[PATCH 46/47] Update and improve CRISv32 kernel/time.c

From: Jesper Nilsson
Date: Thu Dec 06 2007 - 07:57:29 EST


- Shorten include paths to machine dependent header files.
- Register name for first timer is now regi_timer0.
- Remove raw_printk hack, use oops_in_progress instead.
- Add handling of CPU frequency scaling for CRIS.
- Remove regs parameter to timer_interrupt, get them from get_irq_regs instead.
- Whitespace and formatting changes.

Signed-off-by: Jesper Nilsson <jesper.nilsson@xxxxxxxx>
---
arch/cris/arch-v32/kernel/time.c | 237 ++++++++++++++++++++++----------------
1 files changed, 139 insertions(+), 98 deletions(-)

diff --git a/arch/cris/arch-v32/kernel/time.c b/arch/cris/arch-v32/kernel/time.c
index 2f7e8e2..3a13dd6 100644
--- a/arch/cris/arch-v32/kernel/time.c
+++ b/arch/cris/arch-v32/kernel/time.c
@@ -1,8 +1,7 @@
-/* $Id: time.c,v 1.19 2005/04/29 05:40:09 starvik Exp $
- *
+/*
* linux/arch/cris/arch-v32/kernel/time.c
*
- * Copyright (C) 2003 Axis Communications AB
+ * Copyright (C) 2003-2007 Axis Communications AB
*
*/

@@ -14,28 +13,34 @@
#include <linux/sched.h>
#include <linux/init.h>
#include <linux/threads.h>
+#include <linux/cpufreq.h>
#include <asm/types.h>
#include <asm/signal.h>
#include <asm/io.h>
#include <asm/delay.h>
#include <asm/rtc.h>
#include <asm/irq.h>
-
-#include <asm/arch/hwregs/reg_map.h>
-#include <asm/arch/hwregs/reg_rdwr.h>
-#include <asm/arch/hwregs/timer_defs.h>
-#include <asm/arch/hwregs/intr_vect_defs.h>
+#include <asm/irq_regs.h>
+
+#include <hwregs/reg_map.h>
+#include <hwregs/reg_rdwr.h>
+#include <hwregs/timer_defs.h>
+#include <hwregs/intr_vect_defs.h>
+#ifdef CONFIG_CRIS_MACH_ARTPEC3
+#include <hwregs/clkgen_defs.h>
+#endif

/* Watchdog defines */
-#define ETRAX_WD_KEY_MASK 0x7F /* key is 7 bit */
-#define ETRAX_WD_HZ 763 /* watchdog counts at 763 Hz */
-#define ETRAX_WD_CNT ((2*ETRAX_WD_HZ)/HZ + 1) /* Number of 763 counts before watchdog bites */
+#define ETRAX_WD_KEY_MASK 0x7F /* key is 7 bit */
+#define ETRAX_WD_HZ 763 /* watchdog counts at 763 Hz */
+/* Number of 763 counts before watchdog bites */
+#define ETRAX_WD_CNT ((2*ETRAX_WD_HZ)/HZ + 1)

unsigned long timer_regs[NR_CPUS] =
{
- regi_timer,
+ regi_timer0,
#ifdef CONFIG_SMP
- regi_timer2
+ regi_timer2
#endif
};

@@ -44,12 +49,22 @@ extern int set_rtc_mmss(unsigned long nowtime);
extern int setup_irq(int, struct irqaction *);
extern int have_rtc;

+#ifdef CONFIG_CPU_FREQ
+static int
+cris_time_freq_notifier(struct notifier_block *nb, unsigned long val,
+ void *data);
+
+static struct notifier_block cris_time_freq_notifier_block = {
+ .notifier_call = cris_time_freq_notifier,
+};
+#endif
+
unsigned long get_ns_in_jiffie(void)
{
reg_timer_r_tmr0_data data;
unsigned long ns;

- data = REG_RD(timer, regi_timer, r_tmr0_data);
+ data = REG_RD(timer, regi_timer0, r_tmr0_data);
ns = (TIMER0_DIV - data) * 10;
return ns;
}
@@ -59,31 +74,27 @@ unsigned long do_slow_gettimeoffset(void)
unsigned long count;
unsigned long usec_count = 0;

- static unsigned long count_p = TIMER0_DIV;/* for the first call after boot */
+ /* For the first call after boot */
+ static unsigned long count_p = TIMER0_DIV;
static unsigned long jiffies_p = 0;

- /*
- * cache volatile jiffies temporarily; we have IRQs turned off.
- */
+ /* Cache volatile jiffies temporarily; we have IRQs turned off. */
unsigned long jiffies_t;

/* The timer interrupt comes from Etrax timer 0. In order to get
* better precision, we check the current value. It might have
- * underflowed already though.
- */
+ * underflowed already though. */
+ count = REG_RD(timer, regi_timer0, r_tmr0_data);
+ jiffies_t = jiffies;

- count = REG_RD(timer, regi_timer, r_tmr0_data);
- jiffies_t = jiffies;
-
- /*
- * avoiding timer inconsistencies (they are rare, but they happen)...
- * there are one problem that must be avoided here:
- * 1. the timer counter underflows
+ /* Avoiding timer inconsistencies (they are rare, but they happen)
+ * There is one problem that must be avoided here:
+ * 1. the timer counter underflows
*/
if( jiffies_t == jiffies_p ) {
if( count > count_p ) {
- /* Timer wrapped, use new count and prescale
- * increase the time corresponding to one jiffie
+ /* Timer wrapped, use new count and prescale.
+ * Increase the time corresponding to one jiffy.
*/
usec_count = 1000000/HZ;
}
@@ -106,17 +117,15 @@ unsigned long do_slow_gettimeoffset(void)
*/
/* This gives us 1.3 ms to do something useful when the NMI comes */

-/* right now, starting the watchdog is the same as resetting it */
+/* Right now, starting the watchdog is the same as resetting it */
#define start_watchdog reset_watchdog

#if defined(CONFIG_ETRAX_WATCHDOG)
static short int watchdog_key = 42; /* arbitrary 7 bit number */
#endif

-/* number of pages to consider "out of memory". it is normal that the memory
- * is used though, so put this really low.
- */
-
+/* Number of pages to consider "out of memory". It is normal that the memory
+ * is used though, so set this really low. */
#define WATCHDOG_MIN_FREE_PAGES 8

void
@@ -125,14 +134,15 @@ reset_watchdog(void)
#if defined(CONFIG_ETRAX_WATCHDOG)
reg_timer_rw_wd_ctrl wd_ctrl = { 0 };

- /* only keep watchdog happy as long as we have memory left! */
+ /* Only keep watchdog happy as long as we have memory left! */
if(nr_free_pages() > WATCHDOG_MIN_FREE_PAGES) {
- /* reset the watchdog with the inverse of the old key */
- watchdog_key ^= ETRAX_WD_KEY_MASK; /* invert key, which is 7 bits */
+ /* Reset the watchdog with the inverse of the old key */
+ /* Invert key, which is 7 bits */
+ watchdog_key ^= ETRAX_WD_KEY_MASK;
wd_ctrl.cnt = ETRAX_WD_CNT;
wd_ctrl.cmd = regk_timer_start;
wd_ctrl.key = watchdog_key;
- REG_WR(timer, regi_timer, rw_wd_ctrl, wd_ctrl);
+ REG_WR(timer, regi_timer0, rw_wd_ctrl, wd_ctrl);
}
#endif
}
@@ -148,7 +158,7 @@ stop_watchdog(void)
wd_ctrl.cnt = ETRAX_WD_CNT;
wd_ctrl.cmd = regk_timer_stop;
wd_ctrl.key = watchdog_key;
- REG_WR(timer, regi_timer, rw_wd_ctrl, wd_ctrl);
+ REG_WR(timer, regi_timer0, rw_wd_ctrl, wd_ctrl);
#endif
}

@@ -160,17 +170,28 @@ handle_watchdog_bite(struct pt_regs* regs)
#if defined(CONFIG_ETRAX_WATCHDOG)
extern int cause_of_death;

- raw_printk("Watchdog bite\n");
+ oops_in_progress = 1;
+ printk(KERN_WARNING "Watchdog bite\n");

/* Check if forced restart or unexpected watchdog */
if (cause_of_death == 0xbedead) {
+#ifdef CONFIG_CRIS_MACH_ARTPEC3
+ /* There is a bug in Artpec-3 (voodoo TR 78) that requires
+ * us to go to lower frequency for the reset to be reliable
+ */
+ reg_clkgen_rw_clk_ctrl ctrl =
+ REG_RD(clkgen, regi_clkgen, rw_clk_ctrl);
+ ctrl.pll = 0;
+ REG_WR(clkgen, regi_clkgen, rw_clk_ctrl, ctrl);
+#endif
while(1);
}

- /* Unexpected watchdog, stop the watchdog and dump registers*/
+ /* Unexpected watchdog, stop the watchdog and dump registers. */
stop_watchdog();
- raw_printk("Oops: bitten by watchdog\n");
- show_registers(regs);
+ printk(KERN_WARNING "Oops: bitten by watchdog\n");
+ show_registers(regs);
+ oops_in_progress = 0;
#ifndef CONFIG_ETRAX_WATCHDOG_NICE_DOGGY
reset_watchdog();
#endif
@@ -178,21 +199,19 @@ handle_watchdog_bite(struct pt_regs* regs)
#endif
}

-/* last time the cmos clock got updated */
+/* Last time the cmos clock got updated. */
static long last_rtc_update = 0;

/*
* timer_interrupt() needs to keep up the real-time clock,
- * as well as call the "do_timer()" routine every clocktick
+ * as well as call the "do_timer()" routine every clocktick.
*/
-
-//static unsigned short myjiff; /* used by our debug routine print_timestamp */
-
extern void cris_do_profile(struct pt_regs *regs);

static inline irqreturn_t
-timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
+timer_interrupt(int irq, void *dev_id)
{
+ struct pt_regs *regs = get_irq_regs();
int cpu = smp_processor_id();
reg_timer_r_masked_intr masked_intr;
reg_timer_rw_ack_intr ack_intr = { 0 };
@@ -202,11 +221,11 @@ timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
if (!masked_intr.tmr0)
return IRQ_NONE;

- /* acknowledge the timer irq */
+ /* Acknowledge the timer irq. */
ack_intr.tmr0 = 1;
REG_WR(timer, timer_regs[cpu], rw_ack_intr, ack_intr);

- /* reset watchdog otherwise it resets us! */
+ /* Reset watchdog otherwise it resets us! */
reset_watchdog();

/* Update statistics. */
@@ -218,7 +237,7 @@ timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
if (cpu != 0)
return IRQ_HANDLED;

- /* call the real timer interrupt handler */
+ /* Call the real timer interrupt handler */
do_timer(1);

/*
@@ -236,17 +255,17 @@ timer_interrupt(int irq, void *dev_id, struct pt_regs *regs)
if (set_rtc_mmss(xtime.tv_sec) == 0)
last_rtc_update = xtime.tv_sec;
else
- last_rtc_update = xtime.tv_sec - 600; /* do it again in 60 s */
+ /* Do it again in 60 s */
+ last_rtc_update = xtime.tv_sec - 600;
}
return IRQ_HANDLED;
}

-/* timer is IRQF_SHARED so drivers can add stuff to the timer irq chain
- * it needs to be IRQF_DISABLED to make the jiffies update work properly
+/* Timer is IRQF_SHARED so drivers can add stuff to the timer irq chain.
+ * It needs to be IRQF_DISABLED to make the jiffies update work properly.
*/
-
-static struct irqaction irq_timer = {
- .mask = timer_interrupt,
+static struct irqaction irq_timer = {
+ .handler = timer_interrupt,
.flags = IRQF_SHARED | IRQF_DISABLED,
.mask = CPU_MASK_NONE,
.name = "timer"
@@ -256,27 +275,27 @@ void __init
cris_timer_init(void)
{
int cpu = smp_processor_id();
- reg_timer_rw_tmr0_ctrl tmr0_ctrl = { 0 };
- reg_timer_rw_tmr0_div tmr0_div = TIMER0_DIV;
+ reg_timer_rw_tmr0_ctrl tmr0_ctrl = { 0 };
+ reg_timer_rw_tmr0_div tmr0_div = TIMER0_DIV;
reg_timer_rw_intr_mask timer_intr_mask;

- /* Setup the etrax timers
+ /* Setup the etrax timers.
* Base frequency is 100MHz, divider 1000000 -> 100 HZ
* We use timer0, so timer1 is free.
* The trig timer is used by the fasttimer API if enabled.
*/

- tmr0_ctrl.op = regk_timer_ld;
+ tmr0_ctrl.op = regk_timer_ld;
tmr0_ctrl.freq = regk_timer_f100;
- REG_WR(timer, timer_regs[cpu], rw_tmr0_div, tmr0_div);
- REG_WR(timer, timer_regs[cpu], rw_tmr0_ctrl, tmr0_ctrl); /* Load */
- tmr0_ctrl.op = regk_timer_run;
- REG_WR(timer, timer_regs[cpu], rw_tmr0_ctrl, tmr0_ctrl); /* Start */
-
- /* enable the timer irq */
- timer_intr_mask = REG_RD(timer, timer_regs[cpu], rw_intr_mask);
- timer_intr_mask.tmr0 = 1;
- REG_WR(timer, timer_regs[cpu], rw_intr_mask, timer_intr_mask);
+ REG_WR(timer, timer_regs[cpu], rw_tmr0_div, tmr0_div);
+ REG_WR(timer, timer_regs[cpu], rw_tmr0_ctrl, tmr0_ctrl); /* Load */
+ tmr0_ctrl.op = regk_timer_run;
+ REG_WR(timer, timer_regs[cpu], rw_tmr0_ctrl, tmr0_ctrl); /* Start */
+
+ /* Enable the timer irq. */
+ timer_intr_mask = REG_RD(timer, timer_regs[cpu], rw_intr_mask);
+ timer_intr_mask.tmr0 = 1;
+ REG_WR(timer, timer_regs[cpu], rw_intr_mask, timer_intr_mask);
}

void __init
@@ -284,7 +303,7 @@ time_init(void)
{
reg_intr_vect_rw_mask intr_mask;

- /* probe for the RTC and read it if it exists
+ /* Probe for the RTC and read it if it exists.
* Before the RTC can be probed the loops_per_usec variable needs
* to be initialized to make usleep work. A better value for
* loops_per_usec is calculated by the kernel later once the
@@ -293,52 +312,74 @@ time_init(void)
loops_per_usec = 50;

if(RTC_INIT() < 0) {
- /* no RTC, start at 1980 */
+ /* No RTC, start at 1980 */
xtime.tv_sec = 0;
xtime.tv_nsec = 0;
have_rtc = 0;
} else {
- /* get the current time */
+ /* Get the current time */
have_rtc = 1;
update_xtime_from_cmos();
}

/*
- * Initialize wall_to_monotonic such that adding it to xtime will yield zero, the
- * tv_nsec field must be normalized (i.e., 0 <= nsec < NSEC_PER_SEC).
+ * Initialize wall_to_monotonic such that adding it to
+ * xtime will yield zero, the tv_nsec field must be normalized
+ * (i.e., 0 <= nsec < NSEC_PER_SEC).
*/
set_normalized_timespec(&wall_to_monotonic, -xtime.tv_sec, -xtime.tv_nsec);

- /* Start CPU local timer */
+ /* Start CPU local timer. */
cris_timer_init();

- /* enable the timer irq in global config */
- intr_mask = REG_RD(intr_vect, regi_irq, rw_mask);
- intr_mask.timer = 1;
- REG_WR(intr_vect, regi_irq, rw_mask, intr_mask);
-
- /* now actually register the timer irq handler that calls timer_interrupt() */
+ /* Enable the timer irq in global config. */
+ intr_mask = REG_RD_VECT(intr_vect, regi_irq, rw_mask, 1);
+ intr_mask.timer0 = 1;
+ REG_WR_VECT(intr_vect, regi_irq, rw_mask, 1, intr_mask);

- setup_irq(TIMER_INTR_VECT, &irq_timer);
+ /* Now actually register the timer irq handler that calls
+ * timer_interrupt(). */
+ setup_irq(TIMER0_INTR_VECT, &irq_timer);

- /* enable watchdog if we should use one */
+ /* Enable watchdog if we should use one. */

#if defined(CONFIG_ETRAX_WATCHDOG)
- printk("Enabling watchdog...\n");
+ printk(KERN_INFO "Enabling watchdog...\n");
start_watchdog();

/* If we use the hardware watchdog, we want to trap it as an NMI
- and dump registers before it resets us. For this to happen, we
- must set the "m" NMI enable flag (which once set, is unset only
- when an NMI is taken).
-
- The same goes for the external NMI, but that doesn't have any
- driver or infrastructure support yet. */
- {
- unsigned long flags;
- local_save_flags(flags);
- flags |= (1<<30); /* NMI M flag is at bit 30 */
- local_irq_restore(flags);
- }
+ * and dump registers before it resets us. For this to happen, we
+ * must set the "m" NMI enable flag (which once set, is unset only
+ * when an NMI is taken). */
+ {
+ unsigned long flags;
+ local_save_flags(flags);
+ flags |= (1<<30); /* NMI M flag is at bit 30 */
+ local_irq_restore(flags);
+ }
+#endif
+
+#ifdef CONFIG_CPU_FREQ
+ cpufreq_register_notifier(&cris_time_freq_notifier_block,
+ CPUFREQ_TRANSITION_NOTIFIER);
#endif
}
+
+#ifdef CONFIG_CPU_FREQ
+static int
+cris_time_freq_notifier(struct notifier_block *nb, unsigned long val,
+ void *data)
+{
+ struct cpufreq_freqs *freqs = data;
+ if (val == CPUFREQ_POSTCHANGE) {
+ reg_timer_r_tmr0_data data;
+ reg_timer_rw_tmr0_div div = (freqs->new * 500) / HZ;
+ do {
+ data = REG_RD(timer, timer_regs[freqs->cpu],
+ r_tmr0_data);
+ } while (data > 20);
+ REG_WR(timer, timer_regs[freqs->cpu], rw_tmr0_div, div);
+ }
+ return 0;
+}
+#endif
--
1.5.3.6.970.gd25430

--
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/