All of lore.kernel.org
 help / color / mirror / Atom feed
* klibc sparc trouble with gcc > 4.0
@ 2007-11-05 13:55 maximilian attems
  2007-11-06  2:49 ` Oleg Verych
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: maximilian attems @ 2007-11-05 13:55 UTC (permalink / raw)
  To: sparclinux

tried below fix, that is a partial revert of klibc
6fbd8fafdcc793135988733996d72b0298afe934
see git://git.kernel.org/pub/scm/libs/klibc/klibc.git
with belows fix klibc compiles fine on sparc with newer gcc,
but segfaults:


titan:~# strace -vfF /usr/lib/klibc/bin/fstype
execve("/usr/lib/klibc/bin/fstype", ["/usr/lib/klibc/bin/fstype"],
["SHELL=/bin/bash", "TERM=xterm", "SSH_CLIENT=[myip] 39403"...,
"SSH_TTY=/dev/pts/0", "USER=root",
"LS_COLORS=no\0:fi\0:di\x01;34:l"...,
"PATH=/usr/local/sbin:/usr/local/"..., "MAIL=/var/mail/root",
"PWD=/root", "LANG=en_US.UTF-8", "PS1=\\h:\\w\\$ ", "HOME=/root",
"SHLVL=2", "LS_OPTIONS=--color=auto", "LOGNAME=root",
"SSH_CONNECTION=[myip] 3"..., "_=/usr/bin/strace", "OLDPWD=/"]) = 0
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
+++ killed by SIGSEGV +++
Process 3636 detached

thanks for looking at it.

-- 
maks



use our own clzdi on sparc instead the one of libgcc

Signed-off-by: maximilian attems <max@stro.at>

diff --git a/usr/klibc/arch/sparc64/Kbuild b/usr/klibc/arch/sparc64/Kbuild
index 2854f69..f2450a2 100644
--- a/usr/klibc/arch/sparc64/Kbuild
+++ b/usr/klibc/arch/sparc64/Kbuild
@@ -2,7 +2,7 @@
 # klibc files for sparc64
 #
 
-klib-y := pipe.o setjmp.o syscall.o sysfork.o
+klib-y := pipe.o setjmp.o syscall.o sysfork.o ../../libgcc/__clzdi2.o
 
 always  := crt0.o
 targets := crt0.o
diff --git a/usr/klibc/libgcc/__clzdi2.c b/usr/klibc/libgcc/__clzdi2.c
new file mode 100644
index 0000000..13e636c
--- /dev/null
+++ b/usr/klibc/libgcc/__clzdi2.c
@@ -0,0 +1,23 @@
+/*
+ * __clzdi2 - Returns the leading number of 0 bits in the argument
+ */
+
+#include <stdint.h>
+#include <stddef.h>
+
+uint64_t __clzdi2(uint64_t v)
+{
+	uint32_t vhi = (uint32_t) (v >> 32);
+	uint32_t vx;
+	int dp;
+
+	if (vhi) {
+		vx = vhi;
+		dp = 0;
+	} else {
+		vx = (uint32_t) v;
+		dp = 32;
+	}
+
+	return dp + __builtin_clz(vx);
+}

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: klibc sparc trouble with gcc > 4.0
  2007-11-05 13:55 klibc sparc trouble with gcc > 4.0 maximilian attems
@ 2007-11-06  2:49 ` Oleg Verych
  2007-11-06  8:30 ` [klibc] " H. Peter Anvin
  2007-11-09 17:59 ` Bernd Zeimetz
  2 siblings, 0 replies; 4+ messages in thread
From: Oleg Verych @ 2007-11-06  2:49 UTC (permalink / raw)
  To: sparclinux

= Mon, Nov 05, 2007 at 02:55:45PM +0100, maximilian attems =
[]
> titan:~# strace -vfF /usr/lib/klibc/bin/fstype
> execve("/usr/lib/klibc/bin/fstype", ["/usr/lib/klibc/bin/fstype"],
> ["SHELL=/bin/bash", "TERM=xterm", "SSH_CLIENT=[myip] 39403"...,
> "SSH_TTY=/dev/pts/0", "USER=root",
> "LS_COLORS=no\0:fi\0:di\x01;34:l"...,
> "PATH=/usr/local/sbin:/usr/local/"..., "MAIL=/var/mail/root",
> "PWD=/root", "LANG=en_US.UTF-8", "PS1=\\h:\\w\\$ ", "HOME=/root",
> "SHLVL=2", "LS_OPTIONS=--color=auto", "LOGNAME=root",
> "SSH_CONNECTION=[myip] 3"..., "_=/usr/bin/strace", "OLDPWD=/"]) = 0
> --- SIGSEGV (Segmentation fault) @ 0 (0) ---
> +++ killed by SIGSEGV +++

gdb doesn't work/help?

[]
> +++ b/usr/klibc/libgcc/__clzdi2.c
> @@ -0,0 +1,23 @@
> +/*
> + * __clzdi2 - Returns the leading number of 0 bits in the argument
> + */
> +
> +#include <stdint.h>
> +#include <stddef.h>
> +
> +uint64_t __clzdi2(uint64_t v)
> +{
> +	uint32_t vhi = (uint32_t) (v >> 32);

`v' is a data in memory loaded to register to perform shift, thus this
operation yields high 32-bits part of `v'

> +	uint32_t vx;
> +	int dp;
> +
> +	if (vhi) {
> +		vx = vhi;
> +		dp = 0;
> +	} else {
> +		vx = (uint32_t) v;

`v' is a data in memory, accessed with 32-bits cast, thus operation
yields high 32-bits part of `v'.

> +		dp = 32;
> +	}
> +
> +	return dp + __builtin_clz(vx);
> +}

While i didn't see/used any big-endian machine/asm_dump, i just
speculate, since noone replied yet.
____

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [klibc] klibc sparc trouble with gcc > 4.0
  2007-11-05 13:55 klibc sparc trouble with gcc > 4.0 maximilian attems
  2007-11-06  2:49 ` Oleg Verych
@ 2007-11-06  8:30 ` H. Peter Anvin
  2007-11-09 17:59 ` Bernd Zeimetz
  2 siblings, 0 replies; 4+ messages in thread
From: H. Peter Anvin @ 2007-11-06  8:30 UTC (permalink / raw)
  To: sparclinux

Oleg Verych wrote:
>> +	uint32_t vx;
>> +	int dp;
>> +
>> +	if (vhi) {
>> +		vx = vhi;
>> +		dp = 0;
>> +	} else {
>> +		vx = (uint32_t) v;
> 
> `v' is a data in memory, accessed with 32-bits cast, thus operation
> yields high 32-bits part of `v'.
> 

Bullsh*t.  That would be *(uint32_t *)&v.  Casting between integer types 
truncate the high bits regardless of platform endianness.

	-hpa

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: klibc sparc trouble with gcc > 4.0
  2007-11-05 13:55 klibc sparc trouble with gcc > 4.0 maximilian attems
  2007-11-06  2:49 ` Oleg Verych
  2007-11-06  8:30 ` [klibc] " H. Peter Anvin
@ 2007-11-09 17:59 ` Bernd Zeimetz
  2 siblings, 0 replies; 4+ messages in thread
From: Bernd Zeimetz @ 2007-11-09 17:59 UTC (permalink / raw)
  To: sparclinux

Oleg Verych wrote:
> = Mon, Nov 05, 2007 at 02:55:45PM +0100, maximilian attems =
> []
>> titan:~# strace -vfF /usr/lib/klibc/bin/fstype
>> execve("/usr/lib/klibc/bin/fstype", ["/usr/lib/klibc/bin/fstype"],
>> ["SHELL=/bin/bash", "TERM=xterm", "SSH_CLIENT=[myip] 39403"...,
>> "SSH_TTY=/dev/pts/0", "USER=root",
>> "LS_COLORS=no\0:fi\0:di\x01;34:l"...,
>> "PATH=/usr/local/sbin:/usr/local/"..., "MAIL=/var/mail/root",
>> "PWD=/root", "LANG=en_US.UTF-8", "PS1=\\h:\\w\\$ ", "HOME=/root",
>> "SHLVL=2", "LS_OPTIONS=--color=auto", "LOGNAME=root",
>> "SSH_CONNECTION=[myip] 3"..., "_=/usr/bin/strace", "OLDPWD=/"]) = 0
>> --- SIGSEGV (Segmentation fault) @ 0 (0) ---
>> +++ killed by SIGSEGV +++
> 
> gdb doesn't work/help?

(gdb) where
#0  0x8000faac in ?? ()
#1  0x8000facc in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

Not sure if this is a gdb problem, though - never even tried to debug klibc.
With the mentioned patch klibc compiles, but all utils just segfault,
strace is as short as seen above.

> 
> []
>> +++ b/usr/klibc/libgcc/__clzdi2.c
>> @@ -0,0 +1,23 @@
>> +/*
>> + * __clzdi2 - Returns the leading number of 0 bits in the argument
>> + */
>> +


without this patch it doesn't compile at all:


  KLIBCLD usr/klibc/libc.so
ld: sparc architecture of input file
`/usr/lib/gcc/sparc-linux-gnu/4.2.3/libgcc.a(_clzdi2.o)' is incompatible
with sparc:v9 output
ld: sparc architecture of input file
`/usr/lib/gcc/sparc-linux-gnu/4.2.3/libgcc.a(_clz.o)' is incompatible
with sparc:v9 output
/usr/lib/gcc/sparc-linux-gnu/4.2.3/libgcc.a(_clzdi2.o): In function
`__clzdi2':
(.text+0xc): undefined reference to `_GLOBAL_OFFSET_TABLE_'
/usr/lib/gcc/sparc-linux-gnu/4.2.3/libgcc.a(_clzdi2.o): In function
`__clzdi2':
(.text+0x14): undefined reference to `_GLOBAL_OFFSET_TABLE_'
make[3]: *** [usr/klibc/libc.so] Error 1
make[2]: *** [all] Error 2
make[1]: *** [klibc] Error 2
make[1]: Leaving directory `/root/klibc-1.5.7'

-- 
Bernd Zeimetz
<bernd@bzed.de>                         <http://bzed.de/>

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2007-11-09 17:59 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-11-05 13:55 klibc sparc trouble with gcc > 4.0 maximilian attems
2007-11-06  2:49 ` Oleg Verych
2007-11-06  8:30 ` [klibc] " H. Peter Anvin
2007-11-09 17:59 ` Bernd Zeimetz

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.