Skip to contentSkip to conversation

symbols: don't omit "end" symbols upon mixed code / data aliases

PATCH In review Jan Beulich 2 participants 2 messages Jan 21

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 symbols: don't omit "end" symbols upon mixed code / data aliases

At the example of _sinitext - that symbol has four aliases on x86:
__init_begin, __2M_init_start, __2M_rodata_end, and whatever the first function in .init.text. With the GNU toolchain all of them are marked 'T' or 't'. With Clang/LLVM, however, some are marked 'r'. Since, to save space, we want fake "end" symbols only for text, right now want_symbol_end() has a respective check. That is getting in the way, however, when the final of those symbols is 'r'. Remove the check and instead zap the size for anything that is non-code.

Fixes 6eede548df21 ('symbols: avoid emitting "end" symbols for data items')

---
Roger, just fyi that I think that this change would mask the other issue that you reported, without actually adressing the underlying problem. Hence both changes will be wanted.

xen/tools/symbols.c +4 -1
@@ -164,6 +164,10 @@ static int read_symbol(FILE *in, struct
 	else if (str[0] == '$')
 		goto skip_tail;
 
+	/* We want fake "end" symbols only for text / code. */
+	if (toupper((uint8_t)stype) != 'T')
+		s->size = 0;
+
 	/* include the type field in the symbol name, so that it gets
 	 * compressed together */
 	s->len = strlen(str) + 1;
@@ -312,7 +316,6 @@ static int compare_name_orig(const void
 static bool want_symbol_end(unsigned int idx)
 {
 	return table[idx].size &&
-	       toupper(table[idx].type) == 'T' &&
 	       (idx + 1 == table_cnt ||
 	        table[idx].addr + table[idx].size < table[idx + 1].addr);
 }
Acked-by on v1 1/1
#1 lore raw
Quoted 11 lines · Jan Beulich

At the example of _sinitext - that symbol has four aliases on x86:
__init_begin, __2M_init_start, __2M_rodata_end, and whatever the first function in .init.text. With the GNU toolchain all of them are marked 'T' or 't'. With Clang/LLVM, however, some are marked 'r'. Since, to save space, we want fake "end" symbols only for text, right now want_symbol_end() has a respective check. That is getting in the way, however, when the final of those symbols is 'r'. Remove the check and instead zap the size for anything that is non-code.

Fixes 6eede548df21 ('symbols: avoid emitting "end" symbols for data items')
Quoted 4 lines

---
Roger, just fyi that I think that this change would mask the other issue that you reported, without actually adressing the underlying problem. Hence both changes will be wanted.

Yes, it does indeed mask the other issue.

Thanks, Roger.