nsrexecd core dumps

[Nils Andreas Hakansson]

Linux version 6.0.1 and 6.1.3 of nsrexecd dies with  a segmentation fault
on FreeBSD 5.1-RELEASE. The application works on FreeBSD 4.8-RELEASE.

I've tried all linux_base variants from ports with the same result.
After creating a linux_nosys that doesn't send a  SIGSYS on 5.1-RELEASE
I was able to upgrade linux_base-debian to sid (libc-2.3.1.so).

So all in all I've tried with libc-2.1.3.so libc-2.2.5.so
libc-2.3.1.so with a segmentation fault at the sameplace.

The application dies at the same address (check with gdb).
Ktrace is identical up to the SIGSEGV. On 4.8 the application
continues after linux_getrlimit by closing file descriptors (prepareing
to fork). There is no access to proc.

Assistance and suggestions are appreciated.
I think it has something to do with the elf_loader
or possibly the stacksegment but I don't know howto debug those things.

From gdb:
#0  0x080b1c05 in __lgto_xdr_rmtcallres ()
#1  0x080a6d51 in __lgto_xdr_rmtcallres ()
#2  0x080a6dc6 in __lgto_xdr_rmtcallres ()
#3  0x08055e8b in svcexec_parms ()
#4  0x2813614f in ?? ()


From ktrace (read with linux_kdump):
 20965 nsrexecd CALL  getegid
 20965 nsrexecd RET   getegid 0
 20965 nsrexecd CALL  linux_getpid
 20965 nsrexecd RET   linux_getpid 20965/0x51e5
 20965 nsrexecd CALL  geteuid
 20965 nsrexecd RET   geteuid 0
 20965 nsrexecd CALL  linux_getrlimit(0x7,0xbfbffa90)
 20965 nsrexecd RET   linux_getrlimit 0
 20965 nsrexecd PSIG  SIGSEGV SIG_DFL


By the way using nwclient602 from ports is not an option.


/Andreas Håkansson