
> !DSPAM:4697c636297881804284693!
_______________________________________________
Tinyos-help mailing list
[email protected]
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
