Documentation: locking: Describe seqlock design and usage

Proper documentation for the design and usage of sequence counters and
sequential locks does not exist. Complete the seqlock.h documentation as
follows:

  - Divide all documentation on a seqcount_t vs. seqlock_t basis. The
    description for both mechanisms was intermingled, which is incorrect
    since the usage constrains for each type are vastly different.

  - Add an introductory paragraph describing the internal design of, and
    rationale for, sequence counters.

  - Document seqcount_t writer non-preemptibility requirement, which was
    not previously documented anywhere, and provide a clear rationale.

  - Provide template code for seqcount_t and seqlock_t initialization
    and reader/writer critical sections.

  - Recommend using seqlock_t by default. It implicitly handles the
    serialization and non-preemptibility requirements of writers.

At seqlock.h:

  - Remove references to brlocks as they've long been removed from the
    kernel.

  - Remove references to gcc-3.x since the kernel's minimum supported
    gcc version is 4.9.

References: 0f6ed63b17 ("no need to keep brlock macros anymore...")
References: 6ec4476ac8 ("Raise gcc version requirement to 4.9")
Signed-off-by: Ahmed S. Darwish <a.darwish@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20200720155530.1173732-2-a.darwish@linutronix.de

Documentation/locking/index.rst

@@ -14,6 +14,7 @@ locking
     mutex-design
     rt-mutex-design
     rt-mutex
+    seqlock
     spinlocks
     ww-mutex-design
     preempt-locking

Documentation/locking/seqlock.rst

@@ -0,0 +1,170 @@
+======================================
+Sequence counters and sequential locks
+======================================
+
+Introduction
+============
+
+Sequence counters are a reader-writer consistency mechanism with
+lockless readers (read-only retry loops), and no writer starvation. They
+are used for data that's rarely written to (e.g. system time), where the
+reader wants a consistent set of information and is willing to retry if
+that information changes.
+
+A data set is consistent when the sequence count at the beginning of the
+read side critical section is even and the same sequence count value is
+read again at the end of the critical section. The data in the set must
+be copied out inside the read side critical section. If the sequence
+count has changed between the start and the end of the critical section,
+the reader must retry.
+
+Writers increment the sequence count at the start and the end of their
+critical section. After starting the critical section the sequence count
+is odd and indicates to the readers that an update is in progress. At
+the end of the write side critical section the sequence count becomes
+even again which lets readers make progress.
+
+A sequence counter write side critical section must never be preempted
+or interrupted by read side sections. Otherwise the reader will spin for
+the entire scheduler tick due to the odd sequence count value and the
+interrupted writer. If that reader belongs to a real-time scheduling
+class, it can spin forever and the kernel will livelock.
+
+This mechanism cannot be used if the protected data contains pointers,
+as the writer can invalidate a pointer that the reader is following.
+
+
+.. _seqcount_t:
+
+Sequence counters (``seqcount_t``)
+==================================
+
+This is the the raw counting mechanism, which does not protect against
+multiple writers.  Write side critical sections must thus be serialized
+by an external lock.
+
+If the write serialization primitive is not implicitly disabling
+preemption, preemption must be explicitly disabled before entering the
+write side section. If the read section can be invoked from hardirq or
+softirq contexts, interrupts or bottom halves must also be respectively
+disabled before entering the write section.
+
+If it's desired to automatically handle the sequence counter
+requirements of writer serialization and non-preemptibility, use
+:ref:`seqlock_t` instead.
+
+Initialization::
+
+	/* dynamic */
+	seqcount_t foo_seqcount;
+	seqcount_init(&foo_seqcount);
+
+	/* static */
+	static seqcount_t foo_seqcount = SEQCNT_ZERO(foo_seqcount);
+
+	/* C99 struct init */
+	struct {
+		.seq   = SEQCNT_ZERO(foo.seq),
+	} foo;
+
+Write path::
+
+	/* Serialized context with disabled preemption */
+
+	write_seqcount_begin(&foo_seqcount);
+
+	/* ... [[write-side critical section]] ... */
+
+	write_seqcount_end(&foo_seqcount);
+
+Read path::
+
+	do {
+		seq = read_seqcount_begin(&foo_seqcount);
+
+		/* ... [[read-side critical section]] ... */
+
+	} while (read_seqcount_retry(&foo_seqcount, seq));
+
+
+.. _seqlock_t:
+
+Sequential locks (``seqlock_t``)
+================================
+
+This contains the :ref:`seqcount_t` mechanism earlier discussed, plus an
+embedded spinlock for writer serialization and non-preemptibility.
+
+If the read side section can be invoked from hardirq or softirq context,
+use the write side function variants which disable interrupts or bottom
+halves respectively.
+
+Initialization::
+
+	/* dynamic */
+	seqlock_t foo_seqlock;
+	seqlock_init(&foo_seqlock);
+
+	/* static */
+	static DEFINE_SEQLOCK(foo_seqlock);
+
+	/* C99 struct init */
+	struct {
+		.seql   = __SEQLOCK_UNLOCKED(foo.seql)
+	} foo;
+
+Write path::
+
+	write_seqlock(&foo_seqlock);
+
+	/* ... [[write-side critical section]] ... */
+
+	write_sequnlock(&foo_seqlock);
+
+Read path, three categories:
+
+1. Normal Sequence readers which never block a writer but they must
+   retry if a writer is in progress by detecting change in the sequence
+   number.  Writers do not wait for a sequence reader::
+
+	do {
+		seq = read_seqbegin(&foo_seqlock);
+
+		/* ... [[read-side critical section]] ... */
+
+	} while (read_seqretry(&foo_seqlock, seq));
+
+2. Locking readers which will wait if a writer or another locking reader
+   is in progress. A locking reader in progress will also block a writer
+   from entering its critical section. This read lock is
+   exclusive. Unlike rwlock_t, only one locking reader can acquire it::
+
+	read_seqlock_excl(&foo_seqlock);
+
+	/* ... [[read-side critical section]] ... */
+
+	read_sequnlock_excl(&foo_seqlock);
+
+3. Conditional lockless reader (as in 1), or locking reader (as in 2),
+   according to a passed marker. This is used to avoid lockless readers
+   starvation (too much retry loops) in case of a sharp spike in write
+   activity. First, a lockless read is tried (even marker passed). If
+   that trial fails (odd sequence counter is returned, which is used as
+   the next iteration marker), the lockless read is transformed to a
+   full locking read and no retry loop is necessary::
+
+	/* marker; even initialization */
+	int seq = 0;
+	do {
+		read_seqbegin_or_lock(&foo_seqlock, &seq);
+
+		/* ... [[read-side critical section]] ... */
+
+	} while (need_seqretry(&foo_seqlock, seq));
+	done_seqretry(&foo_seqlock, seq);
+
+
+API documentation
+=================
+
+.. kernel-doc:: include/linux/seqlock.h

include/linux/seqlock.h

@@ -1,36 +1,15 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #ifndef __LINUX_SEQLOCK_H
 #define __LINUX_SEQLOCK_H
+
 /*
- * Reader/writer consistent mechanism without starving writers. This type of
+ * seqcount_t / seqlock_t - a reader-writer consistency mechanism with
- * lock for data where the reader wants a consistent set of information
+ * lockless readers (read-only retry loops), and no writer starvation.
- * and is willing to retry if the information changes. There are two types
- * of readers:
- * 1. Sequence readers which never block a writer but they may have to retry
- *    if a writer is in progress by detecting change in sequence number.
- *    Writers do not wait for a sequence reader.
- * 2. Locking readers which will wait if a writer or another locking reader
- *    is in progress. A locking reader in progress will also block a writer
- *    from going forward. Unlike the regular rwlock, the read lock here is
- *    exclusive so that only one locking reader can get it.
  *
- * This is not as cache friendly as brlock. Also, this may not work well
+ * See Documentation/locking/seqlock.rst
- * for data that contains pointers, because any writer could
- * invalidate a pointer that a reader was following.
  *
- * Expected non-blocking reader usage:
+ * Copyrights:
- * 	do {
+ * - Based on x86_64 vsyscall gettimeofday: Keith Owens, Andrea Arcangeli
- *	    seq = read_seqbegin(&foo);
- * 	...
- *      } while (read_seqretry(&foo, seq));
- *
- *
- * On non-SMP the spin locks disappear but the writer still needs
- * to increment the sequence variables because an interrupt routine could
- * change the state of the data.
- *
- * Based on x86_64 vsyscall gettimeofday 
- * by Keith Owens and Andrea Arcangeli
  */
 
 #include <linux/spinlock.h>
@@ -41,8 +20,8 @@
 #include <asm/processor.h>
 
 /*
- * The seqlock interface does not prescribe a precise sequence of read
+ * The seqlock seqcount_t interface does not prescribe a precise sequence of
- * begin/retry/end. For readers, typically there is a call to
+ * read begin/retry/end. For readers, typically there is a call to
  * read_seqcount_begin() and read_seqcount_retry(), however, there are more
  * esoteric cases which do not follow this pattern.
  *
@@ -50,16 +29,30 @@
  * via seqcount_t under KCSAN: upon beginning a seq-reader critical section,
  * pessimistically mark the next KCSAN_SEQLOCK_REGION_MAX memory accesses as
  * atomics; if there is a matching read_seqcount_retry() call, no following
- * memory operations are considered atomic. Usage of seqlocks via seqlock_t
+ * memory operations are considered atomic. Usage of the seqlock_t interface
- * interface is not affected.
+ * is not affected.
  */
 #define KCSAN_SEQLOCK_REGION_MAX 1000
 
 /*
- * Version using sequence counter only.
+ * Sequence counters (seqcount_t)
- * This can be used when code has its own mutex protecting the
+ *
- * updating starting before the write_seqcountbeqin() and ending
+ * This is the raw counting mechanism, without any writer protection.
- * after the write_seqcount_end().
+ *
+ * Write side critical sections must be serialized and non-preemptible.
+ *
+ * If readers can be invoked from hardirq or softirq contexts,
+ * interrupts or bottom halves must also be respectively disabled before
+ * entering the write section.
+ *
+ * This mechanism can't be used if the protected data contains pointers,
+ * as the writer can invalidate a pointer that a reader is following.
+ *
+ * If it's desired to automatically handle the sequence counter writer
+ * serialization and non-preemptibility requirements, use a sequential
+ * lock (seqlock_t) instead.
+ *
+ * See Documentation/locking/seqlock.rst
  */
 typedef struct seqcount {
 	unsigned sequence;
@@ -398,10 +391,6 @@ static inline void raw_write_seqcount_latch(seqcount_t *s)
        smp_wmb();      /* increment "sequence" before following stores */
 }
 
-/*
- * Sequence counter only version assumes that callers are using their
- * own mutexing.
- */
 static inline void write_seqcount_begin_nested(seqcount_t *s, int subclass)
 {
 	raw_write_seqcount_begin(s);
@@ -434,15 +423,21 @@ static inline void write_seqcount_invalidate(seqcount_t *s)
 	kcsan_nestable_atomic_end();
 }
 
+/*
+ * Sequential locks (seqlock_t)
+ *
+ * Sequence counters with an embedded spinlock for writer serialization
+ * and non-preemptibility.
+ *
+ * For more info, see:
+ *    - Comments on top of seqcount_t
+ *    - Documentation/locking/seqlock.rst
+ */
 typedef struct {
 	struct seqcount seqcount;
 	spinlock_t lock;
 } seqlock_t;
 
-/*
- * These macros triggered gcc-3.x compile-time problems.  We think these are
- * OK now.  Be cautious.
- */
 #define __SEQLOCK_UNLOCKED(lockname)			\
 	{						\
 		.seqcount = SEQCNT_ZERO(lockname),	\