Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

/drivers/net/wireless/rtl8192cu is missing from 3.17 & 3.18 #755

Closed
DougieLawson opened this issue Jan 5, 2015 · 12 comments
Closed

/drivers/net/wireless/rtl8192cu is missing from 3.17 & 3.18 #755

DougieLawson opened this issue Jan 5, 2015 · 12 comments

Comments

@DougieLawson
Copy link

/drivers/net/wireless/rtl8192cu is in 3.16 but missing from 3.17 & 3.18.

With a 3.18 kernel my Edimax RTL8188CUS dongle keeps dropping out and losing the network. Can the 8192cu stuff be added back in to the 3.18 source tree?

@amtssp
Copy link

amtssp commented Jan 5, 2015

I have a similar problem with 3.18 and rtl8192cu Edimax wifi.

@popcornmix
Copy link
Collaborator

It appears some users prefer the upstream rtl8192cu driver and some the out-of-tree driver. E.g.
#369 #566

There were conflicts with the out-of-tree driver when moving to 3.18 so I chose to give the upstream driver a try. Really I would very much prefer to use the upstream driver. If there are bugs with it, then they should be reported upstream, fixed there and I can pull in the fixes.

However I accept that the upstream driver has been broken (at least for some users) for a long time, and it is a very widely used driver on the Pi, so some flexibility is possible.
I'll see if there is a newer version of the out-of-tree driver that works with 3.18.

It would be useful if you could file a upstream bug, as making the upstream driver reliable is the best solution. See https://www.kernel.org/pub/linux/docs/lkml/reporting-bugs.html

@DougieLawson
Copy link
Author

At the moment this is a 100% show stopper.

If you were to try to push this kernel on the normal branch (not BRANCH=next) then close to half of the WiFi attached Raspberry Pis in the World will stop working (because the Edimax dongle is one of the highly recommended dongles and has been since at least August 2012 when I got my first RPi).

There's probably a larger user base with Raspberry Pis than on any upstream machines. How many folks outside of "Raspberry Pi Land" wouldn't have some other brand of WiFi on-board and be using USB dongles now?

The upstream driver runs for a short time, but then silently fails, it's so flaky that there's now a long cat5 cable running from my switch to the Raspberry Pi that lives in my conservatory to keep it connected. If I could work out what parms to pass to the module then I'm happy to run it with very verbose debugging enabled. I've been unable to find out what parms can be set for power management either (with this "new" driver).

The out-of-tree driver has some well documented parms that are commonly offered on the RPi forum as a solution to the WiFi dongle randomly disconnecting and most "punters" go away happy with those.

I'm running with a self-built (cross compiled) 3.18.1 kernel and not adverse to patching code.

@DougieLawson
Copy link
Author

I've sent an email to [email protected] reporting this bug. If you'd like a copy of what I've sent please ask.

@popcornmix
Copy link
Collaborator

@DougieLawson
If you can find source for rtl8192cu that builds for 3.18 that would be useful. I believe I originally got the driver source from here:
http://www.realtek.com.tw/downloads/downloadsView.aspx?Langid=1&PNid=48&PFid=48&Level=5&Conn=4&DownTypeID=3&GetDown=false&Downloads=true#RTL8192CU

but that's over a year old and only claims to support up to kernel 3.9.

I seem to remember finding a github which maintained a newer version of that code.

@MrEngman
Copy link

MrEngman commented Jan 6, 2015

I set up an SD card with 3.18.1 then I copied the rtl8192cu source from 3.12.35+ #730 into a separate directory and cloned linux and firmware from github and checked them out for 3.18.y

/home/pi/src/linux
             firmware
             rtl8192cu

Then setup linux using

cd /home/pi/src/Linux
sudo ln -s /home/pi/src/linux /lib/modules/$(uname -r)/build
sudo ln -s /home/pi/src/linux/arch/arm /home/pi/src/linux/arch/armv6l

make mrproper && zcat /proc/config/gz > .config && make modules_prepare && cp ../firmware/extra/Modules.symvers .

then patched rtl8192cu/Makefile to get it to compile.

--- Makefile.old        2015-01-06 13:03:53.876742260 +0000
+++ Makefile    2015-01-06 02:32:10.181163732 +0000
@@ -70,6 +70,7 @@
 CONFIG_PLATFORM_MSTAR_A3 = n
 CONFIG_PLATFORM_ARM_SUNxI = n
 CONFIG_PLATFORM_ARM_SUN6I = n
+CONFIG_PLATFORM_RPI_BCM2708 = y

 CONFIG_DRVEXT_MODULE = n

@@ -236,6 +237,16 @@
 EXTRA_CFLAGS += -DCONFIG_INTEL_WIDI
 endif

+ifeq ($(CONFIG_PLATFORM_RPI_BCM2708), y)
+EXTRA_CFLAGS += -DCONFIG_LITTLE_ENDIAN
+ARCH := arm
+CROSS_COMPILE ?=
+KVER  := $(shell uname -r)
+KSRC := /lib/modules/$(KVER)/build
+MODDESTDIR := /lib/modules/$(KVER)/kernel/drivers/net/wireless/
+INSTALL_PREFIX :=
+endif
+
 ifeq ($(CONFIG_PLATFORM_I386_PC), y)
 EXTRA_CFLAGS += -DCONFIG_LITTLE_ENDIAN
 SUBARCH := $(shell uname -m | sed -e s/i.86/i386/)

Compiled the driver using

cd /home/pi/src/rtl8192cu
make clean && make && sudo make install

Came up with a single warning but compiled and appears to run without problems as in 3.12.xx.

@popcornmix
Copy link
Collaborator

Thanks @MrEngman
I'll add it back in.

Seems the build problem I was hitting was the change in procfs. I had to disable CONFIG_PROC_DEBUG to get it to build.

@MrEngman
Copy link

MrEngman commented Jan 6, 2015

Thanks, that will make a lot of people happy.

re: procfs. I've no idea what that is but glad you've fixed it. Note,I had to compile using gcc 4.7 as gcc 4.8 caused command make modules_prepare to fail with an error. see issue #758

One thing I forgot. To get the 8192cu driver to load in 3.18.1 #731 it's necessary to blacklist the rtl8192cu driver that is currently compiled in the 3.18.1 image.

Add the line

blacklist rtl8192cu

to file /etc/modprobe.d/raspi-blacklist.conf. Note no spaces before blacklist rtl8192cu.

@popcornmix
Copy link
Collaborator

I've pushed it. I didn't need the blacklist - I'm getting 8192cu as default.
Not sure what the kernel rules are for when two modules both support a VID/PID, probably undefined.

It may make sense to blacklist the less preferred (rtl8192cu) as a default on raspbian. If someone wants to switch they can swap the blacklisted one.

popcornmix added a commit to raspberrypi/firmware that referenced this issue Jan 6, 2015
See: raspberrypi/linux#755

kernel: Make LEDS_GPIO builtin rather than a module
raspberrypi/linux#757

kernel: Fix the activity LED in DT mode
See: raspberrypi/linux#757

firmware: arm_loader: DT support for initramfs usage
See: http://www.raspberrypi.org/forums/viewtopic.php?f=29&t=93015

firmware: hdmi: Accept EDID even if checksum is wrong on final read
See: http://openelec.tv/forum/124-raspberry-pi/74408-problems-with-hdmi-audio-on-openelec-5-0

firmware: video_decode: don't include frames without callback structures in timestamp statistics
See: http://forum.kodi.tv/showthread.php?tid=211501&pid=1876654#pid1876654
popcornmix added a commit to Hexxeh/rpi-firmware that referenced this issue Jan 6, 2015
See: raspberrypi/linux#755

kernel: Make LEDS_GPIO builtin rather than a module
raspberrypi/linux#757

kernel: Fix the activity LED in DT mode
See: raspberrypi/linux#757

firmware: arm_loader: DT support for initramfs usage
See: http://www.raspberrypi.org/forums/viewtopic.php?f=29&t=93015

firmware: hdmi: Accept EDID even if checksum is wrong on final read
See: http://openelec.tv/forum/124-raspberry-pi/74408-problems-with-hdmi-audio-on-openelec-5-0

firmware: video_decode: don't include frames without callback structures in timestamp statistics
See: http://forum.kodi.tv/showthread.php?tid=211501&pid=1876654#pid1876654
@MrEngman
Copy link

MrEngman commented Jan 7, 2015

Yes, works just fine, thanks.

@popcornmix
Copy link
Collaborator

@DougieLawson okay for you?

@DougieLawson
Copy link
Author

I'll trust the excellent work from Mr Engman. He is the "Dongle Driver King".

I got a reply from the upstream maintainer, but it was only suggesting updating the stuff in /lib/firmware. So I won't be pursuing that.

Thanks everyone. This can be closed.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants