Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.10-20170630' into staging

ppc patch queue 2017-06-30

  * More DRC cleanups, these now actually fix a few bugs
  * Properly implements the openpic timers (they now count and
    generate interrupts)
  * Fixes for XICS migration
  * Fixes for migration of POWER9 RPT guests
  * The last of the compatibility mode rework

# gpg: Signature made Fri 30 Jun 2017 10:52:25 BST
# gpg:                using RSA key 0x6C38CACA20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>"
# gpg:                 aka "David Gibson (Red Hat) <dgibson@redhat.com>"
# gpg:                 aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>"
# gpg:                 aka "David Gibson (kernel.org) <dwg@kernel.org>"
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E  87DC 6C38 CACA 20D9 B392

* remotes/dgibson/tags/ppc-for-2.10-20170630: (21 commits)
  spapr: Clean up DRC set_isolation_state() path
  spapr: Clean up DRC set_allocation_state path
  spapr: Make DRC reset force DRC into known state
  spapr: Split DRC release from DRC detach
  spapr: Eliminate DRC 'signalled' state variable
  spapr: Start hotplugged PCI devices in ISOLATED state
  target-ppc: Enable open-pic timers to count and generate interrupts
  hw/ppc/spapr.c: consecutive 'spapr->patb_entry = 0' statements
  spapr: prevent QEMU crash when CPU realization fails
  target/ppc: Proper cleanup when ppc_cpu_realizefn fails
  spapr: fix migration of ICPState objects from/to older QEMU
  xics: directly register ICPState objects to vmstate
  target/ppc: Fix return value in tcg radix mmu fault handler
  target/ppc/excp_helper: Take BQL before calling cpu_interrupt()
  spapr: Fix migration of Radix guests
  spapr: Add a "no HPT" encoding to HTAB migration stream
  ppc: Rework CPU compatibility testing across migration
  pseries: Reset CPU compatibility mode
  pseries: Move CPU compatibility property to machine
  qapi: add explicit null to string input and output visitors
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

2 files changed, 115 insertions, 7 deletions

diff --git a/hw/intc/openpic.c b/hw/intc/openpic.c
index 5595bb2..9dd285b 100644
--- a/hw/intc/openpic.c
+++ b/hw/intc/openpic.c
@@ -45,6 +45,7 @@
 #include "qemu/bitops.h"
 #include "qapi/qmp/qerror.h"
 #include "qemu/log.h"
+#include "qemu/timer.h"
 
 //#define DEBUG_OPENPIC
 
@@ -54,8 +55,10 @@ static const int debug_openpic = 1;
 static const int debug_openpic = 0;
 #endif
 
+static int get_current_cpu(void);
 #define DPRINTF(fmt, ...) do { \
         if (debug_openpic) { \
+            printf("Core%d: ", get_current_cpu()); \
             printf(fmt , ## __VA_ARGS__); \
         } \
     } while (0)
@@ -246,9 +249,31 @@ typedef struct IRQSource {
 #define IDR_EP      0x80000000  /* external pin */
 #define IDR_CI      0x40000000  /* critical interrupt */
 
+/* Convert between openpic clock ticks and nanosecs.  In the hardware the clock
+   frequency is driven by board inputs to the PIC which the PIC would then
+   divide by 4 or 8.  For now hard code to 25MZ.
+*/
+#define OPENPIC_TIMER_FREQ_MHZ 25
+#define OPENPIC_TIMER_NS_PER_TICK (1000 / OPENPIC_TIMER_FREQ_MHZ)
+static inline uint64_t ns_to_ticks(uint64_t ns)
+{
+    return ns    / OPENPIC_TIMER_NS_PER_TICK;
+}
+static inline uint64_t ticks_to_ns(uint64_t ticks)
+{
+    return ticks * OPENPIC_TIMER_NS_PER_TICK;
+}
+
 typedef struct OpenPICTimer {
     uint32_t tccr;  /* Global timer current count register */
     uint32_t tbcr;  /* Global timer base count register */
+    int                   n_IRQ;
+    bool                  qemu_timer_active; /* Is the qemu_timer is running? */
+    struct QEMUTimer     *qemu_timer;
+    struct OpenPICState  *opp;          /* Device timer is part of. */
+    /* The QEMU_CLOCK_VIRTUAL time (in ns) corresponding to the last
+       current_count written or read, only defined if qemu_timer_active. */
+    uint64_t              origin_time;
 } OpenPICTimer;
 
 typedef struct OpenPICMSI {
@@ -795,6 +820,65 @@ static uint64_t openpic_gbl_read(void *opaque, hwaddr addr, unsigned len)
     return retval;
 }
 
+static void openpic_tmr_set_tmr(OpenPICTimer *tmr, uint32_t val, bool enabled);
+
+static void qemu_timer_cb(void *opaque)
+{
+    OpenPICTimer *tmr = opaque;
+    OpenPICState *opp = tmr->opp;
+    uint32_t    n_IRQ = tmr->n_IRQ;
+    uint32_t val =   tmr->tbcr & ~TBCR_CI;
+    uint32_t tog = ((tmr->tccr & TCCR_TOG) ^ TCCR_TOG);  /* invert toggle. */
+
+    DPRINTF("%s n_IRQ=%d\n", __func__, n_IRQ);
+    /* Reload current count from base count and setup timer. */
+    tmr->tccr = val | tog;
+    openpic_tmr_set_tmr(tmr, val, /*enabled=*/true);
+    /* Raise the interrupt. */
+    opp->src[n_IRQ].destmask = read_IRQreg_idr(opp, n_IRQ);
+    openpic_set_irq(opp, n_IRQ, 1);
+    openpic_set_irq(opp, n_IRQ, 0);
+}
+
+/* If enabled is true, arranges for an interrupt to be raised val clocks into
+   the future, if enabled is false cancels the timer. */
+static void openpic_tmr_set_tmr(OpenPICTimer *tmr, uint32_t val, bool enabled)
+{
+    uint64_t ns = ticks_to_ns(val & ~TCCR_TOG);
+    /* A count of zero causes a timer to be set to expire immediately.  This
+       effectively stops the simulation since the timer is constantly expiring
+       which prevents guest code execution, so we don't honor that
+       configuration.  On real hardware, this situation would generate an
+       interrupt on every clock cycle if the interrupt was unmasked. */
+    if ((ns == 0) || !enabled) {
+        tmr->qemu_timer_active = false;
+        tmr->tccr = tmr->tccr & TCCR_TOG;
+        timer_del(tmr->qemu_timer); /* set timer to never expire. */
+    } else {
+        tmr->qemu_timer_active = true;
+        uint64_t now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+        tmr->origin_time = now;
+        timer_mod(tmr->qemu_timer, now + ns);     /* set timer expiration. */
+    }
+}
+
+/* Returns the currrent tccr value, i.e., timer value (in clocks) with
+   appropriate TOG. */
+static uint64_t openpic_tmr_get_timer(OpenPICTimer *tmr)
+{
+    uint64_t retval;
+    if (!tmr->qemu_timer_active) {
+        retval = tmr->tccr;
+    } else {
+        uint64_t now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
+        uint64_t used = now - tmr->origin_time;  /* nsecs */
+        uint32_t used_ticks = (uint32_t)ns_to_ticks(used);
+        uint32_t count = (tmr->tccr & ~TCCR_TOG) - used_ticks;
+        retval = (uint32_t)((tmr->tccr & TCCR_TOG) | (count & ~TCCR_TOG));
+    }
+    return retval;
+}
+
 static void openpic_tmr_write(void *opaque, hwaddr addr, uint64_t val,
                               unsigned len)
 {
@@ -819,10 +903,15 @@ static void openpic_tmr_write(void *opaque, hwaddr addr, uint64_t val,
     case 0x00: /* TCCR */
         break;
     case 0x10: /* TBCR */
-        if ((opp->timers[idx].tccr & TCCR_TOG) != 0 &&
-            (val & TBCR_CI) == 0 &&
-            (opp->timers[idx].tbcr & TBCR_CI) != 0) {
-            opp->timers[idx].tccr &= ~TCCR_TOG;
+        /* Did the enable status change? */
+        if ((opp->timers[idx].tbcr & TBCR_CI) != (val & TBCR_CI)) {
+            /* Did "Count Inhibit" transition from 1 to 0? */
+            if ((val & TBCR_CI) == 0) {
+                opp->timers[idx].tccr = val & ~TCCR_TOG;
+            }
+            openpic_tmr_set_tmr(&opp->timers[idx],
+                                (val & ~TBCR_CI),
+                                /*enabled=*/((val & TBCR_CI) == 0));
         }
         opp->timers[idx].tbcr = val;
         break;
@@ -854,7 +943,7 @@ static uint64_t openpic_tmr_read(void *opaque, hwaddr addr, unsigned len)
     idx = (addr >> 6) & 0x3;
     switch (addr & 0x30) {
     case 0x00: /* TCCR */
-        retval = opp->timers[idx].tccr;
+        retval = openpic_tmr_get_timer(&opp->timers[idx]);
         break;
     case 0x10: /* TBCR */
         retval = opp->timers[idx].tbcr;
@@ -1136,7 +1225,10 @@ static uint32_t openpic_iack(OpenPICState *opp, IRQDest *dst, int cpu)
         IRQ_resetbit(&dst->raised, irq);
     }
 
-    if ((irq >= opp->irq_ipi0) &&  (irq < (opp->irq_ipi0 + OPENPIC_MAX_IPI))) {
+    /* Timers and IPIs support multicast. */
+    if (((irq >= opp->irq_ipi0) && (irq < (opp->irq_ipi0 + OPENPIC_MAX_IPI))) ||
+        ((irq >= opp->irq_tim0) && (irq < (opp->irq_tim0 + OPENPIC_MAX_TMR)))) {
+        DPRINTF("irq is IPI or TMR\n");
         src->destmask &= ~(1 << cpu);
         if (src->destmask && !src->level) {
             /* trigger on CPUs that didn't know about it yet */
@@ -1341,6 +1433,10 @@ static void openpic_reset(DeviceState *d)
     for (i = 0; i < OPENPIC_MAX_TMR; i++) {
         opp->timers[i].tccr = 0;
         opp->timers[i].tbcr = TBCR_CI;
+        if (opp->timers[i].qemu_timer_active) {
+            timer_del(opp->timers[i].qemu_timer);  /* Inhibit timer */
+            opp->timers[i].qemu_timer_active = false;
+        }
     }
     /* Go out of RESET state */
     opp->gcr = 0;
@@ -1391,6 +1487,15 @@ static void fsl_common_init(OpenPICState *opp)
         opp->src[i].type = IRQ_TYPE_FSLSPECIAL;
         opp->src[i].level = false;
     }
+
+    for (i = 0; i < OPENPIC_MAX_TMR; i++) {
+        opp->timers[i].n_IRQ = opp->irq_tim0 + i;
+        opp->timers[i].qemu_timer_active = false;
+        opp->timers[i].qemu_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL,
+                                                 &qemu_timer_cb,
+                                                 &opp->timers[i]);
+        opp->timers[i].opp = opp;
+    }
 }
 
 static void map_list(OpenPICState *opp, const MemReg *list, int *count)
diff --git a/hw/intc/xics.c b/hw/intc/xics.c
index d4194d6..a84ba51 100644
--- a/hw/intc/xics.c
+++ b/hw/intc/xics.c
@@ -344,10 +344,14 @@ static void icp_realize(DeviceState *dev, Error **errp)
     }
 
     qemu_register_reset(icp_reset, dev);
+    vmstate_register(NULL, icp->cs->cpu_index, &vmstate_icp_server, icp);
 }
 
 static void icp_unrealize(DeviceState *dev, Error **errp)
 {
+    ICPState *icp = ICP(dev);
+
+    vmstate_unregister(NULL, &vmstate_icp_server, icp);
     qemu_unregister_reset(icp_reset, dev);
 }
 
@@ -355,7 +359,6 @@ static void icp_class_init(ObjectClass *klass, void *data)
 {
     DeviceClass *dc = DEVICE_CLASS(klass);
 
-    dc->vmsd = &vmstate_icp_server;
     dc->realize = icp_realize;
     dc->unrealize = icp_unrealize;
 }