Conference tallis::alpha_migration_tools

953.0. "how to log ioctl() calls ?" by RHETT::SHEPPARD () Tue Feb 25 1997 09:35

[Ultrix v4.4 / Unix V3.2C / migrate V1.2-1]

A customer successfully translated several parts of a 3rd-party ultrix
application (Sybase database app with an X-windows GUI) using decmigrate.
But after translating another utility he gets fatal runtime errors:

   mxr: interpreter attempted to branch to invalid mips address 0x0

We explored some simple reasons why the app might not translate --
FAQ for decmigrate claims following possible problems:
     "You cannot translate the executable with mx if:
         It contains broken code.
         It is a driver.
         It uses R4000 and R6000 extensions to the R2000 and R3000
                instruction set.
         It uses setuid or setgid.
         It uses privileged opcodes or SYSCALLs.
         It uses SYSCALLs unsupported on Digital UNIX (this is rare).
         It accesses /dev/mem or /dev/kmem.
         It depends on the ULTRIX file system hierarchy or system file
                formats where they differ from
         the Digital UNIX file system hierarchy or system file formats.
         It uses $ra to examine the MIPS code stream. "
but none of these rings a bell with him.

He found some data in decmigrate release notes/user-guide on known
problems including a list of system calls that could cause problems.

To see if he was making such calls, he set the env variable
MXR_VERBOSE to '9', expecting to print come ioctl() system call
details and check them against the list of known problems.

Instead, it logs much info on other system calls but nothing on
ioctl() calls.  It gives a count of how many system calls in each
category were made (ie: he sees 2000+ ioctl() calls were made) yet
individual calls are not recorded in the log file.

Some quick research here turned up few examples of use of the
MXR_VERBOSE variable but found a few mentions of the variable
MXR_TRACE_SYSCALLS -- I'm not sure how (or if) they might interact.

Perhaps some usage problem in setting those env variables ?

Can you suggest any other way to narrow down usage of ioctl() or other
system calls to see if they are on the list of known problems ?



By the way, I noticed the update to note #4 announcing the latest kit info.
Thanks for the update !  I may forward the updated bits to him.


                        Steve Sheppard
                        Digital Unix / Ultrix Applications Support Team
                        [email protected]