Author Topic: Debuging why KVM can't boot OS 9.  (Read 4568 times)

Offline darthnVader

  • Platinum Member
  • *****
  • Posts: 679
  • New Member
Debuging why KVM can't boot OS 9.
« on: October 24, 2018, 07:27:31 AM »
KVM is more than a bit buggy on the G4 CPU, tho it will boot, and sometimes install OS X, I normally only get one boot from the installed Tiger system, and then the boot image gets corrupt and will no longer boot.

That's the state of OS X, as far as Qemu/KVM. I haven't tried MOL with KVM on this system.

However, here, we like all things OS 9, so let's see if we can figure out why we can't boot OS 9 with KVM enabled;

Code: [Select]
qemu-system-ppc64  --enable-kvm -M mac99 -m 512 -cpu host -cdrom /dev/sr0 -boot c  -hda ~/Documents/os9.img -bios '/home/jam/OB/openbios-qemu2.elf'  -prom-env boot-args=-v -prom-env "auto-boot?=false" -prom-env "aapl,debug=3013FFF" -serial stdio
Running QEMU with SDL 1.2 is deprecated, and will be removed
in a future release. Please switch to SDL 2.0 instead

>> =============================================================
>> OpenBIOS 1.1 [Oct 16 2018 18:18]
>> Configuration device id QEMU version 1 machine id 1
>> CPUs: 1
>> Memory: 512M
>> UUID: 00000000-0000-0000-0000-000000000000
>> CPU type PowerPC,7455
milliseconds isn't unique.
>> switching to new context:
Nanodebugger activated.
Init ready queue 00000000 00000000 00008652
Init ready queue 00000001 00000000 00043290
Init ready queue 00000002 00000000 00219480
Init ready queue 00000003 00000000 010ca400
System context at 0x1f7ff100  Vector save area at 0x1f7fb880  SDR1 0x1f80003f
Adding blue task 00050001 to the ready queue
Starting timeslicing
Adding idle task 00070001 to the ready queue
NKCreateAddressSpaceSub - group at 0x1f7fc1c0 00010001
Priming the system free list with 81 pages.
VMMaxVirtualPages: 0005fffe VMLogicalPages: 00000000
Interrupt handler kind: 06
Converting PMDTs to areas
 CreateArea [ 00000000 1f7c2fff ] ID 000a0001 placed ... created
 CreateArea [ 5fffe000 5fffefff ] ID 000b0001 placed ... created
 CreateArea [ 64000000 6417ffff ] ID 000c0001 placed ... created
 CreateArea [ 68fef000 68feffff ] ID 000d0001 placed ... created
 CreateArea [ 68ff5000 68ffefff ] ID 000e0001 placed ... created
 CreateArea [ 68fff000 68ffffff ] ID 000f0001 placed ... created
 CreateArea [ 80000000 8005ffff ] ID 00100001 placed ... created
 CreateArea [ 80060000 80060fff ] ID 00110001 placed ... created
 CreateArea [ 80061000 8007ffff ] ID 00120001 placed ... created
 CreateArea [ 80080000 80080fff ] ID 00130001 placed ... created
 CreateArea [ 81000000 81ffffff ] ID 00140001 placed ... created
 CreateArea [ 82000000 82000fff ] ID 00150001 placed ... created
 CreateArea [ f0000000 ffffffff ] ID 00160001 placed ... created
 CreateArea [ 68f16000 68f16fff ] ID 00170001 placed ... created
 CreateArea [ deadb000 deadbfff ] ID 00180001 placed ... created
Reset system - Into the 68K fire: 0000d032 6806e8c0

System just hangs there...........

Offline ELN

  • Gold Member
  • *****
  • Posts: 295
  • new to the forums
Re: Debuging why KVM can't boot OS 9.
« Reply #1 on: October 24, 2018, 08:22:38 AM »
Once the NanoKernel hands control over to the Emulator (the “68k fire”), a 68k RESET in StartInit is executed within a few instructions. This causes the NK to repeat its initialisation log messages. They are not being repeated, so I believe that the problem occurs just before the NanoKernel makes it to userspace.

It is very, very likely that the NK is crashing while trying to access the emulated interrupt controller. Max is the expert on this. Can anyone guess what KVM emulates, and whether the NK should already support it?

Offline darthnVader

  • Platinum Member
  • *****
  • Posts: 679
  • New Member
Re: Debuging why KVM can't boot OS 9.
« Reply #2 on: October 24, 2018, 12:13:55 PM »
Also, I built MOL with KVM support, however it just hangs when loading mach_kernel or the OS 9 ELF( Trampoline ).

Something I've been wondering is what is kvm-pr, I know it's  another version of the kvm module but I don't get any option to build it.

Is it only available  on PPC64?

 

Offline darthnVader

  • Platinum Member
  • *****
  • Posts: 679
  • New Member
Re: Debuging why KVM can't boot OS 9.
« Reply #3 on: October 24, 2018, 02:55:02 PM »
Once the NanoKernel hands control over to the Emulator (the “68k fire”), a 68k RESET in StartInit is executed within a few instructions. This causes the NK to repeat its initialisation log messages. They are not being repeated, so I believe that the problem occurs just before the NanoKernel makes it to userspace.

It is very, very likely that the NK is crashing while trying to access the emulated interrupt controller. Max is the expert on this. Can anyone guess what KVM emulates, and whether the NK should already support it?

Not sure what's going wrong, the .properties for the interrupt controller are the same on the emu G4/G4 with KVM/ Real QS G4.

Code: [Select]
name                      "interrupt-controller"
    device_type               "open-pic"
    compatible                "chrp,open-pic"
    built-in                  <empty>
    reg                       00040000   00040000
    #interrupt-cells          2
    #address-cells            0
    interrupt-controller      <empty>
    clock-frequency           3f940a

Offline darthnVader

  • Platinum Member
  • *****
  • Posts: 679
  • New Member
Re: Debuging why KVM can't boot OS 9.
« Reply #4 on: October 24, 2018, 04:03:17 PM »
Not sure if this has anything to do with it:

qemu-3.0/qemu/hw/ppc/ppc_booke.c

Code: [Select]
/*
 * QEMU PowerPC Booke hardware System Emulator
 *
 * Copyright (c) 2011 AdaCore
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 */
#include "qemu/osdep.h"
#include "qemu-common.h"
#include "cpu.h"
#include "hw/hw.h"
#include "hw/ppc/ppc.h"
#include "qemu/timer.h"
#include "sysemu/sysemu.h"
#include "hw/timer/m48t59.h"
#include "qemu/log.h"
#include "hw/loader.h"
#include "kvm_ppc.h"


/* Timer Control Register */

#define TCR_WP_SHIFT  30        /* Watchdog Timer Period */
#define TCR_WP_MASK   (0x3U << TCR_WP_SHIFT)
#define TCR_WRC_SHIFT 28        /* Watchdog Timer Reset Control */
#define TCR_WRC_MASK  (0x3U << TCR_WRC_SHIFT)
#define TCR_WIE       (1U << 27) /* Watchdog Timer Interrupt Enable */
#define TCR_DIE       (1U << 26) /* Decrementer Interrupt Enable */
#define TCR_FP_SHIFT  24        /* Fixed-Interval Timer Period */
#define TCR_FP_MASK   (0x3U << TCR_FP_SHIFT)
#define TCR_FIE       (1U << 23) /* Fixed-Interval Timer Interrupt Enable */
#define TCR_ARE       (1U << 22) /* Auto-Reload Enable */

/* Timer Control Register (e500 specific fields) */

#define TCR_E500_FPEXT_SHIFT 13 /* Fixed-Interval Timer Period Extension */
#define TCR_E500_FPEXT_MASK  (0xf << TCR_E500_FPEXT_SHIFT)
#define TCR_E500_WPEXT_SHIFT 17 /* Watchdog Timer Period Extension */
#define TCR_E500_WPEXT_MASK  (0xf << TCR_E500_WPEXT_SHIFT)

/* Timer Status Register  */

#define TSR_FIS       (1U << 26) /* Fixed-Interval Timer Interrupt Status */
#define TSR_DIS       (1U << 27) /* Decrementer Interrupt Status */
#define TSR_WRS_SHIFT 28        /* Watchdog Timer Reset Status */
#define TSR_WRS_MASK  (0x3U << TSR_WRS_SHIFT)
#define TSR_WIS       (1U << 30) /* Watchdog Timer Interrupt Status */
#define TSR_ENW       (1U << 31) /* Enable Next Watchdog Timer */

typedef struct booke_timer_t booke_timer_t;
struct booke_timer_t {

    uint64_t fit_next;
    QEMUTimer *fit_timer;

    uint64_t wdt_next;
    QEMUTimer *wdt_timer;

    uint32_t flags;
};

static void booke_update_irq(PowerPCCPU *cpu)
{
    CPUPPCState *env = &cpu->env;

    ppc_set_irq(cpu, PPC_INTERRUPT_DECR,
                (env->spr[SPR_BOOKE_TSR] & TSR_DIS
                 && env->spr[SPR_BOOKE_TCR] & TCR_DIE));

    ppc_set_irq(cpu, PPC_INTERRUPT_WDT,
                (env->spr[SPR_BOOKE_TSR] & TSR_WIS
                 && env->spr[SPR_BOOKE_TCR] & TCR_WIE));

    ppc_set_irq(cpu, PPC_INTERRUPT_FIT,
                (env->spr[SPR_BOOKE_TSR] & TSR_FIS
                 && env->spr[SPR_BOOKE_TCR] & TCR_FIE));
}

/* Return the location of the bit of time base at which the FIT will raise an
   interrupt */
static uint8_t booke_get_fit_target(CPUPPCState *env, ppc_tb_t *tb_env)
{
    uint8_t fp = (env->spr[SPR_BOOKE_TCR] & TCR_FP_MASK) >> TCR_FP_SHIFT;

    if (tb_env->flags & PPC_TIMER_E500) {
        /* e500 Fixed-interval timer period extension */
        uint32_t fpext = (env->spr[SPR_BOOKE_TCR] & TCR_E500_FPEXT_MASK)
            >> TCR_E500_FPEXT_SHIFT;
        fp = 63 - (fp | fpext << 2);
    } else {
        fp = env->fit_period[fp];
    }

    return fp;
}

/* Return the location of the bit of time base at which the WDT will raise an
   interrupt */
static uint8_t booke_get_wdt_target(CPUPPCState *env, ppc_tb_t *tb_env)
{
    uint8_t wp = (env->spr[SPR_BOOKE_TCR] & TCR_WP_MASK) >> TCR_WP_SHIFT;

    if (tb_env->flags & PPC_TIMER_E500) {
        /* e500 Watchdog timer period extension */
        uint32_t wpext = (env->spr[SPR_BOOKE_TCR] & TCR_E500_WPEXT_MASK)
            >> TCR_E500_WPEXT_SHIFT;
        wp = 63 - (wp | wpext << 2);
    } else {
        wp = env->wdt_period[wp];
    }

    return wp;
}

static void booke_update_fixed_timer(CPUPPCState         *env,
                                     uint8_t           target_bit,
                                     uint64_t          *next,
                                     QEMUTimer         *timer,
                                     int               tsr_bit)
{
    ppc_tb_t *tb_env = env->tb_env;
    uint64_t delta_tick, ticks = 0;
    uint64_t tb;
    uint64_t period;
    uint64_t now;

    if (!(env->spr[SPR_BOOKE_TSR] & tsr_bit)) {
        /*
         * Don't arm the timer again when the guest has the current
         * interrupt still pending. Wait for it to ack it.
         */
        return;
    }

    now = qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL);
    tb  = cpu_ppc_get_tb(tb_env, now, tb_env->tb_offset);
    period = 1ULL << target_bit;
    delta_tick = period - (tb & (period - 1));

    /* the timer triggers only when the selected bit toggles from 0 to 1 */
    if (tb & period) {
        ticks = period;
    }

    if (ticks + delta_tick < ticks) {
        /* Overflow, so assume the biggest number we can express. */
        ticks = UINT64_MAX;
    } else {
        ticks += delta_tick;
    }

    *next = now + muldiv64(ticks, NANOSECONDS_PER_SECOND, tb_env->tb_freq);
    if ((*next < now) || (*next > INT64_MAX)) {
        /* Overflow, so assume the biggest number the qemu timer supports. */
        *next = INT64_MAX;
    }

    /* XXX: If expire time is now. We can't run the callback because we don't
     * have access to it. So we just set the timer one nanosecond later.
     */

    if (*next == now) {
        (*next)++;
    } else {
        /*
         * There's no point to fake any granularity that's more fine grained
         * than milliseconds. Anything beyond that just overloads the system.
         */
        *next = MAX(*next, now + SCALE_MS);
    }

    /* Fire the next timer */
    timer_mod(timer, *next);
}

static void booke_decr_cb(void *opaque)
{
    PowerPCCPU *cpu = opaque;
    CPUPPCState *env = &cpu->env;

    env->spr[SPR_BOOKE_TSR] |= TSR_DIS;
    booke_update_irq(cpu);

    if (env->spr[SPR_BOOKE_TCR] & TCR_ARE) {
        /* Do not reload 0, it is already there. It would just trigger
         * the timer again and lead to infinite loop */
        if (env->spr[SPR_BOOKE_DECAR] != 0) {
            /* Auto Reload */
            cpu_ppc_store_decr(env, env->spr[SPR_BOOKE_DECAR]);
        }
    }
}

static void booke_fit_cb(void *opaque)
{
    PowerPCCPU *cpu = opaque;
    CPUPPCState *env = &cpu->env;
    ppc_tb_t *tb_env;
    booke_timer_t *booke_timer;

    tb_env = env->tb_env;
    booke_timer = tb_env->opaque;
    env->spr[SPR_BOOKE_TSR] |= TSR_FIS;

    booke_update_irq(cpu);

    booke_update_fixed_timer(env,
                             booke_get_fit_target(env, tb_env),
                             &booke_timer->fit_next,
                             booke_timer->fit_timer,
                             TSR_FIS);
}

static void booke_wdt_cb(void *opaque)
{
    PowerPCCPU *cpu = opaque;
    CPUPPCState *env = &cpu->env;
    ppc_tb_t *tb_env;
    booke_timer_t *booke_timer;

    tb_env = env->tb_env;
    booke_timer = tb_env->opaque;

    /* TODO: There's lots of complicated stuff to do here */

    booke_update_irq(cpu);

    booke_update_fixed_timer(env,
                             booke_get_wdt_target(env, tb_env),
                             &booke_timer->wdt_next,
                             booke_timer->wdt_timer,
                             TSR_WIS);
}

void store_booke_tsr(CPUPPCState *env, target_ulong val)
{
    PowerPCCPU *cpu = ppc_env_get_cpu(env);
    ppc_tb_t *tb_env = env->tb_env;
    booke_timer_t *booke_timer = tb_env->opaque;

    env->spr[SPR_BOOKE_TSR] &= ~val;
    kvmppc_clear_tsr_bits(cpu, val);

    if (val & TSR_FIS) {
        booke_update_fixed_timer(env,
                                 booke_get_fit_target(env, tb_env),
                                 &booke_timer->fit_next,
                                 booke_timer->fit_timer,
                                 TSR_FIS);
    }

    if (val & TSR_WIS) {
        booke_update_fixed_timer(env,
                                 booke_get_wdt_target(env, tb_env),
                                 &booke_timer->wdt_next,
                                 booke_timer->wdt_timer,
                                 TSR_WIS);
    }

    booke_update_irq(cpu);
}

void store_booke_tcr(CPUPPCState *env, target_ulong val)
{
    PowerPCCPU *cpu = ppc_env_get_cpu(env);
    ppc_tb_t *tb_env = env->tb_env;
    booke_timer_t *booke_timer = tb_env->opaque;

    env->spr[SPR_BOOKE_TCR] = val;
    kvmppc_set_tcr(cpu);

    booke_update_irq(cpu);

    booke_update_fixed_timer(env,
                             booke_get_fit_target(env, tb_env),
                             &booke_timer->fit_next,
                             booke_timer->fit_timer,
                             TSR_FIS);

    booke_update_fixed_timer(env,
                             booke_get_wdt_target(env, tb_env),
                             &booke_timer->wdt_next,
                             booke_timer->wdt_timer,
                             TSR_WIS);
}

static void ppc_booke_timer_reset_handle(void *opaque)
{
    PowerPCCPU *cpu = opaque;
    CPUPPCState *env = &cpu->env;

    store_booke_tcr(env, 0);
    store_booke_tsr(env, -1);
}

/*
 * This function will be called whenever the CPU state changes.
 * CPU states are defined "typedef enum RunState".
 * Regarding timer, When CPU state changes to running after debug halt
 * or similar cases which takes time then in between final watchdog
 * expiry happenes. This will cause exit to QEMU and configured watchdog
 * action will be taken. To avoid this we always clear the watchdog state when
 * state changes to running.
 */
static void cpu_state_change_handler(void *opaque, int running, RunState state)
{
    PowerPCCPU *cpu = opaque;
    CPUPPCState *env = &cpu->env;

    if (!running) {
        return;
    }

    /*
     * Clear watchdog interrupt condition by clearing TSR.
     */
    store_booke_tsr(env, TSR_ENW | TSR_WIS | TSR_WRS_MASK);
}

void ppc_booke_timers_init(PowerPCCPU *cpu, uint32_t freq, uint32_t flags)
{
    ppc_tb_t *tb_env;
    booke_timer_t *booke_timer;
    int ret = 0;

    tb_env      = g_malloc0(sizeof(ppc_tb_t));
    booke_timer = g_malloc0(sizeof(booke_timer_t));

    cpu->env.tb_env = tb_env;
    tb_env->flags = flags | PPC_TIMER_BOOKE | PPC_DECR_ZERO_TRIGGERED;

    tb_env->tb_freq    = freq;
    tb_env->decr_freq  = freq;
    tb_env->opaque     = booke_timer;
    tb_env->decr_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, &booke_decr_cb, cpu);

    booke_timer->fit_timer =
        timer_new_ns(QEMU_CLOCK_VIRTUAL, &booke_fit_cb, cpu);
    booke_timer->wdt_timer =
        timer_new_ns(QEMU_CLOCK_VIRTUAL, &booke_wdt_cb, cpu);

    ret = kvmppc_booke_watchdog_enable(cpu);

    if (ret) {
        /* TODO: Start the QEMU emulated watchdog if not running on KVM.
         * Also start the QEMU emulated watchdog if KVM does not support
         * emulated watchdog or somehow it is not enabled (supported but
         * not enabled is though some bug and requires debugging :)).
         */
    }

    qemu_add_vm_change_state_handler(cpu_state_change_handler, cpu);

    qemu_register_reset(ppc_booke_timer_reset_handle, cpu);
}

Offline darthnVader

  • Platinum Member
  • *****
  • Posts: 679
  • New Member
Re: Debuging why KVM can't boot OS 9.
« Reply #5 on: October 24, 2018, 04:19:11 PM »
Booting OS 9 in QEMU or on the bare metal seems to add this .property to the interrupt controller:

Code: [Select]
AAPL,address 80040000
but encoding this property with KVM doesn't seem to change anything as far as booting:

Code: [Select]
dev /pci/mac-io/@40000
80040000 encode-int " AAPL,address" property