[Qemu-devel] SH4 issue and fstat64

[Qemu-devel] SH4 issue and fstat64

[Laurent Desnogues]

Hello,

after spending two hours to find and fix a simple hello world
program failed on sh4, I found the issue had already been found:

http://www.nabble.com/sh4-linux-user-emulation-on-an-x86_64-system-td19998891.html

I can confirm that the packing of the structure proposed by Takashi Yoshii
and Michael Trimarchi fixes the problem on my x86_64 system.  I also
checked it does not break i386 qemu on my machine.

If someone needs a reproducer feel free to contact me.

This could have been frustrating, but now I know a little bit of sh4,
so that was not useless :-)


Laurent

[Qemu-devel] SH4 issue and fstat64

[Laurent Desnogues]

Hello,

after spending two hours to find and fix a simple hello world
program failed on sh4, I found the issue had already been found:

http://www.nabble.com/sh4-linux-user-emulation-on-an-x86_64-system-td19998891.html

I can confirm that the packing of the structure proposed by Takashi Yoshii
and Michael Trimarchi fixes the problem on my x86_64 system.  I also
checked it does not break i386 qemu on my machine.

If someone needs a reproducer feel free to contact me.

This could have been frustrating, but now I know a little bit of sh4,
so that was not useless :-)


Laurent

[Qemu-devel] Re: SH4 issue and fstat64

[Måns Rullgård]

"Laurent Desnogues" <laurent.desnogues@gmail.com> writes:

> Hello,
>
> after spending two hours to find and fix a simple hello world
> program failed on sh4, I found the issue had already been found:
>
> http://www.nabble.com/sh4-linux-user-emulation-on-an-x86_64-system-td19998891.html
>
> I can confirm that the packing of the structure proposed by Takashi Yoshii
> and Michael Trimarchi fixes the problem on my x86_64 system.  I also
> checked it does not break i386 qemu on my machine.

I can confirm the fix as well.  For the record, this is the patch I
used:

diff --git a/linux-user/syscall_defs.h b/linux-user/syscall_defs.h
index 86c9d82..ba9ef8d 100644
--- a/linux-user/syscall_defs.h
+++ b/linux-user/syscall_defs.h
@@ -1475,7 +1475,7 @@ struct target_stat {
 /* This matches struct stat64 in glibc2.1, hence the absolutely
  * insane amounts of padding around dev_t's.
  */
-struct target_stat64 {
+struct __attribute__((__packed__)) target_stat64 {
        unsigned long long      st_dev;
        unsigned char   __pad0[4];
 


-- 
Måns Rullgård
mans@mansr.com