From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from majordomo by infradead.org with local (Exim 3.03 #1) id 11xv0y-00031q-00 for mtd-list@infradead.org; Tue, 14 Dec 1999 16:41:28 +0000 Received: from mail1.danielind.com ([12.19.96.6]) by infradead.org with esmtp (Exim 3.03 #1) id 11xv0w-00031k-00 for mtd@imladris.mvhi.com; Tue, 14 Dec 1999 16:41:26 +0000 Message-ID: <385673EA.BB0EE434@danielind.com> Date: Tue, 14 Dec 1999 10:44:26 -0600 From: Vipin Malik MIME-Version: 1.0 To: MTD Subject: [Fwd: [Fwd: Power Down]] Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-mtd@imladris.demon.co.uk List-ID: David Woodhouse wrote: > > vmalik@danielind.com said: > > Is ext3 already available in beta form? Is it included in the latest > > 2.3.x kernels? > > yes, no. > > ftp.uk.linux.org:/pub/linux/sct/fs/jfs/ext3-0.0.2c.tar.gz Thanks for the link. I checked ext3 out. what I found... 1. It is still in pre form. Not too bad, but probably at least 6 months (or more) away from being production quality. Not too big a deal for projects about 1+yr out. 2. This is the big one: It needs 1024 blocks for the journal file. At 1k/block, this translates into 1 meg journal file. This could be a *significant* overhead for small file systems (<10M). I'm going to download the pre and play with it for a while to get a first hand impression of it. But the 1 meg journal file overhead is too big for medium size embedded systems. Unfortunately journalling still dosen't solve the problem (that I say first hand) of one entire (512byte) flash block "go bad". The only solution I found to recover this block resulted in the erasure of the ENTIRE block. Now this was at the lowest level (below the file system). I don't know how even a journaled file system would recover stuff (inodes etc.) disappearing from under its feet. Dosen't give me a warm and fuzzy about it. > > > What is the overhead of Journalling? (CPU AND flash space). Of course > > if that solves the problem, then that is the most important thing (for > > me at least). > > I don't know - I haven't looked into it much yet. > > vmalik@danielind.com said: > > One dosen't have to write to the flash every second, even if logs are > > being generated that fast. You can accumulate writes in RAM (even > > backud up RAM), and then archive them to FLASH every 5 minutes etc. > > vmalik@danielind.com said: > > When I write a log, I want it written and saved. I guess one could do > > sync() etc. every time but not very elegant. > > Err... which do you want? You can't have both. LOL :) Actually what I meant was that in the write to RAM and then in batch mode save to FLASH situation, I meant that when I want to store the "batch" of logs accumulated in RAM to FLASH I want the thing "done!" rather than "hang around" in VFS buffers etc. > > -- > dwmw2 To unsubscribe, send "unsubscribe mtd" to majordomo@infradead.org