[...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1

[...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1

[Miroslav Skoric]

Hi,

One of my two machines in the LAN goes into 'endless' repeating of the 
following lines when I want to shut it down. (Ok, not really endless but 
it does take some half an hour.) That comp runs an older version of ax25 
subsystem and on top of it runs FPAC node and FBB mailbox. The issue 
happens whenever the *other* machine has some netrom activity such as a 
BPQ node/bbs or like.

[...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
[...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
[...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
...
...


after some 30 minutes there appear two other lines:


[TIME] Timed out starting Power-Off.
[ !! ] Forcibly powering off: job timed out


and it resumes repeating the above for a while:


[...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
[...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1


and it finally powers off after a minute or so.

It makes me wonder why it behaves like that. I tested that behaviour 
thoroughly (by having the other machine running, or not running some 
packet node stuff) and realized that it shuts down properly, without any 
delay only when the other machine is not having any packet node active. 
(The other comp can be switched on, but without any node/bbs running.)

Any idea?

Misko YT7MPB

Re: [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1

[Lars Kellogg-Stedman]

On Sun, Apr 28, 2024 at 05:57:30PM +0200, Miroslav Skoric wrote:
> subsystem and on top of it runs FPAC node and FBB mailbox. The issue happens
> whenever the *other* machine has some netrom activity such as a BPQ node/bbs
> or like.
> 
> [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
> [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
> [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1

That sounds remarkably like the problem I was asking about a few weeks
ago; that threads start with
https://marc.info/?l=linux-hams&m=171271178900331&w=2. I'm curious if
you see any of the kernel traces I reported in your kernel log? That
would help identify the problem.

I just posted a patch that resolves it for me [1], but that patch was
against the current development kernel (6.9.0rc5).

[1]: https://marc.info/?l=linux-hams&m=171417563600765&w=2

-- 
Lars Kellogg-Stedman <lars@oddbit.com> | larsks @ {irc,twitter,github}
http://blog.oddbit.com/                | N1LKS

Re: [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1

[David Ranch]

Hello Misko,

This is a kernel bug and has been discussed in the list (see the 
archives): http://vger.kernel.org/vger-lists.html#linux-hams

    I recently responded with: 
https://www.spinics.net/lists/linux-hams/msg05486.html

it seems to me that different kernel versions can hit this issue in 
different ways but I can confirm that the Ubuntu 20.04 distro's 5.15.0 
kernel hits this issue.  I'm also starting to believe that if 
Canonical's kernel "LivePatch" feature is enabled and you have stale 
AX.25 connections, the kernel starts to have issues and maybe after 
?5-7? days of decay, the machine will eventually hard lock up.  You 
won't see the panic output if you're running X as the GUI displays hides 
the output.  Btw, it's also becoming painfully obvious that different 
Linux distro's version of "their" kernel might or might not be taking on 
some of these new kernel "fixes" that break or fix parts of the AX.25 
stack.  It's very clear to me, these patches are coming in w/o any real 
testing and as long as it passes a visual review and it compiles, it's 
committed into the tree.

I've since moved from u20.04 over to a Raspberry Pi OS Bookworm 6.6.20 
kernel setup for ~2wks.  So far I haven't seen any similar issues 
(except I found a new issue with the axparms tool now corrupting the 
callsign+ssid stack).

--David
KI6ZHD







Re: unregister_netdevice: waiting for ax0 to become free


On 04/28/2024 08:57 AM, Miroslav Skoric wrote:
> Hi,
>
> One of my two machines in the LAN goes into 'endless' repeating of the 
> following lines when I want to shut it down. (Ok, not really endless 
> but it does take some half an hour.) That comp runs an older version 
> of ax25 subsystem and on top of it runs FPAC node and FBB mailbox. The 
> issue happens whenever the *other* machine has some netrom activity 
> such as a BPQ node/bbs or like.
>
> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
> count = -1
> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
> count = -1
> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
> count = -1
> ...
> ...
>
>
> after some 30 minutes there appear two other lines:
>
>
> [TIME] Timed out starting Power-Off.
> [ !! ] Forcibly powering off: job timed out
>
>
> and it resumes repeating the above for a while:
>
>
> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
> count = -1
> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
> count = -1
>
>
> and it finally powers off after a minute or so.
>
> It makes me wonder why it behaves like that. I tested that behaviour 
> thoroughly (by having the other machine running, or not running some 
> packet node stuff) and realized that it shuts down properly, without 
> any delay only when the other machine is not having any packet node 
> active. (The other comp can be switched on, but without any node/bbs 
> running.)
>
> Any idea?
>
> Misko YT7MPB
>

Re: [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1

[Miroslav Skoric]

Thank you, David. I'll check the link.

The machine experiencing the bug is Ubuntu 18.04 (i386 version that 
cannot be upgraded to any newer one.) And the kernel is an old one 
4.15.0-212-generic #223-Ubuntu SMP.

I suppose that any change will not be possible any time soon because 
that machine is not on the Internet, however serves me well for other 
purposes.

I just wanted to check if the issue was my fault or not :-)

Misko YT7MPB



On 4/28/24 10:03 PM, David Ranch wrote:
> 
> Hello Misko,
> 
> This is a kernel bug and has been discussed in the list (see the 
> archives): http://vger.kernel.org/vger-lists.html#linux-hams
> 
>     I recently responded with: 
> https://www.spinics.net/lists/linux-hams/msg05486.html
> 
> it seems to me that different kernel versions can hit this issue in 
> different ways but I can confirm that the Ubuntu 20.04 distro's 5.15.0 
> kernel hits this issue.  I'm also starting to believe that if 
> Canonical's kernel "LivePatch" feature is enabled and you have stale 
> AX.25 connections, the kernel starts to have issues and maybe after 
> ?5-7? days of decay, the machine will eventually hard lock up.  You 
> won't see the panic output if you're running X as the GUI displays hides 
> the output.  Btw, it's also becoming painfully obvious that different 
> Linux distro's version of "their" kernel might or might not be taking on 
> some of these new kernel "fixes" that break or fix parts of the AX.25 
> stack.  It's very clear to me, these patches are coming in w/o any real 
> testing and as long as it passes a visual review and it compiles, it's 
> committed into the tree.
> 
> I've since moved from u20.04 over to a Raspberry Pi OS Bookworm 6.6.20 
> kernel setup for ~2wks.  So far I haven't seen any similar issues 
> (except I found a new issue with the axparms tool now corrupting the 
> callsign+ssid stack).
> 
> --David
> KI6ZHD
> 
> 
> 
> 
> 
> 
> 
> Re: unregister_netdevice: waiting for ax0 to become free
> 
> 
> On 04/28/2024 08:57 AM, Miroslav Skoric wrote:
>> Hi,
>>
>> One of my two machines in the LAN goes into 'endless' repeating of the 
>> following lines when I want to shut it down. (Ok, not really endless 
>> but it does take some half an hour.) That comp runs an older version 
>> of ax25 subsystem and on top of it runs FPAC node and FBB mailbox. The 
>> issue happens whenever the *other* machine has some netrom activity 
>> such as a BPQ node/bbs or like.
>>
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
>> count = -1
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
>> count = -1
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
>> count = -1
>> ...
>> ...
>>
>>
>> after some 30 minutes there appear two other lines:
>>
>>
>> [TIME] Timed out starting Power-Off.
>> [ !! ] Forcibly powering off: job timed out
>>
>>
>> and it resumes repeating the above for a while:
>>
>>
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
>> count = -1
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage 
>> count = -1
>>
>>
>> and it finally powers off after a minute or so.
>>
>> It makes me wonder why it behaves like that. I tested that behaviour 
>> thoroughly (by having the other machine running, or not running some 
>> packet node stuff) and realized that it shuts down properly, without 
>> any delay only when the other machine is not having any packet node 
>> active. (The other comp can be switched on, but without any node/bbs 
>> running.)
>>
>> Any idea?
>>
>> Misko YT7MPB
>>
> 
> 
> 

Re: [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1

[Miroslav Skoric]

Thank you, Lars. I'll check your sites for details. However, my kernel 
is rather old so I am unsure whether your approach would work kere.

Misko YT7MPB


On 4/28/24 9:57 PM, Lars Kellogg-Stedman wrote:
> On Sun, Apr 28, 2024 at 05:57:30PM +0200, Miroslav Skoric wrote:
>> subsystem and on top of it runs FPAC node and FBB mailbox. The issue happens
>> whenever the *other* machine has some netrom activity such as a BPQ node/bbs
>> or like.
>>
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
>> [...] unregister_netdevice: waiting for ax0 to become free. Usage count = -1
> 
> That sounds remarkably like the problem I was asking about a few weeks
> ago; that threads start with
> https://marc.info/?l=linux-hams&m=171271178900331&w=2. I'm curious if
> you see any of the kernel traces I reported in your kernel log? That
> would help identify the problem.
> 
> I just posted a patch that resolves it for me [1], but that patch was
> against the current development kernel (6.9.0rc5).
> 
> [1]: https://marc.info/?l=linux-hams&m=171417563600765&w=2
>