X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;ds=sidebyside;f=lustre%2Flvfs%2Flvfs_linux.c;h=7e010513133de29fc7481d372880c7b5365fec22;hb=d031e92fe730792e3a4dba2f1e8ae90a085c96c5;hp=53147a4a9f0f69fd815a7227021cab8011c0d24e;hpb=492125c2d1e3e77c5c2ec31e888ce3eb070f2120;p=fs%2Flustre-release.git diff --git a/lustre/lvfs/lvfs_linux.c b/lustre/lvfs/lvfs_linux.c index 53147a4..7e01051 100644 --- a/lustre/lvfs/lvfs_linux.c +++ b/lustre/lvfs/lvfs_linux.c @@ -1,26 +1,41 @@ /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- * vim:expandtab:shiftwidth=8:tabstop=8: * - * lustre/lib/lvfs_linux.c - * Lustre filesystem abstraction routines + * GPL HEADER START * - * Copyright (C) 2002, 2003 Cluster File Systems, Inc. - * Author: Andreas Dilger + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. * - * This file is part of Lustre, http://www.lustre.org. + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. * - * Lustre is free software; you can redistribute it and/or - * modify it under the terms of version 2 of the GNU General Public - * License as published by the Free Software Foundation. + * This program 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 version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). * - * Lustre 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 + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf * - * You should have received a copy of the GNU General Public License - * along with Lustre; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * lustre/lvfs/lvfs_linux.c + * + * Author: Andreas Dilger */ #ifndef EXPORT_SYMTAB @@ -36,10 +51,9 @@ #include #include #include -#include +#include #include #include -#include #include #include #include @@ -50,8 +64,11 @@ #include #include -atomic_t obd_memory; -int obd_memmax; +__u64 obd_max_pages = 0; +__u64 obd_max_alloc = 0; +struct lprocfs_stats *obd_memory = NULL; +cfs_spinlock_t obd_updatemax_lock = CFS_SPIN_LOCK_UNLOCKED; +/* refine later and change to seqlock or simlar from libcfs */ /* Debugging check only needed during development */ #ifdef OBD_CTXT_DEBUG @@ -66,54 +83,28 @@ int obd_memmax; #endif static void push_group_info(struct lvfs_run_ctxt *save, - struct upcall_cache_entry *uce) + struct group_info *ginfo) { - struct group_info *ginfo = uce ? uce->ue_group_info : NULL; - if (!ginfo) { save->ngroups = current_ngroups; current_ngroups = 0; } else { -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,4) task_lock(current); save->group_info = current->group_info; current->group_info = ginfo; task_unlock(current); -#else - LASSERT(ginfo->ngroups <= NGROUPS); - LASSERT(current->ngroups <= NGROUPS_SMALL); - /* save old */ - save->group_info.ngroups = current->ngroups; - if (current->ngroups) - memcpy(save->group_info.small_block, current->groups, - current->ngroups * sizeof(gid_t)); - /* push new */ - current->ngroups = ginfo->ngroups; - if (ginfo->ngroups) - memcpy(current->groups, ginfo->small_block, - current->ngroups * sizeof(gid_t)); -#endif } } static void pop_group_info(struct lvfs_run_ctxt *save, - struct upcall_cache_entry *uce) + struct group_info *ginfo) { - struct group_info *ginfo = uce ? uce->ue_group_info : NULL; - if (!ginfo) { current_ngroups = save->ngroups; } else { -#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,4) task_lock(current); current->group_info = save->group_info; task_unlock(current); -#else - current->ngroups = save->group_info.ngroups; - if (current->ngroups) - memcpy(current->groups, save->group_info.small_block, - current->ngroups * sizeof(gid_t)); -#endif } } @@ -125,23 +116,13 @@ void push_ctxt(struct lvfs_run_ctxt *save, struct lvfs_run_ctxt *new_ctx, ASSERT_CTXT_MAGIC(new_ctx->magic); OBD_SET_CTXT_MAGIC(save); - /* - CDEBUG(D_INFO, - "= push %p->%p = cur fs %p pwd %p:d%d:i%d (%.*s), pwdmnt %p:%d\n", - save, current, current->fs, current->fs->pwd, - atomic_read(¤t->fs->pwd->d_count), - atomic_read(¤t->fs->pwd->d_inode->i_count), - current->fs->pwd->d_name.len, current->fs->pwd->d_name.name, - current->fs->pwdmnt, - atomic_read(¤t->fs->pwdmnt->mnt_count)); - */ - save->fs = get_fs(); - LASSERT(atomic_read(¤t->fs->pwd->d_count)); - LASSERT(atomic_read(&new_ctx->pwd->d_count)); - save->pwd = dget(current->fs->pwd); - save->pwdmnt = mntget(current->fs->pwdmnt); + LASSERT(cfs_atomic_read(&cfs_fs_pwd(current->fs)->d_count)); + LASSERT(cfs_atomic_read(&new_ctx->pwd->d_count)); + save->pwd = dget(cfs_fs_pwd(current->fs)); + save->pwdmnt = mntget(cfs_fs_mnt(current->fs)); save->luc.luc_umask = current->fs->umask; + save->ngroups = current->group_info->ngroups; LASSERT(save->pwd); LASSERT(save->pwdmnt); @@ -149,55 +130,39 @@ void push_ctxt(struct lvfs_run_ctxt *save, struct lvfs_run_ctxt *new_ctx, LASSERT(new_ctx->pwdmnt); if (uc) { + save->luc.luc_uid = current->uid; + save->luc.luc_gid = current->gid; save->luc.luc_fsuid = current->fsuid; save->luc.luc_fsgid = current->fsgid; save->luc.luc_cap = current->cap_effective; + current->uid = uc->luc_uid; + current->gid = uc->luc_gid; current->fsuid = uc->luc_fsuid; current->fsgid = uc->luc_fsgid; current->cap_effective = uc->luc_cap; - push_group_info(save, uc->luc_uce); + + push_group_info(save, + uc->luc_ginfo ?: + uc->luc_identity ? uc->luc_identity->mi_ginfo : + NULL); } current->fs->umask = 0; /* umask already applied on client */ set_fs(new_ctx->fs); ll_set_fs_pwd(current->fs, new_ctx->pwdmnt, new_ctx->pwd); - - /* - CDEBUG(D_INFO, - "= push %p->%p = cur fs %p pwd %p:d%d:i%d (%.*s), pwdmnt %p:%d\n", - new_ctx, current, current->fs, current->fs->pwd, - atomic_read(¤t->fs->pwd->d_count), - atomic_read(¤t->fs->pwd->d_inode->i_count), - current->fs->pwd->d_name.len, current->fs->pwd->d_name.name, - current->fs->pwdmnt, - atomic_read(¤t->fs->pwdmnt->mnt_count)); - */ } EXPORT_SYMBOL(push_ctxt); void pop_ctxt(struct lvfs_run_ctxt *saved, struct lvfs_run_ctxt *new_ctx, struct lvfs_ucred *uc) { - //printk("pc0"); ASSERT_CTXT_MAGIC(saved->magic); - //printk("pc1"); ASSERT_KERNEL_CTXT("popping non-kernel context!\n"); - /* - CDEBUG(D_INFO, - " = pop %p==%p = cur %p pwd %p:d%d:i%d (%.*s), pwdmnt %p:%d\n", - new_ctx, current, current->fs, current->fs->pwd, - atomic_read(¤t->fs->pwd->d_count), - atomic_read(¤t->fs->pwd->d_inode->i_count), - current->fs->pwd->d_name.len, current->fs->pwd->d_name.name, - current->fs->pwdmnt, - atomic_read(¤t->fs->pwdmnt->mnt_count)); - */ - - LASSERTF(current->fs->pwd == new_ctx->pwd, "%p != %p\n", - current->fs->pwd, new_ctx->pwd); - LASSERTF(current->fs->pwdmnt == new_ctx->pwdmnt, "%p != %p\n", - current->fs->pwdmnt, new_ctx->pwdmnt); + LASSERTF(cfs_fs_pwd(current->fs) == new_ctx->pwd, "%p != %p\n", + cfs_fs_pwd(current->fs), new_ctx->pwd); + LASSERTF(cfs_fs_mnt(current->fs) == new_ctx->pwdmnt, "%p != %p\n", + cfs_fs_mnt(current->fs), new_ctx->pwdmnt); set_fs(saved->fs); ll_set_fs_pwd(current->fs, saved->pwdmnt, saved->pwd); @@ -206,22 +171,16 @@ void pop_ctxt(struct lvfs_run_ctxt *saved, struct lvfs_run_ctxt *new_ctx, mntput(saved->pwdmnt); current->fs->umask = saved->luc.luc_umask; if (uc) { + current->uid = saved->luc.luc_uid; + current->gid = saved->luc.luc_gid; current->fsuid = saved->luc.luc_fsuid; current->fsgid = saved->luc.luc_fsgid; current->cap_effective = saved->luc.luc_cap; - pop_group_info(saved, uc->luc_uce); + pop_group_info(saved, + uc->luc_ginfo ?: + uc->luc_identity ? uc->luc_identity->mi_ginfo : + NULL); } - - /* - CDEBUG(D_INFO, - "= pop %p->%p = cur fs %p pwd %p:d%d:i%d (%.*s), pwdmnt %p:%d\n", - saved, current, current->fs, current->fs->pwd, - atomic_read(¤t->fs->pwd->d_count), - atomic_read(¤t->fs->pwd->d_inode->i_count), - current->fs->pwd->d_name.len, current->fs->pwd->d_name.name, - current->fs->pwdmnt, - atomic_read(¤t->fs->pwdmnt->mnt_count)); - */ } EXPORT_SYMBOL(pop_ctxt); @@ -232,7 +191,7 @@ struct dentry *simple_mknod(struct dentry *dir, char *name, int mode, int fix) int err = 0; ENTRY; - ASSERT_KERNEL_CTXT("kernel doing mknod outside kernel context\n"); + // ASSERT_KERNEL_CTXT("kernel doing mknod outside kernel context\n"); CDEBUG(D_INODE, "creating file %.*s\n", (int)strlen(name), name); dchild = ll_lookup_one_len(name, dir, strlen(name)); @@ -271,13 +230,14 @@ out_up: EXPORT_SYMBOL(simple_mknod); /* utility to make a directory */ -struct dentry *simple_mkdir(struct dentry *dir, char *name, int mode, int fix) +struct dentry *simple_mkdir(struct dentry *dir, struct vfsmount *mnt, + const char *name, int mode, int fix) { struct dentry *dchild; int err = 0; ENTRY; - ASSERT_KERNEL_CTXT("kernel doing mkdir outside kernel context\n"); + // ASSERT_KERNEL_CTXT("kernel doing mkdir outside kernel context\n"); CDEBUG(D_INODE, "creating directory %.*s\n", (int)strlen(name), name); dchild = ll_lookup_one_len(name, dir, strlen(name)); if (IS_ERR(dchild)) @@ -294,7 +254,7 @@ struct dentry *simple_mkdir(struct dentry *dir, char *name, int mode, int fix) /* Fixup directory permissions if necessary */ if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) { - CDEBUG(D_CONFIG, + CDEBUG(D_CONFIG, "fixing permissions on %s from %o to %o\n", name, old_mode, mode); dchild->d_inode->i_mode = (mode & S_IALLUGO) | @@ -304,7 +264,7 @@ struct dentry *simple_mkdir(struct dentry *dir, char *name, int mode, int fix) GOTO(out_up, dchild); } - err = vfs_mkdir(dir->d_inode, dchild, mode); + err = ll_vfs_mkdir(dir->d_inode, dchild, mnt, mode); if (err) GOTO(out_err, err); @@ -319,28 +279,30 @@ out_up: EXPORT_SYMBOL(simple_mkdir); /* utility to rename a file */ -int lustre_rename(struct dentry *dir, char *oldname, char *newname) +int lustre_rename(struct dentry *dir, struct vfsmount *mnt, + char *oldname, char *newname) { struct dentry *dchild_old, *dchild_new; int err = 0; ENTRY; ASSERT_KERNEL_CTXT("kernel doing rename outside kernel context\n"); - CDEBUG(D_INODE, "renaming file %.*s to %.*s\n", + CDEBUG(D_INODE, "renaming file %.*s to %.*s\n", (int)strlen(oldname), oldname, (int)strlen(newname), newname); dchild_old = ll_lookup_one_len(oldname, dir, strlen(oldname)); if (IS_ERR(dchild_old)) RETURN(PTR_ERR(dchild_old)); - if (!dchild_old->d_inode) + if (!dchild_old->d_inode) GOTO(put_old, err = -ENOENT); dchild_new = ll_lookup_one_len(newname, dir, strlen(newname)); if (IS_ERR(dchild_new)) GOTO(put_old, err = PTR_ERR(dchild_new)); - err = vfs_rename(dir->d_inode, dchild_old, dir->d_inode, dchild_new); + err = ll_vfs_rename(dir->d_inode, dchild_old, mnt, + dir->d_inode, dchild_new, mnt); dput(dchild_new); put_old: @@ -400,6 +362,7 @@ int lustre_fsync(struct file *file) } EXPORT_SYMBOL(lustre_fsync); +/* Note: dput(dchild) will be called if there is an error */ struct l_file *l_dentry_open(struct lvfs_run_ctxt *ctxt, struct l_dentry *de, int flags) { @@ -428,7 +391,7 @@ static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset, if (!dirent) return -ENOMEM; - list_add_tail(&dirent->lld_list, buf->lrc_list); + cfs_list_add_tail(&dirent->lld_list, buf->lrc_list); buf->lrc_dirent = dirent; dirent->lld_ino = ino; @@ -438,14 +401,14 @@ static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset, return 0; } -long l_readdir(struct file *file, struct list_head *dentry_list) +long l_readdir(struct file *file, cfs_list_t *dentry_list) { struct l_linux_dirent *lastdirent; struct l_readdir_callback buf; int error; buf.lrc_dirent = NULL; - buf.lrc_list = dentry_list; + buf.lrc_list = dentry_list; error = vfs_readdir(file, l_filldir, &buf); if (error < 0) @@ -455,23 +418,71 @@ long l_readdir(struct file *file, struct list_head *dentry_list) if (lastdirent) lastdirent->lld_off = file->f_pos; - return 0; + return 0; } EXPORT_SYMBOL(l_readdir); -EXPORT_SYMBOL(obd_memory); -EXPORT_SYMBOL(obd_memmax); + +int l_notify_change(struct vfsmount *mnt, struct dentry *dchild, + struct iattr *newattrs) +{ + int rc; + + LOCK_INODE_MUTEX(dchild->d_inode); +#ifdef HAVE_SECURITY_PLUG + rc = notify_change(dchild, mnt, newattrs); +#else + rc = notify_change(dchild, newattrs); +#endif + UNLOCK_INODE_MUTEX(dchild->d_inode); + return rc; +} +EXPORT_SYMBOL(l_notify_change); + +/* utility to truncate a file */ +int simple_truncate(struct dentry *dir, struct vfsmount *mnt, + char *name, loff_t length) +{ + struct dentry *dchild; + struct iattr newattrs; + int err = 0; + ENTRY; + + CDEBUG(D_INODE, "truncating file %.*s to %lld\n", (int)strlen(name), + name, (long long)length); + dchild = ll_lookup_one_len(name, dir, strlen(name)); + if (IS_ERR(dchild)) + GOTO(out, err = PTR_ERR(dchild)); + + if (dchild->d_inode) { + int old_mode = dchild->d_inode->i_mode; + if (S_ISDIR(old_mode)) { + CERROR("found %s (%lu/%u) is mode %o\n", name, + dchild->d_inode->i_ino, + dchild->d_inode->i_generation, old_mode); + GOTO(out_dput, err = -EISDIR); + } + + newattrs.ia_size = length; + newattrs.ia_valid = ATTR_SIZE; + err = l_notify_change(mnt, dchild, &newattrs); + } + EXIT; +out_dput: + dput(dchild); +out: + return err; +} +EXPORT_SYMBOL(simple_truncate); #ifdef LUSTRE_KERNEL_VERSION #ifndef HAVE_CLEAR_RDONLY_ON_PUT #error rdonly patchset must be updated [cfs bz11248] #endif - void dev_set_rdonly(lvfs_sbdev_type dev); int dev_check_rdonly(lvfs_sbdev_type dev); void __lvfs_set_rdonly(lvfs_sbdev_type dev, lvfs_sbdev_type jdev) { - lvfs_sbdev_sync(dev); if (jdev && (jdev != dev)) { CDEBUG(D_IOCTL | D_HA, "set journal dev %lx rdonly\n", (long)jdev); @@ -488,7 +499,6 @@ int lvfs_check_rdonly(lvfs_sbdev_type dev) EXPORT_SYMBOL(__lvfs_set_rdonly); EXPORT_SYMBOL(lvfs_check_rdonly); -#endif /* LUSTRE_KERNEL_VERSION */ int lvfs_check_io_health(struct obd_device *obd, struct file *file) { @@ -500,37 +510,105 @@ int lvfs_check_io_health(struct obd_device *obd, struct file *file) OBD_ALLOC(write_page, CFS_PAGE_SIZE); if (!write_page) RETURN(-ENOMEM); - + rc = fsfilt_write_record(obd, file, write_page, CFS_PAGE_SIZE, &offset, 1); - + OBD_FREE(write_page, CFS_PAGE_SIZE); CDEBUG(D_INFO, "write 1 page synchronously for checking io rc %d\n",rc); - RETURN(rc); + RETURN(rc); } EXPORT_SYMBOL(lvfs_check_io_health); +#endif /* LUSTRE_KERNEL_VERSION */ -static int __init lvfs_linux_init(void) +void obd_update_maxusage() { - RETURN(0); + __u64 max1, max2; + + max1 = obd_pages_sum(); + max2 = obd_memory_sum(); + + cfs_spin_lock(&obd_updatemax_lock); + if (max1 > obd_max_pages) + obd_max_pages = max1; + if (max2 > obd_max_alloc) + obd_max_alloc = max2; + cfs_spin_unlock(&obd_updatemax_lock); + } -static void __exit lvfs_linux_exit(void) +__u64 obd_memory_max(void) { - int leaked; - ENTRY; + __u64 ret; - leaked = atomic_read(&obd_memory); - CDEBUG(leaked ? D_ERROR : D_INFO, - "obd mem max: %d leaked: %d\n", obd_memmax, leaked); + cfs_spin_lock(&obd_updatemax_lock); + ret = obd_max_alloc; + cfs_spin_unlock(&obd_updatemax_lock); - EXIT; - return; + return ret; +} + +__u64 obd_pages_max(void) +{ + __u64 ret; + + cfs_spin_lock(&obd_updatemax_lock); + ret = obd_max_pages; + cfs_spin_unlock(&obd_updatemax_lock); + + return ret; +} + +EXPORT_SYMBOL(obd_update_maxusage); +EXPORT_SYMBOL(obd_pages_max); +EXPORT_SYMBOL(obd_memory_max); +EXPORT_SYMBOL(obd_memory); + +#ifdef LPROCFS +__s64 lprocfs_read_helper(struct lprocfs_counter *lc, + enum lprocfs_fields_flags field) +{ + __s64 ret = 0; + int centry; + + if (!lc) + RETURN(0); + do { + centry = cfs_atomic_read(&lc->lc_cntl.la_entry); + + switch (field) { + case LPROCFS_FIELDS_FLAGS_CONFIG: + ret = lc->lc_config; + break; + case LPROCFS_FIELDS_FLAGS_SUM: + ret = lc->lc_sum + lc->lc_sum_irq; + break; + case LPROCFS_FIELDS_FLAGS_MIN: + ret = lc->lc_min; + break; + case LPROCFS_FIELDS_FLAGS_MAX: + ret = lc->lc_max; + break; + case LPROCFS_FIELDS_FLAGS_AVG: + ret = (lc->lc_max - lc->lc_min)/2; + break; + case LPROCFS_FIELDS_FLAGS_SUMSQUARE: + ret = lc->lc_sumsquare; + break; + case LPROCFS_FIELDS_FLAGS_COUNT: + ret = lc->lc_count; + break; + default: + break; + }; + } while (centry != cfs_atomic_read(&lc->lc_cntl.la_entry) && + centry != cfs_atomic_read(&lc->lc_cntl.la_exit)); + + RETURN(ret); } +EXPORT_SYMBOL(lprocfs_read_helper); +#endif /* LPROCFS */ -MODULE_AUTHOR("Cluster File Systems, Inc. "); +MODULE_AUTHOR("Sun Microsystems, Inc. "); MODULE_DESCRIPTION("Lustre VFS Filesystem Helper v0.1"); MODULE_LICENSE("GPL"); - -module_init(lvfs_linux_init); -module_exit(lvfs_linux_exit);