x86: UV, SGI RTC: add UV RTC clocksource/clockevents
authorDimitri Sivanich <sivanich@sgi.com>
Wed, 4 Mar 2009 18:59:18 +0000 (12:59 -0600)
committerIngo Molnar <mingo@elte.hu>
Wed, 4 Mar 2009 19:25:38 +0000 (20:25 +0100)
This patch provides a high resolution clock/timer source using the
SGI UV system-wide synchronized RTC clock/timer hardware.

Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: john stultz <johnstul@us.ibm.com>
LKML-Reference: <20090304185918.GC24419@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
arch/x86/kernel/Makefile
arch/x86/kernel/uv_time.c [new file with mode: 0644]

index 95f216b..339ce35 100644 (file)
@@ -111,7 +111,7 @@ obj-$(CONFIG_SWIOTLB)                       += pci-swiotlb_64.o # NB rename without _64
 ###
 # 64 bit specific files
 ifeq ($(CONFIG_X86_64),y)
-       obj-$(CONFIG_X86_UV)            += tlb_uv.o bios_uv.o uv_irq.o uv_sysfs.o
+       obj-$(CONFIG_X86_UV)            += tlb_uv.o bios_uv.o uv_irq.o uv_sysfs.o uv_time.o
        obj-$(CONFIG_X86_PM_TIMER)      += pmtimer_64.o
        obj-$(CONFIG_AUDIT)             += audit_64.o
 
diff --git a/arch/x86/kernel/uv_time.c b/arch/x86/kernel/uv_time.c
new file mode 100644 (file)
index 0000000..6f8e325
--- /dev/null
@@ -0,0 +1,391 @@
+/*
+ * SGI RTC clock/timer routines.
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307 USA
+ *
+ *  Copyright (c) 2009 Silicon Graphics, Inc.  All Rights Reserved.
+ *  Copyright (c) Dimitri Sivanich
+ */
+#include <linux/clockchips.h>
+
+#include <asm/uv/uv_mmrs.h>
+#include <asm/uv/uv_hub.h>
+#include <asm/uv/bios.h>
+#include <asm/uv/uv.h>
+
+#define RTC_NAME               "sgi_rtc"
+
+static cycle_t uv_read_rtc(void);
+static int uv_rtc_next_event(unsigned long, struct clock_event_device *);
+static void uv_rtc_timer_setup(enum clock_event_mode,
+                               struct clock_event_device *);
+
+static struct clocksource clocksource_uv = {
+       .name           = RTC_NAME,
+       .rating         = 400,
+       .read           = uv_read_rtc,
+       .mask           = (cycle_t)UVH_RTC_REAL_TIME_CLOCK_MASK,
+       .shift          = 10,
+       .flags          = CLOCK_SOURCE_IS_CONTINUOUS,
+};
+
+static struct clock_event_device clock_event_device_uv = {
+       .name           = RTC_NAME,
+       .features       = CLOCK_EVT_FEAT_ONESHOT,
+       .shift          = 20,
+       .rating         = 400,
+       .irq            = -1,
+       .set_next_event = uv_rtc_next_event,
+       .set_mode       = uv_rtc_timer_setup,
+       .event_handler  = NULL,
+};
+
+static DEFINE_PER_CPU(struct clock_event_device, cpu_ced);
+
+/* There is one of these allocated per node */
+struct uv_rtc_timer_head {
+       spinlock_t      lock;
+       /* next cpu waiting for timer, local node relative: */
+       int             next_cpu;
+       /* number of cpus on this node: */
+       int             ncpus;
+       struct {
+               int     lcpu;           /* systemwide logical cpu number */
+               u64     expires;        /* next timer expiration for this cpu */
+       } cpu[1];
+};
+
+/*
+ * Access to uv_rtc_timer_head via blade id.
+ */
+static struct uv_rtc_timer_head                **blade_info __read_mostly;
+
+static int                             uv_rtc_enable;
+
+/*
+ * Hardware interface routines
+ */
+
+/* Send IPIs to another node */
+static void uv_rtc_send_IPI(int cpu)
+{
+       unsigned long apicid, val;
+       int pnode;
+
+       apicid = per_cpu(x86_cpu_to_apicid, cpu);
+       pnode = uv_apicid_to_pnode(apicid);
+       val = (1UL << UVH_IPI_INT_SEND_SHFT) |
+             (apicid << UVH_IPI_INT_APIC_ID_SHFT) |
+             (GENERIC_INTERRUPT_VECTOR << UVH_IPI_INT_VECTOR_SHFT);
+
+       uv_write_global_mmr64(pnode, UVH_IPI_INT, val);
+}
+
+/* Check for an RTC interrupt pending */
+static int uv_intr_pending(int pnode)
+{
+       return uv_read_global_mmr64(pnode, UVH_EVENT_OCCURRED0) &
+               UVH_EVENT_OCCURRED0_RTC1_MASK;
+}
+
+/* Setup interrupt and return non-zero if early expiration occurred. */
+static int uv_setup_intr(int cpu, u64 expires)
+{
+       u64 val;
+       int pnode = uv_cpu_to_pnode(cpu);
+
+       uv_write_global_mmr64(pnode, UVH_RTC1_INT_CONFIG,
+               UVH_RTC1_INT_CONFIG_M_MASK);
+       uv_write_global_mmr64(pnode, UVH_INT_CMPB, -1L);
+
+       uv_write_global_mmr64(pnode, UVH_EVENT_OCCURRED0_ALIAS,
+               UVH_EVENT_OCCURRED0_RTC1_MASK);
+
+       val = (GENERIC_INTERRUPT_VECTOR << UVH_RTC1_INT_CONFIG_VECTOR_SHFT) |
+               ((u64)cpu_physical_id(cpu) << UVH_RTC1_INT_CONFIG_APIC_ID_SHFT);
+
+       /* Set configuration */
+       uv_write_global_mmr64(pnode, UVH_RTC1_INT_CONFIG, val);
+       /* Initialize comparator value */
+       uv_write_global_mmr64(pnode, UVH_INT_CMPB, expires);
+
+       return (expires < uv_read_rtc() && !uv_intr_pending(pnode));
+}
+
+/*
+ * Per-cpu timer tracking routines
+ */
+
+static __init void uv_rtc_deallocate_timers(void)
+{
+       int bid;
+
+       for_each_possible_blade(bid) {
+               kfree(blade_info[bid]);
+       }
+       kfree(blade_info);
+}
+
+/* Allocate per-node list of cpu timer expiration times. */
+static __init int uv_rtc_allocate_timers(void)
+{
+       int cpu;
+
+       blade_info = kmalloc(uv_possible_blades * sizeof(void *), GFP_KERNEL);
+       if (!blade_info)
+               return -ENOMEM;
+       memset(blade_info, 0, uv_possible_blades * sizeof(void *));
+
+       for_each_present_cpu(cpu) {
+               int nid = cpu_to_node(cpu);
+               int bid = uv_cpu_to_blade_id(cpu);
+               int bcpu = uv_cpu_hub_info(cpu)->blade_processor_id;
+               struct uv_rtc_timer_head *head = blade_info[bid];
+
+               if (!head) {
+                       head = kmalloc_node(sizeof(struct uv_rtc_timer_head) +
+                               (uv_blade_nr_possible_cpus(bid) *
+                                       2 * sizeof(u64)),
+                               GFP_KERNEL, nid);
+                       if (!head) {
+                               uv_rtc_deallocate_timers();
+                               return -ENOMEM;
+                       }
+                       spin_lock_init(&head->lock);
+                       head->ncpus = uv_blade_nr_possible_cpus(bid);
+                       head->next_cpu = -1;
+                       blade_info[bid] = head;
+               }
+
+               head->cpu[bcpu].lcpu = cpu;
+               head->cpu[bcpu].expires = ULLONG_MAX;
+       }
+
+       return 0;
+}
+
+/* Find and set the next expiring timer.  */
+static void uv_rtc_find_next_timer(struct uv_rtc_timer_head *head, int pnode)
+{
+       u64 lowest = ULLONG_MAX;
+       int c, bcpu = -1;
+
+       head->next_cpu = -1;
+       for (c = 0; c < head->ncpus; c++) {
+               u64 exp = head->cpu[c].expires;
+               if (exp < lowest) {
+                       bcpu = c;
+                       lowest = exp;
+               }
+       }
+       if (bcpu >= 0) {
+               head->next_cpu = bcpu;
+               c = head->cpu[bcpu].lcpu;
+               if (uv_setup_intr(c, lowest))
+                       /* If we didn't set it up in time, trigger */
+                       uv_rtc_send_IPI(c);
+       } else {
+               uv_write_global_mmr64(pnode, UVH_RTC1_INT_CONFIG,
+                       UVH_RTC1_INT_CONFIG_M_MASK);
+       }
+}
+
+/*
+ * Set expiration time for current cpu.
+ *
+ * Returns 1 if we missed the expiration time.
+ */
+static int uv_rtc_set_timer(int cpu, u64 expires)
+{
+       int pnode = uv_cpu_to_pnode(cpu);
+       int bid = uv_cpu_to_blade_id(cpu);
+       struct uv_rtc_timer_head *head = blade_info[bid];
+       int bcpu = uv_cpu_hub_info(cpu)->blade_processor_id;
+       u64 *t = &head->cpu[bcpu].expires;
+       unsigned long flags;
+       int next_cpu;
+
+       spin_lock_irqsave(&head->lock, flags);
+
+       next_cpu = head->next_cpu;
+       *t = expires;
+       /* Will this one be next to go off? */
+       if (next_cpu < 0 || bcpu == next_cpu ||
+                       expires < head->cpu[next_cpu].expires) {
+               head->next_cpu = bcpu;
+               if (uv_setup_intr(cpu, expires)) {
+                       *t = ULLONG_MAX;
+                       uv_rtc_find_next_timer(head, pnode);
+                       spin_unlock_irqrestore(&head->lock, flags);
+                       return 1;
+               }
+       }
+
+       spin_unlock_irqrestore(&head->lock, flags);
+       return 0;
+}
+
+/*
+ * Unset expiration time for current cpu.
+ *
+ * Returns 1 if this timer was pending.
+ */
+static int uv_rtc_unset_timer(int cpu)
+{
+       int pnode = uv_cpu_to_pnode(cpu);
+       int bid = uv_cpu_to_blade_id(cpu);
+       struct uv_rtc_timer_head *head = blade_info[bid];
+       int bcpu = uv_cpu_hub_info(cpu)->blade_processor_id;
+       u64 *t = &head->cpu[bcpu].expires;
+       unsigned long flags;
+       int rc = 0;
+
+       spin_lock_irqsave(&head->lock, flags);
+
+       if (head->next_cpu == bcpu && uv_read_rtc() >= *t)
+               rc = 1;
+
+       *t = ULLONG_MAX;
+
+       /* Was the hardware setup for this timer? */
+       if (head->next_cpu == bcpu)
+               uv_rtc_find_next_timer(head, pnode);
+
+       spin_unlock_irqrestore(&head->lock, flags);
+
+       return rc;
+}
+
+
+/*
+ * Kernel interface routines.
+ */
+
+/*
+ * Read the RTC.
+ */
+static cycle_t uv_read_rtc(void)
+{
+       return (cycle_t)uv_read_local_mmr(UVH_RTC);
+}
+
+/*
+ * Program the next event, relative to now
+ */
+static int uv_rtc_next_event(unsigned long delta,
+                            struct clock_event_device *ced)
+{
+       int ced_cpu = cpumask_first(ced->cpumask);
+
+       return uv_rtc_set_timer(ced_cpu, delta + uv_read_rtc());
+}
+
+/*
+ * Setup the RTC timer in oneshot mode
+ */
+static void uv_rtc_timer_setup(enum clock_event_mode mode,
+                              struct clock_event_device *evt)
+{
+       int ced_cpu = cpumask_first(evt->cpumask);
+
+       switch (mode) {
+       case CLOCK_EVT_MODE_PERIODIC:
+       case CLOCK_EVT_MODE_ONESHOT:
+       case CLOCK_EVT_MODE_RESUME:
+               /* Nothing to do here yet */
+               break;
+       case CLOCK_EVT_MODE_UNUSED:
+       case CLOCK_EVT_MODE_SHUTDOWN:
+               uv_rtc_unset_timer(ced_cpu);
+               break;
+       }
+}
+
+static void uv_rtc_interrupt(void)
+{
+       struct clock_event_device *ced = &__get_cpu_var(cpu_ced);
+       int cpu = smp_processor_id();
+
+       if (!ced || !ced->event_handler)
+               return;
+
+       if (uv_rtc_unset_timer(cpu) != 1)
+               return;
+
+       ced->event_handler(ced);
+}
+
+static int __init uv_enable_rtc(char *str)
+{
+       uv_rtc_enable = 1;
+
+       return 1;
+}
+__setup("uvrtc", uv_enable_rtc);
+
+static __init void uv_rtc_register_clockevents(struct work_struct *dummy)
+{
+       struct clock_event_device *ced = &__get_cpu_var(cpu_ced);
+
+       *ced = clock_event_device_uv;
+       ced->cpumask = cpumask_of(smp_processor_id());
+       clockevents_register_device(ced);
+}
+
+static __init int uv_rtc_setup_clock(void)
+{
+       int rc;
+
+       if (!uv_rtc_enable || !is_uv_system() || generic_interrupt_extension)
+               return -ENODEV;
+
+       generic_interrupt_extension = uv_rtc_interrupt;
+
+       clocksource_uv.mult = clocksource_hz2mult(sn_rtc_cycles_per_second,
+                               clocksource_uv.shift);
+
+       rc = clocksource_register(&clocksource_uv);
+       if (rc) {
+               generic_interrupt_extension = NULL;
+               return rc;
+       }
+
+       /* Setup and register clockevents */
+       rc = uv_rtc_allocate_timers();
+       if (rc) {
+               clocksource_unregister(&clocksource_uv);
+               generic_interrupt_extension = NULL;
+               return rc;
+       }
+
+       clock_event_device_uv.mult = div_sc(sn_rtc_cycles_per_second,
+                               NSEC_PER_SEC, clock_event_device_uv.shift);
+
+       clock_event_device_uv.min_delta_ns = NSEC_PER_SEC /
+                                               sn_rtc_cycles_per_second;
+
+       clock_event_device_uv.max_delta_ns = clocksource_uv.mask *
+                               (NSEC_PER_SEC / sn_rtc_cycles_per_second);
+
+       rc = schedule_on_each_cpu(uv_rtc_register_clockevents);
+       if (rc) {
+               clocksource_unregister(&clocksource_uv);
+               generic_interrupt_extension = NULL;
+               uv_rtc_deallocate_timers();
+       }
+
+       return rc;
+}
+arch_initcall(uv_rtc_setup_clock);