Roaming problems on NVIDIA Jetson TX2

Manuel Wagesreither ManWag at FastMail.FM
Wed Jun 16 14:54:09 PDT 2021


Hi Steve,

Am Di, 15. Jun 2021, um 22:26, schrieb Steve deRosier:
> Hi Manuel,
> 
> On Tue, Jun 15, 2021 at 8:25 AM Manuel Wagesreither <ManWag at fastmail.fm> wrote:
> >
> > Hello all,
> >
> > I'm working on a project which is using wpa_supplicant 2.9 with NetworkManager 1.22.10 on a NVIDIA Jetson TX2 system-on-module using a CYW4354=BCM4354. The GNU/Linux distribution is a custom one created with Yocto; the kernel used is 4.9.140-l4t-r32.4+g3924d6fccfbf provided by NVIDIA.
> >
> 
> I expect you aren't going to get a lot of help here.  You're asking
> about a 5+ year old vendor kernel with a Broadcom vendor driver, ie
> not fully open source and upstream stuff.  And you're asking about
> largely driver oriented stuff on the hostapd list.  Granted, a lot of
> the same people are on multiple lists, so you'll hit a subset of the
> linux-wireless folk too.  Since your stuff is provided by NVIDIA, I'd
> suggest asking them for support.

I agree I should seek contact with NVIDIA support. Need to check first if the problem occurs with their official Ubuntu Fork as well, otherwise they won't give support. As of linux-wireless, I was unaware of that list, so thank you for pointing that out.


> 
> 
> > I observed the following behaviour:
> > 1. The error message `Failed to enable signal strength monitoring` shows up in our logs.
> 
> It's not necessarily an error message.  Do you need to enable "signal
> strength monitoring"?  The drivers often output lots of information,
> you still need to evaluate if you actually care.  I'm guessing
> everything works fine and you don't care.
> 
> > 2. When roaming to a access point with a similar SSID, roaming works. (WebRTC video stream stutters but doesn't break.)
> > 3. When roaming to an access point with a different SSID, the "Dongle Host Driver" (which is a kernel module I guess?) restarts.
> 
> Typically switching to a different SSID is not considered "roaming."
> And thus would typically have a more intense change. Indeed, if you
> look at your kernel log above your marked point, it's clear that the
> stack is being fully torn down and it looks like the WiFi chip's bus
> is being powered down, which usually results in an unload of the
> driver. And then powered back up reloads the driver which reloads the
> firmware.
> 
> DHD or "Dongle Host Driver" is the Broadcom proprietary vendor driver
> for their WiFi chips.
> 
> I hope that helps you understand at least a little bit.
> 
> - Steve
> 
Much appreciated, it indeed does.

- Manuel



More information about the Hostap mailing list