x86/svm: Separate STI and VMRUN instructions in svm_asm_do_resume()

There is a corner case in the VMRUN instruction where its INTR_SHADOW state
leaks into guest state if a VMExit occurs before the VMRUN is complete.  An
example of this could be taking #NPF due to event injection.

Xen can safely execute STI anywhere between CLGI and VMRUN, as CLGI blocks
external interrupts too.  However, an exception (while fatal) will appear to
be in an irqs-on region (as GIF isn't considered), so position the STI after
the speculation actions but prior to the GPR pops.

Link: https://lore.kernel.org/all/CADH9ctBs1YPmE4aCfGPNBwA10cA8RuAk2gO7542DjMZgs4uzJQ@mail.gmail.com/
Fixes: 66b245d9eaeb ("SVM: limit GIF=0 region")
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: c989ff614f6bad48b3bd4b32694f711b31c7b2d6
master date: 2025-02-19 12:45:48 +0000

diff --git a/xen/arch/x86/hvm/svm/entry.S b/xen/arch/x86/hvm/svm/entry.S
index 5f318629b4a727fc3876d44bf0643f74914736ff..9fb457ad958ee81976649515b74386922020c74f 100644 (file)
--- a/xen/arch/x86/hvm/svm/entry.S
+++ b/xen/arch/x86/hvm/svm/entry.S
@@ -74,6 +74,14 @@ __UNLIKELY_END(nsvm_hap)
         ALTERNATIVE "", svm_vmentry_spec_ctrl, X86_FEATURE_SC_MSR_HVM
         ALTERNATIVE "", DO_SPEC_CTRL_DIV, X86_FEATURE_SC_DIV
 
+        /*
+         * Set EFLAGS.IF after CLGI covers us from real interrupts, but not
+         * immediately prior to VMRUN.  The VMRUN instruction leaks it's
+         * INTR_SHADOW into guest state if a VMExit occurs before VMRUN
+         * completes (e.g. taking #NPF during event injecting.)
+         */
+        sti
+
         pop  %r15
         pop  %r14
         pop  %r13
@@ -91,7 +99,6 @@ __UNLIKELY_END(nsvm_hap)
         pop  %rsi
         pop  %rdi
 
-        sti
         vmrun
 
         SAVE_ALL