Re: [PATCH] fix vsyscall unwind information

Re: [PATCH] fix vsyscall unwind information

[David Mosberger-Tang]

>>>>> On Sun, 04 May 2003 08:30:10 +0200, Richard Henderson <rth@twiddle.net> said:

  Richard> Also adds unwind info for the sigreturn entry points.  This
  Richard> can be used instead of special-case hacks currently in
  Richard> libgcc and gdb, and by extension allows the kernel to
  Richard> change these entry points without breaking userland.

Is there a marker or some other way to identify the sigreturn as such?
If not, could one be added?

	--david

Re: [PATCH] fix vsyscall unwind information

[Richard Henderson]

On Sun, May 04, 2003 at 11:49:31PM -0700, David Mosberger-Tang wrote:
> Is there a marker or some other way to identify the sigreturn as such?

No.

> If not, could one be added?

Why?  Certainly it isn't needed for x86.


r~

Re: [PATCH] fix vsyscall unwind information

[David Mosberger-Tang]

>>>>> On Mon, 5 May 2003 00:42:48 -0700, Richard Henderson <rth@twiddle.net> said:

  >> If not, could one be added?

  Richard> Why?  Certainly it isn't needed for x86.

Certain applications (such as debuggers) want to know.  Sure, you can
do symbol matching (if you have the symbol table) or code-reading
(assuming you know the exact sigreturn sequence), but having a marker
would be more reliable and faster.

	--david

Re: [PATCH] fix vsyscall unwind information

[Richard Henderson]

On Mon, May 05, 2003 at 08:05:58AM -0700, David Mosberger-Tang wrote:
>   Richard> Why?  Certainly it isn't needed for x86.
> 
> Certain applications (such as debuggers) want to know.  Sure, you can
> do symbol matching (if you have the symbol table) or code-reading
> (assuming you know the exact sigreturn sequence), but having a marker
> would be more reliable and faster.

Eh.  The whole point was to *eliminate* the special cases.

If the debugger does nothing special now, it'll see the symbol
from the VDSO in the backtrace and print __kernel_sigreturn.
Isn't this sufficient for the user to recognize what's going on?
Does it really need to print <signal frame>?



r~

Re: [PATCH] fix vsyscall unwind information

[David Mosberger-Tang]

>>>>> On Mon, 5 May 2003 09:34:44 -0700, Richard Henderson <rth@twiddle.net> said:

  Richard> On Mon, May 05, 2003 at 08:05:58AM -0700, David
  Richard> Mosberger-Tang wrote: Why?  Certainly it isn't needed for
  Richard> x86.
  >>  Certain applications (such as debuggers) want to know.  Sure,
  >> you can do symbol matching (if you have the symbol table) or
  >> code-reading (assuming you know the exact sigreturn sequence),
  >> but having a marker would be more reliable and faster.

  Richard> Eh.  The whole point was to *eliminate* the special cases.

Signal handlers have special significance on UNIX-like operating
systems.  An application might want to know when it's in a signal
frame.  If it's a problem to do this with DWARF2 info, fine.  If not,
please consider adding a marker (my current version of libunwind for
x86 does code-reading to detect signal-frames; not pretty, but it
works reasonable well in practice; it would be nicer to get rid of the
code-reading though, in the future).

	--david

Re: [PATCH] fix vsyscall unwind information

[Mark Kettenis]

Richard Henderson <rth@twiddle.net> writes:

> On Mon, May 05, 2003 at 08:05:58AM -0700, David Mosberger-Tang wrote:
> >   Richard> Why?  Certainly it isn't needed for x86.
> > 
> > Certain applications (such as debuggers) want to know.  Sure, you can
> > do symbol matching (if you have the symbol table) or code-reading
> > (assuming you know the exact sigreturn sequence), but having a marker
> > would be more reliable and faster.
> 
> Eh.  The whole point was to *eliminate* the special cases.
> 
> If the debugger does nothing special now, it'll see the symbol
> from the VDSO in the backtrace and print __kernel_sigreturn.
> Isn't this sufficient for the user to recognize what's going on?
> Does it really need to print <signal frame>?

Unfortunately, GDB needs to be able to recognize signal trampolines in
order to be able to single step correctly when a signal arrives.  At
least on some platforms.  Could be that the code-path in question
isn't used for Linux/i386, but I vaguely remember it does.

Anyway, signal trampolines could be marked with a special augmentation
in their CIE.

Mark

Re: [PATCH] fix vsyscall unwind information

[Richard Henderson]

On Tue, May 06, 2003 at 01:10:37AM +0200, Mark Kettenis wrote:
> Unfortunately, GDB needs to be able to recognize signal trampolines in
> order to be able to single step correctly when a signal arrives.

If it actually used the dwarf2 unwind information as written,
I wouldn't expect this to be true.

> Anyway, signal trampolines could be marked with a special augmentation
> in their CIE.

I'd prefer not, if at all possible.


r~

Re: [PATCH] fix vsyscall unwind information

[David Mosberger-Tang]

>>>>> On Mon, 5 May 2003 17:21:00 -0700, Richard Henderson <rth@twiddle.net> said:

  >> Anyway, signal trampolines could be marked with a special
  >> augmentation in their CIE.

  Richard> I'd prefer not, if at all possible.

Why?  I bet you'll live to regret it if you don't add it now. ;-)

	--david

Re: [PATCH] fix vsyscall unwind information

[Ulrich Drepper]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Richard Henderson wrote:

> Also adds unwind info for the sigreturn entry points.

Sweet.


> This can
> be used instead of special-case hacks currently in libgcc and 
> gdb, and by extension allows the kernel to change these entry
> points without breaking userland.

Exactly.  This means we can get rid of the int $0x80.  I.e., move the
sigreturn code in the two .S files for sysenter and int80.  Reducing the
cost of signals is always good.

I'm not sure how this use of sigreturn has to look like.  Linus, maybe
you should do it yourself.


- -- 
- --------------.                        ,-.            444 Castro Street
Ulrich Drepper \    ,-----------------'   \ Mountain View, CA 94041 USA
Red Hat         `--' drepper at redhat.com `---------------------------
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+tMKL2ijCOnn/RHQRAoDbAJ4nnP3Hcb8xAeqjRHxgq0YZZmcs0wCdE4BE
Obuu/wRdq70aooscs1JOhto=
=5Sx4
-----END PGP SIGNATURE-----

[PATCH] fix vsyscall unwind information

[Richard Henderson]

Re-adds the %ebp save/restore correction I'd given Uli when the 
unwind information went into sysenter.c, which somehow got lost
with Roland's reorg.

Also adds unwind info for the sigreturn entry points.  This can
be used instead of special-case hacks currently in libgcc and 
gdb, and by extension allows the kernel to change these entry
points without breaking userland.

Tested with Roland's patch to make the VDSO available in glibc,
plus a set of fixes to libgcc, since location expressions hadn't
ever been tested before in this context.

Linus, please pull from 

	bk://are.twiddle.net/unwind-2.5


r~



 arch/i386/Makefile                    |    9 ++
 arch/i386/kernel/asm-offsets.c        |   31 +++++++++
 arch/i386/kernel/sigframe.h           |   21 ++++++
 arch/i386/kernel/signal.c             |   23 -------
 arch/i386/kernel/vsyscall-sigreturn.S |  110 +++++++++++++++++++++++++++++++++-
 arch/i386/kernel/vsyscall-sysenter.S  |    2 
 6 files changed, 171 insertions(+), 25 deletions(-)

through these ChangeSets:

<rth@eeyore.twiddle.net> (03/05/03 1.1209)
   Fix unwind info for sysenter entry point.
   Add unwind info for sigreturn entry points.

# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher.
# This patch includes the following deltas:
#	           ChangeSet	1.1208  -> 1.1209 
#	  arch/i386/Makefile	1.49    -> 1.50   
#	arch/i386/kernel/vsyscall-sysenter.S	1.1     -> 1.2    
#	arch/i386/kernel/signal.c	1.30    -> 1.31   
#	arch/i386/kernel/vsyscall-sigreturn.S	1.1     -> 1.2    
#	               (new)	        -> 1.1     arch/i386/kernel/sigframe.h
#	               (new)	        -> 1.1     arch/i386/kernel/asm-offsets.c
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 03/05/03	rth@eeyore.twiddle.net	1.1209
# Fix unwind info for sysenter entry point.
# Add unwind info for sigreturn entry points.
# --------------------------------------------
#
diff -Nru a/arch/i386/Makefile b/arch/i386/Makefile
--- a/arch/i386/Makefile	Sat May  3 23:03:13 2003
+++ b/arch/i386/Makefile	Sat May  3 23:03:13 2003
@@ -114,6 +114,15 @@
 install fdimage fdimage144 fdimage288: vmlinux
 	$(Q)$(MAKE) $(build)=$(boot) BOOTIMAGE=$(BOOTIMAGE) $@
 
+prepare: include/asm-$(ARCH)/asm_offsets.h
+CLEAN_FILES += include/asm-$(ARCH)/asm_offsets.h
+
+arch/$(ARCH)/kernel/asm-offsets.s: include/asm include/linux/version.h \
+				   include/config/MARKER
+
+include/asm-$(ARCH)/asm_offsets.h: arch/$(ARCH)/kernel/asm-offsets.s
+	$(call filechk,gen-asm-offsets)
+
 archclean:
 	$(Q)$(MAKE) $(clean)=arch/i386/boot
 
diff -Nru a/arch/i386/kernel/asm-offsets.c b/arch/i386/kernel/asm-offsets.c
--- /dev/null	Wed Dec 31 16:00:00 1969
+++ b/arch/i386/kernel/asm-offsets.c	Sat May  3 23:03:13 2003
@@ -0,0 +1,31 @@
+/*
+ * Generate definitions needed by assembly language modules.
+ * This code generates raw asm output which is post-processed
+ * to extract and format the required data.
+ */
+
+#include <linux/signal.h>
+#include <asm/ucontext.h>
+#include "sigframe.h"
+
+#define DEFINE(sym, val) \
+        asm volatile("\n->" #sym " %0 " #val : : "i" (val))
+
+#define BLANK() asm volatile("\n->" : : )
+
+void foo(void)
+{
+	DEFINE(SIGCONTEXT_eax, offsetof (struct sigcontext, eax));
+	DEFINE(SIGCONTEXT_ebx, offsetof (struct sigcontext, ebx));
+	DEFINE(SIGCONTEXT_ecx, offsetof (struct sigcontext, ecx));
+	DEFINE(SIGCONTEXT_edx, offsetof (struct sigcontext, edx));
+	DEFINE(SIGCONTEXT_esi, offsetof (struct sigcontext, esi));
+	DEFINE(SIGCONTEXT_edi, offsetof (struct sigcontext, edi));
+	DEFINE(SIGCONTEXT_ebp, offsetof (struct sigcontext, ebp));
+	DEFINE(SIGCONTEXT_esp, offsetof (struct sigcontext, esp));
+	DEFINE(SIGCONTEXT_eip, offsetof (struct sigcontext, eip));
+	BLANK();
+
+	DEFINE(RT_SIGFRAME_sigcontext,
+	       offsetof (struct rt_sigframe, uc.uc_mcontext));
+}
diff -Nru a/arch/i386/kernel/sigframe.h b/arch/i386/kernel/sigframe.h
--- /dev/null	Wed Dec 31 16:00:00 1969
+++ b/arch/i386/kernel/sigframe.h	Sat May  3 23:03:13 2003
@@ -0,0 +1,21 @@
+struct sigframe
+{
+	char *pretcode;
+	int sig;
+	struct sigcontext sc;
+	struct _fpstate fpstate;
+	unsigned long extramask[_NSIG_WORDS-1];
+	char retcode[8];
+};
+
+struct rt_sigframe
+{
+	char *pretcode;
+	int sig;
+	struct siginfo *pinfo;
+	void *puc;
+	struct siginfo info;
+	struct ucontext uc;
+	struct _fpstate fpstate;
+	char retcode[8];
+};
diff -Nru a/arch/i386/kernel/signal.c b/arch/i386/kernel/signal.c
--- a/arch/i386/kernel/signal.c	Sat May  3 23:03:13 2003
+++ b/arch/i386/kernel/signal.c	Sat May  3 23:03:13 2003
@@ -23,6 +23,7 @@
 #include <asm/ucontext.h>
 #include <asm/uaccess.h>
 #include <asm/i387.h>
+#include "sigframe.h"
 
 #define DEBUG_SIG 0
 
@@ -125,28 +126,6 @@
 /*
  * Do a signal return; undo the signal stack.
  */
-
-struct sigframe
-{
-	char *pretcode;
-	int sig;
-	struct sigcontext sc;
-	struct _fpstate fpstate;
-	unsigned long extramask[_NSIG_WORDS-1];
-	char retcode[8];
-};
-
-struct rt_sigframe
-{
-	char *pretcode;
-	int sig;
-	struct siginfo *pinfo;
-	void *puc;
-	struct siginfo info;
-	struct ucontext uc;
-	struct _fpstate fpstate;
-	char retcode[8];
-};
 
 static int
 restore_sigcontext(struct pt_regs *regs, struct sigcontext __user *sc, int *peax)
diff -Nru a/arch/i386/kernel/vsyscall-sigreturn.S b/arch/i386/kernel/vsyscall-sigreturn.S
--- a/arch/i386/kernel/vsyscall-sigreturn.S	Sat May  3 23:03:13 2003
+++ b/arch/i386/kernel/vsyscall-sigreturn.S	Sat May  3 23:03:13 2003
@@ -7,6 +7,7 @@
  */
 
 #include <asm/unistd.h>
+#include <asm/asm_offsets.h>
 
 
 /* XXX
@@ -18,21 +19,124 @@
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
 __kernel_sigreturn:
-.LSTART_kernel_sigreturn:
+.LSTART_sigreturn:
 	popl %eax		/* XXX does this mean it needs unwind info? */
 	movl $__NR_sigreturn, %eax
 	int $0x80
 .LEND_sigreturn:
 	.size __kernel_sigreturn,.-.LSTART_sigreturn
 
-	.text
 	.balign 32
 	.globl __kernel_rt_sigreturn
 	.type __kernel_rt_sigreturn,@function
 __kernel_rt_sigreturn:
-.LSTART_kernel_rt_sigreturn:
+.LSTART_rt_sigreturn:
 	movl $__NR_rt_sigreturn, %eax
 	int $0x80
 .LEND_rt_sigreturn:
 	.size __kernel_rt_sigreturn,.-.LSTART_rt_sigreturn
+	.previous
+
+	.section .eh_frame,"a",@progbits
+.LSTARTFRAMEDLSI1:
+	.long .LENDCIEDLSI1-.LSTARTCIEDLSI1
+.LSTARTCIEDLSI1:
+	.long 0			/* CIE ID */
+	.byte 1			/* Version number */
+	.string "zR"		/* NUL-terminated augmentation string */
+	.uleb128 1		/* Code alignment factor */
+	.sleb128 -4		/* Data alignment factor */
+	.byte 8			/* Return address register column */
+	.uleb128 1		/* Augmentation value length */
+	.byte 0x1b		/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
+	.byte 0			/* DW_CFA_nop */
+	.align 4
+.LENDCIEDLSI1:
+	.long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */
+.LSTARTFDEDLSI1:
+	.long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */
+	/* HACK: The dwarf2 unwind routines will subtract 1 from the
+	   return address to get an address in the middle of the
+	   presumed call instruction.  Since we didn't get here via
+	   a call, we need to include the nop before the real start
+	   to make up for it.  */
+	.long .LSTART_sigreturn-1-.	/* PC-relative start address */
+	.long .LEND_sigreturn-.LSTART_sigreturn+1
+	.uleb128 0			/* Augmentation */
+	/* What follows are the instructions for the table generation.
+	   We record the locations of each register saved.  This is
+	   complicated by the fact that the "CFA" is always assumed to
+	   be the value of the stack pointer in the caller.  This means
+	   that we must define the CFA of this body of code to be the
+	   saved value of the stack pointer in the sigcontext.  Which
+	   also means that there is no fixed relation to the other 
+	   saved registers, which means that we must use DW_CFA_expression
+	   to compute their addresses.  It also means that when we 
+	   adjust the stack with the popl, we have to do it all over again.  */
+
+#define do_cfa_expr(offset)						\
+	.byte 0x0f;			/* DW_CFA_def_cfa_expression */	\
+	.uleb128 1f-0f;			/*   length */			\
+0:	.byte 0x74;			/*     DW_OP_breg4 */		\
+	.sleb128 offset;		/*      offset */		\
+	.byte 0x06;			/*     DW_OP_deref */		\
+1:
+
+#define do_expr(regno, offset)						\
+	.byte 0x10;			/* DW_CFA_expression */		\
+	.uleb128 regno;			/*   regno */			\
+	.uleb128 1f-0f;			/*   length */			\
+0:	.byte 0x74;			/*     DW_OP_breg4 */		\
+	.sleb128 offset;		/*       offset */		\
+1:
+
+	do_cfa_expr(SIGCONTEXT_esp+4)
+	do_expr(0, SIGCONTEXT_eax+4)
+	do_expr(1, SIGCONTEXT_ecx+4)
+	do_expr(2, SIGCONTEXT_edx+4)
+	do_expr(3, SIGCONTEXT_ebx+4)
+	do_expr(5, SIGCONTEXT_ebp+4)
+	do_expr(6, SIGCONTEXT_esi+4)
+	do_expr(7, SIGCONTEXT_edi+4)
+	do_expr(8, SIGCONTEXT_eip+4)
+
+	.byte 0x42	/* DW_CFA_advance_loc 2 -- nop; popl eax. */
+
+	do_cfa_expr(SIGCONTEXT_esp)
+	do_expr(0, SIGCONTEXT_eax)
+	do_expr(1, SIGCONTEXT_ecx)
+	do_expr(2, SIGCONTEXT_edx)
+	do_expr(3, SIGCONTEXT_ebx)
+	do_expr(5, SIGCONTEXT_ebp)
+	do_expr(6, SIGCONTEXT_esi)
+	do_expr(7, SIGCONTEXT_edi)
+	do_expr(8, SIGCONTEXT_eip)
+
+	.align 4
+.LENDFDEDLSI1:
+
+	.long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */
+.LSTARTFDEDLSI2:
+	.long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */
+	/* HACK: See above wrt unwind library assumptions.  */
+	.long .LSTART_rt_sigreturn-1-.	/* PC-relative start address */
+	.long .LEND_rt_sigreturn-.LSTART_rt_sigreturn+1
+	.uleb128 0			/* Augmentation */
+	/* What follows are the instructions for the table generation.
+	   We record the locations of each register saved.  This is
+	   slightly less complicated than the above, since we don't
+	   modify the stack pointer in the process.  */
+
+	do_cfa_expr(RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_esp)
+	do_expr(0, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_eax)
+	do_expr(1, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_ecx)
+	do_expr(2, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_edx)
+	do_expr(3, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_ebx)
+	do_expr(5, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_ebp)
+	do_expr(6, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_esi)
+	do_expr(7, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_edi)
+	do_expr(8, RT_SIGFRAME_sigcontext-4 + SIGCONTEXT_eip)
+
+	.align 4
+.LENDFDEDLSI2:
 	.previous
diff -Nru a/arch/i386/kernel/vsyscall-sysenter.S b/arch/i386/kernel/vsyscall-sysenter.S
--- a/arch/i386/kernel/vsyscall-sysenter.S	Sat May  3 23:03:13 2003
+++ b/arch/i386/kernel/vsyscall-sysenter.S	Sat May  3 23:03:13 2003
@@ -74,11 +74,13 @@
 	.long .Lenter_kernel-.Lpush_edx
 	.byte 0x0e		/* DW_CFA_def_cfa_offset */
 	.byte 0x10		/* RA at offset 16 now */
+	.byte 0x85, 0x04	/* DW_CFA_offset %ebp -16 */
 	/* Finally the epilogue.  */
 	.byte 0x04		/* DW_CFA_advance_loc4 */
 	.long .Lpop_ebp-.Lenter_kernel
 	.byte 0x0e		/* DW_CFA_def_cfa_offset */
 	.byte 0x12		/* RA at offset 12 now */
+	.byte 0xc5		/* DW_CFA_restore %ebp */
 	.byte 0x04		/* DW_CFA_advance_loc4 */
 	.long .Lpop_edx-.Lpop_ebp
 	.byte 0x0e		/* DW_CFA_def_cfa_offset */