Linux 5.16 compat: don't use XSTATE_XSAVE to save FPU state

Linux 5.16 moved XSTATE_XSAVE and XSTATE_XRESTORE out of our reach,
so add our own XSAVE{,OPT,S} code and use it for Linux 5.16.

Please note that this differs from previous behavior in that it
won't handle exceptions created by XSAVE an XRSTOR. This is sensible
for three reasons.

 - Exceptions during XSAVE and XRSTOR can only occur if the feature
   is not supported or enabled or the memory operand isn't aligned
   on a 64 byte boundary. If this happens something else went
   terribly wrong, and it may be better to stop execution.

 - Previously we just printed a warning and didn't handle the fault,
   this is arguable for the above reason.

 - All other *SAVE instruction also don't handle exceptions, so this
   at least aligns behavior.

Finally add a test to catch such a regression in the future.

Reviewed-by: Tony Hutter <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Attila Fülöp <[email protected]>
Closes #13042
Closes #13059 

9 files changed, 300 insertions, 6 deletions

diff --git a/config/kernel-fpu.m4 b/config/kernel-fpu.m4
index faa64f1ec..06280239d 100644
--- a/config/kernel-fpu.m4
+++ b/config/kernel-fpu.m4
@@ -1,9 +1,12 @@
-dnl # 
+dnl #
 dnl # Handle differences in kernel FPU code.
 dnl #
 dnl # Kernel
 dnl # 5.16:	XCR code put into asm/fpu/xcr.h
-dnl # 		HAVE_KERNEL_FPU_XCR_HEADER
+dnl #		HAVE_KERNEL_FPU_XCR_HEADER
+dnl #
+dnl #		XSTATE_XSAVE and XSTATE_XRESTORE aren't accessible any more
+dnl #		HAVE_KERNEL_FPU_XSAVE_INTERNAL
 dnl #
 dnl # 5.0:	Wrappers have been introduced to save/restore the FPU state.
 dnl #		This change was made to the 4.19.38 and 4.14.120 LTS kernels.
@@ -107,6 +110,36 @@ AC_DEFUN([ZFS_AC_KERNEL_SRC_FPU], [
 		struct fxregs_state *fxr __attribute__ ((unused)) = &st->fxsave;
 		struct xregs_state *xr __attribute__ ((unused)) = &st->xsave;
 	])
+
+	ZFS_LINUX_TEST_SRC([fpu_xsave_internal], [
+		#include <linux/sched.h>
+		#if defined(__x86_64) || defined(__x86_64__) || \
+		    defined(__i386) || defined(__i386__)
+		#if !defined(__x86)
+		#define __x86
+		#endif
+		#endif
+
+		#if !defined(__x86)
+		#error Unsupported architecture
+		#endif
+
+		#include <linux/types.h>
+		#ifdef HAVE_KERNEL_FPU_API_HEADER
+		#include <asm/fpu/api.h>
+		#include <asm/fpu/internal.h>
+		#else
+		#include <asm/i387.h>
+		#include <asm/xcr.h>
+		#endif
+
+	],[
+		struct fpu *fpu = &current->thread.fpu;
+		union fpregs_state *st = &fpu->fpstate->regs;
+		struct fregs_state *fr __attribute__ ((unused)) = &st->fsave;
+		struct fxregs_state *fxr __attribute__ ((unused)) = &st->fxsave;
+		struct xregs_state *xr __attribute__ ((unused)) = &st->xsave;
+	])
 ])
 
 AC_DEFUN([ZFS_AC_KERNEL_FPU], [
@@ -139,7 +172,13 @@ AC_DEFUN([ZFS_AC_KERNEL_FPU], [
 				AC_DEFINE(HAVE_KERNEL_FPU_INTERNAL, 1,
 				    [kernel fpu internal])
 			],[
+				ZFS_LINUX_TEST_RESULT([fpu_xsave_internal], [
+				    AC_MSG_RESULT(internal with internal XSAVE)
+				    AC_DEFINE(HAVE_KERNEL_FPU_XSAVE_INTERNAL, 1,
+					[kernel fpu and XSAVE internal])
+			    ],[
 				AC_MSG_RESULT(unavailable)
+			    ])
 			])
 		])
 	])
diff --git a/config/toolchain-simd.m4 b/config/toolchain-simd.m4
index 1153cd694..061576fd9 100644
--- a/config/toolchain-simd.m4
+++ b/config/toolchain-simd.m4
@@ -24,6 +24,9 @@ AC_DEFUN([ZFS_AC_CONFIG_ALWAYS_TOOLCHAIN_SIMD], [
 			ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_AES
 			ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_PCLMULQDQ
 			ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_MOVBE
+			ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVE
+			ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVEOPT
+			ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVES
 			;;
 	esac
 ])
@@ -422,3 +425,66 @@ AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_MOVBE], [
 		AC_MSG_RESULT([no])
 	])
 ])
+
+dnl #
+dnl # ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVE
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVE], [
+	AC_MSG_CHECKING([whether host toolchain supports XSAVE])
+
+	AC_LINK_IFELSE([AC_LANG_SOURCE([
+	[
+		void main()
+		{
+		  char b[4096] __attribute__ ((aligned (64)));
+		  __asm__ __volatile__("xsave %[b]\n" : : [b] "m" (*b) : "memory");
+		}
+	]])], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE([HAVE_XSAVE], 1, [Define if host toolchain supports XSAVE])
+	], [
+		AC_MSG_RESULT([no])
+	])
+])
+
+dnl #
+dnl # ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVEOPT
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVEOPT], [
+	AC_MSG_CHECKING([whether host toolchain supports XSAVEOPT])
+
+	AC_LINK_IFELSE([AC_LANG_SOURCE([
+	[
+		void main()
+		{
+		  char b[4096] __attribute__ ((aligned (64)));
+		  __asm__ __volatile__("xsaveopt %[b]\n" : : [b] "m" (*b) : "memory");
+		}
+	]])], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE([HAVE_XSAVEOPT], 1, [Define if host toolchain supports XSAVEOPT])
+	], [
+		AC_MSG_RESULT([no])
+	])
+])
+
+dnl #
+dnl # ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVES
+dnl #
+AC_DEFUN([ZFS_AC_CONFIG_TOOLCHAIN_CAN_BUILD_XSAVES], [
+	AC_MSG_CHECKING([whether host toolchain supports XSAVES])
+
+	AC_LINK_IFELSE([AC_LANG_SOURCE([
+	[
+		void main()
+		{
+		  char b[4096] __attribute__ ((aligned (64)));
+		  __asm__ __volatile__("xsaves %[b]\n" : : [b] "m" (*b) : "memory");
+		}
+	]])], [
+		AC_MSG_RESULT([yes])
+		AC_DEFINE([HAVE_XSAVES], 1, [Define if host toolchain supports XSAVES])
+	], [
+		AC_MSG_RESULT([no])
+	])
+])
diff --git a/configure.ac b/configure.ac
index 7b6113c1b..7037c06b2 100644
--- a/configure.ac
+++ b/configure.ac
@@ -383,6 +383,7 @@ AC_CONFIG_FILES([
 	tests/zfs-tests/tests/functional/rootpool/Makefile
 	tests/zfs-tests/tests/functional/rsend/Makefile
 	tests/zfs-tests/tests/functional/scrub_mirror/Makefile
+	tests/zfs-tests/tests/functional/simd/Makefile
 	tests/zfs-tests/tests/functional/slog/Makefile
 	tests/zfs-tests/tests/functional/snapshot/Makefile
 	tests/zfs-tests/tests/functional/snapused/Makefile
diff --git a/include/os/linux/kernel/linux/simd_x86.h b/include/os/linux/kernel/linux/simd_x86.h
index f2ae0fcbc..7c806bf2c 100644
--- a/include/os/linux/kernel/linux/simd_x86.h
+++ b/include/os/linux/kernel/linux/simd_x86.h
@@ -136,9 +136,27 @@
  * When the kernel_fpu_* symbols are unavailable then provide our own
  * versions which allow the FPU to be safely used.
  */
+#if defined(HAVE_KERNEL_FPU_INTERNAL) || defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL)
+
+#if defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL)
+/*
+ * Some sanity checks.
+ * HAVE_KERNEL_FPU_INTERNAL and HAVE_KERNEL_FPU_XSAVE_INTERNAL are exclusive.
+ */
 #if defined(HAVE_KERNEL_FPU_INTERNAL)
+#error "HAVE_KERNEL_FPU_INTERNAL and HAVE_KERNEL_FPU_XSAVE_INTERNAL defined"
+#endif
+/*
+ * For kernels >= 5.16 we have to use inline assembly with the XSAVE{,OPT,S}
+ * instructions, so we need the toolchain to support at least XSAVE.
+ */
+#if !defined(HAVE_XSAVE)
+#error "Toolchain needs to support the XSAVE assembler instruction"
+#endif
+#endif
 
 #include <linux/mm.h>
+#include <linux/slab.h>
 
 extern union fpregs_state **zfs_kfpu_fpregs;
 
@@ -191,7 +209,9 @@ kfpu_init(void)
 }
 
 #define	kfpu_allowed()		1
+#if defined(HAVE_KERNEL_FPU_INTERNAL)
 #define	ex_handler_fprestore	ex_handler_default
+#endif
 
 /*
  * FPU save and restore instructions.
@@ -206,6 +226,7 @@ kfpu_init(void)
 #define	kfpu_fxsr_clean(rval)	__asm("fnclex; emms; fildl %P[addr]" \
 				    : : [addr] "m" (rval));
 
+#if defined(HAVE_KERNEL_FPU_INTERNAL)
 static inline void
 kfpu_save_xsave(struct xregs_state *addr, uint64_t mask)
 {
@@ -217,6 +238,21 @@ kfpu_save_xsave(struct xregs_state *addr, uint64_t mask)
 	XSTATE_XSAVE(addr, low, hi, err);
 	WARN_ON_ONCE(err);
 }
+#endif /* defined(HAVE_KERNEL_FPU_INTERNAL) */
+
+#if defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL)
+#define	kfpu_do_xsave(instruction, addr, mask)			\
+{								\
+	uint32_t low, hi;					\
+								\
+	low = mask;						\
+	hi = (uint64_t)(mask) >> 32;				\
+	__asm(instruction " %[dst]\n\t"				\
+	    :							\
+	    : [dst] "m" (*(addr)), "a" (low), "d" (hi)		\
+	    : "memory");					\
+}
+#endif /* defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL) */
 
 static inline void
 kfpu_save_fxsr(struct fxregs_state *addr)
@@ -233,6 +269,7 @@ kfpu_save_fsave(struct fregs_state *addr)
 	kfpu_fnsave(addr);
 }
 
+#if defined(HAVE_KERNEL_FPU_INTERNAL)
 static inline void
 kfpu_begin(void)
 {
@@ -250,7 +287,6 @@ kfpu_begin(void)
 	 * FPU state to be correctly preserved and restored.
 	 */
 	union fpregs_state *state = zfs_kfpu_fpregs[smp_processor_id()];
-
 	if (static_cpu_has(X86_FEATURE_XSAVE)) {
 		kfpu_save_xsave(&state->xsave, ~0);
 	} else if (static_cpu_has(X86_FEATURE_FXSR)) {
@@ -259,7 +295,50 @@ kfpu_begin(void)
 		kfpu_save_fsave(&state->fsave);
 	}
 }
+#endif /* defined(HAVE_KERNEL_FPU_INTERNAL) */
+
+#if defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL)
+static inline void
+kfpu_begin(void)
+{
+	/*
+	 * Preemption and interrupts must be disabled for the critical
+	 * region where the FPU state is being modified.
+	 */
+	preempt_disable();
+	local_irq_disable();
+
+	/*
+	 * The current FPU registers need to be preserved by kfpu_begin()
+	 * and restored by kfpu_end().  They are stored in a dedicated
+	 * per-cpu variable, not in the task struct, this allows any user
+	 * FPU state to be correctly preserved and restored.
+	 */
+	union fpregs_state *state = zfs_kfpu_fpregs[smp_processor_id()];
+#if defined(HAVE_XSAVES)
+	if (static_cpu_has(X86_FEATURE_XSAVES)) {
+		kfpu_do_xsave("xsaves", &state->xsave, ~0);
+		goto out;
+	}
+#endif
+#if defined(HAVE_XSAVEOPT)
+	if (static_cpu_has(X86_FEATURE_XSAVEOPT)) {
+		kfpu_do_xsave("xsaveopt", &state->xsave, ~0);
+		goto out;
+	}
+#endif
+	if (static_cpu_has(X86_FEATURE_XSAVE)) {
+		kfpu_do_xsave("xsave", &state->xsave, ~0);
+	} else if (static_cpu_has(X86_FEATURE_FXSR)) {
+		kfpu_save_fxsr(&state->fxsave);
+	} else {
+		kfpu_save_fsave(&state->fsave);
+	}
+out:
+}
+#endif /* defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL) */
 
+#if defined(HAVE_KERNEL_FPU_INTERNAL)
 static inline void
 kfpu_restore_xsave(struct xregs_state *addr, uint64_t mask)
 {
@@ -269,6 +348,21 @@ kfpu_restore_xsave(struct xregs_state *addr, uint64_t mask)
 	hi = mask >> 32;
 	XSTATE_XRESTORE(addr, low, hi);
 }
+#endif /* defined(HAVE_KERNEL_FPU_INTERNAL) */
+
+#if defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL)
+#define	kfpu_do_xrstor(instruction, addr, mask)			\
+{								\
+	uint32_t low, hi;					\
+								\
+	low = mask;						\
+	hi = (uint64_t)(mask) >> 32;				\
+	__asm(instruction " %[src]"				\
+	    :							\
+	    : [src] "m" (*(addr)), "a" (low), "d" (hi)		\
+	    : "memory");					\
+}
+#endif /* defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL) */
 
 static inline void
 kfpu_restore_fxsr(struct fxregs_state *addr)
@@ -294,6 +388,7 @@ kfpu_restore_fsave(struct fregs_state *addr)
 	kfpu_frstor(addr);
 }
 
+#if defined(HAVE_KERNEL_FPU_INTERNAL)
 static inline void
 kfpu_end(void)
 {
@@ -310,6 +405,32 @@ kfpu_end(void)
 	local_irq_enable();
 	preempt_enable();
 }
+#endif /* defined(HAVE_KERNEL_FPU_INTERNAL) */
+
+#if defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL)
+static inline void
+kfpu_end(void)
+{
+	union fpregs_state *state = zfs_kfpu_fpregs[smp_processor_id()];
+#if defined(HAVE_XSAVES)
+	if (static_cpu_has(X86_FEATURE_XSAVES)) {
+		kfpu_do_xrstor("xrstors", &state->xsave, ~0);
+		goto out;
+	}
+#endif
+	if (static_cpu_has(X86_FEATURE_XSAVE)) {
+		kfpu_do_xrstor("xrstor", &state->xsave, ~0);
+	} else if (static_cpu_has(X86_FEATURE_FXSR)) {
+		kfpu_save_fxsr(&state->fxsave);
+	} else {
+		kfpu_save_fsave(&state->fsave);
+	}
+out:
+	local_irq_enable();
+	preempt_enable();
+
+}
+#endif /* defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL) */
 
 #else
 
@@ -322,7 +443,7 @@ kfpu_end(void)
 #define	kfpu_init()		0
 #define	kfpu_fini()		((void) 0)
 
-#endif /* defined(HAVE_KERNEL_FPU_INTERNAL) */
+#endif /* defined(HAVE_KERNEL_FPU_INTERNAL || HAVE_KERNEL_FPU_XSAVE_INTERNAL) */
 #endif /* defined(KERNEL_EXPORTS_X86_FPU) */
 
 /*
diff --git a/module/zcommon/zfs_prop.c b/module/zcommon/zfs_prop.c
index 36f30859d..895a86bb4 100644
--- a/module/zcommon/zfs_prop.c
+++ b/module/zcommon/zfs_prop.c
@@ -993,10 +993,10 @@ zfs_prop_align_right(zfs_prop_t prop)
 
 #include <sys/simd.h>
 
-#if defined(HAVE_KERNEL_FPU_INTERNAL)
+#if defined(HAVE_KERNEL_FPU_INTERNAL) || defined(HAVE_KERNEL_FPU_XSAVE_INTERNAL)
 union fpregs_state **zfs_kfpu_fpregs;
 EXPORT_SYMBOL(zfs_kfpu_fpregs);
-#endif /* HAVE_KERNEL_FPU_INTERNAL */
+#endif /* HAVE_KERNEL_FPU_INTERNAL || HAVE_KERNEL_FPU_XSAVE_INTERNAL */
 
 static int __init
 zcommon_init(void)
diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run
index 58eefec31..6ce78a3fc 100644
--- a/tests/runfiles/linux.run
+++ b/tests/runfiles/linux.run
@@ -147,6 +147,12 @@ tags = ['functional', 'projectquota']
 tests = ['send_realloc_dnode_size', 'send_encrypted_files']
 tags = ['functional', 'rsend']
 
+[tests/functional/simd:Linux]
+pre =
+post =
+tests = ['simd_supported']
+tags = ['functional', 'simd']
+
 [tests/functional/snapshot:Linux]
 tests = ['snapshot_015_pos', 'snapshot_016_pos']
 tags = ['functional', 'snapshot']
diff --git a/tests/zfs-tests/tests/functional/Makefile.am b/tests/zfs-tests/tests/functional/Makefile.am
index 1412e3d81..e71172b8e 100644
--- a/tests/zfs-tests/tests/functional/Makefile.am
+++ b/tests/zfs-tests/tests/functional/Makefile.am
@@ -89,5 +89,6 @@ SUBDIRS = \
 
 if BUILD_LINUX
 SUBDIRS += \
+	simd \
 	tmpfile
 endif
diff --git a/tests/zfs-tests/tests/functional/simd/Makefile.am b/tests/zfs-tests/tests/functional/simd/Makefile.am
new file mode 100644
index 000000000..bfc288680
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/simd/Makefile.am
@@ -0,0 +1,2 @@
+pkgdatadir = $(datadir)/@PACKAGE@/zfs-tests/tests/functional/simd
+dist_pkgdata_SCRIPTS = simd_supported.ksh
diff --git a/tests/zfs-tests/tests/functional/simd/simd_supported.ksh b/tests/zfs-tests/tests/functional/simd/simd_supported.ksh
new file mode 100755
index 000000000..d88bc582b
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/simd/simd_supported.ksh
@@ -0,0 +1,58 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# The contents of this file are subject to the terms of the
+# Common Development and Distribution License (the "License").
+# You may not use this file except in compliance with the License.
+#
+# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
+# or http://www.opensolaris.org/os/licensing.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+#
+# When distributing Covered Code, include this CDDL HEADER in each
+# file and include the License file at usr/src/OPENSOLARIS.LICENSE.
+# If applicable, add the following below this CDDL HEADER, with the
+# fields enclosed by brackets "[]" replaced with your own identifying
+# information: Portions Copyright [yyyy] [name of copyright owner]
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2022 by Attila Fülöp <[email protected]>
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+#
+# DESCRIPTION:
+#       Make sure we have SIMD support, so it will not go away without notice
+#
+# STRATEGY:
+#	1. Test if we are running on a Linux x86 system with SSE support
+#       2. If so, check if the zfs_fletcher_4_impl module parameter contains
+#	   a sse implementation
+#	3. If not fail the test, otherwise pass it
+
+log_note "Testing if we support SIMD instructions (Linux x86 only)"
+
+if !is_linux; then
+    log_unsupported "Not a Linux System"
+fi
+
+case "$(uname -m)" in
+i386|i686|x86_64)
+	typeset -R modparam="/sys/module/zcommon/parameters/zfs_fletcher_4_impl"
+	if cat /proc/cpuinfo | awk '/^flags/ {print; exit;}' | grep -q sse; then
+		log_must grep -q sse "$modparam"
+		log_pass "SIMD instructions supported"
+	else
+		log_unsupported "No FPU present"
+	fi
+	;;
+*)
+	log_unsupported "Not a x86 CPU"
+	;;
+esac