[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Autoloading WAS: normal/cmdline bug & patch

From: Marco Gerards
Subject: Re: Autoloading WAS: normal/cmdline bug & patch
Date: Sat, 19 Jun 2004 18:01:45 +0200
User-agent: Gnus/5.1006 (Gnus v5.10.6) Emacs/21.3 (gnu/linux)

"Yoshinori K. Okuji" <address@hidden> writes:

> for each loaded filesystem module:
>   try the filesystem with a specified partition/disk
>   return if successful
> for each non-loaded filesystem module:
>   load the filesystem module
>   try the filesystem
>   return if successful
>   unload the filesystem module

This is like what I proposed, with the difference of unloading the
module.  I don't think the memory usage is *that* high.

For example, there are 15 filesystems that should be tried.  The
filesystems that we have now (ext2, fat, ufs and minixfs) are all
about 4KB.  I want to make some changes (sharing code) so that will
become smaller, but there will be more complex filesystem.  Let's
assume for now the average filesystem module will be 5KB.  In that
case (15*5) 75KB will be used as max.

To me that looks like low memory usage, less than reloading those
modules all the time.  For the apple it is no problem to use a lot of
memory.  I don't think there are problems for the pc, unless we have
to use <1MB.

So I think it is not needed to unload those filesystems and that
keeping those loaded will only boost performance.  If it is required
we can use (as Tomas proposed, IIRC) a user setable variable to
indicate if modules should be unloaded or not.

> I prefer this to C, because I've seen the command "mount" in GNU/Linux 
> not maintained all the time very well. I guess this is because the 
> author of code for a filesystem may not use "mount -t auto ..." with 
> the filesystem. This would happen even in GRUB, since you wouldn't 
> notice that the autoload of your filesystem module is broken, if you 
> preload it in core or explicitly load it manually or in a config file.

I agree.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]