lkml.org 
[lkml]   [1996]   [Sep]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: NFS as a module / kerneld question
David Flood writes:

> And if these processes are needed, how come they just go to sleep
> and never wake up?

Because they're kernel mode processes, not user mode processes.
Normal rules don't actually apply.

What they do is NFS asynchronous reading. Read fs/nfs/README in the
kernel source tree for more details. The number of nfsiod processes
is equivalent to the number of asynchronous reads that can be in
progress at once.

They aren't actually asleep; just all the code they execute is in
kernel space rather than user space.

> And when another mount command is executed, a new one is forked to
> handle the new mount rather than using one of the existing
> processes? I can see nfs.o forking one process since it gets
> loaded due to a mount command and that one nfsiod is necessary to
> handle that mount. But what about the other 3? And when every nfs
> mount is unmounted, the original 4 don't terminate. They just hang
> around. As a result the use count of nfs.o is always some number
> that is equal or greater than 4.

The number of them is nothing to do with the number of mounts, but the
number of async reads that can occur at once. The fact that each one
adds 1 to the nfs.o use count seems like a bug to me, though.

Basically they're a NFS performance enhancement. Most other UNIX
systems have them too; for example, SunOS 4.1 has 'biod' which does a
similar job.

-Matt

\
 
 \ /
  Last update: 2005-03-22 13:37    [W:0.091 / U:9.564 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site