Hello,
2013/9/1 José Vázquez <ppvazquez...@gmail.com>:
> Each time i select block-mount some config files in /etc appeared
> renamed after first reboot. The file content remains unchanged.
> - Fresh install:
> /etc/config:
> 6relayd dropbear fstab network ucitrack wireless
> dhcp firewall luci system uhttpd
> - After reboot:
> /etc/config:
> 6relayd cici dhcp firewall luci system twtwork uhttpd
> 6reld?d cpcp dropbear fstab rerellll ttttpd ucitrack wireless
>
> Also tested with ext4 and nothing strange happens.
>
> According to the jffs2 debug messages, that filesystem reads wrongly
> the name of the files, but the length of the file names remains the
> same as the original.
> The only way i have found to prevent this issue was replace line 77 of prom.c
> Original:
> } else if (BCMCPU_IS_3368() || BCMCPU_IS_6358()) {
> with:
> } else if (BCMCPU_IS_3368() || BCMCPU_IS_6358() ||
> BCMCPU_IS_6368()) {
Thus disabling the actual SMP code for 6368.
> because this issue happens with both smp and generic subtargets.
This looks strange to me, so even an uniprocessor kernel build fails?
Are you sure this is the correct image/kernel that you are flashing?
> The problem was noticed at the beginning of august with two different
> boards, but not made the right tests:
> https://lists.openwrt.org/pipermail/openwrt-devel/2013-August/020909.html
> One month later i can say that selecting bcm6368 + block-mount is the
> condition to reproduce the issue, and does not depend on the hos PC
> linux distro (Debian 6.0.7 and 7.1.0).
> Don't understand why this happens only with bcm6368, because bcm6358
> is free of this problem.
> Tested different kernel configurations, GCC versions, eglibc, ... and
> the result is always the same when block-mount is selected. Some
> people told that this happens few times without block-mount selected,
> but this point couldn't be verified.
There might be some data cache aliasing issue popping up for this
platform, although BCM6358 should very likely also be affected if that
was case. Can you make sure that the "generic" image you flash really
is the one you built? Then we can start investigating on the actual
BCM6368 SMP implementation issues.
--
Florian
_______________________________________________
openwrt-devel mailing list
openwrt-devel@lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
