mass boot octeon nics from dhcp/tftp

dnsmasq-d-lio.conf

# add one entry per port per card
dhcp-host=00:0f:mac1,192.168.236.100,lios1octeth0,12h
dhcp-host=00:0f:mac2,192.168.236.101,lios1octeth1,12h

# match the vendor pattern to a tag
# everything later matches to this tag
dhcp-mac=set:lio,00:0f:b7:*:*:*

# doku zu options: https://blog.abysm.org/2020/06/human-readable-dhcp-options-for-dnsmasq/

# sntp support etc ermoeglichen
dhcp-option = tag:lio, option:router, 192.168.236.1
dhcp-option = tag:lio, option:ntp-server, 192.168.236.1
# aber auch einfach eine uhrzeit hergeben (nicht supported im build
# dhcp-option = tag:lio, option:time-offset


# bootcommand:
# "  * Added back NetOurBootCommand environment variable as DHCP option 224." braucht u-boot aus sdk 5.1!
dhcp-option-force=224,"tftp ; bootoctlinux $(loadaddr) numcores=16 ip=dhcp console=ttyS0,115200 console=ttyPCI0 rootwait=10 panic

# bootfile: dieses bootet das embedded rootfs mit 4.9 kernel (this works, but success also depends on the console settings
# having a pci stdout blocks boot
# having no pci stdout means you can't interrupt autoboot
# you'll figure it out, it just sucks because we don't know how to do it right
dhcp-boot=tag:lio,erfs/vmlinux.64

# umstellen auf mkimage uboot script
#dhcp-boot = tag:lio, ...



#bootoctlinux ${loadaddr} coremask=f root=/dev/nfs rw console=ttyS0,115200 ip=bootp
#bootoctlinux ${loadaddr} numcores=n root=/dev/ram console=ttyS0,115200
# file:///u_boot.pdf
# idk what that pdf was, seems I thought it matters.

env-reset-script

# docs: https://docs.u-boot.org/en/latest/usage/environment.html

export OCTEON_REMOTE_PROTOCOL=PCI:2

LINES="setenv autoload n
setenv bootdelay -1; setenv baudrate 115200; setenv env_size 2000
setenv fdtaddr 80000; setenv fileaddr 0x20000000; setenv filesize 2000
setenv flash_base_addr 1fb00000; setenv flash_size 100000;
setenv liofsize 0x142840; setenv lionicaddr 0x400000; setenv loadaddr 0x20000000
setenv named_block_addr 0xfffeff80; setenv named_block_size 0x10000
setenv octeth0_speed 10
setenv octeth1_speed 25
setenv qlm4_mode sata"


oct-remote-bootcmd "env default -a"

echo "${LINES}" | \
while read _line ; do
   oct-remote-bootcmd "${_line}"
   sleep 3
done

oct-remote-bootcmd "saveenv"

#setenv bootcmd 'run uefi_bootcmd; sf probe; sf read ${fileaddr} ${lionicaddr} ${liofsize}; bootoct forceboot numcores=${numcores}'
#setenv uefi_bootcmd 'namedalloc __stage2_uefi 0x10000; cp64.b 0x1fb20000 ${named_block_addr} ${named_block_size}'

env.dhcp

stdout=serial
stderr=serial
stdin=serial,bootcmd
bootdelay=0
autoboot=0
autostart=yes
bootretry=100
octeth0_speed=10
octeth1_speed=10

env.tftp

bootdelay=10
autoboot=0
autostart=yes
bootretry=100
pci_console_active
serial_console_active=1
stdin=pci,serial,bootoct
stdout=serial
stderr=serial
octeth0_speed=10
octeth1_speed=10
tftpwindowsize=16
tftpblocksize=1500
bootcmd=dhcp ; tftp ; bootoctlinux coremask=ffff root=/dev/ram rootwait=5 console=ttyS0,115200 console=ttyPCI0 ip=dhcp reboot=pci rootwait=10 panic=10 numcores=16 mem=8G
saveenv

env.tftp-pci

bootdelay=10
autoboot=0
pci_console_active=1
serial_console_active=1
stdin=pci,serial,bootoct
stdout=serial,pci
stderr=serial,pci
octeth0_speed=10
octeth1_speed=10
bootcmd=mmc init; dhcp ; tftp ; bootoctlinux numcores=14 root=/dev/ram rootwait=5  console=ttyS0,115200 ip=dhcp reboot=pci rootwait=10 panic=10 mem=7G
saveenv
[root@lion firmware]# cat env.tftp
bootdelay=10
autoboot=0
autostart=yes
bootretry=100
pci_console_active
serial_console_active=1
stdin=pci,serial,bootoct
stdout=serial
stderr=serial
octeth0_speed=10
octeth1_speed=10
tftpwindowsize=16
tftpblocksize=1500
bootcmd=dhcp ; tftp ; bootoctlinux coremask=ffff root=/dev/ram rootwait=5 console=ttyS0,115200 console=ttyPCI0 ip=dhcp reboot=pci rootwait=10 panic=10 numcores=16 mem=8G
saveenv

load_fw_all.sh

#!/usr/bin/env bash

load_fw(){
    cd /root/firmware
    # lade und warte 20s mit dem boot
    OCTEON_REMOTE_PROTOCOL=PCI:${1} oct-remote-boot /tftpboot/u-boot-octeon_nic225e.bin --envfile=env.dhcp
    #if [ $? != 0 ]; then
    #  OCTEON_REMOTE_PROTOCOL=PCI:${1} timeout 20 oct-remote-reset
    #  OCTEON_REMOTE_PROTOCOL=PCI:${1} oct-remote-boot /tftpboot/u-boot-octeon_nic225e.bin --envfile=env.dhcp
    #fi
    # reset falls sfp xaui problem mit erkennung dac da waere, anderes envfile
}

run_boot(){
    OCTEON_REMOTE_PROTOCOL=PCI:${1} oct-remote-bootcmd "dhcp"
    sleep 10
    OCTEON_REMOTE_PROTOCOL=PCI:${1} oct-remote-bootcmd "run NetOurBootCommand"
}


# reihenfolge im treiber entspricht nicht der in den slots
# und kann sich aendern
for _cnt in 0 2 4 6 8 10 12 14 16 18 ; do
#for _cnt in 0; do
   sleep 2
   clear
   echo "lade auf adapter #${_cnt}"
   load_fw ${_cnt} &&
   echo "geladen auf adapter #${_cnt}" &&
   sleep 5
   #echo "boote adapter #${_cnt}" &&
   #run_boot ${_cnt} &
done

# alt
#    OCTEON_REMOTE_PROTOCOL=PCI:${1} oct-remote-bootcmd "dhcp ; tftp ; bootoctlinux coremask=ffff root=/dev/ram rootwait=5  console=ttyS0,115200 console=ttyPCI0 ip=dhcp reboot=pci rootwait=10 panic=10 numcores=16 mem=8G"

if you wana make progress on this, pay attention to and learn about that dhcp option 224 they use to get a uboot bootscript.

the second largest wound is rebooting. so booting, and rebooting both still suck. after that things will get better.
long-term suck is power management, I have some exact power draw with the cards at u-boot prompt and running kernel 4.9. maybe in 6.10+ it finally works

further suck is the port LEDs, not working on 4.9, can be a device tree or a driver issue. no clue

I'll just keep adding to this random gist instead of my issue tracker

zip offloading does not yet work (i left out 2 cores, what else do you need!)

/examples # ./zip abc 
CVMX_SHARED: 0x102d0000-0x103b0000
Active coremask = 0x3fff
ERROR: cvmx_fpa3_reserve_aura: node=0 desired aura=5
ERROR: cvmx_fpa3_setup_aura_and_pool: Failed in AURA not available
appmain: ERROR: in ZIP buffer pool

this matters i.e. if you wanted to use zram (gzip compressed on multiple paging devices should be fast and low-lat)

crypto example works but ofc /proc/crypto doesn't indicate any offloading in the rest of the system.

"with an engineering mindset" this is a great state because it's just individual, unconnected small issues which are supposed to be easier to be solved.

let us celebrate, let us all make merry.
everyone!
let's shout!
scream out!