Re: [RFC] Future of INP3 patch

Re: [RFC] Future of INP3 patch

[Matt VK2RQ]

> Hi, 
> 
> I haven't seen any replies after posting about the inp3 patches for 2.6 so 
> either the spontanious combustion problems are fixed or nobody cares..... 
> At the moment the patch seems to be pretty stable and running without 
> problems so I was wondering what to do next: 
> 
> - Keep making new patches for the 2.6 kernels and try to get it in in 2.7 
>   (Should be very safe, although less testing) 
> 
> - Try to get it in 2.6. 
>   (The patch only affects netrom and is for 99% configurable with 
> CONFIG_NETROM_INP with the exeption of the netrom node sorting but this 
> shouldn't be a problem) 
> 
> Any comments? 

Hi Jeroen,

I just tried installing version 007 of your INP3 patch to kernel 2.6.32. There have
been some changes to the kernel since 2.6.4, so I needed to tweak it a bit to
get it to compile:
http://qsl.net/vk2rq/2.6.32-inp3_007.patch

However, there are some deadlock problems arising from this patch. Specifically,
when inp3_nodes_neg() is called, neigh_list will already be locked, and if
it was called from inp3_route_neg(), then node_list will be locked as well.

However, inp3_nodes_neg() calls nr_del_node_found(), which assumes
there are no locks in place. If nr_del_node_found() places a call to
nr_remove_neigh() or nr_remove_node(), then those functions will
try to acquire a lock on neigh_list or node_list, which will lead to a
deadlock situation if those locks are already in place.

This thread is already quite old, so not sure if you or anyone else is
still maintaining this code?

73, Matt VK2RQ

Re: [RFC] Future of INP3 patch

[Jeroen Vreeken]

Hi Matt,

On 12/16/2012 12:08 AM, Matt VK2RQ wrote:
>> Hi,
>>
>> I haven't seen any replies after posting about the inp3 patches for 2.6 so
>> either the spontanious combustion problems are fixed or nobody cares.....
>> At the moment the patch seems to be pretty stable and running without
>> problems so I was wondering what to do next:
>>
>> - Keep making new patches for the 2.6 kernels and try to get it in in 2.7
>>    (Should be very safe, although less testing)
>>
>> - Try to get it in 2.6.
>>    (The patch only affects netrom and is for 99% configurable with
>> CONFIG_NETROM_INP with the exeption of the netrom node sorting but this
>> shouldn't be a problem)
>>
>> Any comments?
> Hi Jeroen,
>
> I just tried installing version 007 of your INP3 patch to kernel 2.6.32. There have
> been some changes to the kernel since 2.6.4, so I needed to tweak it a bit to
> get it to compile:
> http://qsl.net/vk2rq/2.6.32-inp3_007.patch
>
> However, there are some deadlock problems arising from this patch. Specifically,
> when inp3_nodes_neg() is called, neigh_list will already be locked, and if
> it was called from inp3_route_neg(), then node_list will be locked as well.
>
> However, inp3_nodes_neg() calls nr_del_node_found(), which assumes
> there are no locks in place. If nr_del_node_found() places a call to
> nr_remove_neigh() or nr_remove_node(), then those functions will
> try to acquire a lock on neigh_list or node_list, which will lead to a
> deadlock situation if those locks are already in place.
>
> This thread is already quite old, so not sure if you or anyone else is
> still maintaining this code?
I have not been actively maintaining the code. However I am currently 
re-building the local packet access point in Eindhoven, and I was 
planning on updating the patch to a more recent kernel. (probably 3.7 based)
I haven't got around to this since I am still fighting with scc cards... 
but I hope to get to the netrom part soon.

73,
Jeroen PE1RXQ

[RFC] Future of INP3 patch

[Jeroen Vreeken]

Hi,

I haven't seen any replies after posting about the inp3 patches for 2.6 so
either the spontanious combustion problems are fixed or nobody cares.....
At the moment the patch seems to be pretty stable and running without
problems so I was wondering what to do next:

- Keep making new patches for the 2.6 kernels and try to get it in in 2.7
  (Should be very safe, although less testing)

- Try to get it in 2.6.
  (The patch only affects netrom and is for 99% configurable with
CONFIG_NETROM_INP with the exeption of the netrom node sorting but this
shouldn't be a problem)


Any comments?
Jeroen

Re: [RFC] Future of INP3 patch

[Wilbert Knol]

> I haven't seen any replies after posting about the inp3 patches for 2.6 so
> either the spontanious combustion problems are fixed or nobody cares.....
> At the moment the patch seems to be pretty stable and running without
> problems so I was wondering what to do next:

It is not a case of not caring. I haven't the guts to try the 2.6
kernels...

My experience upgrading 2.2 to  2.4 wasn't a happy one. I rely on a
working AX25 system for my E-mail, and the current 2.4.22 kernel
(without NETROM) is reasonably stable.

I suspect several list subscribers (those running full-service DX
Clusters, BBS's, gateways etc etc) are in the same boat.

However, if the 2.6.4 kernel with Jeroen's inp3 patches applied is OK
for AX25, then I am quite prepared to give it a try. What is the
verdict?


Wilbert, ZL2BSJ

Re: [RFC] Future of INP3 patch

[Ralf Baechle DL5RB]

On Wed, Mar 31, 2004 at 12:27:33AM +1200, Wilbert Knol wrote:

> It is not a case of not caring. I haven't the guts to try the 2.6
> kernels...
> 
> My experience upgrading 2.2 to  2.4 wasn't a happy one. I rely on a
> working AX25 system for my E-mail, and the current 2.4.22 kernel
> (without NETROM) is reasonably stable.

Recall for how long nobody at all was looking after that code - it's quite
amazing 2.4 was working as well as it does.  Here we're trying to balance
the risk of including the patch with the advantages it's going to deliver.

> I suspect several list subscribers (those running full-service DX
> Clusters, BBS's, gateways etc etc) are in the same boat.
> 
> However, if the 2.6.4 kernel with Jeroen's inp3 patches applied is OK
> for AX25, then I am quite prepared to give it a try. What is the
> verdict?

The potencial of Jeroen's patch to disturb netrom systems when enabled is
fairly low; it's zero if CONFIG_NETROM is disabled or the module not loaded.
It'll need minor changes to make it bulletproof against optimizatios of
modern compiler (the extern vs. static inline thing) but with that applied
I'm very inclined to apply it.

73 de DL5RB op Ralf

--
Loc. JN47BS / CQ 14 / ITU 28 / DOK A21

Re: [RFC] Future of INP3 patch

[Øyvind Hanssen]

I havent dared to switch to the 2.6 kernel yet, use the INP3 patch for the 2.4 
kernel. We really appreciate it (and depend on it) here in my area, so it would be nice 
if it was included in the mainstream kernel, of course. 

73 de LA7ECA @LA3T

> - Try to get it in 2.6.
>   (The patch only affects netrom and is for 99% configurable with
> CONFIG_NETROM_INP with the exeption of the netrom node sorting but this
> shouldn't be a problem)
> 
-- 
Computer Science dept. University of Tromsø,  ohanssen@acm.org,  Cell: +47 9511 7457
-
To unsubscribe from this list: send the line "unsubscribe linux-hams" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [RFC] Future of INP3 patch

[Ralf Baechle DL5RB]

On Tue, Mar 30, 2004 at 03:35:35PM +0200, ?yvind Hanssen wrote:

> I havent dared to switch to the 2.6 kernel yet, use the INP3 patch for
> the 2.4 kernel. We really appreciate it (and depend on it) here in my
> area, so it would be nice if it was included in the mainstream kernel,
> of course. 

The switch to 2.6 was moderatly rough - but the roughness was entirely
outside the area of amateur radio and should be handled by Linux
distributions.  I hope that's going to convince some people to try their
luck.

73 de DL5RB op Ralf

--
Loc. JN47BS / CQ 14 / ITU 28 / DOK A21