]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/i386/fpu.c
xnu-4903.270.47.tar.gz
[apple/xnu.git] / osfmk / i386 / fpu.c
index df870f71e5aa2e524cff136905f12554b08f3d7c..0ac53d48c30e63f6e7e0e8cf80a044cc92f3730e 100644 (file)
@@ -1,8 +1,8 @@
 /*
 /*
- * Copyright (c) 2000-2012 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2018 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
- * 
+ *
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
  * unlawful or unlicensed copies of an Apple operating system, or to
  * circumvent, violate, or enable the circumvention or violation of, any
  * terms of an Apple operating system software license agreement.
  * unlawful or unlicensed copies of an Apple operating system, or to
  * circumvent, violate, or enable the circumvention or violation of, any
  * terms of an Apple operating system software license agreement.
- * 
+ *
  * Please obtain a copy of the License at
  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  * Please obtain a copy of the License at
  * http://www.opensource.apple.com/apsl/ and read it before using this file.
- * 
+ *
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  * Please see the License for the specific language governing rights and
  * limitations under the License.
- * 
+ *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_COPYRIGHT@
  */
  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_COPYRIGHT@
  */
-/* 
+/*
  * Mach Operating System
  * Copyright (c) 1992-1990 Carnegie Mellon University
  * All Rights Reserved.
  * Mach Operating System
  * Copyright (c) 1992-1990 Carnegie Mellon University
  * All Rights Reserved.
- * 
+ *
  * Permission to use, copy, modify and distribute this software and its
  * documentation is hereby granted, provided that both the copyright
  * notice and this permission notice appear in all copies of the
  * software, derivative works or modified versions, and any portions
  * thereof, and that both notices appear in supporting documentation.
  * Permission to use, copy, modify and distribute this software and its
  * documentation is hereby granted, provided that both the copyright
  * notice and this permission notice appear in all copies of the
  * software, derivative works or modified versions, and any portions
  * thereof, and that both notices appear in supporting documentation.
- * 
+ *
  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
- * 
+ *
  * Carnegie Mellon requests users of this software to return to
  * Carnegie Mellon requests users of this software to return to
- * 
+ *
  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
  *  School of Computer Science
  *  Carnegie Mellon University
  *  Pittsburgh PA 15213-3890
  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
  *  School of Computer Science
  *  Carnegie Mellon University
  *  Pittsburgh PA 15213-3890
- * 
+ *
  * any improvements or extensions that they make and grant Carnegie Mellon
  * the rights to redistribute these changes.
  */
  * any improvements or extensions that they make and grant Carnegie Mellon
  * the rights to redistribute these changes.
  */
-/*
- */
-
 
 #include <mach/exception_types.h>
 #include <mach/i386/thread_status.h>
 #include <mach/i386/fp_reg.h>
 
 #include <mach/exception_types.h>
 #include <mach/i386/thread_status.h>
 #include <mach/i386/fp_reg.h>
-#include <mach/branch_predicates.h>
 
 #include <kern/mach_param.h>
 #include <kern/processor.h>
 
 #include <kern/mach_param.h>
 #include <kern/processor.h>
 #include <i386/thread.h>
 #include <i386/trap.h>
 
 #include <i386/thread.h>
 #include <i386/trap.h>
 
-int            fp_kind = FP_NO;        /* not inited */
-zone_t         ifps_zone;              /* zone for FPU save area */
+xstate_t        fpu_capability = UNDEFINED;     /* extended state capability */
+xstate_t        fpu_default = UNDEFINED;        /* default extended state */
 
 
-#define ALIGNED(addr,size)     (((uintptr_t)(addr)&((size)-1))==0)
+#define ALIGNED(addr, size)      (((uintptr_t)(addr)&((size)-1))==0)
 
 /* Forward */
 
 
 /* Forward */
 
-extern void            fpinit(void);
-extern void            fp_save(
-                               thread_t        thr_act);
-extern void            fp_load(
-                               thread_t        thr_act);
-
-static void configure_mxcsr_capability_mask(struct x86_avx_thread_state *fps);
+extern void             fpinit(void);
+extern void             fp_save(
+       thread_t        thr_act);
+extern void             fp_load(
+       thread_t        thr_act);
 
 
-struct x86_avx_thread_state initial_fp_state __attribute((aligned(64)));
+static void configure_mxcsr_capability_mask(x86_ext_thread_state_t *fps);
+static xstate_t thread_xstate(thread_t);
 
 
+x86_ext_thread_state_t  initial_fp_state __attribute((aligned(64)));
+x86_ext_thread_state_t  default_avx512_state __attribute((aligned(64)));
+x86_ext_thread_state_t  default_avx_state __attribute((aligned(64)));
+x86_ext_thread_state_t  default_fx_state __attribute((aligned(64)));
 
 /* Global MXCSR capability bitmask */
 static unsigned int mxcsr_capability_mask;
 
 
 /* Global MXCSR capability bitmask */
 static unsigned int mxcsr_capability_mask;
 
-#define        fninit() \
+#define fninit() \
        __asm__ volatile("fninit")
 
        __asm__ volatile("fninit")
 
-#define        fnstcw(control) \
+#define fnstcw(control) \
        __asm__("fnstcw %0" : "=m" (*(unsigned short *)(control)))
 
        __asm__("fnstcw %0" : "=m" (*(unsigned short *)(control)))
 
-#define        fldcw(control) \
+#define fldcw(control) \
        __asm__ volatile("fldcw %0" : : "m" (*(unsigned short *) &(control)) )
 
        __asm__ volatile("fldcw %0" : : "m" (*(unsigned short *) &(control)) )
 
-#define        fnclex() \
+#define fnclex() \
        __asm__ volatile("fnclex")
 
        __asm__ volatile("fnclex")
 
-#define        fnsave(state)  \
+#define fnsave(state)  \
        __asm__ volatile("fnsave %0" : "=m" (*state))
 
        __asm__ volatile("fnsave %0" : "=m" (*state))
 
-#define        frstor(state) \
+#define frstor(state) \
        __asm__ volatile("frstor %0" : : "m" (state))
 
 #define fwait() \
        __asm__ volatile("frstor %0" : : "m" (state))
 
 #define fwait() \
-       __asm__("fwait");
+       __asm__("fwait");
+
+static inline void
+fxrstor(struct x86_fx_thread_state *a)
+{
+       __asm__ __volatile__ ("fxrstor %0" ::  "m" (*a));
+}
+
+static inline void
+fxsave(struct x86_fx_thread_state *a)
+{
+       __asm__ __volatile__ ("fxsave %0" : "=m" (*a));
+}
 
 
-#define fxrstor(addr)           __asm__ __volatile__("fxrstor %0" : : "m" (*(addr)))     
-#define fxsave(addr)            __asm__ __volatile__("fxsave %0" : "=m" (*(addr)))
+static inline void
+fxrstor64(struct x86_fx_thread_state *a)
+{
+       __asm__ __volatile__ ("fxrstor64 %0" ::  "m" (*a));
+}
 
 
-static uint32_t        fp_register_state_size = 0;
-static uint32_t fpu_YMM_present        = FALSE;
-static uint32_t        cpuid_reevaluated = 0;
+static inline void
+fxsave64(struct x86_fx_thread_state *a)
+{
+       __asm__ __volatile__ ("fxsave64 %0" : "=m" (*a));
+}
+
+#if !defined(RC_HIDE_XNU_J137)
+#define IS_VALID_XSTATE(x)      ((x) == FP || (x) == AVX || (x) == AVX512)
+#else
+#define IS_VALID_XSTATE(x)      ((x) == FP || (x) == AVX)
+#endif
+
+zone_t          ifps_zone[] = {
+       [FP]     = NULL,
+       [AVX]    = NULL,
+#if !defined(RC_HIDE_XNU_J137)
+       [AVX512] = NULL
+#endif
+};
+static uint32_t fp_state_size[] = {
+       [FP]     = sizeof(struct x86_fx_thread_state),
+       [AVX]    = sizeof(struct x86_avx_thread_state),
+#if !defined(RC_HIDE_XNU_J137)
+       [AVX512] = sizeof(struct x86_avx512_thread_state)
+#endif
+};
+
+static const char *xstate_name[] = {
+       [UNDEFINED] = "UNDEFINED",
+       [FP] = "FP",
+       [AVX] = "AVX",
+#if !defined(RC_HIDE_XNU_J137)
+       [AVX512] = "AVX512"
+#endif
+};
+
+#if !defined(RC_HIDE_XNU_J137)
+#define fpu_ZMM_capable (fpu_capability == AVX512)
+#define fpu_YMM_capable (fpu_capability == AVX || fpu_capability == AVX512)
+/*
+ * On-demand AVX512 support
+ * ------------------------
+ * On machines with AVX512 support, by default, threads are created with
+ * AVX512 masked off in XCR0 and an AVX-sized savearea is used. However, AVX512
+ * capabilities are advertised in the commpage and via sysctl. If a thread
+ * opts to use AVX512 instructions, the first will result in a #UD exception.
+ * Faulting AVX512 intructions are recognizable by their unique prefix.
+ * This exception results in the thread being promoted to use an AVX512-sized
+ * savearea and for the AVX512 bit masks being set in its XCR0. The faulting
+ * instruction is re-driven and the thread can proceed to perform AVX512
+ * operations.
+ *
+ * In addition to AVX512 instructions causing promotion, the thread_set_state()
+ * primitive with an AVX512 state flavor result in promotion.
+ *
+ * AVX512 promotion of the first thread in a task causes the default xstate
+ * of the task to be promoted so that any subsequently created or subsequently
+ * DNA-faulted thread will have AVX512 xstate and it will not need to fault-in
+ * a promoted xstate.
+ *
+ * Two savearea zones are used: the default pool of AVX-sized (832 byte) areas
+ * and a second pool of larger AVX512-sized (2688 byte) areas.
+ *
+ * Note the initial state value is an AVX512 object but that the AVX initial
+ * value is a subset of it.
+ */
+#else
+#define fpu_YMM_capable (fpu_capability == AVX)
+#endif
+static uint32_t cpuid_reevaluated = 0;
 
 static void fpu_store_registers(void *, boolean_t);
 static void fpu_load_registers(void *);
 
 
 static void fpu_store_registers(void *, boolean_t);
 static void fpu_load_registers(void *);
 
-extern void xsave64o(void);
-extern void xrstor64o(void);
+#if !defined(RC_HIDE_XNU_J137)
+static const uint32_t xstate_xmask[] = {
+       [FP] =          FP_XMASK,
+       [AVX] =         AVX_XMASK,
+       [AVX512] =      AVX512_XMASK
+};
+#else
+static const uint32_t xstate_xmask[] = {
+       [FP] =          FP_XMASK,
+       [AVX] =         AVX_XMASK,
+};
+#endif
 
 
-#define XMASK ((uint32_t) (XFEM_X87 | XFEM_SSE | XFEM_YMM))
+static inline void
+xsave(struct x86_fx_thread_state *a, uint32_t rfbm)
+{
+       __asm__ __volatile__ ("xsave %0" :"=m" (*a) : "a"(rfbm), "d"(0));
+}
 
 
-static inline void xsetbv(uint32_t mask_hi, uint32_t mask_lo) {
-       __asm__ __volatile__("xsetbv" :: "a"(mask_lo), "d"(mask_hi), "c" (XCR0));
+static inline void
+xsave64(struct x86_fx_thread_state *a, uint32_t rfbm)
+{
+       __asm__ __volatile__ ("xsave64 %0" :"=m" (*a) : "a"(rfbm), "d"(0));
 }
 
 }
 
-static inline void xsave(struct x86_fx_thread_state *a) {
-       __asm__ __volatile__("xsave %0" :"=m" (*a) : "a"(XMASK), "d"(0));
+static inline void
+xrstor(struct x86_fx_thread_state *a, uint32_t rfbm)
+{
+       __asm__ __volatile__ ("xrstor %0" ::  "m" (*a), "a"(rfbm), "d"(0));
+}
+
+static inline void
+xrstor64(struct x86_fx_thread_state *a, uint32_t rfbm)
+{
+       __asm__ __volatile__ ("xrstor64 %0" ::  "m" (*a), "a"(rfbm), "d"(0));
+}
+
+#if !defined(RC_HIDE_XNU_J137)
+__unused static inline void
+vzeroupper(void)
+{
+       __asm__ __volatile__ ("vzeroupper" ::);
+}
+
+static boolean_t fpu_thread_promote_avx512(thread_t);   /* Forward */
+
+/*
+ * Define a wrapper for bcopy to defeat destination size checka.
+ * This is needed to treat repeated objects such as
+ *     _STRUCT_XMM_REG         fpu_ymmh0;
+ *     ...
+ *     _STRUCT_XMM_REG         fpu_ymmh7;
+ * as an array and to copy like so:
+ *     bcopy_nockch(src,&dst->fpu_ymmh0,8*sizeof(_STRUCT_XMM_REG));
+ * without the compiler throwing a __builtin__memmove_chk error.
+ */
+static inline void
+bcopy_nochk(void *_src, void *_dst, size_t _len)
+{
+       bcopy(_src, _dst, _len);
 }
 
 }
 
-static inline void xrstor(struct x86_fx_thread_state *a) {
-       __asm__ __volatile__("xrstor %0" ::  "m" (*a), "a"(XMASK), "d"(0));
+/*
+ * Furthermore, make compile-time asserts that no padding creeps into structures
+ * for which we're doing this.
+ */
+#define ASSERT_PACKED(t, m1, m2, n, mt)                 \
+extern char assert_packed_ ## t ## _ ## m1 ## _ ## m2   \
+       [(offsetof(t,m2) - offsetof(t,m1) == (n - 1)*sizeof(mt)) ? 1 : -1]
+
+ASSERT_PACKED(x86_avx_state32_t, fpu_ymmh0, fpu_ymmh7, 8, _STRUCT_XMM_REG);
+
+ASSERT_PACKED(x86_avx_state64_t, fpu_ymmh0, fpu_ymmh15, 16, _STRUCT_XMM_REG);
+
+ASSERT_PACKED(x86_avx512_state32_t, fpu_k0, fpu_k7, 8, _STRUCT_OPMASK_REG);
+ASSERT_PACKED(x86_avx512_state32_t, fpu_ymmh0, fpu_ymmh7, 8, _STRUCT_XMM_REG);
+ASSERT_PACKED(x86_avx512_state32_t, fpu_zmmh0, fpu_zmmh7, 8, _STRUCT_YMM_REG);
+
+ASSERT_PACKED(x86_avx512_state64_t, fpu_k0, fpu_k7, 8, _STRUCT_OPMASK_REG);
+ASSERT_PACKED(x86_avx512_state64_t, fpu_ymmh0, fpu_ymmh15, 16, _STRUCT_XMM_REG);
+ASSERT_PACKED(x86_avx512_state64_t, fpu_zmmh0, fpu_zmmh15, 16, _STRUCT_YMM_REG);
+ASSERT_PACKED(x86_avx512_state64_t, fpu_zmm16, fpu_zmm31, 16, _STRUCT_ZMM_REG);
+
+#if defined(DEBUG_AVX512)
+
+#define DBG(x...)       kprintf("DBG: " x)
+
+typedef struct { uint8_t byte[8]; }  opmask_t;
+typedef struct { uint8_t byte[16]; } xmm_t;
+typedef struct { uint8_t byte[32]; } ymm_t;
+typedef struct { uint8_t byte[64]; } zmm_t;
+
+static void
+DBG_AVX512_STATE(struct x86_avx512_thread_state *sp)
+{
+       int     i, j;
+       xmm_t *xmm  = (xmm_t *) &sp->fp.fx_XMM_reg;
+       xmm_t *ymmh = (xmm_t *) &sp->x_YMM_Hi128;
+       ymm_t *zmmh = (ymm_t *) &sp->x_ZMM_Hi256;
+       zmm_t *zmm  = (zmm_t *) &sp->x_Hi16_ZMM;
+       opmask_t *k = (opmask_t *) &sp->x_Opmask;
+
+       kprintf("x_YMM_Hi128: %lu\n", offsetof(struct x86_avx512_thread_state, x_YMM_Hi128));
+       kprintf("x_Opmask:    %lu\n", offsetof(struct x86_avx512_thread_state, x_Opmask));
+       kprintf("x_ZMM_Hi256: %lu\n", offsetof(struct x86_avx512_thread_state, x_ZMM_Hi256));
+       kprintf("x_Hi16_ZMM:  %lu\n", offsetof(struct x86_avx512_thread_state, x_Hi16_ZMM));
+
+       kprintf("XCR0:   0x%016llx\n", xgetbv(XCR0));
+       kprintf("XINUSE: 0x%016llx\n", xgetbv(1));
+
+       /* Print all ZMM registers */
+       for (i = 0; i < 16; i++) {
+               kprintf("zmm%d:\t0x", i);
+               for (j = 0; j < 16; j++) {
+                       kprintf("%02x", xmm[i].byte[j]);
+               }
+               for (j = 0; j < 16; j++) {
+                       kprintf("%02x", ymmh[i].byte[j]);
+               }
+               for (j = 0; j < 32; j++) {
+                       kprintf("%02x", zmmh[i].byte[j]);
+               }
+               kprintf("\n");
+       }
+       for (i = 0; i < 16; i++) {
+               kprintf("zmm%d:\t0x", 16 + i);
+               for (j = 0; j < 64; j++) {
+                       kprintf("%02x", zmm[i].byte[j]);
+               }
+               kprintf("\n");
+       }
+       for (i = 0; i < 8; i++) {
+               kprintf("k%d:\t0x", i);
+               for (j = 0; j < 8; j++) {
+                       kprintf("%02x", k[i].byte[j]);
+               }
+               kprintf("\n");
+       }
+
+       kprintf("xstate_bv: 0x%016llx\n", sp->_xh.xstate_bv);
+       kprintf("xcomp_bv:  0x%016llx\n", sp->_xh.xcomp_bv);
+}
+#else
+#define DBG(x...)
+static void
+DBG_AVX512_STATE(__unused struct x86_avx512_thread_state *sp)
+{
+       return;
 }
 }
+#endif /* DEBUG_AVX512 */
+
+#endif
 
 
-#if    DEBUG
+#if     DEBUG
 static inline unsigned short
 fnstsw(void)
 {
        unsigned short status;
 static inline unsigned short
 fnstsw(void)
 {
        unsigned short status;
-       __asm__ volatile("fnstsw %0" : "=ma" (status));
-       return(status);
+       __asm__ volatile ("fnstsw %0" : "=ma" (status));
+       return status;
 }
 #endif
 
 }
 #endif
 
@@ -167,7 +384,7 @@ fnstsw(void)
  */
 
 static void
  */
 
 static void
-configure_mxcsr_capability_mask(struct x86_avx_thread_state *fps)
+configure_mxcsr_capability_mask(x86_ext_thread_state_t *fps)
 {
        /* XSAVE requires a 64 byte aligned store */
        assert(ALIGNED(fps, 64));
 {
        /* XSAVE requires a 64 byte aligned store */
        assert(ALIGNED(fps, 64));
@@ -177,29 +394,46 @@ configure_mxcsr_capability_mask(struct x86_avx_thread_state *fps)
        fpinit();
        fpu_store_registers(fps, FALSE);
 
        fpinit();
        fpu_store_registers(fps, FALSE);
 
-       mxcsr_capability_mask = fps->fx_MXCSR_MASK;
+       mxcsr_capability_mask = fps->fx.fx_MXCSR_MASK;
 
        /* Set default mask value if necessary */
 
        /* Set default mask value if necessary */
-       if (mxcsr_capability_mask == 0)
+       if (mxcsr_capability_mask == 0) {
                mxcsr_capability_mask = 0xffbf;
                mxcsr_capability_mask = 0xffbf;
-       
+       }
+
        /* Clear vector register store */
        /* Clear vector register store */
-       bzero(&fps->fx_XMM_reg[0][0], sizeof(fps->fx_XMM_reg));
-       bzero(&fps->x_YMMH_reg[0][0], sizeof(fps->x_YMMH_reg));
+       bzero(&fps->fx.fx_XMM_reg[0][0], sizeof(fps->fx.fx_XMM_reg));
+       bzero(fps->avx.x_YMM_Hi128, sizeof(fps->avx.x_YMM_Hi128));
+#if !defined(RC_HIDE_XNU_J137)
+       if (fpu_ZMM_capable) {
+               bzero(fps->avx512.x_ZMM_Hi256, sizeof(fps->avx512.x_ZMM_Hi256));
+               bzero(fps->avx512.x_Hi16_ZMM, sizeof(fps->avx512.x_Hi16_ZMM));
+               bzero(fps->avx512.x_Opmask, sizeof(fps->avx512.x_Opmask));
+       }
+#endif
 
 
-       fps->fp_valid = TRUE;
-       fps->fp_save_layout = fpu_YMM_present ? XSAVE32: FXSAVE32;
+       fps->fx.fp_valid = TRUE;
+       fps->fx.fp_save_layout = fpu_YMM_capable ? XSAVE32: FXSAVE32;
        fpu_load_registers(fps);
 
        fpu_load_registers(fps);
 
+       if (fpu_ZMM_capable) {
+               xsave64((struct x86_fx_thread_state *)&default_avx512_state, xstate_xmask[AVX512]);
+       }
+       if (fpu_YMM_capable) {
+               xsave64((struct x86_fx_thread_state *)&default_avx_state, xstate_xmask[AVX]);
+       } else {
+               fxsave64((struct x86_fx_thread_state *)&default_fx_state);
+       }
+
        /* Poison values to trap unsafe usage */
        /* Poison values to trap unsafe usage */
-       fps->fp_valid = 0xFFFFFFFF;
-       fps->fp_save_layout = FP_UNUSED;
+       fps->fx.fp_valid = 0xFFFFFFFF;
+       fps->fx.fp_save_layout = FP_UNUSED;
 
        /* Re-enable FPU/SSE DNA exceptions */
        set_ts();
 }
 
 
        /* Re-enable FPU/SSE DNA exceptions */
        set_ts();
 }
 
-
+int fpsimd_fault_popc = 0;
 /*
  * Look for FPU and initialize it.
  * Called on each CPU.
 /*
  * Look for FPU and initialize it.
  * Called on each CPU.
@@ -207,59 +441,101 @@ configure_mxcsr_capability_mask(struct x86_avx_thread_state *fps)
 void
 init_fpu(void)
 {
 void
 init_fpu(void)
 {
-#if    DEBUG   
-       unsigned short  status;
-       unsigned short  control;
+#if     DEBUG
+       unsigned short  status;
+       unsigned short  control;
 #endif
        /*
         * Check for FPU by initializing it,
         * then trying to read the correct bit patterns from
         * the control and status registers.
         */
 #endif
        /*
         * Check for FPU by initializing it,
         * then trying to read the correct bit patterns from
         * the control and status registers.
         */
-       set_cr0((get_cr0() & ~(CR0_EM|CR0_TS)) | CR0_NE);       /* allow use of FPU */
+       set_cr0((get_cr0() & ~(CR0_EM | CR0_TS)) | CR0_NE);       /* allow use of FPU */
        fninit();
        fninit();
-#if    DEBUG   
+#if     DEBUG
        status = fnstsw();
        fnstcw(&control);
        status = fnstsw();
        fnstcw(&control);
-       
+
        assert(((status & 0xff) == 0) && ((control & 0x103f) == 0x3f));
 #endif
        /* Advertise SSE support */
        if (cpuid_features() & CPUID_FEATURE_FXSR) {
        assert(((status & 0xff) == 0) && ((control & 0x103f) == 0x3f));
 #endif
        /* Advertise SSE support */
        if (cpuid_features() & CPUID_FEATURE_FXSR) {
-               fp_kind = FP_FXSR;
                set_cr4(get_cr4() | CR4_OSFXS);
                /* And allow SIMD exceptions if present */
                if (cpuid_features() & CPUID_FEATURE_SSE) {
                        set_cr4(get_cr4() | CR4_OSXMM);
                }
                set_cr4(get_cr4() | CR4_OSFXS);
                /* And allow SIMD exceptions if present */
                if (cpuid_features() & CPUID_FEATURE_SSE) {
                        set_cr4(get_cr4() | CR4_OSXMM);
                }
-               fp_register_state_size = sizeof(struct x86_fx_thread_state);
-
-       } else
+       } else {
                panic("fpu is not FP_FXSR");
                panic("fpu is not FP_FXSR");
+       }
+
+       fpu_capability = fpu_default = FP;
+
+       PE_parse_boot_argn("fpsimd_fault_popc", &fpsimd_fault_popc, sizeof(fpsimd_fault_popc));
+
+#if !defined(RC_HIDE_XNU_J137)
+       static boolean_t is_avx512_enabled = TRUE;
+       if (cpu_number() == master_cpu) {
+               if (cpuid_leaf7_features() & CPUID_LEAF7_FEATURE_AVX512F) {
+                       PE_parse_boot_argn("avx512", &is_avx512_enabled, sizeof(boolean_t));
+                       kprintf("AVX512 supported %s\n",
+                           is_avx512_enabled ? "and enabled" : "but disabled");
+               }
+       }
+#endif
 
        /* Configure the XSAVE context mechanism if the processor supports
         * AVX/YMM registers
         */
        if (cpuid_features() & CPUID_FEATURE_XSAVE) {
 
        /* Configure the XSAVE context mechanism if the processor supports
         * AVX/YMM registers
         */
        if (cpuid_features() & CPUID_FEATURE_XSAVE) {
-               cpuid_xsave_leaf_t *xsp = &cpuid_info()->cpuid_xsave_leaf[0];
-               if (xsp->extended_state[0] & (uint32_t)XFEM_YMM) {
-                       assert(xsp->extended_state[0] & (uint32_t) XFEM_SSE);
+               cpuid_xsave_leaf_t *xs0p = &cpuid_info()->cpuid_xsave_leaf[0];
+#if !defined(RC_HIDE_XNU_J137)
+               if (is_avx512_enabled &&
+                   (xs0p->extended_state[eax] & XFEM_ZMM) == XFEM_ZMM) {
+                       assert(xs0p->extended_state[eax] & XFEM_SSE);
+                       assert(xs0p->extended_state[eax] & XFEM_YMM);
+                       fpu_capability = AVX512;
                        /* XSAVE container size for all features */
                        /* XSAVE container size for all features */
-                       fp_register_state_size = sizeof(struct x86_avx_thread_state);
-                       fpu_YMM_present = TRUE;
                        set_cr4(get_cr4() | CR4_OSXSAVE);
                        set_cr4(get_cr4() | CR4_OSXSAVE);
-                       xsetbv(0, XMASK);
+                       xsetbv(0, AVX512_XMASK);
                        /* Re-evaluate CPUID, once, to reflect OSXSAVE */
                        /* Re-evaluate CPUID, once, to reflect OSXSAVE */
-                       if (OSCompareAndSwap(0, 1, &cpuid_reevaluated))
+                       if (OSCompareAndSwap(0, 1, &cpuid_reevaluated)) {
                                cpuid_set_info();
                                cpuid_set_info();
+                       }
                        /* Verify that now selected state can be accommodated */
                        /* Verify that now selected state can be accommodated */
-                       assert(xsp->extended_state[1] == fp_register_state_size);
+                       assert(xs0p->extended_state[ebx] == fp_state_size[AVX512]);
+                       /*
+                        * AVX set until AVX512 is used.
+                        * See comment above about on-demand AVX512 support.
+                        */
+                       xsetbv(0, AVX_XMASK);
+                       fpu_default = AVX;
+               } else
+#endif
+               if (xs0p->extended_state[eax] & XFEM_YMM) {
+                       assert(xs0p->extended_state[eax] & XFEM_SSE);
+                       fpu_capability = AVX;
+                       fpu_default = AVX;
+                       /* XSAVE container size for all features */
+                       set_cr4(get_cr4() | CR4_OSXSAVE);
+                       xsetbv(0, AVX_XMASK);
+                       /* Re-evaluate CPUID, once, to reflect OSXSAVE */
+                       if (OSCompareAndSwap(0, 1, &cpuid_reevaluated)) {
+                               cpuid_set_info();
+                       }
+                       /* Verify that now selected state can be accommodated */
+                       assert(xs0p->extended_state[ebx] == fp_state_size[AVX]);
                }
        }
                }
        }
-       else
-               fpu_YMM_present = FALSE;
+
+       if (cpu_number() == master_cpu) {
+               kprintf("fpu_state: %s, state_size: %d\n",
+                   xstate_name[fpu_capability],
+                   fp_state_size[fpu_capability]);
+       }
 
        fpinit();
 
        fpinit();
+       current_cpu_datap()->cpu_xstate = fpu_default;
 
        /*
         * Trap wait instructions.  Turn off FPU for now.
 
        /*
         * Trap wait instructions.  Turn off FPU for now.
@@ -268,76 +544,126 @@ init_fpu(void)
 }
 
 /*
 }
 
 /*
- * Allocate and initialize FP state for current thread.
+ * Allocate and initialize FP state for specified xstate.
  * Don't load state.
  */
 static void *
  * Don't load state.
  */
 static void *
-fp_state_alloc(void)
+fp_state_alloc(xstate_t xs)
 {
 {
-       struct x86_fx_thread_state *ifps = zalloc(ifps_zone);
+       struct x86_fx_thread_state *ifps;
+
+       assert(ifps_zone[xs] != NULL);
+       ifps = zalloc(ifps_zone[xs]);
 
 
-#if    DEBUG   
-       if (!(ALIGNED(ifps,64))) {
-               panic("fp_state_alloc: %p, %u, %p, %u", ifps, (unsigned) ifps_zone->elem_size, (void *) ifps_zone->free_elements, (unsigned) ifps_zone->alloc_size);
+#if     DEBUG
+       if (!(ALIGNED(ifps, 64))) {
+               panic("fp_state_alloc: %p, %u, %p, %u",
+                   ifps, (unsigned) ifps_zone[xs]->elem_size,
+                   (void *) ifps_zone[xs]->free_elements,
+                   (unsigned) ifps_zone[xs]->alloc_size);
        }
 #endif
        }
 #endif
-       bzero(ifps, sizeof(*ifps));
+       bzero(ifps, fp_state_size[xs]);
+
        return ifps;
 }
 
 static inline void
        return ifps;
 }
 
 static inline void
-fp_state_free(void *ifps)
+fp_state_free(void *ifps, xstate_t xs)
 {
 {
-       zfree(ifps_zone, ifps);
+       assert(ifps_zone[xs] != NULL);
+       zfree(ifps_zone[xs], ifps);
 }
 
 }
 
-void clear_fpu(void)
+void
+clear_fpu(void)
 {
        set_ts();
 }
 
 
 {
        set_ts();
 }
 
 
-static void fpu_load_registers(void *fstate) {
+static void
+fpu_load_registers(void *fstate)
+{
        struct x86_fx_thread_state *ifps = fstate;
        fp_save_layout_t layout = ifps->fp_save_layout;
 
        struct x86_fx_thread_state *ifps = fstate;
        fp_save_layout_t layout = ifps->fp_save_layout;
 
-       assert(layout == FXSAVE32 || layout == FXSAVE64 || layout == XSAVE32 || layout == XSAVE64);
+       assert(current_task() == NULL ||                                \
+           (thread_is_64bit_addr(current_thread()) ?                        \
+           (layout == FXSAVE64 || layout == XSAVE64) :     \
+           (layout == FXSAVE32 || layout == XSAVE32)));
        assert(ALIGNED(ifps, 64));
        assert(ml_get_interrupts_enabled() == FALSE);
 
        assert(ALIGNED(ifps, 64));
        assert(ml_get_interrupts_enabled() == FALSE);
 
-#if    DEBUG   
+#if     DEBUG
        if (layout == XSAVE32 || layout == XSAVE64) {
                struct x86_avx_thread_state *iavx = fstate;
                unsigned i;
                /* Verify reserved bits in the XSAVE header*/
        if (layout == XSAVE32 || layout == XSAVE64) {
                struct x86_avx_thread_state *iavx = fstate;
                unsigned i;
                /* Verify reserved bits in the XSAVE header*/
-               if (iavx->_xh.xsbv & ~7)
-                       panic("iavx->_xh.xsbv: 0x%llx", iavx->_xh.xsbv);
-               for (i = 0; i < sizeof(iavx->_xh.xhrsvd); i++)
-                       if (iavx->_xh.xhrsvd[i])
+               if (iavx->_xh.xstate_bv & ~xstate_xmask[current_xstate()]) {
+                       panic("iavx->_xh.xstate_bv: 0x%llx", iavx->_xh.xstate_bv);
+               }
+               for (i = 0; i < sizeof(iavx->_xh.xhrsvd); i++) {
+                       if (iavx->_xh.xhrsvd[i]) {
                                panic("Reserved bit set");
                                panic("Reserved bit set");
+                       }
+               }
        }
        }
-       if (fpu_YMM_present) {
-               if (layout != XSAVE32 && layout != XSAVE64)
+       if (fpu_YMM_capable) {
+               if (layout != XSAVE32 && layout != XSAVE64) {
                        panic("Inappropriate layout: %u\n", layout);
                        panic("Inappropriate layout: %u\n", layout);
+               }
        }
        }
-#endif /* DEBUG */
+#endif  /* DEBUG */
 
 
-       if ((layout == XSAVE64) || (layout == XSAVE32))
-               xrstor(ifps);
-       else
+       switch (layout) {
+       case FXSAVE64:
+               fxrstor64(ifps);
+               break;
+       case FXSAVE32:
                fxrstor(ifps);
                fxrstor(ifps);
+               break;
+       case XSAVE64:
+               xrstor64(ifps, xstate_xmask[current_xstate()]);
+               break;
+       case XSAVE32:
+               xrstor(ifps, xstate_xmask[current_xstate()]);
+               break;
+       default:
+               panic("fpu_load_registers() bad layout: %d\n", layout);
+       }
 }
 
 }
 
-static void fpu_store_registers(void *fstate, boolean_t is64) {
+static void
+fpu_store_registers(void *fstate, boolean_t is64)
+{
        struct x86_fx_thread_state *ifps = fstate;
        assert(ALIGNED(ifps, 64));
        struct x86_fx_thread_state *ifps = fstate;
        assert(ALIGNED(ifps, 64));
-       if (fpu_YMM_present) {
-               xsave(ifps);
-               ifps->fp_save_layout = is64 ? XSAVE64 : XSAVE32;
-       }
-       else {
-               fxsave(ifps);
-               ifps->fp_save_layout = is64 ? FXSAVE64 : FXSAVE32;
+       xstate_t xs = current_xstate();
+       switch (xs) {
+       case FP:
+               if (is64) {
+                       fxsave64(fstate);
+                       ifps->fp_save_layout = FXSAVE64;
+               } else {
+                       fxsave(fstate);
+                       ifps->fp_save_layout = FXSAVE32;
+               }
+               break;
+       case AVX:
+#if !defined(RC_HIDE_XNU_J137)
+       case AVX512:
+#endif
+               if (is64) {
+                       xsave64(ifps, xstate_xmask[xs]);
+                       ifps->fp_save_layout = XSAVE64;
+               } else {
+                       xsave(ifps, xstate_xmask[xs]);
+                       ifps->fp_save_layout = XSAVE32;
+               }
+               break;
+       default:
+               panic("fpu_store_registers() bad xstate: %d\n", xs);
        }
 }
 
        }
 }
 
@@ -348,41 +674,60 @@ static void fpu_store_registers(void *fstate, boolean_t is64) {
 void
 fpu_module_init(void)
 {
 void
 fpu_module_init(void)
 {
-       if ((fp_register_state_size != sizeof(struct x86_fx_thread_state)) &&
-           (fp_register_state_size != sizeof(struct x86_avx_thread_state)))
-               panic("fpu_module_init: incorrect savearea size %u\n", fp_register_state_size);
-
-       assert(fpu_YMM_present != 0xFFFFFFFF);
+       if (!IS_VALID_XSTATE(fpu_default)) {
+               panic("fpu_module_init: invalid extended state %u\n",
+                   fpu_default);
+       }
 
 
-       /* We explicitly choose an allocation size of 64
+       /* We explicitly choose an allocation size of 13 pages = 64 * 832
         * to eliminate waste for the 832 byte sized
         * AVX XSAVE register save area.
         */
         * to eliminate waste for the 832 byte sized
         * AVX XSAVE register save area.
         */
-       ifps_zone = zinit(fp_register_state_size,
-                         thread_max * fp_register_state_size,
-                         64 * fp_register_state_size,
-                         "x86 fpsave state");
+       ifps_zone[fpu_default] = zinit(fp_state_size[fpu_default],
+           thread_max * fp_state_size[fpu_default],
+           64 * fp_state_size[fpu_default],
+           "x86 fpsave state");
 
        /* To maintain the required alignment, disable
         * zone debugging for this zone as that appends
         * 16 bytes to each element.
         */
 
        /* To maintain the required alignment, disable
         * zone debugging for this zone as that appends
         * 16 bytes to each element.
         */
-       zone_change(ifps_zone, Z_ALIGNMENT_REQUIRED, TRUE);
+       zone_change(ifps_zone[fpu_default], Z_ALIGNMENT_REQUIRED, TRUE);
+
+#if !defined(RC_HIDE_XNU_J137)
+       /*
+        * If AVX512 is supported, create a separate savearea zone.
+        * with allocation size: 19 pages = 32 * 2668
+        */
+       if (fpu_capability == AVX512) {
+               ifps_zone[AVX512] = zinit(fp_state_size[AVX512],
+                   thread_max * fp_state_size[AVX512],
+                   32 * fp_state_size[AVX512],
+                   "x86 avx512 save state");
+               zone_change(ifps_zone[AVX512], Z_ALIGNMENT_REQUIRED, TRUE);
+       }
+#endif
+
        /* Determine MXCSR reserved bits and configure initial FPU state*/
        configure_mxcsr_capability_mask(&initial_fp_state);
 }
 
 /*
        /* Determine MXCSR reserved bits and configure initial FPU state*/
        configure_mxcsr_capability_mask(&initial_fp_state);
 }
 
 /*
- * Save thread`s FPU context.
+ * Context switch fpu state.
+ * Always save old thread`s FPU context but don't load new .. allow that to fault-in.
+ * Switch to the new task's xstate.
  */
  */
+
 void
 void
-fpu_save_context(thread_t thread)
+fpu_switch_context(thread_t old, thread_t new)
 {
 {
-       struct x86_fx_thread_state *ifps;
+       struct x86_fx_thread_state      *ifps;
+       cpu_data_t *cdp = current_cpu_datap();
+       xstate_t new_xstate = new ? thread_xstate(new) : fpu_default;
 
        assert(ml_get_interrupts_enabled() == FALSE);
 
        assert(ml_get_interrupts_enabled() == FALSE);
-       ifps = (thread)->machine.ifps;
-#if    DEBUG
+       ifps = (old)->machine.ifps;
+#if     DEBUG
        if (ifps && ((ifps->fp_valid != FALSE) && (ifps->fp_valid != TRUE))) {
                panic("ifps->fp_valid: %u\n", ifps->fp_valid);
        }
        if (ifps && ((ifps->fp_valid != FALSE) && (ifps->fp_valid != TRUE))) {
                panic("ifps->fp_valid: %u\n", ifps->fp_valid);
        }
@@ -395,8 +740,27 @@ fpu_save_context(thread_t thread)
                 */
                clear_ts();
                /* registers are in FPU - save to memory */
                 */
                clear_ts();
                /* registers are in FPU - save to memory */
-               fpu_store_registers(ifps, (thread_is_64bit(thread) && is_saved_state64(thread->machine.iss)));
+               boolean_t is64 = (thread_is_64bit_addr(old) &&
+                   is_saved_state64(old->machine.iss));
+
+               fpu_store_registers(ifps, is64);
                ifps->fp_valid = TRUE;
                ifps->fp_valid = TRUE;
+
+               if (fpu_ZMM_capable && (cdp->cpu_xstate == AVX512)) {
+                       xrstor64((struct x86_fx_thread_state *)&default_avx512_state, xstate_xmask[AVX512]);
+               } else if (fpu_YMM_capable) {
+                       xrstor64((struct x86_fx_thread_state *) &default_avx_state, xstate_xmask[AVX]);
+               } else {
+                       fxrstor64((struct x86_fx_thread_state *)&default_fx_state);
+               }
+       }
+
+       assertf(fpu_YMM_capable ? (xgetbv(XCR0) == xstate_xmask[cdp->cpu_xstate]) : TRUE, "XCR0 mismatch: 0x%llx 0x%x 0x%x", xgetbv(XCR0), cdp->cpu_xstate, xstate_xmask[cdp->cpu_xstate]);
+       if (new_xstate != (xstate_t) cdp->cpu_xstate) {
+               DBG("fpu_switch_context(%p,%p) new xstate: %s\n",
+                   old, new, xstate_name[new_xstate]);
+               xsetbv(0, xstate_xmask[new_xstate]);
+               cdp->cpu_xstate = new_xstate;
        }
        set_ts();
 }
        }
        set_ts();
 }
@@ -407,39 +771,52 @@ fpu_save_context(thread_t thread)
  * Called only when thread terminating - no locking necessary.
  */
 void
  * Called only when thread terminating - no locking necessary.
  */
 void
-fpu_free(void *fps)
+fpu_free(thread_t thread, void *fps)
 {
 {
-       fp_state_free(fps);
+       pcb_t   pcb = THREAD_TO_PCB(thread);
+
+       fp_state_free(fps, pcb->xstate);
+       pcb->xstate = UNDEFINED;
 }
 
 /*
 }
 
 /*
- * Set the floating-point state for a thread based 
- * on the FXSave formatted data. This is basically  
- * the same as fpu_set_state except it uses the 
- * expanded data structure. 
+ * Set the floating-point state for a thread based
+ * on the FXSave formatted data. This is basically
+ * the same as fpu_set_state except it uses the
+ * expanded data structure.
  * If the thread is not the current thread, it is
  * not running (held).  Locking needed against
  * concurrent fpu_set_state or fpu_get_state.
  */
 kern_return_t
 fpu_set_fxstate(
  * If the thread is not the current thread, it is
  * not running (held).  Locking needed against
  * concurrent fpu_set_state or fpu_get_state.
  */
 kern_return_t
 fpu_set_fxstate(
-       thread_t        thr_act,
-       thread_state_t  tstate,
+       thread_t        thr_act,
+       thread_state_t  tstate,
        thread_flavor_t f)
 {
        thread_flavor_t f)
 {
-       struct x86_fx_thread_state *ifps;
-       struct x86_fx_thread_state *new_ifps;
-       x86_float_state64_t     *state;
-       pcb_t   pcb;
-       size_t  state_size = sizeof(struct x86_fx_thread_state);
-       boolean_t       old_valid, fresh_state = FALSE;
+       struct x86_fx_thread_state      *ifps;
+       struct x86_fx_thread_state      *new_ifps;
+       x86_float_state64_t             *state;
+       pcb_t                           pcb;
+       boolean_t                       old_valid, fresh_state = FALSE;
 
 
-       if (fp_kind == FP_NO)
+       if (fpu_capability == UNDEFINED) {
                return KERN_FAILURE;
                return KERN_FAILURE;
+       }
 
        if ((f == x86_AVX_STATE32 || f == x86_AVX_STATE64) &&
 
        if ((f == x86_AVX_STATE32 || f == x86_AVX_STATE64) &&
-           !ml_fpu_avx_enabled())
+           fpu_capability < AVX) {
                return KERN_FAILURE;
                return KERN_FAILURE;
+       }
+
+#if !defined(RC_HIDE_XNU_J137)
+       if ((f == x86_AVX512_STATE32 || f == x86_AVX512_STATE64) &&
+           thread_xstate(thr_act) == AVX) {
+               if (!fpu_thread_promote_avx512(thr_act)) {
+                       return KERN_FAILURE;
+               }
+       }
+#endif
 
        state = (x86_float_state64_t *)tstate;
 
 
        state = (x86_float_state64_t *)tstate;
 
@@ -451,7 +828,7 @@ fpu_set_fxstate(
                 * new FPU state is 'invalid'.
                 * Deallocate the fp state if it exists.
                 */
                 * new FPU state is 'invalid'.
                 * Deallocate the fp state if it exists.
                 */
-               simple_lock(&pcb->lock);
+               simple_lock(&pcb->lock, LCK_GRP_NULL);
 
                ifps = pcb->ifps;
                pcb->ifps = 0;
 
                ifps = pcb->ifps;
                pcb->ifps = 0;
@@ -459,26 +836,27 @@ fpu_set_fxstate(
                simple_unlock(&pcb->lock);
 
                if (ifps != 0) {
                simple_unlock(&pcb->lock);
 
                if (ifps != 0) {
-                       fp_state_free(ifps);
+                       fp_state_free(ifps, thread_xstate(thr_act));
                }
        } else {
                /*
                 * Valid incoming state. Allocate the fp state if there is none.
                 */
                new_ifps = 0;
                }
        } else {
                /*
                 * Valid incoming state. Allocate the fp state if there is none.
                 */
                new_ifps = 0;
-               Retry:
-               simple_lock(&pcb->lock);
+Retry:
+               simple_lock(&pcb->lock, LCK_GRP_NULL);
 
                ifps = pcb->ifps;
                if (ifps == 0) {
                        if (new_ifps == 0) {
                                simple_unlock(&pcb->lock);
 
                ifps = pcb->ifps;
                if (ifps == 0) {
                        if (new_ifps == 0) {
                                simple_unlock(&pcb->lock);
-                               new_ifps = fp_state_alloc();
+                               new_ifps = fp_state_alloc(thread_xstate(thr_act));
                                goto Retry;
                        }
                        ifps = new_ifps;
                        new_ifps = 0;
                        pcb->ifps = ifps;
                                goto Retry;
                        }
                        ifps = new_ifps;
                        new_ifps = 0;
                        pcb->ifps = ifps;
+                       pcb->xstate = thread_xstate(thr_act);
                        fresh_state = TRUE;
                }
 
                        fresh_state = TRUE;
                }
 
@@ -488,7 +866,7 @@ fpu_set_fxstate(
 
                old_valid = ifps->fp_valid;
 
 
                old_valid = ifps->fp_valid;
 
-#if    DEBUG || DEVELOPMENT
+#if     DEBUG || DEVELOPMENT
                if ((fresh_state == FALSE) && (old_valid == FALSE) && (thr_act != current_thread())) {
                        panic("fpu_set_fxstate inconsistency, thread: %p not stopped", thr_act);
                }
                if ((fresh_state == FALSE) && (old_valid == FALSE) && (thr_act != current_thread())) {
                        panic("fpu_set_fxstate inconsistency, thread: %p not stopped", thr_act);
                }
@@ -500,48 +878,102 @@ fpu_set_fxstate(
 
                state->fpu_mxcsr &= mxcsr_capability_mask;
 
 
                state->fpu_mxcsr &= mxcsr_capability_mask;
 
-               bcopy((char *)&state->fpu_fcw, (char *)ifps, state_size);
-
-               if (fpu_YMM_present) {
+               bcopy((char *)&state->fpu_fcw, (char *)ifps, fp_state_size[FP]);
+
+               switch (thread_xstate(thr_act)) {
+               case UNDEFINED_FULL:
+               case FP_FULL:
+               case AVX_FULL:
+               case AVX512_FULL:
+                       panic("fpu_set_fxstate() INVALID xstate: 0x%x", thread_xstate(thr_act));
+                       break;
+
+               case UNDEFINED:
+                       panic("fpu_set_fxstate() UNDEFINED xstate");
+                       break;
+               case FP:
+                       ifps->fp_save_layout = thread_is_64bit_addr(thr_act) ? FXSAVE64 : FXSAVE32;
+                       break;
+               case AVX: {
                        struct x86_avx_thread_state *iavx = (void *) ifps;
                        struct x86_avx_thread_state *iavx = (void *) ifps;
-                       uint32_t fpu_nyreg = 0;
+                       x86_avx_state64_t *xs = (x86_avx_state64_t *) state;
 
 
-                       if (f == x86_AVX_STATE32)
-                               fpu_nyreg = 8;
-                       else if (f == x86_AVX_STATE64)
-                               fpu_nyreg = 16;
+                       iavx->fp.fp_save_layout = thread_is_64bit_addr(thr_act) ? XSAVE64 : XSAVE32;
 
 
-                       if (fpu_nyreg) {
-                               x86_avx_state64_t *ystate = (x86_avx_state64_t *) state;
-                               bcopy(&ystate->__fpu_ymmh0, &iavx->x_YMMH_reg[0][0], fpu_nyreg * sizeof(_STRUCT_XMM_REG));
+                       /* Sanitize XSAVE header */
+                       bzero(&iavx->_xh.xhrsvd[0], sizeof(iavx->_xh.xhrsvd));
+                       iavx->_xh.xstate_bv = AVX_XMASK;
+                       iavx->_xh.xcomp_bv  = 0;
+
+                       if (f == x86_AVX_STATE32) {
+                               bcopy_nochk(&xs->fpu_ymmh0, iavx->x_YMM_Hi128, 8 * sizeof(_STRUCT_XMM_REG));
+                       } else if (f == x86_AVX_STATE64) {
+                               bcopy_nochk(&xs->fpu_ymmh0, iavx->x_YMM_Hi128, 16 * sizeof(_STRUCT_XMM_REG));
+                       } else {
+                               iavx->_xh.xstate_bv = (XFEM_SSE | XFEM_X87);
                        }
                        }
+                       break;
+               }
+#if !defined(RC_HIDE_XNU_J137)
+               case AVX512: {
+                       struct x86_avx512_thread_state *iavx = (void *) ifps;
+                       union {
+                               thread_state_t       ts;
+                               x86_avx512_state32_t *s32;
+                               x86_avx512_state64_t *s64;
+                       } xs = { .ts = tstate };
+
+                       iavx->fp.fp_save_layout = thread_is_64bit_addr(thr_act) ? XSAVE64 : XSAVE32;
 
 
-                       iavx->fp_save_layout = thread_is_64bit(thr_act) ? XSAVE64 : XSAVE32;
                        /* Sanitize XSAVE header */
                        bzero(&iavx->_xh.xhrsvd[0], sizeof(iavx->_xh.xhrsvd));
                        /* Sanitize XSAVE header */
                        bzero(&iavx->_xh.xhrsvd[0], sizeof(iavx->_xh.xhrsvd));
-                       if (fpu_nyreg)
-                               iavx->_xh.xsbv = (XFEM_YMM | XFEM_SSE | XFEM_X87);
-                       else
-                               iavx->_xh.xsbv = (XFEM_SSE | XFEM_X87);
-               } else {
-                       ifps->fp_save_layout = thread_is_64bit(thr_act) ? FXSAVE64 : FXSAVE32;
+                       iavx->_xh.xstate_bv = AVX512_XMASK;
+                       iavx->_xh.xcomp_bv  = 0;
+
+                       switch (f) {
+                       case x86_AVX512_STATE32:
+                               bcopy_nochk(&xs.s32->fpu_k0, iavx->x_Opmask, 8 * sizeof(_STRUCT_OPMASK_REG));
+                               bcopy_nochk(&xs.s32->fpu_zmmh0, iavx->x_ZMM_Hi256, 8 * sizeof(_STRUCT_YMM_REG));
+                               bcopy_nochk(&xs.s32->fpu_ymmh0, iavx->x_YMM_Hi128, 8 * sizeof(_STRUCT_XMM_REG));
+                               DBG_AVX512_STATE(iavx);
+                               break;
+                       case x86_AVX_STATE32:
+                               bcopy_nochk(&xs.s32->fpu_ymmh0, iavx->x_YMM_Hi128, 8 * sizeof(_STRUCT_XMM_REG));
+                               break;
+                       case x86_AVX512_STATE64:
+                               bcopy_nochk(&xs.s64->fpu_k0, iavx->x_Opmask, 8 * sizeof(_STRUCT_OPMASK_REG));
+                               bcopy_nochk(&xs.s64->fpu_zmm16, iavx->x_Hi16_ZMM, 16 * sizeof(_STRUCT_ZMM_REG));
+                               bcopy_nochk(&xs.s64->fpu_zmmh0, iavx->x_ZMM_Hi256, 16 * sizeof(_STRUCT_YMM_REG));
+                               bcopy_nochk(&xs.s64->fpu_ymmh0, iavx->x_YMM_Hi128, 16 * sizeof(_STRUCT_XMM_REG));
+                               DBG_AVX512_STATE(iavx);
+                               break;
+                       case x86_AVX_STATE64:
+                               bcopy_nochk(&xs.s64->fpu_ymmh0, iavx->x_YMM_Hi128, 16 * sizeof(_STRUCT_XMM_REG));
+                               break;
+                       }
+                       break;
                }
                }
+#endif
+               }
+
                ifps->fp_valid = old_valid;
 
                if (old_valid == FALSE) {
                        boolean_t istate = ml_set_interrupts_enabled(FALSE);
                        ifps->fp_valid = TRUE;
                        /* If altering the current thread's state, disable FPU */
                ifps->fp_valid = old_valid;
 
                if (old_valid == FALSE) {
                        boolean_t istate = ml_set_interrupts_enabled(FALSE);
                        ifps->fp_valid = TRUE;
                        /* If altering the current thread's state, disable FPU */
-                       if (thr_act == current_thread())
+                       if (thr_act == current_thread()) {
                                set_ts();
                                set_ts();
+                       }
 
                        ml_set_interrupts_enabled(istate);
                }
 
                simple_unlock(&pcb->lock);
 
 
                        ml_set_interrupts_enabled(istate);
                }
 
                simple_unlock(&pcb->lock);
 
-               if (new_ifps != 0)
-                       fp_state_free(new_ifps);
+               if (new_ifps != 0) {
+                       fp_state_free(new_ifps, thread_xstate(thr_act));
+               }
        }
        return KERN_SUCCESS;
 }
        }
        return KERN_SUCCESS;
 }
@@ -554,29 +986,37 @@ fpu_set_fxstate(
  */
 kern_return_t
 fpu_get_fxstate(
  */
 kern_return_t
 fpu_get_fxstate(
-       thread_t        thr_act,
-       thread_state_t  tstate,
+       thread_t        thr_act,
+       thread_state_t  tstate,
        thread_flavor_t f)
 {
        thread_flavor_t f)
 {
-       struct x86_fx_thread_state      *ifps;
-       x86_float_state64_t     *state;
-       kern_return_t   ret = KERN_FAILURE;
-       pcb_t   pcb;
-       size_t  state_size = sizeof(struct x86_fx_thread_state);
+       struct x86_fx_thread_state      *ifps;
+       x86_float_state64_t             *state;
+       kern_return_t                   ret = KERN_FAILURE;
+       pcb_t                           pcb;
 
 
-       if (fp_kind == FP_NO)
+       if (fpu_capability == UNDEFINED) {
                return KERN_FAILURE;
                return KERN_FAILURE;
+       }
 
        if ((f == x86_AVX_STATE32 || f == x86_AVX_STATE64) &&
 
        if ((f == x86_AVX_STATE32 || f == x86_AVX_STATE64) &&
-           !ml_fpu_avx_enabled())
+           fpu_capability < AVX) {
                return KERN_FAILURE;
                return KERN_FAILURE;
+       }
+
+#if !defined(RC_HIDE_XNU_J137)
+       if ((f == x86_AVX512_STATE32 || f == x86_AVX512_STATE64) &&
+           thread_xstate(thr_act) != AVX512) {
+               return KERN_FAILURE;
+       }
+#endif
 
        state = (x86_float_state64_t *)tstate;
 
        assert(thr_act != THREAD_NULL);
        pcb = THREAD_TO_PCB(thr_act);
 
 
        state = (x86_float_state64_t *)tstate;
 
        assert(thr_act != THREAD_NULL);
        pcb = THREAD_TO_PCB(thr_act);
 
-       simple_lock(&pcb->lock);
+       simple_lock(&pcb->lock, LCK_GRP_NULL);
 
        ifps = pcb->ifps;
        if (ifps == 0) {
 
        ifps = pcb->ifps;
        if (ifps == 0) {
@@ -585,7 +1025,7 @@ fpu_get_fxstate(
                 */
 
                bcopy((char *)&initial_fp_state, (char *)&state->fpu_fcw,
                 */
 
                bcopy((char *)&initial_fp_state, (char *)&state->fpu_fcw,
-                   state_size);
+                   fp_state_size[FP]);
 
                simple_unlock(&pcb->lock);
 
 
                simple_unlock(&pcb->lock);
 
@@ -596,7 +1036,7 @@ fpu_get_fxstate(
         * If the live fpu state belongs to our target
         */
        if (thr_act == current_thread()) {
         * If the live fpu state belongs to our target
         */
        if (thr_act == current_thread()) {
-               boolean_t       intr;
+               boolean_t       intr;
 
                intr = ml_set_interrupts_enabled(FALSE);
 
 
                intr = ml_set_interrupts_enabled(FALSE);
 
@@ -607,20 +1047,62 @@ fpu_get_fxstate(
                (void)ml_set_interrupts_enabled(intr);
        }
        if (ifps->fp_valid) {
                (void)ml_set_interrupts_enabled(intr);
        }
        if (ifps->fp_valid) {
-               bcopy((char *)ifps, (char *)&state->fpu_fcw, state_size);
-               if (fpu_YMM_present) {
+               bcopy((char *)ifps, (char *)&state->fpu_fcw, fp_state_size[FP]);
+               switch (thread_xstate(thr_act)) {
+               case UNDEFINED_FULL:
+               case FP_FULL:
+               case AVX_FULL:
+               case AVX512_FULL:
+                       panic("fpu_get_fxstate() INVALID xstate: 0x%x", thread_xstate(thr_act));
+                       break;
+
+               case UNDEFINED:
+                       panic("fpu_get_fxstate() UNDEFINED xstate");
+                       break;
+               case FP:
+                       break;                  /* already done */
+               case AVX: {
                        struct x86_avx_thread_state *iavx = (void *) ifps;
                        struct x86_avx_thread_state *iavx = (void *) ifps;
-                       uint32_t fpu_nyreg = 0;
-
-                       if (f == x86_AVX_STATE32)
-                               fpu_nyreg = 8;
-                       else if (f == x86_AVX_STATE64)
-                               fpu_nyreg = 16;
-
-                       if (fpu_nyreg) {
-                               x86_avx_state64_t *ystate = (x86_avx_state64_t *) state;
-                               bcopy(&iavx->x_YMMH_reg[0][0], &ystate->__fpu_ymmh0, fpu_nyreg * sizeof(_STRUCT_XMM_REG));
+                       x86_avx_state64_t *xs = (x86_avx_state64_t *) state;
+                       if (f == x86_AVX_STATE32) {
+                               bcopy_nochk(iavx->x_YMM_Hi128, &xs->fpu_ymmh0, 8 * sizeof(_STRUCT_XMM_REG));
+                       } else if (f == x86_AVX_STATE64) {
+                               bcopy_nochk(iavx->x_YMM_Hi128, &xs->fpu_ymmh0, 16 * sizeof(_STRUCT_XMM_REG));
                        }
                        }
+                       break;
+               }
+#if !defined(RC_HIDE_XNU_J137)
+               case AVX512: {
+                       struct x86_avx512_thread_state *iavx = (void *) ifps;
+                       union {
+                               thread_state_t       ts;
+                               x86_avx512_state32_t *s32;
+                               x86_avx512_state64_t *s64;
+                       } xs = { .ts = tstate };
+                       switch (f) {
+                       case x86_AVX512_STATE32:
+                               bcopy_nochk(iavx->x_Opmask, &xs.s32->fpu_k0, 8 * sizeof(_STRUCT_OPMASK_REG));
+                               bcopy_nochk(iavx->x_ZMM_Hi256, &xs.s32->fpu_zmmh0, 8 * sizeof(_STRUCT_YMM_REG));
+                               bcopy_nochk(iavx->x_YMM_Hi128, &xs.s32->fpu_ymmh0, 8 * sizeof(_STRUCT_XMM_REG));
+                               DBG_AVX512_STATE(iavx);
+                               break;
+                       case x86_AVX_STATE32:
+                               bcopy_nochk(iavx->x_YMM_Hi128, &xs.s32->fpu_ymmh0, 8 * sizeof(_STRUCT_XMM_REG));
+                               break;
+                       case x86_AVX512_STATE64:
+                               bcopy_nochk(iavx->x_Opmask, &xs.s64->fpu_k0, 8 * sizeof(_STRUCT_OPMASK_REG));
+                               bcopy_nochk(iavx->x_Hi16_ZMM, &xs.s64->fpu_zmm16, 16 * sizeof(_STRUCT_ZMM_REG));
+                               bcopy_nochk(iavx->x_ZMM_Hi256, &xs.s64->fpu_zmmh0, 16 * sizeof(_STRUCT_YMM_REG));
+                               bcopy_nochk(iavx->x_YMM_Hi128, &xs.s64->fpu_ymmh0, 16 * sizeof(_STRUCT_XMM_REG));
+                               DBG_AVX512_STATE(iavx);
+                               break;
+                       case x86_AVX_STATE64:
+                               bcopy_nochk(iavx->x_YMM_Hi128, &xs.s64->fpu_ymmh0, 16 * sizeof(_STRUCT_XMM_REG));
+                               break;
+                       }
+                       break;
+               }
+#endif
                }
 
                ret = KERN_SUCCESS;
                }
 
                ret = KERN_SUCCESS;
@@ -641,44 +1123,49 @@ fpu_get_fxstate(
 
 void
 fpu_dup_fxstate(
 
 void
 fpu_dup_fxstate(
-       thread_t        parent,
-       thread_t        child)
+       thread_t        parent,
+       thread_t        child)
 {
        struct x86_fx_thread_state *new_ifps = NULL;
 {
        struct x86_fx_thread_state *new_ifps = NULL;
-       boolean_t       intr;
-       pcb_t           ppcb;
+       boolean_t       intr;
+       pcb_t           ppcb;
+       xstate_t        xstate = thread_xstate(parent);
 
        ppcb = THREAD_TO_PCB(parent);
 
 
        ppcb = THREAD_TO_PCB(parent);
 
-       if (ppcb->ifps == NULL)
-               return;
+       if (ppcb->ifps == NULL) {
+               return;
+       }
 
 
-        if (child->machine.ifps)
-               panic("fpu_dup_fxstate: child's ifps non-null");
+       if (child->machine.ifps) {
+               panic("fpu_dup_fxstate: child's ifps non-null");
+       }
 
 
-       new_ifps = fp_state_alloc();
+       new_ifps = fp_state_alloc(xstate);
 
 
-       simple_lock(&ppcb->lock);
+       simple_lock(&ppcb->lock, LCK_GRP_NULL);
 
        if (ppcb->ifps != NULL) {
                struct x86_fx_thread_state *ifps = ppcb->ifps;
 
        if (ppcb->ifps != NULL) {
                struct x86_fx_thread_state *ifps = ppcb->ifps;
-               /*
+               /*
                 * Make sure we`ve got the latest fp state info
                 */
                 * Make sure we`ve got the latest fp state info
                 */
-               intr = ml_set_interrupts_enabled(FALSE);
-               assert(current_thread() == parent);
-               clear_ts();
-               fp_save(parent);
-               clear_fpu();
-
-               (void)ml_set_interrupts_enabled(intr);
+               if (current_thread() == parent) {
+                       intr = ml_set_interrupts_enabled(FALSE);
+                       assert(current_thread() == parent);
+                       clear_ts();
+                       fp_save(parent);
+                       clear_fpu();
+
+                       (void)ml_set_interrupts_enabled(intr);
+               }
 
                if (ifps->fp_valid) {
                        child->machine.ifps = new_ifps;
 
                if (ifps->fp_valid) {
                        child->machine.ifps = new_ifps;
-                       assert((fp_register_state_size == sizeof(struct x86_fx_thread_state)) ||
-                           (fp_register_state_size == sizeof(struct x86_avx_thread_state)));
+                       child->machine.xstate = xstate;
                        bcopy((char *)(ppcb->ifps),
                        bcopy((char *)(ppcb->ifps),
-                           (char *)(child->machine.ifps), fp_register_state_size);
+                           (char *)(child->machine.ifps),
+                           fp_state_size[xstate]);
 
                        /* Mark the new fp saved state as non-live. */
                        /* Temporarily disabled: radar 4647827
 
                        /* Mark the new fp saved state as non-live. */
                        /* Temporarily disabled: radar 4647827
@@ -695,37 +1182,51 @@ fpu_dup_fxstate(
        }
        simple_unlock(&ppcb->lock);
 
        }
        simple_unlock(&ppcb->lock);
 
-       if (new_ifps != NULL)
-               fp_state_free(new_ifps);
+       if (new_ifps != NULL) {
+               fp_state_free(new_ifps, xstate);
+       }
 }
 
 }
 
-
 /*
  * Initialize FPU.
 /*
  * Initialize FPU.
- *
+ * FNINIT programs the x87 control word to 0x37f, which matches
+ * the desired default for macOS.
  */
 
 void
 fpinit(void)
 {
  */
 
 void
 fpinit(void)
 {
-       unsigned short  control;
-
+       boolean_t istate = ml_set_interrupts_enabled(FALSE);
        clear_ts();
        fninit();
        clear_ts();
        fninit();
+#if DEBUG
+       /* We skip this power-on-default verification sequence on
+        * non-DEBUG, as dirtying the x87 control word may slow down
+        * xsave/xrstor and affect energy use.
+        */
+       unsigned short  control, control2;
        fnstcw(&control);
        fnstcw(&control);
-       control &= ~(FPC_PC|FPC_RC); /* Clear precision & rounding control */
-       control |= (FPC_PC_64 |         /* Set precision */ 
-                       FPC_RC_RN |     /* round-to-nearest */
-                       FPC_ZE |        /* Suppress zero-divide */
-                       FPC_OE |        /*  and overflow */
-                       FPC_UE |        /*  underflow */
-                       FPC_IE |        /* Allow NaNQs and +-INF */
-                       FPC_DE |        /* Allow denorms as operands  */
-                       FPC_PE);        /* No trap for precision loss */
+       control2 = control;
+       control &= ~(FPC_PC | FPC_RC); /* Clear precision & rounding control */
+       control |= (FPC_PC_64 |         /* Set precision */
+           FPC_RC_RN |                 /* round-to-nearest */
+           FPC_ZE |                    /* Suppress zero-divide */
+           FPC_OE |                    /*  and overflow */
+           FPC_UE |                    /*  underflow */
+           FPC_IE |                    /* Allow NaNQs and +-INF */
+           FPC_DE |                    /* Allow denorms as operands  */
+           FPC_PE);                    /* No trap for precision loss */
+       assert(control == control2);
        fldcw(control);
        fldcw(control);
-
+#endif
        /* Initialize SSE/SSE2 */
        __builtin_ia32_ldmxcsr(0x1f80);
        /* Initialize SSE/SSE2 */
        __builtin_ia32_ldmxcsr(0x1f80);
+       if (fpu_YMM_capable) {
+               vzeroall();
+       } else {
+               xmmzeroall();
+       }
+       ml_set_interrupts_enabled(istate);
 }
 
 /*
 }
 
 /*
@@ -737,30 +1238,29 @@ uint64_t x86_isr_fp_simd_use;
 void
 fpnoextflt(void)
 {
 void
 fpnoextflt(void)
 {
-       boolean_t       intr;
-       thread_t        thr_act;
-       pcb_t           pcb;
+       boolean_t       intr;
+       thread_t        thr_act;
+       pcb_t           pcb;
        struct x86_fx_thread_state *ifps = 0;
        struct x86_fx_thread_state *ifps = 0;
+       xstate_t        xstate = current_xstate();
 
        thr_act = current_thread();
        pcb = THREAD_TO_PCB(thr_act);
 
 
        thr_act = current_thread();
        pcb = THREAD_TO_PCB(thr_act);
 
-       assert(fp_register_state_size != 0);
-
        if (pcb->ifps == 0 && !get_interrupt_level()) {
        if (pcb->ifps == 0 && !get_interrupt_level()) {
-               ifps = fp_state_alloc();
+               ifps = fp_state_alloc(xstate);
                bcopy((char *)&initial_fp_state, (char *)ifps,
                bcopy((char *)&initial_fp_state, (char *)ifps,
-                   fp_register_state_size);
-               if (!thread_is_64bit(thr_act)) {
-                       ifps->fp_save_layout = fpu_YMM_present ? XSAVE32 : FXSAVE32;
+                   fp_state_size[xstate]);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       ifps->fp_save_layout = fpu_YMM_capable ? XSAVE32 : FXSAVE32;
+               } else {
+                       ifps->fp_save_layout = fpu_YMM_capable ? XSAVE64 : FXSAVE64;
                }
                }
-               else
-                       ifps->fp_save_layout = fpu_YMM_present ? XSAVE64 : FXSAVE64;
                ifps->fp_valid = TRUE;
        }
        intr = ml_set_interrupts_enabled(FALSE);
 
                ifps->fp_valid = TRUE;
        }
        intr = ml_set_interrupts_enabled(FALSE);
 
-       clear_ts();                     /*  Enable FPU use */
+       clear_ts();                     /*  Enable FPU use */
 
        if (__improbable(get_interrupt_level())) {
                /* Track number of #DNA traps at interrupt context,
 
        if (__improbable(get_interrupt_level())) {
                /* Track number of #DNA traps at interrupt context,
@@ -776,8 +1276,9 @@ fpnoextflt(void)
                }
                fpinit();
        } else {
                }
                fpinit();
        } else {
-               if (pcb->ifps == 0) {
-                       pcb->ifps = ifps;
+               if (pcb->ifps == 0) {
+                       pcb->ifps = ifps;
+                       pcb->xstate = xstate;
                        ifps = 0;
                }
                /*
                        ifps = 0;
                }
                /*
@@ -787,8 +1288,9 @@ fpnoextflt(void)
        }
        (void)ml_set_interrupts_enabled(intr);
 
        }
        (void)ml_set_interrupts_enabled(intr);
 
-       if (ifps)
-               fp_state_free(ifps);
+       if (ifps) {
+               fp_state_free(ifps, xstate);
+       }
 }
 
 /*
 }
 
 /*
@@ -799,24 +1301,27 @@ fpnoextflt(void)
 void
 fpextovrflt(void)
 {
 void
 fpextovrflt(void)
 {
-       thread_t        thr_act = current_thread();
-       pcb_t           pcb;
+       thread_t        thr_act = current_thread();
+       pcb_t           pcb;
        struct x86_fx_thread_state *ifps;
        struct x86_fx_thread_state *ifps;
-       boolean_t       intr;
+       boolean_t       intr;
+       xstate_t        xstate = current_xstate();
 
        intr = ml_set_interrupts_enabled(FALSE);
 
 
        intr = ml_set_interrupts_enabled(FALSE);
 
-       if (get_interrupt_level())
+       if (get_interrupt_level()) {
                panic("FPU segment overrun exception  at interrupt context\n");
                panic("FPU segment overrun exception  at interrupt context\n");
-       if (current_task() == kernel_task)
+       }
+       if (current_task() == kernel_task) {
                panic("FPU segment overrun exception in kernel thread context\n");
                panic("FPU segment overrun exception in kernel thread context\n");
+       }
 
        /*
         * This is a non-recoverable error.
         * Invalidate the thread`s FPU state.
         */
        pcb = THREAD_TO_PCB(thr_act);
 
        /*
         * This is a non-recoverable error.
         * Invalidate the thread`s FPU state.
         */
        pcb = THREAD_TO_PCB(thr_act);
-       simple_lock(&pcb->lock);
+       simple_lock(&pcb->lock, LCK_GRP_NULL);
        ifps = pcb->ifps;
        pcb->ifps = 0;
        simple_unlock(&pcb->lock);
        ifps = pcb->ifps;
        pcb->ifps = 0;
        simple_unlock(&pcb->lock);
@@ -834,16 +1339,19 @@ fpextovrflt(void)
 
        (void)ml_set_interrupts_enabled(intr);
 
 
        (void)ml_set_interrupts_enabled(intr);
 
-       if (ifps)
-           zfree(ifps_zone, ifps);
+       if (ifps) {
+               fp_state_free(ifps, xstate);
+       }
 
        /*
         * Raise exception.
         */
 
        /*
         * Raise exception.
         */
-       i386_exception(EXC_BAD_ACCESS, VM_PROT_READ|VM_PROT_EXECUTE, 0);
+       i386_exception(EXC_BAD_ACCESS, VM_PROT_READ | VM_PROT_EXECUTE, 0);
        /*NOTREACHED*/
 }
 
        /*NOTREACHED*/
 }
 
+extern void fpxlog(int, uint32_t, uint32_t, uint32_t);
+
 /*
  * FPU error. Called by AST.
  */
 /*
  * FPU error. Called by AST.
  */
@@ -851,16 +1359,18 @@ fpextovrflt(void)
 void
 fpexterrflt(void)
 {
 void
 fpexterrflt(void)
 {
-       thread_t        thr_act = current_thread();
+       thread_t        thr_act = current_thread();
        struct x86_fx_thread_state *ifps = thr_act->machine.ifps;
        struct x86_fx_thread_state *ifps = thr_act->machine.ifps;
-       boolean_t       intr;
+       boolean_t       intr;
 
        intr = ml_set_interrupts_enabled(FALSE);
 
 
        intr = ml_set_interrupts_enabled(FALSE);
 
-       if (get_interrupt_level())
+       if (get_interrupt_level()) {
                panic("FPU error exception at interrupt context\n");
                panic("FPU error exception at interrupt context\n");
-       if (current_task() == kernel_task)
+       }
+       if (current_task() == kernel_task) {
                panic("FPU error exception in kernel thread context\n");
                panic("FPU error exception in kernel thread context\n");
+       }
 
        /*
         * Save the FPU state and turn off the FPU.
 
        /*
         * Save the FPU state and turn off the FPU.
@@ -869,14 +1379,19 @@ fpexterrflt(void)
 
        (void)ml_set_interrupts_enabled(intr);
 
 
        (void)ml_set_interrupts_enabled(intr);
 
+       const uint32_t mask = ifps->fx_control &
+           (FPC_IM | FPC_DM | FPC_ZM | FPC_OM | FPC_UE | FPC_PE);
+       const uint32_t xcpt = ~mask & (ifps->fx_status &
+           (FPS_IE | FPS_DE | FPS_ZE | FPS_OE | FPS_UE | FPS_PE));
+       fpxlog(EXC_I386_EXTERR, ifps->fx_status, ifps->fx_control, xcpt);
        /*
         * Raise FPU exception.
         * Locking not needed on pcb->ifps,
         * since thread is running.
         */
        i386_exception(EXC_ARITHMETIC,
        /*
         * Raise FPU exception.
         * Locking not needed on pcb->ifps,
         * since thread is running.
         */
        i386_exception(EXC_ARITHMETIC,
-                      EXC_I386_EXTERR,
-                      ifps->fx_status);
+           EXC_I386_EXTERR,
+           ifps->fx_status);
 
        /*NOTREACHED*/
 }
 
        /*NOTREACHED*/
 }
@@ -893,7 +1408,7 @@ fpexterrflt(void)
 
 void
 fp_save(
 
 void
 fp_save(
-       thread_t        thr_act)
+       thread_t        thr_act)
 {
        pcb_t pcb = THREAD_TO_PCB(thr_act);
        struct x86_fx_thread_state *ifps = pcb->ifps;
 {
        pcb_t pcb = THREAD_TO_PCB(thr_act);
        struct x86_fx_thread_state *ifps = pcb->ifps;
@@ -903,7 +1418,7 @@ fp_save(
                assert((get_cr0() & CR0_TS) == 0);
                /* registers are in FPU */
                ifps->fp_valid = TRUE;
                assert((get_cr0() & CR0_TS) == 0);
                /* registers are in FPU */
                ifps->fp_valid = TRUE;
-               fpu_store_registers(ifps, thread_is_64bit(thr_act));
+               fpu_store_registers(ifps, thread_is_64bit_addr(thr_act));
        }
 }
 
        }
 }
 
@@ -915,16 +1430,16 @@ fp_save(
 
 void
 fp_load(
 
 void
 fp_load(
-       thread_t        thr_act)
+       thread_t        thr_act)
 {
        pcb_t pcb = THREAD_TO_PCB(thr_act);
        struct x86_fx_thread_state *ifps = pcb->ifps;
 
        assert(ifps);
 {
        pcb_t pcb = THREAD_TO_PCB(thr_act);
        struct x86_fx_thread_state *ifps = pcb->ifps;
 
        assert(ifps);
-#if    DEBUG
+#if     DEBUG
        if (ifps->fp_valid != FALSE && ifps->fp_valid != TRUE) {
                panic("fp_load() invalid fp_valid: %u, fp_save_layout: %u\n",
        if (ifps->fp_valid != FALSE && ifps->fp_valid != TRUE) {
                panic("fp_load() invalid fp_valid: %u, fp_save_layout: %u\n",
-                     ifps->fp_valid, ifps->fp_save_layout);
+                   ifps->fp_valid, ifps->fp_save_layout);
        }
 #endif
 
        }
 #endif
 
@@ -933,7 +1448,7 @@ fp_load(
        } else {
                fpu_load_registers(ifps);
        }
        } else {
                fpu_load_registers(ifps);
        }
-       ifps->fp_valid = FALSE;         /* in FPU */
+       ifps->fp_valid = FALSE;         /* in FPU */
 }
 
 /*
 }
 
 /*
@@ -944,16 +1459,18 @@ fp_load(
 void
 fpSSEexterrflt(void)
 {
 void
 fpSSEexterrflt(void)
 {
-       thread_t        thr_act = current_thread();
+       thread_t        thr_act = current_thread();
        struct x86_fx_thread_state *ifps = thr_act->machine.ifps;
        struct x86_fx_thread_state *ifps = thr_act->machine.ifps;
-       boolean_t       intr;
+       boolean_t       intr;
 
        intr = ml_set_interrupts_enabled(FALSE);
 
 
        intr = ml_set_interrupts_enabled(FALSE);
 
-       if (get_interrupt_level())
+       if (get_interrupt_level()) {
                panic("SSE exception at interrupt context\n");
                panic("SSE exception at interrupt context\n");
-       if (current_task() == kernel_task)
+       }
+       if (current_task() == kernel_task) {
                panic("SSE exception in kernel thread context\n");
                panic("SSE exception in kernel thread context\n");
+       }
 
        /*
         * Save the FPU state and turn off the FPU.
 
        /*
         * Save the FPU state and turn off the FPU.
@@ -966,30 +1483,332 @@ fpSSEexterrflt(void)
         * Locking not needed on pcb->ifps,
         * since thread is running.
         */
         * Locking not needed on pcb->ifps,
         * since thread is running.
         */
+       const uint32_t mask = (ifps->fx_MXCSR >> 7) &
+           (FPC_IM | FPC_DM | FPC_ZM | FPC_OM | FPC_UE | FPC_PE);
+       const uint32_t xcpt = ~mask & (ifps->fx_MXCSR &
+           (FPS_IE | FPS_DE | FPS_ZE | FPS_OE | FPS_UE | FPS_PE));
+       fpxlog(EXC_I386_SSEEXTERR, ifps->fx_MXCSR, ifps->fx_MXCSR, xcpt);
 
        i386_exception(EXC_ARITHMETIC,
 
        i386_exception(EXC_ARITHMETIC,
-                      EXC_I386_SSEEXTERR,
-                      ifps->fx_MXCSR);
+           EXC_I386_SSEEXTERR,
+           ifps->fx_MXCSR);
        /*NOTREACHED*/
 }
 
        /*NOTREACHED*/
 }
 
+
+#if !defined(RC_HIDE_XNU_J137)
+/*
+ * If a thread is using an AVX-sized savearea:
+ * - allocate a new AVX512-sized  area,
+ * - copy the 256-bit state into the 512-bit area,
+ * - deallocate the smaller area
+ */
+static void
+fpu_savearea_promote_avx512(thread_t thread)
+{
+       struct x86_avx_thread_state     *ifps = NULL;
+       struct x86_avx512_thread_state  *ifps512 = NULL;
+       pcb_t                           pcb = THREAD_TO_PCB(thread);
+       boolean_t                       do_avx512_alloc = FALSE;
+
+       DBG("fpu_upgrade_savearea(%p)\n", thread);
+
+       simple_lock(&pcb->lock, LCK_GRP_NULL);
+
+       ifps = pcb->ifps;
+       if (ifps == NULL) {
+               pcb->xstate = AVX512;
+               simple_unlock(&pcb->lock);
+               if (thread != current_thread()) {
+                       /* nothing to be done */
+
+                       return;
+               }
+               fpnoextflt();
+               return;
+       }
+
+       if (pcb->xstate != AVX512) {
+               do_avx512_alloc = TRUE;
+       }
+       simple_unlock(&pcb->lock);
+
+       if (do_avx512_alloc == TRUE) {
+               ifps512 = fp_state_alloc(AVX512);
+       }
+
+       simple_lock(&pcb->lock, LCK_GRP_NULL);
+       if (thread == current_thread()) {
+               boolean_t       intr;
+
+               intr = ml_set_interrupts_enabled(FALSE);
+
+               clear_ts();
+               fp_save(thread);
+               clear_fpu();
+
+               xsetbv(0, AVX512_XMASK);
+               current_cpu_datap()->cpu_xstate = AVX512;
+               (void)ml_set_interrupts_enabled(intr);
+       }
+       assert(ifps->fp.fp_valid);
+
+       /* Allocate an AVX512 savearea and copy AVX state into it */
+       if (pcb->xstate != AVX512) {
+               bcopy(ifps, ifps512, fp_state_size[AVX]);
+               pcb->ifps = ifps512;
+               pcb->xstate = AVX512;
+               ifps512 = NULL;
+       } else {
+               ifps = NULL;
+       }
+       /* The PCB lock is redundant in some scenarios given the higher level
+        * thread mutex, but its pre-emption disablement is relied upon here
+        */
+       simple_unlock(&pcb->lock);
+
+       if (ifps) {
+               fp_state_free(ifps, AVX);
+       }
+       if (ifps512) {
+               fp_state_free(ifps, AVX512);
+       }
+}
+
+/*
+ * Upgrade the calling thread to AVX512.
+ */
+boolean_t
+fpu_thread_promote_avx512(thread_t thread)
+{
+       task_t          task = current_task();
+
+       if (thread != current_thread()) {
+               return FALSE;
+       }
+       if (!ml_fpu_avx512_enabled()) {
+               return FALSE;
+       }
+
+       fpu_savearea_promote_avx512(thread);
+
+       /* Racy but the task's xstate is only a hint */
+       task->xstate = AVX512;
+
+       return TRUE;
+}
+
+
+/*
+ * Called from user_trap() when an invalid opcode fault is taken.
+ * If the user is attempting an AVX512 instruction on a machine
+ * that supports this, we switch the calling thread to use
+ * a larger savearea, set its XCR0 bit mask to enable AVX512 and
+ * return directly via thread_exception_return().
+ * Otherwise simply return.
+ */
+#define MAX_X86_INSN_LENGTH (16)
+void
+fpUDflt(user_addr_t rip)
+{
+       uint8_t         instruction_prefix;
+       boolean_t       is_AVX512_instruction = FALSE;
+       user_addr_t     original_rip = rip;
+       do {
+               /* TODO: as an optimisation, copy up to the lesser of the
+                * next page boundary or maximal prefix length in one pass
+                * rather than issue multiple copyins
+                */
+               if (copyin(rip, (char *) &instruction_prefix, 1)) {
+                       return;
+               }
+               DBG("fpUDflt(0x%016llx) prefix: 0x%x\n",
+                   rip, instruction_prefix);
+               /* TODO: determine more specifically which prefixes
+                * are sane possibilities for AVX512 insns
+                */
+               switch (instruction_prefix) {
+               case 0x2E:      /* CS segment override */
+               case 0x36:      /* SS segment override */
+               case 0x3E:      /* DS segment override */
+               case 0x26:      /* ES segment override */
+               case 0x64:      /* FS segment override */
+               case 0x65:      /* GS segment override */
+               case 0x66:      /* Operand-size override */
+               case 0x67:      /* address-size override */
+                       /* Skip optional prefixes */
+                       rip++;
+                       if ((rip - original_rip) > MAX_X86_INSN_LENGTH) {
+                               return;
+                       }
+                       break;
+               case 0x62:      /* EVEX */
+               case 0xC5:      /* VEX 2-byte */
+               case 0xC4:      /* VEX 3-byte */
+                       is_AVX512_instruction = TRUE;
+                       break;
+               default:
+                       return;
+               }
+       } while (!is_AVX512_instruction);
+
+       /* Here if we detect attempted execution of an AVX512 instruction */
+
+       /*
+        * Fail if this machine doesn't support AVX512
+        */
+       if (fpu_capability != AVX512) {
+               return;
+       }
+
+       assert(xgetbv(XCR0) == AVX_XMASK);
+
+       DBG("fpUDflt() switching xstate to AVX512\n");
+       (void) fpu_thread_promote_avx512(current_thread());
+
+       thread_exception_return();
+       /* NOT REACHED */
+}
+#endif /* !defined(RC_HIDE_XNU_J137) */
+
 void
 void
-fp_setvalid(boolean_t value) {
-        thread_t       thr_act = current_thread();
+fp_setvalid(boolean_t value)
+{
+       thread_t        thr_act = current_thread();
        struct x86_fx_thread_state *ifps = thr_act->machine.ifps;
 
        if (ifps) {
        struct x86_fx_thread_state *ifps = thr_act->machine.ifps;
 
        if (ifps) {
-               ifps->fp_valid = value;
+               ifps->fp_valid = value;
 
                if (value == TRUE) {
                        boolean_t istate = ml_set_interrupts_enabled(FALSE);
 
                if (value == TRUE) {
                        boolean_t istate = ml_set_interrupts_enabled(FALSE);
-                       clear_fpu();
+                       clear_fpu();
                        ml_set_interrupts_enabled(istate);
                }
        }
 }
 
 boolean_t
                        ml_set_interrupts_enabled(istate);
                }
        }
 }
 
 boolean_t
-ml_fpu_avx_enabled(void) {
-       return (fpu_YMM_present == TRUE);
+ml_fpu_avx_enabled(void)
+{
+       return fpu_capability >= AVX;
+}
+
+#if !defined(RC_HIDE_XNU_J137)
+boolean_t
+ml_fpu_avx512_enabled(void)
+{
+       return fpu_capability == AVX512;
+}
+#endif
+
+static xstate_t
+task_xstate(task_t task)
+{
+       if (task == TASK_NULL) {
+               return fpu_default;
+       } else {
+               return task->xstate;
+       }
+}
+
+static xstate_t
+thread_xstate(thread_t thread)
+{
+       xstate_t xs = THREAD_TO_PCB(thread)->xstate;
+       if (xs == UNDEFINED) {
+               return task_xstate(thread->task);
+       } else {
+               return xs;
+       }
+}
+
+xstate_t
+current_xstate(void)
+{
+       return thread_xstate(current_thread());
+}
+
+/*
+ * Called when exec'ing between bitnesses.
+ * If valid FPU state exists, adjust the layout.
+ */
+void
+fpu_switch_addrmode(thread_t thread, boolean_t is_64bit)
+{
+       struct x86_fx_thread_state *ifps = thread->machine.ifps;
+       mp_disable_preemption();
+
+       if (ifps && ifps->fp_valid) {
+               if (thread_xstate(thread) == FP) {
+                       ifps->fp_save_layout = is_64bit ? FXSAVE64 : FXSAVE32;
+               } else {
+                       ifps->fp_save_layout = is_64bit ? XSAVE64 : XSAVE32;
+               }
+       }
+       mp_enable_preemption();
+}
+
+static inline uint32_t
+fpsimd_pop(uintptr_t ins, int sz)
+{
+       uint32_t rv = 0;
+
+
+       while (sz >= 16) {
+               uint32_t rv1, rv2;
+               uint64_t *ins64 = (uint64_t *) ins;
+               uint64_t *ins642 = (uint64_t *) (ins + 8);
+               rv1 = __builtin_popcountll(*ins64);
+               rv2 = __builtin_popcountll(*ins642);
+               rv += rv1 + rv2;
+               sz -= 16;
+               ins += 16;
+       }
+
+       while (sz >= 4) {
+               uint32_t *ins32 = (uint32_t *) ins;
+               rv += __builtin_popcount(*ins32);
+               sz -= 4;
+               ins += 4;
+       }
+
+       while (sz > 0) {
+               char *ins8 = (char *)ins;
+               rv += __builtin_popcount(*ins8);
+               sz--;
+               ins++;
+       }
+       return rv;
+}
+
+uint32_t
+thread_fpsimd_hash(thread_t ft)
+{
+       if (fpsimd_fault_popc == 0) {
+               return 0;
+       }
+
+       uint32_t prv = 0;
+       boolean_t istate = ml_set_interrupts_enabled(FALSE);
+       struct x86_fx_thread_state *pifps = THREAD_TO_PCB(ft)->ifps;
+
+       if (pifps) {
+               if (pifps->fp_valid) {
+                       prv = fpsimd_pop((uintptr_t) &pifps->fx_XMM_reg[0][0],
+                           sizeof(pifps->fx_XMM_reg));
+               } else {
+                       uintptr_t cr0 = get_cr0();
+                       clear_ts();
+                       fp_save(ft);
+                       prv = fpsimd_pop((uintptr_t) &pifps->fx_XMM_reg[0][0],
+                           sizeof(pifps->fx_XMM_reg));
+                       pifps->fp_valid = FALSE;
+                       if (cr0 & CR0_TS) {
+                               set_cr0(cr0);
+                       }
+               }
+       }
+       ml_set_interrupts_enabled(istate);
+       return prv;
 }
 }