On Wed, Feb 17, 2010 at 3:50 AM, Pavel Roskin <proski@gnu.org> wrote:
On Tue, 2010-02-16 at 10:08 -0800, vishal11 wrote:
> Hi All
> Iam a student and am trying to incorporate fast channel switching in madwifi
> for my project. Currently as measured the time for execution of the function
> ath_set_channel for madwifi_0.9.4_trunkr3314 is about 20ms (as tested on
> Mikrotik RB433AH board running openwrt kamikaze v8.09). However for the
> later releases with HAL code (madwifi_0.9.4_r3980 and later), we have
> measured the execution time for the same function to be about 4-5ms. Also we
> have noticed that there is a function for AR5212 chipset,
> ar5212ChannelChange(), whose execution time as measured is about 500
> microseconds.
> However our problem is that we are not able to make madwifi_0.9.4_r3980 or
> later release (till latest r4119) work on Mikrotik RB433AH boards (employing
> Openwrt kamikaze v8.09). We have been able to compile the latest madwifi
> release on Openwrt, but when modules load, the following error message keeps
> appearing
> wifi0 ath_fatal_tasklet hardware error resetting

I suggest that you do your work on free code, such as madwifi trunk or
better yet, on ath5k.  I assume you use miniPCI cards and don't need AHB

Ok fine we can work on the madwifi trunk version which comes with HAL code, but we are not able to run it on Mikrotik boards employing Openwrt. The hardware and software configuration that we are employing is miniPCI Ubiquiti SR2 cards based on AR5212 chipset on Mikrotik RB433AH boards. The OS that we are using in Openwrt kamikaze v8.09 with linux kernel and we tried with madwifi_0.9.4r3980 till madwifi_0.9.4r4119. As mentioned when the modules load, the following error keeps appearing

wifi0 ath_fatal_tasklet hardware error resetting

I did apply the following patch https://madwifi-project.org/ticket/2322 , but to no avail. Could you point out as to what the problem could be?? If you need me to provide any other debug information please do let me know.
HAL in MadWifi 0.9.4 is essentially a black box.

> Morever we have already incorporated quite a few changes in
> madwifi_0.9.4_trunkr3314 and it will be difficult for us to port these
> changes to latest madwifi release. So would it be possible to make use of
> the function ar5212ChannelChange() in madwifi_0.9.4_trunkr3314. What changes
> will we need to make for that?? We tried to implement the function
> ar5212ChannelChange() in madwifi_0.9.4_trunkr3314 ourselves, but have not
> been successfull so far.
> Any help or suggestion would be greatly appreciated.

I doubt anyone would want to debug problems in non-free HAL at this

Pavel Roskin