John Summerfield wrote:
Andrew Avramenko wrote:
Daniel,
SLES 11 doesn't support old hardware, just only z9 and z10. I'm not
familiar with Hercules, but may be that is the problem. I have
experience with installation SLES 11 on z9 under VM 5.3 and everything
was fine.
I the Hercules folk pride themselves on supporting the latest
incantations of S/360, right up to current.
Actually, Hercules does not support System/360 architecture. It does
support System/370 Basic Control mode which is very close to System/360,
but not identical. And yes, the very latest capabilities of System/370
is supported as defined by the IBM System/370 Principles of Operation,
manual GA22-7000-10 published in September 1987 and still available from
IBM publications. I ordered my copy last year. And, yes, the
intervening architectual enhancements are also supported to the
z/Architecture PoO published last fall. And we are looking at the
implications of the one published in February of this year.
Of course, Hercules has actual users of all of these architectures, so
it does make sense that all of them are supported.
So, I guess this post proves your point, John, we are proud of that.
The Linux kernel developers are not in anyway attempting to restrict
where Linux can run. It is making the boot code sensitive to the
compiler and options used to create the kernel. Those settings
influence instructions selected for the generated code. Rather than an
obscure abend occurring someplace because of a hardware
capabilities/compiler option conflict, the boot logic is trying to
detect that up front. This is standard defensive programming.
While historically there has been slightly less adherence to backward
compatibility for control programs than application programs on IBM
mainframes. For example, backward compatibility was lost for control
programs when the subchannel subsystem was introduced in the 370-XA
architecture. The domain of gcc is application programs, meaning,
non-privileged instructions in mainframe architectural terms. And, on
an IBM mainframe, application program backward compatibility has been
absolute. The Linux kernel's use of privileged instructions in in-line
assembly raises the question of backward compatibility for the a control
program. A separate question from the gcc compiler and options settings
addressed by this boot logic.
The gcc compiler and option detection is also separate from the
decisions of a distributor such as Novell or RedHat or Debian. For the
user, there are both positive and negative implications of those
decisions. The beauty of open source, is, if you don't like those
decisions or absolutely can't live with them, you have options. And,
while these changes don't actually enable such decisions by a
distributor, a distributor could change compiler and option settings at
will, they do attempt to aid a Linux user early detection. And of
course, this becomes a potential differentiator of distributors.
Just as the Hercules developers are sensitive to the users of Hercules,
kernel developers also have to be sensitive to their users. For a
kernel developer, this is not just the users of Linux but also the
distributors of Linux.
Harold Grovesteen
----------------------------------------------------------------------
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
