Merge branches 'expcb.2026.07.24a', 'misc.2026.07.30a', 'rcu-tasks.2026.07.30a', 'srcu.2026.08.11a' and 'torture.2026.08.14a' into HEADgrafted

Changes:

Make expedited grace periods expedite normal RCU callbacks

Miscellaneous fixes:
 * Improve diagnostic output with character task states.
 * Mark accesses to inform KCSAN of concurrency design.
 * Move from kmalloc() to kmalloc_obj().
 * Documentation updates.
 * Improve handling of RCU deferred quiescent states.
 * Clean up unused function arguments and structure fields.
 * Reduce show_rcu_gp_kthreads() stack space.

Tasks RCU updates:
 * Clean up after SRCU re-implementation of Tasks Trace RCU.
 * Mark accesses to inform KCSAN of concurrency design.
 * Add ->lazy_timer status to diagnostic output.
 * Remove an unnecessary memory barrier.
 * Fix a data race, courtesy of KCSAN.
 * Documentation updates.
 * Convert cond_resched_tasks_rcu_qs() from macro to static inline
   function.

SRCU updates:
 * Add Rust helpers for SRCU.
 * Avoid losing queued work at cleanup_srcu_struct() time.

Torture-test updates:
 * Preparation work for immediate RCU priority deboosting.
 * Test RCU readers from real interrupt handlers (as opposed to softirq).
 * Simplify code through use of cpumask_next_wrap().
 * Improve diagnostic output with character task states.
 * Add rcutorture.nwriters parameter to allow lightweight stall testing,
   and rcutorture.stall_only to make doing so easier.
 * Test an RCU Tasks Trace grace period implying an RCU grace period.
 * Make RCU Tasks Trace torturing track reader batches.
 * Fix a data race, courtesy of KCSAN.
 * Plug a shuffle_tmp_mask memory leak on kthread spawn failure.

1 files changed, 166 insertions, 0 deletions

diff --git a/arch/alpha/lib/memcpy.c b/arch/alpha/lib/memcpy.c
new file mode 100644
index 000000000..78b6850a9
--- /dev/null
+++ b/arch/alpha/lib/memcpy.c
@@ -0,0 +1,166 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  linux/arch/alpha/lib/memcpy.c
+ *
+ *  Copyright (C) 1995  Linus Torvalds
+ */
+
+/*
+ * This is a reasonably optimized memcpy() routine.
+ */
+
+/*
+ * Note that the C code is written to be optimized into good assembly. However,
+ * at this point gcc is unable to sanely compile "if (n >= 0)", resulting in a
+ * explicit compare against 0 (instead of just using the proper "blt reg, xx" or
+ * "bge reg, xx"). I hope alpha-gcc will be fixed to notice this eventually..
+ */
+
+#include <linux/types.h>
+#include <linux/export.h>
+#include <linux/string.h>
+
+/*
+ * This should be done in one go with ldq_u*2/mask/stq_u. Do it
+ * with a macro so that we can fix it up later..
+ */
+#define ALIGN_DEST_TO8_UP(d,s,n) \
+	while (d & 7) { \
+		if (n <= 0) return; \
+		n--; \
+		*(char *) d = *(char *) s; \
+		d++; s++; \
+	}
+#define ALIGN_DEST_TO8_DN(d,s,n) \
+	while (d & 7) { \
+		if (n <= 0) return; \
+		n--; \
+		d--; s--; \
+		*(char *) d = *(char *) s; \
+	}
+
+/*
+ * This should similarly be done with ldq_u*2/mask/stq. The destination
+ * is aligned, but we don't fill in a full quad-word
+ */
+#define DO_REST_UP(d,s,n) \
+	while (n > 0) { \
+		n--; \
+		*(char *) d = *(char *) s; \
+		d++; s++; \
+	}
+#define DO_REST_DN(d,s,n) \
+	while (n > 0) { \
+		n--; \
+		d--; s--; \
+		*(char *) d = *(char *) s; \
+	}
+
+/*
+ * This should be done with ldq/mask/stq. The source and destination are
+ * aligned, but we don't fill in a full quad-word
+ */
+#define DO_REST_ALIGNED_UP(d,s,n) DO_REST_UP(d,s,n)
+#define DO_REST_ALIGNED_DN(d,s,n) DO_REST_DN(d,s,n)
+
+/*
+ * This does unaligned memory copies. We want to avoid storing to
+ * an unaligned address, as that would do a read-modify-write cycle.
+ * We also want to avoid double-reading the unaligned reads.
+ *
+ * Note the ordering to try to avoid load (and address generation) latencies.
+ */
+static inline void __memcpy_unaligned_up (unsigned long d, unsigned long s,
+					  long n)
+{
+	ALIGN_DEST_TO8_UP(d,s,n);
+	n -= 8;			/* to avoid compare against 8 in the loop */
+	if (n >= 0) {
+		unsigned long low_word, high_word;
+		__asm__("ldq_u %0,%1":"=r" (low_word):"m" (*(unsigned long *) s));
+		do {
+			unsigned long tmp;
+			__asm__("ldq_u %0,%1":"=r" (high_word):"m" (*(unsigned long *)(s+8)));
+			n -= 8;
+			__asm__("extql %1,%2,%0"
+				:"=r" (low_word)
+				:"r" (low_word), "r" (s));
+			__asm__("extqh %1,%2,%0"
+				:"=r" (tmp)
+				:"r" (high_word), "r" (s));
+			s += 8;
+			*(unsigned long *) d = low_word | tmp;
+			d += 8;
+			low_word = high_word;
+		} while (n >= 0);
+	}
+	n += 8;
+	DO_REST_UP(d,s,n);
+}
+
+static inline void __memcpy_unaligned_dn (unsigned long d, unsigned long s,
+					  long n)
+{
+	/* I don't understand AXP assembler well enough for this. -Tim */
+	s += n;
+	d += n;
+	while (n--)
+		* (char *) --d = * (char *) --s;
+}
+
+/*
+ * Hmm.. Strange. The __asm__ here is there to make gcc use an integer register
+ * for the load-store. I don't know why, but it would seem that using a floating
+ * point register for the move seems to slow things down (very small difference,
+ * though).
+ *
+ * Note the ordering to try to avoid load (and address generation) latencies.
+ */
+static inline void __memcpy_aligned_up (unsigned long d, unsigned long s,
+					long n)
+{
+	ALIGN_DEST_TO8_UP(d,s,n);
+	n -= 8;
+	while (n >= 0) {
+		unsigned long tmp;
+		__asm__("ldq %0,%1":"=r" (tmp):"m" (*(unsigned long *) s));
+		n -= 8;
+		s += 8;
+		*(unsigned long *) d = tmp;
+		d += 8;
+	}
+	n += 8;
+	DO_REST_ALIGNED_UP(d,s,n);
+}
+static inline void __memcpy_aligned_dn (unsigned long d, unsigned long s,
+					long n)
+{
+	s += n;
+	d += n;
+	ALIGN_DEST_TO8_DN(d,s,n);
+	n -= 8;
+	while (n >= 0) {
+		unsigned long tmp;
+		s -= 8;
+		__asm__("ldq %0,%1":"=r" (tmp):"m" (*(unsigned long *) s));
+		n -= 8;
+		d -= 8;
+		*(unsigned long *) d = tmp;
+	}
+	n += 8;
+	DO_REST_ALIGNED_DN(d,s,n);
+}
+
+#undef memcpy
+
+void * memcpy(void * dest, const void *src, size_t n)
+{
+	if (!(((unsigned long) dest ^ (unsigned long) src) & 7)) {
+		__memcpy_aligned_up ((unsigned long) dest, (unsigned long) src,
+				     n);
+		return dest;
+	}
+	__memcpy_unaligned_up ((unsigned long) dest, (unsigned long) src, n);
+	return dest;
+}
+EXPORT_SYMBOL(memcpy);