Skip to contentSkip to conversation

: grub: Fix sparc64 setjmp implementation, update grub_setjmp() attributes.

PATCH In review David Miller 3 participants 8 messages Mar 26, 2009 – Apr 1, 2009

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 : grub: Fix sparc64 setjmp implementation, update grub_setjmp() attributes.

The FSF has received my copyright assignment paperwork so I'm starting to feed fresh copies of my sparc64 changes.

Here, we fix the setjmp assembler implementation for sparc64 and also we fix all of the grub_setjmp declarations to have the proper "returns_twice" attribute.

For normal "setjmp" GCC recognizes this special function name and internally adds the attribute. Amongst other things this makes sure GCC does not emit a tail-call in functions that call grub_setjmp(). Since we use a special name for setjmp in GRUB we need to explicitly add the attribute.

Without this fix, GCC on sparc64 really did emit a tail-call in the nomal module where grub_setjmp() is called, and this led to myserious crashes :-)

2009-03-25 David S. Miller <[email protected]>

* normal/sparc64/setjmp.S: Fix setjmp implementation.
* include/grub/sparc64/setjmp.h (grub_jmp_buf): Update. (grub_setjmp): Mark with 'returns_twice' attribute.
* include/grub/i386/setjmp.h (grub_setjmp): Likewise
* include/grub/powerpc/setjmp.h (grub_setjmp): Likewise.
* include/grub/x86_64/setjmp.h (grub_setjmp): Likewise.

---
 include/grub/i386/setjmp.h    |    5 +++--
 include/grub/powerpc/setjmp.h |    4 ++--
 include/grub/sparc64/setjmp.h |    9 +++++----
 include/grub/x86_64/setjmp.h  |    4 ++--
 normal/sparc64/setjmp.S       |   21 +++++++++++++++------
 5 files changed, 27 insertions(+), 16 deletions(-)
include/grub/i386/setjmp.h +3 -2
index 02b0d8f..c5f94b4 100644
@@ -1,6 +1,6 @@
 /*
  *  GRUB  --  GRand Unified Bootloader
- *  Copyright (C) 2003,2006,2007  Free Software Foundation, Inc.
+ *  Copyright (C) 2003,2006,2007,2009  Free Software Foundation, Inc.
  *
  *  GRUB is free software: you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -21,7 +21,8 @@
 
 typedef unsigned long grub_jmp_buf[6];
 
-int grub_setjmp (grub_jmp_buf env) __attribute__ ((cdecl, regparm (3)));
+int grub_setjmp (grub_jmp_buf env) __attribute__ ((returns_twice, cdecl,
+						   regparm (3)));
 void grub_longjmp (grub_jmp_buf env, int val) __attribute__ ((noreturn, cdecl,
 							      regparm (3)));
 
include/grub/powerpc/setjmp.h +2 -2
index 441e538..fa16f73 100644
@@ -1,6 +1,6 @@
 /*
  *  GRUB  --  GRand Unified Bootloader
- *  Copyright (C) 2002,2004,2006,2007  Free Software Foundation, Inc.
+ *  Copyright (C) 2002,2004,2006,2007,2009  Free Software Foundation, Inc.
  *
  *  GRUB is free software: you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -21,7 +21,7 @@
 
 typedef unsigned long grub_jmp_buf[20];
 
-int grub_setjmp (grub_jmp_buf env);
+int grub_setjmp (grub_jmp_buf env) __attribute__ ((returns_twice));
 void grub_longjmp (grub_jmp_buf env, int val) __attribute__ ((noreturn));
 
 #endif /* ! GRUB_SETJMP_CPU_HEADER */
include/grub/sparc64/setjmp.h +5 -4
index 12d8e01..6096bae 100644
@@ -1,6 +1,6 @@
 /*
  *  GRUB  --  GRand Unified Bootloader
- *  Copyright (C) 2002,2004,2006,2007  Free Software Foundation, Inc.
+ *  Copyright (C) 2002,2004,2006,2007,2009  Free Software Foundation, Inc.
  *
  *  GRUB is free software: you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -19,10 +19,11 @@
 #ifndef GRUB_SETJMP_CPU_HEADER
 #define GRUB_SETJMP_CPU_HEADER	1
 
-/* FIXME (sparc64).  */
-typedef unsigned long grub_jmp_buf[20];
+#include <grub/types.h>
 
-int grub_setjmp (grub_jmp_buf env);
+typedef grub_uint64_t grub_jmp_buf[3];
+
+int grub_setjmp (grub_jmp_buf env) __attribute__ ((returns_twice));
 void grub_longjmp (grub_jmp_buf env, int val) __attribute__ ((noreturn));
 
 #endif /* ! GRUB_SETJMP_CPU_HEADER */
include/grub/x86_64/setjmp.h +2 -2
index e417f65..4ad968e 100644
@@ -1,6 +1,6 @@
 /*
  *  GRUB  --  GRand Unified Bootloader
- *  Copyright (C) 2003,2006,2007  Free Software Foundation, Inc.
+ *  Copyright (C) 2003,2006,2007,2009  Free Software Foundation, Inc.
  *
  *  GRUB is free software: you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -21,7 +21,7 @@
 
 typedef unsigned long grub_jmp_buf[8];
 
-int grub_setjmp (grub_jmp_buf env);
+int grub_setjmp (grub_jmp_buf env) __attribute__ ((returns_twice));
 void grub_longjmp (grub_jmp_buf env, int val) __attribute__ ((noreturn));
 
 #endif /* ! GRUB_SETJMP_CPU_HEADER */
normal/sparc64/setjmp.S +15 -6
index b1a9b6e..0e23ecf 100644
@@ -1,6 +1,6 @@
 /*
  *  GRUB  --  GRand Unified Bootloader
- *  Copyright (C) 2005,2007  Free Software Foundation, Inc.
+ *  Copyright (C) 2005,2007,2009  Free Software Foundation, Inc.
  *
  *  GRUB is free software: you can redistribute it and/or modify
  *  it under the terms of the GNU General Public License as published by
@@ -26,13 +26,22 @@
  * int grub_setjmp (grub_jmp_buf env)
  */
 FUNCTION(grub_setjmp)
-        ret
-         nop
+	stx	%o7, [%o0 + 0x00]
+	stx	%sp, [%o0 + 0x08]
+	stx	%fp, [%o0 + 0x10]
+	retl
+	 clr	%o0
 
 /*
  * int grub_longjmp (grub_jmp_buf env, int val)
  */
 FUNCTION(grub_longjmp)
-        ret
-         nop
-
+	ldx	[%o0 + 0x10], %g1
+	movrz	%o1, 1, %o1
+	flushw
+	ldx	[%o0 + 0x00], %o7
+	ldx	[%o0 + 0x08], %fp
+	sub	%fp, 192, %sp
+	stx	%g1, [%sp + 2047 + (14 * 8)]
+	retl
+	 restore %o1, 0, %o0
Signature · 1 line
1.6.2.1.222.g570cc
#1 lore raw
Quoted 18 lines · David Miller

The FSF has received my copyright assignment paperwork so I'm starting to feed fresh copies of my sparc64 changes.

Here, we fix the setjmp assembler implementation for sparc64 and also we fix all of the grub_setjmp declarations to have the proper "returns_twice" attribute.

For normal "setjmp" GCC recognizes this special function name and internally adds the attribute. Amongst other things this makes sure GCC does not emit a tail-call in functions that call grub_setjmp(). Since we use a special name for setjmp in GRUB we need to explicitly add the attribute.

Without this fix, GCC on sparc64 really did emit a tail-call in the nomal module where grub_setjmp() is called, and this led to myserious crashes :-)

Quoted 8 lines

2009-03-25 David S. Miller <[email protected]>

* normal/sparc64/setjmp.S: Fix setjmp implementation.
* include/grub/sparc64/setjmp.h (grub_jmp_buf): Update. (grub_setjmp): Mark with 'returns_twice' attribute.
* include/grub/i386/setjmp.h (grub_setjmp): Likewise
* include/grub/powerpc/setjmp.h (grub_setjmp): Likewise.
* include/grub/x86_64/setjmp.h (grub_setjmp): Likewise.

Hi,

Unless there's someone else in the list who is knowledgeable about sparc assembly and is going to review this part of David's patch, I'm going to assume it's fine and check it in.

This also goes for future patches sent by David. So if you will speak, do it now :-)

Signature · 5 lines
Robert Millan

  The DRM opt-in fallacy: "Your data belongs to us. We will decide when (and
  how) you may access your data; but nobody's threatening your freedom: we
  still allow you to remove your data and not access it at all."
#2 lore raw
Quoted 27 lines · On Saturday 28 March 2009 22:39:48 Robert Millan wrote:
Quoted 17 lines · David Miller

The FSF has received my copyright assignment paperwork so I'm starting to feed fresh copies of my sparc64 changes.

Here, we fix the setjmp assembler implementation for sparc64 and also we fix all of the grub_setjmp declarations to have the proper "returns_twice" attribute.

For normal "setjmp" GCC recognizes this special function name and internally adds the attribute. Amongst other things this makes sure GCC does not emit a tail-call in functions that call grub_setjmp(). Since we use a special name for setjmp in GRUB we need to explicitly add the attribute.

Without this fix, GCC on sparc64 really did emit a tail-call in the nomal module where grub_setjmp() is called, and this led to myserious crashes :-)

Quoted 8 lines

2009-03-25 David S. Miller <[email protected]>

* normal/sparc64/setjmp.S: Fix setjmp implementation.
* include/grub/sparc64/setjmp.h (grub_jmp_buf): Update. (grub_setjmp): Mark with 'returns_twice' attribute.
* include/grub/i386/setjmp.h (grub_setjmp): Likewise
* include/grub/powerpc/setjmp.h (grub_setjmp): Likewise.
* include/grub/x86_64/setjmp.h (grub_setjmp): Likewise.

Quoted 8 lines

Hi,

Unless there's someone else in the list who is knowledgeable about sparc assembly and is going to review this part of David's patch, I'm going to assume it's fine and check it in.

This also goes for future patches sent by David. So if you will speak, do it now :-)

I know a bit, but not that much. So I vote for that we just trust him. ;)

Regards,
Okuji

#3 lore raw

From: "Yoshinori K. Okuji" <[email protected]>
Date: Sat, 28 Mar 2009 23:34:25 +0900

Quoted 9 lines · On Saturday 28 March 2009 22:39:48 Robert Millan wrote:
Quoted 6 lines

Unless there's someone else in the list who is knowledgeable about sparc assembly and is going to review this part of David's patch, I'm going to assume it's fine and check it in.

This also goes for future patches sent by David. So if you will speak, do it now :-)

I know a bit, but not that much. So I vote for that we just trust him. ;)

Ping?

#4 lore raw
Quoted 8 lines · David Miller

From: "Yoshinori K. Okuji" <[email protected]>
Date: Sat, 28 Mar 2009 23:34:25 +0900

Quoted 5 lines · On Saturday 28 March 2009 22:39:48 Robert Millan wrote:
Quoted 4 lines

Unless there's someone else in the list who is knowledgeable about sparc assembly and is going to review this part of David's patch, I'm going to assume it's fine and check it in.

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

This also goes for future patches sent by David. So if you will speak, do it now :-)

I know a bit, but not that much. So I vote for that we just trust him. ;)

Ping?

Committed.

Signature · 5 lines
Robert Millan

  The DRM opt-in fallacy: "Your data belongs to us. We will decide when (and
  how) you may access your data; but nobody's threatening your freedom: we
  still allow you to remove your data and not access it at all."
#5 lore raw
Quoted 12 lines · On Wednesday 01 April 2009 22:01:11 Robert Millan wrote:
Quoted 11 lines · David Miller

From: "Yoshinori K. Okuji" <[email protected]>
Date: Sat, 28 Mar 2009 23:34:25 +0900

Quoted 8 lines · On Saturday 28 March 2009 22:39:48 Robert Millan wrote:
Quoted 6 lines

Unless there's someone else in the list who is knowledgeable about sparc assembly and is going to review this part of David's patch, I'm going to assume it's fine and check it in.

This also goes for future patches sent by David. So if you will speak, do it now :-)

Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

I know a bit, but not that much. So I vote for that we just trust him. ;)

Ping?

Committed.

David, I can give you a permission, if you tell me your account on Savannah.

Regards,
Okuji

#6 lore raw

From: "Yoshinori K. Okuji" <[email protected]>
Date: Wed, 1 Apr 2009 22:59:03 +0900

Quoted 1 line

David, I can give you a permission, if you tell me your account on Savannah.

It is simply "davem"

#7 lore raw
In reply to Robert Millan

From: Robert Millan <[email protected]>
Date: Wed, 1 Apr 2009 15:01:11 +0200

Quoted 9 lines · David Miller
Quoted 8 lines

From: "Yoshinori K. Okuji" <[email protected]>
Date: Sat, 28 Mar 2009 23:34:25 +0900

Quoted 5 lines · On Saturday 28 March 2009 22:39:48 Robert Millan wrote:
Quoted 4 lines

Unless there's someone else in the list who is knowledgeable about sparc assembly and is going to review this part of David's patch, I'm going to assume it's fine and check it in.

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines

This also goes for future patches sent by David. So if you will speak, do it now :-)

I know a bit, but not that much. So I vote for that we just trust him. ;)

Ping?

Committed.

Thank you.