[CentOS] info wanted about meaning of boot messages ...

Thu Aug 25 21:21:28 UTC 2016
ken <gebser at mousecar.com>

On 08/25/2016 02:42 PM, Walter H. wrote:
> On 25.08.2016 20:24, ken wrote:
>> On 08/25/2016 12:08 PM, Walter H. wrote:
>>> Hello,
>>> I've got CentOS 6.8 x64, updated today to the latest by 'yum update'
>>> this installed a new kernel: 2.6.32-642.4.2.el6.x86_64
>>> in /var/log/boot.log I found these 3 lines ...
>>> No kdump initial ramdisk found. [WARNING]
>>> Rebuilding /boot/initrd-2.6.32-642.4.2.el6.x86_64kdump.img
>>> cp: cannot stat `/lib/firmware/i915/bxt_dmc_ver1.bin': No such file 
>>> or directory
>>> the first two are logic to me, but the 3rd line, did there something 
>>> fail at the update?
>>> Thanks,
>>> Walter
>> 'stat' is a command.  It's like 'ls', but gives more info.  Try it.  
>> The message is saying simply that the file can't be found. It looks 
>> like the install script was trying to 'cp' that file.
> the directory from above shows with 'ls -al /lib/firmware/i915/' this:
> total 156
> drwxr-xr-x.  2 root root   4096 Aug 25 10:08 .
> drwxr-xr-x. 46 root root  12288 Aug 23 17:28 ..
> -rw-r--r--.  1 root root   8824 Aug 23 21:14 skl_dmc_ver1.bin
> -rw-r--r--.  1 root root 128320 Aug 23 21:14 skl_guc_ver4.bin
> means, that the file from above message isn't there ...
> when I do  'cat /etc/rc.d/init.d/* | grep "bxt"' there is nothing 
> shown; from where did this cp come from above's error message?
> Thanks
> Walter 

Walter, it would seem then that one of the boot scripts is calling 
another script [...] which is then calling another script which is 
yielding the boot message.  I gave you the [...] because there could be 
several layers of such wrappers.  So it might well take a bit of 
drilling down and poking around to find the source of that boot message 
from that end.

You might also try 'rpm -qf /lib/firmware/i915' to see if that narrows 
down the sought executable to a specific rpm.  Then do 'rpm -ql 
[package_name]' to get a listing of the files in that package.

In some cases you can use 'trace', but I'm almost always found this to 
be a source of too much information.

If at this point you're saying, "there's gotta be an easier way", well, 
you're right.  Maybe in the sequel to Linux there will be.

Have fun.