Alan Altmark wrote:
There is no need for heroic measures to write a nice message to the console. Just make sure that the addresses in the PSW are unique for each condition. If you can be clever in your use of letters A-F and numbers (and l33t5peak), you can give informative wait state codes.
I'll second that.. That's why I was proposing : 8EADBADC (HEAD detected BAD Cpu) I also noticed there is a flurry of conditions in head.S that also load a disabled wait PSW, incurring from an IPL I/O error.. So I guess 8EAD1030 (For Head I/O ErrOr) The problem of course is that with 8 characters of 1337 (And.. Alan.. it's 13375|*34|< - but I digress) expressing something is - shall we say - bound to be esoteric. My idea would be to put some random magic in there.. and *document* it (possibly in a new ./Documentation/s390/ipl.txt file) so as to allow it to appear prominently in search engines. So if someone gets a 000A0000 8EADBADC disabled wait PSW, chances are good that searching for that will yield useful results quite easily for the unsuspecting victim. We probably have to go through something like this because - contrary to IBM program products (not limited to strictly mainframe OSes.. AIX is the same, as well as about any IBM issued software), which have a centralized managed catalog for message numbers and the documentation that goes with it - linux doesn't have that - so we have to rely on some external source to achieve the same result - that is - finding resolution about a failure with whatever information -is- or -can be- provided at the time of failure. Now, I'm still not 100% convinced that checking the CPUID is the right thing to do.. STFL/STFLE - as proposed earlier- is probably a better approach (but that complicates the code quite a bit). Now, if you generate a say.. z9+ only kernel and run it on a z800.. And gcc generates some z9 only code, we're going to get a program check quite quickly (it won't break anything.. it'll just panic big time).. And if for some reason, IBM decides to port some of the newer functions to older machines (through a driver update - driver in the sense of functional LIC) - then it will run because.. it can ! Finally, maybe this should be moved to the s390 kernel list.. not sure this is the appropriate list to go into such gory details ! --Ivan ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
smime.p7s
Description: S/MIME Cryptographic Signature
