[Cerowrt-devel] some kernel updates

Sebastian Moeller moeller0 at gmx.de
Sat Aug 24 16:51:35 EDT 2013


Hi Dave,


On Aug 23, 2013, at 22:29 , Dave Taht <dave.taht at gmail.com> wrote:

> 
> 
> 
> On Fri, Aug 23, 2013 at 12:56 PM, Sebastian Moeller <moeller0 at gmx.de> wrote:
> Hi Dave,
> 
> I guess I found the culprit:
> 
> once I added $ADSLL to the ingress() in simple.qos:
> 
> 
> 
> I had that in there originally. I ripped it out because it seemed to help with ADSL at the time - as I was unaware the extent that the whole subsystem was busted!

	Ah, and I had added my stab based version to both ingress() and egress() assuming that both links need to be kept under control. So with fixed htb link layer adjustment (LLA) it only worked on the uplink and in retrospect if I look at my initial test data I actually see one of the hallmarks of a working LLA for the upstream. (The upstream good-put was reduced compared to the no LLA test, caused by LLA making the actually sent packets larger so fewer packets fit through the shaped link). But since I was not expecting only half a working system I overlooked that in the data.
	But looking at the latency of the ping RTT probes it becomes quite clear that only doing link layer adjustments on the uplink is even worse than not doing it all (because the latency is still almost as bad as without LLA but the up-link bandwidth is reduced).

> 
> I like to think of the process we've just gone through as "wow, we just fixed the uk, and a few other countries". :) Feels kind of good, doesn't it? (Too bad the pay sucks.)

	Oh, I can not complain about pay, I have a day job in totally different field, so this is more of a hobby for me :) 

> I mean, jeeze, chopping another 30+ms off the latency of that many systems should get medals from economists worldwide monitoring productivity. 
> 
> Does anyone have a date/kernel version on when linklayer overhead compensation stopped working? There was a bug even prior to 3.8 that looked bad. (and RED was busted for 3 years).
> 
> Another step would be trying to improve openwrt's native qos system somewhat in the DSL case. They don't use this subsystem (probably because it didn't work), and it's also broke on ipv6. (They use conn track)

	Oh, in the bql-40 time frame I hacked the stab based LLA into their generate.sh and it worked quite well, even though at time my measurements were quite crude. SInce their qos scripts are HFSC based the HTB private implementation is not going to do them any good. Luckily now that does not seem to matter as both methods now perform identically as they should. (Well, now Jespers last changes are nicer than the old table lookup, but it should be relatively say to implant the same for stab, heck once I got my linux machine up I might take this as my first attempt at making local changes to the kernel :) ). So adding it to openwrt proper should be a piece of cake. Do you know by any chance who would be the best person to contact for that, ?

> 
> At some point I'd like to have a mechanism for saner diffserv classification on egress, and to clamp ingress values to egress ones. There is a ton of work going on on finding sane codepoints on webrtc in the ietf….
> 
> 
> 
> 
> 
> 
> ingress() {
> 
> CEIL=$DOWNLINK
> PRIO_RATE=`expr $CEIL / 3` # Ceiling for prioirty
> BE_RATE=`expr $CEIL / 6`   # Min for best effort
> BK_RATE=`expr $CEIL / 6`   # Min for background
> BE_CEIL=`expr $CEIL - 64`  # A little slop at the top
> 
> LQ="quantum `get_mtu $IFACE`"
> 
> $TC qdisc del dev $IFACE handle ffff: ingress 2> /dev/null
> $TC qdisc add dev $IFACE handle ffff: ingress
> 
> $TC qdisc del dev $DEV root  2> /dev/null
> $TC qdisc add dev $DEV root handle 1: ${STABSTRING} htb default 12
> $TC class add dev $DEV parent 1: classid 1:1 htb $LQ rate ${CEIL}kbit ceil ${CEIL}kbit $ADSLL
> $TC class add dev $DEV parent 1:1 classid 1:10 htb $LQ rate ${CEIL}kbit ceil ${CEIL}kbit prio 0 $ADSLL
> $TC class add dev $DEV parent 1:1 classid 1:11 htb $LQ rate 32kbit ceil ${PRIO_RATE}kbit prio 1 $ADSLL
> $TC class add dev $DEV parent 1:1 classid 1:12 htb $LQ rate ${BE_RATE}kbit ceil ${BE_CEIL}kbit prio 2 $ADSLL
> $TC class add dev $DEV parent 1:1 classid 1:13 htb $LQ rate ${BK_RATE}kbit ceil ${BE_CEIL}kbit prio 3 $ADSLL
> 
> # I'd prefer to use a pre-nat filter but that causes permutation...
> 
> $TC qdisc add dev $DEV parent 1:11 handle 110: $QDISC limit 1000 $ECN `get_quantum 500` `get_flows ${PRIO_RATE}`
> $TC qdisc add dev $DEV parent 1:12 handle 120: $QDISC limit 1000 $ECN `get_quantum 1500` `get_flows ${BE_RATE}`
> $TC qdisc add dev $DEV parent 1:13 handle 130: $QDISC limit 1000 $ECN `get_quantum 1500` `get_flows ${BK_RATE}`
> 
> diffserv $DEV
> 
> ifconfig $DEV up
> 
> # redirect all IP packets arriving in $IFACE to ifb0
> 
> $TC filter add dev $IFACE parent ffff: protocol all prio 10 u32 \
> match u32 0 0 flowid 1:1 action mirred egress redirect dev $DEV
> 
> }
> 
> I get basically the same RRUL ping RTTs for htb_private as for tc_stab. So Jesper was right the patch seems to fix the issue. I guess I should send out my current version of yours and Toke's AQM scripts soon.
> 
> 
> 
> Best
>    Sebastian
> 
> P.S.: I am not sure whether I want to tackle the PIE issue today...
> 
> 
> 
> On Aug 23, 2013, at 21:47 , Dave Taht <dave.taht at gmail.com> wrote:
> 
>> quick note: running this script requires that you
>> 
>> ifconfig ifb0 up
>> 
>> at some point.
> 
>    In my case on cerowrt you took care of that already...
> 
> 
>> 
>> 
>> On Fri, Aug 23, 2013 at 12:38 PM, Sebastian Moeller <moeller0 at gmx.de> wrote:
>> Hi Dave,
>> 
>> On Aug 23, 2013, at 07:13 , Dave Taht <dave.taht at gmail.com> wrote:
>> 
>>> 
>>> 
>>> 
>>> On Thu, Aug 22, 2013 at 5:52 PM, Sebastian Moeller <moeller0 at gmx.de> wrote:
>>> Hi List, hi Jesper,
>>> 
>>> So I tested 3.10.9-1 to assess the status of the HTB atm link layer adjustments to see whether the recent changes resurrected this feature.
>>>    Unfortunately the htb_private link layer adjustments still is broken (RRUL ping RTT against Toke's netperf host in Germany of ~80ms, same as without link layer adjustments). On the bright side the tc_stab method still works as well as before (ping RTT around 40ms).
>>>    I would like to humbly propose to use the tc stab method in cerowrt to perform ATM link layer adjustments as default. To repeat myself, simply telling the kernel a lie about the packet size seems more robust than fudging HTB's rate tables. Especially since the kernel already fudges the packet size to account for the ethernet header and then some, so this path should receive more scrutiny by virtue of having more users?
>>> 
>>> It's my hope that the atm code works but is misconfigured. You can output the tc commands by overriding the TC variable with TC="echo tc" and paste here.
>> 
>>    So I went for TC="logger tc" and used log read to harvest as I could not find the echo output, but I guess that should not matter. So here is the result (slightly edited to get rid of the log timestamps and log level):
>> 
>> tc qdisc del dev ge00 root
>> tc qdisc add dev ge00 root handle 1: htb default 12
>> tc class add dev ge00 parent 1: classid 1:1 htb quantum 1500 rate 2430kbit ceil 2430kbit mpu 0 linklayer adsl overhead 40 mtu 2047
>> tc class add dev ge00 parent 1:1 classid 1:10 htb quantum 1500 rate 2430kbit ceil 2430kbit prio 0 mpu 0 linklayer adsl overhead 40 mtu 2047
>> tc class add dev ge00 parent 1:1 classid 1:11 htb quantum 1500 rate 128kbit ceil 810kbit prio 1 mpu 0 linklayer adsl overhead 40 mtu 2047
>> tc class add dev ge00 parent 1:1 classid 1:12 htb quantum 1500 rate 405kbit ceil 2366kbit prio 2 mpu 0 linklayer adsl overhead 40 mtu 2047
>> tc class add dev ge00 parent 1:1 classid 1:13 htb quantum 1500 rate 405kbit ceil 2366kbit prio 3 mpu 0 linklayer adsl overhead 40 mtu 2047
>> tc qdisc add dev ge00 parent 1:11 handle 110: fq_codel limit 600 noecn quantum 300
>> tc qdisc add dev ge00 parent 1:12 handle 120: fq_codel limit 600 noecn quantum 300
>> tc qdisc add dev ge00 parent 1:13 handle 130: fq_codel limit 600 noecn quantum 300
>> tc filter add dev ge00 parent 1:0 protocol all prio 999 u32 match ip protocol 0 0x00 flowid 1:12
>> tc filter add dev ge00 parent 1:0 protocol ip prio 1 handle 1 fw classid 1:11
>> tc filter add dev ge00 parent 1:0 protocol ip prio 2 handle 2 fw classid 1:12
>> tc filter add dev ge00 parent 1:0 protocol ip prio 3 handle 3 fw classid 1:13
>> tc filter add dev ge00 parent 1:0 protocol ipv6 prio 4 handle 1 fw classid 1:11
>> tc filter add dev ge00 parent 1:0 protocol ipv6 prio 5 handle 2 fw classid 1:12
>> tc filter add dev ge00 parent 1:0 protocol ipv6 prio 6 handle 3 fw classid 1:13
>> tc filter add dev ge00 parent 1:0 protocol arp prio 7 handle 1 fw classid 1:11
>> tc qdisc del dev ge00 handle ffff: ingress
>> tc qdisc add dev ge00 handle ffff: ingress
>> tc qdisc del dev ifb0 root
>> tc qdisc add dev ifb0 root handle 1: htb default 12
>> tc class add dev ifb0 parent 1: classid 1:1 htb quantum 1500 rate 15494kbit ceil 15494kbit
>> tc class add dev ifb0 parent 1:1 classid 1:10 htb quantum 1500 rate 15494kbit ceil 15494kbit prio 0
>> tc class add dev ifb0 parent 1:1 classid 1:11 htb quantum 1500 rate 32kbit ceil 5164kbit prio 1
>> tc class add dev ifb0 parent 1:1 classid 1:12 htb quantum 1500 rate 2582kbit ceil 15430kbit prio 2
>> tc class add dev ifb0 parent 1:1 classid 1:13 htb quantum 1500 rate 2582kbit ceil 15430kbit prio 3
>> tc qdisc add dev ifb0 parent 1:11 handle 110: fq_codel limit 1000 ecn quantum 500
>> tc qdisc add dev ifb0 parent 1:12 handle 120: fq_codel limit 1000 ecn quantum 1500
>> tc qdisc add dev ifb0 parent 1:13 handle 130: fq_codel limit 1000 ecn quantum 1500
>> tc filter add dev ifb0 parent 1:0 protocol all prio 999 u32 match ip protocol 0 0x00 flowid 1:12
>> tc filter add dev ifb0 protocol ip parent 1:0 prio 1 u32 match ip tos 0x00 0xfc classid 1:12
>> tc filter add dev ifb0 protocol ipv6 parent 1:0 prio 2 u32 match ip6 priority 0x00 0xfc classid 1:12
>> tc filter add dev ifb0 protocol ip parent 1:0 prio 3 u32 match ip tos 0x20 0xfc classid 1:13
>> tc filter add dev ifb0 protocol ipv6 parent 1:0 prio 4 u32 match ip6 priority 0x20 0xfc classid 1:13
>> tc filter add dev ifb0 protocol ip parent 1:0 prio 5 u32 match ip tos 0x10 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ipv6 parent 1:0 prio 6 u32 match ip6 priority 0x10 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ip parent 1:0 prio 7 u32 match ip tos 0xb8 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ipv6 parent 1:0 prio 8 u32 match ip6 priority 0xb8 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ip parent 1:0 prio 9 u32 match ip tos 0xc0 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ipv6 parent 1:0 prio 10 u32 match ip6 priority 0xc0 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ip parent 1:0 prio 11 u32 match ip tos 0xe0 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ipv6 parent 1:0 prio 12 u32 match ip6 priority 0xe0 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ip parent 1:0 prio 13 u32 match ip tos 0x90 0xfc classid 1:11
>> tc filter add dev ifb0 protocol ipv6 parent 1:0 prio 14 u32 match ip6 priority 0x90 0xfc classid 1:11
>> tc filter add dev ifb0 parent 1:0 protocol arp prio 15 handle 1 fw classid 1:11
>> tc filter add dev ge00 parent ffff: protocol all prio 10 u32 match u32 0 0 flowid 1:1 action mirred egress redirect dev ifb0
>> 
>> I notice it seem this only shows up for egress(), but looking at simple.qos ingress() is not addend ${ADSLL} at all so that is to be expected. There is nothing in dmesg at all.
>> 
>> So I am off to add ADSLL to ingress() as well and then test RRUL again...
>> 
>> 
>> Jesper please let me know if this looks reasonable, at least to my eye it seems to fit with what "tc disc add htb help" tells me. I tried your:
>> echo "func __detect_linklayer +p" /sys/kernel/debug/dynamic_debug/control
>> but got no output even though debugs was already mounted…
>> 
>> Best
>>    Sebastian
>> 
>>> 
>>>    Now, I have been testing this using Dave's most recent cerowrt alpha version with a 3.10.9 kernel on mips hardware, I think this kernel should contain all htb fixes including commit 8a8e3d84b17 (net_sched: restore "linklayer atm" handling) but am not fully sure.
>>> 
>>> It does.
>>> 
>>> `@Dave is there an easy way to find which patches you applied to the kernels of the cerowrt (testing-)releases?
>>> 
>>> Normally I DO commit stuff that is in testing, but my big push this time around was to get everything important into mainline 3.10, as it will be the "stable" release for a good long time.
>>> 
>>> So I am still mostly working the x86 side at the moment. I WAS kind of hoping that everything I just landed would make it up to 3.10. But for your perusal:
>>> 
>>> http://snapon.lab.bufferbloat.net/~cero2/patches/3.10.9-1/ has most of the kernel patches I used in it. 3.10.9-2 has the ipv6subtrees patch ripped out due to another weird bug I'm looking at. (It also has support for ipv6 nat thx to the ever prolific stephen walker heeding the call for patches...). 100% totally untested, I have this weird bug to figure out how to fix next:
>>> 
>>> http://lists.alioth.debian.org/pipermail/babel-users/2013-August/001419.html
>>> 
>>> I fear it's a comparison gone south, maybe in bradley's optimizations for not kernel trapping, don't know.
>>> 
>>> 3.10.9-2 also disables dnsmasq's dhcpv6 in favor of 6relayd. I HATE losing the close naming integration, but, had to try this....
>>> 
>>> If you guys want me to start committing and pushing patches again, I'll do it, but most of that stuff will end up in 3.10.10, I think, in a couple days. The rest might make 3.12. Pie has to survive scrutiny on the netdev list in particular.
>>> 
>>> While I have you r attention :) I also tested 3.10.9-1's pie and it is way better than 3.10.6-1's (RRUL ping RTTs around 110 ms instead of 3000ms) but still worse than fq_codel (ping RTTs around 40ms with proper atm link layer adjustments).
>>> 
>>> This is with simple.qos I imagine? Simplest should do better than that with pie. Judging from how its estimator works I think it will do badly with multiple queues. But testing will tell...
>>> 
>>> But, yea, this pie is actually usable, and the previous wasn't. Thank you for looking at it!
>>> 
>>> It is different from cisco's last pie drop in that it can do ecn, does local congestion notification, has a better use of net_random, it's mostly KernelStyle, and I forget what else.
>>> 
>>> There is still a major rounding error in the code, and I'd like cisco to fix the api so it uses identical syntax to codel. Right now you specify "target 8" to get "target 7", and the "ms" is implied. target 5 becomes target 3. The default target is a whopping 20 (rounded to 19), which is in part where your 70+ms of extra delay came from.
>>> 
>>> Multiple parties have the delusion that 20ms is "good enough".
>>> 
>>> Part of the remaining delay may also be rounding error. Cisco uses kernels with HZ=1000, cero uses HZ=250.....
>>> 
>>> Anyway, to get more comparable tests... you can fiddle with the two $QDISC lines in simple*.qos to add a target 8 to get closer to a codel 5ms config, but that would break a codel config which treats target 8 as target 8us.
>>> 
>>> I MIGHT, if I get energetic enough, fix the API, the time accounting, and a few other things in pie, the problem is, that ns2_codel seems still more effective on most workloads and *fq_codel smokes absolutely everything. There are a few places where pie is a win over straight codel, notably on packet floods. And it may well be easier to retrofit into existing hardware fast path designs.
>>> 
>>> I worry about interactions between pie and other stuff. It seems inevitable at this point that some form of pie will be widely deployed, and I simply haven't tried enough traffic types and RTTs to draw a firm conclusion, period. Long RTTs are the last big place where codel and pie and fq_codel have to be seriously tested.
>>> 
>>> ns2_codel is looking pretty good now, at the shorter RTTs I've tried. A big problem I have is getting decent long RTT emulation out of netem (some preliminary code is up at github)
>>> 
>>> ... and getting cero stable enough for others to actually use - next up is fixing the userspace problems.
>>> 
>>> ... and trying to make a small dent in the wifi problem along the way (couple commits coming up)
>>> 
>>> ... and find funding to get through the winter.
>>> 
>>> There's probably a few other things that are on that list but I forget. Oh, yea, since the aqm wg was voted on to be formed, I decided I could quit smoking.
>>> 
>>> While I am not able to build kernels, it seems that I am able to quickly test whether link layer adjustments work or not. SO aim happy to help where I can :)
>>> 
>>> Give pie target 8 and target 5 a shot, please? ns2_codel target 3ms and target 7ms, too. fq_codel, same....
>>> 
>>> tc -s qdisc show dev ge00
>>> tc -s qdisc show dev ifb0
>>> 
>>> would be useful info to have in general after each test.
>>> 
>>> TIA.
>>> 
>>> There are also things like tcp_upload and tcp_download and tcp_bidirectional that are useful tests in the rrul suite.
>>> 
>>> Thank you for your efforts on these early alpha releases. I hope things will stablize more soon, and I'll fold your aqm stuff into my next attempt this weekend.
>>> 
>>> This is some of the stuff I know that needs fixing in userspace:
>>> 
>>> * TODO readlink not found
>>> * TODO netdev user missing
>>> * TODO Wed Dec  5 17:14:46 2012 authpriv.error dnsmasq: found already running DHCP-server on interface 'se00' refusing to start, use 'option force 1' to override
>>> * TODO [   18.480468] Mirror/redirect action on
>>> [   18.539062] Failed to load ipt action
>>> * upload and download are reversed in aqm
>>> * BCP38
>>> * Squash CS values
>>> * Replace ntp
>>> * Make ahcp client mode
>>> * Drop more privs for polipo
>>> * upnp
>>> * priv separation
>>> * Review FW rules
>>> * dhcpv6 support
>>> * uci-defaults/make-cert.sh uses a bad path for px5g
>>> * Doesn't configure the web browser either
>>> 
>>> 
>>> 
>>> 
>>> Best
>>>    Sebastian
>>> 
>>> 
>>> 
>>> 
>>> --
>>> Dave Täht
>>> 
>>> Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html
>> 
>> 
>> 
>> 
>> --
>> Dave Täht
>> 
>> Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html
> 
> 
> 
> 
> -- 
> Dave Täht
> 
> Fixing bufferbloat with cerowrt: http://www.teklibre.com/cerowrt/subscribe.html




More information about the Cerowrt-devel mailing list