rt2x00 is still experimental?

rt2x00 is still experimental?

[Pavel Roskin]

Hello!

Can we please have rt2x00 drivers considered stable in Linux 2.6.27?  My
experience running rt73 day by day with WPA on a very crowded network
shows that it's a pretty reliable driver already.

I realize that distros would enable whatever they want, and most people
here have CONFIG_EXPERIMENTAL on anyway, but still, it could help direct
users towards better supported hardware and make CONFIG_EXPERIMENTAL
more relevant.

The latest commits to the drivers/net/wireless/rt2x00 directory seem to
confirm that the major issues have been addressed already.

-- 
Regards,
Pavel Roskin

Re: rt2x00 is still experimental?

[Ivo van Doorn]

Hi,

> Can we please have rt2x00 drivers considered stable in Linux 2.6.27? 

No.

> My experience running rt73 day by day with WPA on a very crowded network
> shows that it's a pretty reliable driver already.

Perhaps managed mode, but not adhoc mode, beaconing is still broken.
In addition the leds are broken as well, and hardware encryption isn't in either.

But to make a very important nore:

rt2400pci: unknown status
rt2500pci: TX rate problems, LED problems
rt2500usb: Completely broken, segfault on unplug, no HW crypto
rt61pci: TX rate problems, LED problems, no HW crypto
rt73usb: TX rate problems, LED problems. no HW crypto

I have 4 open Fedora bugreports, and 3 open Kernel bugreports

When I finally have my testsystem capable of running a >=2.6.26 kernel
then I can test rt2400pci, and _perhaps_ I can mark that one as stable since
it seems to work quite well on 2.6.25.

But marking the drivers as stable with be a per-rt2x00-driver decision, so what
I can do is moving the CONFIG_EXPERIMENTAL dependency to the drivers instead
of globally on CONFIG_RT2X00

> I realize that distros would enable whatever they want, and most people
> here have CONFIG_EXPERIMENTAL on anyway, but still, it could help direct
> users towards better supported hardware and make CONFIG_EXPERIMENTAL
> more relevant.
> 
> The latest commits to the drivers/net/wireless/rt2x00 directory seem to
> confirm that the major issues have been addressed already.

Not sure what you saw in the latest commits, but they are not confirming anything
about stability. ;)

Ivo

Re: rt2x00 is still experimental?

[Pavel Roskin]

Quoting Ivo van Doorn <ivdoorn@gmail.com>:

> Hi,
>
>> Can we please have rt2x00 drivers considered stable in Linux 2.6.27?
>
> No.
>
>> My experience running rt73 day by day with WPA on a very crowded network
>> shows that it's a pretty reliable driver already.
>
> Perhaps managed mode, but not adhoc mode, beaconing is still broken.
> In addition the leds are broken as well, and hardware encryption   
> isn't in either.

I see.  It must be a perception thing.  Things I hack on seem  
unstable, but things that just work seem stable :)

Thank you for the status.  Good luck with fixing the remaining problems!

-- 
Regards,
Pavel Roskin

Re: rt2x00 is still experimental?

[Ivo van Doorn]

On Saturday 12 July 2008, Pavel Roskin wrote:
> Quoting Ivo van Doorn <ivdoorn@gmail.com>:
> 
> > Hi,
> >
> >> Can we please have rt2x00 drivers considered stable in Linux 2.6.27?
> >
> > No.
> >
> >> My experience running rt73 day by day with WPA on a very crowded network
> >> shows that it's a pretty reliable driver already.
> >
> > Perhaps managed mode, but not adhoc mode, beaconing is still broken.
> > In addition the leds are broken as well, and hardware encryption   
> > isn't in either.
> 
> I see.  It must be a perception thing.  Things I hack on seem  
> unstable, but things that just work seem stable :)

Well it is at least good to know managed mode is working well for you. :)

> Thank you for the status.  Good luck with fixing the remaining problems!

Thanks,

Ivo

Re: rt2x00 is still experimental?

[Michael Buesch]

On Saturday 12 July 2008 17:13:54 Ivo van Doorn wrote:
> On Saturday 12 July 2008, Pavel Roskin wrote:
> > Quoting Ivo van Doorn <ivdoorn@gmail.com>:
> > 
> > > Hi,
> > >
> > >> Can we please have rt2x00 drivers considered stable in Linux 2.6.27?
> > >
> > > No.
> > >
> > >> My experience running rt73 day by day with WPA on a very crowded network
> > >> shows that it's a pretty reliable driver already.
> > >
> > > Perhaps managed mode, but not adhoc mode, beaconing is still broken.
> > > In addition the leds are broken as well, and hardware encryption   
> > > isn't in either.
> > 
> > I see.  It must be a perception thing.  Things I hack on seem  
> > unstable, but things that just work seem stable :)
> 
> Well it is at least good to know managed mode is working well for you. :)
> 
> > Thank you for the status.  Good luck with fixing the remaining problems!
> 
> Thanks,

I think the EXPERIMENTAL flag is just meaningless.
You cannot run a kernel without it, as so many required and working
subsystems and drivers depend on it.
So I decided to drop EXPERIMENTAL for b43, as It Works (tm) for most users.
Sure, there are a lot of corners that need fixing, but fixing them all
up before removing EXPERIMENTAL would most likely never happen.

So we removed EXPERIMENTAL from the driver when the driver worked
for most users in STA mode. IMO that improves the overall situation,
as it's one less EXPERIMENTAL dependency, that isn't really one.

-- 
Greetings Michael.

Re: rt2x00 is still experimental?

[Stefanik Gábor]

On Sat, Jul 12, 2008 at 5:31 PM, Michael Buesch <mb@bu3sch.de> wrote:
> I think the EXPERIMENTAL flag is just meaningless.
> You cannot run a kernel without it, as so many required and working
> subsystems and drivers depend on it.
> So I decided to drop EXPERIMENTAL for b43, as It Works (tm) for most users.
> Sure, there are a lot of corners that need fixing, but fixing them all
> up before removing EXPERIMENTAL would most likely never happen.
>
> So we removed EXPERIMENTAL from the driver when the driver worked
> for most users in STA mode. IMO that improves the overall situation,
> as it's one less EXPERIMENTAL dependency, that isn't really one.

I agree that we should reconsider what EXPERIMENTAL really means - the
whole Linux kernel is slowly becoming EXPERIMENTAL, with the original
purpose of the EXPERIMENTAL flag slowly being replaced with dependency
on external patches (see AP mode in mac80211). IMHO in general, if
something is working as most users would expect it, then it shouldn't
be EXPERIMENTAL unless there is a legacy alternative and a good reason
to suggest people to use that alternative. EXPERIMENTAL should mean
something that is not ready for the general public yet, not something
that is simply under active development.

-- 
Vista: [V]iruses, [I]ntruders, [S]pyware, [T]rojans and [A]dware. :-)

Re: rt2x00 is still experimental?

[Stefan Monnier]

> to suggest people to use that alternative. EXPERIMENTAL should mean
> something that is not ready for the general public yet, not something
> that is simply under active development.

It seems that for several drivers the EXPERIMENTAL config flag would be
better replaced by an `experimental' module argument that enables the
in-development features.


        Stefan

Re: rt2x00 is still experimental?

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 237 bytes --]


> I agree that we should reconsider what EXPERIMENTAL really means

I think you get to weigh in on what that means when you do a significant
amount of work.

_fix_ AP mode instead of complaining about me blocking it.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]