Bugzilla Bug 58483

Hi there,

the appended rt2500 ebuild includes the installation of the driver for the RaLink Rt2500 wlan-chipset and its configuration utility.

net-wireless/rt2500 should do it.
it depends on the kernelsources,and on qt (for the config-utility).
Its tested with a 2.6 kernel, but should work with 2.4,too.
I had to apply a patch to a file from the utility, as there was a wrong include, i will try to append this too

Greets,
DD

Created an attachment (id=36236) [details]
rt2500-1.4.1.0.ebuild 

Created an attachment (id=36237) [details]
Patch needed by rt2500-1.4.1.0-gentoo.ebuild

I've not had time too look into this driver much at all, but a few things:

1) If the utility is not *needed* (i noticed from the quick poke at their docs that iwlist/iwconfig are said to work), can you please put all the utility dependancies and build code behind the 'qt' USE flag?

2) As of 2.6.6 and newer, the recommended way to build modules is to use M= instead of SUBDIRS=, can you please change the ebuild to check for a new enough kernel and use the appropriate technique?

Thanks for the ebuild submission, with some small cleanups we'll see about getting it into portage. I'm a tad worried about the license, but we'll cross that bridge when we get there... (:

I just made a new revision of the ebuild with your recommended changes.

With the license i'm not really sure, but I wrote an email to ralinktech yesterday. Hope they will answer soon.

Created an attachment (id=36251) [details]
rt2500-1.4.1.0-r1.ebuild 

(From update of attachment 36237 [details])
This patch is still needed in by new revision, as otherwise the utility wont
compile

There seems to be a problem with the driver and/or the kernel.
Just found this thread on ralinktech.com.tw:
http://61.222.76.235/phpbb2/viewtopic.php?t=12

I don't think this is the "solution", but you can give it a chance... 

Created an attachment (id=36721) [details]
rt2500-1.4.2.0.ebuild 

Ralink released a new version. They did not change anything in the sources, but
added 2 includes (so the patch isnt necessary anymore).

Created an attachment (id=39197) [details]
new release with a patch included

Just read from a fix for the local_bh_enable() problems in ralink forum.
So this ebuild applies the patch
Thanks to Dan Morrill, who figured it out.

Created an attachment (id=39198) [details]
patch for local_bh_enable probs

(From update of attachment 39198 [details])
patch for local_bh_enable probs

Created an attachment (id=39439) [details]
same as above but in unix format; cleaner build 

(From update of attachment 39439 [details])
same as above but in unix format; cleaner build 

(From update of attachment 39198 [details])
replaced

Created an attachment (id=39733) [details]
rt2500-1.4.3.0 new driver release

Ralink released a new driver version, so here it is.
One warning: This ebuild is now masked with -x86, as when inserting the module
here, kernel panics. 
Maybe someone could also try his luck...

Created an attachment (id=40452) [details]
Patch for amd64 compatibility

Thanks for your ebuild. I tried it out on my amd64 machine, but I got a kernel
Oops when inserting the compiled module.
I attached a patch that solves the problem for me. I grabbed it from the RaLink
forums.

Created an attachment (id=40466) [details]
spin_lock patch

Here's another patch I found in the RaLink forums.
It fixes problems with my system being extremely slow during data transfers
through the wireless LAN module.

By the way, these are the forum topics I am referring to:

amd64 compatibility:
http://61.222.76.235/phpbb2/viewtopic.php?t=57

spin_lock:
http://61.222.76.235/phpbb2/viewtopic.php?t=13

here is a gentoo related wiki page for the rt2500 chipset:

http://gentoo-wiki.com/HARDWARE_rt2500

and http://gentoo-wiki.com/HARDWARE_rt2500_on_AMD64 for users of x86_64.

gentoo users (or someone else) can share their expierences with this chipset and drivers there.

The 1.4.3 drivers locks my kernel as soon as I bring the device up (as
described here).
The 1.4.2 drivers work, but it is necessary to disable IO-APIC (and probably
SMP if your capable of this). Might be worth a check and a warning message at
post-install. I don't know if this is card-dependent (I had the problem using a
MSI PC54G).

I had the same problem with the MSI PC54G. Only the 1.4.2 driver worked and I
had to comment the line 'WARN_ON(irqs_disabled());' in softirq.c of my
kernelsource (not a very nice hack...)

In rt2500-1.4.3.0.ebuild I had to add a line to actually perform the package's
Configure-script. Otherwise building for 2.4-Kernel fails:



# diff -ru rt2500-1.4.3.0.ebuild rt2500-1.4.3.0.ebuild.changed
--- rt2500-1.4.3.0.ebuild       2004-11-03 22:28:02.000000000 +0100
+++ rt2500-1.4.3.0.ebuild.changed       2004-11-03 22:30:01.000000000 +0100
@@ -63,6 +63,7 @@
                emake -C /usr/src/linux SUBDIRS=${S}/STA/Module modules || die
        else #this has to be 2.4, as otherwise we wont be here
                chmod +x ./Configure
+               ./Configure
                emake all || die
        fi

is this stable for someone on a 2.6 with a MSI PC54G2? If so, could he/she post
the kernel config?
I'm still getting random freezes (using the spinlock patch and even with
commenting the WARN thingy out).

The driver version 1.4.1 seems to be stable with the MSI PC54G2 (but it is
still necessary to comment the softirq WARN_ON out).

the _bh locks won't work stable here, I needed to change them to _irq:
+ spin_unlock_irq(lock); \
...
+ spin_lock_irq(lock); \

Created an attachment (id=45304) [details]
contains the STA of the unpacked rt2500 ebuild

i couldnt get the patches to work, and i did try to modify the c code.
Luckily I found someone who posted this file and i got it to work, and compile
with only a few warnings, could be my aggressive CFLAGS =P

Anyway, what i did, is i used the latest ebuild, and put it in a portage
overlay directory.next is some sudo code to what i did

ebuild rt2500 unpack
untar RT-2500-Linux-STA-blah to /var/tmp/portage/rt2500.../STA
ebuild rt2500 compile
ebuild rt2500 install
ebuild rt2500 qmerge

and your done, quick, mostly painless, and it works

A few days ago, Ralink released a new driver (1.4.4) and with a few changes of
the .4.3-ebuild I could built it without errors (ebuild is attached). I can
load the module into the kernel, but my system freezes as soon as I try to
'ifconfig ra0 up'. I had the same problem with the 1.4.3 driver and the 1.4.2
driver only works as long as I don't try to copy large files (>50MB). Therefore
I went back to the 1.4.1 driver which is working flawlessly for me.
I never had something like this before, that the driver got worse with every
new version...

(I'm using gentoo-dev-sources 2.6.8-r10 and the MSI-PC54G2 wlan card)

Created an attachment (id=45545) [details]
ebuild for the rt2500-1.4.4.0 driver

Created an attachment (id=45611) [details]
rt2500-1.4.4.0-spin-lock.patch

The new version of the driver includes the two patches I submitted some time
ago. The big change of 1.4.4.0 is the license: RaLink published the driver
under the GPL (Michael, your ebuild still says "freedist" here)!

Michal, the driver causes freezes on my system, too, but only with heavy
network traffic. Please try merging the attached patch file into your ebuild. I
grabbed it from the RaLink forum and it worked fine for me.

I added the spin-lock patch and the changed ebuild compiled without errors, but
my system still freezes when I execute 'ifconfig ra0 up'. Therefore I will
stick with 1.4.1 for now.
Ralink released today version 1.4.5, but it looks quite different to the
releases before and seems to have no qt utility. Maybe someone can get this to
work. I don't really have the time for many tests at the moment :-(

Created an attachment (id=45631) [details]
rt2500-1.4.4.0-r1.ebuild

*** Bug 74453 has been marked as a duplicate of this bug. ***

A generic rt2x00 driver is being developed at sourceforge, supporting both
rt2400 and rt2500 card. Said to be due to release. Status of rt2500 development
is currently just a mirror of the RaLink 1.4.4 (strangely enough called 1.0.0),
but that might improve soon. 

http://rt2x00.serialmonkey.com

A generic rt2x00 driver is being developed at sourceforge, supporting both
rt2400 and rt2500 card. Said to be due to release. Status of rt2500 development
is currently just a mirror of the RaLink 1.4.4 (strangely enough called 1.0.0),
but that might improve soon. 

http://rt2x00.serialmonkey.com

1) utility bit should be conditional on X and qt
2) my libXext is in /usr/X11R6/lib64/ , and there is no way to convince the ebuild of that.

Created an attachment (id=49316) [details]
New version available!!! (very messy ebuild, should be rebuild/rewritten)

Created an attachment (id=49319) [details]
rt2500-1.1.0_beta1.ebuild

update to linux-mod, please test and comment.

Thanx, Stefan!

1. The new 1.1.0_beta1.ebuild tries to install kde-base/kde-env... When I remove 'kde' from inherit, the dependency disappears...
2. Disabling the qt flag doesn't hide the 'dobin' command...

Created an attachment (id=49351) [details]
rt2500-1.1.0_beta1.ebuild

Is this ebuild ok for you?

3. The module isn't autoloaded by the kernel when starting /etc/init.d/net.ra0
(without modprobing rt2500 first). Shouldn't the alias be 'ra0' instead of
'ra'? 

Does "alias ra* rt2500" work?

Perfect, Stefan! 

Should we create a new bug with this version? Or try to get it into portage? According to the rt2x00 group this should be a major rewrite of the origional ralink driver (and much more stable)

in portage, thanks.

Stefan,

New version (1.1.0-b1) werking perfectly here... Could you change the line to 'MODULESD_RT2500_ALIASES=("ra* rt2500")' (change ra to ra*) and contact the kind gentoo people to get this into Portage? I've seen your name at the Prism54 driver, so can you arrange this?

Thanx...

I already added this to portage with this change.

Hi,

This new ebuild works perfectly. Thanks a lot!
But could you do me a favor and add ~amd64 to its keywords?

Thanks,

Alexander

I am not allowed to do it,

amd64 people, can you please add ~amd64 to keywords?

Alexander: you should open a new bug for this issue so that it can be tracked easier .. talking on resolved fixed bugs is no good style imo.

Stefan,

your're absolutely right. It's just that earlier versions of the ebuild that have been posted here already had this flag. On top of that, I have submitted a set of patches for amd64 compatibility for the old RaLink driver and confirmed that it works on my amd64. This is why I wondered why the flag has been removed in the final ebuild without evaluating if the ebuild works on amd64 platforms or not.

But if you want me to, I can open a new bug report for that. :-)

Regards,

Alexander M. Turek

I'm not sure if I can post here, But when emerging the last ebuild I get a
sandbox violation. Not quite sure how to solve it.

Thanks.
Carlos

--------------------------- ACCESS VIOLATION SUMMARY------------------------
LOG FILE = "/tmp/sandbox-net-wireless_-_rt2500-1.1.0_beta1-22393.log"

open_wr:   /usr/qt/3/etc/settings/.qtrc.lock
-----------------------------------------------------------------------------
Sorry, meant to post the error.

Harold Aling: Can you please try if it depends on kde now again?

carlos: thanks, I fixed it, can you please open a new bug for all issues that come up and not pollute this closed one?

I survived emerging rt2500 again without any bulky overhead! (no kde stuff, not
even kde-env)

Thumbs up...