From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Woodhouse To: krishna In-Reply-To: <42E079B8.9070703@globaledgesoft.com> References: <42E079B8.9070703@globaledgesoft.com> Content-Type: text/plain Date: Fri, 22 Jul 2005 11:55:42 -0400 Message-Id: <1122047742.12630.10.camel@localhost.localdomain> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Cc: linux-mtd@lists.infradead.org, Linux Kernel Subject: Re: why is jiffies 128 in jffs2_find_gc_block() in gc.c of jffs2 List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Fri, 2005-07-22 at 10:14 +0530, krishna wrote: > I am not clear why the hardcoded values are 50, 110 and 126 > and why is jiffies moded with 128, why not any other value. It's just a way to achieve 'randomness' which doesn't actually consume entropy and which is quick to obtain. It only needs to be relatively evenly distributed. We use it for selecting the next eraseblock to be garbage-collected. 50/128 of the time we pick a block from the eraseable_list, 60/128 of the time we pick a block from the very_dirty_list, 16/128 of the time we pick a block from the dirty_list, and the remaining 2/128 of the time we pick a block from the clean_list for garbage collection. The precise numbers don't have a huge amount of science behind them; they are mostly guesses about what would achieve evenly-distributed wear levelling over time without garbage-collecting clean blocks more often than is necessary. If you want to conduct tests with various workloads in order to refine the fairly primitive algorithm described above, that could be useful. -- dwmw2