> Weird. For me it was working only at the first boot after flashing the image,
> but any modifications in the fs has been lost after subsequent reboots.
Hm, might be I did not looked at that correctly. Altough, I installed
some packages and I think they were still there after reboot...

Just updated to latest master and flashed the image using Factory Reset
(U-Boot TFTP-Server). The flashing worked fine, but I now get an error
on bootup:
[    2.710000] UBI: attaching mtd8 to ubi0
[    2.790000] UBI error: scan_peb: bad image sequence number 248591051
in PEB 63, expected 172450884
[    2.800000] Erase counter header dump:
[    2.810000]  magic          0x55424923
[    2.810000]  version        1
[    2.810000]  ec             1
[    2.820000]  vid_hdr_offset 2048
[    2.820000]  data_offset    4096
[    2.820000]  image_seq      248591051
[    2.830000]  hdr_crc        0x98ac164
[    2.830000] erase counter header hexdump:
[    2.830000] UBI error: ubi_attach_mtd_dev: failed to attach mtd8,
error -22
[    2.840000] UBI error: ubi_init: cannot attach mtd8
[    2.850000] VFS: Cannot open root device "(null)" or
unknown-block(0,0): error -6

It looks like there's a bug in the new EOF marker support.

I looked into the code, I found this interesting comment just before the
"bad image sequence" error message:
/*                                                               
 * Make sure that all PEBs have the same image sequence number.  
 * This allows us to detect situations when users flash UBI      
 * images incorrectly, so that the flash has the new UBI image   
 * and leftovers from the old one. This feature was added        
 * relatively recently, and the sequence number was always       
 * zero, because old UBI implementations always set it to zero.  
 * For this reasons, we do not panic if some PEBs have zero      
 * sequence number, while other PEBs have non-zero sequence      
 * number.                                                       
 */


--
Stefan
_______________________________________________
openwrt-devel mailing list
[email protected]
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel

Reply via email to