On Thu, Mar 28, 2013 at 2:02 AM, Camaleón <[email protected]> wrote:
> 2013/3/27 Camaleón <[email protected]>:
>> El 2013-03-27 a las 09:41 -0700, Luis R. Rodriguez escribió:
> I tried this: switched off the button for wifi+bt (this is a netbook),
> then manually loaded these modules in this order (mac80211, bcma,
> compat, brcmutil, cfg80211 and cordic). At this point the system is
> fine, no freezes and dmesg shows normal messages.
Great.
> Now, as soon as I proceed to load "brcmsmac" I get a long and
> uninterruptible "beeeeep", system hard-freezes (no mouse nor caps lock
> led reaction) and only restart is possible.
>
> Any clue?
You are on 3.9.0-rc2, so there is nothing for us to backport on these
linux-next based releases except the delta between v3.9 and what is
not yet on Linus' tree, in other worlds just the development delta for
the 802.11 subsystem. We have yet to require any backport work between
3.9 and what we assume will be 3.10. My bet is this is a real driver
bug, not backport bug.
Can you
git clone git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-testing.git
and use your .config, and build and test that kernel ? If it fails to
boot / crash then you know its the driver.
Has brcmsmac ever worked for you?
Luis
El 2013-03-28 a las 03:56 -0700, Luis R. Rodriguez escribió:
> On Thu, Mar 28, 2013 at 3:40 AM, Camaleón <[email protected]> wrote:
> > Now you say... can you please tell me -in user's parlance terms so I can
> > understand it- what's the difference between "brcmsmac" from the stock
> > kernel modules and this one from "compat-drivers"? I mean, what's the
> > difference from a user's POV and from kernel hacker POV. Are they
> > related? Are they going to me merged somehow? What's the goal of having
> > both?
>
> The releases annotated with a date come from linux-next. While Linus
> is whipping people in shape for the RC releases in preparation for the
> first official release we hounds still do some development. During the
> RC cycle only regression fixes get merged into Linus' tree as well a
> shiny new drivers as they do not regress. The development cycle
> continues on without Linus. Stephen Rothwell maintains a tree that
> sucks all development trees together daily, this is called linux-next.
> We use linux-next for compat-drivers given that we backport not just
> wireless but also drm. So by you using dated compat-drivers releases
> you are using what's currently being developed and will likely hit
> 3.10.
Mmm... it resembles to me like Debian's own release cycle where the
"testing" branch would be like the drivers included at stock kernel
and "unstable" would be similar to what compat-drivers are aimed for
→ the next hit.
Well, more or less :-)
Thanks for the explanation. Now awaiting for "git clone" ends its job.
Greetings,
--
Camaleón
On Thu, Mar 28, 2013 at 3:40 AM, Camaleón <[email protected]> wrote:
> Now you say... can you please tell me -in user's parlance terms so I can
> understand it- what's the difference between "brcmsmac" from the stock
> kernel modules and this one from "compat-drivers"? I mean, what's the
> difference from a user's POV and from kernel hacker POV. Are they
> related? Are they going to me merged somehow? What's the goal of having
> both?
The releases annotated with a date come from linux-next. While Linus
is whipping people in shape for the RC releases in preparation for the
first official release we hounds still do some development. During the
RC cycle only regression fixes get merged into Linus' tree as well a
shiny new drivers as they do not regress. The development cycle
continues on without Linus. Stephen Rothwell maintains a tree that
sucks all development trees together daily, this is called linux-next.
We use linux-next for compat-drivers given that we backport not just
wireless but also drm. So by you using dated compat-drivers releases
you are using what's currently being developed and will likely hit
3.10.
Luis
2013/3/28 Luis R. Rodriguez <[email protected]>:
> On Thu, Mar 28, 2013 at 2:02 AM, Camale?n <[email protected]> wrote:
(...)
>> Now, as soon as I proceed to load "brcmsmac" I get a long and
>> uninterruptible "beeeeep", system hard-freezes (no mouse nor caps lock
>> led reaction) and only restart is possible.
>>
>> Any clue?
>
> You are on 3.9.0-rc2, so there is nothing for us to backport on these
> linux-next based releases except the delta between v3.9 and what is
> not yet on Linus' tree, in other worlds just the development delta for
> the 802.11 subsystem.
I have two sets of kernels installed in this system: Debian's stock
kernel (3.2.39-2) and another one I get from kernel.org (3.9.0-rc2).
Now, what happens is that the compat-driver freezes with kernel 3.9
and I get this message with kernel 3.2:
[ 684.589268] Compat-drivers backport release: compat-drivers-2013-03-26-u
[ 684.589279] Backport based on linux-next.git next-20130326
[ 684.589285] compat.git: linux-next.git
[ 684.609187] cfg80211: Unknown symbol
backport_netdev_set_default_ethtool_ops (err 0)
This driver is "driving" me nuts :-)
> We have yet to require any backport work between 3.9 and what we assume will be
> 3.10. My bet is this is a real driver bug, not backport bug.
Curious is that brcmsmac module loads fine from the common kernel
module, I mean, the stock one.
> Can you
>
> git clone git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-testing.git
>
> and use your .config, and build and test that kernel ? If it fails to
> boot / crash then you know its the driver.
I'm not using git (don't even have it installed in my system), I'm
used to compile kernels/modules in the old fashion way. Sorry, I don't
know how git works.
> Has brcmsmac ever worked for you?
Well, yes (this is a long story? with many helpful people doing their
best). It stopped from working since kernel 2.38 or so, but even
though the driver is giving me some headaches, it never froze the
system like is doing now.
? http://marc.info/?t=136359599600001&r=1&w=2
Greetings,
--
Camale?n
On Thu, Mar 28, 2013 at 2:56 AM, Camaleón <[email protected]> wrote:
> 2013/3/28 Luis R. Rodriguez <[email protected]>:
>> On Thu, Mar 28, 2013 at 2:02 AM, Camaleón <[email protected]> wrote:
>
> (...)
>
>>> Now, as soon as I proceed to load "brcmsmac" I get a long and
>>> uninterruptible "beeeeep", system hard-freezes (no mouse nor caps lock
>>> led reaction) and only restart is possible.
>>>
>>> Any clue?
>>
>> You are on 3.9.0-rc2, so there is nothing for us to backport on these
>> linux-next based releases except the delta between v3.9 and what is
>> not yet on Linus' tree, in other worlds just the development delta for
>> the 802.11 subsystem.
>
> I have two sets of kernels installed in this system: Debian's stock
> kernel (3.2.39-2) and another one I get from kernel.org (3.9.0-rc2).
>
> Now, what happens is that the compat-driver freezes with kernel 3.9
> and I get this message with kernel 3.2:
>
> [ 684.589268] Compat-drivers backport release: compat-drivers-2013-03-26-u
> [ 684.589279] Backport based on linux-next.git next-20130326
> [ 684.589285] compat.git: linux-next.git
> [ 684.609187] cfg80211: Unknown symbol
> backport_netdev_set_default_ethtool_ops (err 0)
>
> This driver is "driving" me nuts :-)
I have a fix for this, mind you, this is a Debian thing, that is --
debian cherry picked some code from a future kernel on their own
kernel. The new LINUX_BACKPORT() macro will rescue these. I'll post a
new release tomorrow, but I'm working on backporting some other
collateral evolutions right now.
>> We have yet to require any backport work between 3.9 and what we assume will be
>> 3.10. My bet is this is a real driver bug, not backport bug.
>
> Curious is that brcmsmac module loads fine from the common kernel
> module, I mean, the stock one.
What do you mean? Have you tried the stock driver from 3.9-rc2?
>> Can you
>>
>> git clone git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-testing.git
>>
>> and use your .config, and build and test that kernel ? If it fails to
>> boot / crash then you know its the driver.
>
> I'm not using git (don't even have it installed in my system), I'm
> used to compile kernels/modules in the old fashion way. Sorry, I don't
> know how git works.
make localmodconfig
make
make install
>> Has brcmsmac ever worked for you?
>
> Well, yes (this is a long story¹ with many helpful people doing their
> best). It stopped from working since kernel 2.38 or so, but even
> though the driver is giving me some headaches, it never froze the
> system like is doing now.
>
> ¹ http://marc.info/?t=136359599600001&r=1&w=2
Sorry to hear that, if its a driver issue I hope the maintainers will
jump in as they are Cc'd.
Luis
El 2013-03-28 a las 03:23 -0700, Luis R. Rodriguez escribió:
> On Thu, Mar 28, 2013 at 2:56 AM, Camaleón <[email protected]> wrote:
(...)
> > Now, what happens is that the compat-driver freezes with kernel 3.9
> > and I get this message with kernel 3.2:
> >
> > [ 684.589268] Compat-drivers backport release: compat-drivers-2013-03-26-u
> > [ 684.589279] Backport based on linux-next.git next-20130326
> > [ 684.589285] compat.git: linux-next.git
> > [ 684.609187] cfg80211: Unknown symbol
> > backport_netdev_set_default_ethtool_ops (err 0)
> >
> > This driver is "driving" me nuts :-)
>
> I have a fix for this, mind you, this is a Debian thing, that is --
> debian cherry picked some code from a future kernel on their own
> kernel. The new LINUX_BACKPORT() macro will rescue these. I'll post a
> new release tomorrow, but I'm working on backporting some other
> collateral evolutions right now.
Okay, no problem and thanks for caring about this.
> >> We have yet to require any backport work between 3.9 and what we assume will be
> >> 3.10. My bet is this is a real driver bug, not backport bug.
> >
> > Curious is that brcmsmac module loads fine from the common kernel
> > module, I mean, the stock one.
>
> What do you mean? Have you tried the stock driver from 3.9-rc2?
Sure! (full report is on the kernel wireless mailing list)
In fact, the only reason I compiled the kernel is to test the latest brcmsmac
driver because being this a netbook, having to use an external usb card
to be "online" when I'm out of home is a bit annoying.
Now you say... can you please tell me -in user's parlance terms so I can
understand it- what's the difference between "brcmsmac" from the stock
kernel modules and this one from "compat-drivers"? I mean, what's the
difference from a user's POV and from kernel hacker POV. Are they
related? Are they going to me merged somehow? What's the goal of having
both?
> >> Can you
> >>
> >> git clone git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-testing.git
> >>
> >> and use your .config, and build and test that kernel ? If it fails to
> >> boot / crash then you know its the driver.
> >
> > I'm not using git (don't even have it installed in my system), I'm
> > used to compile kernels/modules in the old fashion way. Sorry, I don't
> > know how git works.
>
> make localmodconfig
> make
> make install
(...)
Will try and report back. Thanks, Luis.
Greetings,
--
Camaleón