[OpenBMC] Change the bmcsetup code back to use raw ipmi commands for "access on"

[whowutwut]

Revert the bmcsetup code back to raw commands from access on

I'm not sure the history about why we had changed the raw command back to access on , it seemed like it was working at some later build of OpenBMC...

• We first set it here: enhance bmcsetup logic for openbmc #2990
• Then undid it here: enhance bmcsetup for WSP #3833

But looking at openbmc/openbmc#2492, it seems that certain commands will not take effect without the raw command (Setting IPSRC to DHCP)

[whowutwut]

Doing some testing here regarding https://w3.rchland.ibm.com/projects/bestquest/?defect=SW407037 Also seems like using the raw command does work better than access on

@zet809 Do you remember why we changed it back?

[xCAT Genesis running on mid05tor12cn05 /]# time /bin/bmcsetup                    
ipmitool version 1.8.18
Invalid channel: 16
/bin/bmcsetup: line 150:  2643 Terminated              allowcred.awk
xcat.genesis.bmcsetup: BMC Information obtained from xCAT
xcat.genesis.bmcsetup: NUMBMCS=1 ==> BMC IP=172.11.139.5/255.255.0.0, GW=172.11.253.27, VLAN=11
xcat.genesis.bmcsetup: IPMIVER=2.0, IPMIMFG=42817, XPROD=16975
xcat.genesis.bmcsetup: Auto detecting LAN channel...
xcat.genesis.bmcsetup: Detected LAN channel 1
Setting LAN IP Address to 172.11.139.5
Setting LAN Subnet Mask to 255.255.0.0
Setting LAN Default Gateway IP to 172.11.253.27
Unable to send RAW command (channel=0x0 netfn=0x6 lun=0x0 cmd=0x40 rsp=0xce): Command response could not be provided

real	0m22.008s
user	0m0.220s
sys	0m0.066s

By doing ipmitool -d 0 raw 0x06 0x40 $LANCHAN 0x42 0x44 instead of access on the networking DOES in fact

[whowutwut]

Here's a patch file:
bmcsetup.4242.patch.txt

To patch your system running 2.13.8

1. Download patch file to: /tmp/bmcsetup.4242.patch.txt
2. cd /opt/xcat/share/xcat/netboot/genesis/ppc64/fs/usr/bin
3. patch -p0 < /tmp/bmcsetup.4242.patch.txt

[hu-weihua]

relate to task #4352

[hu-weihua]

@whowutwut, we use manual test to cover node discovery test, including bmcsetup. This test won't run daily. So even we add test step for this issue, this problem won't be found as quick as daily regression.

[whowutwut]

@hu-weihua Thanks for the comments, I removed the label, I am trying to think of a way to test it and it's not that easy for automation