Merge patch series "RISC-V: hwprobe: Introduce which-cpus"

Andrew Jones <ajones@ventanamicro.com> says:

This series introduces a flag for the hwprobe syscall which effectively
reverses its behavior from getting the values of keys for a set of cpus
to getting the cpus for a set of key-value pairs.

* b4-shazam-merge:
  RISC-V: selftests: Add which-cpus hwprobe test
  RISC-V: hwprobe: Introduce which-cpus flag
  RISC-V: Move the hwprobe syscall to its own file
  RISC-V: hwprobe: Clarify cpus size parameter

Link: https://lore.kernel.org/r/20231122164700.127954-6-ajones@ventanamicro.comSigned-off-by: Palmer Dabbelt <palmer@rivosinc.com>

Documentation/arch/riscv/hwprobe.rst

@@ -12,7 +12,7 @@ is defined in <asm/hwprobe.h>::
     };
 
     long sys_riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
-                           size_t cpu_count, cpu_set_t *cpus,
+                           size_t cpusetsize, cpu_set_t *cpus,
                            unsigned int flags);
 
 The arguments are split into three groups: an array of key-value pairs, a CPU
@@ -20,12 +20,26 @@ set, and some flags. The key-value pairs are supplied with a count. Userspace
 must prepopulate the key field for each element, and the kernel will fill in the
 value if the key is recognized. If a key is unknown to the kernel, its key field
 will be cleared to -1, and its value set to 0. The CPU set is defined by
-CPU_SET(3). For value-like keys (eg. vendor/arch/impl), the returned value will
-be only be valid if all CPUs in the given set have the same value. Otherwise -1
-will be returned. For boolean-like keys, the value returned will be a logical
-AND of the values for the specified CPUs. Usermode can supply NULL for cpus and
-0 for cpu_count as a shortcut for all online CPUs. There are currently no flags,
-this value must be zero for future compatibility.
+CPU_SET(3) with size ``cpusetsize`` bytes. For value-like keys (eg. vendor,
+arch, impl), the returned value will only be valid if all CPUs in the given set
+have the same value. Otherwise -1 will be returned. For boolean-like keys, the
+value returned will be a logical AND of the values for the specified CPUs.
+Usermode can supply NULL for ``cpus`` and 0 for ``cpusetsize`` as a shortcut for
+all online CPUs. The currently supported flags are:
+
+* :c:macro:`RISCV_HWPROBE_WHICH_CPUS`: This flag basically reverses the behavior
+  of sys_riscv_hwprobe().  Instead of populating the values of keys for a given
+  set of CPUs, the values of each key are given and the set of CPUs is reduced
+  by sys_riscv_hwprobe() to only those which match each of the key-value pairs.
+  How matching is done depends on the key type.  For value-like keys, matching
+  means to be the exact same as the value.  For boolean-like keys, matching
+  means the result of a logical AND of the pair's value with the CPU's value is
+  exactly the same as the pair's value.  Additionally, when ``cpus`` is an empty
+  set, then it is initialized to all online CPUs which fit within it, i.e. the
+  CPU set returned is the reduction of all the online CPUs which can be
+  represented with a CPU set of size ``cpusetsize``.
+
+All other flags are reserved for future compatibility and must be zero.
 
 On success 0 is returned, on failure a negative error code is returned.
 

arch/riscv/include/asm/hwprobe.h

@@ -15,4 +15,28 @@ static inline bool riscv_hwprobe_key_is_valid(__s64 key)
 	return key >= 0 && key <= RISCV_HWPROBE_MAX_KEY;
 }
 
+static inline bool hwprobe_key_is_bitmask(__s64 key)
+{
+	switch (key) {
+	case RISCV_HWPROBE_KEY_BASE_BEHAVIOR:
+	case RISCV_HWPROBE_KEY_IMA_EXT_0:
+	case RISCV_HWPROBE_KEY_CPUPERF_0:
+		return true;
+	}
+
+	return false;
+}
+
+static inline bool riscv_hwprobe_pair_cmp(struct riscv_hwprobe *pair,
+					  struct riscv_hwprobe *other_pair)
+{
+	if (pair->key != other_pair->key)
+		return false;
+
+	if (hwprobe_key_is_bitmask(pair->key))
+		return (pair->value & other_pair->value) == other_pair->value;
+
+	return pair->value == other_pair->value;
+}
+
 #endif

arch/riscv/include/uapi/asm/hwprobe.h

@@ -66,4 +66,7 @@ struct riscv_hwprobe {
 #define RISCV_HWPROBE_KEY_ZICBOZ_BLOCK_SIZE	6
 /* Increase RISCV_HWPROBE_MAX_KEY when adding items. */
 
+/* Flags */
+#define RISCV_HWPROBE_WHICH_CPUS	(1 << 0)
+
 #endif

arch/riscv/kernel/Makefile

@@ -50,6 +50,7 @@ obj-y	+= setup.o
 obj-y	+= signal.o
 obj-y	+= syscall_table.o
 obj-y	+= sys_riscv.o
+obj-y	+= sys_hwprobe.o
 obj-y	+= time.o
 obj-y	+= traps.o
 obj-y	+= riscv_ksyms.o

arch/riscv/kernel/sys_hwprobe.c

+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * The hwprobe interface, for allowing userspace to probe to see which features
+ * are supported by the hardware.  See Documentation/arch/riscv/hwprobe.rst for
+ * more details.
+ */
+#include <linux/syscalls.h>
+#include <asm/cacheflush.h>
+#include <asm/cpufeature.h>
+#include <asm/hwprobe.h>
+#include <asm/sbi.h>
+#include <asm/switch_to.h>
+#include <asm/uaccess.h>
+#include <asm/unistd.h>
+#include <asm/vector.h>
+#include <vdso/vsyscall.h>
+
+
+static void hwprobe_arch_id(struct riscv_hwprobe *pair,
+			    const struct cpumask *cpus)
+{
+	u64 id = -1ULL;
+	bool first = true;
+	int cpu;
+
+	for_each_cpu(cpu, cpus) {
+		u64 cpu_id;
+
+		switch (pair->key) {
+		case RISCV_HWPROBE_KEY_MVENDORID:
+			cpu_id = riscv_cached_mvendorid(cpu);
+			break;
+		case RISCV_HWPROBE_KEY_MIMPID:
+			cpu_id = riscv_cached_mimpid(cpu);
+			break;
+		case RISCV_HWPROBE_KEY_MARCHID:
+			cpu_id = riscv_cached_marchid(cpu);
+			break;
+		}
+
+		if (first) {
+			id = cpu_id;
+			first = false;
+		}
+
+		/*
+		 * If there's a mismatch for the given set, return -1 in the
+		 * value.
+		 */
+		if (id != cpu_id) {
+			id = -1ULL;
+			break;
+		}
+	}
+
+	pair->value = id;
+}
+
+static void hwprobe_isa_ext0(struct riscv_hwprobe *pair,
+			     const struct cpumask *cpus)
+{
+	int cpu;
+	u64 missing = 0;
+
+	pair->value = 0;
+	if (has_fpu())
+		pair->value |= RISCV_HWPROBE_IMA_FD;
+
+	if (riscv_isa_extension_available(NULL, c))
+		pair->value |= RISCV_HWPROBE_IMA_C;
+
+	if (has_vector())
+		pair->value |= RISCV_HWPROBE_IMA_V;
+
+	/*
+	 * Loop through and record extensions that 1) anyone has, and 2) anyone
+	 * doesn't have.
+	 */
+	for_each_cpu(cpu, cpus) {
+		struct riscv_isainfo *isainfo = &hart_isa[cpu];
+
+#define EXT_KEY(ext)									\
+	do {										\
+		if (__riscv_isa_extension_available(isainfo->isa, RISCV_ISA_EXT_##ext))	\
+			pair->value |= RISCV_HWPROBE_EXT_##ext;				\
+		else									\
+			missing |= RISCV_HWPROBE_EXT_##ext;				\
+	} while (false)
+
+		/*
+		 * Only use EXT_KEY() for extensions which can be exposed to userspace,
+		 * regardless of the kernel's configuration, as no other checks, besides
+		 * presence in the hart_isa bitmap, are made.
+		 */
+		EXT_KEY(ZBA);
+		EXT_KEY(ZBB);
+		EXT_KEY(ZBS);
+		EXT_KEY(ZICBOZ);
+		EXT_KEY(ZBC);
+
+		EXT_KEY(ZBKB);
+		EXT_KEY(ZBKC);
+		EXT_KEY(ZBKX);
+		EXT_KEY(ZKND);
+		EXT_KEY(ZKNE);
+		EXT_KEY(ZKNH);
+		EXT_KEY(ZKSED);
+		EXT_KEY(ZKSH);
+		EXT_KEY(ZKT);
+		EXT_KEY(ZIHINTNTL);
+
+		if (has_vector()) {
+			EXT_KEY(ZVBB);
+			EXT_KEY(ZVBC);
+			EXT_KEY(ZVKB);
+			EXT_KEY(ZVKG);
+			EXT_KEY(ZVKNED);
+			EXT_KEY(ZVKNHA);
+			EXT_KEY(ZVKNHB);
+			EXT_KEY(ZVKSED);
+			EXT_KEY(ZVKSH);
+			EXT_KEY(ZVKT);
+			EXT_KEY(ZVFH);
+			EXT_KEY(ZVFHMIN);
+		}
+
+		if (has_fpu()) {
+			EXT_KEY(ZFH);
+			EXT_KEY(ZFHMIN);
+			EXT_KEY(ZFA);
+		}
+#undef EXT_KEY
+	}
+
+	/* Now turn off reporting features if any CPU is missing it. */
+	pair->value &= ~missing;
+}
+
+static bool hwprobe_ext0_has(const struct cpumask *cpus, unsigned long ext)
+{
+	struct riscv_hwprobe pair;
+
+	hwprobe_isa_ext0(&pair, cpus);
+	return (pair.value & ext);
+}
+
+static u64 hwprobe_misaligned(const struct cpumask *cpus)
+{
+	int cpu;
+	u64 perf = -1ULL;
+
+	for_each_cpu(cpu, cpus) {
+		int this_perf = per_cpu(misaligned_access_speed, cpu);
+
+		if (perf == -1ULL)
+			perf = this_perf;
+
+		if (perf != this_perf) {
+			perf = RISCV_HWPROBE_MISALIGNED_UNKNOWN;
+			break;
+		}
+	}
+
+	if (perf == -1ULL)
+		return RISCV_HWPROBE_MISALIGNED_UNKNOWN;
+
+	return perf;
+}
+
+static void hwprobe_one_pair(struct riscv_hwprobe *pair,
+			     const struct cpumask *cpus)
+{
+	switch (pair->key) {
+	case RISCV_HWPROBE_KEY_MVENDORID:
+	case RISCV_HWPROBE_KEY_MARCHID:
+	case RISCV_HWPROBE_KEY_MIMPID:
+		hwprobe_arch_id(pair, cpus);
+		break;
+	/*
+	 * The kernel already assumes that the base single-letter ISA
+	 * extensions are supported on all harts, and only supports the
+	 * IMA base, so just cheat a bit here and tell that to
+	 * userspace.
+	 */
+	case RISCV_HWPROBE_KEY_BASE_BEHAVIOR:
+		pair->value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA;
+		break;
+
+	case RISCV_HWPROBE_KEY_IMA_EXT_0:
+		hwprobe_isa_ext0(pair, cpus);
+		break;
+
+	case RISCV_HWPROBE_KEY_CPUPERF_0:
+		pair->value = hwprobe_misaligned(cpus);
+		break;
+
+	case RISCV_HWPROBE_KEY_ZICBOZ_BLOCK_SIZE:
+		pair->value = 0;
+		if (hwprobe_ext0_has(cpus, RISCV_HWPROBE_EXT_ZICBOZ))
+			pair->value = riscv_cboz_block_size;
+		break;
+
+	/*
+	 * For forward compatibility, unknown keys don't fail the whole
+	 * call, but get their element key set to -1 and value set to 0
+	 * indicating they're unrecognized.
+	 */
+	default:
+		pair->key = -1;
+		pair->value = 0;
+		break;
+	}
+}
+
+static int hwprobe_get_values(struct riscv_hwprobe __user *pairs,
+			      size_t pair_count, size_t cpusetsize,
+			      unsigned long __user *cpus_user,
+			      unsigned int flags)
+{
+	size_t out;
+	int ret;
+	cpumask_t cpus;
+
+	/* Check the reserved flags. */
+	if (flags != 0)
+		return -EINVAL;
+
+	/*
+	 * The interface supports taking in a CPU mask, and returns values that
+	 * are consistent across that mask. Allow userspace to specify NULL and
+	 * 0 as a shortcut to all online CPUs.
+	 */
+	cpumask_clear(&cpus);
+	if (!cpusetsize && !cpus_user) {
+		cpumask_copy(&cpus, cpu_online_mask);
+	} else {
+		if (cpusetsize > cpumask_size())
+			cpusetsize = cpumask_size();
+
+		ret = copy_from_user(&cpus, cpus_user, cpusetsize);
+		if (ret)
+			return -EFAULT;
+
+		/*
+		 * Userspace must provide at least one online CPU, without that
+		 * there's no way to define what is supported.
+		 */
+		cpumask_and(&cpus, &cpus, cpu_online_mask);
+		if (cpumask_empty(&cpus))
+			return -EINVAL;
+	}
+
+	for (out = 0; out < pair_count; out++, pairs++) {
+		struct riscv_hwprobe pair;
+
+		if (get_user(pair.key, &pairs->key))
+			return -EFAULT;
+
+		pair.value = 0;
+		hwprobe_one_pair(&pair, &cpus);
+		ret = put_user(pair.key, &pairs->key);
+		if (ret == 0)
+			ret = put_user(pair.value, &pairs->value);
+
+		if (ret)
+			return -EFAULT;
+	}
+
+	return 0;
+}
+
+static int hwprobe_get_cpus(struct riscv_hwprobe __user *pairs,
+			    size_t pair_count, size_t cpusetsize,
+			    unsigned long __user *cpus_user,
+			    unsigned int flags)
+{
+	cpumask_t cpus, one_cpu;
+	bool clear_all = false;
+	size_t i;
+	int ret;
+
+	if (flags != RISCV_HWPROBE_WHICH_CPUS)
+		return -EINVAL;
+
+	if (!cpusetsize || !cpus_user)
+		return -EINVAL;
+
+	if (cpusetsize > cpumask_size())
+		cpusetsize = cpumask_size();
+
+	ret = copy_from_user(&cpus, cpus_user, cpusetsize);
+	if (ret)
+		return -EFAULT;
+
+	if (cpumask_empty(&cpus))
+		cpumask_copy(&cpus, cpu_online_mask);
+
+	cpumask_and(&cpus, &cpus, cpu_online_mask);
+
+	cpumask_clear(&one_cpu);
+
+	for (i = 0; i < pair_count; i++) {
+		struct riscv_hwprobe pair, tmp;
+		int cpu;
+
+		ret = copy_from_user(&pair, &pairs[i], sizeof(pair));
+		if (ret)
+			return -EFAULT;
+
+		if (!riscv_hwprobe_key_is_valid(pair.key)) {
+			clear_all = true;
+			pair = (struct riscv_hwprobe){ .key = -1, };
+			ret = copy_to_user(&pairs[i], &pair, sizeof(pair));
+			if (ret)
+				return -EFAULT;
+		}
+
+		if (clear_all)
+			continue;
+
+		tmp = (struct riscv_hwprobe){ .key = pair.key, };
+
+		for_each_cpu(cpu, &cpus) {
+			cpumask_set_cpu(cpu, &one_cpu);
+
+			hwprobe_one_pair(&tmp, &one_cpu);
+
+			if (!riscv_hwprobe_pair_cmp(&tmp, &pair))
+				cpumask_clear_cpu(cpu, &cpus);
+
+			cpumask_clear_cpu(cpu, &one_cpu);
+		}
+	}
+
+	if (clear_all)
+		cpumask_clear(&cpus);
+
+	ret = copy_to_user(cpus_user, &cpus, cpusetsize);
+	if (ret)
+		return -EFAULT;
+
+	return 0;
+}
+
+static int do_riscv_hwprobe(struct riscv_hwprobe __user *pairs,
+			    size_t pair_count, size_t cpusetsize,
+			    unsigned long __user *cpus_user,
+			    unsigned int flags)
+{
+	if (flags & RISCV_HWPROBE_WHICH_CPUS)
+		return hwprobe_get_cpus(pairs, pair_count, cpusetsize,
+					cpus_user, flags);
+
+	return hwprobe_get_values(pairs, pair_count, cpusetsize,
+				  cpus_user, flags);
+}
+
+#ifdef CONFIG_MMU
+
+static int __init init_hwprobe_vdso_data(void)
+{
+	struct vdso_data *vd = __arch_get_k_vdso_data();
+	struct arch_vdso_data *avd = &vd->arch_data;
+	u64 id_bitsmash = 0;
+	struct riscv_hwprobe pair;
+	int key;
+
+	/*
+	 * Initialize vDSO data with the answers for the "all CPUs" case, to
+	 * save a syscall in the common case.
+	 */
+	for (key = 0; key <= RISCV_HWPROBE_MAX_KEY; key++) {
+		pair.key = key;
+		hwprobe_one_pair(&pair, cpu_online_mask);
+
+		WARN_ON_ONCE(pair.key < 0);
+
+		avd->all_cpu_hwprobe_values[key] = pair.value;
+		/*
+		 * Smash together the vendor, arch, and impl IDs to see if
+		 * they're all 0 or any negative.
+		 */
+		if (key <= RISCV_HWPROBE_KEY_MIMPID)
+			id_bitsmash |= pair.value;
+	}
+
+	/*
+	 * If the arch, vendor, and implementation ID are all the same across
+	 * all harts, then assume all CPUs are the same, and allow the vDSO to
+	 * answer queries for arbitrary masks. However if all values are 0 (not
+	 * populated) or any value returns -1 (varies across CPUs), then the
+	 * vDSO should defer to the kernel for exotic cpu masks.
+	 */
+	avd->homogeneous_cpus = id_bitsmash != 0 && id_bitsmash != -1;
+	return 0;
+}
+
+arch_initcall_sync(init_hwprobe_vdso_data);
+
+#endif /* CONFIG_MMU */
+
+SYSCALL_DEFINE5(riscv_hwprobe, struct riscv_hwprobe __user *, pairs,
+		size_t, pair_count, size_t, cpusetsize, unsigned long __user *,
+		cpus, unsigned int, flags)
+{
+	return do_riscv_hwprobe(pairs, pair_count, cpusetsize,
+				cpus, flags);
+}

arch/riscv/kernel/sys_riscv.c

@@ -7,15 +7,7 @@
 
 #include <linux/syscalls.h>
 #include <asm/cacheflush.h>
-#include <asm/cpufeature.h>
-#include <asm/hwprobe.h>
-#include <asm/sbi.h>
-#include <asm/vector.h>
-#include <asm/switch_to.h>
-#include <asm/uaccess.h>
-#include <asm/unistd.h>
 #include <asm-generic/mman-common.h>
-#include <vdso/vsyscall.h>
 
 static long riscv_sys_mmap(unsigned long addr, unsigned long len,
 			   unsigned long prot, unsigned long flags,
@@ -77,316 +69,6 @@ SYSCALL_DEFINE3(riscv_flush_icache, uintptr_t, start, uintptr_t, end,
 	return 0;
 }
 
-/*
- * The hwprobe interface, for allowing userspace to probe to see which features
- * are supported by the hardware.  See Documentation/arch/riscv/hwprobe.rst for more
- * details.
- */
-static void hwprobe_arch_id(struct riscv_hwprobe *pair,
-			    const struct cpumask *cpus)
-{
-	u64 id = -1ULL;
-	bool first = true;
-	int cpu;
-
-	for_each_cpu(cpu, cpus) {
-		u64 cpu_id;
-
-		switch (pair->key) {
-		case RISCV_HWPROBE_KEY_MVENDORID:
-			cpu_id = riscv_cached_mvendorid(cpu);
-			break;
-		case RISCV_HWPROBE_KEY_MIMPID:
-			cpu_id = riscv_cached_mimpid(cpu);
-			break;
-		case RISCV_HWPROBE_KEY_MARCHID:
-			cpu_id = riscv_cached_marchid(cpu);
-			break;
-		}
-
-		if (first) {
-			id = cpu_id;
-			first = false;
-		}
-
-		/*
-		 * If there's a mismatch for the given set, return -1 in the
-		 * value.
-		 */
-		if (id != cpu_id) {
-			id = -1ULL;
-			break;
-		}
-	}
-
-	pair->value = id;
-}
-
-static void hwprobe_isa_ext0(struct riscv_hwprobe *pair,
-			     const struct cpumask *cpus)
-{
-	int cpu;
-	u64 missing = 0;
-
-	pair->value = 0;
-	if (has_fpu())
-		pair->value |= RISCV_HWPROBE_IMA_FD;
-
-	if (riscv_isa_extension_available(NULL, c))
-		pair->value |= RISCV_HWPROBE_IMA_C;
-
-	if (has_vector())
-		pair->value |= RISCV_HWPROBE_IMA_V;
-
-	/*
-	 * Loop through and record extensions that 1) anyone has, and 2) anyone
-	 * doesn't have.
-	 */
-	for_each_cpu(cpu, cpus) {
-		struct riscv_isainfo *isainfo = &hart_isa[cpu];
-
-#define EXT_KEY(ext)									\
-	do {										\
-		if (__riscv_isa_extension_available(isainfo->isa, RISCV_ISA_EXT_##ext))	\
-			pair->value |= RISCV_HWPROBE_EXT_##ext;				\
-		else									\
-			missing |= RISCV_HWPROBE_EXT_##ext;				\
-	} while (false)
-
-		/*
-		 * Only use EXT_KEY() for extensions which can be exposed to userspace,
-		 * regardless of the kernel's configuration, as no other checks, besides
-		 * presence in the hart_isa bitmap, are made.
-		 */
-		EXT_KEY(ZBA);
-		EXT_KEY(ZBB);
-		EXT_KEY(ZBS);
-		EXT_KEY(ZICBOZ);
-		EXT_KEY(ZBC);
-
-		EXT_KEY(ZBKB);
-		EXT_KEY(ZBKC);
-		EXT_KEY(ZBKX);
-		EXT_KEY(ZKND);
-		EXT_KEY(ZKNE);
-		EXT_KEY(ZKNH);
-		EXT_KEY(ZKSED);
-		EXT_KEY(ZKSH);
-		EXT_KEY(ZKT);
-		EXT_KEY(ZIHINTNTL);
-
-		if (has_vector()) {
-			EXT_KEY(ZVBB);
-			EXT_KEY(ZVBC);
-			EXT_KEY(ZVKB);
-			EXT_KEY(ZVKG);
-			EXT_KEY(ZVKNED);
-			EXT_KEY(ZVKNHA);
-			EXT_KEY(ZVKNHB);
-			EXT_KEY(ZVKSED);
-			EXT_KEY(ZVKSH);
-			EXT_KEY(ZVKT);
-			EXT_KEY(ZVFH);
-			EXT_KEY(ZVFHMIN);
-		}
-
-		if (has_fpu()) {
-			EXT_KEY(ZFH);
-			EXT_KEY(ZFHMIN);
-			EXT_KEY(ZFA);
-		}
-#undef EXT_KEY
-	}
-
-	/* Now turn off reporting features if any CPU is missing it. */
-	pair->value &= ~missing;
-}
-
-static bool hwprobe_ext0_has(const struct cpumask *cpus, unsigned long ext)
-{
-	struct riscv_hwprobe pair;
-
-	hwprobe_isa_ext0(&pair, cpus);
-	return (pair.value & ext);
-}
-
-static u64 hwprobe_misaligned(const struct cpumask *cpus)
-{
-	int cpu;
-	u64 perf = -1ULL;
-
-	for_each_cpu(cpu, cpus) {
-		int this_perf = per_cpu(misaligned_access_speed, cpu);
-
-		if (perf == -1ULL)
-			perf = this_perf;
-
-		if (perf != this_perf) {
-			perf = RISCV_HWPROBE_MISALIGNED_UNKNOWN;
-			break;
-		}
-	}
-
-	if (perf == -1ULL)
-		return RISCV_HWPROBE_MISALIGNED_UNKNOWN;
-
-	return perf;
-}
-
-static void hwprobe_one_pair(struct riscv_hwprobe *pair,
-			     const struct cpumask *cpus)
-{
-	switch (pair->key) {
-	case RISCV_HWPROBE_KEY_MVENDORID:
-	case RISCV_HWPROBE_KEY_MARCHID:
-	case RISCV_HWPROBE_KEY_MIMPID:
-		hwprobe_arch_id(pair, cpus);
-		break;
-	/*
-	 * The kernel already assumes that the base single-letter ISA
-	 * extensions are supported on all harts, and only supports the
-	 * IMA base, so just cheat a bit here and tell that to
-	 * userspace.
-	 */
-	case RISCV_HWPROBE_KEY_BASE_BEHAVIOR:
-		pair->value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA;
-		break;
-
-	case RISCV_HWPROBE_KEY_IMA_EXT_0:
-		hwprobe_isa_ext0(pair, cpus);
-		break;
-
-	case RISCV_HWPROBE_KEY_CPUPERF_0:
-		pair->value = hwprobe_misaligned(cpus);
-		break;
-
-	case RISCV_HWPROBE_KEY_ZICBOZ_BLOCK_SIZE:
-		pair->value = 0;
-		if (hwprobe_ext0_has(cpus, RISCV_HWPROBE_EXT_ZICBOZ))
-			pair->value = riscv_cboz_block_size;
-		break;
-
-	/*
-	 * For forward compatibility, unknown keys don't fail the whole
-	 * call, but get their element key set to -1 and value set to 0
-	 * indicating they're unrecognized.
-	 */
-	default:
-		pair->key = -1;
-		pair->value = 0;
-		break;
-	}
-}
-
-static int do_riscv_hwprobe(struct riscv_hwprobe __user *pairs,
-			    size_t pair_count, size_t cpu_count,
-			    unsigned long __user *cpus_user,
-			    unsigned int flags)
-{
-	size_t out;
-	int ret;
-	cpumask_t cpus;
-
-	/* Check the reserved flags. */
-	if (flags != 0)
-		return -EINVAL;
-
-	/*
-	 * The interface supports taking in a CPU mask, and returns values that
-	 * are consistent across that mask. Allow userspace to specify NULL and
-	 * 0 as a shortcut to all online CPUs.
-	 */
-	cpumask_clear(&cpus);
-	if (!cpu_count && !cpus_user) {
-		cpumask_copy(&cpus, cpu_online_mask);
-	} else {
-		if (cpu_count > cpumask_size())
-			cpu_count = cpumask_size();
-
-		ret = copy_from_user(&cpus, cpus_user, cpu_count);
-		if (ret)
-			return -EFAULT;
-
-		/*
-		 * Userspace must provide at least one online CPU, without that
-		 * there's no way to define what is supported.
-		 */
-		cpumask_and(&cpus, &cpus, cpu_online_mask);
-		if (cpumask_empty(&cpus))
-			return -EINVAL;
-	}
-
-	for (out = 0; out < pair_count; out++, pairs++) {
-		struct riscv_hwprobe pair;
-
-		if (get_user(pair.key, &pairs->key))
-			return -EFAULT;
-
-		pair.value = 0;
-		hwprobe_one_pair(&pair, &cpus);
-		ret = put_user(pair.key, &pairs->key);
-		if (ret == 0)
-			ret = put_user(pair.value, &pairs->value);
-
-		if (ret)
-			return -EFAULT;
-	}
-
-	return 0;
-}
-
-#ifdef CONFIG_MMU
-
-static int __init init_hwprobe_vdso_data(void)
-{
-	struct vdso_data *vd = __arch_get_k_vdso_data();
-	struct arch_vdso_data *avd = &vd->arch_data;
-	u64 id_bitsmash = 0;
-	struct riscv_hwprobe pair;
-	int key;
-
-	/*
-	 * Initialize vDSO data with the answers for the "all CPUs" case, to
-	 * save a syscall in the common case.
-	 */
-	for (key = 0; key <= RISCV_HWPROBE_MAX_KEY; key++) {
-		pair.key = key;
-		hwprobe_one_pair(&pair, cpu_online_mask);
-
-		WARN_ON_ONCE(pair.key < 0);
-
-		avd->all_cpu_hwprobe_values[key] = pair.value;
-		/*
-		 * Smash together the vendor, arch, and impl IDs to see if
-		 * they're all 0 or any negative.
-		 */
-		if (key <= RISCV_HWPROBE_KEY_MIMPID)
-			id_bitsmash |= pair.value;
-	}
-
-	/*
-	 * If the arch, vendor, and implementation ID are all the same across
-	 * all harts, then assume all CPUs are the same, and allow the vDSO to
-	 * answer queries for arbitrary masks. However if all values are 0 (not
-	 * populated) or any value returns -1 (varies across CPUs), then the
-	 * vDSO should defer to the kernel for exotic cpu masks.
-	 */
-	avd->homogeneous_cpus = id_bitsmash != 0 && id_bitsmash != -1;
-	return 0;
-}
-
-arch_initcall_sync(init_hwprobe_vdso_data);
-
-#endif /* CONFIG_MMU */
-
-SYSCALL_DEFINE5(riscv_hwprobe, struct riscv_hwprobe __user *, pairs,
-		size_t, pair_count, size_t, cpu_count, unsigned long __user *,
-		cpus, unsigned int, flags)
-{
-	return do_riscv_hwprobe(pairs, pair_count, cpu_count,
-				cpus, flags);
-}
-
 /* Not defined using SYSCALL_DEFINE0 to avoid error injection */
 asmlinkage long __riscv_sys_ni_syscall(const struct pt_regs *__unused)
 {

arch/riscv/kernel/vdso/hwprobe.c

@@ -3,26 +3,22 @@
  * Copyright 2023 Rivos, Inc
  */
 
+#include <linux/string.h>
 #include <linux/types.h>
 #include <vdso/datapage.h>
 #include <vdso/helpers.h>
 
 extern int riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
-			 size_t cpu_count, unsigned long *cpus,
+			 size_t cpusetsize, unsigned long *cpus,
 			 unsigned int flags);
 
-/* Add a prototype to avoid -Wmissing-prototypes warning. */
-int __vdso_riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
-			 size_t cpu_count, unsigned long *cpus,
-			 unsigned int flags);
-
-int __vdso_riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
-			 size_t cpu_count, unsigned long *cpus,
+static int riscv_vdso_get_values(struct riscv_hwprobe *pairs, size_t pair_count,
+				 size_t cpusetsize, unsigned long *cpus,
 				 unsigned int flags)
 {
 	const struct vdso_data *vd = __arch_get_vdso_data();
 	const struct arch_vdso_data *avd = &vd->arch_data;
-	bool all_cpus = !cpu_count && !cpus;
+	bool all_cpus = !cpusetsize && !cpus;
 	struct riscv_hwprobe *p = pairs;
 	struct riscv_hwprobe *end = pairs + pair_count;
 
@@ -33,7 +29,7 @@ int __vdso_riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
 	 * masks.
 	 */
 	if ((flags != 0) || (!all_cpus && !avd->homogeneous_cpus))
-		return riscv_hwprobe(pairs, pair_count, cpu_count, cpus, flags);
+		return riscv_hwprobe(pairs, pair_count, cpusetsize, cpus, flags);
 
 	/* This is something we can handle, fill out the pairs. */
 	while (p < end) {
@@ -50,3 +46,71 @@ int __vdso_riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
 
 	return 0;
 }
+
+static int riscv_vdso_get_cpus(struct riscv_hwprobe *pairs, size_t pair_count,
+			       size_t cpusetsize, unsigned long *cpus,
+			       unsigned int flags)
+{
+	const struct vdso_data *vd = __arch_get_vdso_data();
+	const struct arch_vdso_data *avd = &vd->arch_data;
+	struct riscv_hwprobe *p = pairs;
+	struct riscv_hwprobe *end = pairs + pair_count;
+	unsigned char *c = (unsigned char *)cpus;
+	bool empty_cpus = true;
+	bool clear_all = false;
+	int i;
+
+	if (!cpusetsize || !cpus)
+		return -EINVAL;
+
+	for (i = 0; i < cpusetsize; i++) {
+		if (c[i]) {
+			empty_cpus = false;
+			break;
+		}
+	}
+
+	if (empty_cpus || flags != RISCV_HWPROBE_WHICH_CPUS || !avd->homogeneous_cpus)
+		return riscv_hwprobe(pairs, pair_count, cpusetsize, cpus, flags);
+
+	while (p < end) {
+		if (riscv_hwprobe_key_is_valid(p->key)) {
+			struct riscv_hwprobe t = {
+				.key = p->key,
+				.value = avd->all_cpu_hwprobe_values[p->key],
+			};
+
+			if (!riscv_hwprobe_pair_cmp(&t, p))
+				clear_all = true;
+		} else {
+			clear_all = true;
+			p->key = -1;
+			p->value = 0;
+		}
+		p++;
+	}
+
+	if (clear_all) {
+		for (i = 0; i < cpusetsize; i++)
+			c[i] = 0;
+	}
+
+	return 0;
+}
+
+/* Add a prototype to avoid -Wmissing-prototypes warning. */
+int __vdso_riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
+			 size_t cpusetsize, unsigned long *cpus,
+			 unsigned int flags);
+
+int __vdso_riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
+			 size_t cpusetsize, unsigned long *cpus,
+			 unsigned int flags)
+{
+	if (flags & RISCV_HWPROBE_WHICH_CPUS)
+		return riscv_vdso_get_cpus(pairs, pair_count, cpusetsize,
+					   cpus, flags);
+
+	return riscv_vdso_get_values(pairs, pair_count, cpusetsize,
+				     cpus, flags);
+}

tools/testing/selftests/riscv/hwprobe/Makefile

@@ -4,7 +4,7 @@
 
 CFLAGS += -I$(top_srcdir)/tools/include
 
-TEST_GEN_PROGS := hwprobe cbo
+TEST_GEN_PROGS := hwprobe cbo which-cpus
 
 include ../../lib.mk
 
@@ -13,3 +13,6 @@ $(OUTPUT)/hwprobe: hwprobe.c sys_hwprobe.S
 
 $(OUTPUT)/cbo: cbo.c sys_hwprobe.S
 	$(CC) -static -o$@ $(CFLAGS) $(LDFLAGS) $^
+
+$(OUTPUT)/which-cpus: which-cpus.c sys_hwprobe.S
+	$(CC) -static -o$@ $(CFLAGS) $(LDFLAGS) $^

tools/testing/selftests/riscv/hwprobe/hwprobe.c

@@ -47,7 +47,7 @@ int main(int argc, char **argv)
 	ksft_test_result(out != 0, "Bad CPU set\n");
 
 	out = riscv_hwprobe(pairs, 8, 1, 0, 0);
-	ksft_test_result(out != 0, "NULL CPU set with non-zero count\n");
+	ksft_test_result(out != 0, "NULL CPU set with non-zero size\n");
 
 	pairs[0].key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR;
 	out = riscv_hwprobe(pairs, 1, 1, &cpus, 0);

tools/testing/selftests/riscv/hwprobe/hwprobe.h

@@ -10,6 +10,6 @@
  * contain the call.
  */
 long riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
-		   size_t cpu_count, unsigned long *cpus, unsigned int flags);
+		   size_t cpusetsize, unsigned long *cpus, unsigned int flags);
 
 #endif

tools/testing/selftests/riscv/hwprobe/which-cpus.c

+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2023 Ventana Micro Systems Inc.
+ *
+ * Test the RISCV_HWPROBE_WHICH_CPUS flag of hwprobe. Also provides a command
+ * line interface to get the cpu list for arbitrary hwprobe pairs.
+ */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sched.h>
+#include <unistd.h>
+#include <assert.h>
+
+#include "hwprobe.h"
+#include "../../kselftest.h"
+
+static void help(void)
+{
+	printf("\n"
+	       "which-cpus: [-h] [<key=value> [<key=value> ...]]\n\n"
+	       "   Without parameters, tests the RISCV_HWPROBE_WHICH_CPUS flag of hwprobe.\n"
+	       "   With parameters, where each parameter is a hwprobe pair written as\n"
+	       "   <key=value>, outputs the cpulist for cpus which all match the given set\n"
+	       "   of pairs.  'key' and 'value' should be in numeric form, e.g. 4=0x3b\n");
+}
+
+static void print_cpulist(cpu_set_t *cpus)
+{
+	int start = 0, end = 0;
+
+	if (!CPU_COUNT(cpus)) {
+		printf("cpus: None\n");
+		return;
+	}
+
+	printf("cpus:");
+	for (int i = 0, c = 0; i < CPU_COUNT(cpus); i++, c++) {
+		if (start != end && !CPU_ISSET(c, cpus))
+			printf("-%d", end);
+
+		while (!CPU_ISSET(c, cpus))
+			++c;
+
+		if (i != 0 && c == end + 1) {
+			end = c;
+			continue;
+		}
+
+		printf("%c%d", i == 0 ? ' ' : ',', c);
+		start = end = c;
+	}
+	if (start != end)
+		printf("-%d", end);
+	printf("\n");
+}
+
+static void do_which_cpus(int argc, char **argv, cpu_set_t *cpus)
+{
+	struct riscv_hwprobe *pairs;
+	int nr_pairs = argc - 1;
+	char *start, *end;
+	int rc;
+
+	pairs = malloc(nr_pairs * sizeof(struct riscv_hwprobe));
+	assert(pairs);
+
+	for (int i = 0; i < nr_pairs; i++) {
+		start = argv[i + 1];
+		pairs[i].key = strtol(start, &end, 0);
+		assert(end != start && *end == '=');
+		start = end + 1;
+		pairs[i].value = strtoul(start, &end, 0);
+		assert(end != start && *end == '\0');
+	}
+
+	rc = riscv_hwprobe(pairs, nr_pairs, sizeof(cpu_set_t), (unsigned long *)cpus, RISCV_HWPROBE_WHICH_CPUS);
+	assert(rc == 0);
+	print_cpulist(cpus);
+	free(pairs);
+}
+
+int main(int argc, char **argv)
+{
+	struct riscv_hwprobe pairs[2];
+	cpu_set_t cpus_aff, cpus;
+	__u64 ext0_all;
+	long rc;
+
+	rc = sched_getaffinity(0, sizeof(cpu_set_t), &cpus_aff);
+	assert(rc == 0);
+
+	if (argc > 1) {
+		if (!strcmp(argv[1], "-h"))
+			help();
+		else
+			do_which_cpus(argc, argv, &cpus_aff);
+		return 0;
+	}
+
+	ksft_print_header();
+	ksft_set_plan(7);
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, };
+	rc = riscv_hwprobe(pairs, 1, 0, NULL, 0);
+	assert(rc == 0 && pairs[0].key == RISCV_HWPROBE_KEY_BASE_BEHAVIOR &&
+	       pairs[0].value == RISCV_HWPROBE_BASE_BEHAVIOR_IMA);
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_IMA_EXT_0, };
+	rc = riscv_hwprobe(pairs, 1, 0, NULL, 0);
+	assert(rc == 0 && pairs[0].key == RISCV_HWPROBE_KEY_IMA_EXT_0);
+	ext0_all = pairs[0].value;
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, .value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA, };
+	CPU_ZERO(&cpus);
+	rc = riscv_hwprobe(pairs, 1, 0, (unsigned long *)&cpus, RISCV_HWPROBE_WHICH_CPUS);
+	ksft_test_result(rc == -EINVAL, "no cpusetsize\n");
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, .value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA, };
+	rc = riscv_hwprobe(pairs, 1, sizeof(cpu_set_t), NULL, RISCV_HWPROBE_WHICH_CPUS);
+	ksft_test_result(rc == -EINVAL, "NULL cpus\n");
+
+	pairs[0] = (struct riscv_hwprobe){ .key = 0xbadc0de, };
+	CPU_ZERO(&cpus);
+	rc = riscv_hwprobe(pairs, 1, sizeof(cpu_set_t), (unsigned long *)&cpus, RISCV_HWPROBE_WHICH_CPUS);
+	ksft_test_result(rc == 0 && CPU_COUNT(&cpus) == 0, "unknown key\n");
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, .value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA, };
+	pairs[1] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, .value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA, };
+	CPU_ZERO(&cpus);
+	rc = riscv_hwprobe(pairs, 2, sizeof(cpu_set_t), (unsigned long *)&cpus, RISCV_HWPROBE_WHICH_CPUS);
+	ksft_test_result(rc == 0, "duplicate keys\n");
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, .value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA, };
+	pairs[1] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_IMA_EXT_0, .value = ext0_all, };
+	CPU_ZERO(&cpus);
+	rc = riscv_hwprobe(pairs, 2, sizeof(cpu_set_t), (unsigned long *)&cpus, RISCV_HWPROBE_WHICH_CPUS);
+	ksft_test_result(rc == 0 && CPU_COUNT(&cpus) == sysconf(_SC_NPROCESSORS_ONLN), "set all cpus\n");
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, .value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA, };
+	pairs[1] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_IMA_EXT_0, .value = ext0_all, };
+	memcpy(&cpus, &cpus_aff, sizeof(cpu_set_t));
+	rc = riscv_hwprobe(pairs, 2, sizeof(cpu_set_t), (unsigned long *)&cpus, RISCV_HWPROBE_WHICH_CPUS);
+	ksft_test_result(rc == 0 && CPU_EQUAL(&cpus, &cpus_aff), "set all affinity cpus\n");
+
+	pairs[0] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_BASE_BEHAVIOR, .value = RISCV_HWPROBE_BASE_BEHAVIOR_IMA, };
+	pairs[1] = (struct riscv_hwprobe){ .key = RISCV_HWPROBE_KEY_IMA_EXT_0, .value = ~ext0_all, };
+	memcpy(&cpus, &cpus_aff, sizeof(cpu_set_t));
+	rc = riscv_hwprobe(pairs, 2, sizeof(cpu_set_t), (unsigned long *)&cpus, RISCV_HWPROBE_WHICH_CPUS);
+	ksft_test_result(rc == 0 && CPU_COUNT(&cpus) == 0, "clear all cpus\n");
+
+	ksft_finished();
+}

tools/testing/selftests/riscv/vector/vstate_prctl.c

 // SPDX-License-Identifier: GPL-2.0-only
 #include <sys/prctl.h>
 #include <unistd.h>
-#include <asm/hwprobe.h>
 #include <errno.h>
 #include <sys/wait.h>
 
+#include "../hwprobe/hwprobe.h"
 #include "../../kselftest.h"
 
-/*
- * Rather than relying on having a new enough libc to define this, just do it
- * ourselves.  This way we don't need to be coupled to a new-enough libc to
- * contain the call.
- */
-long riscv_hwprobe(struct riscv_hwprobe *pairs, size_t pair_count,
-		   size_t cpu_count, unsigned long *cpus, unsigned int flags);
-
 #define NEXT_PROGRAM "./vstate_exec_nolibc"
 static int launch_test(int test_inherit)
 {