Skip to contentSkip to conversation

linkage.h: fix build breakage due to symbol prefix handling

PATCH In review James Hogan 4 participants 9 messages May 1, 2013 – May 8, 2013

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 linkage.h: fix build breakage due to symbol prefix handling

Al's commit e1b5bb6d1236d4ad2084c53aa83dde7cdf6f8eea ("consolidate cond_syscall and SYSCALL_ALIAS declarations") broke the build on blackfin and metag due to the following code:

  #ifndef SYMBOL_NAME
  #ifdef CONFIG_SYMBOL_PREFIX
  #define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
  #else
  #define SYMBOL_NAME(x) x
  #endif
  #endif
  #define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))

__stringify literally stringifies CONFIG_SYMBOL_PREFIX ##x, so you get lines like this in kernel/sys_ni.s:

.weak CONFIG_SYMBOL_PREFIXsys_quotactl
.set CONFIG_SYMBOL_PREFIXsys_quotactl,CONFIG_SYMBOL_PREFIXsys_ni_syscall

The patches in Rusty's modules-next tree such as "CONFIG_SYMBOL_PREFIX:
cleanup." clean up the whole mess around symbol prefixes, so this patch just attempts to fix the build in the mean time. The intermediate definition of SYMBOL_NAME above isn't used and is incorrect when CONFIG_SYMBOL_PREFIX is defined as CONFIG_SYMBOL_PREFIX is a quoted string literal, so define __SYMBOL_NAME directly depending on CONFIG_SYMBOL_PREFIX.

---
 include/linux/linkage.h | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
include/linux/linkage.h +2 -5
index 829d66c..de09dec 100644
@@ -15,14 +15,11 @@
 #define asmlinkage CPP_ASMLINKAGE
 #endif
 
-#ifndef SYMBOL_NAME
 #ifdef CONFIG_SYMBOL_PREFIX
-#define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
+#define __SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX __stringify(x)
 #else
-#define SYMBOL_NAME(x) x
+#define __SYMBOL_NAME(x) __stringify(x)
 #endif
-#endif
-#define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))
 
 #ifndef cond_syscall
 #define cond_syscall(x) asm(".weak\t" __SYMBOL_NAME(x) \
Signature · 1 line
1.8.1.2
#1 lore raw
Quoted 19 lines · James Hogan

Al's commit e1b5bb6d1236d4ad2084c53aa83dde7cdf6f8eea ("consolidate cond_syscall and SYSCALL_ALIAS declarations") broke the build on blackfin and metag due to the following code:

  #ifndef SYMBOL_NAME
  #ifdef CONFIG_SYMBOL_PREFIX
  #define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
  #else
  #define SYMBOL_NAME(x) x
  #endif
  #endif
  #define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))

__stringify literally stringifies CONFIG_SYMBOL_PREFIX ##x, so you get lines like this in kernel/sys_ni.s:

.weak CONFIG_SYMBOL_PREFIXsys_quotactl
.set CONFIG_SYMBOL_PREFIXsys_quotactl,CONFIG_SYMBOL_PREFIXsys_ni_syscall

Quoted 7 lines

The patches in Rusty's modules-next tree such as "CONFIG_SYMBOL_PREFIX:
cleanup." clean up the whole mess around symbol prefixes, so this patch just attempts to fix the build in the mean time. The intermediate definition of SYMBOL_NAME above isn't used and is incorrect when CONFIG_SYMBOL_PREFIX is defined as CONFIG_SYMBOL_PREFIX is a quoted string literal, so define __SYMBOL_NAME directly depending on CONFIG_SYMBOL_PREFIX.

Oh, hell... Guys, my deep apologies - what happened is that this thing has been caught in -next, rebase done here (on top of Rusty's commit) and *not* pushed to linux-next. Unfortunately, folks had been too polite and didn't keep kicking me - mistake, since the whole thing got forgotten ;-/ Sigh...

#2 lore raw
Quoted 28 lines · Al Viro
Quoted 19 lines · James Hogan

Al's commit e1b5bb6d1236d4ad2084c53aa83dde7cdf6f8eea ("consolidate cond_syscall and SYSCALL_ALIAS declarations") broke the build on blackfin and metag due to the following code:

  #ifndef SYMBOL_NAME
  #ifdef CONFIG_SYMBOL_PREFIX
  #define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
  #else
  #define SYMBOL_NAME(x) x
  #endif
  #endif
  #define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))

__stringify literally stringifies CONFIG_SYMBOL_PREFIX ##x, so you get lines like this in kernel/sys_ni.s:

.weak CONFIG_SYMBOL_PREFIXsys_quotactl
.set CONFIG_SYMBOL_PREFIXsys_quotactl,CONFIG_SYMBOL_PREFIXsys_ni_syscall

Quoted 7 lines

The patches in Rusty's modules-next tree such as "CONFIG_SYMBOL_PREFIX:
cleanup." clean up the whole mess around symbol prefixes, so this patch just attempts to fix the build in the mean time. The intermediate definition of SYMBOL_NAME above isn't used and is incorrect when CONFIG_SYMBOL_PREFIX is defined as CONFIG_SYMBOL_PREFIX is a quoted string literal, so define __SYMBOL_NAME directly depending on CONFIG_SYMBOL_PREFIX.

Quoted 5 lines

Oh, hell... Guys, my deep apologies - what happened is that this thing has been caught in -next, rebase done here (on top of Rusty's commit) and *not* pushed to linux-next. Unfortunately, folks had been too polite and didn't keep kicking me - mistake, since the whole thing got forgotten ;-/ Sigh...

No worries. Unfortunately after the tip of linux-next started working again with Rusty's commit it sort of fell off my radar.

Cheers
James

#3 lore raw
In reply to Al Viro
Quoted 3 lines · Al Viro

Oh, hell... Guys, my deep apologies - what happened is that this thing has been caught in -next, rebase done here (on top of Rusty's commit) and *not* pushed to linux-next.

Curiouser and curiouser... FWIW, what I have is

$ cat .git/refs/heads/linkage-fixed
898429c05290b8c05d035199f65a3bb034b34bb7
$ cat .git/refs/heads/remotes/next
898429c05290b8c05d035199f65a3bb034b34bb7
both being the aforementioned rebase;  I really wonder WTF had I been doing
that could've produced that...  How does git deal with that sort of
ambiguities?
$ git log remotes/next | head -3
commit 898429c05290b8c05d035199f65a3bb034b34bb7
Author: Al Viro <[email protected]>
Date:   Tue Mar 19 14:25:51 2013 -0400
$ git log remotes/next/master | head -3
commit d7c35d45ce991ae01b6008abef120173fc9c3814
Author: Stephen Rothwell <[email protected]>
Date:   Tue Apr 30 16:26:41 2013 +1000
(the latter, of course, coming from remote for linux-next mirror)
#4 lore raw
In reply to James Hogan
Quoted 19 lines · James Hogan

Al's commit e1b5bb6d1236d4ad2084c53aa83dde7cdf6f8eea ("consolidate cond_syscall and SYSCALL_ALIAS declarations") broke the build on blackfin and metag due to the following code:

  #ifndef SYMBOL_NAME
  #ifdef CONFIG_SYMBOL_PREFIX
  #define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
  #else
  #define SYMBOL_NAME(x) x
  #endif
  #endif
  #define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))

__stringify literally stringifies CONFIG_SYMBOL_PREFIX ##x, so you get lines like this in kernel/sys_ni.s:

.weak CONFIG_SYMBOL_PREFIXsys_quotactl
.set CONFIG_SYMBOL_PREFIXsys_quotactl,CONFIG_SYMBOL_PREFIXsys_ni_syscall

Quoted 7 lines

The patches in Rusty's modules-next tree such as "CONFIG_SYMBOL_PREFIX:
cleanup." clean up the whole mess around symbol prefixes, so this patch just attempts to fix the build in the mean time. The intermediate definition of SYMBOL_NAME above isn't used and is incorrect when CONFIG_SYMBOL_PREFIX is defined as CONFIG_SYMBOL_PREFIX is a quoted string literal, so define __SYMBOL_NAME directly depending on CONFIG_SYMBOL_PREFIX.

I just pushed my modules-next tree to Linus, so hopefully the nightmare is over soon!

Cheers,
Rusty.

#5 lore raw

Hi all,

Quoted 29 lines · On Thu, 02 May 2013 09:58:08 +0930 Rusty Russell <[email protected]> wrote:
Quoted 19 lines · James Hogan

Al's commit e1b5bb6d1236d4ad2084c53aa83dde7cdf6f8eea ("consolidate cond_syscall and SYSCALL_ALIAS declarations") broke the build on blackfin and metag due to the following code:

  #ifndef SYMBOL_NAME
  #ifdef CONFIG_SYMBOL_PREFIX
  #define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
  #else
  #define SYMBOL_NAME(x) x
  #endif
  #endif
  #define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))

__stringify literally stringifies CONFIG_SYMBOL_PREFIX ##x, so you get lines like this in kernel/sys_ni.s:

.weak CONFIG_SYMBOL_PREFIXsys_quotactl
.set CONFIG_SYMBOL_PREFIXsys_quotactl,CONFIG_SYMBOL_PREFIXsys_ni_syscall

Quoted 7 lines

The patches in Rusty's modules-next tree such as "CONFIG_SYMBOL_PREFIX:
cleanup." clean up the whole mess around symbol prefixes, so this patch just attempts to fix the build in the mean time. The intermediate definition of SYMBOL_NAME above isn't used and is incorrect when CONFIG_SYMBOL_PREFIX is defined as CONFIG_SYMBOL_PREFIX is a quoted string literal, so define __SYMBOL_NAME directly depending on CONFIG_SYMBOL_PREFIX.

Quoted 2 lines

I just pushed my modules-next tree to Linus, so hopefully the nightmare is over soon!

This is the merge fix patch I have been carrying in linux-next since March 14: (this probably does not quite apply any more but, hey ...)

From: Stephen Rothwell <[email protected]>
Date: Thu, 14 Mar 2013 17:14:41 +1100
Subject: [PATCH] cond_syscall and SYSCALL_ALIAS merge fixup

---
 include/linux/linkage.h | 20 ++++++--------------
 1 file changed, 6 insertions(+), 14 deletions(-)
include/linux/linkage.h +6 -14
index 829d66c..bedcddf 100644
@@ -2,7 +2,7 @@
 #define _LINUX_LINKAGE_H
 
 #include <linux/compiler.h>
-#include <linux/stringify.h>
+#include <linux/export.h>
 #include <asm/linkage.h>
 
 #ifdef __cplusplus
@@ -15,24 +15,16 @@
 #define asmlinkage CPP_ASMLINKAGE
 #endif
 
-#ifndef SYMBOL_NAME
-#ifdef CONFIG_SYMBOL_PREFIX
-#define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
-#else
-#define SYMBOL_NAME(x) x
-#endif
-#endif
-#define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))
-
 #ifndef cond_syscall
-#define cond_syscall(x) asm(".weak\t" __SYMBOL_NAME(x) \
-	"\n\t.set\t" __SYMBOL_NAME(x) "," __SYMBOL_NAME(sys_ni_syscall));
+#define cond_syscall(x) asm(".weak\t" VMLINUX_SYMBOL_STR(x) "\n\t"	\
+			    ".set\t" VMLINUX_SYMBOL_STR(x) ","	\
+			    VMLINUX_SYMBOL_STR(sys_ni_syscall))
 #endif
 
 #ifndef SYSCALL_ALIAS
 #define SYSCALL_ALIAS(alias, name)				\
-	asm ("\t.globl " __SYMBOL_NAME(alias)			\
-	"\n\t.set\t" __SYMBOL_NAME(alias) "," __SYMBOL_NAME(name))
+	asm ("\t.globl " VMLINUX_SYMBOL_STR(alias)			\
+	"\n\t.set\t" VMLINUX_SYMBOL_STR(alias) "," VMLINUX_SYMBOL_STR(name))
 #endif
 
 #define __page_aligned_data	__section(.data..page_aligned) __aligned(PAGE_SIZE)
Signature · 5 lines
1.8.1

-- 
Cheers,
Stephen Rothwell                    [email protected]
#6 lore raw

Hi Rusty,

Quoted 58 lines · On Thu, 2 May 2013 13:37:37 +1000 Stephen Rothwell <[email protected]> wrote:

This is the merge fix patch I have been carrying in linux-next since March 14: (this probably does not quite apply any more but, hey ...)

From: Stephen Rothwell <[email protected]>
Date: Thu, 14 Mar 2013 17:14:41 +1100
Subject: [PATCH] cond_syscall and SYSCALL_ALIAS merge fixup

---
 include/linux/linkage.h | 20 ++++++--------------
 1 file changed, 6 insertions(+), 14 deletions(-)
include/linux/linkage.h +6 -14
index 829d66c..bedcddf 100644
@@ -2,7 +2,7 @@
 #define _LINUX_LINKAGE_H
 
 #include <linux/compiler.h>
-#include <linux/stringify.h>
+#include <linux/export.h>
 #include <asm/linkage.h>
 
 #ifdef __cplusplus
@@ -15,24 +15,16 @@
 #define asmlinkage CPP_ASMLINKAGE
 #endif
 
-#ifndef SYMBOL_NAME
-#ifdef CONFIG_SYMBOL_PREFIX
-#define SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX ## x
-#else
-#define SYMBOL_NAME(x) x
-#endif
-#endif
-#define __SYMBOL_NAME(x) __stringify(SYMBOL_NAME(x))
-
 #ifndef cond_syscall
-#define cond_syscall(x) asm(".weak\t" __SYMBOL_NAME(x) \
-	"\n\t.set\t" __SYMBOL_NAME(x) "," __SYMBOL_NAME(sys_ni_syscall));
+#define cond_syscall(x) asm(".weak\t" VMLINUX_SYMBOL_STR(x) "\n\t"	\
+			    ".set\t" VMLINUX_SYMBOL_STR(x) ","	\
+			    VMLINUX_SYMBOL_STR(sys_ni_syscall))
 #endif
 
 #ifndef SYSCALL_ALIAS
 #define SYSCALL_ALIAS(alias, name)				\
-	asm ("\t.globl " __SYMBOL_NAME(alias)			\
-	"\n\t.set\t" __SYMBOL_NAME(alias) "," __SYMBOL_NAME(name))
+	asm ("\t.globl " VMLINUX_SYMBOL_STR(alias)			\
+	"\n\t.set\t" VMLINUX_SYMBOL_STR(alias) "," VMLINUX_SYMBOL_STR(name))
 #endif
 
 #define __page_aligned_data	__section(.data..page_aligned) __aligned(PAGE_SIZE)
Signature · 1 line
1.8.1

Version from today's merge fix. Rusty, you should show this to Linus when you ask him to merge your modules tree (assuming it looks right :-)).

From: Stephen Rothwell <[email protected]>
Date: Thu, 14 Mar 2013 17:14:41 +1100
Subject: [PATCH] cond_syscall and SYSCALL_ALIAS merge fixup

---
 include/linux/linkage.h |   17 ++++++-----------
 1 file changed, 6 insertions(+), 11 deletions(-)
include/linux/linkage.h +6 -11
index de09dec..bedcddf 100644
@@ -2,7 +2,7 @@
 #define _LINUX_LINKAGE_H
 
 #include <linux/compiler.h>
-#include <linux/stringify.h>
+#include <linux/export.h>
 #include <asm/linkage.h>
 
 #ifdef __cplusplus
@@ -15,21 +15,16 @@
 #define asmlinkage CPP_ASMLINKAGE
 #endif
 
-#ifdef CONFIG_SYMBOL_PREFIX
-#define __SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX __stringify(x)
-#else
-#define __SYMBOL_NAME(x) __stringify(x)
-#endif
-
 #ifndef cond_syscall
-#define cond_syscall(x) asm(".weak\t" __SYMBOL_NAME(x) \
-	"\n\t.set\t" __SYMBOL_NAME(x) "," __SYMBOL_NAME(sys_ni_syscall));
+#define cond_syscall(x) asm(".weak\t" VMLINUX_SYMBOL_STR(x) "\n\t"	\
+			    ".set\t" VMLINUX_SYMBOL_STR(x) ","	\
+			    VMLINUX_SYMBOL_STR(sys_ni_syscall))
 #endif
 
 #ifndef SYSCALL_ALIAS
 #define SYSCALL_ALIAS(alias, name)				\
-	asm ("\t.globl " __SYMBOL_NAME(alias)			\
-	"\n\t.set\t" __SYMBOL_NAME(alias) "," __SYMBOL_NAME(name))
+	asm ("\t.globl " VMLINUX_SYMBOL_STR(alias)			\
+	"\n\t.set\t" VMLINUX_SYMBOL_STR(alias) "," VMLINUX_SYMBOL_STR(name))
 #endif
 
 #define __page_aligned_data	__section(.data..page_aligned) __aligned(PAGE_SIZE)
Signature · 2 lines
Cheers,
Stephen Rothwell                    [email protected]
#7 lore raw
Quoted 8 lines · Stephen Rothwell

Version from today's merge fix. Rusty, you should show this to Linus when you ask him to merge your modules tree (assuming it looks right :-)).

From: Stephen Rothwell <[email protected]>
Date: Thu, 14 Mar 2013 17:14:41 +1100
Subject: [PATCH] cond_syscall and SYSCALL_ALIAS merge fixup

I ended up with something similar, except I didn't remove linux/stringify.h from the includes.

Thanks!
Rusty.

Quoted 48 lines
---
 include/linux/linkage.h |   17 ++++++-----------
 1 file changed, 6 insertions(+), 11 deletions(-)
include/linux/linkage.h +6 -11
index de09dec..bedcddf 100644
@@ -2,7 +2,7 @@
 #define _LINUX_LINKAGE_H
 
 #include <linux/compiler.h>
-#include <linux/stringify.h>
+#include <linux/export.h>
 #include <asm/linkage.h>
 
 #ifdef __cplusplus
@@ -15,21 +15,16 @@
 #define asmlinkage CPP_ASMLINKAGE
 #endif
 
-#ifdef CONFIG_SYMBOL_PREFIX
-#define __SYMBOL_NAME(x) CONFIG_SYMBOL_PREFIX __stringify(x)
-#else
-#define __SYMBOL_NAME(x) __stringify(x)
-#endif
-
 #ifndef cond_syscall
-#define cond_syscall(x) asm(".weak\t" __SYMBOL_NAME(x) \
-	"\n\t.set\t" __SYMBOL_NAME(x) "," __SYMBOL_NAME(sys_ni_syscall));
+#define cond_syscall(x) asm(".weak\t" VMLINUX_SYMBOL_STR(x) "\n\t"	\
+			    ".set\t" VMLINUX_SYMBOL_STR(x) ","	\
+			    VMLINUX_SYMBOL_STR(sys_ni_syscall))
 #endif
 
 #ifndef SYSCALL_ALIAS
 #define SYSCALL_ALIAS(alias, name)				\
-	asm ("\t.globl " __SYMBOL_NAME(alias)			\
-	"\n\t.set\t" __SYMBOL_NAME(alias) "," __SYMBOL_NAME(name))
+	asm ("\t.globl " VMLINUX_SYMBOL_STR(alias)			\
+	"\n\t.set\t" VMLINUX_SYMBOL_STR(alias) "," VMLINUX_SYMBOL_STR(name))
 #endif
 
 #define __page_aligned_data	__section(.data..page_aligned) __aligned(PAGE_SIZE)
Signature · 3 lines
-- 
Cheers,
Stephen Rothwell                    [email protected]
#8 lore raw

Hi Rusty,

Quoted 5 lines · On Mon, 06 May 2013 14:59:05 +0930 Rusty Russell <[email protected]> wrote:
Quoted 3 lines · Stephen Rothwell

Version from today's merge fix. Rusty, you should show this to Linus when you ask him to merge your modules tree (assuming it looks right :-)).

Quoted 8 lines
Quoted 5 lines

From: Stephen Rothwell <[email protected]>
Date: Thu, 14 Mar 2013 17:14:41 +1100
Subject: [PATCH] cond_syscall and SYSCALL_ALIAS merge fixup

I ended up with something similar, except I didn't remove linux/stringify.h from the includes.

Well, the stringify include was only added in Al's patch during this merge window, so it should be perfectly safe to remove again (now that all the "stringify" calls have been removed again).

Signature · 2 lines
Cheers,
Stephen Rothwell                    [email protected]