projects / thirdparty / kernel / stable-queue.git / blame
| Commit | Line | Data |
|---|---|---|
| 103e869f GKH |
1 | From 4ff6f8e61eb7f96d3ca535c6d240f863ccd6fb7d Mon Sep 17 00:00:00 2001 |
| 2 | From: Paolo Bonzini <pbonzini@redhat.com> | |
| 3 | Date: Thu, 12 Feb 2015 17:04:47 +0100 | |
| 4 | Subject: KVM: emulate: fix CMPXCHG8B on 32-bit hosts | |
| 5 | ||
| 6 | From: Paolo Bonzini <pbonzini@redhat.com> | |
| 7 | ||
| 8 | commit 4ff6f8e61eb7f96d3ca535c6d240f863ccd6fb7d upstream. | |
| 9 | ||
| 10 | This has been broken for a long time: it broke first in 2.6.35, then was | |
| 11 | almost fixed in 2.6.36 but this one-liner slipped through the cracks. | |
| 12 | The bug shows up as an infinite loop in Windows 7 (and newer) boot on | |
| 13 | 32-bit hosts without EPT. | |
| 14 | ||
| 15 | Windows uses CMPXCHG8B to write to page tables, which causes a | |
| 16 | page fault if running without EPT; the emulator is then called from | |
| 17 | kvm_mmu_page_fault. The loop then happens if the higher 4 bytes are | |
| 18 | not 0; the common case for this is that the NX bit (bit 63) is 1. | |
| 19 | ||
| 20 | Fixes: 6550e1f165f384f3a46b60a1be9aba4bc3c2adad | |
| 21 | Fixes: 16518d5ada690643453eb0aef3cc7841d3623c2d | |
| 22 | Reported-by: Erik Rull <erik.rull@rdsoftware.de> | |
| 23 | Tested-by: Erik Rull <erik.rull@rdsoftware.de> | |
| 24 | Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> | |
| 25 | Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> | |
| 26 | ||
| 27 | --- | |
| 28 | arch/x86/kvm/emulate.c | 3 ++- | |
| 29 | 1 file changed, 2 insertions(+), 1 deletion(-) | |
| 30 | ||
| 31 | --- a/arch/x86/kvm/emulate.c | |
| 32 | +++ b/arch/x86/kvm/emulate.c | |
| 33 | @@ -4863,7 +4863,8 @@ int x86_emulate_insn(struct x86_emulate_ | |
| 34 | if (rc != X86EMUL_CONTINUE) | |
| 35 | goto done; | |
| 36 | } | |
| 37 | - ctxt->dst.orig_val = ctxt->dst.val; | |
| 38 | + /* Copy full 64-bit value for CMPXCHG8B. */ | |
| 39 | + ctxt->dst.orig_val64 = ctxt->dst.val64; | |
| 40 | ||
| 41 | special_insn: | |
| 42 |