Re: DevFS and "c00lness" (Re: Grub)
On Tue, 6 Nov 2001 10:42, Ethan Benson wrote:
> On Tue, Nov 06, 2001 at 08:25:08PM +1100, Daniel Stone wrote:
> > You're claiming this is optimal? 1221 is still far, far too many to sort
> who cares!
> eb@plato ~$ time ls -l /usr/bin | wc -l
> real 0m0.427s
> user 0m0.120s
> sys 0m0.210s
> there zero length files ffs! FUD!
In normal operation you don't need to do "ls -l /usr/bin".
All entries in /usr/bin have (or should have) man pages describing them and
how to use them.
The only documentation for /dev is /usr/src/linux/Documentation/devices.txt,
and it is not very useful documentation unless you do kernel development or
maintain the makedev or devfsd package.
On the Portslave list a user had a problem that only /dev/ttyS0 ..
/dev/ttyS31 were working. He had created all the device nodes for all 192
devices listed in devices.txt but they didn't work.
After 20 minutes reading the serial.c source I figured out that the 2.2.19
kernel only supports 32 ttyS* devices, and the 2.4.13 kernel supports 64.
This would have been obvious to the user if they had used devfs. I made some
suggestions as to how to patch the 2.2.19 kernel. But then when testing the
kernel the user will have a more difficult time because they can't do
"ls -l /dev/tts" to see how many ports were recognised.
Now on the old fashioned /dev if the node doesn't exist then you need to
create it (which means reading devices.txt or the documentation that came
with the device driver). Then if the device driver isn't loaded you get
messages in syslog like char-major-108 not found which is not nearly as clear
as an error referencing "/dev/ppp"! This saves time for experienced users as
well as making things possible for newbies.
http://www.coker.com.au/bonnie++/ Bonnie++ hard drive benchmark
http://www.coker.com.au/postal/ Postal SMTP/POP benchmark
http://www.coker.com.au/projects.html Projects I am working on
http://www.coker.com.au/~russell/ My home page