[Cerowrt-devel] A stable cerowrt release in sight

[Cerowrt-devel] A stable cerowrt release in sight

[Dave Taht]

it is now mid-may, about 3 months after I'd hoped to finally have a
stable release.

and I've spent about 8 months longer on this than I really could
afford. I'm glad that we seem to have a stable release of dnsmasq, the
sqm system is stable, and
aside from bug 442 I haven't hard crashed a box in months and months.

I considered bug 442 to be a showstopper and have spent most of the
last months trying to isolate it. There's some cool code for blowing
up queues in a repo of mine in particular.

So as for fixing 442...

I give up.

Disabling 802.11e entirely appears to fix it in some limited testing,
and as 802.11e is both broken conceptually and operationally, I don't
feel like trying to fix it anymore, I just want a stable release and
to move on to doing something - anything! else.

So hopefully the upcoming 3.10.40-4 build will be stable enough to
freeze on. If there is anything else that people want done before
freezing, let me know soonest.

In other news, availablilty of the 3800 series has dried up almost completely:

http://www.amazon.com/Netgear-WNDR3800-Premium-Gigabit-Wireless/dp/B0058NN6CS/ref=sr_1_1?s=electronics&ie=UTF8&qid=1400221440&sr=1-1&keywords=wndr3800

I have been evaluating multiple pieces of alternate hardware for years
now. In terms of a "continue with cerowrt-with-minimal-changes" play,
the

1) "tp-link archer c7 v2 ac1750" (the v2 part is critical) - uses a
mildly more powerful processor than the wndr3800, has the same amount
of flash and memory, 2 usb ports, and comes with an ath10k chip that
theoretically does 802.11ac with a proprietary firmware blob (but has
an ath9k for 2.4ghz operation).  The performance of the ath10k sucks
rocks presently. It costs 99 bucks at frys.

I have been testing using one for a week now (with straight openwrt) and it's
pretty stable. It also doesn't live up to its billing - it can't
forward packets faster than 350mbits, only slightly better than the
wndr3800, and I figure it won't come close to 802.11ac speeds either.

In terms of finding something faster...

2) Am still waiting on the belkin wrt product to show up

3) am mostly hacking on x86 gear to get the speed I wanted to have

I just want a stable release, to sip boat drinks for a while, and try to come up
with a strategy to move forward from here that makes sense.

I'm open to ideas as to how to spend the next year more productively.

In other, other news, the ubnt edgerouter fq_codel test is going fairly well.


-- 
Dave Täht

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Török Edwin]

On 05/16/2014 09:36 AM, Dave Taht wrote:
> it is now mid-may, about 3 months after I'd hoped to finally have a
> stable release.
[...]
> 
> So hopefully the upcoming 3.10.40-4 build will be stable enough to
> freeze on. If there is anything else that people want done before
> freezing, let me know soonest.
> 

Excellent news!
So if I want to upgrade my router this weekend is 3.10.40-4 
ready for testing yet, or should I wait a bit more?

Best regards,
--Edwin

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Dave Taht]

I think it finished compiling at 1AM, and I have been busy all day. Might work.

I also finished an archer test compile, which of course, failed to
boot entirely. I am taking saturday off... I will try to flash
3.10.40-4 on a 3800 before this evening, otherwise sunday.

On Fri, May 16, 2014 at 12:59 PM, Török Edwin
<edwin+ml-cerowrt@etorok.net> wrote:
> On 05/16/2014 09:36 AM, Dave Taht wrote:
>> it is now mid-may, about 3 months after I'd hoped to finally have a
>> stable release.
> [...]
>>
>> So hopefully the upcoming 3.10.40-4 build will be stable enough to
>> freeze on. If there is anything else that people want done before
>> freezing, let me know soonest.
>>
>
> Excellent news!
> So if I want to upgrade my router this weekend is 3.10.40-4
> ready for testing yet, or should I wait a bit more?
>
> Best regards,
> --Edwin
>
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel



-- 
Dave Täht

NSFW: https://w2.eff.org/Censorship/Internet_censorship_bills/russell_0296_indecent.article

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 1010 bytes --]

On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:

> So hopefully the upcoming 3.10.40-4 build will be stable enough to
> freeze on. If there is anything else that people want done before
> freezing, let me know soonest.

Upgraded just fine for me, been running for several hours on a wndr3800 without
any big problems.

Minor nit 1:  It doesn't seem to save the 'Start UPNP now' setting, so that's
something I need to re-set each time I upgrade. Seems to have been this way
for a long time, 3.10.32-9 did it to me too.  My PS3 says UPNP works once
I start it.

Minor nit 2: It grants IPv4 leases for 24 hours, but IPv6 for only 1 hour.
That's OK - except my laptop running Fedora seems to think the lease is for 24
hours, not 1, so the lease ends up expiring.  Haven't chased this to see who's
to blame.  It's not a show stopper for me, because my laptop's
SLAAC-autoconfigured address and RFC3041 privacy addresses keep working anyhow.
It's probably been as long as the other one, 



[-- Attachment #2: Type: application/pgp-signature, Size: 848 bytes --]

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Aristar]

I also have one thing to add, the "associated stations" hasn't been
working properly in LUCI, it only shows the master station instead of
the clients connected like it used to in 3.7.5, not sure who's to
blame here. Can others confirm this? Not a huge deal, but thought I
might throw this out there anyways.

On Sat, May 17, 2014 at 2:54 AM,  <Valdis.Kletnieks@vt.edu> wrote:
> On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:
>
>> So hopefully the upcoming 3.10.40-4 build will be stable enough to
>> freeze on. If there is anything else that people want done before
>> freezing, let me know soonest.
>
> Upgraded just fine for me, been running for several hours on a wndr3800 without
> any big problems.
>
> Minor nit 1:  It doesn't seem to save the 'Start UPNP now' setting, so that's
> something I need to re-set each time I upgrade. Seems to have been this way
> for a long time, 3.10.32-9 did it to me too.  My PS3 says UPNP works once
> I start it.
>
> Minor nit 2: It grants IPv4 leases for 24 hours, but IPv6 for only 1 hour.
> That's OK - except my laptop running Fedora seems to think the lease is for 24
> hours, not 1, so the lease ends up expiring.  Haven't chased this to see who's
> to blame.  It's not a show stopper for me, because my laptop's
> SLAAC-autoconfigured address and RFC3041 privacy addresses keep working anyhow.
> It's probably been as long as the other one,
>
>
>
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 685 bytes --]

On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:

> So hopefully the upcoming 3.10.40-4 build will be stable enough to
> freeze on. If there is anything else that people want done before
> freezing, let me know soonest.

I just noticed a "wait, what?".  Luci is telling me:

Hostname	bogon-gateway-1
Model	NETGEAR WNDR3800
Firmware Version	CeroWrt Toronto 3.10.40-4 / LuCI Trunk (git-c5a9089)
Kernel Version	3.10.36
Local Time	Sun May 18 09:44:00 2014
Uptime	1d 12h 36m 54s

and sure enough, ssh'ing in I see:

root@bogon-gateway-1:~# uname -a
Linux bogon-gateway-1 3.10.36 #1 Fri May 16 01:00:30 PDT 2014 mips GNU/Linux

Something has gone astray somewhere...

[-- Attachment #2: Type: application/pgp-signature, Size: 848 bytes --]

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Dave Taht]

On Sun, May 18, 2014 at 6:48 AM,  <Valdis.Kletnieks@vt.edu> wrote:
> On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:
>
>> So hopefully the upcoming 3.10.40-4 build will be stable enough to
>> freeze on. If there is anything else that people want done before
>> freezing, let me know soonest.
>
> I just noticed a "wait, what?".  Luci is telling me:
>
> Hostname        bogon-gateway-1
> Model   NETGEAR WNDR3800
> Firmware Version        CeroWrt Toronto 3.10.40-4 / LuCI Trunk (git-c5a9089)
> Kernel Version  3.10.36
> Local Time      Sun May 18 09:44:00 2014
> Uptime  1d 12h 36m 54s
>
> and sure enough, ssh'ing in I see:
>
> root@bogon-gateway-1:~# uname -a
> Linux bogon-gateway-1 3.10.36 #1 Fri May 16 01:00:30 PDT 2014 mips GNU/Linux
>
> Something has gone astray somewhere...

I am running a few versions ahead of openwrt, and I apparently forgot to
commit the last kernel change makefile, so on a pull from openwrt
I'd done a git stash, and it reverted to openwrt head...
(I DID get it right on the ubnt build I did and was puzzled as to
 why the pull from cero main was the wrong kernel)

The patch, however, that hopefully stablizes wifi, IS in this build, so
please feel free to keep at blowing it up.

Obviously I'll do another release...

(it is the patch, not the kernel, that is important)

THX for the spot! I would never have noticed til
.41 came out.

-- 
Dave Täht

NSFW: https://w2.eff.org/Censorship/Internet_censorship_bills/russell_0296_indecent.article

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Maciej Soltysiak]

On Sun, May 18, 2014 at 5:29 PM, Dave Taht <dave.taht@gmail.com> wrote:
> On Sun, May 18, 2014 at 6:48 AM,  <Valdis.Kletnieks@vt.edu> wrote:
>> On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:
>> root@bogon-gateway-1:~# uname -a
>> Linux bogon-gateway-1 3.10.36 #1 Fri May 16 01:00:30 PDT 2014 mips GNU/Linux
Noticed the same after upgrade yesterday.
BTW. sysupgrade didn't preserver a single setting, is that expected?
Was upgrading form 3.10.28 or something.

> (it is the patch, not the kernel, that is important)
Well, if you look closely the 3.10.40 contains a bugfix for
n_tty_write() : CVE-2014-0196
That particular kernel might be important. Sorry for my ignorance, but
are we affected?

Best regards,
Maciej Soltysiak

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Sebastian Moeller]

Hi Maciej, hi list,

\
On May 19, 2014, at 09:30 , Maciej Soltysiak <maciej@soltysiak.com> wrote:

> On Sun, May 18, 2014 at 5:29 PM, Dave Taht <dave.taht@gmail.com> wrote:
>> On Sun, May 18, 2014 at 6:48 AM,  <Valdis.Kletnieks@vt.edu> wrote:
>>> On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:
>>> root@bogon-gateway-1:~# uname -a
>>> Linux bogon-gateway-1 3.10.36 #1 Fri May 16 01:00:30 PDT 2014 mips GNU/Linux
> Noticed the same after upgrade yesterday.

	Note 3.10.40-5 is out and sports a shiny 3.10.40 kernel ;) (Installed yesterday, seems to work, but only lightly tested; did not expose 11an for the 5GHz radio or allow HT40 from the GUI, copying the relevant lines from /etc/config/wireless both made 40MHz wifi work again and exposed the settings in the GUI; I seem to recall that this was reported on the openwrt list as well, so is not a cerowrt specific issue...)

> BTW. sysupgrade didn't preserver a single setting, is that expected?
> Was upgrading form 3.10.28 or something.

	INteresting question, this could be related to the constant flux in openwrt trunk, or could just be a plain bug there. (Just as the inability of keeping gthe settings while sys upgrading from the GUI was an openwrt bug, that got fixed there...)

> 
>> (it is the patch, not the kernel, that is important)
> Well, if you look closely the 3.10.40 contains a bugfix for
> n_tty_write() : CVE-2014-0196

	This looks like a local issue only, since the only local user on cerowrt is root I do not think that this really affects us.

best regards
	sebastian

> That particular kernel might be important. Sorry for my ignorance, but
> are we affected?
> 
> Best regards,
> Maciej Soltysiak
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Dave Taht]

On Mon, May 19, 2014 at 2:57 AM, Sebastian Moeller <moeller0@gmx.de> wrote:
> Hi Maciej, hi list,
>
> \
> On May 19, 2014, at 09:30 , Maciej Soltysiak <maciej@soltysiak.com> wrote:
>
>> On Sun, May 18, 2014 at 5:29 PM, Dave Taht <dave.taht@gmail.com> wrote:
>>> On Sun, May 18, 2014 at 6:48 AM,  <Valdis.Kletnieks@vt.edu> wrote:
>>>> On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:
>>>> root@bogon-gateway-1:~# uname -a
>>>> Linux bogon-gateway-1 3.10.36 #1 Fri May 16 01:00:30 PDT 2014 mips GNU/Linux
>> Noticed the same after upgrade yesterday.
>
>         Note 3.10.40-5 is out and sports a shiny 3.10.40 kernel ;) (Installed yesterday, seems to work, but only lightly tested; did not expose 11an for the 5GHz radio or allow HT40 from the GUI, copying the relevant lines from /etc/config/wireless both made 40MHz wifi work again and exposed the settings in the GUI; I seem to recall that this was reported on the openwrt list as well, so is not a cerowrt specific issue...)

? I saw that the wifi recog routine had changed is it not configuring n?

>
>> BTW. sysupgrade didn't preserver a single setting, is that expected?
>> Was upgrading form 3.10.28 or something.
>
>         INteresting question, this could be related to the constant flux in openwrt trunk, or could just be a plain bug there. (Just as the inability of keeping gthe settings while sys upgrading from the GUI was an openwrt bug, that got fixed there...)

I do look forward to BB freezing.
>
>>
>>> (it is the patch, not the kernel, that is important)
>> Well, if you look closely the 3.10.40 contains a bugfix for
>> n_tty_write() : CVE-2014-0196
>
>         This looks like a local issue only, since the only local user on cerowrt is root I do not think that this really affects us.
>
> best regards
>         sebastian
>
>> That particular kernel might be important. Sorry for my ignorance, but
>> are we affected?
>>
>> Best regards,
>> Maciej Soltysiak
>> _______________________________________________
>> Cerowrt-devel mailing list
>> Cerowrt-devel@lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>



-- 
Dave Täht

NSFW: https://w2.eff.org/Censorship/Internet_censorship_bills/russell_0296_indecent.article

Re: [Cerowrt-devel] A stable cerowrt release in sight

[Sebastian Moeller]

Hi Dave, hi list,

On May 19, 2014, at 18:29 , Dave Taht <dave.taht@gmail.com> wrote:

> On Mon, May 19, 2014 at 2:57 AM, Sebastian Moeller <moeller0@gmx.de> wrote:
>> Hi Maciej, hi list,
>> 
>> \
>> On May 19, 2014, at 09:30 , Maciej Soltysiak <maciej@soltysiak.com> wrote:
>> 
>>> On Sun, May 18, 2014 at 5:29 PM, Dave Taht <dave.taht@gmail.com> wrote:
>>>> On Sun, May 18, 2014 at 6:48 AM,  <Valdis.Kletnieks@vt.edu> wrote:
>>>>> On Thu, 15 May 2014 23:36:11 -0700, Dave Taht said:
>>>>> root@bogon-gateway-1:~# uname -a
>>>>> Linux bogon-gateway-1 3.10.36 #1 Fri May 16 01:00:30 PDT 2014 mips GNU/Linux
>>> Noticed the same after upgrade yesterday.
>> 
>>        Note 3.10.40-5 is out and sports a shiny 3.10.40 kernel ;) (Installed yesterday, seems to work, but only lightly tested; did not expose 11an for the 5GHz radio or allow HT40 from the GUI, copying the relevant lines from /etc/config/wireless both made 40MHz wifi work again and exposed the settings in the GUI; I seem to recall that this was reported on the openwrt list as well, so is not a cerowrt specific issue...)
> 
> ? I saw that the wifi recog routine had changed is it not configuring n?

	Nope, it only offered either "auto" or "a" for the 5GHz radio, for radio0 it offered "n" out of the box. I hope this is just transient and will be fixed upstream...

> 
>> 
>>> BTW. sysupgrade didn't preserver a single setting, is that expected?
>>> Was upgrading form 3.10.28 or something.
>> 
>>        INteresting question, this could be related to the constant flux in openwrt trunk, or could just be a plain bug there. (Just as the inability of keeping gthe settings while sys upgrading from the GUI was an openwrt bug, that got fixed there...)
> 
> I do look forward to BB freezing.

	;)


Best Regards
	Sebastian

>> 
>>> 
>>>> (it is the patch, not the kernel, that is important)
>>> Well, if you look closely the 3.10.40 contains a bugfix for
>>> n_tty_write() : CVE-2014-0196
>> 
>>        This looks like a local issue only, since the only local user on cerowrt is root I do not think that this really affects us.
>> 
>> best regards
>>        sebastian
>> 
>>> That particular kernel might be important. Sorry for my ignorance, but
>>> are we affected?
>>> 
>>> Best regards,
>>> Maciej Soltysiak
>>> _______________________________________________
>>> Cerowrt-devel mailing list
>>> Cerowrt-devel@lists.bufferbloat.net
>>> https://lists.bufferbloat.net/listinfo/cerowrt-devel
>> 
> 
> 
> 
> -- 
> Dave Täht
> 
> NSFW: https://w2.eff.org/Censorship/Internet_censorship_bills/russell_0296_indecent.article