Re: [PATCH] w35und: fix usb_control_msg() error handling in wb35_probe()

Previous thread: port bound SAs by Paul Moore on Monday, January 26, 2009 - 3:21 pm. (31 messages)

Next thread: [PATCH] net: add tracepoints to socket api by Neil Horman on Monday, January 26, 2009 - 3:59 pm. (15 messages)
To: Sandro Bonazzola <sandro.bonazzola@...>
Cc: Pavel Machek <pavel@...>, Greg Kroah-Hartman <greg@...>, <linux-kernel@...>, Linux Netdev List <netdev@...>
Date: Monday, January 26, 2009 - 3:40 pm

Hi Sandro,

On Mon, Jan 26, 2009 at 9:31 PM, Sandro Bonazzola

So the driver doesn't work in 2.6.29-rc1 but does in 2.6.28.1? I
wonder what changed in the networking/wireless stack that broke the
winbond driver.... I'm cc'ing netdev even though this is a driver in
staging.

Pekka
--

To: Sandro Bonazzola <sandro.bonazzola@...>
Cc: Pavel Machek <pavel@...>, Greg Kroah-Hartman <greg@...>, <linux-kernel@...>, Linux Netdev List <netdev@...>
Date: Monday, January 26, 2009 - 3:43 pm

Actually, it's probably me who broke the driver. I think Greg merged
most of my patches after 2.6.28 came out. Sandro, if you have some
time, please consider doing git bisect to find out which changeset
causes the problem.

Pekka
--

To: Pekka Enberg <penberg@...>
Cc: Pavel Machek <pavel@...>, Greg Kroah-Hartman <greg@...>, <linux-kernel@...>, Linux Netdev List <netdev@...>
Date: Monday, January 26, 2009 - 4:13 pm

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Well, I don't know if this can be somehow related to a bug in the driver, but on
my system, after I've unplugged the USB trasnsitter, ksoftirqd/0 begin to use
89% of my CPU time (top). I suspect that there are some issue that should be

Well, actually my 2.6.29 copy seems to be broken:
arch/x86/kernel/amd_iommu.o: In function `debugfs_create_size_t':
amd_iommu.c:(.text+0x0): multiple definition of `debugfs_create_size_t'
arch/x86/kernel/kdebugfs.o:kdebugfs.c:(.text+0x0): first defined here
make[1]: *** [arch/x86/kernel/built-in.o] Error 1
make: *** [arch/x86/kernel] Error 2

# git pull
Already up-to-date.

# git status
# On branch master
# Changed but not updated:
# (use "git add <file>..." to update what will be committed)
# (use "git checkout -- <file>..." to discard changes in working directory)
#
# modified: drivers/staging/winbond/wbusb.c
#
no changes added to commit (use "git add" and/or "git commit -a")

I don't know too much how to use git to do what you're asking. I've some years
of experience with svn and cvs but git is quite a new tool for me.
I'll wait a git kernel that compile while I'll try to find out how to do what
you're asking.
I have not so much free time to spend in a checkout / build / test if the test
phase need a kernel reinstall / system reboot. If it's only the driver module I
think I can try to test it for each changeset in a couple of weeks /starting
from the day I know how to bisect :-) ).

- --
Sandro
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkl+GWIACgkQV871CybFezlRXACfQTBrMm2hL9b+o5oU1/6/xZ+f
gaYAnjyyawOFPBPIe/scq5pYupNrRLa3
=d1SV
-----END PGP SIGNATURE-----
--

To: Sandro Bonazzola <sandro.bonazzola@...>
Cc: Pavel Machek <pavel@...>, Greg Kroah-Hartman <greg@...>, <linux-kernel@...>, Linux Netdev List <netdev@...>
Date: Monday, January 26, 2009 - 4:22 pm

Here's a tutorial on git bisect.

Assuming 2.6.29-rc1 is broken, you initialize the bisection with:

git bisect start -- drivers/staging/winbond/
git bisect good v2.6.28
git bisect bad v2.6.29-rc1

Then you continue doing this:

<apply my patch, compile, test>

and

git bisect good

or

git bisect bad

depending on whether the kernel worked or not. Repeat until git bisect
tells you which changeset broke the kernel. The process shouldn't take
more than 5 compiles/reboots.

Pekka
--

Previous thread: port bound SAs by Paul Moore on Monday, January 26, 2009 - 3:21 pm. (31 messages)

Next thread: [PATCH] net: add tracepoints to socket api by Neil Horman on Monday, January 26, 2009 - 3:59 pm. (15 messages)