From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Woodhouse To: "Koenders, Eric - NL" In-Reply-To: References: Content-Type: text/plain Date: Tue, 21 Dec 2004 11:37:37 +0000 Message-Id: <1103629058.6111.53.camel@localhost.localdomain> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Cc: linux-mtd@lists.infradead.org Subject: Re: System blocked for a while by heavily fragmented file on a jffs2 partition List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Thu, 2004-12-09 at 17:17 +0100, Koenders, Eric - NL wrote: > Does anyone have an idea how to solve this problem? Maybe more agressive > defragmentation behaviour of the GC? Or an fsck like tool to defragment a > file system? It would help to concentrate harder on garbage-collecting the files with tiny nodes. Currently we do keep track of whether a node is an optimal representation ('REF_PRISTINE') or not, but we only use that when we're _actually_ garbage collecting; we don't use it to decide which blocks to GC. Also, the oft-suggested optimisation of writing new nodes out to a _separate_ block from GC-rewritten nodes would probably help to improve the GC performance here. -- dwmw2