From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from rwcrmhc13.comcast.net ([204.127.198.39]) by pentafluge.infradead.org with esmtp (Exim 4.30 #5 (Red Hat Linux)) id 1AYVSd-0001hv-1H for linux-mtd@lists.infradead.org; Mon, 22 Dec 2003 19:11:23 +0000 Date: Mon, 22 Dec 2003 14:09:06 -0500 From: "George G. Davis" To: linux-mtd@lists.infradead.org, linux-arm-kernel@lists.arm.linux.org.uk Message-ID: <20031222190906.GB20778@mvista.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Subject: integrator-flash Chip reports voltage low on erase List-Id: Linux MTD discussion mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Greetings, First, I'm intentionlly cross posting this because portions of the relevent code are contained in both MTD CVS HEAD and linux-2.6.0-rmk1 patch. I submitted a fix for this problem against linux-2.6.0-test10-rmk1 here: http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=1728/1 However RMK has requested further discussion for this issue. So here goes... The ARM Integrator/AP (see drivers/mtd/maps/integrator-flash.c) has a single contiguous logical bank of flash memory comprised of 4 physical banks x 2 x 256kib x 16-bit Intel DT28F320 devices. At least this is the case for my ARM Integrator/AP rev D. For my particular hardware rev of the ARM Integrator/AP, I get "Chip reports voltage low on erase" errors while running linux-2.6.0-rmk1 + MTD CVS HEAD as of 19Dec03 +/-. The same problem also exists in past revs of linux 2.4, 2.6 and MTD CVS. So this is not a regression in recent kernels or MTD. Here are the flash details from kernel init messages: armflash0: Found 2 x16 devices at 0x0 in 32-bit mode armflash0: Found 2 x16 devices at 0x800000 in 32-bit mode armflash0: Found 2 x16 devices at 0x1000000 in 32-bit mode armflash0: Found 2 x16 devices at 0x1800000 in 32-bit mode Intel/Sharp Extended Query Table at 0x0031 cfi_cmdset_0001: Erase suspend on write enabled Using buffer write method 4 cmdlinepart partitions found on MTD device armflash0 Creating 4 MTD partitions on "armflash0": 0x00000000-0x00400000 : "kernel" 0x00400000-0x01000000 : "rootfs" 0x01000000-0x01fc0000 : "userfs" 0x01fc0000-0x02000000 : "sib" FWIW, I added MTD cmdlinepart support as follows (because I am not currently using RedBoot firmware on this target and I abhor the ARM afs support 8 ): Index: drivers/mtd/maps/integrator-flash.c =================================================================== RCS file: /home/cvs/mtd/drivers/mtd/maps/integrator-flash.c,v retrieving revision 1.14 diff -u -r1.14 integrator-flash.c --- drivers/mtd/maps/integrator-flash.c 11 Oct 2003 10:00:31 -0000 1.14 +++ drivers/mtd/maps/integrator-flash.c 22 Dec 2003 17:55:05 -0000 @@ -65,7 +65,7 @@ info->plat->set_vpp(on); } -static const char *probes[] = { "RedBoot", "afs", NULL }; +static const char *probes[] = { "cmdlinepart", "RedBoot", "afs", NULL }; static int armflash_probe(struct device *_dev) { Here's the linux cmdline and flash partition details for this test case: cat /proc/cmdline console=ttyAM0,38400n8 noinitrd root=/dev/nfs ip=bootp mtdparts=armflash0:0x00400000@0x00000000(kernel),0x00c00000@0x00400000(rootfs),0x00fc0000@0x01000000(userfs),0x00040000@0x01fc0000(sib) cat /proc/mtd dev: size erasesize name mtd0: 00400000 00020000 "kernel" mtd1: 00c00000 00020000 "rootfs" mtd2: 00fc0000 00020000 "userfs" mtd3: 00040000 00020000 "sib" And here's the test case with results: mkdir -p /initrd mkdir -p /mnt/mtdblock1 eraseall /dev/mtd1 mount -oloop /boot/initrd.ext2fs /initrd/ mount -tjffs2 /dev/mtdblock1 /mnt/mtdblock1 cd /initrd/ find . -mount -depth--print0 | cpio -padmu0 /mnt/mtdblock1 lock1 Chip reports voltage low on erase: status 0xa800a8 Erase at 0x003e0000 failed immediately: errno -5 Chip reports voltage low on erase: status 0xa800a8 Erase at 0x003c0000 failed immediately: errno -5 Chip reports voltage low on erase: status 0xa800a8 Erase at 0x003a0000 failed immediately: errno -5 . . . ad nauseam! A fix for this problem can be found at the URL noted above. Note that the fix is against the linux-2.6.0-rmk1 patch sice the affected code is not contained in MTD CVS. Also note that this problem exists in linux-2.4 as well. Comments/feedback appreciated. TIA! -- Regards, George