Kernel for A64 git

[androsch]

OK, also added longsleeps network-script.sh to /etc/rc.local and set the governor to performance like tkaiser already suggested (still using gmac-kernel from longsleep, with normal kernel its unusable in GbE), so i now have about 500 mbits with still many retries, but i can live with that at the moment. Its much better than FastEthernet now and using this board as server for minecraft is more limitated by cpu than network now.

Thanks everybody for suggestions and support!

Next up, you can apply some local QoS to your interfaces, and see if that doesn't help you out a little.

This might get you started pretty quickly:   https://github.com/akhepcat/qdisc

With qdisc start and running iperf3 -s afterwards, there is no connection possible (as bad as before without gmac-kernel).....

So how does this help now?

[benjanyan]

I was nudged by waldo some weeks ago to give the gmac kernel a try. I previously suffered from intermittent Gbit NIC issues where it'd slow to a crawl somewhat randomly. The board's used as a NAS and VPN gateway and the network connection is now performing close to Gbit/s on iperf and hasn't exhibited any problems I've noticed in its 25 days uptime.

I'm happy. Nice work.

[pfeerick]

I was nudged by waldo some weeks ago to give the gmac kernel a try. I previously suffered from intermittent Gbit NIC issues where it'd slow to a crawl somewhat randomly. The board's used as a NAS and VPN gateway and the network connection is now performing close to Gbit/s on iperf and hasn't exhibited any problems I've noticed in its 25 days uptime.

I'm happy. Nice work.

Thanks for the update... I'm sure TL and the team will be happy to hear that!

[ak_hepcat]

OK, also added longsleeps network-script.sh to /etc/rc.local and set the governor to performance like tkaiser already suggested (still using gmac-kernel from longsleep, with normal kernel its unusable in GbE), so i now have about 500 mbits with still many retries, but i can live with that at the moment. Its much better than FastEthernet now and using this board as server for minecraft is more limitated by cpu than network now.

Thanks everybody for suggestions and support!

Next up, you can apply some local QoS to your interfaces, and see if that doesn't help you out a little.

This might get you started pretty quickly:   https://github.com/akhepcat/qdisc

With qdisc start and running iperf3 -s afterwards, there is no connection possible (as bad as before without gmac-kernel).....

So how does this help now?

Did you adjust the script for your local environment/speeds?   What did you set it to?

I ask, because this script is running on about 35 different systems quite well for my company.

[longsleep]

I just pushed a new awesome Kernel release. It brings a security fix, overlayfs support and the fix for GbE. For details and update instructions, see the first post of this thread.

Get 3.10.102-4-pine64-longsleep-113 from https://www.stdin.xyz/downloads/people/l...ges/linux/

[tllim]

I just pushed a new awesome Kernel release. It brings a security fix, overlayfs support and the fix for GbE. For details and update instructions, see the first post of this thread.

Get 3.10.102-4-pine64-longsleep-113 from https://www.stdin.xyz/downloads/people/l...ges/linux/

Thanks.

[androsch]

I just pushed a new awesome Kernel release. It brings a security fix, overlayfs support and the fix for GbE. For details and update instructions, see the first post of this thread.

Get 3.10.102-4-pine64-longsleep-113 from https://www.stdin.xyz/downloads/people/l...ges/linux/

Thanks.

Hoorrayy, this seems to do the trick now:

Pine with previous problems as server now:

[ ID] Interval           Transfer     Bandwidth       Retr
[  5]   0.00-30.00  sec   667 MBytes   187 Mbits/sec    0             sender
[  5]   0.00-30.00  sec   664 MBytes   186 Mbits/sec                  receiver
[  7]   0.00-30.00  sec   661 MBytes   185 Mbits/sec    0             sender
[  7]   0.00-30.00  sec   658 MBytes   184 Mbits/sec                  receiver
[  9]   0.00-30.00  sec   656 MBytes   183 Mbits/sec    0             sender
[  9]   0.00-30.00  sec   653 MBytes   183 Mbits/sec                  receiver
[ 11]   0.00-30.00  sec   657 MBytes   184 Mbits/sec    0             sender
[ 11]   0.00-30.00  sec   654 MBytes   183 Mbits/sec                  receiver
[SUM]   0.00-30.00  sec  2.58 GBytes   738 Mbits/sec    0             sender
[SUM]   0.00-30.00  sec  2.57 GBytes   735 Mbits/sec                  receiver

..and as client

[ ID] Interval           Transfer     Bandwidth       Retr
[  5]   0.00-30.01  sec   685 MBytes   191 Mbits/sec  487             sender
[  5]   0.00-30.01  sec   681 MBytes   190 Mbits/sec                  receiver
[  7]   0.00-30.01  sec   675 MBytes   189 Mbits/sec  460             sender
[  7]   0.00-30.01  sec   671 MBytes   188 Mbits/sec                  receiver
[  9]   0.00-30.01  sec   574 MBytes   160 Mbits/sec  120             sender
[  9]   0.00-30.01  sec   572 MBytes   160 Mbits/sec                  receiver
[ 11]   0.00-30.01  sec   637 MBytes   178 Mbits/sec  250             sender
[ 11]   0.00-30.01  sec   636 MBytes   178 Mbits/sec                  receiver
[SUM]   0.00-30.01  sec  2.51 GBytes   718 Mbits/sec  1317             sender
[SUM]   0.00-30.01  sec  2.50 GBytes   716 Mbits/sec                  receiver

So with new kernel nearly no retries any more and (not yet perfect, but network and some settings also are not optimized) GbE speed.

Thanks to all for this long and windy road, i can at least change my signature now :-)

[HenryD]

Pardon if this is wrong thread.. it concerns (Linux Pine64_B 3.10.104-1-pine64gmac-longsleep) kernel but on a gentoo install. Note the gmac variant.

Invoking /usr/local/sbin/pine64_update_kernel.sh results in:
------
Pine64_B gentoo # /usr/local/sbin/pine64_update_kernel.sh
Using Kernel variant: gmac
Checking for update ...
You are already on latest version - abort.
------

My sense of the update_kernel.sh script is that the presence of (VARIANTFILE="/var/lib/misc/pine64_update_kernel.variant") will prevent return/update to the standard longsleep kernel series. This may affect others who are using the gmac kernel variant.

Edit: Renaming the 'variant' file did allow an upgrade back to the standard kernel series.

Sincere thanks to the many who push to make all this happen and be fun and interesting.

HenryD

[longsleep]

Edit: Renaming the 'variant' file did allow an upgrade back to the standard kernel series.

No need to edit anything. Just add the version and the variant as parameters like this:

Code:

sudo /usr/local/sbin/pine64_update_kernel.sh latest default

to update to latest default Kernel no matter what variant you currently have installed.

Usage: pine64_update_kernel.sh [<version>] [<variant>]

The default values for these parameters are "latest" for version and "default" for variant. If a variant is pinned that variant is used instead of the default if the parameters is not given.

[default]

Hi, im use pine64+ board and this image:
Linux pine64 3.10.104-2-pine64-longsleep #113 SMP PREEMPT Thu Dec 15 21:46:07 CET 2016 aarch64 aarch64 aarch64 GNU/Linux

I noticed a problem. Packet loss >50% on gigabyt link or the connection does not work.
I believed that this problem has been fixed. That check to fix the problem on my board?

my board BANANA PI M64.
Thanks.