[Gta04-owner] Modem logs

[Neil Jerram]

Neil Jerram <neil at ossau.homelinux.net> writes:

> To complete such an integration, and hopefully help understand the modem
> reset scenarios, I plan now to modify the program so that (a) it saves
> the logs into a circular memory buffer, instead of printing them out,
> and (b) there is some kind of trigger, probably a signal, that causes
> the problem to dump that circular buffer to file and then exit.

Here's an update with (a) implemented.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: gta04-modem-messages.c
Type: text/x-csrc
Size: 9344 bytes
Desc: not available
URL: <http://lists.goldelico.com/pipermail/gta04-owner/attachments/20130106/57570ddb/attachment.bin>
-------------- next part --------------

Sample run:

  root at neo:/opt/qtmoko# bin/gta04-modem-messages -l 2 -c 10000 -k 1
  Message level 2
  Message count 10000
  Memory buffer size 1 Kb
  main: Bad CRC (after 3039 messages)

  [TS 290367849] HIGH (rtr8600.c:685) Rx Tune Mode: 1, Band: 0, Channel: 10761
  [TS 290367849] HIGH (wl1sleep.c:4274) RFM Wakeup - 3538 us, 116 sc (expected 16 
  [TS 290367849] ERROR (clkrgm_msm.c:1481) NULL source for clk 294
  [TS 290367851] HIGH (tlm.c:2220)
  [TS 290367851] HIGH (tlm.c:2288) tlm_debug_message:after recomputing hw 
  [TS 290367851] HIGH (tlm.c:2392) After interuption TFN1 (fr 1943, cx8 93326),
  [TS 290367851] HIGH (tlm.c:2402) After interuption TRUE WRC (fr 791, cx8 
  [TS 290367851] HIGH (srchzz.c:6365) srchzz_wakeup_cells() psc:27 freq:10761 bi
  [TS 290367851] HIGH (srchzz.c:10594) <---- Reacq passed: eng(166096)pos(92755)psc(
  [TS 290367851] HIGH (wsrch.c:7306) For UI:Ecio=9, AGC=-87, Filt_RSCP=175, C
  [TS 290367851] HIGH (srchzz.c:5755) MCR: Cur: E 0x0 Next: E 0x2 MODE CHAN
  [TS 290367851] HIGH (srchzz.c:5598) MCR: PICH/MICH add mode 1 E 2 RES EVA
  [TS 290367852] HIGH (drx.c:2893) PICH_STATS idx=0, PE=0, m_n=19, m_PE=
  [TS 290367852] HIGH (drx.c:3036) [15903] next_drx slp lngr 63744 Cx8 (28 ticks

New observations:

- Message level selection doesn't work at the ttyHS_Diagnostic level -
  i.e. it will deliver LOW messages even if HIGH is requested.  I'm
  working around this by ignoring messages below the requested level.

- There's sometimes a CRC error, which could be genuine or could be a
  coding bug.  Either way, I need to investigate if it works to ignore
  those and carry on reading following messages; at the moment the code
  just gives up when it sees a bad CRC.

- With the memory buffering the program still takes around 30% CPU,
  which is too much for constant usage.  Probably I can reduce that by
  storing raw packets in the memory buffer and parsing them only when
  dumping out the buffer.

Regards,
        Neil