rlm_sql_null holding locks and hanging radsqlrelay.

[Matthew Schumacher]

Alan,

Last year you directed me to patch rad_unlockfd() to use F_UNLCK instead 
of F_WRLCK.  I did that and things have been stable, but while building 
the latest patches for 3.0.20 I noticed that the 3.0.x branch still uses 
F_WRLCK.  Any reason that this change hasn't made it to the 3.0.x 
branch?  I would have thought that would be something I'd see in the 
next release.

Thanks,
schu


On 8/1/18 2:12 PM, Alan DeKok wrote:
>    Nope.  The rad_unlockfd() function didn't use F_UNLCK.  Instead it used F_WRLCK.
>
>    That was a bug which has been around for a *long* time. :(
>
>> On Aug 1, 2018, at 4:36 PM, Matthew Schumacher <matt.s at aptalaska.net> wrote:
>>
>> On 8/1/18 12:28 PM, Alan DeKok wrote:
>>> Grab the code from git, v3.0.x branch. Matthew Newton found a tiny but rather shocking little bug. :(
>> Is this what you are talking about?
>>
>> https://github.com/mcnewton/freeradius-server/commit/8cf46c98e336a69894d2470ba6988ca4042c4985#diff-34bf3f9b619dacda92ac1628fcff86e4
>>
>> Thanks,
>> schu
>>
>>