| Message ID | 13624879.qYkvC9sWa4@diego |
|---|---|
| State | New |
| Headers | show |
On Saturday 22 November 2014, Heiko Stübner wrote: > it looks like not all bootloaders like the recently introduced > rk3288-smp-support. Uninitialized CNTVOFF registers of the > architected timer produce rcu stalls on some boards. > > Both Olof and Kevin were affected by this. On Kevin's board > we could simply upgrade u-boot to a version behaving correctly > while Olof's board uses coreboot where no fix for this exists > currently. Also as far as I know Olof's board is the only one > in circulation using coreboot, but it would still be nice to keep > the boot tests working for the time being. > > There exist an alternative approach for solving issues like this > consisting of > - clocksource: arch_timer: Fix code to use physical timers when requested [0] > - clocksource: arch_timer: Allow the device tree to specify uninitialized timer registers [1] > > > So to keep Olof's boottests working, disable the rk3288-smp until this > gets sorted. Interested parties with a working bootloader can of course > always simply re-insert the enable-method to have working smp. > Pulled into next/dt, thanks! Arnd
On Fri, Nov 28, 2014 at 12:37:01PM +0100, Arnd Bergmann wrote: > On Saturday 22 November 2014, Heiko Stübner wrote: > > it looks like not all bootloaders like the recently introduced > > rk3288-smp-support. Uninitialized CNTVOFF registers of the > > architected timer produce rcu stalls on some boards. > > > > Both Olof and Kevin were affected by this. On Kevin's board > > we could simply upgrade u-boot to a version behaving correctly > > while Olof's board uses coreboot where no fix for this exists > > currently. Also as far as I know Olof's board is the only one > > in circulation using coreboot, but it would still be nice to keep > > the boot tests working for the time being. > > > > There exist an alternative approach for solving issues like this > > consisting of > > - clocksource: arch_timer: Fix code to use physical timers when requested [0] > > - clocksource: arch_timer: Allow the device tree to specify uninitialized timer registers [1] > > > > > > So to keep Olof's boottests working, disable the rk3288-smp until this > > gets sorted. Interested parties with a working bootloader can of course > > always simply re-insert the enable-method to have working smp. > > > > Pulled into next/dt, thanks! And I've just now reverted the disablement patch since I picked up those two dependencies. Thanks, all. -Olof
Hi Arnd, Olof, it looks like not all bootloaders like the recently introduced rk3288-smp-support. Uninitialized CNTVOFF registers of the architected timer produce rcu stalls on some boards. Both Olof and Kevin were affected by this. On Kevin's board we could simply upgrade u-boot to a version behaving correctly while Olof's board uses coreboot where no fix for this exists currently. Also as far as I know Olof's board is the only one in circulation using coreboot, but it would still be nice to keep the boot tests working for the time being. There exist an alternative approach for solving issues like this consisting of - clocksource: arch_timer: Fix code to use physical timers when requested [0] - clocksource: arch_timer: Allow the device tree to specify uninitialized timer registers [1] So to keep Olof's boottests working, disable the rk3288-smp until this gets sorted. Interested parties with a working bootloader can of course always simply re-insert the enable-method to have working smp. [0] https://patchwork.kernel.org/patch/5051901/ [1] https://patchwork.kernel.org/patch/5051881/ The following changes since commit efb66e93d13da5a7b11b3a8dbb24a6fb29141752: ARM: dts: rockchip: enable PWM on Radxa Rock (2014-11-15 21:16:30 +0100) are available in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/mmind/linux-rockchip.git tags/v3.19-rockchip-dts3 for you to fetch changes up to b77d43943ea83997c6c37b8831d1561981d499c5: ARM: dts: rockchip: temporarily disable smp on rk3288 (2014-11-22 16:23:28 +0100) ---------------------------------------------------------------- Disable smp again on rk3288 temporarily to make Olof's boottest run again, until the patches for using the physical architected timers got accepted. ---------------------------------------------------------------- Heiko Stuebner (1): ARM: dts: rockchip: temporarily disable smp on rk3288 arch/arm/boot/dts/rk3288.dtsi | 1 - 1 file changed, 1 deletion(-)