Re: [PATCH] larger inodes for ext3 at install time

[Eric Sandeen <sandeen redhat com>]

Jeremy Katz wrote:
> On Wed, 2007-10-24 at 10:25 -0400, Peter Jones wrote:
>> Eric Sandeen wrote:
>>> Jeremy Katz wrote:
>>>> On Tue, 2007-10-23 at 11:40 -0500, Eric Sandeen wrote:
>>>>> Larger inodes should help out with selinux and/or beagle attributes a bit
>>>>> (xattrs can be stored in-inode) and also makes upgrade to ext4 possible
>>>>> (which needs an in-inode extent list)
>>>> So the obvious question is "if this is so useful, why not have mke2fs do
>>>> the right thing without us having to pass more arguments to it"?
>>> Ok, I can do that if you prefer.  Although if you don't plan to migrate
>>> to ext4, and don't use xattrs, you might want the smaller inodes to save
>>> space, since inode overhead on ext$FOO is nontrivial.
>>>
>>> I'm not sure where upstream is going on this one, but I can sort it out
>>> if you'd rather not have it in anaconda.
>> Why not do something like the patches at 
>> http://people.redhat.com/pjones/fsprofile/ instead, so next time the 
>> ext3 guys come up with a new option, we won't have to explicitly add 
>> support for it as well?
>>
>> Granted, to be truly useful we either need "include" directives in 
>> mke2fs.conf or a command line argument to specify a config file.  Either 
>> (or both) would allow a user to put mke2fs.conf in an updates.img , and 
>> so they could define their own options that way.
> 
> This only makes a difference if e2fsprogs ships a config file that has
> the new options included in it.  If they don't, we *still* have to make
> change, just now it's hidden away in some copy of a config file which is
> going to atrophy over time

I agree, anaconda should do what e2fsprogs does by default.

Sorry for the noise.  :)

-Eric