[RFC PATCH 2/2] selftests/rseq: Implement rseq numa node id field selftest

@@ -22,6 +22,8 @@ void test_cpu_pointer(void)
 	CPU_ZERO(&test_affinity);
 	for (i = 0; i < CPU_SETSIZE; i++) {
 		if (CPU_ISSET(i, &affinity)) {
+			int node;
+
 			CPU_SET(i, &test_affinity);
 			sched_setaffinity(0, sizeof(test_affinity),
 					&test_affinity);

@@ -29,6 +31,10 @@ void test_cpu_pointer(void)
 			assert(rseq_current_cpu() == i);
 			assert(rseq_current_cpu_raw() == i);
 			assert(rseq_cpu_start() == i);
+			node = rseq_fallback_current_node();
+			assert(rseq_current_node() == node);
+			assert(rseq_current_node_raw() == node);
+			assert(rseq_node_start() == node);
 			CPU_CLR(i, &test_affinity);
 		}
 	}

@@ -13,9 +13,9 @@
 #include <linux/types.h>
 #include <asm/byteorder.h>
 
-enum rseq_abi_cpu_id_state {
-	RSEQ_ABI_CPU_ID_UNINITIALIZED			= -1,
-	RSEQ_ABI_CPU_ID_REGISTRATION_FAILED		= -2,
+enum rseq_abi_id_state {
+	RSEQ_ABI_ID_UNINITIALIZED			= -1,
+	RSEQ_ABI_ID_REGISTRATION_FAILED			= -2,
 };
 
 enum rseq_abi_flags {

@@ -146,6 +146,51 @@ struct rseq_abi {
 	 *     this thread.
 	 */
 	__u32 flags;
+	__u32 padding1[3];
+
+	/*
+	 * This is the end of the original rseq ABI.
+	 * This is a valid end of rseq ABI for the purpose of rseq registration
+	 * rseq_len.
+	 * The original rseq ABI use "sizeof(struct rseq)" on registration,
+	 * thus requiring the padding above.
+	 */
+
+	/*
+	 * Restartable sequences node_id_start field. Updated by the
+	 * kernel. Read by user-space with single-copy atomicity
+	 * semantics. This field should only be read by the thread which
+	 * registered this data structure. Aligned on 32-bit. Always
+	 * contains a value in the range of possible NUMA node IDs, although the
+	 * value may not be the actual current NUMA node ID (e.g. if rseq is not
+	 * initialized). This NUMA node ID number value should always be compared
+	 * against the value of the node_id field before performing a rseq
+	 * commit or returning a value read from a data structure indexed using
+	 * the node_id_start value.
+	 */
+	__u32 node_id_start;
+
+	/*
+	 * Restartable sequences node_id field. Updated by the kernel.
+	 * Read by user-space with single-copy atomicity semantics. This
+	 * field should only be read by the thread which registered this
+	 * data structure. Aligned on 32-bit. Values
+	 * RSEQ_ID_UNINITIALIZED and RSEQ_ID_REGISTRATION_FAILED
+	 * have a special semantic: the former means "rseq uninitialized",
+	 * and latter means "rseq initialization failed". This value is
+	 * meant to be read within rseq critical sections and compared
+	 * with the node_id_start value previously read, before performing
+	 * the commit instruction, or read and compared with the
+	 * node_id_start value before returning a value loaded from a data
+	 * structure indexed using the node_id_start value.
+	 */
+	__u32 node_id;
+
+	/*
+	 * This is a valid end of rseq ABI for the purpose of rseq registration
+	 * rseq_len. Use the offset immediately after the node_id field as
+	 * rseq_len.
+	 */
 } __attribute__((aligned(4 * sizeof(__u64))));
 
 #endif /* _RSEQ_ABI_H */

@@ -27,10 +27,20 @@
 #include <signal.h>
 #include <limits.h>
 #include <dlfcn.h>
+#include <stddef.h>
 
 #include "../kselftest.h"
 #include "rseq.h"
 
+#ifndef sizeof_field
+#define sizeof_field(TYPE, MEMBER) sizeof((((TYPE *)0)->MEMBER))
+#endif
+
+#ifndef offsetofend
+#define offsetofend(TYPE, MEMBER) \
+	(offsetof(TYPE, MEMBER)	+ sizeof_field(TYPE, MEMBER))
+#endif
+
 static const int *libc_rseq_offset_p;
 static const unsigned int *libc_rseq_size_p;
 static const unsigned int *libc_rseq_flags_p;

@@ -49,7 +59,8 @@ static int rseq_ownership;
 
 static
 __thread struct rseq_abi __rseq_abi __attribute__((tls_model("initial-exec"))) = {
-	.cpu_id = RSEQ_ABI_CPU_ID_UNINITIALIZED,
+	.cpu_id = RSEQ_ABI_ID_UNINITIALIZED,
+	.node_id = RSEQ_ABI_ID_UNINITIALIZED,
 };
 
 static int sys_rseq(struct rseq_abi *rseq_abi, uint32_t rseq_len,

@@ -58,6 +69,11 @@ static int sys_rseq(struct rseq_abi *rseq_abi, uint32_t rseq_len,
 	return syscall(__NR_rseq, rseq_abi, rseq_len, flags, sig);
 }
 
+static int sys_getcpu(unsigned *cpu, unsigned *node)
+{
+	return syscall(__NR_getcpu, cpu, node, NULL);
+}
+
 int rseq_available(void)
 {
 	int rc;

@@ -83,7 +99,7 @@ int rseq_register_current_thread(void)
 		/* Treat libc's ownership as a successful registration. */
 		return 0;
 	}
-	rc = sys_rseq(&__rseq_abi, sizeof(struct rseq_abi), 0, RSEQ_SIG);
+	rc = sys_rseq(&__rseq_abi, offsetofend(struct rseq_abi, node_id), 0, RSEQ_SIG);
 	if (rc)
 		return -1;
 	assert(rseq_current_cpu_raw() >= 0);

@@ -98,7 +114,7 @@ int rseq_unregister_current_thread(void)
 		/* Treat libc's ownership as a successful unregistration. */
 		return 0;
 	}
-	rc = sys_rseq(&__rseq_abi, sizeof(struct rseq_abi), RSEQ_ABI_FLAG_UNREGISTER, RSEQ_SIG);
+	rc = sys_rseq(&__rseq_abi, offsetofend(struct rseq_abi, node_id), RSEQ_ABI_FLAG_UNREGISTER, RSEQ_SIG);
 	if (rc)
 		return -1;
 	return 0;

@@ -121,7 +137,7 @@ void rseq_init(void)
 		return;
 	rseq_ownership = 1;
 	rseq_offset = (void *)&__rseq_abi - rseq_thread_pointer();
-	rseq_size = sizeof(struct rseq_abi);
+	rseq_size = offsetofend(struct rseq_abi, node_id);
 	rseq_flags = 0;
 }
 

@@ -146,3 +162,16 @@ int32_t rseq_fallback_current_cpu(void)
 	}
 	return cpu;
 }
+
+int32_t rseq_fallback_current_node(void)
+{
+	uint32_t cpu_id, node_id;
+	int ret;
+
+	ret = sys_getcpu(&cpu_id, &node_id);
+	if (ret) {
+		perror("sys_getcpu()");
+		return ret;
+	}
+	return (int32_t) node_id;
+}

@@ -115,6 +115,11 @@ int rseq_unregister_current_thread(void);
  */
 int32_t rseq_fallback_current_cpu(void);
 
+/*
+ * Restartable sequence fallback for reading the current node number.
+ */
+int32_t rseq_fallback_current_node(void);
+
 /*
  * Values returned can be either the current CPU number, -1 (rseq is
  * uninitialized), or -2 (rseq initialization has failed).

@@ -124,6 +129,15 @@ static inline int32_t rseq_current_cpu_raw(void)
 	return RSEQ_ACCESS_ONCE(rseq_get_abi()->cpu_id);
 }
 
+/*
+ * Values returned can be either the current NUMA node number, -1 (rseq is
+ * uninitialized), or -2 (rseq initialization has failed).
+ */
+static inline int32_t rseq_current_node_raw(void)
+{
+	return RSEQ_ACCESS_ONCE(rseq_get_abi()->node_id);
+}
+
 /*
  * Returns a possible CPU number, which is typically the current CPU.
  * The returned CPU number can be used to prepare for an rseq critical

@@ -140,6 +154,22 @@ static inline uint32_t rseq_cpu_start(void)
 	return RSEQ_ACCESS_ONCE(rseq_get_abi()->cpu_id_start);
 }
 
+/*
+ * Returns a possible NUMA node number, which is typically the current NUMA
+ * node.  The returned NUMA node number can be used to prepare for an rseq
+ * critical section, which will confirm whether the NUMA node number is indeed
+ * the current one, and whether rseq is initialized.
+ *
+ * The NUMA node number returned by rseq_node_start should always be validated
+ * by passing it to a rseq asm sequence, or by comparing it to the return value
+ * of rseq_current_node_raw() if the rseq asm sequence does not need to be
+ * invoked.
+ */
+static inline uint32_t rseq_node_start(void)
+{
+	return RSEQ_ACCESS_ONCE(rseq_get_abi()->node_id_start);
+}
+
 static inline uint32_t rseq_current_cpu(void)
 {
 	int32_t cpu;

@@ -150,6 +180,16 @@ static inline uint32_t rseq_current_cpu(void)
 	return cpu;
 }
 
+static inline uint32_t rseq_current_node(void)
+{
+	int32_t node;
+
+	node = rseq_current_node_raw();
+	if (rseq_unlikely(node < 0))
+		node = rseq_fallback_current_node();
+	return node;
+}
+
 static inline void rseq_clear_rseq_cs(void)
 {
 	RSEQ_WRITE_ONCE(rseq_get_abi()->rseq_cs.arch.ptr, 0);