Linux 5.0 compat: SIMD compatibility

Restore the SIMD optimization for 4.19.38 LTS, 4.14.120 LTS,
and 5.0 and newer kernels.  This is accomplished by leveraging
the fact that by definition dedicated kernel threads never need
to concern themselves with saving and restoring the user FPU state.
Therefore, they may use the FPU as long as we can guarantee user
tasks always restore their FPU state before context switching back
to user space.

For the 5.0 and 5.1 kernels disabling preemption and local
interrupts is sufficient to allow the FPU to be used.  All non-kernel
threads will restore the preserved user FPU state.

For 5.2 and latter kernels the user FPU state restoration will be
skipped if the kernel determines the registers have not changed.
Therefore, for these kernels we need to perform the additional
step of saving and restoring the FPU registers.  Invalidating the
per-cpu global tracking the FPU state would force a restore but
that functionality is private to the core x86 FPU implementation
and unavailable.

In practice, restricting SIMD to kernel threads is not a major
restriction for ZFS.  The vast majority of SIMD operations are
already performed by the IO pipeline.  The remaining cases are
relatively infrequent and can be handled by the generic code
without significant impact.  The two most noteworthy cases are:

  1) Decrypting the wrapping key for an encrypted dataset,
     i.e. `zfs load-key`.  All other encryption and decryption
     operations will use the SIMD optimized implementations.

  2) Generating the payload checksums for a `zfs send` stream.

In order to avoid making any changes to the higher layers of ZFS
all of the `*_get_ops()` functions were updated to take in to
consideration the calling context.  This allows for the fastest
implementation to be used as appropriate (see kfpu_allowed()).

The only other notable instance of SIMD operations being used
outside a kernel thread was at module load time.  This code
was moved in to a taskq in order to accommodate the new kernel
thread restriction.

Finally, a few other modifications were made in order to further
harden this code and facilitate testing.  They include updating
each implementations operations structure to be declared as a
constant.  And allowing "cycle" to be set when selecting the
preferred ops in the kernel as well as user space.

Reviewed-by: Tony Hutter <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #8754 
Closes #8793 
Closes #8965 

1 files changed, 65 insertions, 40 deletions

diff --git a/module/zfs/vdev_raidz_math.c b/module/zfs/vdev_raidz_math.c
index 3ef67768f..ef514e9e1 100644
--- a/module/zfs/vdev_raidz_math.c
+++ b/module/zfs/vdev_raidz_math.c
@@ -27,9 +27,9 @@
 #include <sys/zio.h>
 #include <sys/debug.h>
 #include <sys/zfs_debug.h>
-
 #include <sys/vdev_raidz.h>
 #include <sys/vdev_raidz_impl.h>
+#include <linux/simd.h>
 
 extern boolean_t raidz_will_scalar_work(void);
 
@@ -87,6 +87,7 @@ static uint32_t user_sel_impl = IMPL_FASTEST;
 static size_t raidz_supp_impl_cnt = 0;
 static raidz_impl_ops_t *raidz_supp_impl[ARRAY_SIZE(raidz_all_maths)];
 
+#if defined(_KERNEL)
 /*
  * kstats values for supported implementations
  * Values represent per disk throughput of 8 disk+parity raidz vdev [B/s]
@@ -95,14 +96,19 @@ static raidz_impl_kstat_t raidz_impl_kstats[ARRAY_SIZE(raidz_all_maths) + 1];
 
 /* kstat for benchmarked implementations */
 static kstat_t *raidz_math_kstat = NULL;
+#endif
 
 /*
- * Selects the raidz operation for raidz_map
- * If rm_ops is set to NULL original raidz implementation will be used
+ * Returns the RAIDZ operations for raidz_map() parity calculations.   When
+ * a SIMD implementation is not allowed in the current context, then fallback
+ * to the fastest generic implementation.
  */
-raidz_impl_ops_t *
-vdev_raidz_math_get_ops()
+const raidz_impl_ops_t *
+vdev_raidz_math_get_ops(void)
 {
+	if (!kfpu_allowed())
+		return (&vdev_raidz_scalar_impl);
+
 	raidz_impl_ops_t *ops = NULL;
 	const uint32_t impl = RAIDZ_IMPL_READ(zfs_vdev_raidz_impl);
 
@@ -111,18 +117,14 @@ vdev_raidz_math_get_ops()
 		ASSERT(raidz_math_initialized);
 		ops = &vdev_raidz_fastest_impl;
 		break;
-#if !defined(_KERNEL)
 	case IMPL_CYCLE:
-	{
+		/* Cycle through all supported implementations */
 		ASSERT(raidz_math_initialized);
 		ASSERT3U(raidz_supp_impl_cnt, >, 0);
-		/* Cycle through all supported implementations */
 		static size_t cycle_impl_idx = 0;
 		size_t idx = (++cycle_impl_idx) % raidz_supp_impl_cnt;
 		ops = raidz_supp_impl[idx];
-	}
-	break;
-#endif
+		break;
 	case IMPL_ORIGINAL:
 		ops = (raidz_impl_ops_t *)&vdev_raidz_original_impl;
 		break;
@@ -273,6 +275,8 @@ const char *raidz_rec_name[] = {
 	"rec_pq", "rec_pr", "rec_qr", "rec_pqr"
 };
 
+#if defined(_KERNEL)
+
 #define	RAIDZ_KSTAT_LINE_LEN	(17 + 10*12 + 1)
 
 static int
@@ -435,21 +439,21 @@ benchmark_raidz_impl(raidz_map_t *bench_rm, const int fn, benchmark_fn bench_fn)
 		}
 	}
 }
+#endif
 
-void
-vdev_raidz_math_init(void)
+/*
+ * Initialize and benchmark all supported implementations.
+ */
+static void
+benchmark_raidz(void *arg)
 {
 	raidz_impl_ops_t *curr_impl;
-	zio_t *bench_zio = NULL;
-	raidz_map_t *bench_rm = NULL;
-	uint64_t bench_parity;
-	int i, c, fn;
+	int i, c;
 
-	/* move supported impl into raidz_supp_impl */
+	/* Move supported impl into raidz_supp_impl */
 	for (i = 0, c = 0; i < ARRAY_SIZE(raidz_all_maths); i++) {
 		curr_impl = (raidz_impl_ops_t *)raidz_all_maths[i];
 
-		/* initialize impl */
 		if (curr_impl->init)
 			curr_impl->init();
 
@@ -459,18 +463,10 @@ vdev_raidz_math_init(void)
 	membar_producer();		/* complete raidz_supp_impl[] init */
 	raidz_supp_impl_cnt = c;	/* number of supported impl */
 
-#if !defined(_KERNEL)
-	/* Skip benchmarking and use last implementation as fastest */
-	memcpy(&vdev_raidz_fastest_impl, raidz_supp_impl[raidz_supp_impl_cnt-1],
-	    sizeof (vdev_raidz_fastest_impl));
-	strcpy(vdev_raidz_fastest_impl.name, "fastest");
-
-	raidz_math_initialized = B_TRUE;
-
-	/* Use 'cycle' math selection method for userspace */
-	VERIFY0(vdev_raidz_impl_set("cycle"));
-	return;
-#endif
+#if defined(_KERNEL)
+	zio_t *bench_zio = NULL;
+	raidz_map_t *bench_rm = NULL;
+	uint64_t bench_parity;
 
 	/* Fake a zio and run the benchmark on a warmed up buffer */
 	bench_zio = kmem_zalloc(sizeof (zio_t), KM_SLEEP);
@@ -480,7 +476,7 @@ vdev_raidz_math_init(void)
 	memset(abd_to_buf(bench_zio->io_abd), 0xAA, BENCH_ZIO_SIZE);
 
 	/* Benchmark parity generation methods */
-	for (fn = 0; fn < RAIDZ_GEN_NUM; fn++) {
+	for (int fn = 0; fn < RAIDZ_GEN_NUM; fn++) {
 		bench_parity = fn + 1;
 		/* New raidz_map is needed for each generate_p/q/r */
 		bench_rm = vdev_raidz_map_alloc(bench_zio, SPA_MINBLOCKSHIFT,
@@ -495,7 +491,7 @@ vdev_raidz_math_init(void)
 	bench_rm = vdev_raidz_map_alloc(bench_zio, SPA_MINBLOCKSHIFT,
 	    BENCH_COLS, PARITY_PQR);
 
-	for (fn = 0; fn < RAIDZ_REC_NUM; fn++)
+	for (int fn = 0; fn < RAIDZ_REC_NUM; fn++)
 		benchmark_raidz_impl(bench_rm, fn, benchmark_rec_impl);
 
 	vdev_raidz_map_free(bench_rm);
@@ -503,11 +499,39 @@ vdev_raidz_math_init(void)
 	/* cleanup the bench zio */
 	abd_free(bench_zio->io_abd);
 	kmem_free(bench_zio, sizeof (zio_t));
+#else
+	/*
+	 * Skip the benchmark in user space to avoid impacting libzpool
+	 * consumers (zdb, zhack, zinject, ztest).  The last implementation
+	 * is assumed to be the fastest and used by default.
+	 */
+	memcpy(&vdev_raidz_fastest_impl,
+	    raidz_supp_impl[raidz_supp_impl_cnt - 1],
+	    sizeof (vdev_raidz_fastest_impl));
+	strcpy(vdev_raidz_fastest_impl.name, "fastest");
+#endif /* _KERNEL */
+}
 
-	/* install kstats for all impl */
+void
+vdev_raidz_math_init(void)
+{
+#if defined(_KERNEL)
+	/*
+	 * For 5.0 and latter Linux kernels the fletcher 4 benchmarks are
+	 * run in a kernel threads.  This is needed to take advantage of the
+	 * SIMD functionality, see include/linux/simd_x86.h for details.
+	 */
+	taskqid_t id = taskq_dispatch(system_taskq, benchmark_raidz,
+	    NULL, TQ_SLEEP);
+	if (id != TASKQID_INVALID) {
+		taskq_wait_id(system_taskq, id);
+	} else {
+		benchmark_raidz(NULL);
+	}
+
+	/* Install kstats for all implementations */
 	raidz_math_kstat = kstat_create("zfs", 0, "vdev_raidz_bench", "misc",
 	    KSTAT_TYPE_RAW, 0, KSTAT_FLAG_VIRTUAL);
-
 	if (raidz_math_kstat != NULL) {
 		raidz_math_kstat->ks_data = NULL;
 		raidz_math_kstat->ks_ndata = UINT32_MAX;
@@ -517,6 +541,9 @@ vdev_raidz_math_init(void)
 		    raidz_math_kstat_addr);
 		kstat_install(raidz_math_kstat);
 	}
+#else
+	benchmark_raidz(NULL);
+#endif
 
 	/* Finish initialization */
 	atomic_swap_32(&zfs_vdev_raidz_impl, user_sel_impl);
@@ -527,15 +554,15 @@ void
 vdev_raidz_math_fini(void)
 {
 	raidz_impl_ops_t const *curr_impl;
-	int i;
 
+#if defined(_KERNEL)
 	if (raidz_math_kstat != NULL) {
 		kstat_delete(raidz_math_kstat);
 		raidz_math_kstat = NULL;
 	}
+#endif
 
-	/* fini impl */
-	for (i = 0; i < ARRAY_SIZE(raidz_all_maths); i++) {
+	for (int i = 0; i < ARRAY_SIZE(raidz_all_maths); i++) {
 		curr_impl = raidz_all_maths[i];
 		if (curr_impl->fini)
 			curr_impl->fini();
@@ -546,9 +573,7 @@ static const struct {
 	char *name;
 	uint32_t sel;
 } math_impl_opts[] = {
-#if !defined(_KERNEL)
 		{ "cycle",	IMPL_CYCLE },
-#endif
 		{ "fastest",	IMPL_FASTEST },
 		{ "original",	IMPL_ORIGINAL },
 		{ "scalar",	IMPL_SCALAR }