On May 20, 2013, at 10:29 AM, David Daney wrote: > On 05/20/2013 09:58 AM, Sanjay Lal wrote: >> >> On May 20, 2013, at 8:50 AM, David Daney wrote: >> >>> On 05/18/2013 10:47 PM, Sanjay Lal wrote: >>>> The following patch set adds support for the recently announced virtualization >>>> extensions for the MIPS32 architecture and allows running unmodified kernels in >>>> Guest Mode. >>>> >>>> For more info please refer to : >>>> MIPS Document #: MD00846 >>>> Volume IV-i: Virtualization Module of the MIPS32 Architecture >>>> >>>> which can be accessed @: http://www.mips.com/auth/MD00846-2B-VZMIPS32-AFP-01.03.pdf >>>> >>>> The patch is agains Linux-3.10-rc1. >>>> >>>> KVM/MIPS now supports 2 modes of operation: >>>> >>>> (1) VZ mode: Unmodified kernels running in Guest Mode. The processor now provides >>>> an almost complete COP0 context in Guest mode. This greatly reduces VM exits. >>> >>> Two questions: >>> >>> 1) How are you handling not clobbering the Guest K0/K1 registers when a Root exception occurs? It is not obvious to me from inspecting the code. >>> >>> 2) What environment are you using to test this stuff? >>> >>> David Daney >>> >> >> (1) Newer versions of the MIPS architecture define scratch registers for just this purpose, but since we have to support standard MIPS32R2 processors, we use the DDataLo Register (CP0 Register 28, Select 3) as a scratch register to save k0 and save k1 @ a known offset from EBASE. >> > > Right, I understand that. But I am looking at arch/mips/mm/tlbex.c, and I don't see the code that does that for TLBRefill exceptions. > > Where is it done for interrupts? I would expect code in arch/mips/kernel/genex.S and/or stackframe.h would handle this. But I don't see where it is. > > Am I missing something? > > David Daney > arch/mips/kvm/kvm_locore.S -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
