[Cake] new patchset for upstream net-next

[Cake] new patchset for upstream net-next

[Dave Taht]

[-- Attachment #1: Type: text/plain, Size: 465 bytes --]

I fired off a build of this with the changes to allocate_src/dst that kevin
just committed... which will take long enough for me to do laundry.
Hopefully. Regardless, I'm out of action for 3 hrs.

I'd certainly love it if others integrated this against net-next and
tried it. In particular nobody (including me) has ever tried a "y"
rather than a "m" for building it....

-- 

Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619

[-- Attachment #2: netnext_cake.tgz --]
[-- Type: application/x-gzip, Size: 23188 bytes --]

Re: [Cake] new patchset for upstream net-next

[George Amanakis]

I tried Kevin's latest commit, now it fails with:

   CHK     include/config/kernel.release
   CHK     include/generated/uapi/linux/version.h
   DESCEND  objtool
   CHK     include/generated/utsrelease.h
   CHK     scripts/mod/devicetable-offsets.h
   CHK     include/generated/bounds.h
   CHK     include/generated/timeconst.h
   CHK     include/generated/asm-offsets.h
   CALL    scripts/checksyscalls.sh
   CHK     include/generated/compile.h
   CHK     kernel/config_data.h
   GEN     .version
   CHK     include/generated/compile.h
   UPD     include/generated/compile.h
   CC      init/version.o
   AR      init/built-in.o
   AR      built-in.o
   LD      vmlinux.o
   MODPOST vmlinux.o
net/sched/sch_cake.o: In function `cake_hash':
sch_cake.c:(.text+0x23d0): undefined reference to `nf_ct_get_tuplepr'
sch_cake.c:(.text+0x23f3): undefined reference to `nf_conntrack_find_get'
make: *** [Makefile:993: vmlinux] Error 1

Everything is selected (Y or M) under "Core Netfilter Configuration". 
Are these functions deprecated in net-next?

George


On 11/25/2017 4:42 PM, Dave Taht wrote:
> bad merge. darn it.
>
> On Sat, Nov 25, 2017 at 11:57 AM, George Amanakis <gamanakis@gmail.com> wrote:
>> After correcting this, it fails with:
>>
>>    CHK     include/config/kernel.release
>>    CHK     include/generated/uapi/linux/version.h
>>    DESCEND  objtool
>>    CHK     include/generated/utsrelease.h
>>    CHK     scripts/mod/devicetable-offsets.h
>>    CHK     include/generated/bounds.h
>>    CHK     include/generated/timeconst.h
>>    CHK     include/generated/asm-offsets.h
>>    CALL    scripts/checksyscalls.sh
>>    CHK     include/generated/compile.h
>>    CHK     kernel/config_data.h
>>    CC      net/sched/sch_cake.o
>> net/sched/sch_cake.c: In function ‘cake_hash’:
>> net/sched/sch_cake.c:746:7: error: ‘allocate_host’ undeclared (first use in
>> this function); did you mean ‘allocate_dst’?
>>     if (allocate_host) {
>>         ^~~~~~~~~~~~~
>>         allocate_dst
>> net/sched/sch_cake.c:746:7: note: each undeclared identifier is reported
>> only once for each function it appears in
>> make[2]: *** [scripts/Makefile.build:310: net/sched/sch_cake.o] Error 1
>> make[1]: *** [scripts/Makefile.build:569: net/sched] Error 2
>> make[1]: *** Waiting for unfinished jobs....
>> make: *** [Makefile:1012: net] Error 2
>> make: *** Waiting for unfinished jobs....
>> ==> ERROR: A failure occurred in build().
>>      Aborting...
>>
>> George
>
>

Re: [Cake] new patchset for upstream net-next

[George Amanakis]

grep'ing in net-next for nf_ct_get_tuplepr reveals these are still in use.


On 11/25/2017 7:49 PM, George Amanakis wrote:
> I tried Kevin's latest commit, now it fails with:
>
>   CHK     include/config/kernel.release
>   CHK     include/generated/uapi/linux/version.h
>   DESCEND  objtool
>   CHK     include/generated/utsrelease.h
>   CHK     scripts/mod/devicetable-offsets.h
>   CHK     include/generated/bounds.h
>   CHK     include/generated/timeconst.h
>   CHK     include/generated/asm-offsets.h
>   CALL    scripts/checksyscalls.sh
>   CHK     include/generated/compile.h
>   CHK     kernel/config_data.h
>   GEN     .version
>   CHK     include/generated/compile.h
>   UPD     include/generated/compile.h
>   CC      init/version.o
>   AR      init/built-in.o
>   AR      built-in.o
>   LD      vmlinux.o
>   MODPOST vmlinux.o
> net/sched/sch_cake.o: In function `cake_hash':
> sch_cake.c:(.text+0x23d0): undefined reference to `nf_ct_get_tuplepr'
> sch_cake.c:(.text+0x23f3): undefined reference to `nf_conntrack_find_get'
> make: *** [Makefile:993: vmlinux] Error 1
>
> Everything is selected (Y or M) under "Core Netfilter Configuration". 
> Are these functions deprecated in net-next?
>
> George
>
>
> On 11/25/2017 4:42 PM, Dave Taht wrote:
>> bad merge. darn it.
>>
>> On Sat, Nov 25, 2017 at 11:57 AM, George Amanakis 
>> <gamanakis@gmail.com> wrote:
>>> After correcting this, it fails with:
>>>
>>>    CHK     include/config/kernel.release
>>>    CHK     include/generated/uapi/linux/version.h
>>>    DESCEND  objtool
>>>    CHK     include/generated/utsrelease.h
>>>    CHK     scripts/mod/devicetable-offsets.h
>>>    CHK     include/generated/bounds.h
>>>    CHK     include/generated/timeconst.h
>>>    CHK     include/generated/asm-offsets.h
>>>    CALL    scripts/checksyscalls.sh
>>>    CHK     include/generated/compile.h
>>>    CHK     kernel/config_data.h
>>>    CC      net/sched/sch_cake.o
>>> net/sched/sch_cake.c: In function ‘cake_hash’:
>>> net/sched/sch_cake.c:746:7: error: ‘allocate_host’ undeclared (first 
>>> use in
>>> this function); did you mean ‘allocate_dst’?
>>>     if (allocate_host) {
>>>         ^~~~~~~~~~~~~
>>>         allocate_dst
>>> net/sched/sch_cake.c:746:7: note: each undeclared identifier is 
>>> reported
>>> only once for each function it appears in
>>> make[2]: *** [scripts/Makefile.build:310: net/sched/sch_cake.o] Error 1
>>> make[1]: *** [scripts/Makefile.build:569: net/sched] Error 2
>>> make[1]: *** Waiting for unfinished jobs....
>>> make: *** [Makefile:1012: net] Error 2
>>> make: *** Waiting for unfinished jobs....
>>> ==> ERROR: A failure occurred in build().
>>>      Aborting...
>>>
>>> George
>>
>>
>

Re: [Cake] new patchset for upstream net-next

[Dave Taht]

I just got back from town and pushed a version that builds into the
for_upstream_4.16 branch, with kevins latest.

Trying another build here, with "m", takes hours. thx for trying 'y'!

As for what's going wrong... is nf_conntrack being built? as a module
? as integral? We've always built cake and nf_conntrack as modules
before.



On Sat, Nov 25, 2017 at 4:51 PM, George Amanakis <gamanakis@gmail.com> wrote:
> grep'ing in net-next for nf_ct_get_tuplepr reveals these are still in use.
>
>
>
> On 11/25/2017 7:49 PM, George Amanakis wrote:
>>
>> I tried Kevin's latest commit, now it fails with:
>>
>>   CHK     include/config/kernel.release
>>   CHK     include/generated/uapi/linux/version.h
>>   DESCEND  objtool
>>   CHK     include/generated/utsrelease.h
>>   CHK     scripts/mod/devicetable-offsets.h
>>   CHK     include/generated/bounds.h
>>   CHK     include/generated/timeconst.h
>>   CHK     include/generated/asm-offsets.h
>>   CALL    scripts/checksyscalls.sh
>>   CHK     include/generated/compile.h
>>   CHK     kernel/config_data.h
>>   GEN     .version
>>   CHK     include/generated/compile.h
>>   UPD     include/generated/compile.h
>>   CC      init/version.o
>>   AR      init/built-in.o
>>   AR      built-in.o
>>   LD      vmlinux.o
>>   MODPOST vmlinux.o
>> net/sched/sch_cake.o: In function `cake_hash':
>> sch_cake.c:(.text+0x23d0): undefined reference to `nf_ct_get_tuplepr'
>> sch_cake.c:(.text+0x23f3): undefined reference to `nf_conntrack_find_get'
>> make: *** [Makefile:993: vmlinux] Error 1
>>
>> Everything is selected (Y or M) under "Core Netfilter Configuration". Are
>> these functions deprecated in net-next?
>>
>> George
>>
>>
>> On 11/25/2017 4:42 PM, Dave Taht wrote:
>>>
>>> bad merge. darn it.
>>>
>>> On Sat, Nov 25, 2017 at 11:57 AM, George Amanakis <gamanakis@gmail.com>
>>> wrote:
>>>>
>>>> After correcting this, it fails with:
>>>>
>>>>    CHK     include/config/kernel.release
>>>>    CHK     include/generated/uapi/linux/version.h
>>>>    DESCEND  objtool
>>>>    CHK     include/generated/utsrelease.h
>>>>    CHK     scripts/mod/devicetable-offsets.h
>>>>    CHK     include/generated/bounds.h
>>>>    CHK     include/generated/timeconst.h
>>>>    CHK     include/generated/asm-offsets.h
>>>>    CALL    scripts/checksyscalls.sh
>>>>    CHK     include/generated/compile.h
>>>>    CHK     kernel/config_data.h
>>>>    CC      net/sched/sch_cake.o
>>>> net/sched/sch_cake.c: In function ‘cake_hash’:
>>>> net/sched/sch_cake.c:746:7: error: ‘allocate_host’ undeclared (first use
>>>> in
>>>> this function); did you mean ‘allocate_dst’?
>>>>     if (allocate_host) {
>>>>         ^~~~~~~~~~~~~
>>>>         allocate_dst
>>>> net/sched/sch_cake.c:746:7: note: each undeclared identifier is reported
>>>> only once for each function it appears in
>>>> make[2]: *** [scripts/Makefile.build:310: net/sched/sch_cake.o] Error 1
>>>> make[1]: *** [scripts/Makefile.build:569: net/sched] Error 2
>>>> make[1]: *** Waiting for unfinished jobs....
>>>> make: *** [Makefile:1012: net] Error 2
>>>> make: *** Waiting for unfinished jobs....
>>>> ==> ERROR: A failure occurred in build().
>>>>      Aborting...
>>>>
>>>> George
>>>
>>>
>>>
>>
>



-- 

Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619

Re: [Cake] new patchset for upstream net-next

[George Amanakis]

I am giving it another try. nf_conntrack was built as a module. I am 
rebuilding on a clean dir, nf_conntrack set as integral this time. Will 
report soon.

George


On 11/25/2017 8:00 PM, Dave Taht wrote:
> I just got back from town and pushed a version that builds into the
> for_upstream_4.16 branch, with kevins latest.
>
> Trying another build here, with "m", takes hours. thx for trying 'y'!
>
> As for what's going wrong... is nf_conntrack being built? as a module
> ? as integral? We've always built cake and nf_conntrack as modules
> before.
>
>
>
> On Sat, Nov 25, 2017 at 4:51 PM, George Amanakis <gamanakis@gmail.com> wrote:
>> grep'ing in net-next for nf_ct_get_tuplepr reveals these are still in use.
>>
>>
>>
>> On 11/25/2017 7:49 PM, George Amanakis wrote:
>>> I tried Kevin's latest commit, now it fails with:
>>>
>>>    CHK     include/config/kernel.release
>>>    CHK     include/generated/uapi/linux/version.h
>>>    DESCEND  objtool
>>>    CHK     include/generated/utsrelease.h
>>>    CHK     scripts/mod/devicetable-offsets.h
>>>    CHK     include/generated/bounds.h
>>>    CHK     include/generated/timeconst.h
>>>    CHK     include/generated/asm-offsets.h
>>>    CALL    scripts/checksyscalls.sh
>>>    CHK     include/generated/compile.h
>>>    CHK     kernel/config_data.h
>>>    GEN     .version
>>>    CHK     include/generated/compile.h
>>>    UPD     include/generated/compile.h
>>>    CC      init/version.o
>>>    AR      init/built-in.o
>>>    AR      built-in.o
>>>    LD      vmlinux.o
>>>    MODPOST vmlinux.o
>>> net/sched/sch_cake.o: In function `cake_hash':
>>> sch_cake.c:(.text+0x23d0): undefined reference to `nf_ct_get_tuplepr'
>>> sch_cake.c:(.text+0x23f3): undefined reference to `nf_conntrack_find_get'
>>> make: *** [Makefile:993: vmlinux] Error 1
>>>
>>> Everything is selected (Y or M) under "Core Netfilter Configuration". Are
>>> these functions deprecated in net-next?
>>>
>>> George
>>>
>>>
>>> On 11/25/2017 4:42 PM, Dave Taht wrote:
>>>> bad merge. darn it.
>>>>
>>>> On Sat, Nov 25, 2017 at 11:57 AM, George Amanakis <gamanakis@gmail.com>
>>>> wrote:
>>>>> After correcting this, it fails with:
>>>>>
>>>>>     CHK     include/config/kernel.release
>>>>>     CHK     include/generated/uapi/linux/version.h
>>>>>     DESCEND  objtool
>>>>>     CHK     include/generated/utsrelease.h
>>>>>     CHK     scripts/mod/devicetable-offsets.h
>>>>>     CHK     include/generated/bounds.h
>>>>>     CHK     include/generated/timeconst.h
>>>>>     CHK     include/generated/asm-offsets.h
>>>>>     CALL    scripts/checksyscalls.sh
>>>>>     CHK     include/generated/compile.h
>>>>>     CHK     kernel/config_data.h
>>>>>     CC      net/sched/sch_cake.o
>>>>> net/sched/sch_cake.c: In function ‘cake_hash’:
>>>>> net/sched/sch_cake.c:746:7: error: ‘allocate_host’ undeclared (first use
>>>>> in
>>>>> this function); did you mean ‘allocate_dst’?
>>>>>      if (allocate_host) {
>>>>>          ^~~~~~~~~~~~~
>>>>>          allocate_dst
>>>>> net/sched/sch_cake.c:746:7: note: each undeclared identifier is reported
>>>>> only once for each function it appears in
>>>>> make[2]: *** [scripts/Makefile.build:310: net/sched/sch_cake.o] Error 1
>>>>> make[1]: *** [scripts/Makefile.build:569: net/sched] Error 2
>>>>> make[1]: *** Waiting for unfinished jobs....
>>>>> make: *** [Makefile:1012: net] Error 2
>>>>> make: *** Waiting for unfinished jobs....
>>>>> ==> ERROR: A failure occurred in build().
>>>>>       Aborting...
>>>>>
>>>>> George
>>>>
>>>>
>
>

Re: [Cake] new patchset for upstream net-next

[gamanakis]

Just finished building, setting both sch_cake and nf_conntrack as
integral succeeds. Setting nf_conntrack as module fails with sch_cake
as integral (makes sense).

I am remote right now and cannot test the newly built kernel, hopefully
I will get it done tonight.

George

On Sat, 2017-11-25 at 17:00 -0800, Dave Taht wrote:
> I just got back from town and pushed a version that builds into the
> for_upstream_4.16 branch, with kevins latest.
> 
> Trying another build here, with "m", takes hours. thx for trying 'y'!
> 
> As for what's going wrong... is nf_conntrack being built? as a module
> ? as integral? We've always built cake and nf_conntrack as modules
> before.
> 
> 

Re: [Cake] new patchset for upstream net-next

[gamanakis]

[-- Attachment #1: Type: text/plain, Size: 561 bytes --]

I am currently testing latest net-next with both sch_cake and
nf_conntrack compiled as integrals (Y instead of M). So far it works as
expected (previous build was 4.9.64, sch_cake out-of-tree, and
nf_conntrack as module).

Are there any specific tests you would like me to run?

I am attaching the patch I used.

George

On Sat, 2017-11-25 at 20:27 -0500, George Amanakis wrote:
> I am giving it another try. nf_conntrack was built as a module. I am 
> rebuilding on a clean dir, nf_conntrack set as integral this time.
> Will 
> report soon.
> 
> George
> 
> 

[-- Attachment #2: netnext_cake.george.tgz --]
[-- Type: application/x-compressed-tar, Size: 23218 bytes --]

Re: [Cake] new patchset for upstream net-next

[Dave Taht]

awesome.

On Sat, Nov 25, 2017 at 8:38 PM,  <gamanakis@gmail.com> wrote:
> I am currently testing latest net-next with both sch_cake and
> nf_conntrack compiled as integrals (Y instead of M). So far it works as
> expected (previous build was 4.9.64, sch_cake out-of-tree, and
> nf_conntrack as module).
>
> Are there any specific tests you would like me to run?

Whatever your primary use case is? My biggest concern is that it
simply not crash -  300 second long tests, 1200 seconds, all night
long over and over, again, pounding it flat.

rrul_be, rrul, 100 flows, 1000 flows, multiple stations, udp_flood,
bittorrent, steam, with/without ack-filter, ack-filter-aggressive,
asymmetric networks, as the default qdisc, unlimited, on mq
hardware.... triple-isolate, dual-whatever, besteffort... ingress...
every combination of keywords,

I lost track of the QA matrix once it started rivaling string theory
in complexity.

I am comforted it works with 4.9.

>
> I am attaching the patch I used.


> George
>
> On Sat, 2017-11-25 at 20:27 -0500, George Amanakis wrote:
>> I am giving it another try. nf_conntrack was built as a module. I am
>> rebuilding on a clean dir, nf_conntrack set as integral this time.
>> Will
>> report soon.
>>
>> George
>>
>>



-- 

Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619

Re: [Cake] new patchset for upstream net-next

[Pete Heist]

[-- Attachment #1: Type: text/plain, Size: 722 bytes --]


> On Nov 26, 2017, at 6:46 AM, Dave Taht <dave.taht@gmail.com> wrote:
> 
> Whatever your primary use case is? My biggest concern is that it
> simply not crash -  300 second long tests, 1200 seconds, all night
> long over and over, again, pounding it flat.
> 
> rrul_be, rrul, 100 flows, 1000 flows, multiple stations, udp_flood,
> bittorrent, steam, with/without ack-filter, ack-filter-aggressive,
> asymmetric networks, as the default qdisc, unlimited, on mq
> hardware.... triple-isolate, dual-whatever, besteffort... ingress...
> every combination of keywords,
> 
> I lost track of the QA matrix once it started rivaling string theory
> in complexity.

I volunteer to cover the 'dual-whatever’ tests.

[-- Attachment #2: Type: text/html, Size: 8798 bytes --]

Re: [Cake] new patchset for upstream net-next

[Toke Høiland-Jørgensen]

gamanakis@gmail.com writes:

> Just finished building, setting both sch_cake and nf_conntrack as
> integral succeeds. Setting nf_conntrack as module fails with sch_cake
> as integral (makes sense).

So this probably needs either a Kconfig dependency or an ifdef?

-Toke

Re: [Cake] new patchset for upstream net-next

[Dave Taht]

Toke Høiland-Jørgensen <toke@toke.dk> writes:

> gamanakis@gmail.com writes:
>
>> Just finished building, setting both sch_cake and nf_conntrack as
>> integral succeeds. Setting nf_conntrack as module fails with sch_cake
>> as integral (makes sense).
>
> So this probably needs either a Kconfig dependency or an ifdef?

I don't really know. We want cake to compile without conntrack
(it's expensive).

>
> -Toke
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake

Re: [Cake] new patchset for upstream net-next

[Toke Høiland-Jørgensen]

On 26 November 2017 19:22:35 CET, Dave Taht <dave@taht.net> wrote:
>Toke Høiland-Jørgensen <toke@toke.dk> writes:
>
>> gamanakis@gmail.com writes:
>>
>>> Just finished building, setting both sch_cake and nf_conntrack as
>>> integral succeeds. Setting nf_conntrack as module fails with
>sch_cake
>>> as integral (makes sense).
>>
>> So this probably needs either a Kconfig dependency or an ifdef?
>
>I don't really know. We want cake to compile without conntrack
>(it's expensive).

So ifdef then? 

What happens if the conntrack module is unloaded (currently)?

-Toke

Re: [Cake] new patchset for upstream net-next

[Kevin Darbyshire-Bryant]

> On 26 Nov 2017, at 18:22, Dave Taht <dave@taht.net> wrote:
> 
> Toke Høiland-Jørgensen <toke@toke.dk> writes:
> 
>> gamanakis@gmail.com writes:
>> 
>>> Just finished building, setting both sch_cake and nf_conntrack as
>>> integral succeeds. Setting nf_conntrack as module fails with sch_cake
>>> as integral (makes sense).
>> 
>> So this probably needs either a Kconfig dependency or an ifdef?
> 
> I don't really know. We want cake to compile without conntrack
> (it's expensive).

I’m slightly confused by that statement - The conntrack lookup is controlled per instance and is one conditional in the code path:

if (flow_mode & CAKE_FLOW_NAT_FLAG)
                cake_update_flowkeys(&keys, skb);

Admittedly cake_update_flowkeys is pretty gnarly but it’s only called if you configure the qdisc with ‘nat’.  What am I misunderstanding?

> 
>> 
>> -Toke
>> _______________________________________________
>> Cake mailing list
>> Cake@lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/cake
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake

Re: [Cake] new patchset for upstream net-next

[Jonathan Morton]

[-- Attachment #1: Type: text/plain, Size: 303 bytes --]

It won't link unless conntrack is in the kernel, and *that* is costly for
some.

What we could do is make NAT support optional in Kconfig and have that
option depend on conntrack.  Would need a little fettling of cake itself to
make the NAT support properly optional at compile time.

- Jonathan Morton

[-- Attachment #2: Type: text/html, Size: 367 bytes --]

Re: [Cake] new patchset for upstream net-next

[George Amanakis]

> Whatever your primary use case is? My biggest concern is that it
> simply not crash -  300 second long tests, 1200 seconds, all night
> long over and over, again, pounding it flat.

My home router runs x86_64 Archlinux on net-next with cake and 
nf_conntrack compiled as integrals. TSO, GSO and GRO are turned off. Up 
until now (24h) it has not crashed and everything seems stable. Dmesg 
does not report anything unusual. Ingress traffic is about 15 GiB, 
egress about 1 GiB.

Cake configuration (Comcast cable, advertised 10Mbps/2Mbps):
qdisc cake 8001: dev ens4 root refcnt 2 bandwidth 12200Kbit diffserv3 
dual-dsthost wash ingress ack-filter rtt 100.0ms noatm overhead 18 
via-ethernet mpu 64
qdisc cake 8002: dev ens3 root refcnt 2 bandwidth 2500Kbit diffserv3 
dual-srchost nat wash ack-filter rtt 100.0ms noatm overhead 18 
via-ethernet mpu 64

George

Re: [Cake] new patchset for upstream net-next

[David Lang]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 570 bytes --]

On Sun, 26 Nov 2017, Kevin Darbyshire-Bryant wrote:

>> I don't really know. We want cake to compile without conntrack
>> (it's expensive).
>
> I’m slightly confused by that statement - The conntrack lookup is controlled per instance and is one conditional in the code path:

It's not the cake lookup that's expensive, it's the overhead of running 
conntrack on the system that's expensive.

Yes, most distros enable it by default, but we don't want cake to not work on a 
system that has it disabled (although some things may not work as well, 
gracefull degredation)

Re: [Cake] new patchset for upstream net-next

[Dave Taht]

George Amanakis <gamanakis@gmail.com> writes:

>> Whatever your primary use case is? My biggest concern is that it
>> simply not crash -  300 second long tests, 1200 seconds, all night
>> long over and over, again, pounding it flat.
>
> My home router runs x86_64 Archlinux on net-next with cake and nf_conntrack
> compiled as integrals. TSO, GSO and GRO are turned off. Up until now (24h) it
> has not crashed and everything seems stable. Dmesg does not report anything
> unusual. Ingress traffic is about 15 GiB, egress about 1 GiB.

Groovy. Can you re-enable TSO/GSO/GRO?

>
> Cake configuration (Comcast cable, advertised 10Mbps/2Mbps):
> qdisc cake 8001: dev ens4 root refcnt 2 bandwidth 12200Kbit diffserv3
> dual-dsthost wash ingress ack-filter rtt 100.0ms noatm overhead 18 via-ethernet
> mpu 64
> qdisc cake 8002: dev ens3 root refcnt 2 bandwidth 2500Kbit diffserv3
> dual-srchost nat wash ack-filter rtt 100.0ms noatm overhead 18 via-ethernet mpu
> 64

tc -s qdisc show

>
> George
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake

Re: [Cake] new patchset for upstream net-next

[Sebastian Moeller]

> On Nov 27, 2017, at 04:33, George Amanakis <gamanakis@gmail.com> wrote:
> 
> 
>> Whatever your primary use case is? My biggest concern is that it
>> simply not crash -  300 second long tests, 1200 seconds, all night
>> long over and over, again, pounding it flat.
> 
> My home router runs x86_64 Archlinux on net-next with cake and nf_conntrack compiled as integrals. TSO, GSO and GRO are turned off. Up until now (24h) it has not crashed and everything seems stable. Dmesg does not report anything unusual. Ingress traffic is about 15 GiB, egress about 1 GiB.
> 
> Cake configuration (Comcast cable, advertised 10Mbps/2Mbps):
> qdisc cake 8001: dev ens4 root refcnt 2 bandwidth 12200Kbit diffserv3 dual-dsthost wash ingress ack-filter rtt 100.0ms noatm overhead 18 via-ethernet mpu 64
> qdisc cake 8002: dev ens3 root refcnt 2 bandwidth 2500Kbit diffserv3 dual-srchost nat wash ack-filter rtt 100.0ms noatm overhead 18 via-ethernet mpu 64

Just like ECN ack-filter really only makes sense if the ACKs never hit the bottleneck link; I guess what I want to say, besides for testing, ack-filter on your ingress fro internet leg does not make too much sense...

Best Regards


> 
> George
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake

[Cake] Fwd: new patchset for upstream net-next

[Georgios Amanakis]

[-- Attachment #1: Type: text/plain, Size: 3351 bytes --]

---------- Forwarded message ----------
From: Georgios Amanakis <gamanakis@gmail.com>
Date: Mon, Nov 27, 2017 at 9:27 AM
Subject: Re: [Cake] new patchset for upstream net-next
To: Sebastian Moeller <moeller0@gmx.de>


> Groovy. Can you re-enable TSO/GSO/GRO?
Of course! I have them disabled generally, not just for testing.

> Just like ECN ack-filter really only makes sense if the ACKs never hit
the bottleneck link; I guess what I want to say, besides for testing,
ack-filter on your ingress fro internet leg does not make too much >
sense...
I was getting to the same conclusion Sebastian, thanks for pointing this
out.

I rebooted the router last night in order to integrate the latest commits.
This morning's "tc -s qdisc show":
===============8<===============
qdisc noqueue 0: dev lo root refcnt 2
 Sent 0 bytes 0 pkt (dropped 0, overlimits 0 requeues 0)
 backlog 0b 0p requeues 0
qdisc cake 8001: dev ens4 root refcnt 2 bandwidth 12200Kbit diffserv3
dual-dsthost wash ingress ack-filter rtt 100.0ms noatm overhead 18
via-ethernet mpu 64
 Sent 446349530 bytes 363341 pkt (dropped 1363, overlimits 617167 requeues
0)
 backlog 0b 0p requeues 0
 memory used: 324352b of 4Mb
 capacity estimate: 12200Kbit
                 Bulk   Best Effort      Voice
  thresh     762496bit   12200Kbit    3050Kbit
  target        23.8ms       5.0ms       6.0ms
  interval     118.8ms     100.0ms     101.0ms
  pk_delay       7.2ms       581us       1.1ms
  av_delay       5.5ms        33us        78us
  sp_delay         2us         1us         2us
  pkts          342129       18980        3595
  bytes      434159057    13724645      395648
  way_inds         138           0           0
  way_miss        4549        6707          17
  way_cols           0           0           0
  drops           1069         214           0
  marks              2           0           0
  ack_drop          22          58           0
  sp_flows           0           0           0
  bk_flows           0           0           1
  un_flows           0           0           0
  max_len         1514        1514        1514

qdisc cake 8002: dev ens3 root refcnt 2 bandwidth 2500Kbit diffserv3
dual-srchost nat wash ack-filter rtt 100.0ms noatm overhead 18 via-ethernet
mpu 64
 Sent 34023411 bytes 188901 pkt (dropped 3390, overlimits 125227 requeues 0)
 backlog 0b 0p requeues 0
 memory used: 241920b of 4Mb
 capacity estimate: 2500Kbit
                 Bulk   Best Effort      Voice
  thresh     156248bit    2500Kbit     625Kbit
  target       116.3ms       7.3ms      29.1ms
  interval     232.6ms     102.3ms     124.1ms
  pk_delay         3us      27.4ms       9.8ms
  av_delay         0us       5.7ms       746us
  sp_delay         0us         4us        56us
  pkts              12      190371        1908
  bytes            600    35696740      190789
  way_inds           0         301           0
  way_miss          12        9736          79
  way_cols           0           0           0
  drops              0        1363           0
  marks              0          10           0
  ack_drop           0        2027           0
  sp_flows           0           0           0
  bk_flows           0           0           1
  un_flows           0           0           0
  max_len           74        1514        1514
===============8<===============


George

[-- Attachment #2: Type: text/html, Size: 4495 bytes --]