yuan.fan wrote at Mon Mar 30 2009 22:14:52 GMT-0700 (PDT):
>
>>
>> Thanks. I was looking at the latest auto_nicdrv_0330.tar.bz2
>> and indeed it downloads the latest version of the test suite.
>> That is great. However, as Tom Chen pointed out on the other
>> thread, it does appear that the install.ksh script is running
>> stf_configure with root permissions. I am not sure why this
>> is necessary -- STF should be able to handle stf_configure
>> being run as a non-root user (it does, for every other test
>> suite). Why is it necessary to run stf_configure as root for
>> the nicdrv test suite?
>>
>> -UVR.
>
> Hi Ravindra
> In the stf_configure process,NICDRV disable/enable some system services
> and do some
> configuration on the peer machine with rsh.
> So nicdrv needs users have the root permission.
> thanks
> -yuan.fan

I'm sorry, the explanation does not seem to be valid.  I am not
convinced that stf_configure must be invoked as root.

I assume that the script which you are describing is running the
remote disable/enable on the peer machine is this:
http://cvs.opensolaris.org/source/xref/test/stcnv/usr/src/suites/net/nicdrv/configure.ksh

If so, it is already declared as an STF_ROOT_CONFIGURE script:
http://cvs.opensolaris.org/source/xref/test/stcnv/usr/src/suites/net/nicdrv/Makefile

What this says is that 'stf_configure' should be invoked as
non-root.  stf_configure will prompt the user for the root
password during execution.  After that, STF will take care
of switching to 'root' user; it will run the configure script
on the local host and peer as 'root'.  At least, this is what
STF is designed to do.

If you have tried to run stf_configure as non-root and have
experienced failures, that would be reason to report a bug
against STF.  But the mere fact that the configure script
needs to be run as root does not imply that stf_configure
should be run as root.  In general, stf_configure and
stf_execute should NEVER be run as root.

-UVR.

Reply via email to