Re: [PATCH v8 17/21] clocksource / arch_timer: Parse GTDT to initialize arch timer

From: Hanjun Guo
Date: Thu Feb 05 2015 - 05:12:10 EST


Hi Lorenzo,

On 2015å02æ05æ 02:59, Lorenzo Pieralisi wrote:
On Mon, Feb 02, 2015 at 12:45:45PM +0000, Hanjun Guo wrote:
Using the information presented by GTDT (Generic Timer Description Table)
to initialize the arch timer (not memory-mapped).

Why are you not initializing the memory mapped timer ?

We left it for later work because no need for that to boot
available ARM64 hardware at now, and we have no hardware to
test unless I missed some of platforms.


CC: Daniel Lezcano <daniel.lezcano@xxxxxxxxxx>
Originally-by: Amit Daniel Kachhap <amit.daniel@xxxxxxxxxxx>
Tested-by: Suravee Suthikulpanit <Suravee.Suthikulpanit@xxxxxxx>
Tested-by: Yijing Wang <wangyijing@xxxxxxxxxx>
Tested-by: Mark Langsdorf <mlangsdo@xxxxxxxxxx>
Tested-by: Jon Masters <jcm@xxxxxxxxxx>
Tested-by: Timur Tabi <timur@xxxxxxxxxxxxxx>
Signed-off-by: Hanjun Guo <hanjun.guo@xxxxxxxxxx>
---
arch/arm64/kernel/time.c | 7 ++
drivers/clocksource/arm_arch_timer.c | 132 ++++++++++++++++++++++++++++-------
include/linux/clocksource.h | 6 ++
3 files changed, 118 insertions(+), 27 deletions(-)

diff --git a/arch/arm64/kernel/time.c b/arch/arm64/kernel/time.c
index 1a7125c..42f9195 100644
--- a/arch/arm64/kernel/time.c
+++ b/arch/arm64/kernel/time.c
@@ -35,6 +35,7 @@
#include <linux/delay.h>
#include <linux/clocksource.h>
#include <linux/clk-provider.h>
+#include <linux/acpi.h>

#include <clocksource/arm_arch_timer.h>

@@ -72,6 +73,12 @@ void __init time_init(void)

tick_setup_hrtimer_broadcast();

+ /*
+ * Since ACPI or FDT will only one be available in the system,
+ * we can use acpi_generic_timer_init() here safely
+ */
+ acpi_generic_timer_init();
+
arch_timer_rate = arch_timer_get_rate();
if (!arch_timer_rate)
panic("Unable to initialise architected timer.\n");
diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 095c177..407aa63 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -21,6 +21,7 @@
#include <linux/io.h>
#include <linux/slab.h>
#include <linux/sched_clock.h>
+#include <linux/acpi.h>

#include <asm/arch_timer.h>
#include <asm/virt.h>
@@ -370,8 +371,12 @@ arch_timer_detect_rate(void __iomem *cntbase, struct device_node *np)
if (arch_timer_rate)
return;

- /* Try to determine the frequency from the device tree or CNTFRQ */
- if (of_property_read_u32(np, "clock-frequency", &arch_timer_rate)) {
+ /*
+ * Try to determine the frequency from the device tree or CNTFRQ,
+ * if ACPI is enabled, get the frequency from CNTFRQ ONLY.
+ */
+ if (!acpi_disabled ||
+ of_property_read_u32(np, "clock-frequency", &arch_timer_rate)) {

This is getting a mess. cntbase tells you it is a memory mapped timer,
node pointer that you are probing through DT, and to top it all
acpi_disabled detects if you are probing in ACPI or DT mode.

I think this function should be simplified, this driver is also
pending a refactoring to split arch timer and the memory mapped one
so I think you'd better wait that work to make things simpler.

Does anyone working on this now?


[...]

+/* Initialize all the generic timers presented in GTDT */
+void __init acpi_generic_timer_init(void)
+{
+ if (acpi_disabled)
+ return;

acpi_disabled used again here, I repeat myself this is going to be
hard to track. You should try to organize the code something like:

if (acpi_disabled)
timer_dt_probe();
else
timer_acpi_probe();

mixing the code paths is getting unwieldy, see above to see my
reasoning.

Olof is unhappy with such approach, I think if (acpi_disabled) is
self-contained because we only get DT or ACPI in the system, we
can call this function time_init() without more if (acpi_disabled).


+
+ acpi_table_parse(ACPI_SIG_GTDT, arch_timer_acpi_init);
+}
+#endif
diff --git a/include/linux/clocksource.h b/include/linux/clocksource.h
index abcafaa..af6155a 100644
--- a/include/linux/clocksource.h
+++ b/include/linux/clocksource.h
@@ -346,4 +346,10 @@ extern void clocksource_of_init(void);
static inline void clocksource_of_init(void) {}
#endif

+#ifdef CONFIG_ACPI
+void acpi_generic_timer_init(void);
+#else
+static inline void acpi_generic_timer_init(void) { }
+#endif
+

That's not nice, it is a generic header, arch specific stuff should be
avoided. I think you should have an ACPI generic layer similar to
clocksource_of_init(), and probe from there when matching the respective
timers.

I think I'm OK with it, but do we really need to introduce a heavy
framework to init just arm arch timer (memory mapped or not) ?

Thanks
Hanjun
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/