1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
2 // vim: ts=8 sw=2 smarttab
4 * Ceph - scalable distributed file system
6 * Copyright (C) 2004-2006 Sage Weil <sage@newdream.net>
8 * This is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License version 2.1, as published by the Free Software
11 * Foundation. See file COPYING.
17 * This class stripes a serial log over objects on the store. Four
20 * write_pos - where we're writing new entries
21 * unused_field - where we're reading old entires
22 * expire_pos - what is deemed "old" by user
23 * trimmed_pos - where we're expiring old items
25 * trimmed_pos <= expire_pos <= unused_field <= write_pos.
27 * Often, unused_field <= write_pos (as with MDS log). During
28 * recovery, write_pos is undefined until the end of the log is
31 * A "head" struct at the beginning of the log is used to store
32 * metadata at regular intervals. The basic invariants include:
34 * head.unused_field <= unused_field -- the head may "lag", since
35 * it's updated lazily.
36 * head.write_pos <= write_pos
37 * head.expire_pos <= expire_pos
38 * head.trimmed_pos <= trimmed_pos
42 * head.expire_pos >= trimmed_pos -- this ensures we can find the
43 * "beginning" of the log as last
44 * recorded, before it is trimmed.
45 * trimming will block until a
46 * sufficiently current expire_pos
49 * To recover log state, we simply start at the last write_pos in the
50 * head, and probe the object sequence sizes until we read the end.
52 * Head struct is stored in the first object. Actual journal starts
53 * after layout.period() bytes.
57 #ifndef CEPH_JOURNALER_H
58 #define CEPH_JOURNALER_H
66 #include "common/Timer.h"
67 #include "common/Throttle.h"
75 typedef __u8 stream_format_t;
77 // Legacy envelope is leading uint32_t size
79 JOURNAL_FORMAT_LEGACY = 0,
80 JOURNAL_FORMAT_RESILIENT = 1,
81 // Insert new formats here, before COUNT
85 // Highest journal format version that we support
86 #define JOURNAL_FORMAT_MAX (JOURNAL_FORMAT_COUNT - 1)
88 // Legacy envelope is leading uint32_t size
89 #define JOURNAL_ENVELOPE_LEGACY (sizeof(uint32_t))
91 // Resilient envelope is leading uint64_t sentinel, uint32_t size,
92 // trailing uint64_t start_ptr
93 #define JOURNAL_ENVELOPE_RESILIENT (sizeof(uint32_t) + sizeof(uint64_t) + \
97 * Represents a collection of entries serialized in a byte stream.
99 * Each entry consists of:
100 * - a blob (used by the next level up as a serialized LogEvent)
101 * - a uint64_t (used by the next level up as a pointer to the start
102 * of the entry in the collection bytestream)
106 stream_format_t format;
109 JournalStream(stream_format_t format_) : format(format_) {}
111 void set_format(stream_format_t format_) {format = format_;}
113 bool readable(bufferlist &bl, uint64_t *need) const;
114 size_t read(bufferlist &from, bufferlist *to, uint64_t *start_ptr);
115 size_t write(bufferlist &entry, bufferlist *to, uint64_t const &start_ptr);
116 size_t get_envelope_size() const {
117 if (format >= JOURNAL_FORMAT_RESILIENT) {
118 return JOURNAL_ENVELOPE_RESILIENT;
120 return JOURNAL_ENVELOPE_LEGACY;
124 // A magic number for the start of journal entries, so that we can
125 // identify them in damaged journals.
126 static const uint64_t sentinel = 0x3141592653589793;
132 // this goes at the head of the log "file".
135 uint64_t trimmed_pos;
137 uint64_t unused_field;
140 file_layout_t layout; //< The mapping from byte stream offsets
142 stream_format_t stream_format; //< The encoding of LogEvents
143 // within the journal byte stream
145 Header(const char *m="") :
146 trimmed_pos(0), expire_pos(0), unused_field(0), write_pos(0), magic(m),
150 void encode(bufferlist &bl) const {
151 ENCODE_START(2, 2, bl);
153 ::encode(trimmed_pos, bl);
154 ::encode(expire_pos, bl);
155 ::encode(unused_field, bl);
156 ::encode(write_pos, bl);
157 ::encode(layout, bl, 0); // encode in legacy format
158 ::encode(stream_format, bl);
161 void decode(bufferlist::iterator &bl) {
162 DECODE_START_LEGACY_COMPAT_LEN(2, 2, 2, bl);
164 ::decode(trimmed_pos, bl);
165 ::decode(expire_pos, bl);
166 ::decode(unused_field, bl);
167 ::decode(write_pos, bl);
168 ::decode(layout, bl);
170 ::decode(stream_format, bl);
172 stream_format = JOURNAL_FORMAT_LEGACY;
177 void dump(Formatter *f) const {
178 f->open_object_section("journal_header");
180 f->dump_string("magic", magic);
181 f->dump_unsigned("write_pos", write_pos);
182 f->dump_unsigned("expire_pos", expire_pos);
183 f->dump_unsigned("trimmed_pos", trimmed_pos);
184 f->dump_unsigned("stream_format", stream_format);
185 f->dump_object("layout", layout);
187 f->close_section(); // journal_header
190 static void generate_test_instances(list<Header*> &ls)
192 ls.push_back(new Header());
194 ls.push_back(new Header());
195 ls.back()->trimmed_pos = 1;
196 ls.back()->expire_pos = 2;
197 ls.back()->unused_field = 3;
198 ls.back()->write_pos = 4;
199 ls.back()->magic = "magique";
201 ls.push_back(new Header());
202 ls.back()->stream_format = JOURNAL_FORMAT_RESILIENT;
205 WRITE_CLASS_ENCODER(Header)
207 uint32_t get_stream_format() const {
208 return stream_format;
211 Header last_committed;
217 const std::string name;
218 typedef std::lock_guard<std::mutex> lock_guard;
219 typedef std::unique_lock<std::mutex> unique_lock;
225 file_layout_t layout;
226 uint32_t stream_format;
227 JournalStream journal_stream;
233 PerfCounters *logger;
237 C_DelayFlush *delay_flush_event;
239 * Do a flush as a result of a C_DelayFlush context.
241 void _do_delayed_flush()
243 assert(delay_flush_event != NULL);
245 delay_flush_event = NULL;
250 static const int STATE_UNDEF = 0;
251 static const int STATE_READHEAD = 1;
252 static const int STATE_PROBING = 2;
253 static const int STATE_ACTIVE = 3;
254 static const int STATE_REREADHEAD = 4;
255 static const int STATE_REPROBING = 5;
260 void _write_head(Context *oncommit=NULL);
261 void _wait_for_flush(Context *onsafe);
265 ceph::real_time last_wrote_head;
266 void _finish_write_head(int r, Header &wrote, C_OnFinisher *oncommit);
268 friend class C_WriteHead;
270 void _reread_head(Context *onfinish);
271 void _set_layout(file_layout_t const *l);
272 list<Context*> waitfor_recover;
273 void _read_head(Context *on_finish, bufferlist *bl);
274 void _finish_read_head(int r, bufferlist& bl);
275 void _finish_reread_head(int r, bufferlist& bl, Context *finish);
276 void _probe(Context *finish, uint64_t *end);
277 void _finish_probe_end(int r, uint64_t end);
278 void _reprobe(C_OnFinisher *onfinish);
279 void _finish_reprobe(int r, uint64_t end, C_OnFinisher *onfinish);
280 void _finish_reread_head_and_probe(int r, C_OnFinisher *onfinish);
282 friend class C_ReadHead;
284 friend class C_ProbeEnd;
286 friend class C_RereadHead;
288 friend class C_ReProbe;
289 class C_RereadHeadProbe;
290 friend class C_RereadHeadProbe;
293 uint64_t prezeroing_pos;
294 uint64_t prezero_pos; ///< we zero journal space ahead of write_pos to
295 // avoid problems with tail probing
296 uint64_t write_pos; ///< logical write position, where next entry
298 uint64_t flush_pos; ///< where we will flush. if
299 /// write_pos>flush_pos, we're buffering writes.
300 uint64_t safe_pos; ///< what has been committed safely to disk.
302 uint64_t next_safe_pos; /// start postion of the first entry that isn't
303 /// being fully flushed. If we don't flush any
304 // partial entry, it's equal to flush_pos.
306 bufferlist write_buf; ///< write buffer. flush_pos +
307 /// write_buf.length() == write_pos.
309 // protect write_buf from bufferlist _len overflow
310 Throttle write_buf_throttle;
312 bool waiting_for_zero;
313 interval_set<uint64_t> pending_zero; // non-contig bits we've zeroed
314 std::map<uint64_t, uint64_t> pending_safe; // flush_pos -> safe_pos
315 // when safe through given offset
316 std::map<uint64_t, std::list<Context*> > waitfor_safe;
318 void _flush(C_OnFinisher *onsafe);
319 void _do_flush(unsigned amount=0);
320 void _finish_flush(int r, uint64_t start, ceph::real_time stamp);
322 friend class C_Flush;
325 uint64_t read_pos; // logical read position, where next entry starts.
326 uint64_t requested_pos; // what we've requested from OSD.
327 uint64_t received_pos; // what we've received from OSD.
328 // read buffer. unused_field + read_buf.length() == prefetch_pos.
331 map<uint64_t,bufferlist> prefetch_buf;
333 uint64_t fetch_len; // how much to read at a time
334 uint64_t temp_fetch_len;
336 // for wait_for_readable()
337 C_OnFinisher *on_readable;
338 C_OnFinisher *on_write_error;
339 bool called_write_error;
341 // read completion callback
342 void _finish_read(int r, uint64_t offset, uint64_t length, bufferlist &bl);
343 void _finish_retry_read(int r);
344 void _assimilate_prefetch();
345 void _issue_read(uint64_t len); // read some more
346 void _prefetch(); // maybe read ahead
350 friend class C_RetryRead;
353 uint64_t expire_pos; // what we're allowed to trim to
354 uint64_t trimming_pos; // what we've requested to trim through
355 uint64_t trimmed_pos; // what has been trimmed
359 void _finish_trim(int r, uint64_t to);
363 void _issue_prezero();
364 void _finish_prezero(int r, uint64_t from, uint64_t len);
365 friend struct C_Journaler_Prezero;
367 // only init_headers when following or first reading off-disk
368 void init_headers(Header& h) {
370 state == STATE_READHEAD ||
371 state == STATE_REREADHEAD);
372 last_written = last_committed = h;
376 * handle a write error
378 * called when we get an objecter error on a write.
380 * @param r error code
382 void handle_write_error(int r);
386 void _finish_erase(int data_result, C_OnFinisher *completion);
388 friend class C_EraseFinish;
390 C_OnFinisher *wrap_finisher(Context *c);
392 uint32_t write_iohint; // the fadvise flags for write op, see
393 // CEPH_OSD_OP_FADIVSE_*
396 Journaler(const std::string &name_, inodeno_t ino_, int64_t pool,
397 const char *mag, Objecter *obj, PerfCounters *l, int lkey, Finisher *f) :
399 cct(obj->cct), name(name_), finisher(f), last_written(mag),
400 ino(ino_), pg_pool(pool), readonly(true),
401 stream_format(-1), journal_stream(-1),
403 objecter(obj), filer(objecter, f), logger(l), logger_key_lat(lkey),
404 delay_flush_event(0),
405 state(STATE_UNDEF), error(0),
406 prezeroing_pos(0), prezero_pos(0), write_pos(0), flush_pos(0),
407 safe_pos(0), next_safe_pos(0),
408 write_buf_throttle(cct, "write_buf_throttle", UINT_MAX - (UINT_MAX >> 3)),
409 waiting_for_zero(false),
410 read_pos(0), requested_pos(0), received_pos(0),
411 fetch_len(0), temp_fetch_len(0),
412 on_readable(0), on_write_error(NULL), called_write_error(false),
413 expire_pos(0), trimming_pos(0), trimmed_pos(0), readable(false),
414 write_iohint(0), stopping(false)
420 * NOTE: we assume the caller knows/has ensured that any objects in
421 * our sequence do not exist.. e.g. after a MKFS. this is _not_ an
426 assert(state == STATE_ACTIVE);
429 delay_flush_event = NULL;
442 assert(!on_readable);
446 waiting_for_zero = false;
449 // Asynchronous operations
450 // =======================
451 void erase(Context *completion);
452 void create(file_layout_t *layout, stream_format_t const sf);
453 void recover(Context *onfinish);
454 void reread_head(Context *onfinish);
455 void reread_head_and_probe(Context *onfinish);
456 void write_head(Context *onsave=0);
457 void wait_for_flush(Context *onsafe = 0);
458 void flush(Context *onsafe = 0);
459 void wait_for_readable(Context *onfinish);
460 bool have_waiter() const;
462 // Synchronous setters
463 // ===================
464 void set_layout(file_layout_t const *l);
466 void set_writeable();
467 void set_write_pos(uint64_t p) {
469 prezeroing_pos = prezero_pos = write_pos = flush_pos = safe_pos = next_safe_pos = p;
471 void set_read_pos(uint64_t p) {
473 // we can't cope w/ in-progress read right now.
474 assert(requested_pos == received_pos);
475 read_pos = requested_pos = received_pos = p;
478 uint64_t append_entry(bufferlist& bl);
479 void set_expire_pos(uint64_t ep) {
483 void set_trimmed_pos(uint64_t p) {
485 trimming_pos = trimmed_pos = p;
488 bool _write_head_needed();
489 bool write_head_needed() {
491 return _write_head_needed();
503 void set_write_error_handler(Context *c);
505 void set_write_iohint(uint32_t iohint_flags) {
506 write_iohint = iohint_flags;
509 * Cause any ongoing waits to error out with -EAGAIN, set error
517 // Synchronous getters
518 // ===================
519 // TODO: need some locks on reads for true safety
520 uint64_t get_layout_period() const {
521 return layout.get_period();
523 file_layout_t& get_layout() { return layout; }
524 bool is_active() { return state == STATE_ACTIVE; }
525 int get_error() { return error; }
526 bool is_readonly() { return readonly; }
528 bool try_read_entry(bufferlist& bl);
529 uint64_t get_write_pos() const { return write_pos; }
530 uint64_t get_write_safe_pos() const { return safe_pos; }
531 uint64_t get_read_pos() const { return read_pos; }
532 uint64_t get_expire_pos() const { return expire_pos; }
533 uint64_t get_trimmed_pos() const { return trimmed_pos; }
535 WRITE_CLASS_ENCODER(Journaler::Header)