kqemu support

git-svn-id: svn://svn.savannah.nongnu.org/qemu/trunk@1283 c046a42c-6fe2-441c-8c8c-71466251a162

5 files changed, 527 insertions, 11 deletions

diff --git a/cpu-exec.c b/cpu-exec.c
index ff548af7af..4adfb81aa1 100644
--- a/cpu-exec.c
+++ b/cpu-exec.c
@@ -209,7 +209,33 @@ int cpu_exec(CPUState *env1)
 #endif
                 }
                 env->exception_index = -1;
+            } 
+#ifdef USE_KQEMU
+            if (kqemu_is_ok(env) && env->interrupt_request == 0) {
+                int ret;
+                env->eflags = env->eflags | cc_table[CC_OP].compute_all() | (DF & DF_MASK);
+                ret = kqemu_cpu_exec(env);
+                /* put eflags in CPU temporary format */
+                CC_SRC = env->eflags & (CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
+                DF = 1 - (2 * ((env->eflags >> 10) & 1));
+                CC_OP = CC_OP_EFLAGS;
+                env->eflags &= ~(DF_MASK | CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
+                if (ret == 1) {
+                    /* exception */
+                    longjmp(env->jmp_env, 1);
+                } else if (ret == 2) {
+                    /* softmmu execution needed */
+                } else {
+                    if (env->interrupt_request != 0) {
+                        /* hardware interrupt will be executed just after */
+                    } else {
+                        /* otherwise, we restart */
+                        longjmp(env->jmp_env, 1);
+                    }
+                }
             }
+#endif
+
             T0 = 0; /* force lookup of first TB */
             for(;;) {
 #ifdef __sparc__
diff --git a/exec-all.h b/exec-all.h
index b1f488cd0a..3065858f3a 100644
--- a/exec-all.h
+++ b/exec-all.h
@@ -586,3 +586,25 @@ static inline target_ulong get_phys_addr_code(CPUState *env, target_ulong addr)
     return addr + env->tlb_read[is_user][index].addend - (unsigned long)phys_ram_base;
 }
 #endif
+
+
+#ifdef USE_KQEMU
+extern int kqemu_fd;
+extern int kqemu_flushed;
+
+int kqemu_init(CPUState *env);
+int kqemu_cpu_exec(CPUState *env);
+void kqemu_flush_page(CPUState *env, target_ulong addr);
+void kqemu_flush(CPUState *env, int global);
+
+static inline int kqemu_is_ok(CPUState *env)
+{
+    return(env->kqemu_enabled &&
+           (env->hflags & HF_CPL_MASK) == 3 &&
+           (env->eflags & IOPL_MASK) != IOPL_MASK &&
+           (env->cr[0] & CR0_PE_MASK) && 
+           (env->eflags & IF_MASK) &&
+           !(env->eflags & VM_MASK));
+}
+
+#endif
diff --git a/kqemu.c b/kqemu.c
new file mode 100644
index 0000000000..20aaf6376b
--- /dev/null
+++ b/kqemu.c
@@ -0,0 +1,427 @@
+/*
+ *  KQEMU support
+ * 
+ *  Copyright (c) 2005 Fabrice Bellard
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+#include "config.h"
+#ifdef _WIN32
+#include <windows.h>
+#else
+#include <sys/types.h>
+#include <sys/mman.h>
+#endif
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdarg.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#include <inttypes.h>
+
+#include "cpu.h"
+#include "exec-all.h"
+
+#ifdef USE_KQEMU
+
+#define DEBUG
+
+#include <unistd.h>
+#include <fcntl.h>
+#include <sys/ioctl.h>
+#include "kqemu/kqemu.h"
+
+#define KQEMU_DEVICE "/dev/kqemu"
+
+int kqemu_allowed = 1;
+int kqemu_fd = -1;
+unsigned long *pages_to_flush;
+unsigned int nb_pages_to_flush;
+extern uint32_t **l1_phys_map;
+
+#define cpuid(index, eax, ebx, ecx, edx) \
+  asm volatile ("cpuid" \
+                : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx) \
+                : "0" (index))
+
+static int is_cpuid_supported(void)
+{
+    int v0, v1;
+    asm volatile ("pushf\n"
+                  "popl %0\n"
+                  "movl %0, %1\n"
+                  "xorl $0x00200000, %0\n"
+                  "pushl %0\n"
+                  "popf\n"
+                  "pushf\n"
+                  "popl %0\n"
+                  : "=a" (v0), "=d" (v1)
+                  :
+                  : "cc");
+    return (v0 != v1);
+}
+
+static void kqemu_update_cpuid(CPUState *env)
+{
+    int critical_features_mask, features;
+    uint32_t eax, ebx, ecx, edx;
+
+    /* the following features are kept identical on the host and
+       target cpus because they are important for user code. Strictly
+       speaking, only SSE really matters because the OS must support
+       it if the user code uses it. */
+    critical_features_mask = 
+        CPUID_CMOV | CPUID_CX8 | 
+        CPUID_FXSR | CPUID_MMX | CPUID_SSE | 
+        CPUID_SSE2;
+    if (!is_cpuid_supported()) {
+        features = 0;
+    } else {
+        cpuid(1, eax, ebx, ecx, edx);
+        features = edx;
+    }
+    env->cpuid_features = (env->cpuid_features & ~critical_features_mask) |
+        (features & critical_features_mask);
+    /* XXX: we could update more of the target CPUID state so that the
+       non accelerated code sees exactly the same CPU features as the
+       accelerated code */
+}
+
+int kqemu_init(CPUState *env)
+{
+    struct kqemu_init init;
+    int ret, version;
+
+    if (!kqemu_allowed)
+        return -1;
+
+    kqemu_fd = open(KQEMU_DEVICE, O_RDWR);
+    if (kqemu_fd < 0) {
+        fprintf(stderr, "Could not open '%s' - QEMU acceleration layer not activated\n", KQEMU_DEVICE);
+        return -1;
+    }
+    version = 0;
+    ioctl(kqemu_fd, KQEMU_GET_VERSION, &version);
+    if (version != KQEMU_VERSION) {
+        fprintf(stderr, "Version mismatch between kqemu module and qemu (%08x %08x) - disabling kqemu use\n",
+                version, KQEMU_VERSION);
+        goto fail;
+    }
+
+    pages_to_flush = qemu_vmalloc(KQEMU_MAX_PAGES_TO_FLUSH * 
+                                  sizeof(unsigned long));
+    if (!pages_to_flush)
+        goto fail;
+
+    init.ram_base = phys_ram_base;
+    init.ram_size = phys_ram_size;
+    init.ram_dirty = phys_ram_dirty;
+    init.phys_to_ram_map = l1_phys_map;
+    init.pages_to_flush = pages_to_flush;
+    ret = ioctl(kqemu_fd, KQEMU_INIT, &init);
+    if (ret < 0) {
+        fprintf(stderr, "Error %d while initializing QEMU acceleration layer - disabling it for now\n", ret);
+    fail:
+        close(kqemu_fd);
+        kqemu_fd = -1;
+        return -1;
+    }
+    kqemu_update_cpuid(env);
+    env->kqemu_enabled = 1;
+    nb_pages_to_flush = 0;
+    return 0;
+}
+
+void kqemu_flush_page(CPUState *env, target_ulong addr)
+{
+#ifdef DEBUG
+    if (loglevel & CPU_LOG_INT) {
+        fprintf(logfile, "kqemu_flush_page: addr=" TARGET_FMT_lx "\n", addr);
+    }
+#endif
+    if (nb_pages_to_flush >= KQEMU_MAX_PAGES_TO_FLUSH)
+        nb_pages_to_flush = KQEMU_FLUSH_ALL;
+    else
+        pages_to_flush[nb_pages_to_flush++] = addr;
+}
+
+void kqemu_flush(CPUState *env, int global)
+{
+#ifdef DEBUG
+    if (loglevel & CPU_LOG_INT) {
+        fprintf(logfile, "kqemu_flush:\n");
+    }
+#endif
+    nb_pages_to_flush = KQEMU_FLUSH_ALL;
+}
+
+struct fpstate {
+    uint16_t fpuc;
+    uint16_t dummy1;
+    uint16_t fpus;
+    uint16_t dummy2;
+    uint16_t fptag;
+    uint16_t dummy3;
+
+    uint32_t fpip;
+    uint32_t fpcs;
+    uint32_t fpoo;
+    uint32_t fpos;
+    uint8_t fpregs1[8 * 10];
+};
+
+struct fpxstate {
+    uint16_t fpuc;
+    uint16_t fpus;
+    uint16_t fptag;
+    uint16_t fop;
+    uint32_t fpuip;
+    uint16_t cs_sel;
+    uint16_t dummy0;
+    uint32_t fpudp;
+    uint16_t ds_sel;
+    uint16_t dummy1;
+    uint32_t mxcsr;
+    uint32_t mxcsr_mask;
+    uint8_t fpregs1[8 * 16];
+    uint8_t xmm_regs[8 * 16];
+    uint8_t dummy2[224];
+};
+
+static struct fpxstate fpx1 __attribute__((aligned(16)));
+
+static void restore_native_fp_frstor(CPUState *env)
+{
+    int fptag, i, j;
+    struct fpstate fp1, *fp = &fp1;
+    
+    fp->fpuc = env->fpuc;
+    fp->fpus = (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11;
+    fptag = 0;
+    for (i=7; i>=0; i--) {
+	fptag <<= 2;
+	if (env->fptags[i]) {
+            fptag |= 3;
+        } else {
+            /* the FPU automatically computes it */
+        }
+    }
+    fp->fptag = fptag;
+    j = env->fpstt;
+    for(i = 0;i < 8; i++) {
+        memcpy(&fp->fpregs1[i * 10], &env->fpregs[j].d, 10);
+        j = (j + 1) & 7;
+    }
+    asm volatile ("frstor %0" : "=m" (*fp));
+}
+ 
+static void save_native_fp_fsave(CPUState *env)
+{
+    int fptag, i, j;
+    uint16_t fpuc;
+    struct fpstate fp1, *fp = &fp1;
+
+    asm volatile ("fsave %0" : : "m" (*fp));
+    env->fpuc = fp->fpuc;
+    env->fpstt = (fp->fpus >> 11) & 7;
+    env->fpus = fp->fpus & ~0x3800;
+    fptag = fp->fptag;
+    for(i = 0;i < 8; i++) {
+        env->fptags[i] = ((fptag & 3) == 3);
+        fptag >>= 2;
+    }
+    j = env->fpstt;
+    for(i = 0;i < 8; i++) {
+        memcpy(&env->fpregs[j].d, &fp->fpregs1[i * 10], 10);
+        j = (j + 1) & 7;
+    }
+    /* we must restore the default rounding state */
+    fpuc = 0x037f | (env->fpuc & (3 << 10));
+    asm volatile("fldcw %0" : : "m" (fpuc));
+}
+
+static void restore_native_fp_fxrstor(CPUState *env)
+{
+    struct fpxstate *fp = &fpx1;
+    int i, j, fptag;
+
+    fp->fpuc = env->fpuc;
+    fp->fpus = (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11;
+    fptag = 0;
+    for(i = 0; i < 8; i++)
+        fptag |= (env->fptags[i] << i);
+    fp->fptag = fptag ^ 0xff;
+
+    j = env->fpstt;
+    for(i = 0;i < 8; i++) {
+        memcpy(&fp->fpregs1[i * 16], &env->fpregs[j].d, 10);
+        j = (j + 1) & 7;
+    }
+    if (env->cpuid_features & CPUID_SSE) {
+        fp->mxcsr = env->mxcsr;
+        /* XXX: check if DAZ is not available */
+        fp->mxcsr_mask = 0xffff;
+        memcpy(fp->xmm_regs, env->xmm_regs, 8 * 16);
+    }
+    asm volatile ("fxrstor %0" : "=m" (*fp));
+}
+
+static void save_native_fp_fxsave(CPUState *env)
+{
+    struct fpxstate *fp = &fpx1;
+    int fptag, i, j;
+    uint16_t fpuc;
+
+    asm volatile ("fxsave %0" : : "m" (*fp));
+    env->fpuc = fp->fpuc;
+    env->fpstt = (fp->fpus >> 11) & 7;
+    env->fpus = fp->fpus & ~0x3800;
+    fptag = fp->fptag ^ 0xff;
+    for(i = 0;i < 8; i++) {
+        env->fptags[i] = (fptag >> i) & 1;
+    }
+    j = env->fpstt;
+    for(i = 0;i < 8; i++) {
+        memcpy(&env->fpregs[j].d, &fp->fpregs1[i * 16], 10);
+        j = (j + 1) & 7;
+    }
+    if (env->cpuid_features & CPUID_SSE) {
+        env->mxcsr = fp->mxcsr;
+        memcpy(env->xmm_regs, fp->xmm_regs, 8 * 16);
+    }
+
+    /* we must restore the default rounding state */
+    asm volatile ("fninit");
+    fpuc = 0x037f | (env->fpuc & (3 << 10));
+    asm volatile("fldcw %0" : : "m" (fpuc));
+}
+
+int kqemu_cpu_exec(CPUState *env)
+{
+    struct kqemu_cpu_state kcpu_state, *kenv = &kcpu_state;
+    int ret;
+
+#ifdef DEBUG
+    if (loglevel & CPU_LOG_INT) {
+        fprintf(logfile, "kqemu: cpu_exec: enter\n");
+        cpu_dump_state(env, logfile, fprintf, 0);
+    }
+#endif
+    memcpy(kenv->regs, env->regs, sizeof(kenv->regs));
+    kenv->eip = env->eip;
+    kenv->eflags = env->eflags;
+    memcpy(&kenv->segs, &env->segs, sizeof(env->segs));
+    memcpy(&kenv->ldt, &env->ldt, sizeof(env->ldt));
+    memcpy(&kenv->tr, &env->tr, sizeof(env->tr));
+    memcpy(&kenv->gdt, &env->gdt, sizeof(env->gdt));
+    memcpy(&kenv->idt, &env->idt, sizeof(env->idt));
+    kenv->cr0 = env->cr[0];
+    kenv->cr2 = env->cr[2];
+    kenv->cr3 = env->cr[3];
+    kenv->cr4 = env->cr[4];
+    kenv->a20_mask = env->a20_mask;
+    if (env->dr[7] & 0xff) {
+        kenv->dr7 = env->dr[7];
+        kenv->dr0 = env->dr[0];
+        kenv->dr1 = env->dr[1];
+        kenv->dr2 = env->dr[2];
+        kenv->dr3 = env->dr[3];
+    } else {
+        kenv->dr7 = 0;
+    }
+    kenv->dr6 = env->dr[6];
+    kenv->cpl = 3;
+    kenv->nb_pages_to_flush = nb_pages_to_flush;
+    nb_pages_to_flush = 0;
+    
+    if (!(kenv->cr0 & CR0_TS_MASK)) {
+        if (env->cpuid_features & CPUID_FXSR)
+            restore_native_fp_fxrstor(env);
+        else
+            restore_native_fp_frstor(env);
+    }
+
+    ret = ioctl(kqemu_fd, KQEMU_EXEC, kenv);
+
+    if (!(kenv->cr0 & CR0_TS_MASK)) {
+        if (env->cpuid_features & CPUID_FXSR)
+            save_native_fp_fxsave(env);
+        else
+            save_native_fp_fsave(env);
+    }
+
+    memcpy(env->regs, kenv->regs, sizeof(env->regs));
+    env->eip = kenv->eip;
+    env->eflags = kenv->eflags;
+    memcpy(env->segs, kenv->segs, sizeof(env->segs));
+#if 0
+    /* no need to restore that */
+    memcpy(env->ldt, kenv->ldt, sizeof(env->ldt));
+    memcpy(env->tr, kenv->tr, sizeof(env->tr));
+    memcpy(env->gdt, kenv->gdt, sizeof(env->gdt));
+    memcpy(env->idt, kenv->idt, sizeof(env->idt));
+    env->cr[0] = kenv->cr0;
+    env->cr[3] = kenv->cr3;
+    env->cr[4] = kenv->cr4;
+    env->a20_mask = kenv->a20_mask;
+#endif
+    env->cr[2] = kenv->cr2;
+    env->dr[6] = kenv->dr6;
+
+#ifdef DEBUG
+    if (loglevel & CPU_LOG_INT) {
+        fprintf(logfile, "kqemu: kqemu_cpu_exec: ret=0x%x\n", ret);
+    }
+#endif
+    if ((ret & 0xff00) == KQEMU_RET_INT) {
+        env->exception_index = ret & 0xff;
+        env->error_code = 0;
+        env->exception_is_int = 1;
+        env->exception_next_eip = kenv->next_eip;
+#ifdef DEBUG
+    if (loglevel & CPU_LOG_INT) {
+        fprintf(logfile, "kqemu: interrupt v=%02x:\n", 
+                env->exception_index);
+        cpu_dump_state(env, logfile, fprintf, 0);
+    }
+#endif
+        return 1;
+    } else if ((ret & 0xff00) == KQEMU_RET_EXCEPTION) {
+        env->exception_index = ret & 0xff;
+        env->error_code = kenv->error_code;
+        env->exception_is_int = 0;
+        env->exception_next_eip = 0;
+#ifdef DEBUG
+        if (loglevel & CPU_LOG_INT) {
+            fprintf(logfile, "kqemu: exception v=%02x e=%04x:\n",
+                    env->exception_index, env->error_code);
+            cpu_dump_state(env, logfile, fprintf, 0);
+        }
+#endif
+        return 1;
+    } else if (ret == KQEMU_RET_INTR) {
+        return 0;
+    } else if (ret == KQEMU_RET_SOFTMMU) { 
+        return 2;
+    } else {
+        cpu_dump_state(env, stderr, fprintf, 0);
+        fprintf(stderr, "Unsupported return value: 0x%x\n", ret);
+        exit(1);
+    }
+    return 0;
+}
+
+#endif
diff --git a/target-i386/cpu.h b/target-i386/cpu.h
index 781be2838e..74c882b867 100644
--- a/target-i386/cpu.h
+++ b/target-i386/cpu.h
@@ -39,6 +39,9 @@
 #if defined(__i386__) && !defined(CONFIG_SOFTMMU)
 #define USE_CODE_COPY
 #endif
+#if defined(__linux__) && defined(CONFIG_SOFTMMU) && defined(__i386__) && !defined(TARGET_X86_64)
+#define USE_KQEMU
+#endif
 
 #define R_EAX 0
 #define R_ECX 1
@@ -248,6 +251,14 @@
 #define CPUID_SSE  (1 << 25)
 #define CPUID_SSE2 (1 << 26)
 
+#define CPUID_EXT_SS3      (1 << 0)
+#define CPUID_EXT_MONITOR  (1 << 3)
+#define CPUID_EXT_CX16     (1 << 13)
+
+#define CPUID_EXT2_SYSCALL (1 << 11)
+#define CPUID_EXT2_NX      (1 << 20)
+#define CPUID_EXT2_LM      (1 << 29)
+
 #define EXCP00_DIVZ	0
 #define EXCP01_SSTP	1
 #define EXCP02_NMI	2
@@ -408,6 +419,16 @@ typedef struct CPUX86State {
     int32_t df; /* D flag : 1 if D = 0, -1 if D = 1 */
     uint32_t hflags; /* hidden flags, see HF_xxx constants */
 
+    /* segments */
+    SegmentCache segs[6]; /* selector values */
+    SegmentCache ldt;
+    SegmentCache tr;
+    SegmentCache gdt; /* only base and limit are used */
+    SegmentCache idt; /* only base and limit are used */
+
+    target_ulong cr[5]; /* NOTE: cr1 is unused */
+    uint32_t a20_mask;
+
     /* FPU state */
     unsigned int fpstt; /* top of stack index */
     unsigned int fpus;
@@ -431,13 +452,6 @@ typedef struct CPUX86State {
         int64_t i64;
     } fp_convert;
     
-    /* segments */
-    SegmentCache segs[6]; /* selector values */
-    SegmentCache ldt;
-    SegmentCache tr;
-    SegmentCache gdt; /* only base and limit are used */
-    SegmentCache idt; /* only base and limit are used */
-
     uint32_t mxcsr;
     XMMReg xmm_regs[CPU_NB_REGS];
     XMMReg xmm_t0;
@@ -470,13 +484,10 @@ typedef struct CPUX86State {
     int exception_is_int;
     target_ulong exception_next_eip;
     struct TranslationBlock *current_tb; /* currently executing TB */
-    target_ulong cr[5]; /* NOTE: cr1 is unused */
     target_ulong dr[8]; /* debug registers */
     int interrupt_request; 
     int user_mode_only; /* user mode only simulation */
 
-    uint32_t a20_mask;
-
     /* soft mmu support */
     /* in order to avoid passing too many arguments to the memory
        write helpers, we store some rarely used information in the CPU
@@ -501,7 +512,11 @@ typedef struct CPUX86State {
     uint32_t cpuid_vendor3;
     uint32_t cpuid_version;
     uint32_t cpuid_features;
+    uint32_t cpuid_ext_features;
 
+#ifdef USE_KQEMU
+    int kqemu_enabled;
+#endif
     /* in order to simplify APIC support, we leave this pointer to the
        user */
     struct APICState *apic_state;
diff --git a/target-i386/helper.c b/target-i386/helper.c
index ea2ef5e027..b727e93d66 100644
--- a/target-i386/helper.c
+++ b/target-i386/helper.c
@@ -1274,7 +1274,7 @@ void helper_cpuid(void)
     case 1:
         EAX = env->cpuid_version;
         EBX = 0;
-        ECX = 0;
+        ECX = env->cpuid_ext_features;
         EDX = env->cpuid_features;
         break;
     default:
@@ -1828,6 +1828,12 @@ void helper_lcall_protected_T0_T1(int shift, int next_eip)
         ESP = (ESP & ~sp_mask) | (sp & sp_mask);
         EIP = offset;
     }
+#ifdef USE_KQEMU
+    if (kqemu_is_ok(env)) {
+        env->exception_index = -1;
+        cpu_loop_exit();
+    }
+#endif
 }
 
 /* real and vm86 mode iret */
@@ -2097,11 +2103,25 @@ void helper_iret_protected(int shift, int next_eip)
     } else {
         helper_ret_protected(shift, 1, 0);
     }
+#ifdef USE_KQEMU
+    if (kqemu_is_ok(env)) {
+        CC_OP = CC_OP_EFLAGS;
+        env->exception_index = -1;
+        cpu_loop_exit();
+    }
+#endif
 }
 
 void helper_lret_protected(int shift, int addend)
 {
     helper_ret_protected(shift, 0, addend);
+#ifdef USE_KQEMU
+    if (kqemu_is_ok(env)) {
+        CC_OP = CC_OP_EFLAGS;
+        env->exception_index = -1;
+        cpu_loop_exit();
+    }
+#endif
 }
 
 void helper_sysenter(void)
@@ -2146,6 +2166,12 @@ void helper_sysexit(void)
                            DESC_W_MASK | DESC_A_MASK);
     ESP = ECX;
     EIP = EDX;
+#ifdef USE_KQEMU
+    if (kqemu_is_ok(env)) {
+        env->exception_index = -1;
+        cpu_loop_exit();
+    }
+#endif
 }
 
 void helper_movl_crN_T0(int reg)