Jerone Young
2005-May-12 03:55 UTC
[Xen-devel] VMX check in /xen/arch/x86/domain.c is invalid on x86-64 (at least AMD)
if (!(c->flags & ECF_VMX_GUEST)) if ( ((c->cpu_ctxt.cs & 3) == 0) || ((c->cpu_ctxt.ss & 3) == 0) ) return -EINVAL; This check in /xen/arch/x86/domain.c (line 394) is invalid for AMD x86-64 as this field 1<<1 is a reserved field in the rflags (name for 64bit eflags) that reads as 1. This can be found in the AMD x86-64 Programming manual vol 2 on page 65. Below is the code, I''m not sure how to put an ifdef around since I''m not sure if Intel''s EMT64 with VMX cpus have the same rflag reserved (which if this check is correct they do)? This helps break x86-64 building domU domains. -- Jerone Young IBM Linux Technology Center jyoung5@us.ibm.com 512-838-1157 (T/L: 678-1157) _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Arun Sharma
2005-May-12 05:47 UTC
Re: [Xen-devel] VMX check in /xen/arch/x86/domain.c is invalid on x86-64 (at least AMD)
Jerone Young wrote:> if (!(c->flags & ECF_VMX_GUEST)) > if ( ((c->cpu_ctxt.cs & 3) == 0) || > ((c->cpu_ctxt.ss & 3) == 0) ) > return -EINVAL; > > This check in /xen/arch/x86/domain.c (line 394) is invalid for AMD > x86-64 as this field 1<<1 is a reserved field in the rflags (name for > 64bit eflags) that reads as 1. This can be found in the AMD x86-64 > Programming manual vol 2 on page 65. Below is the code, I''m not sure how > to put an ifdef around since I''m not sure if Intel''s EMT64 with VMX cpus > have the same rflag reserved (which if this check is correct they do)? > This helps break x86-64 building domU domains. >It is a reserved bit on Intel as well. Look at: VMCS_EFLAGS_RESERVED_1 and the IA-32 vol1 Fig 3-7. Are you mixing up ctxt->user_regs.rflags and ctxt->flags in xc_linux_build.c? -Arun _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Jerone Young
2005-May-12 05:58 UTC
Re: [Xen-devel] VMX check in /xen/arch/x86/domain.c is invalid on x86-64 (at least AMD)
Yes I am actually. Sorry guy. I got that one screwed up in my head. Pressed the send button too quickly. Thanks for the respsonses though. On 5/12/05, Arun Sharma <arun.sharma@intel.com> wrote:> Jerone Young wrote: > > if (!(c->flags & ECF_VMX_GUEST)) > > if ( ((c->cpu_ctxt.cs & 3) == 0) || > > ((c->cpu_ctxt.ss & 3) == 0) ) > > return -EINVAL; > > > > This check in /xen/arch/x86/domain.c (line 394) is invalid for AMD > > x86-64 as this field 1<<1 is a reserved field in the rflags (name for > > 64bit eflags) that reads as 1. This can be found in the AMD x86-64 > > Programming manual vol 2 on page 65. Below is the code, I''m not sure how > > to put an ifdef around since I''m not sure if Intel''s EMT64 with VMX cpus > > have the same rflag reserved (which if this check is correct they do)? > > This helps break x86-64 building domU domains. > > > > It is a reserved bit on Intel as well. Look at: > > VMCS_EFLAGS_RESERVED_1 and the IA-32 vol1 Fig 3-7. > > Are you mixing up ctxt->user_regs.rflags and ctxt->flags in > xc_linux_build.c? > > -Arun > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel