[Qemu-devel] [PATCH] tcg: Use the GDB JIT debugging interface.

[Richard Henderson <rth@twiddle.net>]

This allows us to generate unwind info for the dynamicly generated
code in the code_gen_buffer.  Only i386 is converted at this point.

Signed-off-by: Richard Henderson <rth@twiddle.net>
---
 elf.h                 |    1 +
 exec.c                |    1 +
 tcg/i386/tcg-target.c |  114 +++++++++++++++++++--
 tcg/tcg.c             |  277 +++++++++++++++++++++++++++++++++++++++++++++++++
 tcg/tcg.h             |    2 +
 5 files changed, 386 insertions(+), 9 deletions(-)

diff --git a/elf.h b/elf.h
index 2e05d34..310e05a 100644
--- a/elf.h
+++ b/elf.h
@@ -216,6 +216,7 @@ typedef int64_t  Elf64_Sxword;
 
 #define ELF_ST_BIND(x)		((x) >> 4)
 #define ELF_ST_TYPE(x)		(((unsigned int) x) & 0xf)
+#define ELF_ST_INFO(bind, type) (((bind) << 4) | ((type) & 0xf))
 #define ELF32_ST_BIND(x)	ELF_ST_BIND(x)
 #define ELF32_ST_TYPE(x)	ELF_ST_TYPE(x)
 #define ELF64_ST_BIND(x)	ELF_ST_BIND(x)
diff --git a/exec.c b/exec.c
index be392e2..dd24939 100644
--- a/exec.c
+++ b/exec.c
@@ -636,6 +636,7 @@ void tcg_exec_init(unsigned long tb_size)
     cpu_gen_init();
     code_gen_alloc(tb_size);
     code_gen_ptr = code_gen_buffer;
+    tcg_register_jit(code_gen_buffer, code_gen_buffer_size);
     page_init();
 #if !defined(CONFIG_USER_ONLY) || !defined(CONFIG_USE_GUEST_BASE)
     /* There's no guest base to take into account, so go ahead and
diff --git a/tcg/i386/tcg-target.c b/tcg/i386/tcg-target.c
index 43a51a1..871a7e7 100644
--- a/tcg/i386/tcg-target.c
+++ b/tcg/i386/tcg-target.c
@@ -1989,22 +1989,29 @@ static int tcg_target_callee_save_regs[] = {
 #endif
 };
 
+/* Compute frame size via macros, to share between tcg_target_qemu_prologue
+   and tcg_register_jit.  */
+
+#define PUSH_SIZE \
+    ((1 + ARRAY_SIZE(tcg_target_callee_save_regs)) \
+     * (TCG_TARGET_REG_BITS / 8))
+
+#define FRAME_SIZE \
+    ((PUSH_SIZE \
+      + TCG_STATIC_CALL_ARGS_SIZE \
+      + CPU_TEMP_BUF_NLONGS * sizeof(long) \
+      + TCG_TARGET_STACK_ALIGN - 1) \
+     & ~(TCG_TARGET_STACK_ALIGN - 1))
+
 /* Generate global QEMU prologue and epilogue code */
 static void tcg_target_qemu_prologue(TCGContext *s)
 {
-    int i, frame_size, push_size, stack_addend;
+    int i, stack_addend;
 
     /* TB prologue */
 
     /* Reserve some stack space, also for TCG temps.  */
-    push_size = 1 + ARRAY_SIZE(tcg_target_callee_save_regs);
-    push_size *= TCG_TARGET_REG_BITS / 8;
-
-    frame_size = push_size + TCG_STATIC_CALL_ARGS_SIZE +
-        CPU_TEMP_BUF_NLONGS * sizeof(long);
-    frame_size = (frame_size + TCG_TARGET_STACK_ALIGN - 1) &
-        ~(TCG_TARGET_STACK_ALIGN - 1);
-    stack_addend = frame_size - push_size;
+    stack_addend = FRAME_SIZE - PUSH_SIZE;
     tcg_set_frame(s, TCG_REG_CALL_STACK, TCG_STATIC_CALL_ARGS_SIZE,
                   CPU_TEMP_BUF_NLONGS * sizeof(long));
 
@@ -2070,3 +2077,92 @@ static void tcg_target_init(TCGContext *s)
 
     tcg_add_target_add_op_defs(x86_op_defs);
 }
+
+typedef struct {
+    uint32_t len __attribute__((aligned((sizeof(void *)))));
+    uint32_t id;
+    uint8_t version;
+    char augmentation[1];
+    uint8_t code_align;
+    uint8_t data_align;
+    uint8_t return_column;
+} DebugFrameCIE;
+
+typedef struct {
+    uint32_t len __attribute__((aligned((sizeof(void *)))));
+    uint32_t cie_offset;
+    tcg_target_long func_start __attribute__((packed));
+    tcg_target_long func_len __attribute__((packed));
+    uint8_t def_cfa[4];
+    uint8_t reg_ofs[14];
+} DebugFrameFDE;
+
+typedef struct {
+    DebugFrameCIE cie;
+    DebugFrameFDE fde;
+} DebugFrame;
+
+#if TCG_TARGET_REG_BITS == 64
+#define ELF_HOST_MACHINE EM_X86_64
+static DebugFrame debug_frame = {
+    .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */
+    .cie.id = -1,
+    .cie.version = 1,
+    .cie.code_align = 1,
+    .cie.data_align = 0x78,             /* sleb128 -8 */
+    .cie.return_column = 16,
+
+    .fde.len = sizeof(DebugFrameFDE)-4, /* length after .len member */
+    .fde.def_cfa = {
+        12, 7,                          /* DW_CFA_def_cfa %rsp, ... */
+        (FRAME_SIZE & 0x7f) | 0x80,     /* ... uleb128 FRAME_SIZE */
+        (FRAME_SIZE >> 7)
+    },
+    .fde.reg_ofs = {
+        0x90, 1,                        /* DW_CFA_offset, %rip, -8 */
+        /* The following ordering must match tcg_target_callee_save_regs.  */
+        0x86, 2,                        /* DW_CFA_offset, %rbp, -16 */
+        0x83, 3,                        /* DW_CFA_offset, %rbx, -24 */
+        0x8c, 4,                        /* DW_CFA_offset, %r12, -32 */
+        0x8d, 5,                        /* DW_CFA_offset, %r13, -40 */
+        0x8e, 6,                        /* DW_CFA_offset, %r14, -48 */
+        0x8f, 7,                        /* DW_CFA_offset, %r15, -56 */
+    }
+};
+#else
+#define ELF_HOST_MACHINE EM_386
+static DebugFrame debug_frame = {
+    .cie.len = sizeof(DebugFrameCIE)-4, /* length after .len member */
+    .cie.id = -1,
+    .cie.version = 1,
+    .cie.code_align = 1,
+    .cie.data_align = 0x7c,             /* sleb128 -4 */
+    .cie.return_column = 8,
+
+    .fde.len = sizeof(DebugFrameFDE)-4, /* length after .len member */
+    .fde.def_cfa = {
+        12, 4,                          /* DW_CFA_def_cfa %esp, ... */
+        (FRAME_SIZE & 0x7f) | 0x80,     /* ... uleb128 FRAME_SIZE */
+        (FRAME_SIZE >> 7)
+    },
+    .fde.reg_ofs = {
+        0x88, 1,                        /* DW_CFA_offset, %eip, -4 */
+        /* The following ordering must match tcg_target_callee_save_regs.  */
+        0x85, 2,                        /* DW_CFA_offset, %ebp, -8 */
+        0x83, 3,                        /* DW_CFA_offset, %ebx, -12 */
+        0x86, 4,                        /* DW_CFA_offset, %esi, -16 */
+        0x87, 5,                        /* DW_CFA_offset, %edi, -20 */
+    }
+};
+#endif
+
+void tcg_register_jit(void *buf, size_t buf_size)
+{
+    /* We're expecting a 2 byte uleb128 encoded value.  */
+    assert(FRAME_SIZE >> 14 == 0);
+
+    debug_frame.fde.func_start = (tcg_target_long) buf;
+    debug_frame.fde.func_len = buf_size;
+
+    tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame));
+}
diff --git a/tcg/tcg.c b/tcg/tcg.c
index ccfcd1a..df4edc0 100644
--- a/tcg/tcg.c
+++ b/tcg/tcg.c
@@ -28,6 +28,9 @@
 
 #include "config.h"
 
+/* Define to jump the ELF file used to communicate with GDB.  */
+#undef DEBUG_JIT
+
 #if !defined(CONFIG_DEBUG_TCG) && !defined(NDEBUG)
 /* define it to suppress various consistency checks (faster) */
 #define NDEBUG
@@ -45,6 +48,18 @@
 #include "cpu.h"
 
 #include "tcg-op.h"
+
+#if TCG_TARGET_REG_BITS == 64
+# define ELF_CLASS  ELFCLASS64
+#else
+# define ELF_CLASS  ELFCLASS32
+#endif
+#ifdef HOST_WORDS_BIGENDIAN
+# define ELF_DATA   ELFDATA2MSB
+#else
+# define ELF_DATA   ELFDATA2LSB
+#endif
+
 #include "elf.h"
 
 #if defined(CONFIG_USE_GUEST_BASE) && !defined(TCG_TARGET_HAS_GUEST_BASE)
@@ -57,6 +72,10 @@ static void tcg_target_qemu_prologue(TCGContext *s);
 static void patch_reloc(uint8_t *code_ptr, int type, 
                         tcg_target_long value, tcg_target_long addend);
 
+static void tcg_register_jit_int(void *buf, size_t size,
+                                 void *debug_frame, size_t debug_frame_size)
+    __attribute__((unused));
+
 /* Forward declarations for functions declared and used in tcg-target.c. */
 static int target_parse_constraint(TCGArgConstraint *ct, const char **pct_str);
 static void tcg_out_ld(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg1,
@@ -2231,3 +2250,261 @@ void tcg_dump_info(FILE *f, fprintf_function cpu_fprintf)
     cpu_fprintf(f, "[TCG profiler not compiled]\n");
 }
 #endif
+
+#ifdef ELF_HOST_MACHINE
+/* In order to use this feature, the backend needs to do three things:
+
+   (1) Define ELF_HOST_MACHINE to indicate both what value to
+       put into the ELF image and to indicate support for the feature.
+
+   (2) Define tcg_register_jit.  This should create a buffer containing
+       the contents of a .debug_frame section that describes the post-
+       prologue unwind info for the tcg machine.
+
+   (3) Call tcg_register_jit_int, with the constructed .debug_frame.
+*/
+
+/* Begin GDB interface.  THE FOLLOWING MUST MATCH GDB DOCS.  */
+typedef enum {
+    JIT_NOACTION = 0,
+    JIT_REGISTER_FN,
+    JIT_UNREGISTER_FN
+} jit_actions_t;
+
+struct jit_code_entry {
+    struct jit_code_entry *next_entry;
+    struct jit_code_entry *prev_entry;
+    const void *symfile_addr;
+    uint64_t symfile_size;
+};
+
+struct jit_descriptor {
+    uint32_t version;
+    uint32_t action_flag;
+    struct jit_code_entry *relevant_entry;
+    struct jit_code_entry *first_entry;
+};
+
+void __jit_debug_register_code(void) __attribute__((noinline));
+void __jit_debug_register_code(void)
+{
+    asm("");
+}
+
+/* Must statically initialize the version, because GDB may check
+   the version before we can set it.  */
+struct jit_descriptor __jit_debug_descriptor = { 1, 0, 0, 0 };
+
+/* End GDB interface.  */
+
+static int find_string(const char *strtab, const char *str)
+{
+    const char *p = strtab + 1;
+
+    while (1) {
+        if (strcmp(p, str) == 0) {
+            return p - strtab;
+        }
+        p += strlen(p) + 1;
+    }
+}
+
+static void tcg_register_jit_int(void *buf_ptr, size_t buf_size,
+                                 void *debug_frame, size_t debug_frame_size)
+{
+    struct __attribute__((packed)) DebugInfo {
+        uint32_t  len;
+        uint16_t  version;
+        uint32_t  abbrev;
+        uint8_t   ptr_size;
+        uint8_t   cu_die;
+        uint16_t  cu_lang;
+        uintptr_t cu_low_pc;
+        uintptr_t cu_high_pc;
+        uint8_t   fn_die;
+        char      fn_name[16];
+        uintptr_t fn_low_pc;
+        uintptr_t fn_high_pc;
+        uint8_t   cu_eoc;
+    };
+
+    struct ElfImage {
+        ElfW(Ehdr) ehdr;
+        ElfW(Phdr) phdr;
+        ElfW(Shdr) shdr[7];
+        ElfW(Sym)  sym[2];
+        struct DebugInfo di;
+        uint8_t    da[24];
+        char       str[80];
+    };
+
+    struct ElfImage *img;
+
+    static const struct ElfImage img_template = {
+        .ehdr = {
+            .e_ident[EI_MAG0] = ELFMAG0,
+            .e_ident[EI_MAG1] = ELFMAG1,
+            .e_ident[EI_MAG2] = ELFMAG2,
+            .e_ident[EI_MAG3] = ELFMAG3,
+            .e_ident[EI_CLASS] = ELF_CLASS,
+            .e_ident[EI_DATA] = ELF_DATA,
+            .e_ident[EI_VERSION] = EV_CURRENT,
+            .e_type = ET_EXEC,
+            .e_machine = ELF_HOST_MACHINE,
+            .e_version = EV_CURRENT,
+            .e_phoff = offsetof(struct ElfImage, phdr),
+            .e_shoff = offsetof(struct ElfImage, shdr),
+            .e_ehsize = sizeof(ElfW(Shdr)),
+            .e_phentsize = sizeof(ElfW(Phdr)),
+            .e_phnum = 1,
+            .e_shentsize = sizeof(ElfW(Shdr)),
+            .e_shnum = ARRAY_SIZE(img->shdr),
+            .e_shstrndx = ARRAY_SIZE(img->shdr) - 1,
+        },
+        .phdr = {
+            .p_type = PT_LOAD,
+            .p_flags = PF_X,
+        },
+        .shdr = {
+            [0] = { .sh_type = SHT_NULL },
+            /* Trick: The contents of code_gen_buffer are not present in
+               this fake ELF file; that got allocated elsewhere.  Therefore
+               we mark .text as SHT_NOBITS (similar to .bss) so that readers
+               will not look for contents.  We can record any address.  */
+            [1] = { /* .text */
+                .sh_type = SHT_NOBITS,
+                .sh_flags = SHF_EXECINSTR | SHF_ALLOC,
+            },
+            [2] = { /* .debug_info */
+                .sh_type = SHT_PROGBITS,
+                .sh_offset = offsetof(struct ElfImage, di),
+                .sh_size = sizeof(struct DebugInfo),
+            },
+            [3] = { /* .debug_abbrev */
+                .sh_type = SHT_PROGBITS,
+                .sh_offset = offsetof(struct ElfImage, da),
+                .sh_size = sizeof(img->da),
+            },
+            [4] = { /* .debug_frame */
+                .sh_type = SHT_PROGBITS,
+                .sh_offset = sizeof(struct ElfImage),
+            },
+            [5] = { /* .symtab */
+                .sh_type = SHT_SYMTAB,
+                .sh_offset = offsetof(struct ElfImage, sym),
+                .sh_size = sizeof(img->sym),
+                .sh_info = 1,
+                .sh_link = ARRAY_SIZE(img->shdr) - 1,
+                .sh_entsize = sizeof(ElfW(Sym)),
+            },
+            [6] = { /* .strtab */
+                .sh_type = SHT_STRTAB,
+                .sh_offset = offsetof(struct ElfImage, str),
+                .sh_size = sizeof(img->str),
+            }
+        },
+        .sym = {
+            [1] = { /* code_gen_buffer */
+                .st_info = ELF_ST_INFO(STB_GLOBAL, STT_FUNC),
+                .st_shndx = 1,
+            }
+        },
+        .di = {
+            .len = sizeof(struct DebugInfo) - 4,
+            .version = 2,
+            .ptr_size = sizeof(void *),
+            .cu_die = 1,
+            .cu_lang = 0x8001,  /* DW_LANG_Mips_Assembler */
+            .fn_die = 2,
+            .fn_name = "code_gen_buffer"
+        },
+        .da = {
+            1,          /* abbrev number (the cu) */
+            0x11, 1,    /* DW_TAG_compile_unit, has children */
+            0x13, 0x5,  /* DW_AT_language, DW_FORM_data2 */
+            0x11, 0x1,  /* DW_AT_low_pc, DW_FORM_addr */
+            0x12, 0x1,  /* DW_AT_high_pc, DW_FORM_addr */
+            0, 0,       /* end of abbrev */
+            2,          /* abbrev number (the fn) */
+            0x2e, 0,    /* DW_TAG_subprogram, no children */
+            0x3, 0x8,   /* DW_AT_name, DW_FORM_string */
+            0x11, 0x1,  /* DW_AT_low_pc, DW_FORM_addr */
+            0x12, 0x1,  /* DW_AT_high_pc, DW_FORM_addr */
+            0, 0,       /* end of abbrev */
+            0           /* no more abbrev */
+        },
+        .str = "\0" ".text\0" ".debug_info\0" ".debug_abbrev\0"
+               ".debug_frame\0" ".symtab\0" ".strtab\0" "code_gen_buffer",
+    };
+
+    /* We only need a single jit entry; statically allocate it.  */
+    static struct jit_code_entry one_entry;
+
+    uintptr_t buf = (uintptr_t)buf_ptr;
+    size_t img_size = sizeof(struct ElfImage) + debug_frame_size;
+
+    img = g_malloc(img_size);
+    *img = img_template;
+    memcpy(img + 1, debug_frame, debug_frame_size);
+
+    img->phdr.p_vaddr = buf;
+    img->phdr.p_paddr = buf;
+    img->phdr.p_memsz = buf_size;
+
+    img->shdr[1].sh_name = find_string(img->str, ".text");
+    img->shdr[1].sh_addr = buf;
+    img->shdr[1].sh_size = buf_size;
+
+    img->shdr[2].sh_name = find_string(img->str, ".debug_info");
+    img->shdr[3].sh_name = find_string(img->str, ".debug_abbrev");
+
+    img->shdr[4].sh_name = find_string(img->str, ".debug_frame");
+    img->shdr[4].sh_size = debug_frame_size;
+
+    img->shdr[5].sh_name = find_string(img->str, ".symtab");
+    img->shdr[6].sh_name = find_string(img->str, ".strtab");
+
+    img->sym[1].st_name = find_string(img->str, "code_gen_buffer");
+    img->sym[1].st_value = buf;
+    img->sym[1].st_size = buf_size;
+
+    img->di.cu_low_pc = buf;
+    img->di.cu_high_pc = buf_size;
+    img->di.fn_low_pc = buf;
+    img->di.fn_high_pc = buf_size;
+
+#ifdef DEBUG_JIT
+    /* Enable this block to be able to debug the ELF image file creation.
+       One can use readelf, objdump, or other inspection utilities.  */
+    {
+        FILE *f = fopen("/tmp/qemu.jit", "w+b");
+        if (f) {
+            if (fwrite(img, img_size, 1, f) != img_size) {
+                /* Avoid stupid unused return value warning for fwrite.  */
+            }
+            fclose(f);
+        }
+    }
+#endif
+
+    one_entry.symfile_addr = img;
+    one_entry.symfile_size = img_size;
+
+    __jit_debug_descriptor.action_flag = JIT_REGISTER_FN;
+    __jit_debug_descriptor.relevant_entry = &one_entry;
+    __jit_debug_descriptor.first_entry = &one_entry;
+    __jit_debug_register_code();
+}
+#else
+/* No support for the feature.  Provide the entry point expected by exec.c,
+   and implement the internal function we declared earlier.  */
+
+static void tcg_register_jit_int(void *buf, size_t size,
+                                 void *debug_frame, size_t debug_frame_size)
+{
+}
+
+void tcg_register_jit(void *buf, size_t buf_size)
+{
+}
+#endif /* ELF_HOST_MACHINE */
diff --git a/tcg/tcg.h b/tcg/tcg.h
index 5f6c647..a83bddd 100644
--- a/tcg/tcg.h
+++ b/tcg/tcg.h
@@ -586,3 +586,5 @@ extern uint8_t code_gen_prologue[];
 # define tcg_qemu_tb_exec(env, tb_ptr) \
     ((tcg_target_ulong (*)(void *, void *))code_gen_prologue)(env, tb_ptr)
 #endif
+
+void tcg_register_jit(void *buf, size_t buf_size);
-- 
1.7.7.6