Bug#413444: [Fwd: Re: [Rt2400-devel] rt2x00: Current issues #7]

[Ben Hutchings]

This appears to explain the bug as being the fault of mac80211.

Ben.

-------- Forwarded Message --------
From: Ivo van Doorn <ivdoorn at gmail.com>
To: Ben Hutchings <ben at decadent.org.uk>
Cc: Mattias Nissler <mattias.nissler at gmx.de>,
rt2400-devel at lists.sourceforge.net
Subject: Re: [Rt2400-devel] rt2x00: Current issues #7
Date: Tue, 2 Oct 2007 20:22:07 +0200

On Tuesday 02 October 2007, Ben Hutchings wrote:
> On Mon, 2007-10-01 at 15:44 +0200, Mattias Nissler wrote:
> > On Sun, 2007-09-30 at 18:44 +0200, Ivo van Doorn wrote:
> > > Please keep rt2400-devel in the CC list
> > > 
> > > On Sunday 30 September 2007, Ben Hutchings wrote:
> > > > On Sun, 2007-09-30 at 17:07 +0200, Ivo van Doorn wrote:
> > > > > On Sunday 30 September 2007, Ben Hutchings wrote:
> > > > > > On Sun, 2007-09-30 at 16:29 +0200, Ivo van Doorn wrote:
> > > > > > > 9 - Hang on resume
> > > > > > > - When resuming, the driver rt73usb driver will hang on the first register access
> > > > > > > - Does rt2500usb work?
> > > > > > > - Does PCI work?
> > > > > > 
> > > > > > rt2500pci is broken on resume; you need to reload the module to make it
> > > > > > work again.  It doesn't hang.
> > > > > 
> > > > > How exactly is it broken? Does it display any errors?
> > > > > Could you create a debugfs dump of before suspend and one after resume?
> > > > 
> > > > Is there a script I should use to create a debugfs dump?
> > > 
> > > See attachment
> 
> I'm attaching the dumps.

Thanks, I'll look into them.

> <snip my kernel log>
> > Hm, I don't have rt2500 hardware, but this looks fine to me. Please
> > remember that mac80211 is currently missing suspend/resume handlers, so
> > you cannot expect it to reinitialize the device properly.
> 
> Why would mac80211 need them, when this is a low level hardware
> function?  The rt2x00lib_resume function is obviously being called as
> expected.

Because you would be suspending while being associated. This means
that there will be no disassociation message, and mac80211 still thinks it can
freely configure the device and send out frames.
During resume, mac80211 will not be aware of the lost association which was
caused by the suspend, and will therefor not rescan/reassociate.

> > What happens if you trigger a scan after resume? If that works, I'd say
> > the driver is ok.
> 
> Scanning is broken after resume.  I can bring the interface back to life
> by running:
> 
> # iwconfig wlan0 essid $ESSID
> # iwconfig wlan0 ap $AP
> # dhclient wlan0
> 
> (All three commands appear to be needed.)  However, scanning remains
> broken after this and cannot be fixed except by removing and reinserting
> the card or the module.
> 
> There is a more nasty bug that rt2500pci causes a hang on resume if the
> card has been removed.  I will try to provide more detail if I can.

Ok,

Ivo

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.alioth.debian.org/pipermail/pkg-ralink-maintainers/attachments/20071117/d8f48e42/attachment.pgp