diff mbox series

x86/kvm: Fix state restore in em_rsm

Message ID 20221026215255.1063662-1-lk@c--e.de
State New
Headers
Series x86/kvm: Fix state restore in em_rsm |

Commit Message

Christian A. Ehrhardt Oct. 26, 2022, 9:52 p.m. UTC 
  Syzkaller reports a stack-out-of-bounds access when
emulating RSM (return from system management mode).

Assume that a 64-bit capable host (i.e. CONFIG_X86_64 is true)
emulates a guest cpu that does not support 64-bit mode. In this case
RSM must use the 32-bit version of the SMM state map which only
contains space for 8 general purpose registers.
However, NR_EMULATOR_GPRS is defined to 16 due to CONFIG_X86_64.

As a result rsm_load_state_32 will try to restore 16
registers from the state save area which only contains 8
registers. Manual offset calculation easily shows that
memory beyond the end of the smstate buffer is accessed in
this case.

Revert the relevant parts of b443183a25ab and use explicit constants
for the number of general purpose registers, again. This
also ensures that the code in rsm_load_state_{32,64} matches
what is done in enter_smm_save_state_{32,64}.

Fixes: b443183a25ab ("KVM: x86: Reduce the number of emulator GPRs to '8' for 32-bit KVM")
Signed-off-by: Christian A. Ehrhardt <lk@c--e.de>
---
 arch/x86/kvm/emulate.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Sean Christopherson Oct. 26, 2022, 10:33 p.m. UTC | #1 
On Wed, Oct 26, 2022, Christian A. Ehrhardt wrote:
> Syzkaller reports a stack-out-of-bounds access when
> emulating RSM (return from system management mode).
> 
> Assume that a 64-bit capable host (i.e. CONFIG_X86_64 is true)
> emulates a guest cpu that does not support 64-bit mode. In this case
> RSM must use the 32-bit version of the SMM state map which only
> contains space for 8 general purpose registers.
> However, NR_EMULATOR_GPRS is defined to 16 due to CONFIG_X86_64.
> 
> As a result rsm_load_state_32 will try to restore 16
> registers from the state save area which only contains 8
> registers. Manual offset calculation easily shows that
> memory beyond the end of the smstate buffer is accessed in
> this case.
> 
> Revert the relevant parts of b443183a25ab and use explicit constants
> for the number of general purpose registers, again. This
> also ensures that the code in rsm_load_state_{32,64} matches
> what is done in enter_smm_save_state_{32,64}.
> 
> Fixes: b443183a25ab ("KVM: x86: Reduce the number of emulator GPRs to '8' for 32-bit KVM")
> Signed-off-by: Christian A. Ehrhardt <lk@c--e.de>

A fix has already been posted[*], we've just been abysmally slow getting it merged.
I'll make sure it gets queued up for 6.1.

Thanks!

[*] https://lore.kernel.org/all/20221025124741.228045-15-mlevitsk@redhat.com
diff mbox series

Patch

diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index 3b27622d4642..05355ebaf4f3 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -2432,7 +2432,7 @@  static int rsm_load_state_32(struct x86_emulate_ctxt *ctxt,
 	ctxt->eflags =             GET_SMSTATE(u32, smstate, 0x7ff4) | X86_EFLAGS_FIXED;
 	ctxt->_eip =               GET_SMSTATE(u32, smstate, 0x7ff0);
 
-	for (i = 0; i < NR_EMULATOR_GPRS; i++)
+	for (i = 0; i < 8; i++)
 		*reg_write(ctxt, i) = GET_SMSTATE(u32, smstate, 0x7fd0 + i * 4);
 
 	val = GET_SMSTATE(u32, smstate, 0x7fcc);
@@ -2489,7 +2489,7 @@  static int rsm_load_state_64(struct x86_emulate_ctxt *ctxt,
 	u16 selector;
 	int i, r;
 
-	for (i = 0; i < NR_EMULATOR_GPRS; i++)
+	for (i = 0; i < 16; i++)
 		*reg_write(ctxt, i) = GET_SMSTATE(u64, smstate, 0x7ff8 - i * 8);
 
 	ctxt->_eip   = GET_SMSTATE(u64, smstate, 0x7f78);