From 77adf4cd648dce41f89469dd185deec6b6b53a0b Mon Sep 17 00:00:00 2001 From: Vikas Gorur Date: Wed, 18 Feb 2009 17:36:07 +0530 Subject: Added all files --- booster/Makefile.am | 1 + booster/src/Makefile.am | 17 + booster/src/booster.c | 920 ++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 938 insertions(+) create mode 100644 booster/Makefile.am create mode 100644 booster/src/Makefile.am create mode 100644 booster/src/booster.c (limited to 'booster') diff --git a/booster/Makefile.am b/booster/Makefile.am new file mode 100644 index 00000000..e1c45f30 --- /dev/null +++ b/booster/Makefile.am @@ -0,0 +1 @@ +SUBDIRS=src \ No newline at end of file diff --git a/booster/src/Makefile.am b/booster/src/Makefile.am new file mode 100644 index 00000000..9b6e77f9 --- /dev/null +++ b/booster/src/Makefile.am @@ -0,0 +1,17 @@ +xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/performance + +ldpreload_PROGRAMS = glusterfs-booster.so +ldpreloaddir = $(libdir)/glusterfs/ +glusterfs_booster_so_SOURCES = booster.c +glusterfs_booster_so_CFLAGS = -I$(top_srcdir)/libglusterfsclient/src/ -D_GNU_SOURCE -D$(GF_HOST_OS) -fPIC -Wall \ + -pthread $(GF_BOOSTER_CFLAGS) +glusterfs_booster_so_CPPFLAGS = -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE \ + -I$(top_srcdir)/libglusterfsclient/src \ + -I$(top_srcdir)/libglusterfs/src -DDATADIR=\"$(localstatedir)\" \ + -DCONFDIR=\"$(sysconfdir)/glusterfs\" +glusterfs_booster_so_LDFLAGS = -shared -nostartfiles +glusterfs_booster_so_LDADD = -L$(top_builddir)/libglusterfs/src -lglusterfs \ + -L$(top_builddir)/libglusterfsclient/src -lglusterfsclient + +CLEANFILES = + diff --git a/booster/src/booster.c b/booster/src/booster.c new file mode 100644 index 00000000..cf5f7883 --- /dev/null +++ b/booster/src/booster.c @@ -0,0 +1,920 @@ +/* + Copyright (c) 2007, 2008 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 + . +*/ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#ifndef GF_UNIT_KB +#define GF_UNIT_KB 1024 +#endif + +#ifndef UNIX_PATH_MAX +#define UNIX_PATH_MAX 108 +#endif + +struct _inode; +struct _dict; +struct _fd { + pid_t pid; + struct list_head inode_list; + struct _inode *inode; + struct _dict *ctx; + int32_t refcount; +}; + +typedef struct _fdtable fdtable_t; +typedef struct _fd fd_t; + + +inline void +gf_fd_put (struct _fdtable *fdtable, int64_t fd); + +struct _fd * +gf_fd_fdptr_get (struct _fdtable *fdtable, int64_t fd); + +struct _fdtable * +gf_fd_fdtable_alloc (void); + +void +gf_fd_fdtable_destroy (struct _fdtable *); + +int32_t +gf_fd_unused_get (struct _fdtable *fdtable, struct _fd *fdptr); + +int32_t +gf_fd_unused_get2 (struct _fdtable *fdtable, struct _fd *fdptr, int64_t fd); + +void +fd_unref (struct _fd *fd); + +fd_t * +fd_ref (struct _fd *fd); + +pid_t +getpid (void); + +ssize_t +write (int fd, const void *buf, size_t count); + +/* open, open64, creat */ +static int (*real_open) (const char *pathname, int flags, ...); +static int (*real_open64) (const char *pathname, int flags, ...); +static int (*real_creat) (const char *pathname, mode_t mode); + +/* read, readv, pread, pread64 */ +static ssize_t (*real_read) (int fd, void *buf, size_t count); +static ssize_t (*real_readv) (int fd, const struct iovec *vector, int count); +static ssize_t (*real_pread) (int fd, void *buf, size_t count, unsigned long offset); +static ssize_t (*real_pread64) (int fd, void *buf, size_t count, uint64_t offset); + +/* write, writev, pwrite, pwrite64 */ +static ssize_t (*real_write) (int fd, const void *buf, size_t count); +static ssize_t (*real_writev) (int fd, const struct iovec *vector, int count); +static ssize_t (*real_pwrite) (int fd, const void *buf, size_t count, unsigned long offset); +static ssize_t (*real_pwrite64) (int fd, const void *buf, size_t count, uint64_t offset); + +/* lseek, llseek, lseek64 */ +static off_t (*real_lseek) (int fildes, unsigned long offset, int whence); +static off_t (*real_lseek64) (int fildes, uint64_t offset, int whence); + +/* close */ +static int (*real_close) (int fd); + +/* dup dup2 */ +static int (*real_dup) (int fd); +static int (*real_dup2) (int oldfd, int newfd); + +static pid_t (*real_fork) (void); + +#define RESOLVE(sym) do { \ + if (!real_##sym) \ + real_##sym = dlsym (RTLD_NEXT, #sym); \ + } while (0) + +/*TODO: set proper value */ +#define MOUNT_HASH_SIZE 256 + +struct booster_mount { + dev_t st_dev; + libglusterfs_handle_t handle; + struct list_head device_list; +}; +typedef struct booster_mount booster_mount_t; + +struct booster_mount_table { + pthread_mutex_t lock; + struct list_head *mounts; + int32_t hash_size; +}; +typedef struct booster_mount_table booster_mount_table_t; + +static fdtable_t *booster_glfs_fdtable = NULL; +static booster_mount_table_t *booster_mount_table = NULL; + +static int32_t +booster_put_handle (booster_mount_table_t *table, + dev_t st_dev, + libglusterfs_handle_t handle) +{ + int32_t hash = 0; + booster_mount_t *mount = NULL, *tmp = NULL; + int32_t ret = 0; + + mount = calloc (1, sizeof (*mount)); + if (!mount) { + return -1; + } + + // ERR_ABORT (mount); + INIT_LIST_HEAD (&mount->device_list); + mount->st_dev = st_dev; + mount->handle = handle; + + hash = st_dev % table->hash_size; + + pthread_mutex_lock (&table->lock); + { + list_for_each_entry (tmp, &table->mounts[hash], device_list) { + if (tmp->st_dev == st_dev) { + ret = -1; + errno = EEXIST; + goto unlock; + } + } + + list_add (&mount->device_list, &table->mounts[hash]); + } +unlock: + pthread_mutex_unlock (&table->lock); + + return ret; +} + + +static inline long +booster_get_glfs_fd (fdtable_t *fdtable, int fd) +{ + fd_t *glfs_fd = NULL; + + glfs_fd = gf_fd_fdptr_get (fdtable, fd); + return (long) glfs_fd; +} + + +static inline void +booster_put_glfs_fd (long glfs_fd) +{ + fd_unref ((fd_t *)glfs_fd); +} + + +static inline int32_t +booster_get_unused_fd (fdtable_t *fdtable, long glfs_fd, int fd) +{ + int32_t ret = -1; + ret = gf_fd_unused_get2 (fdtable, (fd_t *)glfs_fd, fd); + return ret; +} + + +static inline void +booster_put_fd (fdtable_t *fdtable, int fd) +{ + gf_fd_put (fdtable, fd); +} + + +static libglusterfs_handle_t +booster_get_handle (booster_mount_table_t *table, dev_t st_dev) +{ + int32_t hash = 0; + booster_mount_t *mount = NULL; + libglusterfs_handle_t handle = NULL; + + hash = st_dev % table->hash_size; + + pthread_mutex_lock (&table->lock); + { + list_for_each_entry (mount, &table->mounts[hash], device_list) { + if (mount->st_dev == st_dev) { + handle = mount->handle; + break; + } + } + } + pthread_mutex_unlock (&table->lock); + + return handle; +} + + +void +do_open (int fd, int flags, mode_t mode) +{ + char *specfile = NULL; + libglusterfs_handle_t handle; + int32_t file_size; + struct stat st = {0,}; + int32_t ret = -1; + + ret = fstat (fd, &st); + if (ret == -1) { + return; + } + + if (!booster_mount_table) { + return; + } + + handle = booster_get_handle (booster_mount_table, st.st_dev); + if (!handle) { + FILE *specfp = NULL; + + glusterfs_init_ctx_t ctx = { + .loglevel = "critical", + .lookup_timeout = 600, + .stat_timeout = 600, + }; + + file_size = fgetxattr (fd, "user.glusterfs-booster-volfile", NULL, 0); + if (file_size == -1) { + return; + } + + specfile = calloc (1, file_size); + if (!specfile) { + fprintf (stderr, "cannot allocate memory: %s\n", strerror (errno)); + return; + } + + ret = fgetxattr (fd, "user.glusterfs-booster-volfile", specfile, file_size); + if (ret == -1) { + free (specfile); + return ; + } + + specfp = tmpfile (); + if (!specfp) { + free (specfile); + return; + } + + ret = fwrite (specfile, file_size, 1, specfp); + if (ret != 1) { + fclose (specfp); + free (specfile); + } + + fseek (specfp, 0L, SEEK_SET); + + ctx.logfile = getenv ("GLFS_BOOSTER_LOGFILE"); + ctx.specfp = specfp; + + handle = glusterfs_init (&ctx); + + free (specfile); + fclose (specfp); + + if (!handle) { + return; + } + + ret = booster_put_handle (booster_mount_table, st.st_dev, handle); + if (ret == -1) { + glusterfs_fini (handle); + if (errno != EEXIST) { + return; + } + } + } + + if (handle) { + long glfs_fd; + char path [UNIX_PATH_MAX]; + ret = fgetxattr (fd, "user.glusterfs-booster-path", path, UNIX_PATH_MAX); + if (ret == -1) { + return; + } + + glfs_fd = glusterfs_open (handle, path, flags, mode); + if (glfs_fd) { + ret = booster_get_unused_fd (booster_glfs_fdtable, glfs_fd, fd); + if (ret == -1) { + glusterfs_close (glfs_fd); + return; + } + } + } + + return; +} + +#ifndef __USE_FILE_OFFSET64 +int +open (const char *pathname, int flags, ...) +{ + int ret; + mode_t mode = 0; + va_list ap; + + if (flags & O_CREAT) { + va_start (ap, flags); + mode = va_arg (ap, mode_t); + va_end (ap); + + ret = real_open (pathname, flags, mode); + } else { + ret = real_open (pathname, flags); + } + + if (ret != -1) { + flags &= ~ O_CREAT; + do_open (ret, flags, mode); + } + + return ret; +} +#endif + +#if defined (__USE_LARGEFILE64) || !defined (__USE_FILE_OFFSET64) +int +open64 (const char *pathname, int flags, ...) +{ + int ret; + mode_t mode = 0; + va_list ap; + + if (flags & O_CREAT) { + va_start (ap, flags); + mode = va_arg (ap, mode_t); + va_end (ap); + + ret = real_open64 (pathname, flags, mode); + } else { + ret = real_open64 (pathname, flags); + } + + if (ret != -1) { + flags &= ~O_CREAT; + do_open (ret, flags, mode); + } + + return ret; +} +#endif + +int +creat (const char *pathname, mode_t mode) +{ + int ret; + + ret = real_creat (pathname, mode); + + if (ret != -1) { + do_open (ret, O_WRONLY | O_TRUNC, mode); + } + + return ret; +} + + +/* pread */ + +ssize_t +pread (int fd, void *buf, size_t count, unsigned long offset) +{ + ssize_t ret; + long glfs_fd = 0; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + if (!glfs_fd) { + ret = real_pread (fd, buf, count, offset); + } else { + ret = glusterfs_pread (glfs_fd, buf, count, offset); + if (ret == -1) { + ret = real_pread (fd, buf, count, offset); + } + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} + + +ssize_t +pread64 (int fd, void *buf, size_t count, uint64_t offset) +{ + ssize_t ret; + long glfs_fd = 0; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + if (!glfs_fd) { + ret = real_pread (fd, buf, count, offset); + } else { + ret = glusterfs_pread (glfs_fd, buf, count, offset); + if (ret == -1) { + ret = real_pread (fd, buf, count, offset); + } + } + + return ret; +} + + +ssize_t +read (int fd, void *buf, size_t count) +{ + int ret; + long glfs_fd; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + if (!glfs_fd) { + ret = real_read (fd, buf, count); + } else { + uint64_t offset = 0; + offset = real_lseek64 (fd, 0L, SEEK_CUR); + if ((int64_t)offset != -1) { + ret = glusterfs_lseek (glfs_fd, offset, SEEK_SET); + if (ret != -1) { + ret = glusterfs_read (glfs_fd, buf, count); + } + } else { + ret = -1; + } + + if (ret == -1) { + ret = real_read (fd, buf, count); + } + + if (ret > 0 && ((int64_t) offset) >= 0) { + real_lseek64 (fd, ret + offset, SEEK_SET); + } + + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} + + +ssize_t +readv (int fd, const struct iovec *vector, int count) +{ + int ret; + long glfs_fd = 0; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + if (!glfs_fd) { + ret = real_readv (fd, vector, count); + } else { + uint64_t offset = 0; + offset = real_lseek64 (fd, 0L, SEEK_CUR); + if ((int64_t)offset != -1) { + ret = glusterfs_lseek (glfs_fd, offset, SEEK_SET); + if (ret != -1) { + ret = glusterfs_readv (glfs_fd, vector, count); + } + } else { + ret = -1; + } + + ret = glusterfs_readv (glfs_fd, vector, count); + if (ret > 0) { + real_lseek64 (fd, offset + ret, SEEK_SET); + } + + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} + + +ssize_t +write (int fd, const void *buf, size_t count) +{ + int ret; + long glfs_fd = 0; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + + if (!glfs_fd) { + ret = real_write (fd, buf, count); + } else { + uint64_t offset = 0; + offset = real_lseek64 (fd, 0L, SEEK_CUR); + if (((int64_t) offset) != -1) { + ret = glusterfs_lseek (glfs_fd, offset, SEEK_SET); + if (ret != -1) { + ret = glusterfs_write (glfs_fd, buf, count); + } + } else { + ret = -1; + } + + if (ret == -1) { + ret = real_write (fd, buf, count); + } + + if (ret > 0 && ((int64_t) offset) >= 0) { + real_lseek64 (fd, offset + ret, SEEK_SET); + } + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} + +ssize_t +writev (int fd, const struct iovec *vector, int count) +{ + int ret = 0; + long glfs_fd = 0; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + + if (!glfs_fd) { + ret = real_writev (fd, vector, count); + } else { + uint64_t offset = 0; + offset = real_lseek64 (fd, 0L, SEEK_CUR); + + if (((int64_t) offset) != -1) { + ret = glusterfs_lseek (glfs_fd, offset, SEEK_SET); + if (ret != -1) { + ret = glusterfs_writev (glfs_fd, vector, count); + } + } else { + ret = -1; + } + +/* ret = glusterfs_writev (glfs_fd, vector, count); */ + if (ret == -1) { + ret = real_writev (fd, vector, count); + } + + if (ret > 0 && ((int64_t)offset) >= 0) { + real_lseek64 (fd, offset + ret, SEEK_SET); + } + + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} + + +ssize_t +pwrite (int fd, const void *buf, size_t count, unsigned long offset) +{ + int ret; + long glfs_fd = 0; + + assert (real_pwrite != NULL); + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + + if (!glfs_fd) { + ret = real_pwrite (fd, buf, count, offset); + } else { + ret = glusterfs_pwrite (glfs_fd, buf, count, offset); + if (ret == -1) { + ret = real_pwrite (fd, buf, count, offset); + } + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} + + +ssize_t +pwrite64 (int fd, const void *buf, size_t count, uint64_t offset) +{ + int ret; + long glfs_fd = 0; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + + if (!glfs_fd) { + ret = real_pwrite64 (fd, buf, count, offset); + } else { + ret = glusterfs_pwrite (glfs_fd, buf, count, offset); + if (ret == -1) { + ret = real_pwrite64 (fd, buf, count, offset); + } + } + + return ret; +} + + +int +close (int fd) +{ + int ret = -1; + long glfs_fd = 0; +/* struct stat st = {0,}; */ + +/* ret = fstat (fd, &st); + if (ret != -1) { + libglusterfs_handle_t handle = 0; + handle = booster_get_handle (booster_mount_table, st.st_dev); + if (handle) { */ + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, fd); + + if (glfs_fd) { + booster_put_fd (booster_glfs_fdtable, fd); + ret = glusterfs_close (glfs_fd); + booster_put_glfs_fd (glfs_fd); + } +/*} + }*/ + + ret = real_close (fd); + + return ret; +} + +#ifndef _LSEEK_DECLARED +#define _LSEEK_DECLARED +off_t +lseek (int filedes, unsigned long offset, int whence) +{ + int ret; + long glfs_fd = 0; + + ret = real_lseek (filedes, offset, whence); + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, filedes); + if (glfs_fd) { + ret = glusterfs_lseek (glfs_fd, offset, whence); + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} +#endif + +off_t +lseek64 (int filedes, uint64_t offset, int whence) +{ + int ret; + long glfs_fd = 0; + + ret = real_lseek64 (filedes, offset, whence); + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, filedes); + if (glfs_fd) { + ret = glusterfs_lseek (glfs_fd, offset, whence); + booster_put_glfs_fd (glfs_fd); + } + + return ret; +} + +int +dup (int oldfd) +{ + int ret = -1, new_fd = -1; + long glfs_fd = 0; + + glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, oldfd); + new_fd = real_dup (oldfd); + + if (new_fd >=0 && glfs_fd) { + ret = booster_get_unused_fd (booster_glfs_fdtable, glfs_fd, new_fd); + fd_ref ((fd_t *)glfs_fd); + if (ret == -1) { + real_close (new_fd); + } + } + + if (glfs_fd) { + booster_put_glfs_fd (glfs_fd); + } + + return new_fd; +} + + +int +dup2 (int oldfd, int newfd) +{ + int ret = -1; + long old_glfs_fd = 0, new_glfs_fd = 0; + + if (oldfd == newfd) { + return newfd; + } + + old_glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, oldfd); + new_glfs_fd = booster_get_glfs_fd (booster_glfs_fdtable, newfd); + + ret = real_dup2 (oldfd, newfd); + if (ret >= 0) { + if (new_glfs_fd) { + glusterfs_close (new_glfs_fd); + booster_put_glfs_fd (new_glfs_fd); + booster_put_fd (booster_glfs_fdtable, newfd); + new_glfs_fd = 0; + } + + if (old_glfs_fd) { + ret = booster_get_unused_fd (booster_glfs_fdtable, old_glfs_fd, newfd); + fd_ref ((fd_t *)old_glfs_fd); + if (ret == -1) { + real_close (newfd); + } + } + } + + if (old_glfs_fd) { + booster_put_glfs_fd (old_glfs_fd); + } + + if (new_glfs_fd) { + booster_put_glfs_fd (new_glfs_fd); + } + + return ret; +} + + +#define MOUNT_TABLE_HASH_SIZE 256 + + +static int +booster_init (void) +{ + int i = 0; + booster_glfs_fdtable = gf_fd_fdtable_alloc (); + if (!booster_glfs_fdtable) { + fprintf (stderr, "cannot allocate fdtable: %s\n", strerror (errno)); + goto err; + } + + booster_mount_table = calloc (1, sizeof (*booster_mount_table)); + if (!booster_mount_table) { + fprintf (stderr, "cannot allocate memory: %s\n", strerror (errno)); + goto err; + } + + pthread_mutex_init (&booster_mount_table->lock, NULL); + booster_mount_table->hash_size = MOUNT_TABLE_HASH_SIZE; + booster_mount_table->mounts = calloc (booster_mount_table->hash_size, sizeof (*booster_mount_table->mounts)); + if (!booster_mount_table->mounts) { + fprintf (stderr, "cannot allocate memory: %s\n", strerror (errno)); + goto err; + } + + for (i = 0; i < booster_mount_table->hash_size; i++) + { + INIT_LIST_HEAD (&booster_mount_table->mounts[i]); + } + + return 0; + +err: + if (booster_glfs_fdtable) { + gf_fd_fdtable_destroy (booster_glfs_fdtable); + booster_glfs_fdtable = NULL; + } + + if (booster_mount_table) { + if (booster_mount_table->mounts) { + free (booster_mount_table->mounts); + } + + free (booster_mount_table); + booster_mount_table = NULL; + } + return -1; +} + + +static void +booster_cleanup (void) +{ + int i; + booster_mount_t *mount = NULL, *tmp = NULL; + + /* gf_fd_fdtable_destroy (booster_glfs_fdtable);*/ + /*for (i=0; i < booster_glfs_fdtable->max_fds; i++) { + if (booster_glfs_fdtable->fds[i]) { + fd_t *fd = booster_glfs_fdtable->fds[i]; + free (fd); + } + }*/ + + free (booster_glfs_fdtable); + booster_glfs_fdtable = NULL; + + pthread_mutex_lock (&booster_mount_table->lock); + { + for (i = 0; i < booster_mount_table->hash_size; i++) + { + list_for_each_entry_safe (mount, tmp, + &booster_mount_table->mounts[i], device_list) { + list_del (&mount->device_list); + glusterfs_fini (mount->handle); + free (mount); + } + } + free (booster_mount_table->mounts); + } + pthread_mutex_unlock (&booster_mount_table->lock); + + glusterfs_reset (); + free (booster_mount_table); + booster_mount_table = NULL; +} + + + +pid_t +fork (void) +{ + pid_t pid = 0; + char child = 0; + + glusterfs_log_lock (); + { + pid = real_fork (); + } + glusterfs_log_unlock (); + + child = (pid == 0); + if (child) { + booster_cleanup (); + booster_init (); + } + + return pid; +} + + +void +_init (void) +{ + booster_init (); + + RESOLVE (open); + RESOLVE (open64); + RESOLVE (creat); + + RESOLVE (read); + RESOLVE (readv); + RESOLVE (pread); + RESOLVE (pread64); + + RESOLVE (write); + RESOLVE (writev); + RESOLVE (pwrite); + RESOLVE (pwrite64); + + RESOLVE (lseek); + RESOLVE (lseek64); + + RESOLVE (close); + + RESOLVE (dup); + RESOLVE (dup2); + + RESOLVE (fork); +} + -- cgit