Summary: | net-wireless/broadcom-sta-6.30.223.248-r1 with kernel 3.17 and 3.18 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | mike <mike> |
Component: | [OLD] Core system | Assignee: | Matt Turner <mattst88> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | bpshacklett, darwinskernel, proxy-maint, v.simankin |
Priority: | Normal | Keywords: | PATCH |
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
broadcom-sta-6.30.223.248-linux-3.17.patch
ebuild patch broadcom-sta-6.30.223.248-linux-3.18.patch ebuild patch w/ 3.18 ebuild patch w/ 3.18 (typo fix) gcc 4.9 patch Patch for panic of broadcom-sta-6.30.223.248 on kernel 3.18 |
Description
mike@marineau.org
2014-09-20 20:12:35 UTC
Created attachment 385200 [details, diff]
ebuild patch
confirmed working on 3.17.0 tested on 3.17.1 and it works Tested, build fine with kernel 3.17.1-r1, please tell admin and add patch to portage tree. (In reply to mike@marineau.org from comment #0) > Created attachment 385198 [details, diff] [details, diff] > broadcom-sta-6.30.223.248-linux-3.17.patch > > Linux 3.17-rc introduces a minor API change: > > http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/ > ?id=c835a677331495cf137a7f8a023463afd9f032f8 > > Attached patch has been tested against 3.17-rc5 could you please test for 802.11n? i have: 02:00.0 Network controller [0280]: Broadcom Corporation BCM4321 802.11a/b/g/n [14e4:4328] (rev 03) and it doesn't do 11n here with 3.16.x series. i can provide more info/logs if need be. (In reply to darwinskernel from comment #5) > could you please test for 802.11n? > > i have: > > 02:00.0 Network controller [0280]: Broadcom Corporation BCM4321 > 802.11a/b/g/n [14e4:4328] (rev 03) > > and it doesn't do 11n here with 3.16.x series. i can provide more info/logs > if need be. The patch posted here does not have any impact on 3.16 kernels and n/ac work just fine for me with a BCM4360. Your issue should be unrelated to this bug. Created attachment 389138 [details, diff] broadcom-sta-6.30.223.248-linux-3.18.patch Another kernel cycle, another small API change: http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=5bc8c1f2b070bab82ed738f98ecfac725e33c57f Fixes build w/ 3.18-rc4 Created attachment 389140 [details, diff]
ebuild patch w/ 3.18
Created attachment 389142 [details, diff]
ebuild patch w/ 3.18 (typo fix)
Patch for 3.17 works fine for me since I updated my laptop to 3.17. (In reply to mike@marineau.org from comment #6) > (In reply to darwinskernel from comment #5) > > could you please test for 802.11n? > > > > i have: > > > > 02:00.0 Network controller [0280]: Broadcom Corporation BCM4321 > > 802.11a/b/g/n [14e4:4328] (rev 03) > > > > and it doesn't do 11n here with 3.16.x series. i can provide more info/logs > > if need be. > > The patch posted here does not have any impact on 3.16 kernels and n/ac work > just fine for me with a BCM4360. Your issue should be unrelated to this bug. how did you check for n/ac? with 3.17 patch on 3.17.3 kernel none of the popular wireless tools (iwlist, wavemon) work here! the link is up though but i don't know how to tell what the bitrate/mode is. The 3.17 patch works for me on sys-kernel/gentoo-sources-3.17.4. I think it'd be nice if this fix could be pushed out as soon as possible now that 3.16 is EOL. hello maintainer, please voice, we can help push this the 3.18 patch can wait, since no 3.18 kernel has hit into portage tree Hi, guys, Kernel 3.18 has been released for several days, but our patch is still not in portage, why? +*broadcom-sta-6.30.223.248-r2 (15 Dec 2014) + + 15 Dec 2014; Yixun Lan <dlan@gentoo.org> + +broadcom-sta-6.30.223.248-r2.ebuild, + +files/broadcom-sta-6.30.223.248-linux-3.17.patch, + +files/broadcom-sta-6.30.223.248-linux-3.18.patch: + fix build err with kernel 3.17/3.18, bug 523326 maintainer timeout, so pushed this. you may still have problem to build with gcc-4.9 + kernel3.18 (should open as another bug) /var/tmp/portage/net-wireless/broadcom-sta-6.30.223.248-r2/work/src/wl/sys/wl_linux.c:2054:3: error: macro "__DATE__" might prevent reproducible builds [-Werror=date-time] __DATE__, __TIME__, EPI_VERSION_STR); ^ /var/tmp/portage/net-wireless/broadcom-sta-6.30.223.248-r2/work/src/wl/sys/wl_linux.c:2054:13: error: macro "__TIME__" might prevent reproducible builds [-Werror=date-time] __DATE__, __TIME__, EPI_VERSION_STR); Created attachment 391798 [details, diff]
gcc 4.9 patch
(In reply to Yixun Lan from comment #15) > +*broadcom-sta-6.30.223.248-r2 (15 Dec 2014) > + > + 15 Dec 2014; Yixun Lan <dlan@gentoo.org> > + +broadcom-sta-6.30.223.248-r2.ebuild, > + +files/broadcom-sta-6.30.223.248-linux-3.17.patch, > + +files/broadcom-sta-6.30.223.248-linux-3.18.patch: > + fix build err with kernel 3.17/3.18, bug 523326 > > maintainer timeout, so pushed this. > > you may still have problem to build with gcc-4.9 + kernel3.18 (should open > as another bug) > > /var/tmp/portage/net-wireless/broadcom-sta-6.30.223.248-r2/work/src/wl/sys/ > wl_linux.c:2054:3: error: macro "__DATE__" might prevent reproducible builds > [-Werror=date-time] > __DATE__, __TIME__, EPI_VERSION_STR); > ^ > /var/tmp/portage/net-wireless/broadcom-sta-6.30.223.248-r2/work/src/wl/sys/ > wl_linux.c:2054:13: error: macro "__TIME__" might prevent reproducible > builds [-Werror=date-time] > __DATE__, __TIME__, EPI_VERSION_STR); Thanks for pushing this into portage tree. I think -Werror=date-time can be bypassed by adding an cflags which is Wno-date-time. see gcc 4.9 patch from package "broadcom-wl" in archlinux AUR. Anyone else seeing a kernel lockup with this + 3.18? http://imgur.com/4LLkGtW here is the lockup (In reply to Vladi from comment #19) > http://imgur.com/4LLkGtW > > here is the lockup A null pointer dereference during a mprotect() systemc call, that seems unlikely to be this driver unless it is scribbling over random memory. Is it repoducable? Do you get similar stack traces or is the crash entirely different/random looking each time? In any case the report should be a new bug, not this old one. not sure that seems to come before the actual lockup right after i modprobe wl module. I also get a kernel panic on 3.18 with these patches. (In reply to Brennan Shacklett from comment #22) > I also get a kernel panic on 3.18 with these patches. Hi,did you blacklist these modules(b43 b44 ssb bcma brcmsmac brcmfmac) in /etc/modprobe.d/ ? These modules are quite conflicted with wl driver. Created attachment 393374 [details, diff]
Patch for panic of broadcom-sta-6.30.223.248 on kernel 3.18
I too had the panic/lockup of wl.ko from version 6.30.223.248 after upgrading to 3.18 from 3.17.4. I bisected and arrived at kernel commit bec3cfdca36bf43cfa3751ad7b56db1a307e0760 by Eric Dumazet that first triggered this bug. I contacted him and he was so nice to look at the Broadcom source and proposed the fix that I attached as a patch here. This fixes it for me.
I can confirm that the gcc-4.9 and kernel panic patches are needed too. I can confirm that the patch suplied by Bolko Maass fixes the kernel panic in 3.18.1 and 3.18.2. But I compiled it with gcc-3.8.4 and it runs OK for me. The patch above is also working for me on kernel 3.18.3. (In reply to Bolko Maass from comment #24) > Created attachment 393374 [details, diff] [details, diff] > Patch for panic of broadcom-sta-6.30.223.248 on kernel 3.18 > > I too had the panic/lockup of wl.ko from version 6.30.223.248 after > upgrading to 3.18 from 3.17.4. I bisected and arrived at kernel commit > bec3cfdca36bf43cfa3751ad7b56db1a307e0760 by Eric Dumazet that first > triggered this bug. I contacted him and he was so nice to look at the > Broadcom source and proposed the fix that I attached as a patch here. This > fixes it for me. Thanks, that's great work. Much appreciated! It looks like the Arch maintainers found this bug report and the patch you attached and rolled it into a nice patch that takes care of the API changes since 3.15. I've committed that, so this bug should legitimately be fixed now. I've been using 3.18 and this patch for about a week now without any trouble. |