x86-64: Combine buffers of sections before we call tcc_run().

- Now we can run tcc -run tcc.c successfully, though there are some bugs.
- Remove jmp_table and got_table and use text_section for got and plt entries.
- Combine buffers in tcc_relocate().
- Use R_X86_64_64 instead of R_X86_64_32 for R_DATA_32 (now the name R_DATA_32 is inappropriate...).

tcc.c

@@ -542,15 +542,6 @@ struct TCCState {
 
     /* output file for preprocessing */
     FILE *outfile;
-
-#ifdef TCC_TARGET_X86_64
-    /* buffer to store jump tables used when the output is memory */
-    char *jmp_table;
-    int jmp_table_num;
-    /* buffer to store got tables used when the output is memory */
-    void **got_table;
-    int got_table_num;
-#endif
 };
 
 /* The current value can be: */
@@ -1315,7 +1306,12 @@ Section *new_section(TCCState *s1, const char *name, int sh_type, int sh_flags)
 
 static void free_section(Section *s)
 {
-    tcc_free(s->data);
+#ifdef TCC_TARGET_X86_64
+    /* after tcc_relocate(), some sections share the data buffer.
+       let's check if the data is allocated not to free the shared buffers */
+    if (s->data_allocated)
+#endif
+        tcc_free(s->data);
 }
 
 /* realloc section and set its content to zero */
@@ -10276,6 +10272,51 @@ int tcc_relocate(TCCState *s1)
 #ifndef TCC_TARGET_PE
     build_got_entries(s1);
 #endif
+
+#ifdef TCC_TARGET_X86_64
+    {
+        /* If the distance of two sections are longer than 32bit, our
+           program will crash. Let's combine all sections which are
+           necessary to run the program into a single buffer in the
+           text section */
+        unsigned char *p;
+        int size;
+        /* calculate the size of buffers we need */
+        size = 0;
+        for(i = 1; i < s1->nb_sections; i++) {
+            s = s1->sections[i];
+            if (s->sh_flags & SHF_ALLOC) {
+                size += s->data_offset;
+            }
+        }
+        /* double the size of the buffer for got and plt entries
+           XXX: calculate exact size for them? */
+        section_realloc(text_section, size * 2);
+        p = text_section->data + text_section->data_offset;
+        /* we will put got and plt after this offset */
+        text_section->data_offset = size;
+
+        for(i = 1; i < s1->nb_sections; i++) {
+            s = s1->sections[i];
+            if (s->sh_flags & SHF_ALLOC) {
+                if (s != text_section && s->data_offset) {
+                    if (s->sh_type == SHT_NOBITS) {
+                        /* for bss section */
+                        memset(p, 0, s->data_offset);
+                    } else {
+                        memcpy(p, s->data, s->data_offset);
+                        tcc_free(s->data);
+                    }
+                    s->data = p;
+                    /* we won't free s->data for this section */
+                    s->data_allocated = 0;
+                    p += s->data_offset;
+                }
+                s->sh_addr = (unsigned long)s->data;
+            }
+        }
+    }
+#else
     /* compute relocation address : section are relocated in place. We
        also alloc the bss space */
     for(i = 1; i < s1->nb_sections; i++) {
@@ -10286,6 +10327,7 @@ int tcc_relocate(TCCState *s1)
             s->sh_addr = (unsigned long)s->data;
         }
     }
+#endif
 
     relocate_syms(s1, 1);
 
@@ -10505,11 +10547,6 @@ TCCState *tcc_new(void)
     /* XXX: currently the PE linker is not ready to support that */
     s->leading_underscore = 1;
 #endif
-
-#ifdef TCC_TARGET_X86_64
-    s->jmp_table = NULL;
-    s->got_table = NULL;
-#endif
     return s;
 }
 
@@ -10546,10 +10583,6 @@ void tcc_delete(TCCState *s1)
     dynarray_reset(&s1->include_paths, &s1->nb_include_paths);
     dynarray_reset(&s1->sysinclude_paths, &s1->nb_sysinclude_paths);
 
-#ifdef TCC_TARGET_X86_64
-    tcc_free(s1->jmp_table);
-    tcc_free(s1->got_table);
-#endif
     tcc_free(s1);
 }
 

tccelf.c

@@ -479,22 +479,9 @@ static void relocate_syms(TCCState *s1, int do_resolve)
 
 #ifdef TCC_TARGET_X86_64
 #define JMP_TABLE_ENTRY_SIZE 14
-#define JMP_TABLE_ENTRY_MAX_NUM 4096
 static unsigned long add_jmp_table(TCCState *s1, unsigned long val)
 {
-    char *p;
-    if (!s1->jmp_table) {
-        int size = JMP_TABLE_ENTRY_SIZE * JMP_TABLE_ENTRY_MAX_NUM;
-        s1->jmp_table_num = 0;
-        s1->jmp_table = (char *)tcc_malloc(size);
-        set_pages_executable(s1->jmp_table, size);
-    }
-    if (s1->jmp_table_num == JMP_TABLE_ENTRY_MAX_NUM) {
-        error("relocating >%d symbols are not supported",
-              JMP_TABLE_ENTRY_MAX_NUM);
-    }
-    p = s1->jmp_table + s1->jmp_table_num * JMP_TABLE_ENTRY_SIZE;
-    s1->jmp_table_num++;
+    char *p = (char *)section_ptr_add(text_section, JMP_TABLE_ENTRY_SIZE);
     /* jmp *0x0(%rip) */
     p[0] = 0xff;
     p[1] = 0x25;
@@ -503,21 +490,10 @@ static unsigned long add_jmp_table(TCCState *s1, unsigned long val)
     return (unsigned long)p;
 }
 
-#define GOT_TABLE_ENTRY_MAX_NUM 4096
 static unsigned long add_got_table(TCCState *s1, unsigned long val)
 {
-    unsigned long *p;
-    if (!s1->got_table) {
-        int size = sizeof(void *) * GOT_TABLE_ENTRY_MAX_NUM;
-        s1->got_table_num = 0;
-        s1->got_table = (char *)tcc_malloc(size);
-    }
-    if (s1->got_table_num == GOT_TABLE_ENTRY_MAX_NUM) {
-        error("relocating >%d symbols are not supported",
-              GOT_TABLE_ENTRY_MAX_NUM);
-    }
-    p = s1->got_table + s1->got_table_num;
-    s1->got_table_num++;
+    unsigned long *p =
+        (unsigned long *)section_ptr_add(text_section, sizeof(void *));
     *p = val;
     return (unsigned long)p;
 }

x86_64-gen.c

@@ -91,7 +91,7 @@ int reg_classes[NB_REGS] = {
 #define EM_TCC_TARGET EM_X86_64
 
 /* relocation type for 32 bit data relocation */
-#define R_DATA_32   R_X86_64_32
+#define R_DATA_32   R_X86_64_64
 #define R_JMP_SLOT  R_X86_64_JUMP_SLOT
 #define R_COPY      R_X86_64_COPY