kernel_optimize_test/kernel/locking/lockdep_internals.h
Waiman Long ecd384c436 locking/lockdep: Iterate lock_classes directly when reading lockdep files
[ Upstream commit fb7275acd6fb988313dddd8d3d19efa70d9015ad ]

When dumping lock_classes information via /proc/lockdep, we can't take
the lockdep lock as the lock hold time is indeterminate. Iterating
over all_lock_classes without holding lock can be dangerous as there
is a slight chance that it may branch off to other lists leading to
infinite loop or even access invalid memory if changes are made to
all_lock_classes list in parallel.

To avoid this problem, iteration of lock classes is now done directly
on the lock_classes array itself. The lock_classes_in_use bitmap is
checked to see if the lock class is being used. To avoid iterating
the full array all the times, a new max_lock_class_idx value is added
to track the maximum lock_class index that is currently being used.

We can theoretically take the lockdep lock for iterating all_lock_classes
when other lockdep files (lockdep_stats and lock_stat) are accessed as
the lock hold time will be shorter for them. For consistency, they are
also modified to iterate the lock_classes array directly.

Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20220211035526.1329503-2-longman@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2022-04-08 14:40:32 +02:00

262 lines
7.1 KiB
C

/* SPDX-License-Identifier: GPL-2.0 */
/*
* kernel/lockdep_internals.h
*
* Runtime locking correctness validator
*
* lockdep subsystem internal functions and variables.
*/
/*
* Lock-class usage-state bits:
*/
enum lock_usage_bit {
#define LOCKDEP_STATE(__STATE) \
LOCK_USED_IN_##__STATE, \
LOCK_USED_IN_##__STATE##_READ, \
LOCK_ENABLED_##__STATE, \
LOCK_ENABLED_##__STATE##_READ,
#include "lockdep_states.h"
#undef LOCKDEP_STATE
LOCK_USED,
LOCK_USED_READ,
LOCK_USAGE_STATES,
};
/* states after LOCK_USED_READ are not traced and printed */
static_assert(LOCK_TRACE_STATES == LOCK_USAGE_STATES);
#define LOCK_USAGE_READ_MASK 1
#define LOCK_USAGE_DIR_MASK 2
#define LOCK_USAGE_STATE_MASK (~(LOCK_USAGE_READ_MASK | LOCK_USAGE_DIR_MASK))
/*
* Usage-state bitmasks:
*/
#define __LOCKF(__STATE) LOCKF_##__STATE = (1 << LOCK_##__STATE),
enum {
#define LOCKDEP_STATE(__STATE) \
__LOCKF(USED_IN_##__STATE) \
__LOCKF(USED_IN_##__STATE##_READ) \
__LOCKF(ENABLED_##__STATE) \
__LOCKF(ENABLED_##__STATE##_READ)
#include "lockdep_states.h"
#undef LOCKDEP_STATE
__LOCKF(USED)
__LOCKF(USED_READ)
};
#define LOCKDEP_STATE(__STATE) LOCKF_ENABLED_##__STATE |
static const unsigned long LOCKF_ENABLED_IRQ =
#include "lockdep_states.h"
0;
#undef LOCKDEP_STATE
#define LOCKDEP_STATE(__STATE) LOCKF_USED_IN_##__STATE |
static const unsigned long LOCKF_USED_IN_IRQ =
#include "lockdep_states.h"
0;
#undef LOCKDEP_STATE
#define LOCKDEP_STATE(__STATE) LOCKF_ENABLED_##__STATE##_READ |
static const unsigned long LOCKF_ENABLED_IRQ_READ =
#include "lockdep_states.h"
0;
#undef LOCKDEP_STATE
#define LOCKDEP_STATE(__STATE) LOCKF_USED_IN_##__STATE##_READ |
static const unsigned long LOCKF_USED_IN_IRQ_READ =
#include "lockdep_states.h"
0;
#undef LOCKDEP_STATE
#define LOCKF_ENABLED_IRQ_ALL (LOCKF_ENABLED_IRQ | LOCKF_ENABLED_IRQ_READ)
#define LOCKF_USED_IN_IRQ_ALL (LOCKF_USED_IN_IRQ | LOCKF_USED_IN_IRQ_READ)
#define LOCKF_IRQ (LOCKF_ENABLED_IRQ | LOCKF_USED_IN_IRQ)
#define LOCKF_IRQ_READ (LOCKF_ENABLED_IRQ_READ | LOCKF_USED_IN_IRQ_READ)
/*
* CONFIG_LOCKDEP_SMALL is defined for sparc. Sparc requires .text,
* .data and .bss to fit in required 32MB limit for the kernel. With
* CONFIG_LOCKDEP we could go over this limit and cause system boot-up problems.
* So, reduce the static allocations for lockdeps related structures so that
* everything fits in current required size limit.
*/
#ifdef CONFIG_LOCKDEP_SMALL
/*
* MAX_LOCKDEP_ENTRIES is the maximum number of lock dependencies
* we track.
*
* We use the per-lock dependency maps in two ways: we grow it by adding
* every to-be-taken lock to all currently held lock's own dependency
* table (if it's not there yet), and we check it for lock order
* conflicts and deadlocks.
*/
#define MAX_LOCKDEP_ENTRIES 16384UL
#define MAX_LOCKDEP_CHAINS_BITS 15
#define MAX_STACK_TRACE_ENTRIES 262144UL
#define STACK_TRACE_HASH_SIZE 8192
#else
#define MAX_LOCKDEP_ENTRIES 32768UL
#define MAX_LOCKDEP_CHAINS_BITS 16
/*
* Stack-trace: tightly packed array of stack backtrace
* addresses. Protected by the hash_lock.
*/
#define MAX_STACK_TRACE_ENTRIES 524288UL
#define STACK_TRACE_HASH_SIZE 16384
#endif
/*
* Bit definitions for lock_chain.irq_context
*/
#define LOCK_CHAIN_SOFTIRQ_CONTEXT (1 << 0)
#define LOCK_CHAIN_HARDIRQ_CONTEXT (1 << 1)
#define MAX_LOCKDEP_CHAINS (1UL << MAX_LOCKDEP_CHAINS_BITS)
#define MAX_LOCKDEP_CHAIN_HLOCKS (MAX_LOCKDEP_CHAINS*5)
extern struct lock_chain lock_chains[];
#define LOCK_USAGE_CHARS (2*XXX_LOCK_USAGE_STATES + 1)
extern void get_usage_chars(struct lock_class *class,
char usage[LOCK_USAGE_CHARS]);
extern const char *__get_key_name(const struct lockdep_subclass_key *key,
char *str);
struct lock_class *lock_chain_get_class(struct lock_chain *chain, int i);
extern unsigned long nr_lock_classes;
extern unsigned long nr_zapped_classes;
extern unsigned long nr_zapped_lock_chains;
extern unsigned long nr_list_entries;
long lockdep_next_lockchain(long i);
unsigned long lock_chain_count(void);
extern unsigned long nr_stack_trace_entries;
extern unsigned int nr_hardirq_chains;
extern unsigned int nr_softirq_chains;
extern unsigned int nr_process_chains;
extern unsigned int nr_free_chain_hlocks;
extern unsigned int nr_lost_chain_hlocks;
extern unsigned int nr_large_chain_blocks;
extern unsigned int max_lockdep_depth;
extern unsigned int max_bfs_queue_depth;
extern unsigned long max_lock_class_idx;
extern struct lock_class lock_classes[MAX_LOCKDEP_KEYS];
extern unsigned long lock_classes_in_use[];
#ifdef CONFIG_PROVE_LOCKING
extern unsigned long lockdep_count_forward_deps(struct lock_class *);
extern unsigned long lockdep_count_backward_deps(struct lock_class *);
#ifdef CONFIG_TRACE_IRQFLAGS
u64 lockdep_stack_trace_count(void);
u64 lockdep_stack_hash_count(void);
#endif
#else
static inline unsigned long
lockdep_count_forward_deps(struct lock_class *class)
{
return 0;
}
static inline unsigned long
lockdep_count_backward_deps(struct lock_class *class)
{
return 0;
}
#endif
#ifdef CONFIG_DEBUG_LOCKDEP
#include <asm/local.h>
/*
* Various lockdep statistics.
* We want them per cpu as they are often accessed in fast path
* and we want to avoid too much cache bouncing.
*/
struct lockdep_stats {
unsigned long chain_lookup_hits;
unsigned int chain_lookup_misses;
unsigned long hardirqs_on_events;
unsigned long hardirqs_off_events;
unsigned long redundant_hardirqs_on;
unsigned long redundant_hardirqs_off;
unsigned long softirqs_on_events;
unsigned long softirqs_off_events;
unsigned long redundant_softirqs_on;
unsigned long redundant_softirqs_off;
int nr_unused_locks;
unsigned int nr_redundant_checks;
unsigned int nr_redundant;
unsigned int nr_cyclic_checks;
unsigned int nr_find_usage_forwards_checks;
unsigned int nr_find_usage_backwards_checks;
/*
* Per lock class locking operation stat counts
*/
unsigned long lock_class_ops[MAX_LOCKDEP_KEYS];
};
DECLARE_PER_CPU(struct lockdep_stats, lockdep_stats);
#define __debug_atomic_inc(ptr) \
this_cpu_inc(lockdep_stats.ptr);
#define debug_atomic_inc(ptr) { \
WARN_ON_ONCE(!irqs_disabled()); \
__this_cpu_inc(lockdep_stats.ptr); \
}
#define debug_atomic_dec(ptr) { \
WARN_ON_ONCE(!irqs_disabled()); \
__this_cpu_dec(lockdep_stats.ptr); \
}
#define debug_atomic_read(ptr) ({ \
struct lockdep_stats *__cpu_lockdep_stats; \
unsigned long long __total = 0; \
int __cpu; \
for_each_possible_cpu(__cpu) { \
__cpu_lockdep_stats = &per_cpu(lockdep_stats, __cpu); \
__total += __cpu_lockdep_stats->ptr; \
} \
__total; \
})
static inline void debug_class_ops_inc(struct lock_class *class)
{
int idx;
idx = class - lock_classes;
__debug_atomic_inc(lock_class_ops[idx]);
}
static inline unsigned long debug_class_ops_read(struct lock_class *class)
{
int idx, cpu;
unsigned long ops = 0;
idx = class - lock_classes;
for_each_possible_cpu(cpu)
ops += per_cpu(lockdep_stats.lock_class_ops[idx], cpu);
return ops;
}
#else
# define __debug_atomic_inc(ptr) do { } while (0)
# define debug_atomic_inc(ptr) do { } while (0)
# define debug_atomic_dec(ptr) do { } while (0)
# define debug_atomic_read(ptr) 0
# define debug_class_ops_inc(ptr) do { } while (0)
#endif