On Thu, 24 Feb 2000, Tigran Aivazian wrote:
>On Wed, 23 Feb 2000, George wrote:
>> The alternative to devfs is what we already have, a working (maybe not
>> perfect, but working) /dev. And if you already have /dev (and use it),
>> what's the point of wasting memory on devfs which would duplicate things?
>
>Once upon a time we had a working (not perfect but working) aout-based
>Linux systems. What was the point of switching to ELF? Lots of points. So
>it is with devfs - it solves a lot of problems, all nicely documented in
>Documentation/filesystems/devfs/README.
Read it, wasn't impressed.
"On a modest system with 64 MBytes of RAM, the cost is under 0.1 percent.
The accusations of "bloatware" levelled at devfs are not justified."
By my calculations, that's 67,109 bytes. That is larger than even
'scsi_mod'.
Module Size Used by
sb 33908 0 (autoclean)
uart401 5968 0 (autoclean) [sb]
sound 57368 0 (autoclean) [sb uart401]
soundlow 224 0 (autoclean) [sound]
soundcore 2340 6 (autoclean) [sb sound]
bttv 33784 1 (autoclean)
tuner 2092 1 (autoclean)
i2c 3328 2 (autoclean) [bttv tuner]
videodev 2368 3 (autoclean) [bttv]
ppp_deflate 40548 1 (autoclean)
ppp 20716 2 (autoclean) [ppp_deflate]
slhc 4268 1 (autoclean) [ppp]
ide-cd 23220 1 (autoclean)
sr_mod 17992 0 (autoclean) (unused)
cdrom 13336 0 (autoclean) [ide-cd sr_mod]
scsi_mod 52032 1 (autoclean) [sr_mod]
af_packet 5836 1 (autoclean)
tulip 31680 1 (autoclean)
Even if we say 0.5%, it's still 33,554 bytes. I think one of Alan Cox's
estimates (I could be wildly wrong, it's from memory) for /proc was ~68 kB.
So I fail to see a large improvement being 0.1% of a 64 MB machine. (I'd
give numbers but I haven't run the 2.3.x kernel with devfs yet.)
"A typical /dev has over 1200 nodes!"
So the admin is lazy. Besides, it takes 0 memory when not being used.
Redhat 6.1 default /dev:
root@moving:~# ls -l /dev/ | wc
2311 23131 141939
After deleting most of what I don't need:
root@moving:~# ls -l /dev/ | wc
263 2640 16265
Currently all options suck equally. I'm not advocating /proc for the data
(that'd be silly) but I am against a wholesale change from one dump to
another in the name of being "pretty" wit a more significant fix.
-George Greer
-
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.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 21:00:11 EST