Merge patch series "membarrier: riscv: Core serializing command"

palmer-dabbelt · palmer-dabbelt · commit 0420af54c2c2 · 2024-02-15T08:04:23.000-08:00
RISC-V was lacking a membarrier implementation for the store/fetch ordering, which is a bit tricky because of the deferred icache flushing we use in RISC-V. * b4-shazam-merge: membarrier: riscv: Provide core serializing command locking: Introduce prepare_sync_core_cmd() membarrier: Create Documentation/scheduler/membarrier.rst membarrier: riscv: Add full memory barrier in switch_mm() Link: https://lore.kernel.org/r/20240131144936.29190-1-parri.andrea@gmail.com Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
diff --git a/Documentation/features/sched/membarrier-sync-core/arch-support.txt b/Documentation/features/sched/membarrier-sync-core/arch-support.txt
@@ -10,6 +10,22 @@
 # Rely on implicit context synchronization as a result of exception return
 # when returning from IPI handler, and when returning to user-space.
 #
+# * riscv
+#
+# riscv uses xRET as return from interrupt and to return to user-space.
+#
+# Given that xRET is not core serializing, we rely on FENCE.I for providing
+# core serialization:
+#
+#  - by calling sync_core_before_usermode() on return from interrupt (cf.
+#    ipi_sync_core()),
+#
+#  - via switch_mm() and sync_core_before_usermode() (respectively, for
+#    uthread->uthread and kthread->uthread transitions) before returning
+#    to user-space.
+#
+#  The serialization in switch_mm() is activated by prepare_sync_core_cmd().
+#
 # * x86
 #
 # x86-32 uses IRET as return from interrupt, which takes care of the IPI.
@@ -43,7 +59,7 @@
     |    openrisc: | TODO |
     |      parisc: | TODO |
     |     powerpc: |  ok  |
-    |       riscv: | TODO |
+    |       riscv: |  ok  |
     |        s390: |  ok  |
     |          sh: | TODO |
     |       sparc: | TODO |
diff --git a/Documentation/scheduler/index.rst b/Documentation/scheduler/index.rst
@@ -7,6 +7,7 @@ Scheduler
 
 
     completion
+    membarrier
     sched-arch
     sched-bwc
     sched-deadline
diff --git a/Documentation/scheduler/membarrier.rst b/Documentation/scheduler/membarrier.rst
@@ -0,0 +1,39 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+========================
+membarrier() System Call
+========================
+
+MEMBARRIER_CMD_{PRIVATE,GLOBAL}_EXPEDITED - Architecture requirements
+=====================================================================
+
+Memory barriers before updating rq->curr
+----------------------------------------
+
+The commands MEMBARRIER_CMD_PRIVATE_EXPEDITED and MEMBARRIER_CMD_GLOBAL_EXPEDITED
+require each architecture to have a full memory barrier after coming from
+user-space, before updating rq->curr.  This barrier is implied by the sequence
+rq_lock(); smp_mb__after_spinlock() in __schedule().  The barrier matches a full
+barrier in the proximity of the membarrier system call exit, cf.
+membarrier_{private,global}_expedited().
+
+Memory barriers after updating rq->curr
+---------------------------------------
+
+The commands MEMBARRIER_CMD_PRIVATE_EXPEDITED and MEMBARRIER_CMD_GLOBAL_EXPEDITED
+require each architecture to have a full memory barrier after updating rq->curr,
+before returning to user-space.  The schemes providing this barrier on the various
+architectures are as follows.
+
+ - alpha, arc, arm, hexagon, mips rely on the full barrier implied by
+   spin_unlock() in finish_lock_switch().
+
+ - arm64 relies on the full barrier implied by switch_to().
+
+ - powerpc, riscv, s390, sparc, x86 rely on the full barrier implied by
+   switch_mm(), if mm is not NULL; they rely on the full barrier implied
+   by mmdrop(), otherwise.  On powerpc and riscv, switch_mm() relies on
+   membarrier_arch_switch_mm().
+
+The barrier matches a full barrier in the proximity of the membarrier system call
+entry, cf. membarrier_{private,global}_expedited().
diff --git a/MAINTAINERS b/MAINTAINERS
@@ -14039,7 +14039,9 @@ M:	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 M:	"Paul E. McKenney" <paulmck@kernel.org>
 L:	linux-kernel@vger.kernel.org
 S:	Supported
-F:	arch/powerpc/include/asm/membarrier.h
+F:	Documentation/scheduler/membarrier.rst
+F:	arch/*/include/asm/membarrier.h
+F:	arch/*/include/asm/sync_core.h
 F:	include/uapi/linux/membarrier.h
 F:	kernel/sched/membarrier.c
 
diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
@@ -27,14 +27,18 @@ config RISCV
 	select ARCH_HAS_GCOV_PROFILE_ALL
 	select ARCH_HAS_GIGANTIC_PAGE
 	select ARCH_HAS_KCOV
+	select ARCH_HAS_MEMBARRIER_CALLBACKS
+	select ARCH_HAS_MEMBARRIER_SYNC_CORE
 	select ARCH_HAS_MMIOWB
 	select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
 	select ARCH_HAS_PMEM_API
+	select ARCH_HAS_PREPARE_SYNC_CORE_CMD
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_SET_DIRECT_MAP if MMU
 	select ARCH_HAS_SET_MEMORY if MMU
 	select ARCH_HAS_STRICT_KERNEL_RWX if MMU && !XIP_KERNEL
 	select ARCH_HAS_STRICT_MODULE_RWX if MMU && !XIP_KERNEL
+	select ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
 	select ARCH_HAS_SYSCALL_WRAPPER
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_HAS_UBSAN_SANITIZE_ALL
diff --git a/arch/riscv/include/asm/membarrier.h b/arch/riscv/include/asm/membarrier.h
@@ -0,0 +1,50 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+#ifndef _ASM_RISCV_MEMBARRIER_H
+#define _ASM_RISCV_MEMBARRIER_H
+
+static inline void membarrier_arch_switch_mm(struct mm_struct *prev,
+					     struct mm_struct *next,
+					     struct task_struct *tsk)
+{
+	/*
+	 * Only need the full barrier when switching between processes.
+	 * Barrier when switching from kernel to userspace is not
+	 * required here, given that it is implied by mmdrop(). Barrier
+	 * when switching from userspace to kernel is not needed after
+	 * store to rq->curr.
+	 */
+	if (IS_ENABLED(CONFIG_SMP) &&
+	    likely(!(atomic_read(&next->membarrier_state) &
+		     (MEMBARRIER_STATE_PRIVATE_EXPEDITED |
+		      MEMBARRIER_STATE_GLOBAL_EXPEDITED)) || !prev))
+		return;
+
+	/*
+	 * The membarrier system call requires a full memory barrier
+	 * after storing to rq->curr, before going back to user-space.
+	 *
+	 * This barrier is also needed for the SYNC_CORE command when
+	 * switching between processes; in particular, on a transition
+	 * from a thread belonging to another mm to a thread belonging
+	 * to the mm for which a membarrier SYNC_CORE is done on CPU0:
+	 *
+	 *   - [CPU0] sets all bits in the mm icache_stale_mask (in
+	 *     prepare_sync_core_cmd());
+	 *
+	 *   - [CPU1] stores to rq->curr (by the scheduler);
+	 *
+	 *   - [CPU0] loads rq->curr within membarrier and observes
+	 *     cpu_rq(1)->curr->mm != mm, so the IPI is skipped on
+	 *     CPU1; this means membarrier relies on switch_mm() to
+	 *     issue the sync-core;
+	 *
+	 *   - [CPU1] switch_mm() loads icache_stale_mask; if the bit
+	 *     is zero, switch_mm() may incorrectly skip the sync-core.
+	 *
+	 * Matches a full barrier in the proximity of the membarrier
+	 * system call entry.
+	 */
+	smp_mb();
+}
+
+#endif /* _ASM_RISCV_MEMBARRIER_H */
diff --git a/arch/riscv/include/asm/sync_core.h b/arch/riscv/include/asm/sync_core.h
@@ -0,0 +1,29 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _ASM_RISCV_SYNC_CORE_H
+#define _ASM_RISCV_SYNC_CORE_H
+
+/*
+ * RISC-V implements return to user-space through an xRET instruction,
+ * which is not core serializing.
+ */
+static inline void sync_core_before_usermode(void)
+{
+	asm volatile ("fence.i" ::: "memory");
+}
+
+#ifdef CONFIG_SMP
+/*
+ * Ensure the next switch_mm() on every CPU issues a core serializing
+ * instruction for the given @mm.
+ */
+static inline void prepare_sync_core_cmd(struct mm_struct *mm)
+{
+	cpumask_setall(&mm->context.icache_stale_mask);
+}
+#else
+static inline void prepare_sync_core_cmd(struct mm_struct *mm)
+{
+}
+#endif /* CONFIG_SMP */
+
+#endif /* _ASM_RISCV_SYNC_CORE_H */
diff --git a/arch/riscv/mm/context.c b/arch/riscv/mm/context.c
@@ -323,6 +323,8 @@ void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	if (unlikely(prev == next))
 		return;
 
+	membarrier_arch_switch_mm(prev, next, task);
+
 	/*
 	 * Mark the current MM context as inactive, and the next as
 	 * active.  This is at least used by the icache flushing
diff --git a/include/linux/sync_core.h b/include/linux/sync_core.h
@@ -17,5 +17,19 @@ static inline void sync_core_before_usermode(void)
 }
 #endif
 
-#endif /* _LINUX_SYNC_CORE_H */
+#ifdef CONFIG_ARCH_HAS_PREPARE_SYNC_CORE_CMD
+#include <asm/sync_core.h>
+#else
+/*
+ * This is a dummy prepare_sync_core_cmd() implementation that can be used on
+ * all architectures which provide unconditional core serializing instructions
+ * in switch_mm().
+ * If your architecture doesn't provide such core serializing instructions in
+ * switch_mm(), you may need to write your own functions.
+ */
+static inline void prepare_sync_core_cmd(struct mm_struct *mm)
+{
+}
+#endif
 
+#endif /* _LINUX_SYNC_CORE_H */
diff --git a/init/Kconfig b/init/Kconfig
@@ -1970,6 +1970,9 @@ source "kernel/Kconfig.locks"
 config ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
 	bool
 
+config ARCH_HAS_PREPARE_SYNC_CORE_CMD
+	bool
+
 config ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
 	bool
 
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
@@ -6638,7 +6638,9 @@ static void __sched notrace __schedule(unsigned int sched_mode)
 	 *     if (signal_pending_state())	    if (p->state & @state)
 	 *
 	 * Also, the membarrier system call requires a full memory barrier
-	 * after coming from user-space, before storing to rq->curr.
+	 * after coming from user-space, before storing to rq->curr; this
+	 * barrier matches a full barrier in the proximity of the membarrier
+	 * system call exit.
 	 */
 	rq_lock(rq, &rf);
 	smp_mb__after_spinlock();
@@ -6709,12 +6711,20 @@ static void __sched notrace __schedule(unsigned int sched_mode)
 		 *
 		 * Here are the schemes providing that barrier on the
 		 * various architectures:
-		 * - mm ? switch_mm() : mmdrop() for x86, s390, sparc, PowerPC.
-		 *   switch_mm() rely on membarrier_arch_switch_mm() on PowerPC.
+		 * - mm ? switch_mm() : mmdrop() for x86, s390, sparc, PowerPC,
+		 *   RISC-V.  switch_mm() relies on membarrier_arch_switch_mm()
+		 *   on PowerPC and on RISC-V.
 		 * - finish_lock_switch() for weakly-ordered
 		 *   architectures where spin_unlock is a full barrier,
 		 * - switch_to() for arm64 (weakly-ordered, spin_unlock
 		 *   is a RELEASE barrier),
+		 *
+		 * The barrier matches a full barrier in the proximity of
+		 * the membarrier system call entry.
+		 *
+		 * On RISC-V, this barrier pairing is also needed for the
+		 * SYNC_CORE command when switching between processes, cf.
+		 * the inline comments in membarrier_arch_switch_mm().
 		 */
 		++*switch_count;
 
diff --git a/kernel/sched/membarrier.c b/kernel/sched/membarrier.c
@@ -251,7 +251,7 @@ static int membarrier_global_expedited(void)
 		return 0;
 
 	/*
-	 * Matches memory barriers around rq->curr modification in
+	 * Matches memory barriers after rq->curr modification in
 	 * scheduler.
 	 */
 	smp_mb();	/* system call entry is not a mb. */
@@ -300,7 +300,7 @@ static int membarrier_global_expedited(void)
 
 	/*
 	 * Memory barrier on the caller thread _after_ we finished
-	 * waiting for the last IPI. Matches memory barriers around
+	 * waiting for the last IPI. Matches memory barriers before
 	 * rq->curr modification in scheduler.
 	 */
 	smp_mb();	/* exit from system call is not a mb */
@@ -320,6 +320,7 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 		      MEMBARRIER_STATE_PRIVATE_EXPEDITED_SYNC_CORE_READY))
 			return -EPERM;
 		ipi_func = ipi_sync_core;
+		prepare_sync_core_cmd(mm);
 	} else if (flags == MEMBARRIER_FLAG_RSEQ) {
 		if (!IS_ENABLED(CONFIG_RSEQ))
 			return -EINVAL;
@@ -339,8 +340,12 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 		return 0;
 
 	/*
-	 * Matches memory barriers around rq->curr modification in
+	 * Matches memory barriers after rq->curr modification in
 	 * scheduler.
+	 *
+	 * On RISC-V, this barrier pairing is also needed for the
+	 * SYNC_CORE command when switching between processes, cf.
+	 * the inline comments in membarrier_arch_switch_mm().
 	 */
 	smp_mb();	/* system call entry is not a mb. */
 
@@ -415,7 +420,7 @@ static int membarrier_private_expedited(int flags, int cpu_id)
 
 	/*
 	 * Memory barrier on the caller thread _after_ we finished
-	 * waiting for the last IPI. Matches memory barriers around
+	 * waiting for the last IPI. Matches memory barriers before
 	 * rq->curr modification in scheduler.
 	 */
 	smp_mb();	/* exit from system call is not a mb */
