#include <linux/quotaops.h>
#include <linux/version.h>
#include <libcfs/kp30.h>
-#include <linux/lustre_fsfilt.h>
-#include <linux/obd.h>
-#include <linux/obd_class.h>
+#include <lustre_fsfilt.h>
+#include <obd.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 <linux/lustre_mds.h> /* for mds_grp_hash_entry */
+#include <obd.h>
+#include <lustre_lib.h>
+#include <lustre_quota.h>
-atomic_t obd_memory;
-int obd_memmax;
+__u64 obd_max_pages = 0;
+__u64 obd_max_alloc = 0;
+struct lprocfs_stats *obd_memory = NULL;
+spinlock_t obd_updatemax_lock = SPIN_LOCK_UNLOCKED;
+/* refine later and change to seqlock or simlar from libcfs */
/* Debugging check only needed during development */
#ifdef OBD_CTXT_DEBUG
# 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)
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);
- /* save old */
- save->group_info.ngroups = current->ngroups;
- if (current->ngroups)
- memcpy(save->group_info.small_block, current->groups,
- current->ngroups);
- /* push new */
- current->ngroups = ginfo->ngroups;
- if (ginfo->ngroups)
- memcpy(current->groups, ginfo->small_block,
- current->ngroups);
-#endif
}
}
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 = ginfo->ngroups;
- if (current->ngroups)
- memcpy(current->groups, save->group_info.small_block,
- current->ngroups);
-#endif
}
}
{
//ASSERT_NOT_KERNEL_CTXT("already in kernel context!\n");
ASSERT_CTXT_MAGIC(new_ctx->magic);
- LASSERT(save->magic != OBD_RUN_CTXT_MAGIC || save->pid != current->pid);
OBD_SET_CTXT_MAGIC(save);
- save->pid = current->pid;
/*
CDEBUG(D_INFO,
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;
+ save->ngroups = current->group_info->ngroups;
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_nid = current->user->nid;
-
+
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;
- current->user->nid = uc->luc_nid;
- push_group_info(save, uc->luc_ginfo);
+ 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);
- 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,
{
//printk("pc0");
ASSERT_CTXT_MAGIC(saved->magic);
- LASSERT(saved->pid == current->pid);
- saved->magic = 0;
- saved->pid = 0;
//printk("pc1");
ASSERT_KERNEL_CTXT("popping non-kernel context!\n");
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->fsuid = saved->luc.luc_fsuid;
current->fsgid = saved->luc.luc_fsgid;
current->cap_effective = saved->luc.luc_cap;
- current->user->nid = saved->luc.luc_nid;
- pop_group_info(saved, uc->luc_ginfo);
+ pop_group_info(saved,
+ uc->luc_ginfo ?:
+ uc->luc_identity ? uc->luc_identity->mi_ginfo :
+ NULL);
}
/*
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));
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))
/* 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;
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)
if (lastdirent)
lastdirent->lld_off = file->f_pos;
- return 0;
+ return 0;
}
EXPORT_SYMBOL(l_readdir);
-EXPORT_SYMBOL(obd_memory);
-EXPORT_SYMBOL(obd_memmax);
#if defined (CONFIG_DEBUG_MEMORY) && defined(__KERNEL__)
static spinlock_t obd_memlist_lock = SPIN_LOCK_UNLOCKED;
-static struct hlist_head *obd_memtable;
-static unsigned long obd_memtable_size;
+static struct hlist_head *obd_memtable = NULL;
+static unsigned long obd_memtable_size = 0;
static int lvfs_memdbg_init(int size)
{
LASSERT(size > sizeof(sizeof(struct hlist_head)));
obd_memtable_size = size / sizeof(struct hlist_head);
- CWARN("allocating %lu malloc entries...\n",
+ CWARN("Allocating %lu memdbg entries.\n",
(unsigned long)obd_memtable_size);
+ LASSERT(obd_memtable == NULL);
obd_memtable = kmalloc(size, GFP_KERNEL);
if (!obd_memtable)
return -ENOMEM;
{
struct hlist_node *node = NULL, *tmp = NULL;
struct hlist_head *head;
- struct mem_track *mt;
+ struct obd_mem_track *mt;
int i;
spin_lock(&obd_memlist_lock);
for (i = 0, head = obd_memtable; i < obd_memtable_size; i++, head++) {
hlist_for_each_safe(node, tmp, head) {
- mt = hlist_entry(node, struct mem_track, m_hash);
- hlist_del_init(&mt->m_hash);
+ mt = hlist_entry(node, struct obd_mem_track, mt_hash);
+ hlist_del_init(&mt->mt_hash);
kfree(mt);
}
}
(obd_memtable_size - 1);
}
-static void __lvfs_memdbg_insert(struct mem_track *mt)
+static void __lvfs_memdbg_insert(struct obd_mem_track *mt)
{
struct hlist_head *head = obd_memtable +
- hashfn(mt->m_ptr);
- hlist_add_head(&mt->m_hash, head);
+ hashfn(mt->mt_ptr);
+ hlist_add_head(&mt->mt_hash, head);
}
-void lvfs_memdbg_insert(struct mem_track *mt)
+void lvfs_memdbg_insert(struct obd_mem_track *mt)
{
spin_lock(&obd_memlist_lock);
__lvfs_memdbg_insert(mt);
}
EXPORT_SYMBOL(lvfs_memdbg_insert);
-static void __lvfs_memdbg_remove(struct mem_track *mt)
+static void __lvfs_memdbg_remove(struct obd_mem_track *mt)
{
- hlist_del_init(&mt->m_hash);
+ hlist_del_init(&mt->mt_hash);
}
-void lvfs_memdbg_remove(struct mem_track *mt)
+void lvfs_memdbg_remove(struct obd_mem_track *mt)
{
spin_lock(&obd_memlist_lock);
__lvfs_memdbg_remove(mt);
}
EXPORT_SYMBOL(lvfs_memdbg_remove);
-static struct mem_track *__lvfs_memdbg_find(void *ptr)
+static struct obd_mem_track *__lvfs_memdbg_find(void *ptr)
{
struct hlist_node *node = NULL;
- struct mem_track *mt = NULL;
+ struct obd_mem_track *mt = NULL;
struct hlist_head *head;
head = obd_memtable + hashfn(ptr);
hlist_for_each(node, head) {
- mt = hlist_entry(node, struct mem_track, m_hash);
- if ((unsigned long)mt->m_ptr == (unsigned long)ptr)
+ mt = hlist_entry(node, struct obd_mem_track, mt_hash);
+ if ((unsigned long)mt->mt_ptr == (unsigned long)ptr)
break;
mt = NULL;
}
return mt;
}
-struct mem_track *lvfs_memdbg_find(void *ptr)
+struct obd_mem_track *lvfs_memdbg_find(void *ptr)
{
- struct mem_track *mt;
+ struct obd_mem_track *mt;
spin_lock(&obd_memlist_lock);
mt = __lvfs_memdbg_find(ptr);
}
EXPORT_SYMBOL(lvfs_memdbg_find);
-int lvfs_memdbg_check_insert(struct mem_track *mt)
+int lvfs_memdbg_check_insert(struct obd_mem_track *mt)
{
+ struct obd_mem_track *tmp;
+
spin_lock(&obd_memlist_lock);
- if (!__lvfs_memdbg_find(mt->m_ptr)) {
+ tmp = __lvfs_memdbg_find(mt->mt_ptr);
+ if (tmp == NULL) {
__lvfs_memdbg_insert(mt);
spin_unlock(&obd_memlist_lock);
return 1;
}
EXPORT_SYMBOL(lvfs_memdbg_check_insert);
-struct mem_track *
+struct obd_mem_track *
lvfs_memdbg_check_remove(void *ptr)
{
- struct mem_track *mt;
+ struct obd_mem_track *mt;
spin_lock(&obd_memlist_lock);
mt = __lvfs_memdbg_find(ptr);
#if defined (CONFIG_DEBUG_MEMORY) && defined(__KERNEL__)
struct hlist_node *node = NULL;
struct hlist_head *head;
- struct mem_track *mt;
+ struct obd_mem_track *mt;
+ int header = 0;
#endif
- int leaked;
#if defined (CONFIG_DEBUG_MEMORY) && defined(__KERNEL__)
int i;
#endif
- leaked = atomic_read(&obd_memory);
-
- if (leaked > 0) {
- CWARN("memory leaks detected (max %d, leaked %d)\n",
- obd_memmax, leaked);
-
+
#if defined (CONFIG_DEBUG_MEMORY) && defined(__KERNEL__)
- spin_lock(&obd_memlist_lock);
- for (i = 0, head = obd_memtable; i < obd_memtable_size; i++, head++) {
- hlist_for_each(node, head) {
- mt = hlist_entry(node, struct mem_track, m_hash);
- CWARN(" [%s] ptr: 0x%p, size: %d, src at \"%s\"\n",
- ((mt->m_flags & MT_FLAGS_WRONG_SIZE) ?
- "wrong ck size" : "leaked memory"),
- mt->m_ptr, mt->m_size, mt->m_loc);
+ spin_lock(&obd_memlist_lock);
+ for (i = 0, head = obd_memtable; i < obd_memtable_size; i++, head++) {
+ hlist_for_each(node, head) {
+ if (header == 0) {
+ CWARN("Abnormal memory activities:\n");
+ header = 1;
}
+ mt = hlist_entry(node, struct obd_mem_track, mt_hash);
+ CWARN(" [%s] ptr: 0x%p, size: %d, src at %s\n",
+ ((mt->mt_flags & OBD_MT_WRONG_SIZE) ?
+ "wrong size" : "leaked memory"),
+ mt->mt_ptr, mt->mt_size, mt->mt_loc);
}
- spin_unlock(&obd_memlist_lock);
-#endif
- /* remove for production */
- portals_debug_dumplog();
}
+ spin_unlock(&obd_memlist_lock);
+#endif
}
EXPORT_SYMBOL(lvfs_memdbg_show);
+#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);
+
+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);
+#endif /* LUSTRE_KERNEL_VERSION */
+
+void obd_update_maxusage()
+{
+ __u64 max1, max2;
+
+ max1 = obd_pages_sum();
+ max2 = obd_memory_sum();
+
+ spin_lock(&obd_updatemax_lock);
+ if (max1 > obd_max_pages)
+ obd_max_pages = max1;
+ if (max2 > obd_max_alloc)
+ obd_max_alloc = max2;
+ spin_unlock(&obd_updatemax_lock);
+
+}
+
+__u64 obd_memory_max(void)
+{
+ __u64 ret;
+
+ spin_lock(&obd_updatemax_lock);
+ ret = obd_max_alloc;
+ spin_unlock(&obd_updatemax_lock);
+
+ return ret;
+}
+
+__u64 obd_pages_max(void)
+{
+ __u64 ret;
+
+ spin_lock(&obd_updatemax_lock);
+ ret = obd_max_pages;
+ 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 = 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;
+ 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 != atomic_read(&lc->lc_cntl.la_entry) &&
+ centry != atomic_read(&lc->lc_cntl.la_exit));
+
+ RETURN(ret);
+}
+EXPORT_SYMBOL(lprocfs_read_helper);
+#endif /* LPROCFS */
+
static int __init lvfs_linux_init(void)
{
ENTRY;
#if defined (CONFIG_DEBUG_MEMORY) && defined(__KERNEL__)
lvfs_memdbg_init(PAGE_SIZE);
#endif
- lvfs_mount_list_init();
RETURN(0);
}
{
ENTRY;
- lvfs_mount_list_cleanup();
lvfs_memdbg_show();
-
+
#if defined (CONFIG_DEBUG_MEMORY) && defined(__KERNEL__)
lvfs_memdbg_cleanup();
#endif
EXIT;
- return;
}
MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");