rocksdb/util/env_posix.cc

// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.

#include <deque>
#include <dirent.h>
#include <errno.h>
#include <fcntl.h>
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mman.h>
#include <sys/stat.h>
#include <sys/time.h>
#include <sys/types.h>
#include <time.h>
#include <unistd.h>
#if defined(LEVELDB_PLATFORM_ANDROID)
#include <sys/stat.h>
#endif
#include "leveldb/env.h"
#include "leveldb/slice.h"
#include "port/port.h"
#include "util/logging.h"

namespace leveldb {

namespace {

class PosixSequentialFile: public SequentialFile {
 private:
  std::string filename_;
  FILE* file_;

 public:
  PosixSequentialFile(const std::string& fname, FILE* f)
      : filename_(fname), file_(f) { }
  virtual ~PosixSequentialFile() { fclose(file_); }

  virtual Status Read(size_t n, Slice* result, char* scratch) {
    Status s;
    size_t r = fread_unlocked(scratch, 1, n, file_);
    *result = Slice(scratch, r);
    if (r < n) {
      if (feof(file_)) {
        // We leave status as ok if we hit the end of the file
      } else {
        // A partial read with an error: return a non-ok status
        s = Status::IOError(filename_, strerror(errno));
      }
    }
    return s;
  }

  virtual Status Skip(uint64_t n) {
    if (fseek(file_, n, SEEK_CUR)) {
      return Status::IOError(filename_, strerror(errno));
    }
    return Status::OK();
  }
};

class PosixRandomAccessFile: public RandomAccessFile {
 private:
  std::string filename_;
  int fd_;

 public:
  PosixRandomAccessFile(const std::string& fname, int fd)
      : filename_(fname), fd_(fd) { }
  virtual ~PosixRandomAccessFile() { close(fd_); }

  virtual Status Read(uint64_t offset, size_t n, Slice* result,
                      char* scratch) const {
    Status s;
    ssize_t r = pread(fd_, scratch, n, static_cast<off_t>(offset));
    *result = Slice(scratch, (r < 0) ? 0 : r);
    if (r < 0) {
      // An error: return a non-ok status
      s = Status::IOError(filename_, strerror(errno));
    }
    return s;
  }
};

// We preallocate up to an extra megabyte and use memcpy to append new
// data to the file.  This is safe since we either properly close the
// file before reading from it, or for log files, the reading code
// knows enough to skip zero suffixes.
class PosixMmapFile : public WritableFile {
 private:
  std::string filename_;
  int fd_;
  size_t page_size_;
  size_t map_size_;       // How much extra memory to map at a time
  char* base_;            // The mapped region
  char* limit_;           // Limit of the mapped region
  char* dst_;             // Where to write next  (in range [base_,limit_])
  char* last_sync_;       // Where have we synced up to
  uint64_t file_offset_;  // Offset of base_ in file

  // Have we done an munmap of unsynced data?
  bool pending_sync_;

  // Roundup x to a multiple of y
  static size_t Roundup(size_t x, size_t y) {
    return ((x + y - 1) / y) * y;
  }

  size_t TruncateToPageBoundary(size_t s) {
    s -= (s & (page_size_ - 1));
    assert((s % page_size_) == 0);
    return s;
  }

  void UnmapCurrentRegion() {
    if (base_ != NULL) {
      if (last_sync_ < limit_) {
        // Defer syncing this data until next Sync() call, if any
        pending_sync_ = true;
      }
      munmap(base_, limit_ - base_);
      file_offset_ += limit_ - base_;
      base_ = NULL;
      limit_ = NULL;
      last_sync_ = NULL;
      dst_ = NULL;

      // Increase the amount we map the next time, but capped at 1MB
      if (map_size_ < (1<<20)) {
        map_size_ *= 2;
      }
    }
  }

  bool MapNewRegion() {
    assert(base_ == NULL);
    if (ftruncate(fd_, file_offset_ + map_size_) < 0) {
      return false;
    }
    void* ptr = mmap(NULL, map_size_, PROT_READ | PROT_WRITE, MAP_SHARED,
                     fd_, file_offset_);
    if (ptr == MAP_FAILED) {
      return false;
    }
    base_ = reinterpret_cast<char*>(ptr);
    limit_ = base_ + map_size_;
    dst_ = base_;
    last_sync_ = base_;
    return true;
  }

 public:
  PosixMmapFile(const std::string& fname, int fd, size_t page_size)
      : filename_(fname),
        fd_(fd),
        page_size_(page_size),
        map_size_(Roundup(65536, page_size)),
        base_(NULL),
        limit_(NULL),
        dst_(NULL),
        last_sync_(NULL),
        file_offset_(0),
        pending_sync_(false) {
    assert((page_size & (page_size - 1)) == 0);
  }


  ~PosixMmapFile() {
    if (fd_ >= 0) {
      PosixMmapFile::Close();
    }
  }

  virtual Status Append(const Slice& data) {
    const char* src = data.data();
    size_t left = data.size();
    while (left > 0) {
      assert(base_ <= dst_);
      assert(dst_ <= limit_);
      size_t avail = limit_ - dst_;
      if (avail == 0) {
        UnmapCurrentRegion();
        MapNewRegion();
      }

      size_t n = (left <= avail) ? left : avail;
      memcpy(dst_, src, n);
      dst_ += n;
      src += n;
      left -= n;
    }
    return Status::OK();
  }

  virtual Status Close() {
    Status s;
    size_t unused = limit_ - dst_;
    UnmapCurrentRegion();
    if (unused > 0) {
      // Trim the extra space at the end of the file
      if (ftruncate(fd_, file_offset_ - unused) < 0) {
        s = Status::IOError(filename_, strerror(errno));
      }
    }

    if (close(fd_) < 0) {
      if (s.ok()) {
        s = Status::IOError(filename_, strerror(errno));
      }
    }

    fd_ = -1;
    base_ = NULL;
    limit_ = NULL;
    return s;
  }

  virtual Status Flush() {
    return Status::OK();
  }

  virtual Status Sync() {
    Status s;

    if (pending_sync_) {
      // Some unmapped data was not synced
      pending_sync_ = false;
      if (fdatasync(fd_) < 0) {
        s = Status::IOError(filename_, strerror(errno));
      }
    }

    if (dst_ > last_sync_) {
      // Find the beginnings of the pages that contain the first and last
      // bytes to be synced.
      size_t p1 = TruncateToPageBoundary(last_sync_ - base_);
      size_t p2 = TruncateToPageBoundary(dst_ - base_ - 1);
      last_sync_ = dst_;
      if (msync(base_ + p1, p2 - p1 + page_size_, MS_SYNC) < 0) {
        s = Status::IOError(filename_, strerror(errno));
      }
    }

    return s;
  }
};

static int LockOrUnlock(int fd, bool lock) {
  errno = 0;
  struct flock f;
  memset(&f, 0, sizeof(f));
  f.l_type = (lock ? F_WRLCK : F_UNLCK);
  f.l_whence = SEEK_SET;
  f.l_start = 0;
  f.l_len = 0;        // Lock/unlock entire file
  return fcntl(fd, F_SETLK, &f);
}

class PosixFileLock : public FileLock {
 public:
  int fd_;
};

class PosixEnv : public Env {
 public:
  PosixEnv();
  virtual ~PosixEnv() {
    fprintf(stderr, "Destroying Env::Default()\n");
    exit(1);
  }

  virtual Status NewSequentialFile(const std::string& fname,
                                   SequentialFile** result) {
    FILE* f = fopen(fname.c_str(), "r");
    if (f == NULL) {
      *result = NULL;
      return Status::IOError(fname, strerror(errno));
    } else {
      *result = new PosixSequentialFile(fname, f);
      return Status::OK();
    }
  }

  virtual Status NewRandomAccessFile(const std::string& fname,
                                     RandomAccessFile** result) {
    int fd = open(fname.c_str(), O_RDONLY);
    if (fd < 0) {
      *result = NULL;
      return Status::IOError(fname, strerror(errno));
    }
    *result = new PosixRandomAccessFile(fname, fd);
    return Status::OK();
  }

  virtual Status NewWritableFile(const std::string& fname,
                                 WritableFile** result) {
    Status s;
    const int fd = open(fname.c_str(), O_CREAT | O_RDWR | O_TRUNC, 0644);
    if (fd < 0) {
      *result = NULL;
      s = Status::IOError(fname, strerror(errno));
    } else {
      *result = new PosixMmapFile(fname, fd, page_size_);
    }
    return s;
  }

  virtual bool FileExists(const std::string& fname) {
    return access(fname.c_str(), F_OK) == 0;
  }

  virtual Status GetChildren(const std::string& dir,
                             std::vector<std::string>* result) {
    result->clear();
    DIR* d = opendir(dir.c_str());
    if (d == NULL) {
      return Status::IOError(dir, strerror(errno));
    }
    struct dirent* entry;
    while ((entry = readdir(d)) != NULL) {
      result->push_back(entry->d_name);
    }
    closedir(d);
    return Status::OK();
  }

  virtual Status DeleteFile(const std::string& fname) {
    Status result;
    if (unlink(fname.c_str()) != 0) {
      result = Status::IOError(fname, strerror(errno));
    }
    return result;
  };

  virtual Status CreateDir(const std::string& name) {
    Status result;
    if (mkdir(name.c_str(), 0755) != 0) {
      result = Status::IOError(name, strerror(errno));
    }
    return result;
  };

  virtual Status DeleteDir(const std::string& name) {
    Status result;
    if (rmdir(name.c_str()) != 0) {
      result = Status::IOError(name, strerror(errno));
    }
    return result;
  };

  virtual Status GetFileSize(const std::string& fname, uint64_t* size) {
    Status s;
    struct stat sbuf;
    if (stat(fname.c_str(), &sbuf) != 0) {
      *size = 0;
      s = Status::IOError(fname, strerror(errno));
    } else {
      *size = sbuf.st_size;
    }
    return s;
  }

  virtual Status RenameFile(const std::string& src, const std::string& target) {
    Status result;
    if (rename(src.c_str(), target.c_str()) != 0) {
      result = Status::IOError(src, strerror(errno));
    }
    return result;
  }

  virtual Status LockFile(const std::string& fname, FileLock** lock) {
    *lock = NULL;
    Status result;
    int fd = open(fname.c_str(), O_RDWR | O_CREAT, 0644);
    if (fd < 0) {
      result = Status::IOError(fname, strerror(errno));
    } else if (LockOrUnlock(fd, true) == -1) {
      result = Status::IOError("lock " + fname, strerror(errno));
      close(fd);
    } else {
      PosixFileLock* my_lock = new PosixFileLock;
      my_lock->fd_ = fd;
      *lock = my_lock;
    }
    return result;
  }

  virtual Status UnlockFile(FileLock* lock) {
    PosixFileLock* my_lock = reinterpret_cast<PosixFileLock*>(lock);
    Status result;
    if (LockOrUnlock(my_lock->fd_, false) == -1) {
      result = Status::IOError(strerror(errno));
    }
    close(my_lock->fd_);
    delete my_lock;
    return result;
  }

  virtual void Schedule(void (*function)(void*), void* arg);

  virtual void StartThread(void (*function)(void* arg), void* arg);

  virtual Status GetTestDirectory(std::string* result) {
    const char* env = getenv("TEST_TMPDIR");
    if (env && env[0] != '\0') {
      *result = env;
    } else {
      char buf[100];
      snprintf(buf, sizeof(buf), "/tmp/leveldbtest-%d", int(geteuid()));
      *result = buf;
    }
    // Directory may already exist
    CreateDir(*result);
    return Status::OK();
  }

  virtual void Logv(WritableFile* info_log, const char* format, va_list ap) {
    pthread_t tid = pthread_self();
    uint64_t thread_id = 0;
    memcpy(&thread_id, &tid, std::min(sizeof(thread_id), sizeof(tid)));

    // We try twice: the first time with a fixed-size stack allocated buffer,
    // and the second time with a much larger dynamically allocated buffer.
    char buffer[500];
    for (int iter = 0; iter < 2; iter++) {
      char* base;
      int bufsize;
      if (iter == 0) {
        bufsize = sizeof(buffer);
        base = buffer;
      } else {
        bufsize = 30000;
        base = new char[bufsize];
      }
      char* p = base;
      char* limit = base + bufsize;

      struct timeval now_tv;
      gettimeofday(&now_tv, NULL);
      const time_t seconds = now_tv.tv_sec;
      struct tm t;
      localtime_r(&seconds, &t);
      p += snprintf(p, limit - p,
                    "%04d/%02d/%02d-%02d:%02d:%02d.%06d %llx ",
                    t.tm_year + 1900,
                    t.tm_mon + 1,
                    t.tm_mday,
                    t.tm_hour,
                    t.tm_min,
                    t.tm_sec,
                    static_cast<int>(now_tv.tv_usec),
                    static_cast<long long unsigned int>(thread_id));

      // Print the message
      if (p < limit) {
        va_list backup_ap;
        va_copy(backup_ap, ap);
        p += vsnprintf(p, limit - p, format, backup_ap);
        va_end(backup_ap);
      }

      // Truncate to available space if necessary
      if (p >= limit) {
        if (iter == 0) {
          continue;       // Try again with larger buffer
        } else {
          p = limit - 1;
        }
      }

      // Add newline if necessary
      if (p == base || p[-1] != '\n') {
        *p++ = '\n';
      }

      assert(p <= limit);
      info_log->Append(Slice(base, p - base));
      info_log->Flush();
      if (base != buffer) {
        delete[] base;
      }
      break;
    }
  }

  virtual uint64_t NowMicros() {
    struct timeval tv;
    gettimeofday(&tv, NULL);
    return static_cast<uint64_t>(tv.tv_sec) * 1000000 + tv.tv_usec;
  }

  virtual void SleepForMicroseconds(int micros) {
    usleep(micros);
  }

 private:
  void PthreadCall(const char* label, int result) {
    if (result != 0) {
      fprintf(stderr, "pthread %s: %s\n", label, strerror(result));
      exit(1);
    }
  }

  // BGThread() is the body of the background thread
  void BGThread();
  static void* BGThreadWrapper(void* arg) {
    reinterpret_cast<PosixEnv*>(arg)->BGThread();
    return NULL;
  }

  size_t page_size_;
  pthread_mutex_t mu_;
  pthread_cond_t bgsignal_;
  pthread_t bgthread_;
  bool started_bgthread_;

  // Entry per Schedule() call
  struct BGItem { void* arg; void (*function)(void*); };
  typedef std::deque<BGItem> BGQueue;
  BGQueue queue_;
};

PosixEnv::PosixEnv() : page_size_(getpagesize()),
                       started_bgthread_(false) {
  PthreadCall("mutex_init", pthread_mutex_init(&mu_, NULL));
  PthreadCall("cvar_init", pthread_cond_init(&bgsignal_, NULL));
}

void PosixEnv::Schedule(void (*function)(void*), void* arg) {
  PthreadCall("lock", pthread_mutex_lock(&mu_));

  // Start background thread if necessary
  if (!started_bgthread_) {
    started_bgthread_ = true;
    PthreadCall(
        "create thread",
        pthread_create(&bgthread_, NULL,  &PosixEnv::BGThreadWrapper, this));
  }

  // If the queue is currently empty, the background thread may currently be
  // waiting.
  if (queue_.empty()) {
    PthreadCall("signal", pthread_cond_signal(&bgsignal_));
  }

  // Add to priority queue
  queue_.push_back(BGItem());
  queue_.back().function = function;
  queue_.back().arg = arg;

  PthreadCall("unlock", pthread_mutex_unlock(&mu_));
}

void PosixEnv::BGThread() {
  while (true) {
    // Wait until there is an item that is ready to run
    PthreadCall("lock", pthread_mutex_lock(&mu_));
    while (queue_.empty()) {
      PthreadCall("wait", pthread_cond_wait(&bgsignal_, &mu_));
    }

    void (*function)(void*) = queue_.front().function;
    void* arg = queue_.front().arg;
    queue_.pop_front();

    PthreadCall("unlock", pthread_mutex_unlock(&mu_));
    (*function)(arg);
  }
}

namespace {
struct StartThreadState {
  void (*user_function)(void*);
  void* arg;
};
}
static void* StartThreadWrapper(void* arg) {
  StartThreadState* state = reinterpret_cast<StartThreadState*>(arg);
  state->user_function(state->arg);
  delete state;
  return NULL;
}

void PosixEnv::StartThread(void (*function)(void* arg), void* arg) {
  pthread_t t;
  StartThreadState* state = new StartThreadState;
  state->user_function = function;
  state->arg = arg;
  PthreadCall("start thread",
              pthread_create(&t, NULL,  &StartThreadWrapper, state));
}

}

static pthread_once_t once = PTHREAD_ONCE_INIT;
static Env* default_env;
static void InitDefaultEnv() { default_env = new PosixEnv; }

Env* Env::Default() {
  pthread_once(&once, InitDefaultEnv);
  return default_env;
}

}