Bug #4875: [WARNING] mpegts: too much queued table input data (over 2MB), discarding new - Tvheadend

Bug #4875

Bug #4851: TVH freezes: mpegts: too much queued table input data (over 2MB), discarding new

[WARNING] mpegts: too much queued table input data (over 2MB), discarding new

Added by Igor Mokrushin almost 7 years ago. Updated over 4 years ago.

Status:

Fixed

Priority:

Normal

Assignee:

Adam Sutton

Category:

EPG - Grabbers

Target version:

Start date:

2018-01-19

Due date:

% Done:

Estimated time:

Found in version:

4.2 commit 1665c355cb14c210fbe647cd2c911d9083061bd0

Affected Versions:

Description

After commit 1665c355cb14c210fbe647cd2c911d9083061bd0(release/4.2) there were problems:

Logs have messages like:
[WARNING] mpegts: too much queued table input data (over 2MB), discarding new
Does not respond web interface.
CPU usage > 53% after restart...

P.S. Testing was conducted on boards MIPS MT7621 with 256MB memory and DVB-T2 USB dongle. With previous commit these problems are not observed.

Files

bad_tvheadend.log (30.5 KB) bad_tvheadend.log	current commit 1665c355cb14c210fbe647cd2c911d9083061bd0	Igor Mokrushin, 2018-01-19 19:57
ok_tvheadend.log (214 KB) ok_tvheadend.log	previous commit 31d5854ccd73fde028fdbbcedd3ad23e13f8d124	Igor Mokrushin, 2018-01-19 19:57
om_retry_time.patch (764 Bytes) om_retry_time.patch		Igor Mokrushin, 2018-01-19 23:02
new_om_retry_time.patch (777 Bytes) new_om_retry_time.patch		Igor Mokrushin, 2018-01-21 23:19

History

Updated by Mark Clarkstone almost 7 years ago

Parent task set to #4851

Updated by Igor Mokrushin almost 7 years ago

File om_retry_time.patch om_retry_time.patch added

New int64_t om_retry_time in structure epggrab_ota_mux causes race condition....
In src/epggrab/otamux.c om_retry_time is not initialized anywhere. If the condition (r! = SM_CODE_NO_ADAPTERS) is not fulfilled, then om->om_retry_time = 0 always!
This condition is almost always satisfied: if (networks_count < ARRAY_SIZE(networks) && om && om->om_retry_time < mclk()) goto next_one;
If set om->om_retry_time > mclk(), then the problem disappears. I made a fix that solves this problem, it's almost like the previous variant...

Updated by Jaroslav Kysela almost 7 years ago

Could you show me really bad traces ? It seems that your uploaded bad_tvheadend.log is ok . Your assumption is incorrect - the zero value in om->om_retry_time is fine, the condition means to not do fast retry for already tried muxes. You changed the condition to 'retry only if the mux failed in 60 seconds'.

Updated by Jaroslav Kysela almost 7 years ago

Does this change help?

--- a/src/epggrab/otamux.c
+++ b/src/epggrab/otamux.c

       if (net->failed) {
         TAILQ_INSERT_TAIL(&epggrab_ota_pending, om, om_q_link);
         om->om_q_type = EPGGRAB_OTA_MUX_PENDING;
+        om->om_retry_time = mclk() + mono2sec(60);
         goto done;
       }
       break;
     }
   }

Updated by saen acro almost 7 years ago

cannot send any log because there is UI restart continuously,
(only UI sub process main stay alive)
but I disable all EPG modules and all go near perfect

Updated by Pim Zandbergen almost 7 years ago

I was seeing the same problem.
Revision commit:945e1d5f appears to have fixed it.

Thanks!

Updated by Anonymous almost 7 years ago

I normally experience the problem 2-3 times a day but 14 hours after applying the patch, I've not yet experienced the problem.

Updated by Igor Mokrushin almost 7 years ago

Hi, Jaroslav!

In bad_tvheadend.log not hited lines:
[WARNING] mpegts: too much queued table input data (over 2MB), discarding new
As system used 100% CPU and I had to turn off the power. This is all that was saved in the file...
I use branch 4.2, applied patches to it from revision 945e1d5f, I do not observe this problem any more. Thanks!