Boot messages not the same as "dmesg" or "/var/log/messages"?
 

Hello,

Being in the process of building a Debian-sid system, I curently finished installing the 2.6 kernel.

However, I need to make sure that it "sits" properly before continuing with the rest of the instalation (X-windows, KDE, etc)

The problem is that inspection of "dmesg" or "/var/log/messages" shows that all is alright.

However, when the system boots and the boot text flies-by I get a glimpse of some errors (some of them even "fatal errors" ?) but they are not registered in /var/log/messages .

Why is this happening? How can I record all the boot-output verbatim to the /var/log ?


Thanks

/var/log/boot.* are the boot up logs

hmm... there is no "/var/log/boot" on this Debian box.

well, its not exactly /var/log/boot, there are three files. boot.log boot.msg boot.omsg

The "var/log" directory does not contain anything with the word "boot".

As far as I know at this point "dmesg" and "var/log/messages" are the only ways to inspect the boot messages.

I know "dmesg" has an (n level) option possibly to control verbosity, but the man pages neglect to tell us what the range of those "levels" is.

Does anyone know the answer to this ? Is this the limit of what can be recovered from boot ? (It is 110% confirmed that the boot messages are more verbose)

Is it a matter of increasing the buffer size ? (I thought the default buffer for dmesg already matches the kernel ring buffer ?)

The loglevel definitions are in the syslog(2) manpage, also apparently in kernel.h. :)

Sounds like your syslog or initlog or whatnot is either getting called with a low log level, or it's not sending its output to the correct files. Either way, I don't know Debian very well, so I can only hazard a guess that your boot log daemon gets called in one of your /etc/rc scripts. I know those files also sometimes pass additional parameters to the kernel... But I can't imagine I'm going to be much help at this point, so I'll shut up now.