[MIPS] SMTC: Fix SMTC dyntick support.
authorKevin D. Kissell <kevink@paralogos.com>
Tue, 9 Sep 2008 19:48:52 +0000 (21:48 +0200)
committerRalf Baechle <ralf@linux-mips.org>
Fri, 3 Oct 2008 16:58:58 +0000 (17:58 +0100)
Rework of SMTC support to make it work with the new clock event system,
allowing "tickless" operation, and to make it compatible with the use of
the "wait_irqoff" idle loop.  The new clocking scheme means that the
previously optional IPI instant replay mechanism is now required, and has
been made more robust.

Signed-off-by: Kevin D. Kissell <kevink@paralogos.com>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
arch/mips/Kconfig
arch/mips/kernel/Makefile
arch/mips/kernel/cevt-r4k.c
arch/mips/kernel/cevt-smtc.c [new file with mode: 0644]
arch/mips/kernel/cpu-probe.c
arch/mips/kernel/genex.S
arch/mips/kernel/smtc.c
arch/mips/mti-malta/malta-smtc.c
include/asm-mips/cevt-r4k.h [new file with mode: 0644]
include/asm-mips/irqflags.h
include/asm-mips/smtc.h

index 49896a2a1d722e0760a866487f8b80b7d8af8fb0..c930b8ceb4182e5c1172003966e43a8a90b93440 100644 (file)
@@ -1403,7 +1403,6 @@ config MIPS_MT_SMTC
        depends on CPU_MIPS32_R2
        #depends on CPU_MIPS64_R2               # once there is hardware ...
        depends on SYS_SUPPORTS_MULTITHREADING
-       select GENERIC_CLOCKEVENTS_BROADCAST
        select CPU_MIPSR2_IRQ_VI
        select CPU_MIPSR2_IRQ_EI
        select MIPS_MT
@@ -1451,32 +1450,17 @@ config MIPS_VPE_LOADER
          Includes a loader for loading an elf relocatable object
          onto another VPE and running it.
 
-config MIPS_MT_SMTC_INSTANT_REPLAY
-       bool "Low-latency Dispatch of Deferred SMTC IPIs"
-       depends on MIPS_MT_SMTC && !PREEMPT
-       default y
-       help
-         SMTC pseudo-interrupts between TCs are deferred and queued
-         if the target TC is interrupt-inhibited (IXMT). In the first
-         SMTC prototypes, these queued IPIs were serviced on return
-         to user mode, or on entry into the kernel idle loop. The
-         INSTANT_REPLAY option dispatches them as part of local_irq_restore()
-         processing, which adds runtime overhead (hence the option to turn
-         it off), but ensures that IPIs are handled promptly even under
-         heavy I/O interrupt load.
-
 config MIPS_MT_SMTC_IM_BACKSTOP
        bool "Use per-TC register bits as backstop for inhibited IM bits"
        depends on MIPS_MT_SMTC
-       default y
+       default n
        help
          To support multiple TC microthreads acting as "CPUs" within
          a VPE, VPE-wide interrupt mask bits must be specially manipulated
          during interrupt handling. To support legacy drivers and interrupt
          controller management code, SMTC has a "backstop" to track and
          if necessary restore the interrupt mask. This has some performance
-         impact on interrupt service overhead. Disable it only if you know
-         what you are doing.
+         impact on interrupt service overhead.
 
 config MIPS_MT_SMTC_IRQAFF
        bool "Support IRQ affinity API"
@@ -1486,10 +1470,8 @@ config MIPS_MT_SMTC_IRQAFF
          Enables SMP IRQ affinity API (/proc/irq/*/smp_affinity, etc.)
          for SMTC Linux kernel. Requires platform support, of which
          an example can be found in the MIPS kernel i8259 and Malta
-         platform code.  It is recommended that MIPS_MT_SMTC_INSTANT_REPLAY
-         be enabled if MIPS_MT_SMTC_IRQAFF is used. Adds overhead to
-         interrupt dispatch, and should be used only if you know what
-         you are doing.
+         platform code.  Adds some overhead to interrupt dispatch, and
+         should be used only if you know what you are doing.
 
 config MIPS_VPE_LOADER_TOM
        bool "Load VPE program into memory hidden from linux"
index 706f9397479770265d69b0307df9d76ea3739178..25775cb54000dd3871673c512e43cf59398c6461 100644 (file)
@@ -10,6 +10,7 @@ obj-y         += cpu-probe.o branch.o entry.o genex.o irq.o process.o \
 
 obj-$(CONFIG_CEVT_BCM1480)     += cevt-bcm1480.o
 obj-$(CONFIG_CEVT_R4K)         += cevt-r4k.o
+obj-$(CONFIG_MIPS_MT_SMTC)     += cevt-smtc.o
 obj-$(CONFIG_CEVT_DS1287)      += cevt-ds1287.o
 obj-$(CONFIG_CEVT_GT641XX)     += cevt-gt641xx.o
 obj-$(CONFIG_CEVT_SB1250)      += cevt-sb1250.o
index 24a2d907aa0de4c022c0d5e89227dceae2a8a572..4a4c59f2737a5d0bcdc8fc48f4f8e28ea217f528 100644 (file)
 
 #include <asm/smtc_ipi.h>
 #include <asm/time.h>
+#include <asm/cevt-r4k.h>
+
+/*
+ * The SMTC Kernel for the 34K, 1004K, et. al. replaces several
+ * of these routines with SMTC-specific variants.
+ */
+
+#ifndef CONFIG_MIPS_MT_SMTC
 
 static int mips_next_event(unsigned long delta,
                            struct clock_event_device *evt)
@@ -19,60 +27,27 @@ static int mips_next_event(unsigned long delta,
        unsigned int cnt;
        int res;
 
-#ifdef CONFIG_MIPS_MT_SMTC
-       {
-       unsigned long flags, vpflags;
-       local_irq_save(flags);
-       vpflags = dvpe();
-#endif
        cnt = read_c0_count();
        cnt += delta;
        write_c0_compare(cnt);
        res = ((int)(read_c0_count() - cnt) > 0) ? -ETIME : 0;
-#ifdef CONFIG_MIPS_MT_SMTC
-       evpe(vpflags);
-       local_irq_restore(flags);
-       }
-#endif
        return res;
 }
 
-static void mips_set_mode(enum clock_event_mode mode,
-                          struct clock_event_device *evt)
+#endif /* CONFIG_MIPS_MT_SMTC */
+
+void mips_set_clock_mode(enum clock_event_mode mode,
+                               struct clock_event_device *evt)
 {
        /* Nothing to do ...  */
 }
 
-static DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device);
-static int cp0_timer_irq_installed;
+DEFINE_PER_CPU(struct clock_event_device, mips_clockevent_device);
+int cp0_timer_irq_installed;
 
-/*
- * Timer ack for an R4k-compatible timer of a known frequency.
- */
-static void c0_timer_ack(void)
-{
-       write_c0_compare(read_c0_compare());
-}
+#ifndef CONFIG_MIPS_MT_SMTC
 
-/*
- * Possibly handle a performance counter interrupt.
- * Return true if the timer interrupt should not be checked
- */
-static inline int handle_perf_irq(int r2)
-{
-       /*
-        * The performance counter overflow interrupt may be shared with the
-        * timer interrupt (cp0_perfcount_irq < 0). If it is and a
-        * performance counter has overflowed (perf_irq() == IRQ_HANDLED)
-        * and we can't reliably determine if a counter interrupt has also
-        * happened (!r2) then don't check for a timer interrupt.
-        */
-       return (cp0_perfcount_irq < 0) &&
-               perf_irq() == IRQ_HANDLED &&
-               !r2;
-}
-
-static irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
+irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
 {
        const int r2 = cpu_has_mips_r2;
        struct clock_event_device *cd;
@@ -93,12 +68,8 @@ static irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
         * interrupt.  Being the paranoiacs we are we check anyway.
         */
        if (!r2 || (read_c0_cause() & (1 << 30))) {
-               c0_timer_ack();
-#ifdef CONFIG_MIPS_MT_SMTC
-               if (cpu_data[cpu].vpe_id)
-                       goto out;
-               cpu = 0;
-#endif
+               /* Clear Count/Compare Interrupt */
+               write_c0_compare(read_c0_compare());
                cd = &per_cpu(mips_clockevent_device, cpu);
                cd->event_handler(cd);
        }
@@ -107,65 +78,16 @@ out:
        return IRQ_HANDLED;
 }
 
-static struct irqaction c0_compare_irqaction = {
+#endif /* Not CONFIG_MIPS_MT_SMTC */
+
+struct irqaction c0_compare_irqaction = {
        .handler = c0_compare_interrupt,
-#ifdef CONFIG_MIPS_MT_SMTC
-       .flags = IRQF_DISABLED,
-#else
        .flags = IRQF_DISABLED | IRQF_PERCPU,
-#endif
        .name = "timer",
 };
 
-#ifdef CONFIG_MIPS_MT_SMTC
-DEFINE_PER_CPU(struct clock_event_device, smtc_dummy_clockevent_device);
-
-static void smtc_set_mode(enum clock_event_mode mode,
-                          struct clock_event_device *evt)
-{
-}
-
-static void mips_broadcast(cpumask_t mask)
-{
-       unsigned int cpu;
-
-       for_each_cpu_mask(cpu, mask)
-               smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0);
-}
-
-static void setup_smtc_dummy_clockevent_device(void)
-{
-       //uint64_t mips_freq = mips_hpt_^frequency;
-       unsigned int cpu = smp_processor_id();
-       struct clock_event_device *cd;
 
-       cd = &per_cpu(smtc_dummy_clockevent_device, cpu);
-
-       cd->name                = "SMTC";
-       cd->features            = CLOCK_EVT_FEAT_DUMMY;
-
-       /* Calculate the min / max delta */
-       cd->mult        = 0; //div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32);
-       cd->shift               = 0; //32;
-       cd->max_delta_ns        = 0; //clockevent_delta2ns(0x7fffffff, cd);
-       cd->min_delta_ns        = 0; //clockevent_delta2ns(0x30, cd);
-
-       cd->rating              = 200;
-       cd->irq                 = 17; //-1;
-//     if (cpu)
-//             cd->cpumask     = CPU_MASK_ALL; // cpumask_of_cpu(cpu);
-//     else
-               cd->cpumask     = cpumask_of_cpu(cpu);
-
-       cd->set_mode            = smtc_set_mode;
-
-       cd->broadcast           = mips_broadcast;
-
-       clockevents_register_device(cd);
-}
-#endif
-
-static void mips_event_handler(struct clock_event_device *dev)
+void mips_event_handler(struct clock_event_device *dev)
 {
 }
 
@@ -177,7 +99,23 @@ static int c0_compare_int_pending(void)
        return (read_c0_cause() >> cp0_compare_irq) & 0x100;
 }
 
-static int c0_compare_int_usable(void)
+/*
+ * Compare interrupt can be routed and latched outside the core,
+ * so a single execution hazard barrier may not be enough to give
+ * it time to clear as seen in the Cause register.  4 time the
+ * pipeline depth seems reasonably conservative, and empirically
+ * works better in configurations with high CPU/bus clock ratios.
+ */
+
+#define compare_change_hazard() \
+       do { \
+               irq_disable_hazard(); \
+               irq_disable_hazard(); \
+               irq_disable_hazard(); \
+               irq_disable_hazard(); \
+       } while (0)
+
+int c0_compare_int_usable(void)
 {
        unsigned int delta;
        unsigned int cnt;
@@ -187,7 +125,7 @@ static int c0_compare_int_usable(void)
         */
        if (c0_compare_int_pending()) {
                write_c0_compare(read_c0_count());
-               irq_disable_hazard();
+               compare_change_hazard();
                if (c0_compare_int_pending())
                        return 0;
        }
@@ -196,7 +134,7 @@ static int c0_compare_int_usable(void)
                cnt = read_c0_count();
                cnt += delta;
                write_c0_compare(cnt);
-               irq_disable_hazard();
+               compare_change_hazard();
                if ((int)(read_c0_count() - cnt) < 0)
                    break;
                /* increase delta if the timer was already expired */
@@ -205,11 +143,12 @@ static int c0_compare_int_usable(void)
        while ((int)(read_c0_count() - cnt) <= 0)
                ;       /* Wait for expiry  */
 
+       compare_change_hazard();
        if (!c0_compare_int_pending())
                return 0;
 
        write_c0_compare(read_c0_count());
-       irq_disable_hazard();
+       compare_change_hazard();
        if (c0_compare_int_pending())
                return 0;
 
@@ -219,6 +158,8 @@ static int c0_compare_int_usable(void)
        return 1;
 }
 
+#ifndef CONFIG_MIPS_MT_SMTC
+
 int __cpuinit mips_clockevent_init(void)
 {
        uint64_t mips_freq = mips_hpt_frequency;
@@ -229,17 +170,6 @@ int __cpuinit mips_clockevent_init(void)
        if (!cpu_has_counter || !mips_hpt_frequency)
                return -ENXIO;
 
-#ifdef CONFIG_MIPS_MT_SMTC
-       setup_smtc_dummy_clockevent_device();
-
-       /*
-        * On SMTC we only register VPE0's compare interrupt as clockevent
-        * device.
-        */
-       if (cpu)
-               return 0;
-#endif
-
        if (!c0_compare_int_usable())
                return -ENXIO;
 
@@ -265,13 +195,9 @@ int __cpuinit mips_clockevent_init(void)
 
        cd->rating              = 300;
        cd->irq                 = irq;
-#ifdef CONFIG_MIPS_MT_SMTC
-       cd->cpumask             = CPU_MASK_ALL;
-#else
        cd->cpumask             = cpumask_of_cpu(cpu);
-#endif
        cd->set_next_event      = mips_next_event;
-       cd->set_mode            = mips_set_mode;
+       cd->set_mode            = mips_set_clock_mode;
        cd->event_handler       = mips_event_handler;
 
        clockevents_register_device(cd);
@@ -281,12 +207,9 @@ int __cpuinit mips_clockevent_init(void)
 
        cp0_timer_irq_installed = 1;
 
-#ifdef CONFIG_MIPS_MT_SMTC
-#define CPUCTR_IMASKBIT (0x100 << cp0_compare_irq)
-       setup_irq_smtc(irq, &c0_compare_irqaction, CPUCTR_IMASKBIT);
-#else
        setup_irq(irq, &c0_compare_irqaction);
-#endif
 
        return 0;
 }
+
+#endif /* Not CONFIG_MIPS_MT_SMTC */
diff --git a/arch/mips/kernel/cevt-smtc.c b/arch/mips/kernel/cevt-smtc.c
new file mode 100644 (file)
index 0000000..5162fe4
--- /dev/null
@@ -0,0 +1,321 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2007 MIPS Technologies, Inc.
+ * Copyright (C) 2007 Ralf Baechle <ralf@linux-mips.org>
+ * Copyright (C) 2008 Kevin D. Kissell, Paralogos sarl
+ */
+#include <linux/clockchips.h>
+#include <linux/interrupt.h>
+#include <linux/percpu.h>
+
+#include <asm/smtc_ipi.h>
+#include <asm/time.h>
+#include <asm/cevt-r4k.h>
+
+/*
+ * Variant clock event timer support for SMTC on MIPS 34K, 1004K
+ * or other MIPS MT cores.
+ *
+ * Notes on SMTC Support:
+ *
+ * SMTC has multiple microthread TCs pretending to be Linux CPUs.
+ * But there's only one Count/Compare pair per VPE, and Compare
+ * interrupts are taken opportunisitically by available TCs
+ * bound to the VPE with the Count register.  The new timer
+ * framework provides for global broadcasts, but we really
+ * want VPE-level multicasts for best behavior. So instead
+ * of invoking the high-level clock-event broadcast code,
+ * this version of SMTC support uses the historical SMTC
+ * multicast mechanisms "under the hood", appearing to the
+ * generic clock layer as if the interrupts are per-CPU.
+ *
+ * The approach taken here is to maintain a set of NR_CPUS
+ * virtual timers, and track which "CPU" needs to be alerted
+ * at each event.
+ *
+ * It's unlikely that we'll see a MIPS MT core with more than
+ * 2 VPEs, but we *know* that we won't need to handle more
+ * VPEs than we have "CPUs".  So NCPUs arrays of NCPUs elements
+ * is always going to be overkill, but always going to be enough.
+ */
+
+unsigned long smtc_nexttime[NR_CPUS][NR_CPUS];
+static int smtc_nextinvpe[NR_CPUS];
+
+/*
+ * Timestamps stored are absolute values to be programmed
+ * into Count register.  Valid timestamps will never be zero.
+ * If a Zero Count value is actually calculated, it is converted
+ * to be a 1, which will introduce 1 or two CPU cycles of error
+ * roughly once every four billion events, which at 1000 HZ means
+ * about once every 50 days.  If that's actually a problem, one
+ * could alternate squashing 0 to 1 and to -1.
+ */
+
+#define MAKEVALID(x) (((x) == 0L) ? 1L : (x))
+#define ISVALID(x) ((x) != 0L)
+
+/*
+ * Time comparison is subtle, as it's really truncated
+ * modular arithmetic.
+ */
+
+#define IS_SOONER(a, b, reference) \
+    (((a) - (unsigned long)(reference)) < ((b) - (unsigned long)(reference)))
+
+/*
+ * CATCHUP_INCREMENT, used when the function falls behind the counter.
+ * Could be an increasing function instead of a constant;
+ */
+
+#define CATCHUP_INCREMENT 64
+
+static int mips_next_event(unsigned long delta,
+                               struct clock_event_device *evt)
+{
+       unsigned long flags;
+       unsigned int mtflags;
+       unsigned long timestamp, reference, previous;
+       unsigned long nextcomp = 0L;
+       int vpe = current_cpu_data.vpe_id;
+       int cpu = smp_processor_id();
+       local_irq_save(flags);
+       mtflags = dmt();
+
+       /*
+        * Maintain the per-TC virtual timer
+        * and program the per-VPE shared Count register
+        * as appropriate here...
+        */
+       reference = (unsigned long)read_c0_count();
+       timestamp = MAKEVALID(reference + delta);
+       /*
+        * To really model the clock, we have to catch the case
+        * where the current next-in-VPE timestamp is the old
+        * timestamp for the calling CPE, but the new value is
+        * in fact later.  In that case, we have to do a full
+        * scan and discover the new next-in-VPE CPU id and
+        * timestamp.
+        */
+       previous = smtc_nexttime[vpe][cpu];
+       if (cpu == smtc_nextinvpe[vpe] && ISVALID(previous)
+           && IS_SOONER(previous, timestamp, reference)) {
+               int i;
+               int soonest = cpu;
+
+               /*
+                * Update timestamp array here, so that new
+                * value gets considered along with those of
+                * other virtual CPUs on the VPE.
+                */
+               smtc_nexttime[vpe][cpu] = timestamp;
+               for_each_online_cpu(i) {
+                       if (ISVALID(smtc_nexttime[vpe][i])
+                           && IS_SOONER(smtc_nexttime[vpe][i],
+                               smtc_nexttime[vpe][soonest], reference)) {
+                                   soonest = i;
+                       }
+               }
+               smtc_nextinvpe[vpe] = soonest;
+               nextcomp = smtc_nexttime[vpe][soonest];
+       /*
+        * Otherwise, we don't have to process the whole array rank,
+        * we just have to see if the event horizon has gotten closer.
+        */
+       } else {
+               if (!ISVALID(smtc_nexttime[vpe][smtc_nextinvpe[vpe]]) ||
+                   IS_SOONER(timestamp,
+                       smtc_nexttime[vpe][smtc_nextinvpe[vpe]], reference)) {
+                           smtc_nextinvpe[vpe] = cpu;
+                           nextcomp = timestamp;
+               }
+               /*
+                * Since next-in-VPE may me the same as the executing
+                * virtual CPU, we update the array *after* checking
+                * its value.
+                */
+               smtc_nexttime[vpe][cpu] = timestamp;
+       }
+
+       /*
+        * It may be that, in fact, we don't need to update Compare,
+        * but if we do, we want to make sure we didn't fall into
+        * a crack just behind Count.
+        */
+       if (ISVALID(nextcomp)) {
+               write_c0_compare(nextcomp);
+               ehb();
+               /*
+                * We never return an error, we just make sure
+                * that we trigger the handlers as quickly as
+                * we can if we fell behind.
+                */
+               while ((nextcomp - (unsigned long)read_c0_count())
+                       > (unsigned long)LONG_MAX) {
+                       nextcomp += CATCHUP_INCREMENT;
+                       write_c0_compare(nextcomp);
+                       ehb();
+               }
+       }
+       emt(mtflags);
+       local_irq_restore(flags);
+       return 0;
+}
+
+
+void smtc_distribute_timer(int vpe)
+{
+       unsigned long flags;
+       unsigned int mtflags;
+       int cpu;
+       struct clock_event_device *cd;
+       unsigned long nextstamp = 0L;
+       unsigned long reference;
+
+
+repeat:
+       for_each_online_cpu(cpu) {
+           /*
+            * Find virtual CPUs within the current VPE who have
+            * unserviced timer requests whose time is now past.
+            */
+           local_irq_save(flags);
+           mtflags = dmt();
+           if (cpu_data[cpu].vpe_id == vpe &&
+               ISVALID(smtc_nexttime[vpe][cpu])) {
+               reference = (unsigned long)read_c0_count();
+               if ((smtc_nexttime[vpe][cpu] - reference)
+                        > (unsigned long)LONG_MAX) {
+                           smtc_nexttime[vpe][cpu] = 0L;
+                           emt(mtflags);
+                           local_irq_restore(flags);
+                           /*
+                            * We don't send IPIs to ourself.
+                            */
+                           if (cpu != smp_processor_id()) {
+                               smtc_send_ipi(cpu, SMTC_CLOCK_TICK, 0);
+                           } else {
+                               cd = &per_cpu(mips_clockevent_device, cpu);
+                               cd->event_handler(cd);
+                           }
+               } else {
+                       /* Local to VPE but Valid Time not yet reached. */
+                       if (!ISVALID(nextstamp) ||
+                           IS_SOONER(smtc_nexttime[vpe][cpu], nextstamp,
+                           reference)) {
+                               smtc_nextinvpe[vpe] = cpu;
+                               nextstamp = smtc_nexttime[vpe][cpu];
+                       }
+                       emt(mtflags);
+                       local_irq_restore(flags);
+               }
+           } else {
+               emt(mtflags);
+               local_irq_restore(flags);
+
+           }
+       }
+       /* Reprogram for interrupt at next soonest timestamp for VPE */
+       if (ISVALID(nextstamp)) {
+               write_c0_compare(nextstamp);
+               ehb();
+               if ((nextstamp - (unsigned long)read_c0_count())
+                       > (unsigned long)LONG_MAX)
+                               goto repeat;
+       }
+}
+
+
+irqreturn_t c0_compare_interrupt(int irq, void *dev_id)
+{
+       int cpu = smp_processor_id();
+
+       /* If we're running SMTC, we've got MIPS MT and therefore MIPS32R2 */
+       handle_perf_irq(1);
+
+       if (read_c0_cause() & (1 << 30)) {
+               /* Clear Count/Compare Interrupt */
+               write_c0_compare(read_c0_compare());
+               smtc_distribute_timer(cpu_data[cpu].vpe_id);
+       }
+       return IRQ_HANDLED;
+}
+
+
+int __cpuinit mips_clockevent_init(void)
+{
+       uint64_t mips_freq = mips_hpt_frequency;
+       unsigned int cpu = smp_processor_id();
+       struct clock_event_device *cd;
+       unsigned int irq;
+       int i;
+       int j;
+
+       if (!cpu_has_counter || !mips_hpt_frequency)
+               return -ENXIO;
+       if (cpu == 0) {
+               for (i = 0; i < num_possible_cpus(); i++) {
+                       smtc_nextinvpe[i] = 0;
+                       for (j = 0; j < num_possible_cpus(); j++)
+                               smtc_nexttime[i][j] = 0L;
+               }
+               /*
+                * SMTC also can't have the usablility test
+                * run by secondary TCs once Compare is in use.
+                */
+               if (!c0_compare_int_usable())
+                       return -ENXIO;
+       }
+
+       /*
+        * With vectored interrupts things are getting platform specific.
+        * get_c0_compare_int is a hook to allow a platform to return the
+        * interrupt number of it's liking.
+        */
+       irq = MIPS_CPU_IRQ_BASE + cp0_compare_irq;
+       if (get_c0_compare_int)
+               irq = get_c0_compare_int();
+
+       cd = &per_cpu(mips_clockevent_device, cpu);
+
+       cd->name                = "MIPS";
+       cd->features            = CLOCK_EVT_FEAT_ONESHOT;
+
+       /* Calculate the min / max delta */
+       cd->mult        = div_sc((unsigned long) mips_freq, NSEC_PER_SEC, 32);
+       cd->shift               = 32;
+       cd->max_delta_ns        = clockevent_delta2ns(0x7fffffff, cd);
+       cd->min_delta_ns        = clockevent_delta2ns(0x300, cd);
+
+       cd->rating              = 300;
+       cd->irq                 = irq;
+       cd->cpumask             = cpumask_of_cpu(cpu);
+       cd->set_next_event      = mips_next_event;
+       cd->set_mode            = mips_set_clock_mode;
+       cd->event_handler       = mips_event_handler;
+
+       clockevents_register_device(cd);
+
+       /*
+        * On SMTC we only want to do the data structure
+        * initialization and IRQ setup once.
+        */
+       if (cpu)
+               return 0;
+       /*
+        * And we need the hwmask associated with the c0_compare
+        * vector to be initialized.
+        */
+       irq_hwmask[irq] = (0x100 << cp0_compare_irq);
+       if (cp0_timer_irq_installed)
+               return 0;
+
+       cp0_timer_irq_installed = 1;
+
+       setup_irq(irq, &c0_compare_irqaction);
+
+       return 0;
+}
index 11c92dc53791bc2d6b62c727c09fe48cc032a7de..e621fda8ab37fad4847cb3a19ee238f415f82a91 100644 (file)
@@ -54,14 +54,18 @@ extern void r4k_wait(void);
  * interrupt is requested" restriction in the MIPS32/MIPS64 architecture makes
  * using this version a gamble.
  */
-static void r4k_wait_irqoff(void)
+void r4k_wait_irqoff(void)
 {
        local_irq_disable();
        if (!need_resched())
-               __asm__("       .set    mips3           \n"
+               __asm__("       .set    push            \n"
+                       "       .set    mips3           \n"
                        "       wait                    \n"
-                       "       .set    mips0           \n");
+                       "       .set    pop             \n");
        local_irq_enable();
+       __asm__("       .globl __pastwait       \n"
+               "__pastwait:                    \n");
+       return;
 }
 
 /*
index f886dd7f708e1d13f833fce5db545b737f15e684..01dcbe38fa019dc20ac8d4999b818f62ec0ff970 100644 (file)
@@ -282,8 +282,8 @@ NESTED(except_vec_vi_handler, 0, sp)
        and     t0, a0, t1
 #ifdef CONFIG_MIPS_MT_SMTC_IM_BACKSTOP
        mfc0    t2, CP0_TCCONTEXT
-       or      t0, t0, t2
-       mtc0    t0, CP0_TCCONTEXT
+       or      t2, t0, t2
+       mtc0    t2, CP0_TCCONTEXT
 #endif /* CONFIG_MIPS_MT_SMTC_IM_BACKSTOP */
        xor     t1, t1, t0
        mtc0    t1, CP0_STATUS
index 39b491b9ad87ab8beeba2cf313d79a8e380dddcf..897fb2b4751c95715af0b798d79db5cf899a5e5d 100644 (file)
@@ -1,4 +1,21 @@
-/* Copyright (C) 2004 Mips Technologies, Inc */
+/*
+ * 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) 2004 Mips Technologies, Inc
+ * Copyright (C) 2008 Kevin D. Kissell
+ */
 
 #include <linux/clockchips.h>
 #include <linux/kernel.h>
@@ -21,7 +38,6 @@
 #include <asm/time.h>
 #include <asm/addrspace.h>
 #include <asm/smtc.h>
-#include <asm/smtc_ipi.h>
 #include <asm/smtc_proc.h>
 
 /*
@@ -58,11 +74,6 @@ unsigned long irq_hwmask[NR_IRQS];
 
 asiduse smtc_live_asid[MAX_SMTC_TLBS][MAX_SMTC_ASIDS];
 
-/*
- * Clock interrupt "latch" buffers, per "CPU"
- */
-
-static atomic_t ipi_timer_latch[NR_CPUS];
 
 /*
  * Number of InterProcessor Interrupt (IPI) message buffers to allocate
@@ -282,7 +293,7 @@ static void smtc_configure_tlb(void)
  * phys_cpu_present_map and the logical/physical mappings.
  */
 
-int __init mipsmt_build_cpu_map(int start_cpu_slot)
+int __init smtc_build_cpu_map(int start_cpu_slot)
 {
        int i, ntcs;
 
@@ -325,7 +336,12 @@ static void smtc_tc_setup(int vpe, int tc, int cpu)
        write_tc_c0_tcstatus((read_tc_c0_tcstatus()
                        & ~(TCSTATUS_TKSU | TCSTATUS_DA | TCSTATUS_IXMT))
                        | TCSTATUS_A);
-       write_tc_c0_tccontext(0);
+       /*
+        * TCContext gets an offset from the base of the IPIQ array
+        * to be used in low-level code to detect the presence of
+        * an active IPI queue
+        */
+       write_tc_c0_tccontext((sizeof(struct smtc_ipi_q) * cpu) << 16);
        /* Bind tc to vpe */
        write_tc_c0_tcbind(vpe);
        /* In general, all TCs should have the same cpu_data indications */
@@ -336,10 +352,18 @@ static void smtc_tc_setup(int vpe, int tc, int cpu)
                cpu_data[cpu].options &= ~MIPS_CPU_FPU;
        cpu_data[cpu].vpe_id = vpe;
        cpu_data[cpu].tc_id = tc;
+       /* Multi-core SMTC hasn't been tested, but be prepared */
+       cpu_data[cpu].core = (read_vpe_c0_ebase() >> 1) & 0xff;
 }
 
+/*
+ * Tweak to get Count registes in as close a sync as possible.
+ * Value seems good for 34K-class cores.
+ */
+
+#define CP0_SKEW 8
 
-void mipsmt_prepare_cpus(void)
+void smtc_prepare_cpus(int cpus)
 {
        int i, vpe, tc, ntc, nvpe, tcpervpe[NR_CPUS], slop, cpu;
        unsigned long flags;
@@ -363,13 +387,13 @@ void mipsmt_prepare_cpus(void)
                IPIQ[i].head = IPIQ[i].tail = NULL;
                spin_lock_init(&IPIQ[i].lock);
                IPIQ[i].depth = 0;
-               atomic_set(&ipi_timer_latch[i], 0);
        }
 
        /* cpu_data index starts at zero */
        cpu = 0;
        cpu_data[cpu].vpe_id = 0;
        cpu_data[cpu].tc_id = 0;
+       cpu_data[cpu].core = (read_c0_ebase() >> 1) & 0xff;
        cpu++;
 
        /* Report on boot-time options */
@@ -484,7 +508,8 @@ void mipsmt_prepare_cpus(void)
                        write_vpe_c0_compare(0);
                        /* Propagate Config7 */
                        write_vpe_c0_config7(read_c0_config7());
-                       write_vpe_c0_count(read_c0_count());
+                       write_vpe_c0_count(read_c0_count() + CP0_SKEW);
+                       ehb();
                }
                /* enable multi-threading within VPE */
                write_vpe_c0_vpecontrol(read_vpe_c0_vpecontrol() | VPECONTROL_TE);
@@ -585,24 +610,22 @@ void __cpuinit smtc_boot_secondary(int cpu, struct task_struct *idle)
 
 void smtc_init_secondary(void)
 {
-       /*
-        * Start timer on secondary VPEs if necessary.
-        * plat_timer_setup has already have been invoked by init/main
-        * on "boot" TC.  Like per_cpu_trap_init() hack, this assumes that
-        * SMTC init code assigns TCs consdecutively and in ascending order
-        * to across available VPEs.
-        */
-       if (((read_c0_tcbind() & TCBIND_CURTC) != 0) &&
-           ((read_c0_tcbind() & TCBIND_CURVPE)
-           != cpu_data[smp_processor_id() - 1].vpe_id)){
-               write_c0_compare(read_c0_count() + mips_hpt_frequency/HZ);
-       }
-
        local_irq_enable();
 }
 
 void smtc_smp_finish(void)
 {
+       int cpu = smp_processor_id();
+
+       /*
+        * Lowest-numbered CPU per VPE starts a clock tick.
+        * Like per_cpu_trap_init() hack, this assumes that
+        * SMTC init code assigns TCs consdecutively and
+        * in ascending order across available VPEs.
+        */
+       if (cpu > 0 && (cpu_data[cpu].vpe_id != cpu_data[cpu - 1].vpe_id))
+               write_c0_compare(read_c0_count() + mips_hpt_frequency/HZ);
+
        printk("TC %d going on-line as CPU %d\n",
                cpu_data[smp_processor_id()].tc_id, smp_processor_id());
 }
@@ -755,6 +778,8 @@ void smtc_send_ipi(int cpu, int type, unsigned int action)
        struct smtc_ipi *pipi;
        unsigned long flags;
        int mtflags;
+       unsigned long tcrestart;
+       extern void r4k_wait_irqoff(void), __pastwait(void);
 
        if (cpu == smp_processor_id()) {
                printk("Cannot Send IPI to self!\n");
@@ -771,8 +796,6 @@ void smtc_send_ipi(int cpu, int type, unsigned int action)
        pipi->arg = (void *)action;
        pipi->dest = cpu;
        if (cpu_data[cpu].vpe_id != cpu_data[smp_processor_id()].vpe_id) {
-               if (type == SMTC_CLOCK_TICK)
-                       atomic_inc(&ipi_timer_latch[cpu]);
                /* If not on same VPE, enqueue and send cross-VPE interrupt */
                smtc_ipi_nq(&IPIQ[cpu], pipi);
                LOCK_CORE_PRA();
@@ -800,22 +823,29 @@ void smtc_send_ipi(int cpu, int type, unsigned int action)
 
                if ((tcstatus & TCSTATUS_IXMT) != 0) {
                        /*
-                        * Spin-waiting here can deadlock,
-                        * so we queue the message for the target TC.
+                        * If we're in the the irq-off version of the wait
+                        * loop, we need to force exit from the wait and
+                        * do a direct post of the IPI.
+                        */
+                       if (cpu_wait == r4k_wait_irqoff) {
+                               tcrestart = read_tc_c0_tcrestart();
+                               if (tcrestart >= (unsigned long)r4k_wait_irqoff
+                                   && tcrestart < (unsigned long)__pastwait) {
+                                       write_tc_c0_tcrestart(__pastwait);
+                                       tcstatus &= ~TCSTATUS_IXMT;
+                                       write_tc_c0_tcstatus(tcstatus);
+                                       goto postdirect;
+                               }
+                       }
+                       /*
+                        * Otherwise we queue the message for the target TC
+                        * to pick up when he does a local_irq_restore()
                         */
                        write_tc_c0_tchalt(0);
                        UNLOCK_CORE_PRA();
-                       /* Try to reduce redundant timer interrupt messages */
-                       if (type == SMTC_CLOCK_TICK) {
-                           if (atomic_postincrement(&ipi_timer_latch[cpu])!=0){
-                               smtc_ipi_nq(&freeIPIq, pipi);
-                               return;
-                           }
-                       }
                        smtc_ipi_nq(&IPIQ[cpu], pipi);
                } else {
-                       if (type == SMTC_CLOCK_TICK)
-                               atomic_inc(&ipi_timer_latch[cpu]);
+postdirect:
                        post_direct_ipi(cpu, pipi);
                        write_tc_c0_tchalt(0);
                        UNLOCK_CORE_PRA();
@@ -883,7 +913,7 @@ static void ipi_call_interrupt(void)
        smp_call_function_interrupt();
 }
 
-DECLARE_PER_CPU(struct clock_event_device, smtc_dummy_clockevent_device);
+DECLARE_PER_CPU(struct clock_event_device, mips_clockevent_device);
 
 void ipi_decode(struct smtc_ipi *pipi)
 {
@@ -891,20 +921,13 @@ void ipi_decode(struct smtc_ipi *pipi)
        struct clock_event_device *cd;
        void *arg_copy = pipi->arg;
        int type_copy = pipi->type;
-       int ticks;
-
        smtc_ipi_nq(&freeIPIq, pipi);
        switch (type_copy) {
        case SMTC_CLOCK_TICK:
                irq_enter();
                kstat_this_cpu.irqs[MIPS_CPU_IRQ_BASE + 1]++;
-               cd = &per_cpu(smtc_dummy_clockevent_device, cpu);
-               ticks = atomic_read(&ipi_timer_latch[cpu]);
-               atomic_sub(ticks, &ipi_timer_latch[cpu]);
-               while (ticks) {
-                       cd->event_handler(cd);
-                       ticks--;
-               }
+               cd = &per_cpu(mips_clockevent_device, cpu);
+               cd->event_handler(cd);
                irq_exit();
                break;
 
@@ -937,24 +960,48 @@ void ipi_decode(struct smtc_ipi *pipi)
        }
 }
 
+/*
+ * Similar to smtc_ipi_replay(), but invoked from context restore,
+ * so it reuses the current exception frame rather than set up a
+ * new one with self_ipi.
+ */
+
 void deferred_smtc_ipi(void)
 {
-       struct smtc_ipi *pipi;
-       unsigned long flags;
-/* DEBUG */
-       int q = smp_processor_id();
+       int cpu = smp_processor_id();
 
        /*
         * Test is not atomic, but much faster than a dequeue,
         * and the vast majority of invocations will have a null queue.
+        * If irq_disabled when this was called, then any IPIs queued
+        * after we test last will be taken on the next irq_enable/restore.
+        * If interrupts were enabled, then any IPIs added after the
+        * last test will be taken directly.
         */
-       if (IPIQ[q].head != NULL) {
-               while((pipi = smtc_ipi_dq(&IPIQ[q])) != NULL) {
-                       /* ipi_decode() should be called with interrupts off */
-                       local_irq_save(flags);
+
+       while (IPIQ[cpu].head != NULL) {
+               struct smtc_ipi_q *q = &IPIQ[cpu];
+               struct smtc_ipi *pipi;
+               unsigned long flags;
+
+               /*
+                * It may be possible we'll come in with interrupts
+                * already enabled.
+                */
+               local_irq_save(flags);
+
+               spin_lock(&q->lock);
+               pipi = __smtc_ipi_dq(q);
+               spin_unlock(&q->lock);
+               if (pipi != NULL)
                        ipi_decode(pipi);
-                       local_irq_restore(flags);
-               }
+               /*
+                * The use of the __raw_local restore isn't
+                * as obviously necessary here as in smtc_ipi_replay(),
+                * but it's more efficient, given that we're already
+                * running down the IPI queue.
+                */
+               __raw_local_irq_restore(flags);
        }
 }
 
@@ -1066,55 +1113,53 @@ static void setup_cross_vpe_interrupts(unsigned int nvpe)
 
 /*
  * SMTC-specific hacks invoked from elsewhere in the kernel.
- *
- * smtc_ipi_replay is called from raw_local_irq_restore which is only ever
- * called with interrupts disabled.  We do rely on interrupts being disabled
- * here because using spin_lock_irqsave()/spin_unlock_irqrestore() would
- * result in a recursive call to raw_local_irq_restore().
  */
 
-static void __smtc_ipi_replay(void)
+ /*
+  * smtc_ipi_replay is called from raw_local_irq_restore
+  */
+
+void smtc_ipi_replay(void)
 {
        unsigned int cpu = smp_processor_id();
 
        /*
         * To the extent that we've ever turned interrupts off,
         * we may have accumulated deferred IPIs.  This is subtle.
-        * If we use the smtc_ipi_qdepth() macro, we'll get an
-        * exact number - but we'll also disable interrupts
-        * and create a window of failure where a new IPI gets
-        * queued after we test the depth but before we re-enable
-        * interrupts. So long as IXMT never gets set, however,
         * we should be OK:  If we pick up something and dispatch
         * it here, that's great. If we see nothing, but concurrent
         * with this operation, another TC sends us an IPI, IXMT
         * is clear, and we'll handle it as a real pseudo-interrupt
-        * and not a pseudo-pseudo interrupt.
+        * and not a pseudo-pseudo interrupt.  The important thing
+        * is to do the last check for queued message *after* the
+        * re-enabling of interrupts.
         */
-       if (IPIQ[cpu].depth > 0) {
-               while (1) {
-                       struct smtc_ipi_q *q = &IPIQ[cpu];
-                       struct smtc_ipi *pipi;
-                       extern void self_ipi(struct smtc_ipi *);
-
-                       spin_lock(&q->lock);
-                       pipi = __smtc_ipi_dq(q);
-                       spin_unlock(&q->lock);
-                       if (!pipi)
-                               break;
+       while (IPIQ[cpu].head != NULL) {
+               struct smtc_ipi_q *q = &IPIQ[cpu];
+               struct smtc_ipi *pipi;
+               unsigned long flags;
 
+               /*
+                * It's just possible we'll come in with interrupts
+                * already enabled.
+                */
+               local_irq_save(flags);
+
+               spin_lock(&q->lock);
+               pipi = __smtc_ipi_dq(q);
+               spin_unlock(&q->lock);
+               /*
+                ** But use a raw restore here to avoid recursion.
+                */
+               __raw_local_irq_restore(flags);
+
+               if (pipi) {
                        self_ipi(pipi);
                        smtc_cpu_stats[cpu].selfipis++;
                }
        }
 }
 
-void smtc_ipi_replay(void)
-{
-       raw_local_irq_disable();
-       __smtc_ipi_replay();
-}
-
 EXPORT_SYMBOL(smtc_ipi_replay);
 
 void smtc_idle_loop_hook(void)
@@ -1193,40 +1238,13 @@ void smtc_idle_loop_hook(void)
                }
        }
 
-       /*
-        * Now that we limit outstanding timer IPIs, check for hung TC
-        */
-       for (tc = 0; tc < NR_CPUS; tc++) {
-               /* Don't check ourself - we'll dequeue IPIs just below */
-               if ((tc != smp_processor_id()) &&
-                   atomic_read(&ipi_timer_latch[tc]) > timerq_limit) {
-                   if (clock_hang_reported[tc] == 0) {
-                       pdb_msg += sprintf(pdb_msg,
-                               "TC %d looks hung with timer latch at %d\n",
-                               tc, atomic_read(&ipi_timer_latch[tc]));
-                       clock_hang_reported[tc]++;
-                       }
-               }
-       }
        emt(mtflags);
        local_irq_restore(flags);
        if (pdb_msg != &id_ho_db_msg[0])
                printk("CPU%d: %s", smp_processor_id(), id_ho_db_msg);
 #endif /* CONFIG_SMTC_IDLE_HOOK_DEBUG */
 
-       /*
-        * Replay any accumulated deferred IPIs. If "Instant Replay"
-        * is in use, there should never be any.
-        */
-#ifndef CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY
-       {
-               unsigned long flags;
-
-               local_irq_save(flags);
-               __smtc_ipi_replay();
-               local_irq_restore(flags);
-       }
-#endif /* CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY */
+       smtc_ipi_replay();
 }
 
 void smtc_soft_dump(void)
@@ -1242,10 +1260,6 @@ void smtc_soft_dump(void)
                printk("%d: %ld\n", i, smtc_cpu_stats[i].selfipis);
        }
        smtc_ipi_qdump();
-       printk("Timer IPI Backlogs:\n");
-       for (i=0; i < NR_CPUS; i++) {
-               printk("%d: %d\n", i, atomic_read(&ipi_timer_latch[i]));
-       }
        printk("%d Recoveries of \"stolen\" FPU\n",
               atomic_read(&smtc_fpu_recoveries));
 }
index 5ea705e49454e3ce49b27fb6b2f80a9f796f6950..f84a46a8ae6e5261d0c8669b2726ed353cda7dc3 100644 (file)
@@ -84,12 +84,17 @@ static void msmtc_cpus_done(void)
 
 static void __init msmtc_smp_setup(void)
 {
-       mipsmt_build_cpu_map(0);
+       /*
+        * we won't get the definitive value until
+        * we've run smtc_prepare_cpus later, but
+        * we would appear to need an upper bound now.
+        */
+       smp_num_siblings = smtc_build_cpu_map(0);
 }
 
 static void __init msmtc_prepare_cpus(unsigned int max_cpus)
 {
-       mipsmt_prepare_cpus();
+       smtc_prepare_cpus(max_cpus);
 }
 
 struct plat_smp_ops msmtc_smp_ops = {
diff --git a/include/asm-mips/cevt-r4k.h b/include/asm-mips/cevt-r4k.h
new file mode 100644 (file)
index 0000000..fa4328f
--- /dev/null
@@ -0,0 +1,46 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2008 Kevin D. Kissell
+ */
+
+/*
+ * Definitions used for common event timer implementation
+ * for MIPS 4K-type processors and their MIPS MT variants.
+ * Avoids unsightly extern declarations in C files.
+ */
+#ifndef __ASM_CEVT_R4K_H
+#define __ASM_CEVT_R4K_H
+
+DECLARE_PER_CPU(struct clock_event_device, mips_clockevent_device);
+
+void mips_event_handler(struct clock_event_device *dev);
+int c0_compare_int_usable(void);
+void mips_set_clock_mode(enum clock_event_mode, struct clock_event_device *);
+irqreturn_t c0_compare_interrupt(int, void *);
+
+extern struct irqaction c0_compare_irqaction;
+extern int cp0_timer_irq_installed;
+
+/*
+ * Possibly handle a performance counter interrupt.
+ * Return true if the timer interrupt should not be checked
+ */
+
+static inline int handle_perf_irq(int r2)
+{
+       /*
+        * The performance counter overflow interrupt may be shared with the
+        * timer interrupt (cp0_perfcount_irq < 0). If it is and a
+        * performance counter has overflowed (perf_irq() == IRQ_HANDLED)
+        * and we can't reliably determine if a counter interrupt has also
+        * happened (!r2) then don't check for a timer interrupt.
+        */
+       return (cp0_perfcount_irq < 0) &&
+               perf_irq() == IRQ_HANDLED &&
+               !r2;
+}
+
+#endif /* __ASM_CEVT_R4K_H */
index 881e8866501d5f2af6b09355227edc620c3ba5b4..701ec0ba8fa9cdd596369b9de3825b284bae1ec9 100644 (file)
@@ -38,8 +38,17 @@ __asm__(
        "       .set    pop                                             \n"
        "       .endm");
 
+extern void smtc_ipi_replay(void);
+
 static inline void raw_local_irq_enable(void)
 {
+#ifdef CONFIG_MIPS_MT_SMTC
+       /*
+        * SMTC kernel needs to do a software replay of queued
+        * IPIs, at the cost of call overhead on each local_irq_enable()
+        */
+       smtc_ipi_replay();
+#endif
        __asm__ __volatile__(
                "raw_local_irq_enable"
                : /* no outputs */
@@ -47,6 +56,7 @@ static inline void raw_local_irq_enable(void)
                : "memory");
 }
 
+
 /*
  * For cli() we have to insert nops to make sure that the new value
  * has actually arrived in the status register before the end of this
@@ -185,15 +195,14 @@ __asm__(
        "       .set    pop                                             \n"
        "       .endm                                                   \n");
 
-extern void smtc_ipi_replay(void);
 
 static inline void raw_local_irq_restore(unsigned long flags)
 {
        unsigned long __tmp1;
 
-#ifdef CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY
+#ifdef CONFIG_MIPS_MT_SMTC
        /*
-        * CONFIG_MIPS_MT_SMTC_INSTANT_REPLAY does prompt replay of deferred
+        * SMTC kernel needs to do a software replay of queued
         * IPIs, at the cost of branch and call overhead on each
         * local_irq_restore()
         */
@@ -208,6 +217,17 @@ static inline void raw_local_irq_restore(unsigned long flags)
                : "memory");
 }
 
+static inline void __raw_local_irq_restore(unsigned long flags)
+{
+       unsigned long __tmp1;
+
+       __asm__ __volatile__(
+               "raw_local_irq_restore\t%0"
+               : "=r" (__tmp1)
+               : "0" (flags)
+               : "memory");
+}
+
 static inline int raw_irqs_disabled_flags(unsigned long flags)
 {
 #ifdef CONFIG_MIPS_MT_SMTC
index 3639b28f80db84acbdea9672eb705fa8976aca7d..ea60bf08dcb0b9792e4433a7dd92117a3b1afd98 100644 (file)
@@ -6,6 +6,7 @@
  */
 
 #include <asm/mips_mt.h>
+#include <asm/smtc_ipi.h>
 
 /*
  * System-wide SMTC status information
@@ -38,14 +39,15 @@ struct mm_struct;
 struct task_struct;
 
 void smtc_get_new_mmu_context(struct mm_struct *mm, unsigned long cpu);
-
+void self_ipi(struct smtc_ipi *);
 void smtc_flush_tlb_asid(unsigned long asid);
-extern int mipsmt_build_cpu_map(int startslot);
-extern void mipsmt_prepare_cpus(void);
+extern int smtc_build_cpu_map(int startslot);
+extern void smtc_prepare_cpus(int cpus);
 extern void smtc_smp_finish(void);
 extern void smtc_boot_secondary(int cpu, struct task_struct *t);
 extern void smtc_cpus_done(void);
 
+
 /*
  * Sharing the TLB between multiple VPEs means that the
  * "random" index selection function is not allowed to