Add support for the XSAVEOPT instruction use. Our XSAVE/XRSTOR usage

mostly meets the guidelines set by the Intel SDM:
1. We use XRSTOR and XSAVE from the same CPL using the same linear
   address for the store area
2. Contrary to the recommendations, we cannot zero the FPU save area
   for a new thread, since fork semantic requires the copy of the
   previous state. This advice seemingly contradicts to the advice
   from the item 6.
3. We do use XSAVEOPT in the context switch code only, and the area
   for XSAVEOPT already always contains the data saved by XSAVE.
4. We do not modify the save area between XRSTOR, when the area is
   loaded into FPU context, and XSAVE. We always spit the fpu context
   into save area and start emulation when directly writing into FPU
   context.
5. We do not use segmented addressing to access save area, or rather,
   always address it using %ds basing.
6. XSAVEOPT can be only executed in the area which was previously
   loaded with XRSTOR, since context switch code checks for FPU use by
   outgoing thread before saving, and thread which stopped emulation
   forcibly get context loaded with XRSTOR.
7. The PCB cannot be paged out while FPU emulation is turned off, since
   stack of the executing thread is never swapped out.

The context switch code is patched to issue XSAVEOPT instead of XSAVE
if supported. This approach eliminates one conditional in the context
switch code, which would be needed otherwise.

For user-visible machine context to have proper data, fpugetregs()
checks for unsaved extension blocks and manually copies pristine FPU
state into them, according to the description provided by CPUID leaf
0xd.

MFC after:  1 month

3 files changed, 73 insertions, 1 deletions

diff --git a/sys/amd64/amd64/cpu_switch.S b/sys/amd64/amd64/cpu_switch.S
index 1254f3f..ed1ccb5 100644
--- a/sys/amd64/amd64/cpu_switch.S
+++ b/sys/amd64/amd64/cpu_switch.S
@@ -122,6 +122,9 @@ done_store_dr:
 1:	movq	%rdx,%rcx
 	movl	xsave_mask,%eax
 	movl	xsave_mask+4,%edx
+	.globl	ctx_switch_xsave
+ctx_switch_xsave:
+	/* This is patched to xsaveopt if supported, see fpuinit_bsp1() */
 	xsave	(%r8)
 	movq	%rcx,%rdx
 2:	smsw	%ax
diff --git a/sys/amd64/amd64/fpu.c b/sys/amd64/amd64/fpu.c
index ca951ad..a7812b7 100644
--- a/sys/amd64/amd64/fpu.c
+++ b/sys/amd64/amd64/fpu.c
@@ -132,10 +132,16 @@ static	void	fpu_clean_state(void);
 SYSCTL_INT(_hw, HW_FLOATINGPT, floatingpoint, CTLFLAG_RD,
     NULL, 1, "Floating point instructions executed in hardware");
 
+static int use_xsaveopt;
 int use_xsave;			/* non-static for cpu_switch.S */
 uint64_t xsave_mask;		/* the same */
 static	struct savefpu *fpu_initialstate;
 
+struct xsave_area_elm_descr {
+	u_int	offset;
+	u_int	size;
+} *xsave_area_desc;
+
 void
 fpusave(void *addr)
 {
@@ -182,6 +188,17 @@ fpuinit_bsp1(void)
 	TUNABLE_ULONG_FETCH("hw.xsave_mask", &xsave_mask_user);
 	xsave_mask_user |= XFEATURE_ENABLED_X87 | XFEATURE_ENABLED_SSE;
 	xsave_mask &= xsave_mask_user;
+
+	cpuid_count(0xd, 0x1, cp);
+	if ((cp[0] & CPUID_EXTSTATE_XSAVEOPT) != 0) {
+		/*
+		 * Patch the XSAVE instruction in the cpu_switch code
+		 * to XSAVEOPT.  We assume that XSAVE encoding used
+		 * REX byte, and set the bit 4 of the r/m byte.
+		 */
+		ctx_switch_xsave[3] |= 0x10;
+		use_xsaveopt = 1;
+	}
 }
 
 /*
@@ -252,6 +269,7 @@ static void
 fpuinitstate(void *arg __unused)
 {
 	register_t saveintr;
+	int cp[4], i, max_ext_n;
 
 	fpu_initialstate = malloc(cpu_max_ext_state_size, M_DEVBUF,
 	    M_WAITOK | M_ZERO);
@@ -273,6 +291,28 @@ fpuinitstate(void *arg __unused)
 	 */
 	bzero(&fpu_initialstate->sv_xmm[0], sizeof(struct xmmacc));
 
+	/*
+	 * Create a table describing the layout of the CPU Extended
+	 * Save Area.
+	 */
+	if (use_xsaveopt) {
+		max_ext_n = flsl(xsave_mask);
+		xsave_area_desc = malloc(max_ext_n * sizeof(struct
+		    xsave_area_elm_descr), M_DEVBUF, M_WAITOK | M_ZERO);
+		/* x87 state */
+		xsave_area_desc[0].offset = 0;
+		xsave_area_desc[0].size = 160;
+		/* XMM */
+		xsave_area_desc[1].offset = 160;
+		xsave_area_desc[1].size = 288 - 160;
+
+		for (i = 2; i < max_ext_n; i++) {
+			cpuid_count(0xd, i, cp);
+			xsave_area_desc[i].offset = cp[1];
+			xsave_area_desc[i].size = cp[0];
+		}
+	}
+
 	start_emulating();
 	intr_restore(saveintr);
 }
@@ -560,8 +600,14 @@ fpudna(void)
 		 * This is the first time this thread has used the FPU or
 		 * the PCB doesn't contain a clean FPU state.  Explicitly
 		 * load an initial state.
+		 *
+		 * We prefer to restore the state from the actual save
+		 * area in PCB instead of directly loading from
+		 * fpu_initialstate, to ignite the XSAVEOPT
+		 * tracking engine.
 		 */
-		fpurestore(fpu_initialstate);
+		bcopy(fpu_initialstate, pcb->pcb_save, cpu_max_ext_state_size);
+		fpurestore(pcb->pcb_save);
 		if (pcb->pcb_initial_fpucw != __INITIAL_FPUCW__)
 			fldcw(pcb->pcb_initial_fpucw);
 		if (PCB_USER_FPU(pcb))
@@ -596,6 +642,9 @@ int
 fpugetregs(struct thread *td)
 {
 	struct pcb *pcb;
+	uint64_t *xstate_bv, bit;
+	char *sa;
+	int max_ext_n, i;
 
 	pcb = td->td_pcb;
 	if ((pcb->pcb_flags & PCB_USERFPUINITDONE) == 0) {
@@ -613,6 +662,25 @@ fpugetregs(struct thread *td)
 		return (_MC_FPOWNED_FPU);
 	} else {
 		critical_exit();
+		if (use_xsaveopt) {
+			/*
+			 * Handle partially saved state.
+			 */
+			sa = (char *)get_pcb_user_save_pcb(pcb);
+			xstate_bv = (uint64_t *)(sa + sizeof(struct savefpu) +
+			    offsetof(struct xstate_hdr, xstate_bv));
+			max_ext_n = flsl(xsave_mask);
+			for (i = 0; i < max_ext_n; i++) {
+				bit = 1 << i;
+				if ((*xstate_bv & bit) != 0)
+					continue;
+				bcopy((char *)fpu_initialstate +
+				    xsave_area_desc[i].offset,
+				    sa + xsave_area_desc[i].offset,
+				    xsave_area_desc[i].size);
+				*xstate_bv |= bit;
+			}
+		}
 		return (_MC_FPOWNED_PCB);
 	}
 }
diff --git a/sys/amd64/include/md_var.h b/sys/amd64/include/md_var.h
index ff11ea1..ff322bb 100644
--- a/sys/amd64/include/md_var.h
+++ b/sys/amd64/include/md_var.h
@@ -57,6 +57,7 @@ extern	u_int	cpu_procinfo;
 extern	u_int	cpu_procinfo2;
 extern	char	cpu_vendor[];
 extern	u_int	cpu_vendor_id;
+extern	char	ctx_switch_xsave[];
 extern	char	kstack[];
 extern	char	sigcode[];
 extern	int	szsigcode;