Richard Peurifoy wrote:

I was curious, so I did a GTF trace and see I/O to the PDS to count the number
of directories. SVC 0 is used and shows a DDNAME of PPPPPPPP, a DCB addr,
and a DEB addr. There is no OPEN, CLOSE, DYNALLOC, or DEBCHK SVC.
There may be branch entries internally for these services, or the routine may 
just build
all the control blocks itself.

I wrote a program of few years back, offlindr, that dumps/restores logically offline dasd (eg z/vm or linux volumes). Looks like one place you can find it is http://www.clueful.co.uk/mbeattie/s390/offlindr.jcl Yes, apf authorization is absolutely required.

A number of years ago we were the object of a political intrigue. Government agents on behalf of a congressman obtained legitimate userids to our system and spent some number of months trying, as a normal user, to get into supervisor state and/or key 0. They were unable to discover such a method in IBM code and (thankfully) our home-grown code, but did discover three methods through vendor software.

One method that I recall is using a particular vendor svc. A key 8 storage area was passed to the svc and the svc used some of that area as an internal save area. The exploit simply issued a stimer for a short duration, and then issued the svc. If the stimer exit got scheduled at the right time, it simply modified the return address in the save area. The method only worked about once in a thousand attempts but considering that we could iterate this procedure tens of thousands of times a second it was virtually instantaneous. I did notify the vendor and the exposure was fixed.

Greg Smith

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html

Reply via email to