* Re: [PATCH] i386/x86-64: adjust unwind information register encodings
@ 2005-05-06 15:30 Jan Beulich
0 siblings, 0 replies; 7+ messages in thread
From: Jan Beulich @ 2005-05-06 15:30 UTC (permalink / raw)
To: hjl; +Cc: binutils
>>> "H. J. Lu" <hjl@lucon.org> 06.05.05 16:00:38 >>>
>On Fri, May 06, 2005 at 02:01:21PM +0200, Jan Beulich wrote:
>> The register numbers used in x86-64's .eh_frame entries neither matched the
>> ABI nor the gcc implementation. Additionally, a significant amount of
>> registers were missing in both 32- and 64-bit modes (even now there are, but
>> only those for which there don't appear to be register numbers assigned in
>> the respective ABI).
>>
>
>Have we been generating wrong unwind info up to now?
Yes, it looks like that.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] i386/x86-64: adjust unwind information register encodings
2005-05-06 16:06 Jan Beulich
@ 2005-05-06 16:14 ` H. J. Lu
0 siblings, 0 replies; 7+ messages in thread
From: H. J. Lu @ 2005-05-06 16:14 UTC (permalink / raw)
To: Jan Beulich; +Cc: binutils
On Fri, May 06, 2005 at 06:04:45PM +0200, Jan Beulich wrote:
> >>> "H. J. Lu" <hjl@lucon.org> 06.05.05 17:33:26 >>>
> >On Fri, May 06, 2005 at 05:05:48PM +0200, Jan Beulich wrote:
> >> >>> "H. J. Lu" <hjl@lucon.org> 06.05.05 16:00:38 >>>
> >> >On Fri, May 06, 2005 at 02:01:21PM +0200, Jan Beulich wrote:
> >> >> The register numbers used in x86-64's .eh_frame entries neither matched the
> >> >> ABI nor the gcc implementation. Additionally, a significant amount of
> >> >> registers were missing in both 32- and 64-bit modes (even now there are, but
> >> >> only those for which there don't appear to be register numbers assigned in
> >> >> the respective ABI).
> >> >>
> >> >
> >> >Have we been generating wrong unwind info up to now?
> >>
> >> Yes, it looks like that.
> >
> >How has it ever worked with glibc and gdb?
>
> Because the CFI directives are rarely used, and (as Daniel pointed out) only some registers were mis-numbered.
>
CFI directives are used in glibc. Have you double checked them?
H.J.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] i386/x86-64: adjust unwind information register encodings
@ 2005-05-06 16:06 Jan Beulich
2005-05-06 16:14 ` H. J. Lu
0 siblings, 1 reply; 7+ messages in thread
From: Jan Beulich @ 2005-05-06 16:06 UTC (permalink / raw)
To: hjl; +Cc: binutils
>>> "H. J. Lu" <hjl@lucon.org> 06.05.05 17:33:26 >>>
>On Fri, May 06, 2005 at 05:05:48PM +0200, Jan Beulich wrote:
>> >>> "H. J. Lu" <hjl@lucon.org> 06.05.05 16:00:38 >>>
>> >On Fri, May 06, 2005 at 02:01:21PM +0200, Jan Beulich wrote:
>> >> The register numbers used in x86-64's .eh_frame entries neither matched the
>> >> ABI nor the gcc implementation. Additionally, a significant amount of
>> >> registers were missing in both 32- and 64-bit modes (even now there are, but
>> >> only those for which there don't appear to be register numbers assigned in
>> >> the respective ABI).
>> >>
>> >
>> >Have we been generating wrong unwind info up to now?
>>
>> Yes, it looks like that.
>
>How has it ever worked with glibc and gdb?
Because the CFI directives are rarely used, and (as Daniel pointed out) only some registers were mis-numbered.
Jan
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] i386/x86-64: adjust unwind information register encodings
2005-05-06 15:37 ` H. J. Lu
@ 2005-05-06 15:42 ` Daniel Jacobowitz
0 siblings, 0 replies; 7+ messages in thread
From: Daniel Jacobowitz @ 2005-05-06 15:42 UTC (permalink / raw)
To: H. J. Lu; +Cc: Jan Beulich, binutils
On Fri, May 06, 2005 at 08:33:26AM -0700, H. J. Lu wrote:
> On Fri, May 06, 2005 at 05:05:48PM +0200, Jan Beulich wrote:
> > >>> "H. J. Lu" <hjl@lucon.org> 06.05.05 16:00:38 >>>
> > >On Fri, May 06, 2005 at 02:01:21PM +0200, Jan Beulich wrote:
> > >> The register numbers used in x86-64's .eh_frame entries neither matched the
> > >> ABI nor the gcc implementation. Additionally, a significant amount of
> > >> registers were missing in both 32- and 64-bit modes (even now there are, but
> > >> only those for which there don't appear to be register numbers assigned in
> > >> the respective ABI).
> > >>
> > >
> > >Have we been generating wrong unwind info up to now?
> >
> > Yes, it looks like that.
>
> How has it ever worked with glibc and gdb?
Take a look at the changes in Jan's patch. The errors only affect a
couple of GPRs, in 64-bit mode, when using gas's .cfi operators.
--
Daniel Jacobowitz
CodeSourcery, LLC
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] i386/x86-64: adjust unwind information register encodings
[not found] <s27b95cc.095@emea1-mh.id2.novell.com>
@ 2005-05-06 15:37 ` H. J. Lu
2005-05-06 15:42 ` Daniel Jacobowitz
0 siblings, 1 reply; 7+ messages in thread
From: H. J. Lu @ 2005-05-06 15:37 UTC (permalink / raw)
To: Jan Beulich; +Cc: binutils
On Fri, May 06, 2005 at 05:05:48PM +0200, Jan Beulich wrote:
> >>> "H. J. Lu" <hjl@lucon.org> 06.05.05 16:00:38 >>>
> >On Fri, May 06, 2005 at 02:01:21PM +0200, Jan Beulich wrote:
> >> The register numbers used in x86-64's .eh_frame entries neither matched the
> >> ABI nor the gcc implementation. Additionally, a significant amount of
> >> registers were missing in both 32- and 64-bit modes (even now there are, but
> >> only those for which there don't appear to be register numbers assigned in
> >> the respective ABI).
> >>
> >
> >Have we been generating wrong unwind info up to now?
>
> Yes, it looks like that.
How has it ever worked with glibc and gdb?
H.J.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [PATCH] i386/x86-64: adjust unwind information register encodings
2005-05-06 12:02 Jan Beulich
@ 2005-05-06 14:00 ` H. J. Lu
0 siblings, 0 replies; 7+ messages in thread
From: H. J. Lu @ 2005-05-06 14:00 UTC (permalink / raw)
To: Jan Beulich; +Cc: binutils
On Fri, May 06, 2005 at 02:01:21PM +0200, Jan Beulich wrote:
> The register numbers used in x86-64's .eh_frame entries neither matched the
> ABI nor the gcc implementation. Additionally, a significant amount of
> registers were missing in both 32- and 64-bit modes (even now there are, but
> only those for which there don't appear to be register numbers assigned in
> the respective ABI).
>
Have we been generating wrong unwind info up to now?
H.J.
^ permalink raw reply [flat|nested] 7+ messages in thread
* [PATCH] i386/x86-64: adjust unwind information register encodings
@ 2005-05-06 12:02 Jan Beulich
2005-05-06 14:00 ` H. J. Lu
0 siblings, 1 reply; 7+ messages in thread
From: Jan Beulich @ 2005-05-06 12:02 UTC (permalink / raw)
To: binutils
[-- Attachment #1: Type: text/plain, Size: 2574 bytes --]
The register numbers used in x86-64's .eh_frame entries neither matched the
ABI nor the gcc implementation. Additionally, a significant amount of
registers were missing in both 32- and 64-bit modes (even now there are, but
only those for which there don't appear to be register numbers assigned in
the respective ABI).
Built and tested on i686-pc-linux-gnu and x86_64-unknown-linux-gnu.
Jan
gas/
2005-05-06 Jan Beulich <jbeulich@novell.com>
* config/tc-i386.c (tc_x86_regname_to_dw2regnum): Correct 64-bit mode
names to match ABI. Add more registers for 32-bit and 64-bit modes.
Make name array static and const. Adjust lookup to account for NULL
entries (standing for unused register numbers).
--- /home/jbeulich/src/binutils/mainline/2005-05-06/gas/config/tc-i386.c 2005-05-06 08:24:28.000000000 +0200
+++ 2005-05-06/gas/config/tc-i386.c 2005-05-06 11:45:55.861726528 +0200
@@ -6777,21 +6796,36 @@ tc_x86_regname_to_dw2regnum (const char
{
unsigned int regnum;
unsigned int regnames_count;
- char *regnames_32[] =
+ static const char *const regnames_32[] =
{
"eax", "ecx", "edx", "ebx",
"esp", "ebp", "esi", "edi",
- "eip"
+ "eip", "eflags", NULL,
+ "st0", "st1", "st2", "st3",
+ "st4", "st5", "st6", "st7",
+ NULL, NULL,
+ "xmm0", "xmm1", "xmm2", "xmm3",
+ "xmm4", "xmm5", "xmm6", "xmm7",
+ "mm0", "mm1", "mm2", "mm3",
+ "mm4", "mm5", "mm6", "mm7"
};
- char *regnames_64[] =
+ static const char *const regnames_64[] =
{
- "rax", "rbx", "rcx", "rdx",
- "rdi", "rsi", "rbp", "rsp",
- "r8", "r9", "r10", "r11",
+ "rax", "rdx", "rcx", "rbx",
+ "rsi", "rdi", "rbp", "rsp",
+ "r8", "r9", "r10", "r11",
"r12", "r13", "r14", "r15",
- "rip"
+ "rip",
+ "xmm0", "xmm1", "xmm2", "xmm3",
+ "xmm4", "xmm5", "xmm6", "xmm7",
+ "xmm8", "xmm9", "xmm10", "xmm11",
+ "xmm12", "xmm13", "xmm14", "xmm15",
+ "st0", "st1", "st2", "st3",
+ "st4", "st5", "st6", "st7",
+ "mm0", "mm1", "mm2", "mm3",
+ "mm4", "mm5", "mm6", "mm7"
};
- char **regnames;
+ const char *const *regnames;
if (flag_code == CODE_64BIT)
{
@@ -6805,7 +6839,8 @@ tc_x86_regname_to_dw2regnum (const char
}
for (regnum = 0; regnum < regnames_count; regnum++)
- if (strcmp (regname, regnames[regnum]) == 0)
+ if (regnames[regnum] != NULL
+ && strcmp (regname, regnames[regnum]) == 0)
return regnum;
return -1;
[-- Attachment #2: binutils-mainline-x86-cfi-regnum.patch --]
[-- Type: text/plain, Size: 2495 bytes --]
The register numbers used in x86-64's .eh_frame entries neither matched the
ABI nor the gcc implementation. Additionally, a significant amount of
registers were missing in both 32- and 64-bit modes (even now there are, but
only those for which there don't appear to be register numbers assigned in
the respective ABI).
Built and tested on i686-pc-linux-gnu and x86_64-unknown-linux-gnu.
Jan
gas/
2005-05-06 Jan Beulich <jbeulich@novell.com>
* config/tc-i386.c (tc_x86_regname_to_dw2regnum): Correct 64-bit mode
names to match ABI. Add more registers for 32-bit and 64-bit modes.
Make name array static and const. Adjust lookup to account for NULL
entries (standing for unused register numbers).
--- /home/jbeulich/src/binutils/mainline/2005-05-06/gas/config/tc-i386.c 2005-05-06 08:24:28.000000000 +0200
+++ 2005-05-06/gas/config/tc-i386.c 2005-05-06 11:45:55.861726528 +0200
@@ -6777,21 +6796,36 @@ tc_x86_regname_to_dw2regnum (const char
{
unsigned int regnum;
unsigned int regnames_count;
- char *regnames_32[] =
+ static const char *const regnames_32[] =
{
"eax", "ecx", "edx", "ebx",
"esp", "ebp", "esi", "edi",
- "eip"
+ "eip", "eflags", NULL,
+ "st0", "st1", "st2", "st3",
+ "st4", "st5", "st6", "st7",
+ NULL, NULL,
+ "xmm0", "xmm1", "xmm2", "xmm3",
+ "xmm4", "xmm5", "xmm6", "xmm7",
+ "mm0", "mm1", "mm2", "mm3",
+ "mm4", "mm5", "mm6", "mm7"
};
- char *regnames_64[] =
+ static const char *const regnames_64[] =
{
- "rax", "rbx", "rcx", "rdx",
- "rdi", "rsi", "rbp", "rsp",
- "r8", "r9", "r10", "r11",
+ "rax", "rdx", "rcx", "rbx",
+ "rsi", "rdi", "rbp", "rsp",
+ "r8", "r9", "r10", "r11",
"r12", "r13", "r14", "r15",
- "rip"
+ "rip",
+ "xmm0", "xmm1", "xmm2", "xmm3",
+ "xmm4", "xmm5", "xmm6", "xmm7",
+ "xmm8", "xmm9", "xmm10", "xmm11",
+ "xmm12", "xmm13", "xmm14", "xmm15",
+ "st0", "st1", "st2", "st3",
+ "st4", "st5", "st6", "st7",
+ "mm0", "mm1", "mm2", "mm3",
+ "mm4", "mm5", "mm6", "mm7"
};
- char **regnames;
+ const char *const *regnames;
if (flag_code == CODE_64BIT)
{
@@ -6805,7 +6839,8 @@ tc_x86_regname_to_dw2regnum (const char
}
for (regnum = 0; regnum < regnames_count; regnum++)
- if (strcmp (regname, regnames[regnum]) == 0)
+ if (regnames[regnum] != NULL
+ && strcmp (regname, regnames[regnum]) == 0)
return regnum;
return -1;
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2005-05-06 16:10 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2005-05-06 15:30 [PATCH] i386/x86-64: adjust unwind information register encodings Jan Beulich
-- strict thread matches above, loose matches on Subject: below --
2005-05-06 16:06 Jan Beulich
2005-05-06 16:14 ` H. J. Lu
[not found] <s27b95cc.095@emea1-mh.id2.novell.com>
2005-05-06 15:37 ` H. J. Lu
2005-05-06 15:42 ` Daniel Jacobowitz
2005-05-06 12:02 Jan Beulich
2005-05-06 14:00 ` H. J. Lu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).