2013/8/15, Florian Fainelli <flor...@openwrt.org>: > Le jeudi 15 août 2013 21:18:53 José Vázquez a écrit : >> 2013/8/11, José Vázquez <ppvazquez...@gmail.com>: >> > Still we have problems with rootfs_data partition if some openwrt and >> > kernel options are changed or added. >> > Because nobody answered i decided to blind shooting after reading all >> > the info i've could find. Also applied Florian's and Jonas' patches >> > published in linux-mips patchwork without success, so step back to >> > trunk r36529 and now can make all the changes without concern. >> > These are the revisions i've tested: >> > - https://github.com/Pteridium/openwrt/tree/bcm63xx-r37180: a lot of >> > problems with renamed files. >> > - trunk r37745: a bit more stable, but if i do modifications in the >> > kernel config the previous problems appear. >> > - https://github.com/Pteridium/openwrt/tree/bcm63xx-r36529: perfect. >> > Tested with and without procd as init process due to obvious reasons. >> > >> > José Vázquez >> >> Finally my colleague and me could establish that enabling block-mount >> in trunk r37180 and r37745 the file rename issue in rootfs_data >> partition happens always with BCM6368 (generic and SMP subtargets) no >> matter the kernel version (3.9.11 or 3.10.4). >> These are the tests made with r37745: >> - Repository used: https://github.com/Pteridium/OpenWRT-SMP-initial >> - Bootlog: http://pastebin.com/gsXqfutD >> - Openwrt config: http://pastebin.com/pMBx1BpE >> - Kernel config: http://pastebin.com/LtCcYAQW >> Also tested in a BCM6358 (Huawei HG556) and worked fine: >> - Bootlog: http://pastebin.com/0KAXi7rv >> >> Additionally we tested trunk r36529 with block-mount enabled and procd >> set as init process and nothing noticeable. >> Bootlog: http://pastebin.com/KHjq69ZA >> >> Due to my limited skills i can't tell you nothing more to help finding >> what it is happening with block-mount and BCM6368, but now we go to >> make more tests with the latest trunk because some modifications were >> made in the init packages. > > After discussing with Jonas on IRC about this issue, we believe that what > you > are seeing might just be some true concurrency issues, BCM6368 is one of the > > only SMP target supported in OpenWrt at the moment. Octeon is also one and > should show similar issues. > -- > Florian > Reviewing the Barrier Breaker targets, found that there are some ARM and PowerPC with SMP support and 2 or more cores enabled by default, but nobody opened a ticket describing an issue like the one we are experiencing. We will continue testing each new trunk revision and linux-mips patches that are being published.
José _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
