[mythtv-commits] Ticket #10414: HDHomeRun: Bad Recordings
MythTV
noreply at mythtv.org
Sun Jul 22 05:34:49 UTC 2012
#10414: HDHomeRun: Bad Recordings
------------------------------------------------+--------------------------
Reporter: Billy Macdonald <billymacdonald@…> | Owner: danielk
Type: Bug Report - General | Status: accepted
Priority: minor | Milestone: 0.26
Component: MythTV - Recording | Version: Master
Severity: medium | Head
Keywords: | Resolution:
| Ticket locked: 0
------------------------------------------------+--------------------------
Comment (by David Walker <David@…>):
I've compiled a new mythbackend with the 20120405 version of libhdhomerun
and ran it with net.core.rmem_max = 1048576. This improved things, but it
was only a compensation, as spartacus06 said. Mythbackend's log still
reported gaps in recorded programs.
While waiting for compiles, though, I did some packet traces with tcpdump
and noticed something pretty suspicious. While a recording is in
progress, there are long, uninterrupted strings of 1328-byte UDP packets,
as expected. A few seconds before the start of a recording gap, though,
the HDHomeRun Prime starts sending ARP requests for the MythTV server once
a second. The server responds to each of the ARP requests. After a few
seconds, the UDP stream stops, causing the recording gap. While the UDP
stream stops, the once-a-second ARP requests continue. I'll attach a
tcpdump output file (tcp.log.gz) showing this behavior; I've edited long
runs of the UDP packets out, as you will see in the file.
I could see that the ARP requests stopped eventually, and I had a hunch
that they stopped when the HDHomeRun Prime received a unicast packet from
the MythTV server, so I tried running
ping -q -n <HDHomeRun Prime address>
from the server to generate a once-a-second stream of unicast ping
packets. I scheduled several recordings, and no gaps! Then I reverted to
the Packman version of mythbackend and the standard default for
net.core.rmem_max (114688), but left the ping running. Still no
recordings gaps. The once-per-second ARP requests still occur, but the
pings seem to keep the UDP stream going.
So, I think the HDHomeRun Prime has an issue. It may be aggravated by
something that MythTV is doing, but it looks like a bug in the HDHomeRun
network code to me.
Is someone from SiliconDust following this? Is this known behavior?
--
Ticket URL: <http://code.mythtv.org/trac/ticket/10414#comment:25>
MythTV <http://code.mythtv.org/trac>
MythTV Media Center
More information about the mythtv-commits
mailing list