[prev in list] [next in list] [prev in thread] [next in thread] 

List:       tux2-dev
Subject:    Re: [tux2-dev] tracking written data
From:       Daniel Phillips <phillips () bonn-fries ! net>
Date:       2001-07-09 23:15:43
[Download RAW message or body]

On Monday 09 July 2001 16:11, Andrew Morton wrote:
> I assume that tux2 needs to know when data (as opposed to metadata)
> blocks have been written to disk?

Yes, that's done in ext2_get_block in the 2.2.13 prototype, the 
"create" parameter being the signal to treat the operation as a write.

> If so, how were you intending to keep track of all the pages/buffers?

The 2.2.13 prototype uses a linked list in malloc'ed memory.  I suppose 
with an additional year of experience behind me I should feel obliged 
to come up with some clever variant that saves this memory by 
repurposing one of the existing list fields, for example, the buffer 
lru list field, which is available because Tux2 submits the buffers 
directly for writing and doesn't just wait around for kflush (oops, 
sorry, bdflush) or kupdate to do its thing.

I will most likely have to hack the vfs a bit to stop it from doing 
irresponsible things to buffers it knows nothing about, but I'm not as 
shy about doing that as I once was.  Most fs junki^H^H^^H^H developers 
would likely agree that this part of buffer.c (sync, fsync) is a little 
broken anyway.

--
Daniel
_______________________________________________
tux2-dev mailing list
tux2-dev@list.innominate.org
http://innominate.org/mailman/listinfo/tux2-dev

[prev in list] [next in list] [prev in thread] [next in thread]