Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 11 Mar 2005 11:12:26 -0800
From:      Sam Leffler <sam@errno.com>
To:        frank@exit.com
Cc:        Eric Anderson <anderson@centtech.com>
Subject:   Re: [Fwd: cvs commit: src/sys/modules Makefile src/sys/conf files src/sys/modules/ath_rate_sample Makefile src/sys/i386/conf NOTES]
Message-ID:  <4231ED9A.3090504@errno.com>
In-Reply-To: <200503111829.j2BITm9x094682@realtime.exit.com>
References:  <200503111829.j2BITm9x094682@realtime.exit.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Frank Mayhar wrote:
> Eric Anderson wrote:
> 
>>Hmm - this was working this morning before I updated, with a -CURRENT as of last week, and I didn't notice anything in /usr/src/UPDATING.  So what I should have is:
>>
>>ath_load="YES"
>>ath_hal_load="YES"
>>ath_rate_load="YES"
> 
> 
> I just have
> 
> if_ath_load="YES"
> 
> in my /boot/loader.conf.  That picks up everything.
> 
> 
>>I guess what I missed (and just noticed), is that there are not individual modules for each rate algorithm now, one module (named ath_rate) that is the algorithm of whatever is built. 
> 
> 
> It has been this way for some time, actually.

Yes, I forgot about this little "detail".  As a result of my change 
folks using modules are now using ath_rate_sample by default.  I'll deal 
with it.

> 
> 
>>>You want to go to sys/modules/ath_rate_{amrr|onoe|sample} and do a 'make
>>>install'.
>>
>>I did that already - but why doesn't a 'make installkernel ...' not accomplish that?  (forgive my ignorance)
> 
> 
> It does.  It installs ath_rate.ko for each of amrr, onoe and (now) sample.
> Obviously, the last one in wins.  You'll have to ask Sam about what shortcoming
> of the FreeBSD module system leads to this.

Actually it's nothing to do with the module system.  When I split the 
rate control code into a separate module various folks didn't want to 
incur the overhead of using indrect function pointers in the fast path 
so we used the symbol names to bind rate control support to the driver. 
  This has many downsides and I'm not reall happy with it but since it's 
unlikely folks will want to run multiple algorithms concurrently I've 
not spent any time on the issue.

> 
> 
>>>Hmm.  Is ath_rate_sample linked properly?
>>
>>How can I help you determine that?   (is readelf -a output what you want?)
> 
> 
> Already fixed; read your -current email.

Yes, Tai-hwa fixed my botch (I'd tested only with a statically linked 
kernel).

	Sam



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4231ED9A.3090504>