JFFS2 issues on OSK?

This is strange. When I try to mount /dev/mtdblock3 as the root FS I get a variety of errors (minimally "point returned len too short: 0") ... which do *not* appear later when I just issue a "mount -t jffs2 /dev/mtdblock3 /mnt"!!

System is an OMAP 5912 OSK board: arm926ej-s, with a pair of 16 MB Micron NOR flash chips concatenated to act like one bigger chip. The relevant parts of the boot log follow.

This particular kernel didn't include JFFS2 summary support, but similar problems happen with summary enabled. It was the same kernel which _wrote_ the filesystem, starting with an "mtd_debug erase /dev/mtd3 ..." and then mounting it before writing the files (only about 8 MBytes).

Does anyone know what may be going on here? This is with u-Boot 1.3.2 from GIT, and a 2.6.25-rc6 kernel. (This one happens to be from the OMAP tree, so cpufreq works, but the same thing happens with kernel.org code.)

Details below ...

- Dave

omapflash.0: Found 1 x16 devices at 0x0 in 16-bit bank
omapflash.0: Found 1 x16 devices at 0x1000000 in 16-bit bank
Intel/Sharp Extended Query Table at 0x0031
Using buffer write method
cfi_cmdset_0001: Erase suspend on write enabled
erase region 0: offset=0x0,size=0x20000,blocks=128
erase region 1: offset=0x1000000,size=0x20000,blocks=128
Creating 4 MTD partitions on "omapflash.0":
0x00000000-0x00020000 : "bootloader"
0x00020000-0x00040000 : "params"
0x00040000-0x00240000 : "kernel"
0x00240000-0x02000000 : "filesystem"
MUX: initialized W11_1610_CF_CD1
MUX: initialized P11_1610_CF_CD2
MUX: initialized R11_1610_CF_IOIS16
MUX: initialized V10_1610_CF_IREQ
MUX: initialized W10_1610_CF_RESET
omap_cf: cs2 on irq 222
omap_rtc omap_rtc: rtc core: registered omap_rtc as rtc0
omap_rtc: already running
OMAP Watchdog Timer Rev 0x10: initial timeout 60 sec
Registered led device: d9
Registered led device: d2
Registered led device: d3
TCP cubic registered
NET: Registered protocol family 1
NET: Registered protocol family 17
RPC: Registered udp transport module.
RPC: Registered tcp transport module.
omap_rtc omap_rtc: setting system clock to 2008-03-17 22:41:27 UTC (1205793687)
VFS: Mounted root (jffs2 filesystem).
Freeing init memory: 96K
JFFS2 warning: (180) check_node_data: MTD point returned len too short: 0 instead of 976. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 627. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 18. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 343. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 905. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2902. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2557. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 14. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 398. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 975. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2275. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2661. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2525. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2625. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2833. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2875. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2653. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2282. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2542.
pccard: PCMCIA card inserted into slot 0
pcmcia: registering new device pcmcia0.0
JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 1740. scsi0 : pata_pcmcia
ata1: PATA max PIO0 cmd 0xc2852000 ctl 0xc285200e irq 222
JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 19. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 1083. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 1604. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 1767. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2257. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 1836. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2184. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 1787. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2531. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2630. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2655. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2538. JFFS2 warning: (1) check_node_data: MTD point returned len too short: 0 instead of 2659. ata1.00: CFA: SAMSUNG CF/ATA, Rev 1.15, max PIO2
ata1.00: 251904 sectors, multi 0: LBA
ata1.00: configured for PIO0
ata1.00: configured for PIO0
ata1: EH complete
scsi 0:0:0:0: Direct-Access ATA SAMSUNG CF/ATA Rev PQ: 0 ANSI: 5
JFFS2 warning: (180) check_node_data: MTD point returned len too short: 0 instead of 2.

OABI Buildroot Linux 2.6.25-rc6-omap1, cpu armv5tejl

osk login: root
Password:

Mar 17 15:41:35 login[194]: root login on 'ttyS0'

When I did this without the CF card inserted, there were not as many "len too short" errors ... but they were stil present.

The (brand new) buildroot image doesn't work too well either ... not clear yet just why. Notice that none of the /etc/init.d/S* scripts ran, for example; no shell seems to work.

Update: the latest kernel.org GIT (nyet OMAP git) has a patch which seems to resolve this issue. It's a NOR flash corruption issue specific to CFI command set 1 (Intel), dating since early
October: c4a9f88daf6c382fedde4cdddef0b30f1d0a20db caused this.

This afflicted OSK, since its Micron flash is compatible with Intel's Strataflash. The failure mode variability was caused by different read and write patterns.

If you have NOR flash issues, you may *NEED* this fix.

- Dave

commit fb6d080c6f75dfd7e23d5a3575334785aa8738eb

Author: Alexey Korolev <[email protected]>
Date:   Fri Apr 4 14:30:01 2008 -0700

mtd: fix broken state in CFI driver caused by FL_SHUTDOWN

THe CFI driver in 2.6.24 kernel is broken. Not so intensive read/write operations cause incomplete writes which lead to kernel panics in JFFS2.

We investigated the issue - it is caused by bug in FL_SHUTDOWN parsing code. Sometimes chip returns -EIO as if it is in FL_SHUTDOWN state when it should wait in FL_PONT (error in order of conditions).

The following patch fixes the bug in state parsing code of CFI. Also I've added comments to notify developers if they want to add new case in future.

index 47794d2..0080452 100644
@@ -718,7 +718,7 @@ static int chip_ready (struct map_info *map, struct flchip *chip, unsigned long
 			/* Someone else might have been playing with it. */
 			return -EAGAIN;
 		}
-
+		/* Fall through */
 	case FL_READY:
 	case FL_CFI_QUERY:
 	case FL_JEDEC_QUERY:

@@ -778,14 +778,14 @@ static int chip_ready (struct map_info *map, struct flchip *chip, unsigned long
 		chip->state = FL_READY;
 		return 0;
 
+	case FL_SHUTDOWN:
+		/* The machine is rebooting now,so no one can get chip anymore */
+		return -EIO;
 	case FL_POINT:
 		/* Only if there's no operation suspended... */
 		if (mode == FL_READY && chip->oldstate == FL_READY)
 			return 0;
-
-	case FL_SHUTDOWN:
-		/* The machine is rebooting now,so no one can get chip anymore */
-		return -EIO;
+		/* Fall through */
 	default:
 	sleep:
 		set_current_state(TASK_UNINTERRUPTIBLE);