]> git.sesse.net Git - bcachefs-tools-debian/blobdiff - libbcachefs/bcachefs.h
Update bcachefs sources to 5d0a6c2b32f1 bcachefs: check_directory_structure() can...
[bcachefs-tools-debian] / libbcachefs / bcachefs.h
index 295efeda12ff8eaaf43e6c88aa9aabea138fb548..bf8d69fbfdb39357fcd088299094a25faa40d88f 100644 (file)
 #include <linux/mutex.h>
 #include <linux/percpu-refcount.h>
 #include <linux/percpu-rwsem.h>
+#include <linux/refcount.h>
 #include <linux/rhashtable.h>
 #include <linux/rwsem.h>
 #include <linux/semaphore.h>
 
 #define race_fault(...)                        dynamic_fault("bcachefs:race")
 
+#define count_event(_c, _name) this_cpu_inc((_c)->counters[BCH_COUNTER_##_name])
+
 #define trace_and_count(_c, _name, ...)                                        \
 do {                                                                   \
-       this_cpu_inc((_c)->counters[BCH_COUNTER_##_name]);              \
+       count_event(_c, _name);                                         \
        trace_##_name(__VA_ARGS__);                                     \
 } while (0)
 
@@ -262,46 +265,76 @@ do {                                                                      \
 
 #define bch2_fmt(_c, fmt)              bch2_log_msg(_c, fmt "\n")
 
+__printf(2, 3)
+void __bch2_print(struct bch_fs *c, const char *fmt, ...);
+
+#define maybe_dev_to_fs(_c)    _Generic((_c),                          \
+       struct bch_dev *:       ((struct bch_dev *) (_c))->fs,          \
+       struct bch_fs *:        (_c))
+
+#define bch2_print(_c, ...) __bch2_print(maybe_dev_to_fs(_c), __VA_ARGS__)
+
+#define bch2_print_ratelimited(_c, ...)                                        \
+do {                                                                   \
+       static DEFINE_RATELIMIT_STATE(_rs,                              \
+                                     DEFAULT_RATELIMIT_INTERVAL,       \
+                                     DEFAULT_RATELIMIT_BURST);         \
+                                                                       \
+       if (__ratelimit(&_rs))                                          \
+               bch2_print(_c, __VA_ARGS__);                            \
+} while (0)
+
 #define bch_info(c, fmt, ...) \
-       printk(KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_INFO bch2_fmt(c, fmt), ##__VA_ARGS__)
 #define bch_notice(c, fmt, ...) \
-       printk(KERN_NOTICE bch2_fmt(c, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_NOTICE bch2_fmt(c, fmt), ##__VA_ARGS__)
 #define bch_warn(c, fmt, ...) \
-       printk(KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
 #define bch_warn_ratelimited(c, fmt, ...) \
-       printk_ratelimited(KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
+       bch2_print_ratelimited(c, KERN_WARNING bch2_fmt(c, fmt), ##__VA_ARGS__)
 
 #define bch_err(c, fmt, ...) \
-       printk(KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
 #define bch_err_dev(ca, fmt, ...) \
-       printk(KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
 #define bch_err_dev_offset(ca, _offset, fmt, ...) \
-       printk(KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
 #define bch_err_inum(c, _inum, fmt, ...) \
-       printk(KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
 #define bch_err_inum_offset(c, _inum, _offset, fmt, ...) \
-       printk(KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
+       bch2_print(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
 
 #define bch_err_ratelimited(c, fmt, ...) \
-       printk_ratelimited(KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
+       bch2_print_ratelimited(c, KERN_ERR bch2_fmt(c, fmt), ##__VA_ARGS__)
 #define bch_err_dev_ratelimited(ca, fmt, ...) \
-       printk_ratelimited(KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
+       bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev(ca, fmt), ##__VA_ARGS__)
 #define bch_err_dev_offset_ratelimited(ca, _offset, fmt, ...) \
-       printk_ratelimited(KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
+       bch2_print_ratelimited(ca, KERN_ERR bch2_fmt_dev_offset(ca, _offset, fmt), ##__VA_ARGS__)
 #define bch_err_inum_ratelimited(c, _inum, fmt, ...) \
-       printk_ratelimited(KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
+       bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum(c, _inum, fmt), ##__VA_ARGS__)
 #define bch_err_inum_offset_ratelimited(c, _inum, _offset, fmt, ...) \
-       printk_ratelimited(KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
+       bch2_print_ratelimited(c, KERN_ERR bch2_fmt_inum_offset(c, _inum, _offset, fmt), ##__VA_ARGS__)
+
+static inline bool should_print_err(int err)
+{
+       return err && !bch2_err_matches(err, BCH_ERR_transaction_restart);
+}
 
 #define bch_err_fn(_c, _ret)                                           \
 do {                                                                   \
-       if (_ret && !bch2_err_matches(_ret, BCH_ERR_transaction_restart))\
+       if (should_print_err(_ret))                                     \
                bch_err(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
 } while (0)
 
+#define bch_err_fn_ratelimited(_c, _ret)                               \
+do {                                                                   \
+       if (should_print_err(_ret))                                     \
+               bch_err_ratelimited(_c, "%s(): error %s", __func__, bch2_err_str(_ret));\
+} while (0)
+
 #define bch_err_msg(_c, _ret, _msg, ...)                               \
 do {                                                                   \
-       if (_ret && !bch2_err_matches(_ret, BCH_ERR_transaction_restart))\
+       if (should_print_err(_ret))                                     \
                bch_err(_c, "%s(): error " _msg " %s", __func__,        \
                        ##__VA_ARGS__, bch2_err_str(_ret));             \
 } while (0)
@@ -445,6 +478,12 @@ enum bch_time_stats {
 
 struct btree;
 
+struct log_output {
+       spinlock_t              lock;
+       wait_queue_head_t       wait;
+       struct printbuf         buf;
+};
+
 enum gc_phase {
        GC_PHASE_NOT_RUNNING,
        GC_PHASE_START,
@@ -567,32 +606,38 @@ struct bch_dev {
        struct io_count __percpu *io_done;
 };
 
-enum {
-       /* startup: */
-       BCH_FS_STARTED,
-       BCH_FS_MAY_GO_RW,
-       BCH_FS_RW,
-       BCH_FS_WAS_RW,
-
-       /* shutdown: */
-       BCH_FS_STOPPING,
-       BCH_FS_EMERGENCY_RO,
-       BCH_FS_GOING_RO,
-       BCH_FS_WRITE_DISABLE_COMPLETE,
-       BCH_FS_CLEAN_SHUTDOWN,
-
-       /* fsck passes: */
-       BCH_FS_FSCK_DONE,
-       BCH_FS_INITIAL_GC_UNFIXED,      /* kill when we enumerate fsck errors */
-       BCH_FS_NEED_ANOTHER_GC,
-
-       BCH_FS_NEED_DELETE_DEAD_SNAPSHOTS,
-
-       /* errors: */
-       BCH_FS_ERROR,
-       BCH_FS_TOPOLOGY_ERROR,
-       BCH_FS_ERRORS_FIXED,
-       BCH_FS_ERRORS_NOT_FIXED,
+/*
+ * fsck_done - kill?
+ *
+ * replace with something more general from enumated fsck passes/errors:
+ * initial_gc_unfixed
+ * error
+ * topology error
+ */
+
+#define BCH_FS_FLAGS()                 \
+       x(started)                      \
+       x(may_go_rw)                    \
+       x(rw)                           \
+       x(was_rw)                       \
+       x(stopping)                     \
+       x(emergency_ro)                 \
+       x(going_ro)                     \
+       x(write_disable_complete)       \
+       x(clean_shutdown)               \
+       x(fsck_done)                    \
+       x(initial_gc_unfixed)           \
+       x(need_another_gc)              \
+       x(need_delete_dead_snapshots)   \
+       x(error)                        \
+       x(topology_error)               \
+       x(errors_fixed)                 \
+       x(errors_not_fixed)
+
+enum bch_fs_flags {
+#define x(n)           BCH_FS_##n,
+       BCH_FS_FLAGS()
+#undef x
 };
 
 struct btree_debug {
@@ -605,7 +650,7 @@ struct btree_transaction_stats {
        struct bch2_time_stats  lock_hold_times;
        struct mutex            lock;
        unsigned                nr_max_paths;
-       unsigned                wb_updates_size;
+       unsigned                journal_entries_size;
        unsigned                max_mem;
        char                    *max_paths_text;
 };
@@ -693,6 +738,8 @@ struct bch_fs {
        struct super_block      *vfs_sb;
        dev_t                   dev;
        char                    name[40];
+       struct log_output       *output;
+       struct task_struct      *output_filter;
 
        /* ro/rw, add/remove/resize devices: */
        struct rw_semaphore     state_lock;
@@ -703,6 +750,13 @@ struct bch_fs {
 #else
        struct percpu_ref       writes;
 #endif
+       /*
+        * Analagous to c->writes, for asynchronous ops that don't necessarily
+        * need fs to be read-write
+        */
+       refcount_t              ro_ref;
+       wait_queue_head_t       ro_ref_wait;
+
        struct work_struct      read_only_work;
 
        struct bch_dev __rcu    *devs[BCH_SB_MEMBERS_MAX];
@@ -1005,10 +1059,21 @@ struct bch_fs {
        /* RECOVERY */
        u64                     journal_replay_seq_start;
        u64                     journal_replay_seq_end;
+       /*
+        * Two different uses:
+        * "Has this fsck pass?" - i.e. should this type of error be an
+        * emergency read-only
+        * And, in certain situations fsck will rewind to an earlier pass: used
+        * for signaling to the toplevel code which pass we want to run now.
+        */
        enum bch_recovery_pass  curr_recovery_pass;
        /* bitmap of explicitly enabled recovery passes: */
        u64                     recovery_passes_explicit;
+       /* bitmask of recovery passes that we actually ran */
        u64                     recovery_passes_complete;
+       /* never rewinds version of curr_recovery_pass */
+       enum bch_recovery_pass  recovery_pass_done;
+       struct semaphore        online_fsck_mutex;
 
        /* DEBUG JUNK */
        struct dentry           *fs_debug_dir;
@@ -1071,7 +1136,7 @@ static inline void bch2_write_ref_get(struct bch_fs *c, enum bch_write_ref ref)
 static inline bool __bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
 {
 #ifdef BCH_WRITE_REF_DEBUG
-       return !test_bit(BCH_FS_GOING_RO, &c->flags) &&
+       return !test_bit(BCH_FS_going_ro, &c->flags) &&
                atomic_long_inc_not_zero(&c->writes[ref]);
 #else
        return percpu_ref_tryget(&c->writes);
@@ -1081,7 +1146,7 @@ static inline bool __bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref
 static inline bool bch2_write_ref_tryget(struct bch_fs *c, enum bch_write_ref ref)
 {
 #ifdef BCH_WRITE_REF_DEBUG
-       return !test_bit(BCH_FS_GOING_RO, &c->flags) &&
+       return !test_bit(BCH_FS_going_ro, &c->flags) &&
                atomic_long_inc_not_zero(&c->writes[ref]);
 #else
        return percpu_ref_tryget_live(&c->writes);
@@ -1100,13 +1165,27 @@ static inline void bch2_write_ref_put(struct bch_fs *c, enum bch_write_ref ref)
                if (atomic_long_read(&c->writes[i]))
                        return;
 
-       set_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags);
+       set_bit(BCH_FS_write_disable_complete, &c->flags);
        wake_up(&bch2_read_only_wait);
 #else
        percpu_ref_put(&c->writes);
 #endif
 }
 
+static inline bool bch2_ro_ref_tryget(struct bch_fs *c)
+{
+       if (test_bit(BCH_FS_stopping, &c->flags))
+               return false;
+
+       return refcount_inc_not_zero(&c->ro_ref);
+}
+
+static inline void bch2_ro_ref_put(struct bch_fs *c)
+{
+       if (refcount_dec_and_test(&c->ro_ref))
+               wake_up(&c->ro_ref_wait);
+}
+
 static inline void bch2_set_ra_pages(struct bch_fs *c, unsigned ra_pages)
 {
 #ifndef NO_BCACHEFS_FS