jffs2_flash_writev(): Non-contiguous write to

Hi 4 All!

I have read in the archives some letter with the same problem :

- JFFS2 Non-Contiguous Write Oops on 2.6.18 - Scott Shumate
- JFFS2 dataflash problem - Peer Georgi
- Jffs2 Oops - z l

... but I haven't found any answer or solution for it. I'm using at91rm9200 based -own designed- board with 2.6.18 linux kernel. I have patched it from mtd csv (http://sourceware.org/jffs2/), and I'm using mtd-utils 1.0.0 (http://www.linux-mtd.infradead.org/) and compiled with jffs2 fs support on debug level 1!

I have partitioned my dataflash as follows
at91_dataflash: AT45DB642 detected [spi0] (8650752 bytes) Creating 4 MTD partitions on "AT45DB642.spi0":
0x00000000-0x00021000 : "boot_area" 0x00021000-0x00129000 : "kernel" 0x00129000-0x002b5000 : "service" 0x002b5000-0x00840000 : "filesystem"

I'm using at45db642 like Scott Shumate!

//I'm doing the following steps :

#mkfs.jffs2 -e0x2100KiB -n -l -o service.img -r ~/service/ #flash_eraseall /dev/mtd2
Erasing 1 Kibyte @ 18bbe0 -- 99 % complete

#flashcp -v service.img /dev/mtd2
Erasing blocks: 6/6 (100%)
Writing data: 6k/6k (100%)
Verifying data: 6k/6k (100%)

#mount -t jffs2 /dev/mtdblock2 /mnt
JFFS2 write-buffering enabled buffer (1056) erasesize (8448)

#ls /mnt
... and I see the files, and can read too - everything seems to okay!

but any write action causes the following :

jffs2_flash_writev(): Non-contiguous write to 00000030 wbuf was previously 000014a0-000018b8
kernel BUG at fs/jffs2/wbuf.c:786!

Unable to handle kernel NULL pointer dereference at virtual address 00000000
pgd = c1030000
[00000000] *pgd=2102f031, *pte=00000000, *ppte=00000000
Internal error: Oops: 817 [#1]
Modules linked in:
CPU: 0
PC is at __bug+0x40/0x54
LR is at 0x1
pc : [<c0024ba4>]    lr : [<00000001>]    Not tainted

sp : c033bc90 ip : 60000093 fp : c033bc9c
r10: 000014a0 r9 : 00000000 r8 : 00000000
r7 : 00000030 r6 : c038ba00 r5 : 00000418 r4 : 00000000 r3 : 00000000 r2 : 00000000 r1 : 0000908f r0 : 00000001
Flags: nZCv IRQs on FIQs on Mode SVC_32 Segment user
Control: C000317F Table: 21030000 DAC: 00000015 Process vi (pid: 699, stack limit = 0xc033a250)

Stack: (0xc033bc90 to 0xc033c000)
...
...
Backtrace:
[<c0024b64>] (__bug+0x0/0x54) from [<c00d63d4>] (jffs2_flash_writev+0x19c/0x470)
[<c00d623c>] (jffs2_flash_writev+0x4/0x470) from [<c00d6700>] (jffs2_flash_write
+0x58/0x60)
[<c00d66a8>] (jffs2_flash_write+0x0/0x60) from [<c00c907c>] (jffs2_mark_node_obs
olete+0x600/0x768)
r6 = C033BD34  r5 = C038BA00  r4 = C033BD2C
[<c00c8a7c>] (jffs2_mark_node_obsolete+0x0/0x768) from [<c00c661c>] (jffs2_obsol
ete_node_frag+0x50/0x64)
[<c00c65cc>] (jffs2_obsolete_node_frag+0x0/0x64) from [<c00c66e4>] (jffs2_trunca
te_fragtree+0xb4/0x11c)

r4 = 00000000
[<c00c6630>] (jffs2_truncate_fragtree+0x0/0x11c) from [<c00d207c>] (jffs2_do_set attr+0x5d8/0x6f0)
r8 = C13C11D4 r7 = C10244F8 r6 = 00000000 r5 = 00000000
r4 = 00000000
[<c00d1aa4>] (jffs2_do_setattr+0x0/0x6f0) from [<c00d21a8>] (jffs2_setattr+0x14/ 0x18)
[<c00d2194>] (jffs2_setattr+0x0/0x18) from [<c008ace4>] (notify_change+0x134/0x2
44)
[<c008abb0>] (notify_change+0x0/0x244) from [<c006e07c>] (do_truncate+0x5c/0x78)

[<c006e020>] (do_truncate+0x0/0x78) from [<c007fd40>] (may_open+0x1a8/0x1e4) r5 = C13C11D4 r4 = 00000242
[<c007fb98>] (may_open+0x0/0x1e4) from [<c007fff8>] (open_namei+0x27c/0x6bc) r7 = C0017E28 r6 = C033BF04 r5 = 00000000 r4 = FFFFFFEB [<c007fd7c>] (open_namei+0x0/0x6bc) from [<c006f05c>] (do_filp_open+0x2c/0x48) [<c006f030>] (do_filp_open+0x0/0x48) from [<c006f3ac>] (do_sys_open+0x54/0xe4) r5 = 000001B4 r4 = 00000241
[<c006f358>] (do_sys_open+0x0/0xe4) from [<c006f460>] (sys_open+0x24/0x28) [<c006f43c>] (sys_open+0x0/0x28) from [<c0020d20>] (ret_fast_syscall+0x0/0x2c)

Code: 1b00444d e59f0014 eb00444b e3a03000 (e5833000)
Segmentation fault
                  # <7>jffs2_write_super()

jffs2_thread_should_wake(): nr_free_blocks 191, nr_erasing_blocks 0, dirty_size 0x0: no

Any help would be great! Thanks in advance!

Best Regards
Zsolt Demi

While I don't think this is the solution to your problem, note that there there now is an mtd-utils 1.0.1 from the same place. I know at least some of the applications in 1.0.0 don't work properly with 2.6.18 due to mtd changes.

/Ricard
--
Ricard Wolf Wanderlöf ricardw(at)axis.com
Axis Communications AB, Lund, Sweden www.axis.com
Phone +46 46 272 2016 Fax +46 46 13 61 30

Well I didn't really think it was a solution, just a longshot since there was an updated version available.

Sounds ok so far.

Yes, in the sense that it is developed by the group of people, but it lives in fs/jffs2 in the kernel tree rather than in drivers/mtd .

/Ricard
--
Ricard Wolf Wanderlöf ricardw(at)axis.com
Axis Communications AB, Lund, Sweden www.axis.com
Phone +46 46 272 2016 Fax +46 46 13 61 30