[SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

eworm · 2013-08-19 19:28:47

akanski wrote:

@eworm,
You can reverse the minstrel_ht commit before compiling the mac80211 module like I did (#45). If you also do such a thing, you won't have that "crappy connectivity". In fact, I'm using Linux 3.10.6 right now and I've had no trouble with my broadcom wireless since then.

Sure... But that does not help debugging the root problem.

Thralas · 2013-08-19 20:43:41

akanski wrote:

Recent fix in mac80211/minstrel_ht is the main cause of broadcom wireless malfunction (fix cck rate sampling).

I was getting panics on both 3.10.6 and 3.10.7 - one contains the commit you linked, the other doesn't.

akanski · 2013-08-19 20:44:56

@eworm,

Of course, that doesn't. However, that could help some developers to focus their attention on cck rate sampling in order to know what's happening with broadcom hardware. Therefore, reversing that commit is only a temporary measure, and not a definite one.

Furthermore, bear in mind that the default kernel is not compiled with options as CONFIG_CRASH_DUMP and CONFIG_DEBUG_INFO. Consequently, we can't do much debugging unless we compile our own kernel to debug that issue.

Last edited by akanski (2013-08-19 23:14:57)

akanski · 2013-08-19 23:11:23

Thralas wrote:

akanski wrote:
Recent fix in mac80211/minstrel_ht is the main cause of broadcom wireless malfunction (fix cck rate sampling).
I was getting panics on both 3.10.6 and 3.10.7 - one contains the commit you linked, the other doesn't.

Could you look at the git log first:

https://git.kernel.org/cgit/linux/kerne … 0.6&ofs=65

In fact, I did not see anything new for the mac80211 wireless stack in Linux 3.10.7:

https://git.kernel.org/cgit/linux/kerne … gs/v3.10.7

On the other hand, in addition to download both packages (1,2), I cannot see any minstrel_ht change either between patch-3.10.6.xz and patch-3.10.7.xz. As a result, that commit since Linux 3.10.6 is still in use.

Unfortunately, I can realize that you did not pore over this matter, did you?

https://www.kernel.org/diff/diffview.cg … -3.10.6.xz
https://www.kernel.org/diff/diffview.cg … -3.10.7.xz
___

N.B. I reversed that commit again just a few minutes after upgrading, and I've got no kernel panic with brcm80211 driver on Linux 3.10.7.

Last edited by akanski (2013-08-23 23:17:15)

pdizzle · 2013-08-20 01:24:58

WonderWoofy wrote:

zigzed wrote:
pdizzle wrote:
the broadcom-wl package from the AUR fixed this for me, if only temporarily
edit: after updating to 3.7.10 this morning, I am no longer able to see my wireless adapter, with or without the Broadcom-wl module.
you need re-compile the broadcom-wl module from source after upgrade to 3.10.7.
This is not true. Most of the functionality of the module comes from the included binary blob (hence the designation as a "hybrid" source, meaning part GPL, and part proprietary). So I have found that most of the time, the build module will simply continue to work through all of a kernel's minor version (ie. through all of 3.10). There was one exception when I had to recompile it going from something like 3.9.6 to 3.9.7 or something. It lives in the "extramodules" directory, so it will be applied to any kernel update that isn't a bump in the minor version number.

this is a bit over my head, but thanks for the response. I've downgraded the kernel and headers to 3.10.6-2 and am happy enough for a short time.

edit: re-upgraded the kernel and then recompiled/installed the broadcom-wl package and everything is working jsut fine for now. thanks!

Last edited by pdizzle (2013-08-20 01:34:00)

WonderWoofy · 2013-08-20 02:19:59

@pdizzle, so what I was trying to tell you is that most of the time, that broadcom-wl module will continue to work when you upgrade from say, 3.10.6 to 3.10.7. But it will almost certainly break when you upgrade to 3.11 when that is released. Though this *might* be simply because of the location of the module.

From time to time, there will be a release that is a "binary breaker" in which things will indeed break during the update. This doesn't happen very often with the wl module, but I did experience it going from (I think) 3.9.6 to 3.9.7.

archan.paul · 2013-08-20 06:43:38

Following seems to solve the crash. Will ArchLinux patch core/linux 3.10.7-x such that crash can be resolved with stock kernel?

http://www.spinics.net/lists/linux-wire … 11229.html
https://git.kernel.org/cgit/linux/kerne … 595897bef6

Last edited by archan.paul (2013-08-20 06:44:45)

Awebb · 2013-08-20 13:58:59

WonderWoofy wrote:

@pdizzle, so what I was trying to tell you is that most of the time, that broadcom-wl module will continue to work when you upgrade from say, 3.10.6 to 3.10.7. But it will almost certainly break when you upgrade to 3.11 when that is released. Though this *might* be simply because of the location of the module.
From time to time, there will be a release that is a "binary breaker" in which things will indeed break during the update. This doesn't happen very often with the wl module, but I did experience it going from (I think) 3.9.6 to 3.9.7.

It should not really matter in this case, if I condensed the information in this thread correctly, then this issue will be solved, by the time 3.11 is out, so broadcom-wl won't be needed anymore.

WonderWoofy · 2013-08-20 14:07:35

@Awebb, this is true. But since there are some in this thread that are going the route of finding a temporary workaround, I figured I'd offer an explanation to pdizzle, who indicated that he/she was confused by my previous explanation.

eworm · 2013-08-20 14:15:27

To date probably the best solution comes from Felix Fietkau:
http://www.gossamer-threads.com/lists/l … 99#1768699

A patched kernel is available:
http://dl.mylinuxtime.de/arch/eworm/x86 … pkg.tar.xz

Again, this is completely untested.

Awebb · 2013-08-20 14:39:59

I would test this, but my netbook runs on 32bit. Would you mind providing your PKGBUILD/tarball, so I can test it?

vincentstorme · 2013-08-20 18:37:57

removed

Last edited by vincentstorme (2020-01-02 22:14:17)

eworm · 2013-08-21 09:07:47

Updated my fixed kernel:
http://dl.mylinuxtime.de/arch/eworm/x86 … pkg.tar.xz

Awebb · 2013-08-21 10:02:47

eworm wrote:

Updated my fixed kernel:
http://dl.mylinuxtime.de/arch/eworm/x86 … pkg.tar.xz

Awebb wrote:

I would test this, but my netbook runs on 32bit. Would you mind providing your PKGBUILD/tarball, so I can test it?

eworm · 2013-08-21 10:11:52

Awebb wrote:

eworm wrote:
Updated my fixed kernel:
http://dl.mylinuxtime.de/arch/eworm/x86 … pkg.tar.xz
Awebb wrote:
I would test this, but my netbook runs on 32bit. Would you mind providing your PKGBUILD/tarball, so I can test it?

Did you receive my mail?

However, here is the updated tarball with PKGBUILD and file for anybody who wants to build:
http://www.eworm.de/tmp/linux.tar.gz

anti-destin · 2013-08-21 17:54:47

@eworm: is that the patch applied to the stock kernel?

eworm · 2013-08-21 20:17:20

anti-destin wrote:

@eworm: is that the patch applied to the stock kernel?

This is the patch applied to the Arch kernel. They patch anything about Logitech mice, but the wlan changes are the only thing I did.

snoppiz · 2013-08-21 20:37:13

@eworm, i took the your kernel-patch and it worked. I am using the brcmsmac driver and it works.

eworm · 2013-08-22 13:52:41

Did anybody file an issue on bugs.archlinux.org? Could not find anything so far.

Methos · 2013-08-22 15:28:22

eworm wrote:

Did anybody file an issue on bugs.archlinux.org? Could not find anything so far.

Neither do I. But one bug report mentions this bug: https://bugs.archlinux.org/task/36539?p … &pagenum=1

eworm · 2013-08-22 15:52:12

Methos wrote:

eworm wrote:
Did anybody file an issue on bugs.archlinux.org? Could not find anything so far.
Neither do I. But one bug report mentions this bug: https://bugs.archlinux.org/task/36539?p … &pagenum=1

You are right... This is unrelated to Radeon.
I posted some hints there... Let's hope this will find its way into Arch kernels soon.

WonderWoofy · 2013-08-22 15:56:14

Maybe it might be a better idea to make a new bug, and link back to the old one indicating that the other is invalid and not related to what the poster had thought. Maybe then, the right eyes will find it since it will (hopefully) have been posted in the right section. As it is now, your post eworm is the only really relevant post there, and there has been no response whatsoever from Thomas Bachler.

akanski · 2013-08-22 19:50:41

@WonderWoofy,

Felix Fietkau's patch limits cck rate sampling in mac80211/minstrel_ht to drivers that can handle it in order to avoid the use of this stuff on hardware that doesn't work with it (e.g. broadcom wireless).

https://lkml.org/lkml/2013/8/20/19

The above means that my assumption about mac80211 wireless stack was indeed right on. Apart from that, @eworm did some excellent work on the matter by searching for it.

Therefore, if we get to the main point here, the problem is not due to right eyes that are unable to find relevant information because it's posted in a wrong section, but because some maintainers haven't provided feedback on this thread; quite the opposite, other ones like, for instance, Allan McRae who did it immediately on glibc 2.18 last week.

Last edited by akanski (2013-08-23 17:49:53)

PedroF · 2013-08-23 01:01:01

eworm wrote:

However, here is the updated tarball with PKGBUILD and file for anybody who wants to build:
http://www.eworm.de/tmp/linux.tar.gz

The second md5sum is not correct (in the PKGBUILD). f614b2847eef5fc22cfc9f1e2d28a3 should be 868d7f5315f95da5e48ed56691a36263 for the file patch-3.10.9.xz

Last edited by PedroF (2013-08-23 01:01:16)

WonderWoofy · 2013-08-23 01:08:31

@PedroF, the pacman package now includes a really handy tool called updpkgsums, which will simply find the mentioned files in the sources array as well as the type of checksumming that is applied (ie. md5sum, sha1sum, sha256sum, etc) and fix things up for you.

Arch Linux

#51 2013-08-19 19:28:47

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#52 2013-08-19 20:43:41

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#53 2013-08-19 20:44:56

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#54 2013-08-19 23:11:23

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#55 2013-08-20 01:24:58

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#56 2013-08-20 02:19:59

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#57 2013-08-20 06:43:38

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#58 2013-08-20 13:58:59

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#59 2013-08-20 14:07:35

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#60 2013-08-20 14:15:27

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#61 2013-08-20 14:39:59

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#62 2013-08-20 18:37:57

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#63 2013-08-21 09:07:47

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#64 2013-08-21 10:02:47

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#65 2013-08-21 10:11:52

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#66 2013-08-21 17:54:47

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#67 2013-08-21 20:17:20

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#68 2013-08-21 20:37:13

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#69 2013-08-22 13:52:41

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#70 2013-08-22 15:28:22

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#71 2013-08-22 15:52:12

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#72 2013-08-22 15:56:14

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#73 2013-08-22 19:50:41

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#74 2013-08-23 01:01:01

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

#75 2013-08-23 01:08:31

Re: [SOLVED] Linux 3.10.6 and brcmsmac: kernel panic

Board footer