[asterisk-bugs] [DAHDI-linux 0017724]: [patch] Race condition in handling writes to proslic LINEFEED register

[Asterisk Bug Tracker]

A NOTE has been added to this issue. 
====================================================================== 
https://issues.asterisk.org/view.php?id=17724 
====================================================================== 
Reported By:                alecdavis
Assigned To:                
====================================================================== 
Project:                    DAHDI-linux
Issue ID:                   17724
Category:                   wctdm24xxp
Reproducibility:            always
Severity:                   minor
Priority:                   normal
Status:                     new
JIRA:                       DAHDI-677 
Reviewboard Link:            
====================================================================== 
Date Submitted:             2010-07-27 06:32 CDT
Last Modified:              2010-07-29 13:18 CDT
====================================================================== 
Summary:                    [patch] Race condition in handling writes to proslic
LINEFEED register
Description: 
set_lasttxhook_interruptible times out waiting for SLIC_LF_OPPENDING on 
fxs->lasttxhook to go.

Even changing the timeout in set_lasttxhook_interruptible to 2 Seconds,
the bottom half of <b>if (SLIC_LF_OPPENDING & fxs->lasttxhook)  {</b>
randomly get executed.




======================================================================
Relationships       ID      Summary
----------------------------------------------------------------------
related to          0017318 [patch] FXS Polarity Reversal on Hangup...
====================================================================== 

---------------------------------------------------------------------- 
 (0125280) alecdavis (manager) - 2010-07-29 13:18
 https://issues.asterisk.org/view.php?id=17724#c125280 
---------------------------------------------------------------------- 
[101824.010533] wctdm24xxp 0000:05:04.0: Setting FXS hook state to 2 (04)
intcount=93196
[101824.026535] wctdm24xxp 0000:05:04.0: SLIC_LF OK: card=3 shadow=14
lasttxhook=04 intcount=93202
[101824.030194] wctdm24xxp 0000:05:04.0: wctdm: Card 3 Going off hook
[101824.031335] wctdm24xxp 0000:05:04.0: Setting FXS hook state to 1 (02)
intcount=93207

The above SLIC_LF times between Setting from 2 to 1 when going off hook,
is 93196 -> 93207 = 11ms, so I'd think there's not too much room to
increase.

I was however thinking, if the 1ms hook detector ran on 'odd' intcounts,
and the oppending check ran on 'even' intcounts, the interrupt load would
be distrubuted. 

Issue History 
Date Modified    Username       Field                    Change               
====================================================================== 
2010-07-29 13:18 alecdavis      Note Added: 0125280                          
======================================================================