[Letux-kernel] [PATCH RFC] net: hso: register netdev later to avoid a race condition

H. Nikolaus Schaller hns at goldelico.com
Wed Apr 26 08:53:18 CEST 2017


> Am 26.04.2017 um 07:27 schrieb Andreas Kemnade <andreas at kemnade.info>:
> 
> Hi,
> 
> On Wed, 26 Apr 2017 06:50:00 +0200
> "H. Nikolaus Schaller" <hns at goldelico.com> wrote:
> 
>> Hi,
>> 
>>> Am 25.04.2017 um 23:11 schrieb Andreas Kemnade <andreas at kemnade.info>:
>>> 
>>> On Tue, 25 Apr 2017 15:29:53 +0200
>>> "H. Nikolaus Schaller" <hns at goldelico.com> wrote:
>>> 
>>>> 
>>>> So my conclusion is:
>>>> 
>>>> * dhcpd is recognising quite quickly that there is a new interface being enabled.
>>>> * and tries to do some ioctl which makes the kernel segfault
>>>> * and crashes dhcpd
>>>> 
>>>> The fix we are talking about makes this being ignored.
>>>> 
>>>> It might simply be that dhcpd is responding faster with an ioctl() than with older kernels.
>>>> So let's keep the patch for the moment.
>>>> 
>>> well, i would like to replace it with something upstreamable. But maybe
>>> it is best if you first send the 20ms-hack as a [PATCH RFC] together
>>> with the backtrace so that we have more brains on that topic.
>>> 
>>> How was dhcpcd started? Just so that I can properly reproduce that.
>> 
>> Just apt-get install isc-dhcp-client on Jessie.
>> This installs /etc/init.d/dhcpcd startup code and defaults.
>> 
>> No special rule(s) for hso.
>> 
> isc-dhcp-client is dhclient stuff.
> 
> The package name is dhcpcd5. I did not know that it does so much
> stuff automatically.

Well, there are always two contradicting trends in the community: monolithic vs. modular
* initd was modular, systemd is more monolithic
* lxde vs. enlightenment
* ...

Both have their good and bad sides. So there is no reason to make an ideology out of it.

> Starting wpa_supplicant without my permission.

Oops.

> That either needs to be tamed

there is certainly some config file.

> or it deserves a place on my personal
> packet banlist.
> 
> But nevertheless I managed to reproduce the kernel oops.

Ok, that is the most important thing for kernel debugging.

BR,
Nikolaus

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.goldelico.com/pipermail/letux-kernel/attachments/20170426/8e25a2b0/attachment.asc>


More information about the Letux-kernel mailing list