> > I'm pretty sure that the jffs2 file name corruption problem doesn't > happen BCM63xx SoCs with spi flash and smp enabled, so, if this is > correct, seems to be a race condition between the flash type, jffs2 > and some Broadcom SoCs, as Jonas pointed some time ago. Can somebody confirm is the problem disappears running openwrt in a smp capable BCM63xx with SPI flash? > > These warnings are easily reproducible enabling the mentioned options > in the kernel. > > Initially my intention was to send this information to linux-mips but, > because these warnings can be repeated easily i think that this should > be discussed by more skilled and experienced people. > > Some other options inside "Kernel hacking" couldn't be tested because > with them enabled the CFE of the VR-3025un was unable to uncompress > the firmware. > > Regards: > > José >
Forgot to mention that, even though BCM6358 and BCM6368 send the mentioned warnings, file name corruption only happens in the BCM6368 when enabling SMP. Besides, danitool, adding isolcpus=0 and activating bit 5 in register $22, 2 made the problem much less serious, but it still remains, and in addition, he achieved a significant increase in network throughput with a Netgear DGND3700 v1. With that configuration almost all the hardware irqs are driven by cpu0. _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel