Skip to contentSkip to conversation

lto, watchdog/hpwdt.c: Make assembler label global

PATCH In review Andi Kleen 2 participants 3 messages Aug 5, 2013 – Aug 7, 2013

Subsystems (2)
lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

PATCH 1/1 lto, watchdog/hpwdt.c: Make assembler label global

From: Andi Kleen <[email protected]>

We cannot assume that the inline assembler code always ends up in the same file as the original C file. So make any assembler labels that are called with "extern" by C global

---
 drivers/watchdog/hpwdt.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/watchdog/hpwdt.c +4 -2
index de7e4f4..5be5e3d 100644
@@ -162,7 +162,8 @@ extern asmlinkage void asminline_call(struct cmn_registers *pi86Regs,
 #define HPWDT_ARCH	32
 
 asm(".text                          \n\t"
-    ".align 4                       \n"
+    ".align 4                       \n\t"
+    ".globl asminline_call	    \n"
     "asminline_call:                \n\t"
     "pushl       %ebp               \n\t"
     "movl        %esp, %ebp         \n\t"
@@ -352,7 +353,8 @@ static int detect_cru_service(void)
 #define HPWDT_ARCH	64
 
 asm(".text                      \n\t"
-    ".align 4                   \n"
+    ".align 4                   \n\t"
+    ".globl asminline_call	\n"
     "asminline_call:            \n\t"
     "pushq      %rbp            \n\t"
     "movq       %rsp, %rbp      \n\t"
Signature · 1 line
1.8.3.1
#1 lore raw
Quoted 2 lines · On Mon, 5 Aug 2013 15:16:39 -0700 Andi Kleen <[email protected]> wrote:

We cannot assume that the inline assembler code always ends up in the same file as the original C file.

um, why not? You put an asm(".text\n...") into a .c file and you expect that assembly code to be emitted into the .s file then assembled into the .o file.

Obviously something somewhere is doing something unexpected. Much more detail needed, please.

Also... what's "lto"? Link Time Optimization? It's unclear how this fits into that. Please provide that additional background info in these patches so that others know what's going on and so that others (such as me) can help people avoid making these mistakes in the future.

author
#2 lore raw
Quoted 11 lines · Andrew Morton
Quoted 2 lines · On Mon, 5 Aug 2013 15:16:39 -0700 Andi Kleen <[email protected]> wrote:

We cannot assume that the inline assembler code always ends up in the same file as the original C file.

um, why not? You put an asm(".text\n...") into a .c file and you expect that assembly code to be emitted into the .s file then assembled into the .o file.

Obviously something somewhere is doing something unexpected. Much more detail needed, please.

Sorry this is an extract from my larger LTO patchkit. I didn't fully adapt the description.

With LTO (yes Link Time Optimization) the compiler repartitions the individual definitions into new files, so yes the assembler statement and the rest of the file can end up separated.

Quoted 4 lines

Also... what's "lto"? Link Time Optimization? It's unclear how this fits into that. Please provide that additional background info in these patches so that others know what's going on and so that others (such as me) can help people avoid making these mistakes in the future.

I catch them by fixing compiler errors in my whole tree (well not all, only the ones that happen in my setup)

https://github.com/andikleen/linux-misc/commits/lto-3.11

There's much more there, I'm currently just trying extract some patches and make my pile smaller.

-Andi