Hello gem5 users (more specifically x86 users),

I ran into an issue today where the panic in tlb.cc (line 225) is triggered by a WRMSR (write model specific register) that writes to some bogus address because it is on the wrong path.

The Intel manual states that "The WRMSR instruction is a serializing instruction (see "Serializing Instructions" in Chapter 8 of the IA-32 Intel Architecture Software Developer's Manual, Volume 3)." and the microcode in src/arch/x86/isa/insts/system/msrs.py declares the WRMSR macroop as serializing, but the O3 model considers it to be SerializeAfter only, while it should also wait for all previous instructions to have executed (so be SerializeBefore).

I'm not yet clear if this is a flag missing in the instruction object or if this is due to the way Rename handles the instruction (looks like the former though).

So I am merely wondering if someone else ran into the issue, and I will try to check if the problem arises on newer versions of gem5, but the src/arch/x86/isa/insts/system/msrs.py did not change between head and the revision I use. If it has already been caught/fixed then my bad (I did not really find anything related on the mailing list).

Best,

--
Arthur Perais
INRIA Bretagne Atlantique
Bâtiment 12E, Bureau E303, Campus de Beaulieu
35042 Rennes, France

_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to