Loading new broadcom-sta-5.100.82.112 DKMS files...
Building only for 3.2-3.towo.1-siduction-amd64
Building initial module for 3.2-3.towo.1-siduction-amd64
/bin/sh: 1: [: Illegal number:
/bin/sh: 1: [: Illegal number:
/bin/sh: 1: [: Illegal number:
/bin/sh: 1: [: Illegal number:
Done.
wl:
Running module version sanity check.
- Original module
- No original module exists within this kernel
- Installation
- Installing to /lib/modules/3.2-3.towo.1-siduction-amd64/updates/dkms/
do_depmod 3.2-3.towo.1-siduction-amd64
Es liegt am Makefile im /usr/src/broadcom-sta Verzeichnis:
if [ "$(VERSION)" -ge "2" -a "$(PATCHLEVEL)" -ge "6" -a "$(SUBLEVEL)" -ge "32" ]; then
besser wäre eine umgekehrte Logik, die dadurch rechtzeitig abbricht:
if [ "$(VERSION)" -lt "3" -a "$(PATCHLEVEL)" -lt "7" -a "$(SUBLEVEL)" -lt "33" ]; then
Quote
neues dkms für broadcom-sta - kann nicht kurze Kernelversio
towo wird schon erraten, was Du meinst...
greetz
devil
Devil, das broadcom dkms Paket ist ganz neu bei Debian eingetrudelt!
Der Sublevel mit -a AND verbunden verursacht bei Towo Kerneln natürlich eine NULL Exception. SLH hatte so kurze Kernel Versionen vorsorglich beim Übergang 2zu3 eingeführt, weil er Linus Torwalds falsch verstanden hatte. Bei mir geht es mit einem Hand-gekürzten Makefile :# written permission of Broadcom Corporation.
#
# $Id: Makefile_kbuild_portsrc,v 1.6.54.4 2011-01-27 00:55:34 Exp $
ifneq ($(KERNELRELEASE),)
APICHOICE := FORCE_CFG80211
$(info CFG80211 API specified in command line)
LINUXVER_GOODFOR_CFG80211 := TRUE
APICHOICE := PREFER_CFG80211
$(info CFG80211 API is prefered for this kernel version)
APIFINAL := CFG80211
endif
EXTRA_CFLAGS :=
ifeq ($(APIFINAL),CFG80211)
EXTRA_CFLAGS += -DUSE_CFG80211
$(info Using CFG80211 API)
endif
ifeq ($(APIFINAL),WEXT)
EXTRA_CFLAGS += -DUSE_IW
$(info Using Wireless Extension API)
endif
obj-m += wl.o
wl-objs :=
wl-objs += src/shared/linux_osl.o
wl-objs += src/wl/sys/wl_linux.o
wl-objs += src/wl/sys/wl_iw.o
wl-objs += src/wl/sys/wl_cfg80211.o
EXTRA_CFLAGS += -I$(src)/src/include
EXTRA_CFLAGS += -I$(src)/src/wl/sys -I$(src)/src/wl/phy
#EXTRA_CFLAGS += -DBCMDBG_ASSERT
#ralul
KVER = `$(ls -1 -t /lib/modules/.|grep '^3'|head -n 1)`
ifeq ($(KVER),)
KVER = $(shell uname -r)
endif
KDIR = /lib/modules/$(KVER)/build
PWD = $(shell pwd)
KARCH := $(shell egrep 'CONFIG_X86_(32|64)=y' $(KDIR)/.config 2>/dev/null| cut -d= -f1)
ifeq ($(KARCH),CONFIG_X86_32)
SHIPPED=wlc_hybrid.o_i386
$(info Kernel architecture is X86_32)
else
KARCH := CONFIG_X86_64
SHIPPED=wlc_hybrid.o_amd64
endif
EXTRA_LDFLAGS := $(src)/lib/$(SHIPPED)
all:
KBUILD_NOPEDANTIC=1 make -C $(KDIR) M=$(PWD)
clean:
KBUILD_NOPEDANTIC=1 make -C $(KDIR) M=$(PWD) clean
install:
install -D -m 755 wl.ko /lib/modules/$(KVER)/kernel/drivers/net/wireless/wl.ko
Ich glaube, dass es doch vielleicht gar nicht an der kurzen Kernelversion liegt. Es scheinen einfach die richtigen Shell-Umgebungsvariablen nicht dem Makefile übergeben zu werden. Dann ist es doch ein upstream Debian Bug des dkms-Pakets !
This issue is triggered when SUBLEVEL is blank in /lib/modules/kernel-version/Makefile (as is the case with 3.6-*). If the kernel version was 3.6.0 (SUBLEVEL = 0), this wouldn't happen.
I posted a patch to the Makefile as a workaround to this issue, on this bug (which I opened to update broadcom-dkms to 6.20.55.19): http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=688823#10
Of course it's only a workaround.
But the kernel bug with 3.6 still exists, even with the first stable release as of yesterday.
http://marc.info/?l=linux-wireless&m=134908356432162&w=2
Besten GruBen :)
Yesterday I had a look at the wireless mailinglist to find something about it: There I saw your report with "towo" ! Thank you for care!
But I wonder
- why I can connect via wl.ko with siduction linux-3.6?
- but using Towos same linux/.config with Gentoo it segfaults like yours?
German: Ich stehe vollkommen auf dem Schlauch :(
And nobody has answered your message in the mailinglist yet :(
Perhaps some simple alias-renamings in broadcom headers are possible:
In linux/include/net/cfg80211.h
I can see simple renames
net_device to wireless_dev
dev to wdev
[edit]But it is too complicated, I cannot find the points where to define aliases in the broadcom source :(
@jasmineaura
Wpasupplicant-1.0 seems to be everywhere the same
Your broadcom-dkms
6.20.55.19 seems to be newer than mine:
5.100.82.112
I analyzed my paradox situation:
I had vastly edited the Makefile of Debian broadcom-sta for that it uses NOT wext. This broadcom sources I copied to my Gentoo installation. But my Debian had successfully updated that broadcom-sta package. In effect I use wext on Debian and cfg80211 on Gentoo. Because the older wext method didn't change with linux-3.6 this method is working on Debian with Linux-3.6.
But not that newer method on Gentoo with Linux-3.6 any more :(
@jasmineaura, this is working with my 3 years old broadcom chip but perhaps not any more wext is usable with your newer one ....
@jasmineaura, as I can see your
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=688823
is not solved yet, perhaps you can use b43 with newer firmware instead of braodcom-sta module building:
http://forum.siduction.org/index.php?msg=24880#24880