x86: make PioBus return BadAddress errors

Information
Submitter:	Steve Reinhardt
Repository:	gem5
Branch:	default
Bugs:
Depends On:
Reviewers
Groups:	Default
People:

Description

Changeset 10239:ef2bd7e9922c
---------------------------
x86: make PioBus return BadAddress errors

Stop setting the use_default_range flag in PioBus in order to
have random bad addresses result in a BadAddress response and
not a gem5 fatal error.  This is necessary in Ruby as Ruby is
connected directly to PioBus, so misspeculated addresses will
be sent there directly.  For the classic memory system, this
change has no effect, as bad addresses are caught by the
memory bus before being sent to the PioBus.

This work was done while Binh was an intern at AMD Research.

Testing Done

Perhaps I am missing something, but why would Ruby forward the packet to the iobus in the first place if the address is not valid?

Steve Reinhardt June 25, 2014, 7:44 a.m. (June 25, 2014, 7:44 a.m.)

This is an FS thing, where you misspeculate in the kernel and generate a bogus address in the TLB bypass region, so you have no way of knowing a priori that it's not valid.  Normally what happens is you get a "bad address" response from the memory bus, which flags the access as a fault, but you never take the fault because the access gets squashed.  Since Ruby doesn't use the memory bus, but sends all non-cacheable requests to the I/O bus, we just need to reproduce that behavior there.

Andreas Hansson June 25, 2014, 9 a.m. (June 25, 2014, 9 a.m.)

Thanks for the clarification. Perhaps this is too heavy weight, but is there any chance Ruby could actually check the address range of the port connected to the piobus before forwarding the packet?

Steve Reinhardt June 25, 2014, 10:22 a.m. (June 25, 2014, 10:22 a.m.)

I suppose anything is possible ;).  Is there a reason you'd prefer that solution?

Andreas Hansson June 25, 2014, 11 a.m. (June 25, 2014, 11 a.m.)

It is mostly down to symmetry. It would be nice to keep things the same way for the Ruby vs not-Ruby case. If it's more than a few lines of code to be added to Ruby, then let's go with the original idea.

Ali Saidi June 25, 2014, 1:05 p.m. (June 25, 2014, 1:05 p.m.)

this is the symmetrical thing to do. For classic it's set on the membus.

As I mentioned before, I'm not opposed this going in as is if adding the check in Ruby is more than a line or two of code.

You have a pending review.

Review Board 2.0.15

This change has been marked as submitted.

Screenshots

Files

Status: Closed (submitted)