[lvm-devel] Why print_log() is doing the memlock check?

[Alasdair G Kergon]

On Tue, Dec 26, 2006 at 03:06:37PM +0900, Shinn'ya Hoshino wrote:
> However, when a trouble occurs in the server currently used 
> for business, it needs to avoid that it becomes impossible to 
> investigate in order that there may be no log.
 
We don't have a safe way of logging yet.

What happens, is people temporarily turn on logging during
activation in the config file while debugging, and put things
back again afterwards.
 
A long time ago I started attempting to write some such code but never
finished it.  Look for init_log_direct().
Another approach might be to cache the log messages in memory
then write them out when exiting the critical section.

Alasdair
-- 
agk at redhat.com