/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
* vim:expandtab:shiftwidth=8:tabstop=8:
*
- * lustre/lib/fsfilt_ext3.c
+ * lustre/lib/lvfs_linux.c
* Lustre filesystem abstraction routines
*
* Copyright (C) 2002, 2003 Cluster File Systems, Inc.
#include <linux/version.h>
#include <linux/fs.h>
#include <asm/unistd.h>
-#include <linux/jbd.h>
#include <linux/slab.h>
#include <linux/pagemap.h>
#include <linux/quotaops.h>
#include <linux/version.h>
-#include <linux/kp30.h>
-#include <linux/lustre_fsfilt.h>
-#include <linux/obd.h>
-#include <linux/obd_class.h>
+#include <libcfs/kp30.h>
+#include <lustre_fsfilt.h>
+#include <obd.h>
+#include <obd_class.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/lustre_compat25.h>
-#include <linux/lvfs.h>
+#include <lvfs.h>
#include "lvfs_internal.h"
-#include <linux/obd.h>
-#include <linux/lustre_lib.h>
+#include <obd.h>
+#include <lustre_lib.h>
+#include <lustre_quota.h>
atomic_t obd_memory;
int obd_memmax;
-
/* Debugging check only needed during development */
#ifdef OBD_CTXT_DEBUG
# define ASSERT_CTXT_MAGIC(magic) LASSERT((magic) == OBD_RUN_CTXT_MAGIC)
-# define ASSERT_NOT_KERNEL_CTXT(msg) LASSERT(!segment_eq(get_fs(), get_ds()))
-# define ASSERT_KERNEL_CTXT(msg) LASSERT(segment_eq(get_fs(), get_ds()))
+# define ASSERT_NOT_KERNEL_CTXT(msg) LASSERTF(!segment_eq(get_fs(), get_ds()),\
+ msg)
+# define ASSERT_KERNEL_CTXT(msg) LASSERTF(segment_eq(get_fs(), get_ds()), msg)
#else
# define ASSERT_CTXT_MAGIC(magic) do {} while(0)
# define ASSERT_NOT_KERNEL_CTXT(msg) do {} while(0)
# define ASSERT_KERNEL_CTXT(msg) do {} while(0)
#endif
-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,0))
-#define current_ngroups current->group_info->ngroups
-#define current_groups current->group_info->small_block
+static void push_group_info(struct lvfs_run_ctxt *save,
+ struct upcall_cache_entry *uce)
+{
+ 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
-#define current_ngroups current->ngroups
-#define current_groups current->groups
+ 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 = 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
+ }
+}
/* push / pop to root of obd store */
void push_ctxt(struct lvfs_run_ctxt *save, struct lvfs_run_ctxt *new_ctx,
/*
CDEBUG(D_INFO,
- "= push %p->%p = cur fs %p pwd %p:d%d:i%d (%*s), pwdmnt %p:%d\n",
+ "= 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),
LASSERT(atomic_read(&new_ctx->pwd->d_count));
save->pwd = dget(current->fs->pwd);
save->pwdmnt = mntget(current->fs->pwdmnt);
- save->ngroups = current_ngroups;
+ save->luc.luc_umask = current->fs->umask;
LASSERT(save->pwd);
LASSERT(save->pwdmnt);
save->luc.luc_fsuid = current->fsuid;
save->luc.luc_fsgid = current->fsgid;
save->luc.luc_cap = current->cap_effective;
- save->luc.luc_suppgid1 = current_groups[0];
- save->luc.luc_suppgid2 = current_groups[1];
current->fsuid = uc->luc_fsuid;
current->fsgid = uc->luc_fsgid;
current->cap_effective = uc->luc_cap;
- current_ngroups = 0;
-
- if (uc->luc_suppgid1 != -1)
- current_groups[current_ngroups++] = uc->luc_suppgid1;
- if (uc->luc_suppgid2 != -1)
- current_groups[current_ngroups++] = uc->luc_suppgid2;
+ push_group_info(save, uc->luc_uce);
}
+ current->fs->umask = 0; /* umask already applied on client */
set_fs(new_ctx->fs);
- set_fs_pwd(current->fs, new_ctx->pwdmnt, new_ctx->pwd);
+ 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",
+ "= 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),
/*
CDEBUG(D_INFO,
- " = pop %p==%p = cur %p pwd %p:d%d:i%d (%*s), pwdmnt %p:%d\n",
+ " = 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),
atomic_read(¤t->fs->pwdmnt->mnt_count));
*/
- LASSERT(current->fs->pwd == new_ctx->pwd);
- LASSERT(current->fs->pwdmnt == new_ctx->pwdmnt);
+ 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);
set_fs(saved->fs);
- set_fs_pwd(current->fs, saved->pwdmnt, saved->pwd);
+ ll_set_fs_pwd(current->fs, saved->pwdmnt, saved->pwd);
dput(saved->pwd);
mntput(saved->pwdmnt);
+ current->fs->umask = saved->luc.luc_umask;
if (uc) {
current->fsuid = saved->luc.luc_fsuid;
current->fsgid = saved->luc.luc_fsgid;
current->cap_effective = saved->luc.luc_cap;
- current_ngroups = saved->ngroups;
- current_groups[0] = saved->luc.luc_suppgid1;
- current_groups[1] = saved->luc.luc_suppgid2;
+ pop_group_info(saved, uc->luc_uce);
}
/*
CDEBUG(D_INFO,
- "= pop %p->%p = cur fs %p pwd %p:d%d:i%d (%*s), pwdmnt %p:%d\n",
+ "= 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),
ENTRY;
ASSERT_KERNEL_CTXT("kernel doing mknod outside kernel context\n");
- CDEBUG(D_INODE, "creating file %*s\n", (int)strlen(name), name);
+ CDEBUG(D_INODE, "creating file %.*s\n", (int)strlen(name), name);
dchild = ll_lookup_one_len(name, dir, strlen(name));
if (IS_ERR(dchild))
ENTRY;
ASSERT_KERNEL_CTXT("kernel doing mkdir outside kernel context\n");
- CDEBUG(D_INODE, "creating directory %*s\n", (int)strlen(name), name);
+ CDEBUG(D_INODE, "creating directory %.*s\n", (int)strlen(name), name);
dchild = ll_lookup_one_len(name, dir, strlen(name));
if (IS_ERR(dchild))
GOTO(out_up, dchild);
if (dchild->d_inode) {
int old_mode = dchild->d_inode->i_mode;
- if (!S_ISDIR(old_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_err, err = -ENOTDIR);
+ }
/* Fixup directory permissions if necessary */
if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) {
- CWARN("fixing permissions on %s from %o to %o\n",
- name, old_mode, mode);
+ CDEBUG(D_CONFIG,
+ "fixing permissions on %s from %o to %o\n",
+ name, old_mode, mode);
dchild->d_inode->i_mode = (mode & S_IALLUGO) |
(old_mode & ~S_IALLUGO);
mark_inode_dirty(dchild->d_inode);
}
EXPORT_SYMBOL(simple_mkdir);
+/* utility to rename a file */
+int lustre_rename(struct dentry *dir, 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",
+ (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)
+ 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);
+
+ dput(dchild_new);
+put_old:
+ dput(dchild_old);
+ RETURN(err);
+}
+EXPORT_SYMBOL(lustre_rename);
+
/*
* Read a file from within kernel context. Prior to calling this
* function we should already have done a push_ctxt().
}
EXPORT_SYMBOL(l_dentry_open);
+#ifdef HAVE_VFS_READDIR_U64_INO
+static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset,
+ u64 ino, unsigned int d_type)
+#else
static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset,
ino_t ino, unsigned int d_type)
+#endif
{
struct l_linux_dirent *dirent;
struct l_readdir_callback *buf = (struct l_readdir_callback *)__buf;
-
+
dirent = buf->lrc_dirent;
if (dirent)
- dirent->lld_off = offset;
+ dirent->lld_off = offset;
OBD_ALLOC(dirent, sizeof(*dirent));
+ if (!dirent)
+ return -ENOMEM;
+
list_add_tail(&dirent->lld_list, buf->lrc_list);
buf->lrc_dirent = dirent;
EXPORT_SYMBOL(obd_memory);
EXPORT_SYMBOL(obd_memmax);
+#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);
+ dev_set_rdonly(jdev);
+ }
+ CDEBUG(D_IOCTL | D_HA, "set dev %lx rdonly\n", (long)dev);
+ dev_set_rdonly(dev);
+}
+
+int lvfs_check_rdonly(lvfs_sbdev_type dev)
+{
+ return dev_check_rdonly(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)
+{
+ char *write_page = NULL;
+ loff_t offset = 0;
+ int rc = 0;
+ ENTRY;
+
+ 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);
+}
+EXPORT_SYMBOL(lvfs_check_io_health);
+
static int __init lvfs_linux_init(void)
{
RETURN(0);
CDEBUG(leaked ? D_ERROR : D_INFO,
"obd mem max: %d leaked: %d\n", obd_memmax, leaked);
+ EXIT;
return;
}