Attasm

libr/arch/p/x86_nz/att2intel.c

@@ -1,168 +1,24 @@
-/* radare - LGPL - Copyright 2011-2022 - pancake */
+/* radare - LGPL - Copyright 2011-2025 - pancake */
 
 #include <r_asm.h>
 
-static int replace(int argc, const char *argv[], char *newstr) {
-	int i,j,k;
-	const struct {
-		const char *op;
-		const char *str;
-	} ops[] = {
-		{ "cmpl", "cmp 2, 1" },
-		{ "testl", "test 2, 1" },
-		{ "leal", "lea 2, 1" },
-		{ "movl", "mov 2, 1" },
-		{ "xorl", "xor 2, 1" },
-		{ "andl", "and 2, 1" },
-		{ "orl", "or 2, 1" },
-		{ "addl", "add 2, 1" },
-		{ "incl", "inc 1" },
-		{ "decl", "dec 1" },
-		{ "subl", "sub 2, 1" },
-		{ "mull", "mul 2, 1" },
-		{ "divl", "div 2, 1" },
-		{ "pushl", "push 1" },
-		{ "popl", "pop 1" },
-		{ "ret", "ret" },
-		{ NULL }
-	};
-
-	for (i = 0; ops[i].op; i++) {
-		if (!strcmp (ops[i].op, argv[0])) {
-			if (newstr) {
-				for (j = k = 0; ops[i].str[j] != '\0'; j++, k++) {
-					if (ops[i].str[j] >= '0' && ops[i].str[j] <= '9') {
-						const char *w = argv[ops[i].str[j] - '0'];
-						if (w) {
-							strcpy (newstr + k, w);
-							k += strlen (w) - 1;
-						}
-					} else {
-						newstr[k] = ops[i].str[j];
-					}
-				}
-				newstr[k]='\0';
-			}
-			return true;
-		}
-	}
-
-	/* TODO: this is slow */
-	if (newstr) {
-		newstr[0] = '\0';
-		for (i = 0; i < argc; i++) {
-			strcat (newstr, argv[i]);
-			strcat (newstr, (i == 0 || i== argc - 1)?" ":",");
-		}
-	}
-
-	return false;
-}
+// x86-specific AT&T to Intel parser plugin
+// Uses the x86-specific r_str_att2intel() from libr/util/str_att.c
+// This plugin is used as a pseudo-disassembler parser to convert AT&T output to Intel
 
 static char *parse(RAsmPluginSession *aps, const char *data) {
-	char w0[32], w1[32], w2[32], w3[32];
-	char *optr, *num;
-	int i, n;
-
-	// malloc can be slow here :?
-	char *buf = strdup (data);
-	if (!buf) {
-		return NULL;
-	}
-	r_str_trim_head (buf);
-
-	char *ptr = strchr (buf, '#');
-	if (ptr) {
-		*ptr = 0;
-		r_str_trim (buf);
-	}
-	if (*buf == '.' || buf[strlen(buf)-1] == ':') {
-		free (buf);
-		return strdup (data);
-	}
-	r_str_replace_char (buf, '$', 0);
-	r_str_replace_char (buf, '%', 0);
-	r_str_replace_char (buf, '\t', ' ');
-	r_str_replace_char (buf, '(', '[');
-	r_str_replace_char (buf, ')', ']');
-	ptr = strchr (buf, '[');
-	if (ptr) {
-		*ptr = 0;
-		num = (char*)r_str_lchr (buf, ' ');
-		if (!num) {
-			num = (char *)r_str_lchr (buf, ',');
-		}
-		if (num) {
-			n = atoi (num + 1);
-			*ptr = '[';
-			r_str_cpy (num + 1, ptr);
-			ptr = (char*)r_str_lchr (buf, ']');
-			if (n && ptr) {
-				char *rest = strdup (ptr+1);
-				size_t dist = strlen (data) + 1 - (ptr - buf);
-				snprintf (ptr, dist, "%+d]%s", n, rest);
-				free (rest);
-			}
-		} else {
-			*ptr = '[';
-		}
-	}
-	char *str = NULL;
-	if (*buf) {
-		*w0 = *w1 = *w2 = *w3 = 0;
-		ptr = strchr (buf, ' ');
-		if (!ptr) {
-			ptr = strchr (buf, '\t');
-		}
-		if (ptr) {
-			*ptr = '\0';
-			for (ptr++; *ptr == ' '; ptr++) {
-				;
-			}
-			strncpy (w0, buf, sizeof (w0) - 1);
-			strncpy (w1, ptr, sizeof (w1) - 1);
+	return r_str_att2intel (data);
+}
 
-			optr = ptr;
-			ptr = strchr (ptr, ',');
-			if (ptr) {
-				*ptr = '\0';
-				for (ptr++; *ptr == ' '; ptr++) {
-					;
-				}
-				strncpy (w1, optr, sizeof (w1)-1);
-				strncpy (w2, ptr, sizeof (w2)-1);
-				ptr = strchr (ptr, ',');
-				if (ptr) {
-					*ptr = '\0';
-					for (ptr++; *ptr == ' '; ptr++) {
-						;
-					}
-					strncpy (w2, optr, sizeof (w2)-1);
-					strncpy (w3, ptr, sizeof (w3)-1);
-				}
-			}
-		}
-		{
-			const char *wa[] = { w0, w1, w2, w3 };
-			int nw = 0;
-			for (i = 0; i < 4; i++) {
-				if (wa[i][0] != '\0') {
-					nw++;
-				}
-			}
-			str = malloc (strlen (data) + 128);
-			strcpy (str, data);
-			replace (nw, wa, str);
-		}
-	}
-	free (buf);
-	return str;
+// Public API wrapper for backwards compatibility
+R_API char *r_asm_att2intel(const char *att_str) {
+	return r_str_att2intel (att_str);
 }
 
 RAsmPlugin r_asm_plugin_att2intel = {
 	.meta = {
 		.name = "att2intel",
-		.desc = "X86 att 2 intel plugin",
+		.desc = "AT&T to Intel syntax converter",
 		.author = "pancake",
 		.license = "LGPL-3.0-only",
 	},

libr/asm/asm.c

@@ -619,6 +619,28 @@ static int r_asm_assemble_single(RAsm *a, RAnalOp *op, const char *buf) {
 	if (!b) {
 		return 0;
 	}
+	// convert AT&T syntax to Intel if needed, but only if the encoder doesn't support ATT natively
+	// The r_str_att2intel function is x86-specific, so only use it for x86 architectures
+	if (a->config->syntax == R_ARCH_SYNTAX_ATT) {
+		const char *arch = R_UNWRAP3 (a, config, arch);
+		// Get the encoder plugin to check its syntax support
+		RArchSession *as = R_UNWRAP4 (a, analb.anal, arch, session);
+		RArchPlugin *encoder_plugin = NULL;
+		if (as) {
+			// If encoder is separate, use that
+			encoder_plugin = as->encoder ? as->encoder->plugin : as->plugin;
+		}
+		// Check if encoder supports ATT syntax natively
+		bool encoder_supports_att = encoder_plugin && (encoder_plugin->encode_syntax & R_ARCH_SYNTAX_MASK_ATT);
+		// Only convert for x86 if encoder doesn't support ATT natively
+		if (!encoder_supports_att && arch && r_str_startswith (arch, "x86")) {
+			char *intel = r_str_att2intel (b);
+			if (intel) {
+				free (b);
+				b = intel;
+			}
+		}
+	}
 	r_str_case (b, false); // to-lower
 	if (a->analb.anal) {
 		ut8 buf[256] = { 0 };

libr/include/r_arch.h

@@ -151,6 +151,11 @@ typedef bool (*RArchPluginInitCallback)(RArchSession *s);
 typedef bool (*RArchPluginFiniCallback)(RArchSession *s);
 typedef bool (*RArchPluginEsilCallback)(RArchSession *s, RArchEsilAction action);
 
+// Bitmask for supported syntax modes in arch plugins
+#define R_ARCH_SYNTAX_MASK_INTEL (1 << R_ARCH_SYNTAX_INTEL)
+#define R_ARCH_SYNTAX_MASK_ATT   (1 << R_ARCH_SYNTAX_ATT)
+#define R_ARCH_SYNTAX_MASK_MASM  (1 << R_ARCH_SYNTAX_MASM)
+
 // TODO: use `const char *const` instead of `char*`
 typedef struct r_arch_plugin_t {
 	RPluginMeta meta;
@@ -161,6 +166,7 @@ typedef struct r_arch_plugin_t {
 	ut32 endian;
 	RSysBits bits;
 	RSysBits addr_bits;
+	ut32 encode_syntax; // bitmask of R_ARCH_SYNTAX_MASK_* for encoder-supported syntaxes (0 = Intel only, default)
 
 	const RArchPluginInitCallback init;
 	const RArchPluginFiniCallback fini;

libr/include/r_asm.h

@@ -137,6 +137,7 @@ R_API bool r_asm_set_big_endian(RAsm *a, bool big_endian);
 
 R_API bool r_asm_set_syntax(RAsm *a, int syntax); // This is in RArchConfig
 R_API int r_asm_syntax_from_string(const char *name);
+R_API char *r_asm_att2intel(const char *att_str); // Convert AT&T to Intel syntax
 R_API int r_asm_set_pc(RAsm *a, ut64 pc);
 R_API int r_asm_disassemble(RAsm *a, RAnalOp *op, const ut8 *buf, int len);
 R_API RAsmCode* r_asm_mdisassemble(RAsm *a, const ut8 *buf, int len);

libr/include/r_util/r_str.h

@@ -344,6 +344,8 @@ R_API void r_string_appendf(RString *a, const char * R_NONNULL fmt, ...);
 R_API char *r_str_pseudo_transform(const char **rules, const char *asm_str);
 R_API char *r_str_pseudo_subvar(char *pseudo, void *varmap);
 
+// AT&T to Intel syntax conversion
+R_API char *r_str_att2intel(const char *att_str);
 
 #ifdef __cplusplus
 }

libr/util/Makefile

@@ -15,7 +15,7 @@ OBJS+=udiff.o bdiff.o stack.o queue.o tree.o idpool.o assert.o bplist.o treemap.
 OBJS+=punycode.o pkcs7.o x509.o asn1.o asn1_str.o json_parser.o json_indent.o skiplist.o
 OBJS+=pj.o rbtree.o intervaltree.o qrcode.o vector.o str_constpool.o str_trim.o
 OBJS+=chartable.o protobuf.o graph_drawable.o axml.o sstext.o new_rbtree.o token.o
-OBJS+=rvc.o rvc_git.o rvc_rvc.o bscanf.o rprintf.o base32.o bloom.o mmap.o rxmldom.o
+OBJS+=rvc.o rvc_git.o rvc_rvc.o bscanf.o rprintf.o base32.o bloom.o mmap.o rxmldom.o str_att.o
 
 ifeq (${HAVE_GPERF},1)
 OBJS+=d/ascii.o

libr/util/meson.build

@@ -78,6 +78,7 @@ r_util_sources = [
   'str_pseudo.c',
   'str_strip.c',
   'str_trim.c',
+  'str_att.c',
   'str.c',
   'strbuf.c',
   'strpool.c',