Interesting MTD (jffs2) problem - magic bitmask not found
hello...
i've posted this on linux-arm mailing list, but i think its also interesting for this list...
i run 2.4.18-rmk7 with mtd from cvs (mid september). problem as follows:
# cat /proc/mtd dev: size erasesize name mtd0: 00040000 00040000 "blob" mtd1: 00040000 00040000 "parm" mtd2: 00100000 00040000 "krnl" mtd3: 00400000 00040000 "rd" mtd4: 00100000 00040000 "krnl2" mtd5: 00400000 00040000 "rd2" mtd6: 01580000 00040000 "jffs2" # mtd_debug info /dev/mtd6 mtd.type = MTD_NORFLASH mtd.flags = MTD_CAP_NORFLASH mtd.size = 22544384 (21M) mtd.erasesize = 262144 (256K) mtd.oobblock = 0 mtd.oobsize = 0 mtd.ecctype = MTD_ECC_NONE regions = 0 # cd /root # ls temp temp1 temp2 # mkfs.jffs2 -r temp -e 262144 -o jffs2img # eraseall /dev/mtd6 Erasing 256 Kibyte @ ...[snip]... 100% complete. # cp jffs2img /dev/mtd6 # mount /dev/mtdblock6 -t jffs2 /mnt/jffs2/ # cd /mnt/jffs2 # ls temp1 temp2 # rm -r temp* ... [this tooks very long time] ... # ls # cd / # umount /mnt/jffs2/ # mount /dev/mtdblock6 -t jffs2 /mnt/jffs2/ ????????????????????????????????????????????????????????????????????? jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00040000: 0x2003 instead jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00040004: 0x000c instead jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00040008: 0xdc6d instead jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080000: 0x2003 instead jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080004: 0x000c instead ...[snip]... ????????????????????????????????????????????????????????????????????? # mount /dev/ramdisk on / type ext2 (rw) none on /proc type proc (rw) /dev/mtdblock6 on /mnt/jffs2 type jffs2 (rw) # cd /mnt/jffs2/ # ls # ...
best regards... Michael Palme
Re: jffs2 problem - magic bitmask not found
hello again :-)
after lots of time playing around that thing put me really into madness...
but could be i am doing something wrong!?
please verify:
i want to use jffs2 on my strongarmboard with 32Mb Intel strata flash. so i think things to do to get a working jffs2 partition are as follows:
* built kernel with mtd and jffs2 support (i have done this ... kernel 2.4.18-rmk7 with patched in mtd-cvs)
* built mtd-utils (successfully done -- crossbuilt for running on arm native)
* create partitions on flash (i have done this via commmand line partition table parsing -- CONFIG_CMDLINE="root=/dev/ram0 mtdparts=sa1100:256k(blob),256k(parm),1M(krnl),4M(rd),1M(krnl2),4M(rd2),-(jffs2)")
* erase the jffs2 partition (done via "eraseall /dev/mtd6" -- also tried to delete it via jtag, so im sure theres no crap in after erasing...)
* verify things ("cat /proc/mtd" and "mtd_debug info /dev/mtd6" -- okay, everything as expected)
* create a jffs2 image (done via "mkfs.jffs2 -r foo -o jffs2.img -p -e 262144" -- image contains only a dummy file and is very small)
* copy the image to the mtd partition ("cp jffs2.img /dev/mtd6")
* mounting the jffs2 partition 1st time (works okay via "mount -t jffs2 /dev/mtdblock6 /mnt/jffs2")
* verify contents (okay "ls /mnt/jffs2" shows the dummy file)
BUT THEN:
* umount jffs2 part ("umount /mnt/jffs2") takes *very* long
* immediately remount of the jffs2 part brings lots of error messages:
jffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00040000:0x2003 insteadjffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00040004:0x000c insteadjffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00040008:0xdc6d insteadjffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080000:0x2003 insteadjffs2_scan_eraseblock(): Magic bitmask 0x1985 not found at 0x00080004:0x000c instead
* but after that, "mount" shows the jffs2 part as mounted and i can see the dummy file again
* now unmounting again -- now very fast!?
i've tried again and again... no success.
please who can help
i need the fs for production use...
thanks in advance ... Michael Palme
RE: jffs2 problem - magic bitmask not found
Quoted 32 lines
hello again :-)
after lots of time playing around that thing put me really into madness...
but could be i am doing something wrong!?
please verify:i want to use jffs2 on my strongarmboard with 32Mb Intel strata flash. so i think things to do to get a working jffs2 partition are as follows:
* built kernel with mtd and jffs2 support (i have done this ... kernel 2.4.18-rmk7 with patched in mtd-cvs)
* built mtd-utils (successfully done -- crossbuilt for running on arm native)
* create partitions on flash (i have done this via commmand line partition table parsing --
CONFIG_CMDLINE="root=/dev/ram0mtdparts=sa1100:256k(blob),256k(parm),1M(krnl),4M(rd),1M(krnl2),4M(rd2),-(jffs2)")* erase the jffs2 partition (done via "eraseall
/dev/mtd6" -- also tried to delete it via jtag, so im sure theres no crap in after erasing...)* verify things ("cat
/proc/mtd" and "mtd_debuginfo/dev/mtd6" -- okay, everything as expected)* create a jffs2 image (done via "mkfs.jffs2 -r foo -o jffs2.img -p -e 262144" -- image contains only a dummy file and is very small)
* copy the image to the mtd partition ("cp jffs2.img
/dev/mtd6")
Quoted 8 lines
* mounting the jffs2 partition 1st time (works okay via "mount -t jffs2
/dev/mtdblock6/mnt/jffs2")* verify contents (okay "ls
/mnt/jffs2" shows the dummy file)BUT THEN:
* umount jffs2 part ("umount
/mnt/jffs2") takes *very* long
JFFS2 is probaly erasing and putting a CLEANMARKER on all unused erase blocks and will block until it's finished. I think this can be fixed.
What happens if you wait 5-10 minutes before you unmount the FS for the time? While waiting, do a couple of 'df' commands and see if the Use field slowly increases to 100% and then drops back your expected usage.
Quoted 13 lines
* immediately remount of the jffs2 part brings lots of error messages:
jffs2_scan_eraseblock(): Magic bitmask0x1985not found at0x00040000:0x2003insteadjffs2_scan_eraseblock(): Magic bitmask0x1985not found at0x00040004:0x000cinsteadjffs2_scan_eraseblock(): Magic bitmask0x1985not found at0x00040008:0xdc6dinsteadjffs2_scan_eraseblock(): Magic bitmask0x1985not found at0x00080000:0x2003insteadjffs2_scan_eraseblock(): Magic bitmask0x1985not found at0x00080004:0x000cinstead
No idea why this happens.
Quoted 5 lines
* but after that, "mount" shows the jffs2 part as mounted and i can see the dummy file again
* now unmounting again -- now very fast!?
That's because JFFS2 already has marked all unused erase blocks with a CLEANMARKER.
Jocke
Quoted 13 lines
i've tried again and again... no success.
please who can help
i need the fs for production use...thanks in advance ... Michael Palme
______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/