ARM: mmp: PXA168 timer fixes
Subsystems (4)
cover ARM: mmp: PXA168 timer fixes
This series contains a couple of bug fixes for the PXA168 timer. The clock wasn't assigned in pxa168.dtsi which eventually resulted in a hang at every boot after the clock was disabled. Also, the timer read function wasn't waiting long enough to capture the new timer value, which resulted in erroneous high CPU usage percent being reported with CONFIG_NO_HZ_IDLE=y.
I don't have any other MMP systems to test with, but I suspect the timer read delay problem also affected them. For example, the OLPC XO-4 kernel disabled CONFIG_NO_HZ due to incorrect high CPU usage reporting:
http://dev.laptop.org/git/olpc-kernel/commit?h=arm-3.5&id=5bd2520f8f51fc44911ec7a86b84f41a1f3e384c
CCing [email protected] because I didn't receive any responses when I submitted these last time and was hoping these fixes could get merged.
Doug Brown (2):
ARM: dts: pxa168: add timer reset and clock
ARM: mmp: fix timer_read delay
arch/arm/boot/dts/pxa168.dtsi | 2 ++ arch/arm/mach-mmp/time.c | 11 +++++++---- 2 files changed, 9 insertions(+), 4 deletions(-)
Signature · 1 line
2.34.1
PATCH 1/2 ARM: dts: pxa168: add timer reset and clock
The timer was missing the clock and reset like the other peripherals. Add them to allow the timer to continue working after boot completes.
--- arch/arm/boot/dts/pxa168.dtsi | 2 ++ 1 file changed, 2 insertions(+)
arch/arm/boot/dts/pxa168.dtsi +2 -0
@@ -53,6 +53,8 @@ timer0: timer@d4014000 { compatible = "mrvl,mmp-timer"; reg = <0xd4014000 0x100>; interrupts = <13>; + clocks = <&soc_clocks PXA168_CLK_TIMER>; + resets = <&soc_clocks PXA168_CLK_TIMER>; }; uart1: serial@d4017000 {
Signature · 1 line
2.34.1
PATCH 2/2 ARM: mmp: fix timer_read delay
In reply to Doug Browntimer_read() was using an empty 100-iteration loop to wait for the TMR_CVWR register to capture the latest timer counter value. The delay wasn't long enough. This resulted in CPU idle time being extremely underreported on PXA168 with CONFIG_NO_HZ_IDLE=y.
Switch to the approach used in the vendor kernel, which implements the capture delay by reading TMR_CVWR a few times instead.
--- arch/arm/mach-mmp/time.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-)
arch/arm/mach-mmp/time.c +7 -4
@@ -43,18 +43,21 @@static void __iomem *mmp_timer_base =TIMERS_VIRT_BASE ;/* -* FIXME: the timer needs some delay to stablize the counter capture +* Read the timer through the CVWR register. Delay is required after requesting +* a read. The CR register cannot be directly read due to metastability issues +* documented in the PXA168 software manual. */ static inline uint32_t timer_read (void ) { -int delay =100 ; +uint32_t val; +int delay =3 ;__raw_writel (1 , mmp_timer_base +TMR_CVWR (1 ));while (delay--) -cpu_relax (); + val =__raw_readl (mmp_timer_base +TMR_CVWR (1 )); -return __raw_readl (mmp_timer_base +TMR_CVWR (1 )); +return val; } static u64 notrace mmp_read_sched_clock(void)
Signature · 1 line
2.34.1
This series was applied to soc/soc.git (for-next) by Arnd Bergmann <[email protected]>:
Hello:
This series was applied to soc/soc.git (for-next)
by Arnd Bergmann <[email protected]>:
Quoted 8 lines · On Sat, 3 Dec 2022 16:51:15 -0800 you wrote:
This series contains a couple of bug fixes for the PXA168 timer. The clock wasn't assigned in
pxa168.dtsiwhich eventually resulted in a hang at every boot after the clock was disabled. Also, the timer read function wasn't waiting long enough to capture the new timer value, which resulted in erroneous high CPU usage percent being reported withCONFIG_NO_HZ_IDLE=y.[...]
Here is the summary with links:
- [1/2] ARM: dts: pxa168: add timer reset and clock
https://git.kernel.org/soc/soc/c/7b3e7df92a06 - [2/2] ARM: mmp: fix timer_read delay https://git.kernel.org/soc/soc/c/e348b4014c31
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html