Joined: 03 Jan 2010 Posts: 7568 Location: YWG, Canada
Posted: Fri May 31, 2019 17:40 Post subject:
that smp affinity is old & for k3.18 i think, need to do cat /proc/interrupts & use the correct one for eth0 its 100 or 101 now _________________ LATEST FIRMWARE(S)
BrainSlayer wrote:
we just do it since we do not like any restrictions enforced by stupid cocaine snorting managers
Kong builds have WAN & wireless interrupts mostly fixed -- BS builds do NOT.
@stephensuley- You can try running some custom commands to move the WAN interface over to the other CPU Core and force the EA8500 CPU's to run at full speed all the time to get the most speed out of the router.
Seems both commands failed to take on startup tab. I entered them from the CLI and I saw the clock jump up to 1400 on the second core as reported in the webgui after I entered the second command. _________________ Location (urban) - 1x Linksys EA8500 (AP wlan0 & wlan1 enabled)
1x Asus 68u (Repeater Bridge w/VAP) - wl0 disabled
1x Asus 87u (Client Bridge) - wl1 disabled
Joined: 21 Jan 2017 Posts: 1783 Location: Illinois Moderator
Posted: Fri May 31, 2019 18:07 Post subject:
stephensuley wrote:
stephensuley wrote:
stephensuley wrote:
msoengineer wrote:
mrjcd wrote:
Kong builds have WAN & wireless interrupts mostly fixed -- BS builds do NOT.
@stephensuley- You can try running some custom commands to move the WAN interface over to the other CPU Core and force the EA8500 CPU's to run at full speed all the time to get the most speed out of the router.
Seems both commands failed to take on startup tab. I entered them from the CLI and I saw the clock jump up to 1400 on the second core as reported in the webgui after I entered the second command.
For the EA8500 it's irq 100. So,
Code:
echo 2 > /proc/irq/100/smp_affinity
for CPUFREQ in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do [ -f $CPUFREQ ] || continue; echo -n performance > $CPUFREQ; done
save that in startup commands, and then reboot the router _________________ FORUM RULES
I reckon if you have real fast internet you probably benefit using Kong and maybe keeping CUP at 1400 'See This' for the difference in Kong & BS for the EA8500
Now days they both have minimum CPU 800MHz
Router: Netgear R7800
Firmware: DD-WRT v3.0-r39855M kongat (05/25/19)
Kernel: Linux 4.9.177 #412 SMP Sat May 25 14:58:57 CEST 2019 armv7l
Status: Working
Reset: No
Previous: 39715M
Errors: No
Temperatures : CPU 53.22 °C / ath0 28 °C / ath1 28 °C
Uptime 7 days, 6:23
Working very well :
Router mode : DHCP
SFE Enable STP Enable
DNSMasq
Cache DNSSEC data
Validate DNS Replies (DNSSEC)
Check unsigned DNS replies
Local DNS
No DNS Rebind
ath0, ath1
Vpn (OpenVPN Client)
I've just installed this version on a fresh R7800, but I am unable to any responses on WAN Pings.. I have disabled (unticked) "Block Anonymous WAN Requests (ping)" under the firewall settings so it should work..? Is this a bug? (My previous dd-wrt on my R7000 was over a year old, so I dont know if its not been working for a while or not.. )
Joined: 03 Jan 2010 Posts: 7568 Location: YWG, Canada
Posted: Wed Jun 05, 2019 6:47 Post subject:
can this get looked at? fq_codel's weak upload gets worked around by either having an entry in netmask or interface priority.
since my network is on 10.150.10.1/24 on br0 if i enter that in netmask, then give it any priority (MUST do priority, a speed of 0/0 will do nothing), then i go from this upload:
my entered qos rates are the exact same of 327680 download (320mbps) and 15974 upload (15.6mbps). this obviously creates more load on the cpu as theres more rules in iptables -t mangle -vnL, but its not my much. both ways of approach work around this bug, but netmask adds more load as theres doubled the amount of extra rules than that of interface br0 addition way.
the priority of the netmask or interface does seem to have an effect on how fairly bandwidth is split even though i have port based entries in services (the same ones multiple seem to like using), an entry of premium seems to be the best option, guaranteeing 50% of entered bandwidth. if i use standard, than a p2p app using max upload with all random port 30000+ tcp, will take 85% of the bandwidth from everything else, even if the other streams are entered in services list above, such as https tcp 443 at premium..
i thought this wasnt affecting other routers like mips based ones on k3.18 but it is as well, so someone fucked up fq_codel across multiple kernels. this difference in upload is happening across all routers using fq_codel, htb/hfsc. all types of load, on all client devices, in the same way.
why does a netmask or interface entry like that (that even gets screwy with services priority) fix the upload to get much closer to what the user enters? _________________ LATEST FIRMWARE(S)
BrainSlayer wrote:
we just do it since we do not like any restrictions enforced by stupid cocaine snorting managers
Joined: 03 Jan 2010 Posts: 7568 Location: YWG, Canada
Posted: Thu Jun 06, 2019 2:20 Post subject:
vit5421 wrote:
Guys,
R u able to use the 802.11r / fast bss transition? Any advice please
client device must support 802.11r, and 802.11w management frame protection MUST be disabled, or this problem happens https://svn.dd-wrt.com/ticket/6655
mobility domain must be the same between all interfaces on the same network, set the nas id of each interface to the mac address of the interface without the ":", eg: if ath0 mac is 12:34:56:78:90:AB set ath0 mobility domain to 1234567890AB same thing for other interfaces. _________________ LATEST FIRMWARE(S)
BrainSlayer wrote:
we just do it since we do not like any restrictions enforced by stupid cocaine snorting managers
I don't know exactly when happen, but seems to be related to another problem that happen when I open a laptop that is in standby, so all the lan/wifi stop working (on all devices) for some time and then start to working again. If I wait the problem internet start again after some time, if I close the laptop lid the internet start immediately. So seems to be correlated but I'm sure.