2013/11/27, José Vázquez <ppvazquez...@gmail.com>: > 2013/11/27, John Crispin <j...@phrozen.org>: >> On 27/11/13 08:48, Florian Fainelli wrote: >>> >>> On Nov 26, 2013 11:42 PM, "John Crispin" <j...@phrozen.org >>> <mailto:j...@phrozen.org>> wrote: >>> > >>> > On 26/11/13 23:29, José Vázquez wrote: >>> >> >>> >> Maybe the problem is that part of the new init process does not work >>> >> well with two cpus:http://pastebin.com/LTUmbxZ6 >>> > >>> > >>> > >>> > who says the new init does not work on smp ? i would claim that bmips >>> is just not supported properly and the maintainers of the code base have >>> let the code rot for too long ... >>> > Nobody said that the problem is due to smp or bmips. Could be a > silicon errata in some BCM6368 SoCs, something related with jffs2 and > smp > (http://lists.infradead.org/pipermail/linux-mtd-cvs/2012-November/008218.html), > flash chips not completely supported or something else. > >>> That is kind of gratuitous, I see no evidence of problems coming from >>> the BMIPS SMP support code. The other BMIPS based platforms I have seen >>> do not have any jffs2 problems but they do not use overlayfs. I still >>> think that there might be some missing cache flushing in the overlay fs >>> code which only pops up with BMIPS CPUs due to how the I/D caches are >>> designed between the two threads. Time to reproduce that on my end too. >>> > What can i do to help? > >> >> >> according to you its a caching problem. as other mips boards have >> working SMP i would expect this to be bmips related. > > A couple of months ago compiled an ext4 image to test but found that, > once the kernel calls preinit, the firm creates a partition with jffs2 > fs and switch to overlayfs; the only way to avoid this behavior was > unsetting jffs2 from the kernel. I go to make a jffs2 image and will > post the logs. > If somebody can tell what i must do to check deeper this issue i'll > try to make my best effort. > > José > I've enabled CPU_HOTPLUG in the kernel and added maxcpus=1 to the CMDLINE. Once openwrt boot is finished i enable the second core. https://www.kernel.org/doc/Documentation/cpu-hotplug.txt The initial tests showed that the jffs2 problem doesn't happen, but still more tests are needed to confirm it.
Now i'm working in a backport of AA to enable smp in bcm63xx (done and working fine) and b53, so i will drop trunk for my bcm6368 based router. https://github.com/Pteridium/Attitude-Adjustment-ARV4518PWR01/commits/devel Best regards: José _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
