1. Hardware & Machines
  3. Driver Boards
  5. Mesa hm2/hm2_7i96s.0: error finishing read

Mesa hm2/hm2_7i96s.0: error finishing read

More
10 Dec 2022 22:02 #259140 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read

rtt min/avg/max/mdev = 0.245/0.465/0.698/0.074 ms

max ping time = 0.510

Nope, max ping time 0.698.
Usually the first one or two replies are longer, then settle.
min/ avg/ max/ mdev
0.245/0.465/0.698/0.074 ms

Tommy, yses .510 is after the first slower ping.
But the point is pings are  worse than Johns and does not get an error finishing read and TMAX values are way lower...
The queston is why?

Please Log in or Create an account to join the conversation.

More
10 Dec 2022 22:22 #259144 by tommylight
Replied by tommylight on topic Mesa hm2/hm2_7i96s.0: error finishing read
I would venture a guess at the way the driver requests interrupts to hardware, but i am not willing to go deep into that, especially since you already did a lot of work in figuring out solutions and what drivers work, so thank you for the time and effort you put into it.
Did anyone try ping with bigger frames/packets?
Pretty sure the size will mess response times quickly on the kernels with issues.
Or maybe, just maybe there is something wrong in those kernels in the way they handle UDP transport?
Ping is TCP/IP/ICMP, Mesa uses UDP if i am not mistaken, and so did Quake 3, the game. Quake had a way to switch to TCP/IP but that was laggy as hell with the limited bandwidth we had.

Please Log in or Create an account to join the conversation.

More
10 Dec 2022 22:32 #259148 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read
I don't think getting a machine to work by accdent really counts moving forward to a solution.
I have wondered about NIC parameters.
I am wondering if the hm2-eth driver can be improved to compensate. Maybe its the way we do things..

Please Log in or Create an account to join the conversation.

More
10 Dec 2022 23:06 #259158 by PCW
Replied by PCW on topic Mesa hm2/hm2_7i96s.0: error finishing read
I think the issue is kernel driver related (though it may be unrelated drivers such as
video drivers)

To test ping times , you must run the test at a real time priority, and run the test for a long time, something like:

sudo chrt 99 ping -i .001 -q 10.10.10.10
The following user(s) said Thank You: tommylight, JT

Please Log in or Create an account to join the conversation.

More
10 Dec 2022 23:15 #259159 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read
For a short time, the average was 0.129. The Max is meaningless as its the first ping and no doubt is distorting the result.
Trying for a long time now.......

Please Log in or Create an account to join the conversation.

More
11 Dec 2022 03:10 #259177 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read
After 4 hours
--- 10.10.10.10 ping statistics ---
14022619 packets transmitted, 14022619 received, 0% packet loss, time 14037118ms
rtt min/avg/max/mdev = 0.056/0.129/0.881/0.054 ms
The max is still likely to be the startup command. I gusee but the average is good @ 0.129

Please Log in or Create an account to join the conversation.

More
12 Dec 2022 10:27 #259247 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read
Moving on, I started to work on another J1900 based industrial PC.
I've had absolutely no luck building kernels on a fresh version of Bookworm with the 6.0 kernel.
After updating the earlier script to use the Release Candidate 7 following advances in the kernel versioning,
It failed because it could not find  System.map

I then tried to build it the Debian way
www.dwarmstrong.org/kernel/
which refers to the  debian handbook
debian-handbook.info/browse/stable/sect.kernel-compilation.html
and it failed too. So I have a couple of PC's now which are essentially useless for use with Mesa cards. Very frustrating
It would be interesting to see if somebody could actually build the kernel the Debian way and share the resulting deb for testing.
(use linuxcnc instead of custom when naming it)

Please Log in or Create an account to join the conversation.

More
13 Dec 2022 18:18 - 13 Dec 2022 18:22 #259339 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read
Try these debs for 6.1 Install headers first or the image might have issues.
linux-headers
drive.google.com/file/d/1BMtApShC1G1vQZK...Ta1/view?usp=sharing
Linux-image
drive.google.com/file/d/1mjv1nKsTdPTDPcL...B0q/view?usp=sharing
Reboot
Then:
set isolcpus=2,3
if needed:
sudo apt-get install r8168-dkms
reboot
So far so good. It was hopeless without isolcpus and the r8168 driver
plasma2@plasma2:~$ halcmd show param hm2*tmax*
Parameters:
Owner   Type  Dir         Value  Name
    36  s32   RW              0  hm2_7i76e.0.read-request.tmax
    36  bit   RO          FALSE  hm2_7i76e.0.read-request.tmax-increased
    36  s32   RW         662184  hm2_7i76e.0.read.tmax
    36  bit   RO          FALSE  hm2_7i76e.0.read.tmax-increased
    36  s32   RW         391176  hm2_7i76e.0.write.tmax
    36  bit   RO          FALSE  hm2_7i76e.0.write.tmax-increased

The journey goes something like this:
sudo apt update
sudo apt install build-essential bin86 libncurses5-dev pkg-config libssl-dev
sudo apt install dwarves bison flex gnupg libelf-dev libssl-dev wget qt5-devel

wget https://cdn.kernel.org/pub/linux/kernel/v6.x/linux-6.1.tar.xz
wget https://cdn.kernel.org/pub/linux/kernel/v6.x/linux-6.1.tar.xz.sign
unxz -c linux-6.1.tar.xz | gpg --verify linux-6.1.tar.sign -
tar xvf linux-6.1.tar.xz
wget https://www.kernel.org/pub/linux/kernel/projects/rt/6.1/patch-6.1-rc7-rt5.patch
wget https://www.kernel.org/pub/linux/kernel/projects/rt/6.1/patch-6.1-rc7-rt5.patch.sign
unxz -c linux-6.1.tar.xz | gpg --verify linux-6.1.tar.sign -
gunzip patch-6.1-rc7-rt5.patch.gz
cp patch-6.1-rc7-rt5.patch linux-6.1
cat patch-6.1-rc7-rt5.patch | patch -p1
cd linux-6.1
make clean && make mrproper
make xconfig
scripts/config --disable SYSTEM_REVOCATION_KEYS
scripts/config --disable DEBUG_INFO
scripts/config --enable DEBUG_INFO_NONE
scripts/config --set-str SYSTEM_TRUSTED_KEYS ""

make -j4 deb-pkg LOCALVERSION=-linuxcnc
cd ..
sudo dpkg -i linux-image-6.1.0-rt5-linuxcnc-rt_6.1.0-rt5-linuxcnc-4_amd64.deb
sudo dpkg -i linux-headers-6.1.0-rt5-linuxcnc-rt_6.1.0-rt5-linuxcnc-4_amd64.deb

Check the deb names as they change with each version
Nice to know that a deb can work... Only one idiot needs to build it.
Last edit: 13 Dec 2022 18:22 by rodw.
The following user(s) said Thank You: IlyaKw, slukas

Please Log in or Create an account to join the conversation.

More
13 Dec 2022 18:31 #259341 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read
Sorry, I did not explain how to install the debs. The last 2 lines of the journey covers it but
hitting tab after each of these lines does it
sudo dpkg -i linux-headers 
sudo dpkg -i linux-image

Please Log in or Create an account to join the conversation.

More
15 Dec 2022 21:24 - 15 Dec 2022 21:34 #259587 by rodw
Replied by rodw on topic Mesa hm2/hm2_7i96s.0: error finishing read
Hmm, this J1900 PC
 
has the best latency I've seen in a Celeron
 

Attachment not found




Custom kernel from above, isolcpus=2,3, R8168-dkms network driver, 7i76e
and the tmax values are not too bad either.

plasma2@plasma2:~$ halcmd show param hm2*tmax*
Parameters:
Owner   Type  Dir         Value  Name
    93  s32   RW              0  hm2_7i76e.0.read-request.tmax
    93  bit   RO          FALSE  hm2_7i76e.0.read-request.tmax-increased
    93  s32   RW         649800  hm2_7i76e.0.read.tmax
    93  bit   RO          FALSE  hm2_7i76e.0.read.tmax-increased
    93  s32   RW         310776  hm2_7i76e.0.write.tmax
    93  bit   RO          FALSE  hm2_7i76e.0.write.tmax-increased

Ping times
plasma2@plasma2:~$ ping 10.10.10.10
PING 10.10.10.10 (10.10.10.10) 56(84) bytes of data.
64 bytes from 10.10.10.10: icmp_seq=1 ttl=64 time=0.664 ms
64 bytes from 10.10.10.10: icmp_seq=2 ttl=64 time=0.244 ms
64 bytes from 10.10.10.10: icmp_seq=3 ttl=64 time=0.229 ms
64 bytes from 10.10.10.10: icmp_seq=4 ttl=64 time=0.227 ms
64 bytes from 10.10.10.10: icmp_seq=5 ttl=64 time=0.229 ms
64 bytes from 10.10.10.10: icmp_seq=6 ttl=64 time=0.229 ms
64 bytes from 10.10.10.10: icmp_seq=7 ttl=64 time=0.232 ms
64 bytes from 10.10.10.10: icmp_seq=8 ttl=64 time=0.238 ms
64 bytes from 10.10.10.10: icmp_seq=9 ttl=64 time=0.230 ms
64 bytes from 10.10.10.10: icmp_seq=10 ttl=64 time=0.267 ms
64 bytes from 10.10.10.10: icmp_seq=11 ttl=64 time=0.241 ms
64 bytes from 10.10.10.10: icmp_seq=12 ttl=64 time=0.231 ms
64 bytes from 10.10.10.10: icmp_seq=13 ttl=64 time=0.235 ms
64 bytes from 10.10.10.10: icmp_seq=14 ttl=64 time=0.226 ms
64 bytes from 10.10.10.10: icmp_seq=15 ttl=64 time=0.231 ms
^C
--- 10.10.10.10 ping statistics ---
15 packets transmitted, 15 received, 0% packet loss, time 14333ms
rtt min/avg/max/mdev = 0.226/0.263/0.664/0.107 ms


Time to get it back on my machine!

The only thing wrong I can see is that if I set it up as an autostart, most of the time QTPlasmac fails to load due to a hm2 50 ms timeout. Can that be increased?
Attachments:
Last edit: 15 Dec 2022 21:34 by rodw.

Please Log in or Create an account to join the conversation.

Moderators: PCW, jmelson
  1. Hardware & Machines
  3. Driver Boards
  5. Mesa hm2/hm2_7i96s.0: error finishing read
Time to create page: 0.169 seconds
Powered by Kunena Forum