On Tue, Jun 24, 2014 at 1:00 AM, Ming Lei <ming.lei@xxxxxxxxxxxxx> wrote: > Hi Jon and Vinayak, > > On Tue, Jun 24, 2014 at 11:42 AM, Jon Masters <jcm@xxxxxxxxxx> wrote: >> On 06/23/2014 10:47 PM, Ming Lei wrote: >> >>> On Thu, Mar 20, 2014 at 5:47 PM, Vinayak Kale <vkale@xxxxxxx> wrote: >>>> This patch adds the PMU device tree node for APM X-Gene Storm SOC. >>>> >>>> Please note that this patch has dependancy on a GIC driver patch [1] which is >>>> yet to be approved by maintainers. >>>> >>>> [1]- https://lkml.org/lkml/2014/2/27/605 >>>> (irqchip:gic: change access of gicc_ctrl register to read modify write) >> >>> No matter if Feng Kan's gic patch is applied or not, perf(PMU) >>> still can't work well with your patch, and the failure[1] will be >>> triggered when 'perf top' is started. >>> >>> The failure happened during the 1st pmu overflow interrupt. >>> >>> I test several mustang boards, and all has the problem with >>> upstream kernel. >> >> Note that the GIC bypass register config is set by the firmware and this >> only preserves the original setting, so you could be running an older >> firmware (such as U-Boot). On our (UEFI-enabled systems, complete with >> correct patch) perf top runs correctly without any such problems. > > I am wondering if it is related with boot or firmware. We have a board > which is running APM internal 3.12 based kernel, and perf(pmu) works > well. But if I just replace the 3.12 based kernel with upstream kernel > and keep others(boot, kernel parameter, fs, ....) not changed, then > the problem still can happen. fyi, newer firmware is confirmed to fix this issue on u-boot based platforms. Tested-by: dann frazier <dann.frazier@xxxxxxxxxxxxx> -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html