From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from relay02.pair.com ([209.68.5.16]) by bombadil.infradead.org with smtp (Exim 4.68 #1 (Red Hat Linux)) id 1KWv8j-0005X7-PV for linux-mtd@lists.infradead.org; Sat, 23 Aug 2008 15:34:58 +0000 Date: Sat, 23 Aug 2008 08:34:38 -0700 Subject: Re: Creating an Empty JFFS2 File System From: Grant Erickson To: "\"linux-mtd@lists.infradead.org\"" Message-ID: In-Reply-To: Mime-version: 1.0 Content-type: text/plain; charset="US-ASCII" Content-transfer-encoding: 7bit Cc: Josh Boyer , Stefan Roese List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On 3/7/08 12:39 PM, Grant Erickson wrote: > On 3/7/08 10:37 AM, Josh Boyer wrote: >> On Fri, 07 Mar 2008 09:55:13 -0800 >> Grant Erickson wrote: >>> Is creating an empty JFFS2 file system supported or is this a deprecated >>> capability? The following steps seem to be incorrect or failing: >>> >>> # flash_eraseall -q -j /dev/mtd9 >>> # mount -t jffs2 /dev/mtdblock9 /mnt >>> Cowardly refusing to erase blocks on filesystem with no valid JFFS2 >>> nodes >>> empty_blocks 0, bad_blocks 0, c->nr_blocks 3840 >>> >>> Or, is the only way to accomplish this to create a NULL JFFS2 image with >>> mkfs.jffs2 and then nandwrite it? >> >> Just erase it completely without the -j option and mount. JFFS2 will go >> through and erase them all again and add the cleanmarkers itself. > > Josh: > > That technique is a no-go as well: > > # flash_eraseall -q /dev/mtd8 > # mount -t jffs2 /dev/mtdblock8 /mnt > Cowardly refusing to erase blocks on filesystem with no valid JFFS2 > nodes > empty_blocks 0, bad_blocks 0, c->nr_blocks 3840 > mount: mounting /dev/mtdblock8 on /mnt failed: Input/output error For those following this thread via Google, I finally had time to debug this upon arrival of our prototype boards and thought I would close the thread out with the root cause. On the prototype board, which uses a Numonyx NAND01GW3B2B (128 KiB block / 2 KiB page), I found that I COULD successfully erase a NAND partition with 'nand erase' in u-boot or 'flash_eraseall' in Linux and then mount it: # flash_eraseall -q /dev/mtd9 # mount -t jffs2 /dev/mtdblock9 /mnt/root0 # df Filesystem 1k-blocks Used Available Use% Mounted on ... /dev/mtdblock9 8192 512 7680 6% /mnt/root0 However, as before, I tried this on Haleakala/Kilauea with its Samsung K9F1208U0B (16 KiB block / 512 B page) part using the same 2.6.25-rc3 DENX kernel, I saw the following: # flash_eraseall -q /dev/mtd9 # mount -t jffs2 /dev/mtdblock9 /mnt/root0 Cowardly refusing to erase blocks on filesystem with no valid JFFS2 nodes empty_blocks 0, bad_blocks 0, c->nr_blocks 512 mount: mounting /dev/mtdblock9 on /mnt/root0 failed: Input/output error My suspicion was that this was somehow related to the custom OOB/ECC layout specified in arch/powerpc/platforms/40x/kilauea-nand.c because an erase from either u-boot with 'nand erase' or Linux with 'flash_eraseall' produced identical results. The larger page part on the prototype board uses a default OOB/ECC layout (similar to Canyonlands, et al) and worked correctly in either case. So, I made the following change in arch/powerpc/platforms/40x/kilauea-nand.c in the DENX-2.6.26.2 kernel: - .ecclayout = &nand_oob_16, + .ecclayout = NULL, and an empty/erased partition mount worked, both with an erase from u-boot using 'nand erase' and from Linux using 'flash erase_all'. Stefan Roese at DENX confirmed this was likely an arch/ppc to arch/powerpc porting error/typo. Regards, Grant Erickson