Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

>>> On 14.02.18 at 17:23, wrote: > The current XPTI implementation isolates the directmap (and therefore a lot > of > guest data), but a large quantity of CPU0's state (including its stack) > remains visible. > > Furthermore, an attacker able to read .text is in a

[Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

The current XPTI implementation isolates the directmap (and therefore a lot of guest data), but a large quantity of CPU0's state (including its stack) remains visible. Furthermore, an attacker able to read .text is in a vastly superior position to normal when it comes to fingerprinting Xen for

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 14/02/18 12:27, Juergen Gross wrote: > On 14/02/18 13:19, Andrew Cooper wrote: >> On 14/02/18 12:15, Juergen Gross wrote: >>> On 14/02/18 13:03, Juergen Gross wrote: On 14/02/18 12:48, Andrew Cooper wrote: > On 14/02/18 07:54, Juergen Gross wrote: >> On 13/02/18 20:45, Andrew

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 14/02/18 09:14, Jan Beulich wrote: On 13.02.18 at 20:45, wrote: >> RFC, because I don't think the stubs handling is particularly sensible. >> >> We allocate 4k of virtual address space per CPU, but squash loads of CPUs >> together onto a single MFN. The stubs

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 14/02/18 13:19, Andrew Cooper wrote: > On 14/02/18 12:15, Juergen Gross wrote: >> On 14/02/18 13:03, Juergen Gross wrote: >>> On 14/02/18 12:48, Andrew Cooper wrote: On 14/02/18 07:54, Juergen Gross wrote: > On 13/02/18 20:45, Andrew Cooper wrote: >> The current XPTI implementation

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 14/02/18 12:15, Juergen Gross wrote: > On 14/02/18 13:03, Juergen Gross wrote: >> On 14/02/18 12:48, Andrew Cooper wrote: >>> On 14/02/18 07:54, Juergen Gross wrote: On 13/02/18 20:45, Andrew Cooper wrote: > The current XPTI implementation isolates the directmap (and therefore a >

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 14/02/18 13:03, Juergen Gross wrote: > On 14/02/18 12:48, Andrew Cooper wrote: >> On 14/02/18 07:54, Juergen Gross wrote: >>> On 13/02/18 20:45, Andrew Cooper wrote: The current XPTI implementation isolates the directmap (and therefore a lot of guest data), but a large quantity

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 14/02/18 12:48, Andrew Cooper wrote: > On 14/02/18 07:54, Juergen Gross wrote: >> On 13/02/18 20:45, Andrew Cooper wrote: >>> The current XPTI implementation isolates the directmap (and therefore a lot >>> of >>> guest data), but a large quantity of CPU0's state (including its stack) >>>

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 14/02/18 07:54, Juergen Gross wrote: > On 13/02/18 20:45, Andrew Cooper wrote: >> The current XPTI implementation isolates the directmap (and therefore a lot >> of >> guest data), but a large quantity of CPU0's state (including its stack) >> remains visible. >> >> Furthermore, an attacker able

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

>>> On 13.02.18 at 20:45, wrote: > RFC, because I don't think the stubs handling is particularly sensible. > > We allocate 4k of virtual address space per CPU, but squash loads of CPUs > together onto a single MFN. The stubs ought to be isolated as well (as they >

Re: [Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

On 13/02/18 20:45, Andrew Cooper wrote: > The current XPTI implementation isolates the directmap (and therefore a lot of > guest data), but a large quantity of CPU0's state (including its stack) > remains visible. > > Furthermore, an attacker able to read .text is in a vastly superior position >

[Xen-devel] [PATCH] x86/xpti: Hide almost all of .text and all .data/.rodata/.bss mappings

The current XPTI implementation isolates the directmap (and therefore a lot of guest data), but a large quantity of CPU0's state (including its stack) remains visible. Furthermore, an attacker able to read .text is in a vastly superior position to normal when it comes to fingerprinting Xen for