[Devel,RHEL7,COMMIT] ms/KVM: x86: emulate FXSAVE and FXRSTOR

Submitted by Konstantin Khorenko on Aug. 7, 2017, 10:41 a.m.

Details

Message ID 201708071041.v77AfBlp008154@finist_cl7.x64_64.work.ct
State New
Series "ms/KVM: x86: emulate FXSAVE and FXRSTOR"
Headers show

Commit Message

Konstantin Khorenko Aug. 7, 2017, 10:41 a.m.
The commit is pushed to "branch-rh7-3.10.0-514.26.1.vz7.35.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-514.26.1.vz7.33.22
------>
commit 0ed8ba6d68a7e77e2c07c47fff2a2f98641df252
Author: Radim Krčmář <rkrcmar@redhat.com>
Date:   Mon Aug 7 14:41:11 2017 +0400

    ms/KVM: x86: emulate FXSAVE and FXRSTOR
    
    Internal errors were reported on 16 bit fxsave and fxrstor with ipxe.
    Old Intels don't have unrestricted_guest, so we have to emulate them.
    
    The patch takes advantage of the hardware implementation.
    
    AMD and Intel differ in saving and restoring other fields in first 32
    bytes.  A test wrote 0xff to the fxsave area, 0 to upper bits of MCSXR
    in the fxsave area, executed fxrstor, rewrote the fxsave area to 0xee,
    and executed fxsave:
    
      Intel (Nehalem):
        7f 1f 7f 7f ff 00 ff 07 ff ff ff ff ff ff 00 00
        ff ff ff ff ff ff 00 00 ff ff 00 00 ff ff 00 00
      Intel (Haswell -- deprecated FPU CS and FPU DS):
        7f 1f 7f 7f ff 00 ff 07 ff ff ff ff 00 00 00 00
        ff ff ff ff 00 00 00 00 ff ff 00 00 ff ff 00 00
      AMD (Opteron 2300-series):
        7f 1f 7f 7f ff 00 ee ee ee ee ee ee ee ee ee ee
        ee ee ee ee ee ee ee ee ff ff 00 00 ff ff 02 00
    
    fxsave/fxrstor will only be emulated on early Intels, so KVM can't do
    much to improve the situation.
    
    Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
    
    https://jira.sw.ru/browse/PSBM-69206
    ms commit: 283c95d0e3891b64087706b344a4b545d04a6e62
    
    Signed-off-by: Evgeny Yakovlev <eyakovlev@virtuozzo.com>
---
 arch/x86/kvm/emulate.c | 129 ++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 128 insertions(+), 1 deletion(-)

Patch hide | download patch | download mbox

diff --git a/arch/x86/kvm/emulate.c b/arch/x86/kvm/emulate.c
index ee1835d..d7e33ac 100644
--- a/arch/x86/kvm/emulate.c
+++ b/arch/x86/kvm/emulate.c
@@ -3931,6 +3931,131 @@  static int em_movsxd(struct x86_emulate_ctxt *ctxt)
 	return X86EMUL_CONTINUE;
 }
 
+static int check_fxsr(struct x86_emulate_ctxt *ctxt)
+{
+	u32 eax = 1, ebx, ecx = 0, edx;
+
+	ctxt->ops->get_cpuid(ctxt, &eax, &ebx, &ecx, &edx);
+	if (!(edx & FFL(FXSR)))
+		return emulate_ud(ctxt);
+
+	if (ctxt->ops->get_cr(ctxt, 0) & (X86_CR0_TS | X86_CR0_EM))
+		return emulate_nm(ctxt);
+
+	/*
+	 * Don't emulate a case that should never be hit, instead of working
+	 * around a lack of fxsave64/fxrstor64 on old compilers.
+	 */
+	if (ctxt->mode >= X86EMUL_MODE_PROT64)
+		return X86EMUL_UNHANDLEABLE;
+
+	return X86EMUL_CONTINUE;
+}
+
+/*
+ * FXSAVE and FXRSTOR have 4 different formats depending on execution mode,
+ *  1) 16 bit mode
+ *  2) 32 bit mode
+ *     - like (1), but FIP and FDP (foo) are only 16 bit.  At least Intel CPUs
+ *       preserve whole 32 bit values, though, so (1) and (2) are the same wrt.
+ *       save and restore
+ *  3) 64-bit mode with REX.W prefix
+ *     - like (2), but XMM 8-15 are being saved and restored
+ *  4) 64-bit mode without REX.W prefix
+ *     - like (3), but FIP and FDP are 64 bit
+ *
+ * Emulation uses (3) for (1) and (2) and preserves XMM 8-15 to reach the
+ * desired result.  (4) is not emulated.
+ *
+ * Note: Guest and host CPUID.(EAX=07H,ECX=0H):EBX[bit 13] (deprecate FPU CS
+ * and FPU DS) should match.
+ */
+static int em_fxsave(struct x86_emulate_ctxt *ctxt)
+{
+	struct i387_fxsave_struct fx_state;
+	size_t size;
+	int rc;
+
+	rc = check_fxsr(ctxt);
+	if (rc != X86EMUL_CONTINUE)
+		return rc;
+
+	ctxt->ops->get_fpu(ctxt);
+
+	rc = asm_safe("fxsave %[fx]", , [fx] "+m"(fx_state));
+
+	ctxt->ops->put_fpu(ctxt);
+
+	if (rc != X86EMUL_CONTINUE)
+		return rc;
+
+	if (ctxt->ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR)
+		size = offsetof(struct i387_fxsave_struct, xmm_space[8 * 16/4]);
+	else
+		size = offsetof(struct i387_fxsave_struct, xmm_space[0]);
+
+	return segmented_write_std(ctxt, ctxt->memop.addr.mem, &fx_state, size);
+}
+
+static int fxrstor_fixup(struct x86_emulate_ctxt *ctxt,
+		struct i387_fxsave_struct *new)
+{
+	int rc = X86EMUL_CONTINUE;
+	struct i387_fxsave_struct old;
+
+	rc = asm_safe("fxsave %[fx]", , [fx] "+m"(old));
+	if (rc != X86EMUL_CONTINUE)
+		return rc;
+
+	/*
+	 * 64 bit host will restore XMM 8-15, which is not correct on non-64
+	 * bit guests.  Load the current values in order to preserve 64 bit
+	 * XMMs after fxrstor.
+	 */
+#ifdef CONFIG_X86_64
+	/* XXX: accessing XMM 8-15 very awkwardly */
+	memcpy(&new->xmm_space[8 * 16/4], &old.xmm_space[8 * 16/4], 8 * 16);
+#endif
+
+	/*
+	 * Hardware doesn't save and restore XMM 0-7 without CR4.OSFXSR, but
+	 * does save and restore MXCSR.
+	 */
+	if (!(ctxt->ops->get_cr(ctxt, 4) & X86_CR4_OSFXSR))
+		memcpy(new->xmm_space, old.xmm_space, 8 * 16);
+
+	return rc;
+}
+
+static int em_fxrstor(struct x86_emulate_ctxt *ctxt)
+{
+	struct i387_fxsave_struct fx_state;
+	int rc;
+
+	rc = check_fxsr(ctxt);
+	if (rc != X86EMUL_CONTINUE)
+		return rc;
+
+	rc = segmented_read_std(ctxt, ctxt->memop.addr.mem, &fx_state, 512);
+	if (rc != X86EMUL_CONTINUE)
+		return rc;
+
+	if (fx_state.mxcsr >> 16)
+		return emulate_gp(ctxt, 0);
+
+	ctxt->ops->get_fpu(ctxt);
+
+	if (ctxt->mode < X86EMUL_MODE_PROT64)
+		rc = fxrstor_fixup(ctxt, &fx_state);
+
+	if (rc == X86EMUL_CONTINUE)
+		rc = asm_safe("fxrstor %[fx]", : [fx] "m"(fx_state));
+
+	ctxt->ops->put_fpu(ctxt);
+
+	return rc;
+}
+
 static bool valid_cr(int nr)
 {
 	switch (nr) {
@@ -4283,7 +4408,9 @@  static const struct gprefix pfx_0f_ae_7 = {
 };
 
 static const struct group_dual group15 = { {
-	N, N, N, N, N, N, N, GP(0, &pfx_0f_ae_7),
+	I(ModRM | Aligned16, em_fxsave),
+	I(ModRM | Aligned16, em_fxrstor),
+	N, N, N, N, N, GP(0, &pfx_0f_ae_7),
 }, {
 	N, N, N, N, N, N, N, N,
 } };