/* Copyright (c) 2006-2009 Z RESEARCH, Inc. This file is part of GlusterFS. GlusterFS is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or (at your option) any later version. GlusterFS is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . */ /* TODO: - handle O_DIRECT - maintain offset, flush on lseek - ensure efficient memory managment in case of random seek */ #ifndef _CONFIG_H #define _CONFIG_H #include "config.h" #endif #include "glusterfs.h" #include "logging.h" #include "dict.h" #include "xlator.h" #include "read-ahead.h" #include #include static void read_ahead (call_frame_t *frame, ra_file_t *file); int ra_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, fd_t *fd) { ra_conf_t *conf = NULL; ra_file_t *file = NULL; int ret = 0; conf = this->private; if (op_ret == -1) { goto unwind; } file = CALLOC (1, sizeof (*file)); if (!file) { op_ret = -1; op_errno = ENOMEM; gf_log (this->name, GF_LOG_ERROR, "out of memory"); goto unwind; } ret = fd_ctx_set (fd, this, (uint64_t)(long)file); /* If mandatory locking has been enabled on this file, we disable caching on it */ if ((fd->inode->st_mode & S_ISGID) && !(fd->inode->st_mode & S_IXGRP)) file->disabled = 1; /* If O_DIRECT open, we disable caching on it */ if ((fd->flags & O_DIRECT) || (fd->flags & O_WRONLY)) file->disabled = 1; file->offset = (unsigned long long) 0; file->conf = conf; file->pages.next = &file->pages; file->pages.prev = &file->pages; file->pages.offset = (unsigned long long) 0; file->pages.file = file; ra_conf_lock (conf); { file->next = conf->files.next; conf->files.next = file; file->next->prev = file; file->prev = &conf->files; } ra_conf_unlock (conf); file->fd = fd; file->page_count = conf->page_count; file->page_size = conf->page_size; pthread_mutex_init (&file->file_lock, NULL); if (!file->disabled) { file->page_count = 1; } unwind: STACK_UNWIND (frame, op_ret, op_errno, fd); return 0; } int ra_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode, struct stat *buf) { ra_conf_t *conf = NULL; ra_file_t *file = NULL; int ret = 0; conf = this->private; if (op_ret == -1) { goto unwind; } file = CALLOC (1, sizeof (*file)); if (!file) { op_ret = -1; op_errno = ENOMEM; gf_log (this->name, GF_LOG_ERROR, "out of memory"); goto unwind; } ret = fd_ctx_set (fd, this, (uint64_t)(long)file); /* If mandatory locking has been enabled on this file, we disable caching on it */ if ((fd->inode->st_mode & S_ISGID) && !(fd->inode->st_mode & S_IXGRP)) file->disabled = 1; /* If O_DIRECT open, we disable caching on it */ if ((fd->flags & O_DIRECT) || (fd->flags & O_WRONLY)) file->disabled = 1; file->offset = (unsigned long long) 0; //file->size = fd->inode->buf.st_size; file->conf = conf; file->pages.next = &file->pages; file->pages.prev = &file->pages; file->pages.offset = (unsigned long long) 0; file->pages.file = file; ra_conf_lock (conf); { file->next = conf->files.next; conf->files.next = file; file->next->prev = file; file->prev = &conf->files; } ra_conf_unlock (conf); file->fd = fd; file->page_count = conf->page_count; file->page_size = conf->page_size; pthread_mutex_init (&file->file_lock, NULL); unwind: STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf); return 0; } int ra_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags, fd_t *fd) { STACK_WIND (frame, ra_open_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->open, loc, flags, fd); return 0; } int ra_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags, mode_t mode, fd_t *fd) { STACK_WIND (frame, ra_create_cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->create, loc, flags, mode, fd); return 0; } /* free cache pages between offset and offset+size, does not touch pages with frames waiting on it */ static void flush_region (call_frame_t *frame, ra_file_t *file, off_t offset, off_t size) { ra_page_t *trav = NULL; ra_page_t *next = NULL; ra_file_lock (file); { trav = file->pages.next; while (trav != &file->pages && trav->offset < (offset + size)) { next = trav->next; if (trav->offset >= offset && !trav->waitq) { ra_page_purge (trav); } trav = next; } } ra_file_unlock (file); } int ra_release (xlator_t *this, fd_t *fd) { uint64_t tmp_file = 0; int ret = 0; ret = fd_ctx_del (fd, this, &tmp_file); if (!ret) { ra_file_destroy ((ra_file_t *)(long)tmp_file); } return 0; } void read_ahead (call_frame_t *frame, ra_file_t *file) { off_t ra_offset = 0; size_t ra_size = 0; off_t trav_offset = 0; ra_page_t *trav = NULL; off_t cap = 0; char fault = 0; if (!file->page_count) return; ra_size = file->page_size * file->page_count; ra_offset = floor (file->offset, file->page_size); cap = file->size ? file->size : file->offset + ra_size; while (ra_offset < min (file->offset + ra_size, cap)) { ra_file_lock (file); { trav = ra_page_get (file, ra_offset); } ra_file_unlock (file); if (!trav) break; ra_offset += file->page_size; } if (trav) /* comfortable enough */ return; trav_offset = ra_offset; trav = file->pages.next; cap = file->size ? file->size : ra_offset + ra_size; while (trav_offset < min(ra_offset + ra_size, cap)) { fault = 0; ra_file_lock (file); { trav = ra_page_get (file, trav_offset); if (!trav) { fault = 1; trav = ra_page_create (file, trav_offset); if (trav) trav->dirty = 1; } } ra_file_unlock (file); if (!trav) { /* OUT OF MEMORY */ break; } if (fault) { gf_log (frame->this->name, GF_LOG_TRACE, "RA at offset=%"PRId64, trav_offset); ra_page_fault (file, frame, trav_offset); } trav_offset += file->page_size; } return; } int ra_need_atime_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, struct iovec *vector, int32_t count, struct stat *stbuf, struct iobref *iobref) { STACK_DESTROY (frame->root); return 0; } static void dispatch_requests (call_frame_t *frame, ra_file_t *file) { ra_local_t *local = NULL; ra_conf_t *conf = NULL; off_t rounded_offset = 0; off_t rounded_end = 0; off_t trav_offset = 0; ra_page_t *trav = NULL; call_frame_t *ra_frame = NULL; char need_atime_update = 1; char fault = 0; local = frame->local; conf = file->conf; rounded_offset = floor (local->offset, file->page_size); rounded_end = roof (local->offset + local->size, file->page_size); trav_offset = rounded_offset; trav = file->pages.next; while (trav_offset < rounded_end) { fault = 0; ra_file_lock (file); { trav = ra_page_get (file, trav_offset); if (!trav) { trav = ra_page_create (file, trav_offset); fault = 1; need_atime_update = 0; } if (!trav) { local->op_ret = -1; local->op_errno = ENOMEM; goto unlock; } if (trav->ready) { gf_log (frame->this->name, GF_LOG_TRACE, "HIT at offset=%"PRId64".", trav_offset); ra_frame_fill (trav, frame); } else { gf_log (frame->this->name, GF_LOG_TRACE, "IN-TRANSIT at offset=%"PRId64".", trav_offset); ra_wait_on_page (trav, frame); need_atime_update = 0; } } unlock: ra_file_unlock (file); if (fault) { gf_log (frame->this->name, GF_LOG_TRACE, "MISS at offset=%"PRId64".", trav_offset); ra_page_fault (file, frame, trav_offset); } trav_offset += file->page_size; } if (need_atime_update && conf->force_atime_update) { /* TODO: use untimens() since readv() can confuse underlying io-cache and others */ ra_frame = copy_frame (frame); if (ra_frame == NULL) { goto out; } STACK_WIND (ra_frame, ra_need_atime_cbk, FIRST_CHILD (frame->this), FIRST_CHILD (frame->this)->fops->readv, file->fd, 1, 1); } out: return ; } int ra_readv_disabled_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, struct iovec *vector, int32_t count, struct stat *stbuf, struct iobref *iobref) { STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf, iobref); return 0; } int ra_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size, off_t offset) { ra_file_t *file = NULL; ra_local_t *local = NULL; ra_conf_t *conf = NULL; int op_errno = 0; int ret = 0; char expected_offset = 1; uint64_t tmp_file = 0; conf = this->private; gf_log (this->name, GF_LOG_TRACE, "NEW REQ at offset=%"PRId64" for size=%"GF_PRI_SIZET"", offset, size); ret = fd_ctx_get (fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (file == NULL) { op_errno = EBADF; gf_log (this->name, GF_LOG_DEBUG, "readv received on fd with no" " file set in its context"); goto unwind; } if (file->offset != offset) { gf_log (this->name, GF_LOG_DEBUG, "unexpected offset (%"PRId64" != %"PRId64") resetting", file->offset, offset); expected_offset = file->expected = file->page_count = 0; } else { gf_log (this->name, GF_LOG_TRACE, "expected offset (%"PRId64") when page_count=%d", offset, file->page_count); if (file->expected < (conf->page_size * conf->page_count)) { file->expected += size; file->page_count = min ((file->expected / file->page_size), conf->page_count); } } if (!expected_offset) { flush_region (frame, file, 0, file->pages.prev->offset + 1); } if (file->disabled) { STACK_WIND (frame, ra_readv_disabled_cbk, FIRST_CHILD (frame->this), FIRST_CHILD (frame->this)->fops->readv, file->fd, size, offset); return 0; } local = (void *) CALLOC (1, sizeof (*local)); if (!local) { gf_log (this->name, GF_LOG_ERROR, "out of memory"); op_errno = ENOMEM; goto unwind; } local->fd = fd; local->offset = offset; local->size = size; local->wait_count = 1; local->fill.next = &local->fill; local->fill.prev = &local->fill; pthread_mutex_init (&local->local_lock, NULL); frame->local = local; dispatch_requests (frame, file); flush_region (frame, file, 0, floor (offset, file->page_size)); read_ahead (frame, file); ra_frame_return (frame); file->offset = offset + size; return 0; unwind: STACK_UNWIND (frame, -1, op_errno, NULL, 0, NULL); return 0; } int ra_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno) { STACK_UNWIND (frame, op_ret, op_errno); return 0; } int ra_flush (call_frame_t *frame, xlator_t *this, fd_t *fd) { ra_file_t *file = NULL; int ret = 0; uint64_t tmp_file = 0; int32_t op_errno = 0; ret = fd_ctx_get (fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (file == NULL) { op_errno = EBADF; gf_log (this->name, GF_LOG_DEBUG, "flush received on fd with no" " file set in its context"); goto unwind; } flush_region (frame, file, 0, file->pages.prev->offset+1); STACK_WIND (frame, ra_flush_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->flush, fd); return 0; unwind: STACK_UNWIND (frame, -1, op_errno); return 0; } int ra_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync) { ra_file_t *file = NULL; int ret = 0; uint64_t tmp_file = 0; int32_t op_errno = 0; ret = fd_ctx_get (fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (file == NULL) { op_errno = EBADF; gf_log (this->name, GF_LOG_DEBUG, "fsync received on fd with no" " file set in its context"); goto unwind; } if (file) { flush_region (frame, file, 0, file->pages.prev->offset+1); } STACK_WIND (frame, ra_flush_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsync, fd, datasync); return 0; unwind: STACK_UNWIND (frame, -1, op_errno); return 0; } int ra_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, struct stat *stbuf) { fd_t *fd = NULL; ra_file_t *file = NULL; int ret = 0; uint64_t tmp_file = 0; fd = frame->local; ret = fd_ctx_get (fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; flush_region (frame, file, 0, file->pages.prev->offset+1); frame->local = NULL; STACK_UNWIND (frame, op_ret, op_errno, stbuf); return 0; } int ra_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector, int32_t count, off_t offset, struct iobref *iobref) { ra_file_t *file = NULL; int ret = 0; uint64_t tmp_file = 0; int32_t op_errno = 0; ret = fd_ctx_get (fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (file == NULL) { op_errno = EBADF; gf_log (this->name, GF_LOG_DEBUG, "writev received on fd with" "no file set in its context"); goto unwind; } flush_region (frame, file, 0, file->pages.prev->offset+1); /* reset the read-ahead counters too */ file->expected = file->page_count = 0; frame->local = fd; STACK_WIND (frame, ra_writev_cbk, FIRST_CHILD(this), FIRST_CHILD(this)->fops->writev, fd, vector, count, offset, iobref); return 0; unwind: STACK_UNWIND (frame, -1, op_errno, NULL); return 0; } int ra_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret, int32_t op_errno, struct stat *buf) { STACK_UNWIND (frame, op_ret, op_errno, buf); return 0; } int ra_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset) { ra_file_t *file = NULL; fd_t *iter_fd = NULL; inode_t *inode = NULL; int ret = 0; uint64_t tmp_file = 0; inode = loc->inode; LOCK (&inode->lock); { list_for_each_entry (iter_fd, &inode->fd_list, inode_list) { ret = fd_ctx_get (iter_fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (!file) continue; flush_region (frame, file, 0, file->pages.prev->offset + 1); } } UNLOCK (&inode->lock); STACK_WIND (frame, ra_attr_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->truncate, loc, offset); return 0; } int ra_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd) { ra_file_t *file = NULL; fd_t *iter_fd = NULL; inode_t *inode = NULL; int ret = 0; uint64_t tmp_file = 0; inode = fd->inode; LOCK (&inode->lock); { list_for_each_entry (iter_fd, &inode->fd_list, inode_list) { ret = fd_ctx_get (iter_fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (!file) continue; flush_region (frame, file, 0, file->pages.prev->offset + 1); } } UNLOCK (&inode->lock); STACK_WIND (frame, ra_attr_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->fstat, fd); return 0; } int ra_fchown (call_frame_t *frame, xlator_t *this, fd_t *fd, uid_t uid, gid_t gid) { ra_file_t *file = NULL; fd_t *iter_fd = NULL; inode_t *inode = NULL; int ret = 0; uint64_t tmp_file = 0; inode = fd->inode; LOCK (&inode->lock); { list_for_each_entry (iter_fd, &inode->fd_list, inode_list) { ret = fd_ctx_get (iter_fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (!file) continue; flush_region (frame, file, 0, file->pages.prev->offset + 1); } } UNLOCK (&inode->lock); STACK_WIND (frame, ra_attr_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->fchown, fd, uid, gid); return 0; } int ra_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset) { ra_file_t *file = NULL; fd_t *iter_fd = NULL; inode_t *inode = NULL; int ret = 0; uint64_t tmp_file = 0; inode = fd->inode; LOCK (&inode->lock); { list_for_each_entry (iter_fd, &inode->fd_list, inode_list) { ret = fd_ctx_get (iter_fd, this, &tmp_file); file = (ra_file_t *)(long)tmp_file; if (!file) continue; flush_region (frame, file, 0, file->pages.prev->offset + 1); } } UNLOCK (&inode->lock); STACK_WIND (frame, ra_attr_cbk, FIRST_CHILD (this), FIRST_CHILD (this)->fops->ftruncate, fd, offset); return 0; } int init (xlator_t *this) { ra_conf_t *conf; dict_t *options = this->options; char *page_count_string = NULL; int32_t ret = -1; if (!this->children || this->children->next) { gf_log (this->name, GF_LOG_ERROR, "FATAL: read-ahead not configured with exactly one" " child"); goto out; } if (!this->parents) { gf_log (this->name, GF_LOG_WARNING, "dangling volume. check volfile "); } conf = (void *) CALLOC (1, sizeof (*conf)); if (conf == NULL) { gf_log (this->name, GF_LOG_ERROR, "FATAL: Out of memory"); goto out; } conf->page_size = this->ctx->page_size; conf->page_count = 4; if (dict_get (options, "page-count")) page_count_string = data_to_str (dict_get (options, "page-count")); if (page_count_string) { if (gf_string2uint_base10 (page_count_string, &conf->page_count) != 0) { gf_log ("read-ahead", GF_LOG_ERROR, "invalid number format \"%s\" of \"option " "page-count\"", page_count_string); goto out; } gf_log (this->name, GF_LOG_DEBUG, "Using conf->page_count = %u", conf->page_count); } if (dict_get (options, "force-atime-update")) { char *force_atime_update_str = data_to_str (dict_get (options, "force-atime-update")); if (gf_string2boolean (force_atime_update_str, &conf->force_atime_update) == -1) { gf_log (this->name, GF_LOG_ERROR, "'force-atime-update' takes only boolean " "options"); goto out; } if (conf->force_atime_update) gf_log (this->name, GF_LOG_DEBUG, "Forcing atime " "updates on cache hit"); } conf->files.next = &conf->files; conf->files.prev = &conf->files; pthread_mutex_init (&conf->conf_lock, NULL); this->private = conf; ret = 0; out: if (ret == -1) { if (conf != NULL) { FREE (conf); } } return ret; } void fini (xlator_t *this) { ra_conf_t *conf = this->private; pthread_mutex_destroy (&conf->conf_lock); FREE (conf); this->private = NULL; return; } struct xlator_fops fops = { .open = ra_open, .create = ra_create, .readv = ra_readv, .writev = ra_writev, .flush = ra_flush, .fsync = ra_fsync, .truncate = ra_truncate, .ftruncate = ra_ftruncate, .fstat = ra_fstat, .fchown = ra_fchown, }; struct xlator_mops mops = { }; struct xlator_cbks cbks = { .release = ra_release, }; struct volume_options options[] = { { .key = {"force-atime-update"}, .type = GF_OPTION_TYPE_BOOL }, { .key = {"page-count"}, .type = GF_OPTION_TYPE_INT, .min = 1, .max = 16 }, { .key = {NULL} }, };