From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from dragon.actrix.co.nz ([203.96.16.164]) by pentafluge.infradead.org with esmtp (Exim 3.22 #1 (Red Hat Linux)) id 18eS2J-000368-00 for ; Fri, 31 Jan 2003 03:40:16 +0000 Content-Type: text/plain; charset="iso-8859-1" From: Charles Manning Reply-To: manningc2@actrix.gen.nz To: "Alex Samoutin" , Subject: Re: NAND subsector writing problem Date: Fri, 31 Jan 2003 17:01:05 +1300 References: <007801c2c7fc$5834da60$1a00a8c0@itc.intrinsyc.com> <20030130210325.0F2BB405E@tiger.actrix.co.nz> <004701c2c8ce$107caea0$1a00a8c0@itc.intrinsyc.com> In-Reply-To: <004701c2c8ce$107caea0$1a00a8c0@itc.intrinsyc.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Message-Id: <20030131041111.5DE53151BD@dragon.actrix.co.nz> Sender: linux-mtd-admin@lists.infradead.org Errors-To: linux-mtd-admin@lists.infradead.org List-Help: List-Post: List-Subscribe: , List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: Alex I assume you're using the TC58128AFT since there is no TC5818AFT. I don't spend time in the mtd, only in YAFFS, but I can think of a few things to look at. No doubt other people can think of more things... When I first looked at the NAND mtd a long time back I noticed that the verify did not always work properly on NAND, giving false errors. This I tracked down to a difference between NOR and NAND. In NOR, a write fails if try to overwrite a zero with a one bit. In NAND, the operation does not fail even though the bit does not change. ie. Say you start with 12FF56. If you write FF34FF You now have 123456. The NAND part will say the write was OK. The verify might fail since FF34FF != 123456. Thus, problems might arise if the page is not clean. I've never used nandtest.c, so I am unsure as to whether it is expected to work with verify enabled. It might be worth double-checking your chip commanding/handling (especially the address handling) to ensure that you're doing the right things for this part. It seems possible to me that there might be some address aliasing issue going on. Perhaps you can hack nandtest to check that the page is empty before you write to it, then read it back in the case of a failure and look at what you got back. I'm sorry I can't speak to the JFFS2 questions as I only understand the YAFFS file system. -- Charles