netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* iproute2 regression -- ss -u returns an empty list
@ 2015-04-28  1:22 Mihai Moldovan
       [not found] ` <20150428040516.GA14129@angus-think.lan>
  2015-04-28 11:21 ` Vadim Kochan
  0 siblings, 2 replies; 8+ messages in thread
From: Mihai Moldovan @ 2015-04-28  1:22 UTC (permalink / raw)
  To: netdev

[-- Attachment #1: Type: text/plain, Size: 632 bytes --]

Hi,

Following up a bug report I received at
http://bugs.x2go.org/cgi-bin/bugreport.cgi?bug=799 it looks like the current
version iproute2's ss utility always returns an empty result set whenever
specifying the -u flag.

The problem has been bisected to 9db7bf15e22b6a1b8bc09c4a1e29571cbca55c94 and is
also affecting version 3.19.0, but most notably 4.0.0+.

I'm not quite sure that's the correct commit, but it does definitely look related.

Even ss -xu, which previously returned a list of UNIX and UDP sockets now
returns nothing on my systems.


Any help with that matter is greatly appreciated.



Mihai


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: iproute2 regression -- ss -u returns an empty list
       [not found] ` <20150428040516.GA14129@angus-think.lan>
@ 2015-04-28  4:33   ` Mihai Moldovan
  2015-04-28  4:47     ` Vadim Kochan
  0 siblings, 1 reply; 8+ messages in thread
From: Mihai Moldovan @ 2015-04-28  4:33 UTC (permalink / raw)
  To: Vadim Kochan; +Cc: netdev

[-- Attachment #1: Type: text/plain, Size: 660 bytes --]

On 28.04.2015 06:05 AM, Vadim Kochan wrote:> So by default ss prints socket with
CONNECTED state and even UDP can be
> in the CONNECTED state on the Linux,

UDP can be in a CONNECTED state... ooookay. That's probably a conntrack thing.


> so you can specify 'ss -ua' (may be
> some explanation should be added to the ss man page) which
> should print UDP sockets in the any state, AFAIK it was the default
> behaviour before my changes.

Could this please be reverted? It's breaking other stuff that uses ss.

Also: ss -lxua returns an empty list, too. Previously, the families seem to have
been OR'd, while they are now ANDed?



Mihai


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: iproute2 regression -- ss -u returns an empty list
  2015-04-28  4:33   ` Mihai Moldovan
@ 2015-04-28  4:47     ` Vadim Kochan
  2015-04-28  5:07       ` Mihai Moldovan
  0 siblings, 1 reply; 8+ messages in thread
From: Vadim Kochan @ 2015-04-28  4:47 UTC (permalink / raw)
  To: Mihai Moldovan; +Cc: Vadim Kochan, netdev

On Tue, Apr 28, 2015 at 06:33:18AM +0200, Mihai Moldovan wrote:
> On 28.04.2015 06:05 AM, Vadim Kochan wrote:> So by default ss prints socket with
> CONNECTED state and even UDP can be
> > in the CONNECTED state on the Linux,
> 
> UDP can be in a CONNECTED state... ooookay. That's probably a conntrack thing.
> 
> 
> > so you can specify 'ss -ua' (may be
> > some explanation should be added to the ss man page) which
> > should print UDP sockets in the any state, AFAIK it was the default
> > behaviour before my changes.
> 
> Could this please be reverted? It's breaking other stuff that uses ss.
> 
> Also: ss -lxua returns an empty list, too. Previously, the families seem to have
> been OR'd, while they are now ANDed?
> 
> 
> 
> Mihai
> 

I will try to fix this ...

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: iproute2 regression -- ss -u returns an empty list
  2015-04-28  4:47     ` Vadim Kochan
@ 2015-04-28  5:07       ` Mihai Moldovan
  2015-04-28  5:11         ` Vadim Kochan
  0 siblings, 1 reply; 8+ messages in thread
From: Mihai Moldovan @ 2015-04-28  5:07 UTC (permalink / raw)
  To: Vadim Kochan; +Cc: netdev, Bernard Cafarelli

[-- Attachment #1: Type: text/plain, Size: 1563 bytes --]

On 28.04.2015 06:47 AM, Vadim Kochan wrote:
> On Tue, Apr 28, 2015 at 06:33:18AM +0200, Mihai Moldovan wrote:
>> On 28.04.2015 06:05 AM, Vadim Kochan wrote:> So by default ss prints
>> socket with CONNECTED state and even UDP can be
>>> in the CONNECTED state on the Linux,
>> 
>> UDP can be in a CONNECTED state... ooookay. That's probably a conntrack 
>> thing.
>> 
>> 
>>> so you can specify 'ss -ua' (may be some explanation should be added to 
>>> the ss man page) which should print UDP sockets in the any state, AFAIK 
>>> it was the default behaviour before my changes.
>> 
>> Could this please be reverted? It's breaking other stuff that uses ss.
>> 
>> Also: ss -lxua returns an empty list, too. Previously, the families seem
>> to have been OR'd, while they are now ANDed?
>> 
>> 
>> 
>> Mihai
>> 
> 
> I will try to fix this ...

Thanks!

To be a bit more verbose about the rationale:

Changing the default behavior is generally a bad idea, unless strictly necessary
(e.g., if the previous default behavior was buggy anyway), because it may break
other software. Please try introducing behavioral changes via new options.

That way, dependencies will still work fine and anyone who wants to use the new
behavior can explicitly do so via a switch.

I could work around this by checking ss' version, but that turns out to be a
pain, too, because ss -V reports something like this:

"ss utility, iproute2-ss150413"

Not exactly a "friendly" version like "3.19.0" or "4.0.0" to check against...



Mihai



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: iproute2 regression -- ss -u returns an empty list
  2015-04-28  5:07       ` Mihai Moldovan
@ 2015-04-28  5:11         ` Vadim Kochan
  0 siblings, 0 replies; 8+ messages in thread
From: Vadim Kochan @ 2015-04-28  5:11 UTC (permalink / raw)
  To: Mihai Moldovan; +Cc: Vadim Kochan, netdev, Bernard Cafarelli

On Tue, Apr 28, 2015 at 07:07:56AM +0200, Mihai Moldovan wrote:
> On 28.04.2015 06:47 AM, Vadim Kochan wrote:
> > On Tue, Apr 28, 2015 at 06:33:18AM +0200, Mihai Moldovan wrote:
> >> On 28.04.2015 06:05 AM, Vadim Kochan wrote:> So by default ss prints
> >> socket with CONNECTED state and even UDP can be
> >>> in the CONNECTED state on the Linux,
> >> 
> >> UDP can be in a CONNECTED state... ooookay. That's probably a conntrack 
> >> thing.
> >> 
> >> 
> >>> so you can specify 'ss -ua' (may be some explanation should be added to 
> >>> the ss man page) which should print UDP sockets in the any state, AFAIK 
> >>> it was the default behaviour before my changes.
> >> 
> >> Could this please be reverted? It's breaking other stuff that uses ss.
> >> 
> >> Also: ss -lxua returns an empty list, too. Previously, the families seem
> >> to have been OR'd, while they are now ANDed?
> >> 
> >> 
> >> 
> >> Mihai
> >> 
> > 
> > I will try to fix this ...
> 
> Thanks!
> 
> To be a bit more verbose about the rationale:
> 
> Changing the default behavior is generally a bad idea, unless strictly necessary
> (e.g., if the previous default behavior was buggy anyway), because it may break

Yes the previous behaviour was buggy ... And that was the reason.

> other software. Please try introducing behavioral changes via new options.
> 
> That way, dependencies will still work fine and anyone who wants to use the new
> behavior can explicitly do so via a switch.
> 
> I could work around this by checking ss' version, but that turns out to be a
> pain, too, because ss -V reports something like this:
> 
> "ss utility, iproute2-ss150413"
> 
> Not exactly a "friendly" version like "3.19.0" or "4.0.0" to check against...
> 
> 
> 
> Mihai
> 
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: iproute2 regression -- ss -u returns an empty list
  2015-04-28  1:22 iproute2 regression -- ss -u returns an empty list Mihai Moldovan
       [not found] ` <20150428040516.GA14129@angus-think.lan>
@ 2015-04-28 11:21 ` Vadim Kochan
  2015-04-28 13:14   ` Mihai Moldovan
  1 sibling, 1 reply; 8+ messages in thread
From: Vadim Kochan @ 2015-04-28 11:21 UTC (permalink / raw)
  To: Mihai Moldovan; +Cc: netdev

On Tue, Apr 28, 2015 at 03:22:33AM +0200, Mihai Moldovan wrote:
> Hi,
> 
> Following up a bug report I received at
> http://bugs.x2go.org/cgi-bin/bugreport.cgi?bug=799 it looks like the current
> version iproute2's ss utility always returns an empty result set whenever
> specifying the -u flag.
> 
> The problem has been bisected to 9db7bf15e22b6a1b8bc09c4a1e29571cbca55c94 and is
> also affecting version 3.19.0, but most notably 4.0.0+.
> 
> I'm not quite sure that's the correct commit, but it does definitely look related.
> 
> Even ss -xu, which previously returned a list of UNIX and UDP sockets now
> returns nothing on my systems.
> 
> 
> Any help with that matter is greatly appreciated.
> 
> 
> 
> Mihai
> 


Hi Mihai,

Seems I have a fixed version, I tested it localy with different
option combinations, and now there should be no issues when specifying
different families.

Would you be able to test it ? How I can provide a fixed version to you -
by patch or I can push to my github and give you a link ?

Thanks,

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: iproute2 regression -- ss -u returns an empty list
  2015-04-28 11:21 ` Vadim Kochan
@ 2015-04-28 13:14   ` Mihai Moldovan
  2015-04-29  9:46     ` Vadim Kochan
  0 siblings, 1 reply; 8+ messages in thread
From: Mihai Moldovan @ 2015-04-28 13:14 UTC (permalink / raw)
  To: Vadim Kochan; +Cc: netdev

[-- Attachment #1: Type: text/plain, Size: 489 bytes --]

Hi Vadim,


On 28.04.2015 01:21 PM, Vadim Kochan wrote:
> Seems I have a fixed version, I tested it localy with different
> option combinations, and now there should be no issues when specifying
> different families.

Thanks for the quick work! I'm happy to test.


> Would you be able to test it ? How I can provide a fixed version to you -
> by patch or I can push to my github and give you a link ?

I don't mind, any way is fine. I can work either way.



Mihai



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: iproute2 regression -- ss -u returns an empty list
  2015-04-28 13:14   ` Mihai Moldovan
@ 2015-04-29  9:46     ` Vadim Kochan
  0 siblings, 0 replies; 8+ messages in thread
From: Vadim Kochan @ 2015-04-29  9:46 UTC (permalink / raw)
  To: Mihai Moldovan; +Cc: Vadim Kochan, netdev, stephen

On Tue, Apr 28, 2015 at 03:14:31PM +0200, Mihai Moldovan wrote:
> Hi Vadim,
> 
> 
> On 28.04.2015 01:21 PM, Vadim Kochan wrote:
> > Seems I have a fixed version, I tested it localy with different
> > option combinations, and now there should be no issues when specifying
> > different families.
> 
> Thanks for the quick work! I'm happy to test.
> 
> 
> > Would you be able to test it ? How I can provide a fixed version to you -
> > by patch or I can push to my github and give you a link ?
> 
> I don't mind, any way is fine. I can work either way.
> 
> 
> 
> Mihai
> 
> 

Hi Stephen,

I just saw that Linux 4.0.1 was released so I assume that you will
prepare the v4.0.1 tag for iproute2 too, but would you please wait until
this bug will be fixed ? I already have a working patch (from my point
of view it works better than before), so I'd like to wait when Mihai
will test the patch which I have sent directly to him.

Thanks,
Vadim

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2015-04-29  9:46 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-04-28  1:22 iproute2 regression -- ss -u returns an empty list Mihai Moldovan
     [not found] ` <20150428040516.GA14129@angus-think.lan>
2015-04-28  4:33   ` Mihai Moldovan
2015-04-28  4:47     ` Vadim Kochan
2015-04-28  5:07       ` Mihai Moldovan
2015-04-28  5:11         ` Vadim Kochan
2015-04-28 11:21 ` Vadim Kochan
2015-04-28 13:14   ` Mihai Moldovan
2015-04-29  9:46     ` Vadim Kochan

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).