lkml.org 
[lkml]   [1998]   [Aug]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: DEVFSv50 and /dev/fb? (or /dev/fb/? ???)
On Tue, Aug 11, 1998 at 08:15:47PM -0400, Andrew J. Anderson wrote:

> So there is precedence for a "different" format for filesystem
> information. (Even though I would call AIX more "like-unix" than
> "unix-like". ;-)

Does AIX have getmntent? If so, what does its manual page say?

> I know of a busy POP server that had to increase their connection
> rate above 40/sec, and their daemon syslog's every connection (via
> /dev/log), so there's at least one documented case. And let's not
> forget sendmail and ftp (in some setups), as well.

I'm not sure why POP3 is using ptys, at least my code here doesn't.

Anyhow, I've since actually tried to measure how often access to
/dev/something occurs (by running strace -f on a live system for a
while).

Something in the sendmail/deliver process opens /dev/null for each
message delivered. Since sendmail is so horrible slow, I hardly
think this is an issue yet.

openlog(3) is used by many many applications, but most daemon, etc.
open the log before forking, so we see only lookup when the daemon
starts.

> My other thought is a busy telnet server and pty's. Once you get
> up there in users, it may be quite often, perhaps as much as
> several times a minute.

Here, I would expect the forking/exec overhead to drown out any
lookup times in /dev/.

> So if sorting though the /dev directory becomes unreasonable with
> many devices, then it truely could become an issue for some
> applications. I believe that I saw a time of up to 0.5 seconds
> wall clock posted earlier for an 'ls' on /dev (just tried it here
> -- 0.47 for 'ls', 3.94 for 'ls -l'). Consider a 0.5 second delay
> for a busy mail hub or pop server or ftp site per message
> syslog'ed. Store and forward, anyone? :)

I get 400ms on a fairly busy /dev/, but if I pipe the output to
/dev/null its only 40ms or so.

Anyhow, /dev/ needs a cleanup. I personally would like to see little
directories within /dev to make things pretty... a flat /dev isn't
necessary or even very useful.

Anyhow, I don't forsee a time when we will ever have thousands of
inode in /dev, ever.

> Or things like updating /etc/mnttab, and other /etc files are what
> I wondered about.

Presently, /etc/fstab and /etc/mtab{~} are just plain wrong when
using a F) filesystem. I think some version of mount (perhaps
RedHat-5.1?) uses /proc/mounts instead of /etc/mtab, not sure about
/etc/fstab.
Maybe its time the whole issue of /dev, /etc/mtab, /etc/fstab, need a
good thorough re-think?



-cw


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html

\
 
 \ /
  Last update: 2005-03-22 13:43    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site