-/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
- * vim:expandtab:shiftwidth=8:tabstop=8:
- *
+/*
* GPL HEADER START
*
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
/*
* Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
+ *
+ * Copyright (c) 2011, 2012, Whamcloud, Inc.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
cfs_mem_cache_t *ll_file_data_slab;
CFS_LIST_HEAD(ll_super_blocks);
-cfs_spinlock_t ll_sb_lock = CFS_SPIN_LOCK_UNLOCKED;
+DEFINE_SPINLOCK(ll_sb_lock);
+#ifndef MS_HAS_NEW_AOPS
extern struct address_space_operations ll_aops;
-extern struct address_space_operations ll_dir_aops;
+#else
+extern struct address_space_operations_ext ll_aops;
+#endif
#ifndef log2
#define log2(n) cfs_ffz(~(n))
static struct ll_sb_info *ll_init_sbi(void)
{
- struct ll_sb_info *sbi = NULL;
- unsigned long pages;
+ struct ll_sb_info *sbi = NULL;
+ unsigned long pages;
+ unsigned long lru_page_max;
struct sysinfo si;
class_uuid_t uuid;
int i;
RETURN(NULL);
cfs_spin_lock_init(&sbi->ll_lock);
- cfs_init_mutex(&sbi->ll_lco.lco_lock);
+ cfs_mutex_init(&sbi->ll_lco.lco_lock);
cfs_spin_lock_init(&sbi->ll_pp_extent_lock);
cfs_spin_lock_init(&sbi->ll_process_lock);
sbi->ll_rw_stats_on = 0;
pages = si.totalram - si.totalhigh;
if (pages >> (20 - CFS_PAGE_SHIFT) < 512) {
#ifdef HAVE_BGL_SUPPORT
- sbi->ll_async_page_max = pages / 4;
+ lru_page_max = pages / 4;
#else
- sbi->ll_async_page_max = pages / 2;
+ lru_page_max = pages / 2;
#endif
- } else {
- sbi->ll_async_page_max = (pages / 4) * 3;
- }
+ } else {
+ lru_page_max = (pages / 4) * 3;
+ }
+
+ /* initialize lru data */
+ cfs_atomic_set(&sbi->ll_lru.ccl_users, 0);
+ sbi->ll_lru.ccl_page_max = lru_page_max;
+ cfs_atomic_set(&sbi->ll_lru.ccl_page_left, lru_page_max);
+ cfs_spin_lock_init(&sbi->ll_lru.ccl_lock);
+ CFS_INIT_LIST_HEAD(&sbi->ll_lru.ccl_list);
sbi->ll_ra_info.ra_max_pages_per_file = min(pages / 32,
SBI_DEFAULT_READAHEAD_MAX);
cfs_list_add_tail(&sbi->ll_list, &ll_super_blocks);
cfs_spin_unlock(&ll_sb_lock);
-#ifdef ENABLE_LLITE_CHECKSUM
+ sbi->ll_flags |= LL_SBI_VERBOSE;
+#ifdef ENABLE_CHECKSUM
sbi->ll_flags |= LL_SBI_CHECKSUM;
#endif
/* metadata statahead is enabled by default */
sbi->ll_sa_max = LL_SA_RPC_DEF;
- atomic_set(&sbi->ll_sa_total, 0);
- atomic_set(&sbi->ll_sa_wrong, 0);
+ cfs_atomic_set(&sbi->ll_sa_total, 0);
+ cfs_atomic_set(&sbi->ll_sa_wrong, 0);
+ cfs_atomic_set(&sbi->ll_agl_total, 0);
+ sbi->ll_flags |= LL_SBI_AGL_ENABLED;
RETURN(sbi);
}
.d_revalidate = ll_revalidate_nd,
};
-static int client_common_fill_super(struct super_block *sb, char *md, char *dt)
+static int client_common_fill_super(struct super_block *sb, char *md, char *dt,
+ struct vfsmount *mnt)
{
struct inode *root = 0;
struct ll_sb_info *sbi = ll_s2sbi(sb);
struct obd_device *obd;
struct obd_capa *oc = NULL;
- struct obd_statfs osfs;
+ struct obd_statfs *osfs = NULL;
struct ptlrpc_request *request = NULL;
struct obd_connect_data *data = NULL;
struct obd_uuid *uuid;
if (data == NULL)
RETURN(-ENOMEM);
+ OBD_ALLOC_PTR(osfs);
+ if (osfs == NULL) {
+ OBD_FREE_PTR(data);
+ RETURN(-ENOMEM);
+ }
+
if (proc_lustre_fs_root) {
err = lprocfs_register_mountpoint(proc_lustre_fs_root, sb,
dt, md);
/* indicate the features supported by this client */
data->ocd_connect_flags = OBD_CONNECT_IBITS | OBD_CONNECT_NODEVOH |
- OBD_CONNECT_JOIN | OBD_CONNECT_ATTRFID |
- OBD_CONNECT_VERSION | OBD_CONNECT_MDS_CAPA |
- OBD_CONNECT_OSS_CAPA | OBD_CONNECT_CANCELSET|
- OBD_CONNECT_FID | OBD_CONNECT_AT |
- OBD_CONNECT_LOV_V3 | OBD_CONNECT_RMT_CLIENT |
- OBD_CONNECT_VBR | OBD_CONNECT_FULL20;
+ OBD_CONNECT_ATTRFID |
+ OBD_CONNECT_VERSION | OBD_CONNECT_BRW_SIZE |
+ OBD_CONNECT_MDS_CAPA | OBD_CONNECT_OSS_CAPA |
+ OBD_CONNECT_CANCELSET | OBD_CONNECT_FID |
+ OBD_CONNECT_AT | OBD_CONNECT_LOV_V3 |
+ OBD_CONNECT_RMT_CLIENT | OBD_CONNECT_VBR |
+ OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH|
+ OBD_CONNECT_EINPROGRESS |
+ OBD_CONNECT_JOBSTATS;
if (sbi->ll_flags & LL_SBI_SOM_PREVIEW)
data->ocd_connect_flags |= OBD_CONNECT_SOM;
#ifdef CONFIG_FS_POSIX_ACL
data->ocd_connect_flags |= OBD_CONNECT_ACL;
#endif
+
+ if (OBD_FAIL_CHECK(OBD_FAIL_MDC_LIGHTWEIGHT))
+ /* flag mdc connection as lightweight, only used for test
+ * purpose, use with care */
+ data->ocd_connect_flags |= OBD_CONNECT_LIGHTWEIGHT;
+
data->ocd_ibits_known = MDS_INODELOCK_FULL;
data->ocd_version = LUSTRE_VERSION_CODE;
/* force vfs to use lustre handler for flock() calls - bug 10743 */
sb->s_flags |= MS_FLOCK_LOCK;
#endif
+#ifdef MS_HAS_NEW_AOPS
+ sb->s_flags |= MS_HAS_NEW_AOPS;
+#endif
if (sbi->ll_flags & LL_SBI_FLOCK)
sbi->ll_fop = &ll_file_operations_flock;
if (sbi->ll_flags & LL_SBI_RMT_CLIENT)
data->ocd_connect_flags |= OBD_CONNECT_RMT_CLIENT_FORCE;
+ data->ocd_brw_size = PTLRPC_MAX_BRW_SIZE;
+
err = obd_connect(NULL, &sbi->ll_md_exp, obd, &sbi->ll_sb_uuid, data, NULL);
if (err == -EBUSY) {
LCONSOLE_ERROR_MSG(0x14f, "An MDT (md %s) is performing "
GOTO(out_md, err);
}
- err = obd_statfs(obd, &osfs,
+ err = obd_statfs(NULL, sbi->ll_md_exp, osfs,
cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS), 0);
if (err)
GOTO(out_md_fid, err);
+ /* This needs to be after statfs to ensure connect has finished.
+ * Note that "data" does NOT contain the valid connect reply.
+ * If connecting to a 1.8 server there will be no LMV device, so
+ * we can access the MDC export directly and exp_connect_flags will
+ * be non-zero, but if accessing an upgraded 2.1 server it will
+ * have the correct flags filled in.
+ * XXX: fill in the LMV exp_connect_flags from MDC(s). */
+ valid = sbi->ll_md_exp->exp_connect_flags & CLIENT_CONNECT_MDT_REQD;
+ if (sbi->ll_md_exp->exp_connect_flags != 0 &&
+ valid != CLIENT_CONNECT_MDT_REQD) {
+ char *buf;
+
+ OBD_ALLOC_WAIT(buf, CFS_PAGE_SIZE);
+ obd_connect_flags2str(buf, CFS_PAGE_SIZE,
+ valid ^ CLIENT_CONNECT_MDT_REQD, ",");
+ LCONSOLE_ERROR_MSG(0x170, "Server %s does not support "
+ "feature(s) needed for correct operation "
+ "of this client (%s). Please upgrade "
+ "server or downgrade client.\n",
+ sbi->ll_md_exp->exp_obd->obd_name, buf);
+ OBD_FREE(buf, CFS_PAGE_SIZE);
+ GOTO(out_md, err = -EPROTO);
+ }
+
size = sizeof(*data);
- err = obd_get_info(sbi->ll_md_exp, sizeof(KEY_CONN_DATA),
+ err = obd_get_info(NULL, sbi->ll_md_exp, sizeof(KEY_CONN_DATA),
KEY_CONN_DATA, &size, data, NULL);
if (err) {
CERROR("Get connect data failed: %d \n", err);
GOTO(out_md, err);
}
- LASSERT(osfs.os_bsize);
- sb->s_blocksize = osfs.os_bsize;
- sb->s_blocksize_bits = log2(osfs.os_bsize);
+ LASSERT(osfs->os_bsize);
+ sb->s_blocksize = osfs->os_bsize;
+ sb->s_blocksize_bits = log2(osfs->os_bsize);
sb->s_magic = LL_SUPER_MAGIC;
-
- /* for bug 11559. in $LINUX/fs/read_write.c, function do_sendfile():
- * retval = in_file->f_op->sendfile(...);
- * if (*ppos > max)
- * retval = -EOVERFLOW;
- *
- * it will check if *ppos is greater than max. However, max equals to
- * s_maxbytes, which is a negative integer in a x86_64 box since loff_t
- * has been defined as a signed long long integer in linux kernel. */
-#if BITS_PER_LONG == 64
- sb->s_maxbytes = PAGE_CACHE_MAXBYTES >> 1;
-#else
- sb->s_maxbytes = PAGE_CACHE_MAXBYTES;
-#endif
- sbi->ll_namelen = osfs.os_namelen;
+ sb->s_maxbytes = MAX_LFS_FILESIZE;
+ sbi->ll_namelen = osfs->os_namelen;
sbi->ll_max_rw_chunk = LL_DEFAULT_MAX_RW_CHUNK;
if ((sbi->ll_flags & LL_SBI_USER_XATTR) &&
sbi->ll_flags |= LL_SBI_OSS_CAPA;
}
+ if (data->ocd_connect_flags & OBD_CONNECT_64BITHASH)
+ sbi->ll_flags |= LL_SBI_64BIT_HASH;
+
+ if (data->ocd_connect_flags & OBD_CONNECT_BRW_SIZE)
+ sbi->ll_md_brw_size = data->ocd_brw_size;
+ else
+ sbi->ll_md_brw_size = CFS_PAGE_SIZE;
+
+ if (data->ocd_connect_flags & OBD_CONNECT_LAYOUTLOCK) {
+ LCONSOLE_INFO("Layout lock feature supported.\n");
+ sbi->ll_flags |= LL_SBI_LAYOUT_LOCK;
+ }
+
obd = class_name2obd(dt);
if (!obd) {
CERROR("DT %s: not setup or attached\n", dt);
OBD_CONNECT_SRVLOCK | OBD_CONNECT_TRUNCLOCK|
OBD_CONNECT_AT | OBD_CONNECT_RMT_CLIENT |
OBD_CONNECT_OSS_CAPA | OBD_CONNECT_VBR|
- OBD_CONNECT_FULL20;
+ OBD_CONNECT_FULL20 | OBD_CONNECT_64BITHASH |
+ OBD_CONNECT_MAXBYTES |
+ OBD_CONNECT_EINPROGRESS |
+ OBD_CONNECT_JOBSTATS;
if (sbi->ll_flags & LL_SBI_SOM_PREVIEW)
data->ocd_connect_flags |= OBD_CONNECT_SOM;
* agreement on the supported algorithms at connect time */
data->ocd_connect_flags |= OBD_CONNECT_CKSUM;
- if (OBD_FAIL_CHECK(OBD_FAIL_OSC_CKSUM_ADLER_ONLY))
- data->ocd_cksum_types = OBD_CKSUM_ADLER;
- else
- /* send the list of supported checksum types */
- data->ocd_cksum_types = OBD_CKSUM_ALL;
+ if (OBD_FAIL_CHECK(OBD_FAIL_OSC_CKSUM_ADLER_ONLY))
+ data->ocd_cksum_types = OBD_CKSUM_ADLER;
+ else
+ data->ocd_cksum_types = cksum_types_supported_client();
}
#ifdef HAVE_LRU_RESIZE_SUPPORT
obd->obd_upcall.onu_owner = &sbi->ll_lco;
obd->obd_upcall.onu_upcall = cl_ocd_update;
- data->ocd_brw_size = PTLRPC_MAX_BRW_PAGES << CFS_PAGE_SHIFT;
+
+ data->ocd_brw_size = PTLRPC_MAX_BRW_SIZE;
err = obd_connect(NULL, &sbi->ll_dt_exp, obd, &sbi->ll_sb_uuid, data, NULL);
if (err == -EBUSY) {
GOTO(out_dt, err);
}
- cfs_mutex_down(&sbi->ll_lco.lco_lock);
+ cfs_mutex_lock(&sbi->ll_lco.lco_lock);
sbi->ll_lco.lco_flags = data->ocd_connect_flags;
sbi->ll_lco.lco_md_exp = sbi->ll_md_exp;
sbi->ll_lco.lco_dt_exp = sbi->ll_dt_exp;
- cfs_mutex_up(&sbi->ll_lco.lco_lock);
+ cfs_mutex_unlock(&sbi->ll_lco.lco_lock);
fid_zero(&sbi->ll_root_fid);
err = md_getstatus(sbi->ll_md_exp, &sbi->ll_root_fid, &oc);
CDEBUG(D_SUPER, "rootfid "DFID"\n", PFID(&sbi->ll_root_fid));
sb->s_op = &lustre_super_operations;
+#if THREAD_SIZE >= 8192 /*b=17630*/
sb->s_export_op = &lustre_export_operations;
+#endif
/* make root inode
* XXX: move this to after cbd setup? */
valid |= OBD_MD_FLACL;
OBD_ALLOC_PTR(op_data);
- if (op_data == NULL)
+ if (op_data == NULL)
GOTO(out_lock_cn_cb, err = -ENOMEM);
op_data->op_fid1 = sbi->ll_root_fid;
CERROR("md_getattr failed for root: rc = %d\n", err);
GOTO(out_lock_cn_cb, err);
}
- memset(&lmd, 0, sizeof(lmd));
err = md_get_lustre_md(sbi->ll_md_exp, request, sbi->ll_dt_exp,
sbi->ll_md_exp, &lmd);
if (err) {
}
LASSERT(fid_is_sane(&sbi->ll_root_fid));
- root = ll_iget(sb, cl_fid_build_ino(&sbi->ll_root_fid), &lmd);
+ root = ll_iget(sb, cl_fid_build_ino(&sbi->ll_root_fid, 0), &lmd);
md_free_lustre_md(sbi->ll_md_exp, &lmd);
ptlrpc_req_finished(request);
#endif
checksum = sbi->ll_flags & LL_SBI_CHECKSUM;
- err = obd_set_info_async(sbi->ll_dt_exp, sizeof(KEY_CHECKSUM),
+ err = obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_CHECKSUM),
KEY_CHECKSUM, sizeof(checksum), &checksum,
NULL);
cl_sb_init(sb);
- sb->s_root = d_alloc_root(root);
- if (data != NULL)
- OBD_FREE(data, sizeof(*data));
+ err = obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_LRU_SET),
+ KEY_LRU_SET, sizeof(sbi->ll_lru),
+ &sbi->ll_lru, NULL);
- sb->s_root->d_op = &ll_d_root_ops;
+ sb->s_root = d_alloc_root(root);
+#ifdef HAVE_DCACHE_LOCK
+ sb->s_root->d_op = &ll_d_root_ops;
+#else
+ /* kernel >= 2.6.38 store dentry operations in sb->s_d_op. */
+ d_set_d_op(sb->s_root, &ll_d_root_ops);
+ sb->s_d_op = &ll_d_ops;
+#endif
sbi->ll_sdev_orig = sb->s_dev;
if (uuid != NULL)
sb->s_dev = get_uuid2int(uuid->uuid, strlen(uuid->uuid));
+ if (data != NULL)
+ OBD_FREE_PTR(data);
+ if (osfs != NULL)
+ OBD_FREE_PTR(osfs);
+
RETURN(err);
out_root:
if (root)
out:
if (data != NULL)
OBD_FREE_PTR(data);
+ if (osfs != NULL)
+ OBD_FREE_PTR(osfs);
lprocfs_unregister_mountpoint(sbi);
return err;
}
*lmmsize = obd_size_diskmd(sbi->ll_dt_exp, NULL);
size = sizeof(int);
- rc = obd_get_info(sbi->ll_md_exp, sizeof(KEY_MAX_EASIZE),
+ rc = obd_get_info(NULL, sbi->ll_md_exp, sizeof(KEY_MAX_EASIZE),
KEY_MAX_EASIZE, &size, lmmsize, NULL);
if (rc)
CERROR("Get max mdsize error rc %d \n", rc);
" flags=0x%x, fsdata=%p, %d subdirs\n", dentry,
dentry->d_name.len, dentry->d_name.name,
dentry->d_parent->d_name.len, dentry->d_parent->d_name.name,
- dentry->d_parent, dentry->d_inode, atomic_read(&dentry->d_count),
+ dentry->d_parent, dentry->d_inode, d_refcount(dentry),
dentry->d_flags, dentry->d_fsdata, subdirs);
if (dentry->d_inode != NULL)
ll_dump_inode(dentry->d_inode);
if (recur == 0)
return;
- list_for_each(tmp, &dentry->d_subdirs) {
- struct dentry *d = list_entry(tmp, struct dentry, d_child);
- lustre_dump_dentry(d, recur - 1);
- }
+ list_for_each(tmp, &dentry->d_subdirs) {
+ struct dentry *d = list_entry(tmp, struct dentry, d_u.d_child);
+ lustre_dump_dentry(d, recur - 1);
+ }
}
void client_common_put_super(struct super_block *sb)
*flags &= ~tmp;
goto next;
}
-#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2,5,50,0)
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 5, 50, 0)
tmp = ll_set_opt("acl", s1, LL_SBI_ACL);
if (tmp) {
/* Ignore deprecated mount option. The client will
*flags |= tmp;
goto next;
}
+ tmp = ll_set_opt("user_fid2path", s1, LL_SBI_USER_FID2PATH);
+ if (tmp) {
+ *flags |= tmp;
+ goto next;
+ }
+ tmp = ll_set_opt("nouser_fid2path", s1, LL_SBI_USER_FID2PATH);
+ if (tmp) {
+ *flags &= ~tmp;
+ goto next;
+ }
tmp = ll_set_opt("checksum", s1, LL_SBI_CHECKSUM);
if (tmp) {
*flags |= tmp;
goto next;
}
-
+ tmp = ll_set_opt("verbose", s1, LL_SBI_VERBOSE);
+ if (tmp) {
+ *flags |= tmp;
+ goto next;
+ }
+ tmp = ll_set_opt("noverbose", s1, LL_SBI_VERBOSE);
+ if (tmp) {
+ *flags &= ~tmp;
+ goto next;
+ }
LCONSOLE_ERROR_MSG(0x152, "Unknown option '%s', won't mount.\n",
s1);
RETURN(-EINVAL);
void ll_lli_init(struct ll_inode_info *lli)
{
lli->lli_inode_magic = LLI_INODE_MAGIC;
- cfs_sema_init(&lli->lli_size_sem, 1);
- cfs_sema_init(&lli->lli_write_sem, 1);
- cfs_sema_init(&lli->lli_trunc_sem, 1);
lli->lli_flags = 0;
- lli->lli_maxbytes = PAGE_CACHE_MAXBYTES;
+ lli->lli_ioepoch = 0;
+ lli->lli_maxbytes = MAX_LFS_FILESIZE;
cfs_spin_lock_init(&lli->lli_lock);
+ lli->lli_posix_acl = NULL;
+ lli->lli_remote_perms = NULL;
+ cfs_mutex_init(&lli->lli_rmtperm_mutex);
+ /* Do not set lli_fid, it has been initialized already. */
+ fid_zero(&lli->lli_pfid);
CFS_INIT_LIST_HEAD(&lli->lli_close_list);
- lli->lli_inode_magic = LLI_INODE_MAGIC;
- cfs_sema_init(&lli->lli_och_sem, 1);
- lli->lli_mds_read_och = lli->lli_mds_write_och = NULL;
+ CFS_INIT_LIST_HEAD(&lli->lli_oss_capas);
+ cfs_atomic_set(&lli->lli_open_count, 0);
+ lli->lli_mds_capa = NULL;
+ lli->lli_rmtperm_time = 0;
+ lli->lli_pending_och = NULL;
+ lli->lli_mds_read_och = NULL;
+ lli->lli_mds_write_och = NULL;
lli->lli_mds_exec_och = NULL;
- lli->lli_open_fd_read_count = lli->lli_open_fd_write_count = 0;
+ lli->lli_open_fd_read_count = 0;
+ lli->lli_open_fd_write_count = 0;
lli->lli_open_fd_exec_count = 0;
- CFS_INIT_LIST_HEAD(&lli->lli_dead_list);
- lli->lli_remote_perms = NULL;
- lli->lli_rmtperm_utime = 0;
- cfs_sema_init(&lli->lli_rmtperm_sem, 1);
- CFS_INIT_LIST_HEAD(&lli->lli_oss_capas);
- cfs_spin_lock_init(&lli->lli_sa_lock);
- CFS_INIT_LIST_HEAD(&lli->lli_sa_dentry);
+ cfs_mutex_init(&lli->lli_och_mutex);
+ cfs_spin_lock_init(&lli->lli_agl_lock);
+ lli->lli_has_smd = false;
+ lli->lli_clob = NULL;
+
+ LASSERT(lli->lli_vfs_inode.i_mode != 0);
+ if (S_ISDIR(lli->lli_vfs_inode.i_mode)) {
+ cfs_mutex_init(&lli->lli_readdir_mutex);
+ lli->lli_opendir_key = NULL;
+ lli->lli_sai = NULL;
+ lli->lli_def_acl = NULL;
+ cfs_spin_lock_init(&lli->lli_sa_lock);
+ lli->lli_opendir_pid = 0;
+ } else {
+ cfs_sema_init(&lli->lli_size_sem, 1);
+ lli->lli_size_sem_owner = NULL;
+ lli->lli_symlink_name = NULL;
+ cfs_init_rwsem(&lli->lli_trunc_sem);
+ cfs_mutex_init(&lli->lli_write_mutex);
+ cfs_init_rwsem(&lli->lli_glimpse_sem);
+ lli->lli_glimpse_time = 0;
+ CFS_INIT_LIST_HEAD(&lli->lli_agl_list);
+ lli->lli_agl_index = 0;
+ lli->lli_async_rc = 0;
+ }
+ cfs_mutex_init(&lli->lli_layout_mutex);
+}
+
+static inline int ll_bdi_register(struct backing_dev_info *bdi)
+{
+#ifdef HAVE_BDI_REGISTER
+ static atomic_t ll_bdi_num = ATOMIC_INIT(0);
+
+#ifdef HAVE_BDI_NAME
+ bdi->name = "lustre";
+#endif
+ return bdi_register(bdi, NULL, "lustre-%d",
+ atomic_inc_return(&ll_bdi_num));
+#else
+ return 0;
+#endif
}
-int ll_fill_super(struct super_block *sb)
+int ll_fill_super(struct super_block *sb, struct vfsmount *mnt)
{
- struct lustre_profile *lprof;
+ struct lustre_profile *lprof = NULL;
struct lustre_sb_info *lsi = s2lsi(sb);
struct ll_sb_info *sbi;
char *dt = NULL, *md = NULL;
char *profilenm = get_profile_name(sb);
- struct config_llog_instance cfg = {0, };
- char ll_instance[sizeof(sb) * 2 + 1];
+ struct config_llog_instance *cfg;
+ /* %p for void* in printf needs 16+2 characters: 0xffffffffffffffff */
+ const int instlen = sizeof(cfg->cfg_instance) * 2 + 2;
int err;
ENTRY;
CDEBUG(D_VFSTRACE, "VFS Op: sb %p\n", sb);
+ OBD_ALLOC_PTR(cfg);
+ if (cfg == NULL)
+ RETURN(-ENOMEM);
+
cfs_module_get();
/* client additional sb info */
lsi->lsi_llsbi = sbi = ll_init_sbi();
if (!sbi) {
cfs_module_put(THIS_MODULE);
+ OBD_FREE_PTR(cfg);
RETURN(-ENOMEM);
}
if (err)
GOTO(out_free, err);
+ err = bdi_init(&lsi->lsi_bdi);
+ if (err)
+ GOTO(out_free, err);
+ lsi->lsi_flags |= LSI_BDI_INITIALIZED;
+ lsi->lsi_bdi.capabilities = BDI_CAP_MAP_COPY;
+ err = ll_bdi_register(&lsi->lsi_bdi);
+ if (err)
+ GOTO(out_free, err);
+
+#ifdef HAVE_SB_BDI
+ sb->s_bdi = &lsi->lsi_bdi;
+#endif
+
/* Generate a string unique to this super, in case some joker tries
to mount the same fs at two mount points.
Use the address of the super itself.*/
- sprintf(ll_instance, "%p", sb);
- cfg.cfg_instance = ll_instance;
- cfg.cfg_uuid = lsi->lsi_llsbi->ll_sb_uuid;
-
+ cfg->cfg_instance = sb;
+ cfg->cfg_uuid = lsi->lsi_llsbi->ll_sb_uuid;
+ cfg->cfg_callback = class_config_llog_handler;
/* set up client obds */
- err = lustre_process_log(sb, profilenm, &cfg);
+ err = lustre_process_log(sb, profilenm, cfg);
if (err < 0) {
CERROR("Unable to process log: %d\n", err);
GOTO(out_free, err);
CDEBUG(D_CONFIG, "Found profile %s: mdc=%s osc=%s\n", profilenm,
lprof->lp_md, lprof->lp_dt);
- OBD_ALLOC(dt, strlen(lprof->lp_dt) +
- strlen(ll_instance) + 2);
+ OBD_ALLOC(dt, strlen(lprof->lp_dt) + instlen + 2);
if (!dt)
GOTO(out_free, err = -ENOMEM);
- sprintf(dt, "%s-%s", lprof->lp_dt, ll_instance);
+ sprintf(dt, "%s-%p", lprof->lp_dt, cfg->cfg_instance);
- OBD_ALLOC(md, strlen(lprof->lp_md) +
- strlen(ll_instance) + 2);
+ OBD_ALLOC(md, strlen(lprof->lp_md) + instlen + 2);
if (!md)
GOTO(out_free, err = -ENOMEM);
- sprintf(md, "%s-%s", lprof->lp_md, ll_instance);
+ sprintf(md, "%s-%p", lprof->lp_md, cfg->cfg_instance);
/* connections, registrations, sb setup */
- err = client_common_fill_super(sb, md, dt);
+ err = client_common_fill_super(sb, md, dt, mnt);
out_free:
if (md)
- OBD_FREE(md, strlen(md) + 1);
+ OBD_FREE(md, strlen(lprof->lp_md) + instlen + 2);
if (dt)
- OBD_FREE(dt, strlen(dt) + 1);
+ OBD_FREE(dt, strlen(lprof->lp_dt) + instlen + 2);
if (err)
ll_put_super(sb);
- else
- LCONSOLE_WARN("Client %s has started\n", profilenm);
+ else if (sbi->ll_flags & LL_SBI_VERBOSE)
+ LCONSOLE_WARN("Mounted %s\n", profilenm);
+ OBD_FREE_PTR(cfg);
RETURN(err);
} /* ll_fill_super */
void ll_put_super(struct super_block *sb)
{
struct config_llog_instance cfg;
- char ll_instance[sizeof(sb) * 2 + 1];
struct obd_device *obd;
struct lustre_sb_info *lsi = s2lsi(sb);
struct ll_sb_info *sbi = ll_s2sbi(sb);
ll_print_capa_stat(sbi);
- sprintf(ll_instance, "%p", sb);
- cfg.cfg_instance = ll_instance;
- lustre_end_log(sb, NULL, &cfg);
+ cfg.cfg_instance = sb;
+ lustre_end_log(sb, profilenm, &cfg);
if (sbi->ll_md_exp) {
obd = class_exp2obd(sbi->ll_md_exp);
class_manual_cleanup(obd);
}
+ if (sbi->ll_flags & LL_SBI_VERBOSE)
+ LCONSOLE_WARN("Unmounted %s\n", profilenm ? profilenm : "");
+
if (profilenm)
class_del_profile(profilenm);
+ if (lsi->lsi_flags & LSI_BDI_INITIALIZED) {
+ bdi_destroy(&lsi->lsi_bdi);
+ lsi->lsi_flags &= ~LSI_BDI_INITIALIZED;
+ }
+
ll_free_sbi(sb);
lsi->lsi_llsbi = NULL;
cl_env_cache_purge(~0);
- LCONSOLE_WARN("client %s umount complete\n", ll_instance);
-
cfs_module_put(THIS_MODULE);
EXIT;
inode = igrab(lock->l_ast_data);
} else {
inode = lock->l_ast_data;
- ldlm_lock_debug(NULL, inode->i_state & I_FREEING ?
- D_INFO : D_WARNING,
- lock, __FILE__, __func__, __LINE__,
- "l_ast_data %p is bogus: magic %08x",
- lock->l_ast_data, lli->lli_inode_magic);
+ LDLM_DEBUG_LIMIT(inode->i_state & I_FREEING ? D_INFO :
+ D_WARNING, lock, "l_ast_data %p is "
+ "bogus: magic %08x", lock->l_ast_data,
+ lli->lli_inode_magic);
inode = NULL;
}
}
if (S_ISDIR(inode->i_mode)) {
/* these should have been cleared in ll_file_release */
- LASSERT(lli->lli_sai == NULL);
LASSERT(lli->lli_opendir_key == NULL);
+ LASSERT(lli->lli_sai == NULL);
LASSERT(lli->lli_opendir_pid == 0);
}
if (lli->lli_mds_read_och)
ll_md_real_close(inode, FMODE_READ);
- if (lli->lli_symlink_name) {
+ if (S_ISLNK(inode->i_mode) && lli->lli_symlink_name) {
OBD_FREE(lli->lli_symlink_name,
strlen(lli->lli_symlink_name) + 1);
lli->lli_symlink_name = NULL;
lli->lli_inode_magic = LLI_INODE_DEAD;
ll_clear_inode_capas(inode);
+ if (!S_ISDIR(inode->i_mode))
+ LASSERT(cfs_list_empty(&lli->lli_agl_list));
+
/*
* XXX This has to be done before lsm is freed below, because
* cl_object still uses inode lsm.
*/
cl_inode_fini(inode);
+ lli->lli_has_smd = false;
- if (lli->lli_smd) {
- obd_free_memmd(sbi->ll_dt_exp, &lli->lli_smd);
- lli->lli_smd = NULL;
- }
-
-
- EXIT;
+ EXIT;
}
-int ll_md_setattr(struct inode *inode, struct md_op_data *op_data,
+int ll_md_setattr(struct dentry *dentry, struct md_op_data *op_data,
struct md_open_data **mod)
{
struct lustre_md md;
+ struct inode *inode = dentry->d_inode;
struct ll_sb_info *sbi = ll_i2sbi(inode);
struct ptlrpc_request *request = NULL;
- int rc;
+ int rc, ia_valid;
ENTRY;
op_data = ll_prep_md_op_data(op_data, inode, NULL, NULL, 0, 0,
rc = md_setattr(sbi->ll_md_exp, op_data, NULL, 0, NULL, 0,
&request, mod);
- if (rc) {
- ptlrpc_req_finished(request);
- if (rc == -ENOENT) {
- inode->i_nlink = 0;
- /* Unlinked special device node? Or just a race?
- * Pretend we done everything. */
- if (!S_ISREG(inode->i_mode) &&
- !S_ISDIR(inode->i_mode))
- rc = inode_setattr(inode, &op_data->op_attr);
- } else if (rc != -EPERM && rc != -EACCES && rc != -ETXTBSY) {
- CERROR("md_setattr fails: rc = %d\n", rc);
- }
- RETURN(rc);
- }
+ if (rc) {
+ ptlrpc_req_finished(request);
+ if (rc == -ENOENT) {
+ clear_nlink(inode);
+ /* Unlinked special device node? Or just a race?
+ * Pretend we done everything. */
+ if (!S_ISREG(inode->i_mode) &&
+ !S_ISDIR(inode->i_mode)) {
+ ia_valid = op_data->op_attr.ia_valid;
+ op_data->op_attr.ia_valid &= ~TIMES_SET_FLAGS;
+ rc = simple_setattr(dentry, &op_data->op_attr);
+ op_data->op_attr.ia_valid = ia_valid;
+ }
+ } else if (rc != -EPERM && rc != -EACCES && rc != -ETXTBSY) {
+ CERROR("md_setattr fails: rc = %d\n", rc);
+ }
+ RETURN(rc);
+ }
rc = md_get_lustre_md(sbi->ll_md_exp, request, sbi->ll_dt_exp,
sbi->ll_md_exp, &md);
RETURN(rc);
}
- /* We call inode_setattr to adjust timestamps.
+ /* We want to adjust timestamps.
* If there is at least some data in file, we cleared ATTR_SIZE
- * above to avoid invoking vmtruncate, otherwise it is important
- * to call vmtruncate in inode_setattr to update inode->i_size
+ * to avoid update size, otherwise it is important to do.(SOM case)
* (bug 6196) */
- rc = inode_setattr(inode, &op_data->op_attr);
+ ia_valid = op_data->op_attr.ia_valid;
+ /* Since we set ATTR_*_SET flags above, and already done permission
+ * check, So don't let inode_change_ok() check it again. */
+ op_data->op_attr.ia_valid &= ~TIMES_SET_FLAGS;
+ rc = simple_setattr(dentry, &op_data->op_attr);
+ op_data->op_attr.ia_valid = ia_valid;
/* Extract epoch data if obtained. */
op_data->op_handle = md.body->handle;
return rc;
}
+#ifndef HAVE_VFS_INODE_NEWSIZE_OK
+/**
+ * inode_newsize_ok - may this inode be truncated to a given size
+ * @inode: the inode to be truncated
+ * @offset: the new size to assign to the inode
+ * @Returns: 0 on success, -ve errno on failure
+ *
+ * inode_newsize_ok will check filesystem limits and ulimits to check that the
+ * new inode size is within limits. inode_newsize_ok will also send SIGXFSZ
+ * when necessary. Caller must not proceed with inode size change if failure is
+ * returned. @inode must be a file (not directory), with appropriate
+ * permissions to allow truncate (inode_newsize_ok does NOT check these
+ * conditions).
+ *
+ * inode_newsize_ok must be called with i_mutex held.
+ */
+int inode_newsize_ok(const struct inode *inode, loff_t offset)
+{
+ if (inode->i_size < offset) {
+ unsigned long limit;
+
+ limit = rlimit(RLIMIT_FSIZE);
+ if (limit != RLIM_INFINITY && offset > limit)
+ goto out_sig;
+ if (offset > inode->i_sb->s_maxbytes)
+ goto out_big;
+ } else {
+ /*
+ * truncation of in-use swapfiles is disallowed - it would
+ * cause subsequent swapout to scribble on the now-freed
+ * blocks.
+ */
+ if (IS_SWAPFILE(inode))
+ return -ETXTBSY;
+ }
+
+ return 0;
+out_sig:
+ send_sig(SIGXFSZ, current, 0);
+out_big:
+ return -EFBIG;
+}
+#endif
+
/* If this inode has objects allocated to it (lsm != NULL), then the OST
* object(s) determine the file size and mtime. Otherwise, the MDS will
* keep these values until such a time that objects are allocated for it.
* I don't believe it is possible to get e.g. ATTR_MTIME_SET and ATTR_SIZE
* at the same time.
*/
-int ll_setattr_raw(struct inode *inode, struct iattr *attr)
+int ll_setattr_raw(struct dentry *dentry, struct iattr *attr)
{
+ struct lov_stripe_md *lsm;
+ struct inode *inode = dentry->d_inode;
struct ll_inode_info *lli = ll_i2info(inode);
- struct lov_stripe_md *lsm = lli->lli_smd;
struct md_op_data *op_data = NULL;
struct md_open_data *mod = NULL;
int ia_valid = attr->ia_valid;
int rc = 0, rc1 = 0;
ENTRY;
- CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu valid %x\n", inode->i_ino,
- attr->ia_valid);
- ll_stats_ops_tally(ll_i2sbi(inode), LPROC_LL_SETATTR, 1);
+ CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu (%p) to %llu, valid %x\n",
+ inode->i_ino,
+ inode,i_size_read(inode),
+ attr->ia_valid);
if (ia_valid & ATTR_SIZE) {
+ /* Check new size against VFS/VM file size limit and rlimit */
+ rc = inode_newsize_ok(inode, attr->ia_size);
+ if (rc)
+ RETURN(rc);
+
+ /* The maximum Lustre file size is variable, based on the
+ * OST maximum object size and number of stripes. This
+ * needs another check in addition to the VFS check above. */
if (attr->ia_size > ll_file_maxbytes(inode)) {
- CDEBUG(D_INODE, "file too large %llu > "LPU64"\n",
- attr->ia_size, ll_file_maxbytes(inode));
+ CDEBUG(D_INODE,"file "DFID" too large %llu > "LPU64"\n",
+ PFID(&lli->lli_fid), attr->ia_size,
+ ll_file_maxbytes(inode));
RETURN(-EFBIG);
}
}
/* POSIX: check before ATTR_*TIME_SET set (from inode_change_ok) */
- if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET)) {
+ if (ia_valid & TIMES_SET_FLAGS) {
if (cfs_curproc_fsuid() != inode->i_uid &&
!cfs_capable(CFS_CAP_FOWNER))
RETURN(-EPERM);
LTIME_S(attr->ia_mtime), LTIME_S(attr->ia_ctime),
cfs_time_current_sec());
- /* NB: ATTR_SIZE will only be set after this point if the size
- * resides on the MDS, ie, this file has no objects. */
- if (lsm)
- attr->ia_valid &= ~ATTR_SIZE;
-
/* We always do an MDS RPC, even if we're only changing the size;
* only the MDS knows whether truncate() should fail with -ETXTBUSY */
if (op_data == NULL)
RETURN(-ENOMEM);
- UNLOCK_INODE_MUTEX(inode);
- if (ia_valid & ATTR_SIZE)
- UP_WRITE_I_ALLOC_SEM(inode);
- cfs_down(&lli->lli_trunc_sem);
- LOCK_INODE_MUTEX(inode);
- if (ia_valid & ATTR_SIZE)
- DOWN_WRITE_I_ALLOC_SEM(inode);
+ if (!S_ISDIR(inode->i_mode)) {
+ if (ia_valid & ATTR_SIZE)
+ inode_dio_write_done(inode);
+ mutex_unlock(&inode->i_mutex);
+ cfs_down_write(&lli->lli_trunc_sem);
+ mutex_lock(&inode->i_mutex);
+ if (ia_valid & ATTR_SIZE)
+ inode_dio_wait(inode);
+ }
+
+ /* We need a steady stripe configuration for setattr to avoid
+ * confusion. */
+ lsm = ccc_inode_lsm_get(inode);
+
+ /* NB: ATTR_SIZE will only be set after this point if the size
+ * resides on the MDS, ie, this file has no objects. */
+ if (lsm != NULL)
+ attr->ia_valid &= ~ATTR_SIZE;
memcpy(&op_data->op_attr, attr, sizeof(*attr));
(ia_valid & (ATTR_SIZE | ATTR_MTIME | ATTR_MTIME_SET)))
op_data->op_flags = MF_EPOCH_OPEN;
- rc = ll_md_setattr(inode, op_data, &mod);
+ rc = ll_md_setattr(dentry, op_data, &mod);
if (rc)
GOTO(out, rc);
ll_ioepoch_open(lli, op_data->op_ioepoch);
- if (!lsm || !S_ISREG(inode->i_mode)) {
+ if (lsm == NULL || !S_ISREG(inode->i_mode)) {
CDEBUG(D_INODE, "no lsm: not setting attrs on OST\n");
GOTO(out, rc = 0);
}
if (ia_valid & ATTR_SIZE)
attr->ia_valid |= ATTR_SIZE;
- if ((ia_valid & ATTR_SIZE) |
- ((ia_valid | ATTR_ATIME | ATTR_ATIME_SET) &&
- LTIME_S(attr->ia_atime) < LTIME_S(attr->ia_ctime)) ||
- ((ia_valid | ATTR_MTIME | ATTR_MTIME_SET) &&
- LTIME_S(attr->ia_mtime) < LTIME_S(attr->ia_ctime)))
- /* perform truncate and setting mtime/atime to past under PW
- * 0:EOF extent lock (new_size:EOF for truncate) */
+ if (ia_valid & (ATTR_SIZE |
+ ATTR_ATIME | ATTR_ATIME_SET |
+ ATTR_MTIME | ATTR_MTIME_SET))
+ /* For truncate and utimes sending attributes to OSTs, setting
+ * mtime/atime to the past will be performed under PW [0:EOF]
+ * extent lock (new_size:EOF for truncate). It may seem
+ * excessive to send mtime/atime updates to OSTs when not
+ * setting times to past, but it is necessary due to possible
+ * time de-synchronization between MDT inode and OST objects */
rc = ll_setattr_ost(inode, attr);
EXIT;
out:
+ ccc_inode_lsm_put(inode, lsm);
if (op_data) {
- if (op_data->op_ioepoch)
+ if (op_data->op_ioepoch) {
rc1 = ll_setattr_done_writing(inode, op_data, mod);
+ if (!rc)
+ rc = rc1;
+ }
ll_finish_md_op_data(op_data);
}
- cfs_up(&lli->lli_trunc_sem);
- return rc ? rc : rc1;
+ if (!S_ISDIR(inode->i_mode))
+ cfs_up_write(&lli->lli_trunc_sem);
+
+ ll_stats_ops_tally(ll_i2sbi(inode), (ia_valid & ATTR_SIZE) ?
+ LPROC_LL_TRUNC : LPROC_LL_SETATTR, 1);
+
+ return rc;
}
int ll_setattr(struct dentry *de, struct iattr *attr)
{
+ int mode = de->d_inode->i_mode;
+
if ((attr->ia_valid & (ATTR_CTIME|ATTR_SIZE|ATTR_MODE)) ==
- (ATTR_CTIME|ATTR_SIZE|ATTR_MODE))
+ (ATTR_CTIME|ATTR_SIZE|ATTR_MODE))
attr->ia_valid |= MDS_OPEN_OWNEROVERRIDE;
- if ((de->d_inode->i_mode & S_ISUID) &&
+ if (((attr->ia_valid & (ATTR_MODE|ATTR_FORCE|ATTR_SIZE)) ==
+ (ATTR_SIZE|ATTR_MODE)) &&
+ (((mode & S_ISUID) && !(attr->ia_mode & S_ISUID)) ||
+ (((mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) &&
+ !(attr->ia_mode & S_ISGID))))
+ attr->ia_valid |= ATTR_FORCE;
+
+ if ((mode & S_ISUID) &&
!(attr->ia_mode & S_ISUID) &&
!(attr->ia_valid & ATTR_KILL_SUID))
attr->ia_valid |= ATTR_KILL_SUID;
- if (((de->d_inode->i_mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) &&
+ if (((mode & (S_ISGID|S_IXGRP)) == (S_ISGID|S_IXGRP)) &&
!(attr->ia_mode & S_ISGID) &&
!(attr->ia_valid & ATTR_KILL_SGID))
attr->ia_valid |= ATTR_KILL_SGID;
- return ll_setattr_raw(de->d_inode, attr);
+ return ll_setattr_raw(de, attr);
}
int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
int rc;
ENTRY;
- rc = obd_statfs(class_exp2obd(sbi->ll_md_exp), osfs, max_age, flags);
+ rc = obd_statfs(NULL, sbi->ll_md_exp, osfs, max_age, flags);
if (rc) {
CERROR("md_statfs fails: rc = %d\n", rc);
RETURN(rc);
if (sbi->ll_flags & LL_SBI_LAZYSTATFS)
flags |= OBD_STATFS_NODELAY;
- rc = obd_statfs_rqset(class_exp2obd(sbi->ll_dt_exp),
- &obd_osfs, max_age, flags);
+ rc = obd_statfs_rqset(sbi->ll_dt_exp, &obd_osfs, max_age, flags);
if (rc) {
CERROR("obd_statfs fails: rc = %d\n", rc);
RETURN(rc);
RETURN(rc);
}
-#ifndef HAVE_STATFS_DENTRY_PARAM
-int ll_statfs(struct super_block *sb, struct kstatfs *sfs)
-{
-#else
int ll_statfs(struct dentry *de, struct kstatfs *sfs)
{
- struct super_block *sb = de->d_sb;
-#endif
- struct obd_statfs osfs;
- int rc;
+ struct super_block *sb = de->d_sb;
+ struct obd_statfs osfs;
+ int rc;
CDEBUG(D_VFSTRACE, "VFS Op: at "LPU64" jiffies\n", get_jiffies_64());
ll_stats_ops_tally(ll_s2sbi(sb), LPROC_LL_STAFS, 1);
return 0;
}
-void ll_inode_size_lock(struct inode *inode, int lock_lsm)
+void ll_inode_size_lock(struct inode *inode)
{
struct ll_inode_info *lli;
- struct lov_stripe_md *lsm;
+
+ LASSERT(!S_ISDIR(inode->i_mode));
lli = ll_i2info(inode);
LASSERT(lli->lli_size_sem_owner != current);
cfs_down(&lli->lli_size_sem);
LASSERT(lli->lli_size_sem_owner == NULL);
lli->lli_size_sem_owner = current;
- lsm = lli->lli_smd;
- LASSERTF(lsm != NULL || lock_lsm == 0, "lsm %p, lock_lsm %d\n",
- lsm, lock_lsm);
- if (lock_lsm)
- lov_stripe_lock(lsm);
}
-void ll_inode_size_unlock(struct inode *inode, int unlock_lsm)
+void ll_inode_size_unlock(struct inode *inode)
{
struct ll_inode_info *lli;
- struct lov_stripe_md *lsm;
lli = ll_i2info(inode);
- lsm = lli->lli_smd;
- LASSERTF(lsm != NULL || unlock_lsm == 0, "lsm %p, lock_lsm %d\n",
- lsm, unlock_lsm);
- if (unlock_lsm)
- lov_stripe_unlock(lsm);
LASSERT(lli->lli_size_sem_owner == current);
lli->lli_size_sem_owner = NULL;
cfs_up(&lli->lli_size_sem);
void ll_update_inode(struct inode *inode, struct lustre_md *md)
{
- struct ll_inode_info *lli = ll_i2info(inode);
- struct mdt_body *body = md->body;
- struct lov_stripe_md *lsm = md->lsm;
- struct ll_sb_info *sbi = ll_i2sbi(inode);
-
- LASSERT ((lsm != NULL) == ((body->valid & OBD_MD_FLEASIZE) != 0));
- if (lsm != NULL) {
- cfs_down(&lli->lli_och_sem);
- if (lli->lli_smd == NULL) {
- if (lsm->lsm_magic != LOV_MAGIC_V1 &&
- lsm->lsm_magic != LOV_MAGIC_V3) {
- dump_lsm(D_ERROR, lsm);
- LBUG();
- }
- CDEBUG(D_INODE, "adding lsm %p to inode %lu/%u(%p)\n",
- lsm, inode->i_ino, inode->i_generation, inode);
- /* cl_inode_init must go before lli_smd or a race is
- * possible where client thinks the file has stripes,
- * but lov raid0 is not setup yet and parallel e.g.
- * glimpse would try to use uninitialized lov */
- cl_inode_init(inode, md);
- lli->lli_smd = lsm;
- cfs_up(&lli->lli_och_sem);
- lli->lli_maxbytes = lsm->lsm_maxbytes;
- if (lli->lli_maxbytes > PAGE_CACHE_MAXBYTES)
- lli->lli_maxbytes = PAGE_CACHE_MAXBYTES;
- } else {
- cfs_up(&lli->lli_och_sem);
- LASSERT(lli->lli_smd->lsm_magic == lsm->lsm_magic &&
- lli->lli_smd->lsm_stripe_count ==
- lsm->lsm_stripe_count);
- if (lov_stripe_md_cmp(lli->lli_smd, lsm)) {
- CERROR("lsm mismatch for inode %ld\n",
- inode->i_ino);
- CERROR("lli_smd:\n");
- dump_lsm(D_ERROR, lli->lli_smd);
- CERROR("lsm:\n");
- dump_lsm(D_ERROR, lsm);
- LBUG();
- }
- }
- if (lli->lli_smd != lsm)
- obd_free_memmd(ll_i2dtexp(inode), &lsm);
- }
+ struct ll_inode_info *lli = ll_i2info(inode);
+ struct mdt_body *body = md->body;
+ struct lov_stripe_md *lsm = md->lsm;
+ struct ll_sb_info *sbi = ll_i2sbi(inode);
+
+ LASSERT ((lsm != NULL) == ((body->valid & OBD_MD_FLEASIZE) != 0));
+ if (lsm != NULL) {
+ LASSERT(S_ISREG(inode->i_mode));
+ cfs_mutex_lock(&lli->lli_och_mutex);
+ CDEBUG(D_INODE, "adding lsm %p to inode %lu/%u(%p)\n",
+ lsm, inode->i_ino, inode->i_generation, inode);
+ /* cl_file_inode_init must go before lli_has_smd or a race
+ * is possible where client thinks the file has stripes,
+ * but lov raid0 is not setup yet and parallel e.g.
+ * glimpse would try to use uninitialized lov */
+ if (cl_file_inode_init(inode, md) == 0)
+ lli->lli_has_smd = true;
+ cfs_mutex_unlock(&lli->lli_och_mutex);
+
+ lli->lli_maxbytes = lsm->lsm_maxbytes;
+ if (lli->lli_maxbytes > MAX_LFS_FILESIZE)
+ lli->lli_maxbytes = MAX_LFS_FILESIZE;
+ if (md->lsm != NULL)
+ obd_free_memmd(ll_i2dtexp(inode), &md->lsm);
+ }
if (sbi->ll_flags & LL_SBI_RMT_CLIENT) {
if (body->valid & OBD_MD_FLRMTPERM)
cfs_spin_unlock(&lli->lli_lock);
}
#endif
- inode->i_ino = cl_fid_build_ino(&body->fid1);
+ inode->i_ino = cl_fid_build_ino(&body->fid1, 0);
inode->i_generation = cl_fid_build_gen(&body->fid1);
if (body->valid & OBD_MD_FLATIME) {
} else {
inode->i_blkbits = inode->i_sb->s_blocksize_bits;
}
-#ifdef HAVE_INODE_BLKSIZE
- inode->i_blksize = 1<<inode->i_blkbits;
-#endif
- if (body->valid & OBD_MD_FLUID)
- inode->i_uid = body->uid;
- if (body->valid & OBD_MD_FLGID)
- inode->i_gid = body->gid;
- if (body->valid & OBD_MD_FLFLAGS)
- inode->i_flags = ll_ext_to_inode_flags(body->flags);
- if (body->valid & OBD_MD_FLNLINK)
- inode->i_nlink = body->nlink;
- if (body->valid & OBD_MD_FLRDEV)
- inode->i_rdev = old_decode_dev(body->rdev);
+ if (body->valid & OBD_MD_FLUID)
+ inode->i_uid = body->uid;
+ if (body->valid & OBD_MD_FLGID)
+ inode->i_gid = body->gid;
+ if (body->valid & OBD_MD_FLFLAGS)
+ inode->i_flags = ll_ext_to_inode_flags(body->flags);
+ if (body->valid & OBD_MD_FLNLINK)
+ set_nlink(inode, body->nlink);
+ if (body->valid & OBD_MD_FLRDEV)
+ inode->i_rdev = old_decode_dev(body->rdev);
if (body->valid & OBD_MD_FLID) {
/* FID shouldn't be changed! */
if (body->valid & OBD_MD_FLSIZE) {
if (exp_connect_som(ll_i2mdexp(inode)) &&
- S_ISREG(inode->i_mode) && lli->lli_smd) {
+ S_ISREG(inode->i_mode) && lli->lli_has_smd) {
struct lustre_handle lockh;
ldlm_mode_t mode;
if (lli->lli_flags & (LLIF_DONE_WRITING |
LLIF_EPOCH_PENDING |
LLIF_SOM_DIRTY)) {
- CERROR("ino %lu flags %lu still has "
+ CERROR("ino %lu flags %u still has "
"size authority! do not trust "
"the size got from MDS\n",
inode->i_ino, lli->lli_flags);
} else {
/* Use old size assignment to avoid
* deadlock bz14138 & bz14326 */
- inode->i_size = body->size;
+ i_size_write(inode, body->size);
lli->lli_flags |= LLIF_MDS_SIZE_LOCK;
}
ldlm_lock_decref(&lockh, mode);
} else {
/* Use old size assignment to avoid
* deadlock bz14138 & bz14326 */
- inode->i_size = body->size;
+ i_size_write(inode, body->size);
CDEBUG(D_VFSTRACE, "inode=%lu, updating i_size %llu\n",
- inode->i_ino, body->size);
+ inode->i_ino, (unsigned long long)body->size);
}
if (body->valid & OBD_MD_FLBLOCKS)
}
}
-static struct backing_dev_info ll_backing_dev_info = {
- .ra_pages = 0, /* No readahead */
-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,12))
- .capabilities = 0, /* Does contribute to dirty memory */
-#else
- .memory_backed = 0, /* Does contribute to dirty memory */
-#endif
-};
-
void ll_read_inode2(struct inode *inode, void *opaque)
{
struct lustre_md *md = opaque;
struct ll_inode_info *lli = ll_i2info(inode);
ENTRY;
- CDEBUG(D_VFSTRACE, "VFS Op:inode=%lu/%u(%p)\n",
- inode->i_ino, inode->i_generation, inode);
+ CDEBUG(D_VFSTRACE, "VFS Op:inode="DFID"(%p)\n",
+ PFID(&lli->lli_fid), inode);
- ll_lli_init(lli);
-
- LASSERT(!lli->lli_smd);
+ LASSERT(!lli->lli_has_smd);
/* Core attributes from the MDS first. This is a new inode, and
* the VFS doesn't zero times in the core inode so we have to do
/* OIDEBUG(inode); */
+ /* initializing backing dev info. */
+ inode->i_mapping->backing_dev_info = &s2lsi(inode->i_sb)->lsi_bdi;
+
+
if (S_ISREG(inode->i_mode)) {
struct ll_sb_info *sbi = ll_i2sbi(inode);
inode->i_op = &ll_file_inode_operations;
inode->i_fop = sbi->ll_fop;
- inode->i_mapping->a_ops = &ll_aops;
+ inode->i_mapping->a_ops = (struct address_space_operations *)&ll_aops;
EXIT;
} else if (S_ISDIR(inode->i_mode)) {
inode->i_op = &ll_dir_inode_operations;
inode->i_fop = &ll_dir_operations;
- inode->i_mapping->a_ops = &ll_dir_aops;
EXIT;
} else if (S_ISLNK(inode->i_mode)) {
inode->i_op = &ll_fast_symlink_inode_operations;
} else {
inode->i_op = &ll_special_inode_operations;
- init_special_inode(inode, inode->i_mode,
- kdev_t_to_nr(inode->i_rdev));
-
- /* initializing backing dev info. */
- inode->i_mapping->backing_dev_info = &ll_backing_dev_info;
+ init_special_inode(inode, inode->i_mode,
+ inode->i_rdev);
EXIT;
}
void ll_delete_inode(struct inode *inode)
{
- struct ll_sb_info *sbi = ll_i2sbi(inode);
- int rc;
- ENTRY;
+ struct cl_inode_info *lli = cl_i2info(inode);
+ struct ll_sb_info *sbi = ll_i2sbi(inode);
+ int rc;
+ ENTRY;
+
+ rc = obd_fid_delete(sbi->ll_md_exp, ll_inode2fid(inode));
+ if (rc)
+ CERROR("fid_delete() failed, rc %d\n", rc);
+
+ if (S_ISREG(inode->i_mode) && lli->lli_clob != NULL)
+ /* discard all dirty pages before truncating them, required by
+ * osc_extent implementation at LU-1030. */
+ cl_sync_file_range(inode, 0, OBD_OBJECT_EOF, CL_FSYNC_DISCARD);
- rc = obd_fid_delete(sbi->ll_md_exp, ll_inode2fid(inode));
- if (rc) {
- CERROR("fid_delete() failed, rc %d\n", rc);
- }
truncate_inode_pages(&inode->i_data, 0);
+
+ /* Workaround for LU-118 */
+ if (inode->i_data.nrpages) {
+ TREE_READ_LOCK_IRQ(&inode->i_data);
+ TREE_READ_UNLOCK_IRQ(&inode->i_data);
+ LASSERTF(inode->i_data.nrpages == 0,
+ "inode=%lu/%u(%p) nrpages=%lu, see "
+ "http://jira.whamcloud.com/browse/LU-118\n",
+ inode->i_ino, inode->i_generation, inode,
+ inode->i_data.nrpages);
+ }
+ /* Workaround end */
+
+#ifdef HAVE_SBOPS_EVICT_INODE
+ ll_clear_inode(inode);
+ end_writeback(inode);
+#else
clear_inode(inode);
+#endif
EXIT;
}
op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL,
0, 0, LUSTRE_OPC_ANY,
NULL);
- if (op_data == NULL)
- RETURN(-ENOMEM);
+ if (IS_ERR(op_data))
+ RETURN(PTR_ERR(op_data));
op_data->op_valid = OBD_MD_FLFLAGS;
rc = md_getattr(sbi->ll_md_exp, op_data, &req);
RETURN(put_user(flags, (int *)arg));
}
case FSFILT_IOC_SETFLAGS: {
- struct lov_stripe_md *lsm = ll_i2info(inode)->lli_smd;
+ struct lov_stripe_md *lsm;
struct obd_info oinfo = { { { 0 } } };
struct md_op_data *op_data;
if (get_user(flags, (int *)arg))
RETURN(-EFAULT);
- oinfo.oi_md = lsm;
- OBDO_ALLOC(oinfo.oi_oa);
- if (!oinfo.oi_oa)
- RETURN(-ENOMEM);
-
op_data = ll_prep_md_op_data(NULL, inode, NULL, NULL, 0, 0,
LUSTRE_OPC_ANY, NULL);
if (IS_ERR(op_data))
NULL, 0, NULL, 0, &req, NULL);
ll_finish_md_op_data(op_data);
ptlrpc_req_finished(req);
- if (rc) {
- OBDO_FREE(oinfo.oi_oa);
- RETURN(rc);
- }
+ if (rc)
+ RETURN(rc);
- if (lsm == NULL) {
- OBDO_FREE(oinfo.oi_oa);
- GOTO(update_cache, rc);
- }
+ inode->i_flags = ll_ext_to_inode_flags(flags);
+ lsm = ccc_inode_lsm_get(inode);
+ if (lsm == NULL)
+ RETURN(0);
+
+ OBDO_ALLOC(oinfo.oi_oa);
+ if (!oinfo.oi_oa) {
+ ccc_inode_lsm_put(inode, lsm);
+ RETURN(-ENOMEM);
+ }
+ oinfo.oi_md = lsm;
oinfo.oi_oa->o_id = lsm->lsm_object_id;
oinfo.oi_oa->o_seq = lsm->lsm_object_seq;
oinfo.oi_oa->o_flags = flags;
oinfo.oi_oa->o_valid = OBD_MD_FLID | OBD_MD_FLFLAGS |
OBD_MD_FLGROUP;
oinfo.oi_capa = ll_mdscapa_get(inode);
- obdo_from_inode(oinfo.oi_oa, inode,
- &ll_i2info(inode)->lli_fid, 0);
+ obdo_set_parent_fid(oinfo.oi_oa, &ll_i2info(inode)->lli_fid);
rc = obd_setattr_rqset(sbi->ll_dt_exp, &oinfo, NULL);
capa_put(oinfo.oi_capa);
OBDO_FREE(oinfo.oi_oa);
- if (rc) {
- if (rc != -EPERM && rc != -EACCES)
- CERROR("osc_setattr_async fails: rc = %d\n",rc);
- RETURN(rc);
- }
+ ccc_inode_lsm_put(inode, lsm);
- EXIT;
-update_cache:
- inode->i_flags = ll_ext_to_inode_flags(flags);
- return 0;
+ if (rc && rc != -EPERM && rc != -EACCES)
+ CERROR("osc_setattr_async fails: rc = %d\n", rc);
+
+ RETURN(rc);
}
default:
RETURN(-ENOSYS);
CDEBUG(D_SEC, "flush context for user %d\n", cfs_curproc_uid());
- obd_set_info_async(sbi->ll_md_exp,
+ obd_set_info_async(NULL, sbi->ll_md_exp,
sizeof(KEY_FLUSH_CTX), KEY_FLUSH_CTX,
0, NULL, NULL);
- obd_set_info_async(sbi->ll_dt_exp,
+ obd_set_info_async(NULL, sbi->ll_dt_exp,
sizeof(KEY_FLUSH_CTX), KEY_FLUSH_CTX,
0, NULL, NULL);
return 0;
void ll_umount_begin(struct super_block *sb)
{
#endif
- struct lustre_sb_info *lsi = s2lsi(sb);
struct ll_sb_info *sbi = ll_s2sbi(sb);
struct obd_device *obd;
struct obd_ioctl_data *ioc_data;
}
#endif
- /* Tell the MGC we got umount -f */
- lsi->lsi_flags |= LSI_UMOUNT_FORCE;
-
CDEBUG(D_VFSTRACE, "VFS Op: superblock %p count %d active %d\n", sb,
sb->s_count, atomic_read(&sb->s_active));
int ll_remount_fs(struct super_block *sb, int *flags, char *data)
{
struct ll_sb_info *sbi = ll_s2sbi(sb);
+ char *profilenm = get_profile_name(sb);
int err;
__u32 read_only;
if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
read_only = *flags & MS_RDONLY;
- err = obd_set_info_async(sbi->ll_md_exp,
+ err = obd_set_info_async(NULL, sbi->ll_md_exp,
sizeof(KEY_READ_ONLY),
KEY_READ_ONLY, sizeof(read_only),
&read_only, NULL);
if (err) {
- CERROR("Failed to change the read-only flag during "
- "remount: %d\n", err);
+ LCONSOLE_WARN("Failed to remount %s %s (%d)\n",
+ profilenm, read_only ?
+ "read-only" : "read-write", err);
return err;
}
sb->s_flags |= MS_RDONLY;
else
sb->s_flags &= ~MS_RDONLY;
+
+ if (sbi->ll_flags & LL_SBI_VERBOSE)
+ LCONSOLE_WARN("Remounted %s %s\n", profilenm,
+ read_only ? "read-only" : "read-write");
}
return 0;
}
struct ptlrpc_request *req,
struct super_block *sb)
{
- struct ll_sb_info *sbi = NULL;
- struct lustre_md md;
+ struct ll_sb_info *sbi = NULL;
+ struct lustre_md md;
+ __u64 ibits;
int rc;
ENTRY;
LASSERT(*inode || sb);
sbi = sb ? ll_s2sbi(sb) : ll_i2sbi(*inode);
- memset(&md, 0, sizeof(struct lustre_md));
-
rc = md_get_lustre_md(sbi->ll_md_exp, req, sbi->ll_dt_exp,
sbi->ll_md_exp, &md);
if (rc)
*/
LASSERT(fid_is_sane(&md.body->fid1));
- *inode = ll_iget(sb, cl_fid_build_ino(&md.body->fid1), &md);
+ *inode = ll_iget(sb, cl_fid_build_ino(&md.body->fid1, 0), &md);
if (*inode == NULL || IS_ERR(*inode)) {
if (md.lsm)
obd_free_memmd(sbi->ll_dt_exp, &md.lsm);
}
}
+ /* sanity check for LAYOUT lock. */
+ ibits = MDS_INODELOCK_LAYOUT;
+ if (S_ISREG(md.body->mode) && sbi->ll_flags & LL_SBI_LAYOUT_LOCK &&
+ md.lsm != NULL && !ll_have_md_lock(*inode, &ibits, LCK_MINMODE)) {
+ CERROR("%s: inode "DFID" (%p) layout lock not granted.\n",
+ ll_get_fsname(sb, NULL, 0),
+ PFID(ll_inode2fid(*inode)), *inode);
+ }
+
out:
- md_free_lustre_md(sbi->ll_md_exp, &md);
- RETURN(rc);
+ md_free_lustre_md(sbi->ll_md_exp, &md);
+ RETURN(rc);
}
int ll_obd_statfs(struct inode *inode, void *arg)
char *buf = NULL;
struct obd_ioctl_data *data = NULL;
__u32 type;
+ __u32 flags;
int len = 0, rc;
if (!inode || !(sbi = ll_i2sbi(inode)))
GOTO(out_statfs, rc = -EINVAL);
memcpy(&type, data->ioc_inlbuf1, sizeof(__u32));
- if (type == LL_STATFS_MDC)
+ if (type & LL_STATFS_LMV)
exp = sbi->ll_md_exp;
- else if (type == LL_STATFS_LOV)
+ else if (type & LL_STATFS_LOV)
exp = sbi->ll_dt_exp;
else
GOTO(out_statfs, rc = -ENODEV);
- rc = obd_iocontrol(IOC_OBD_STATFS, exp, len, buf, NULL);
+ flags = (type & LL_STATFS_NODELAY) ? OBD_STATFS_NODELAY : 0;
+ rc = obd_iocontrol(IOC_OBD_STATFS, exp, len, buf, &flags);
if (rc)
GOTO(out_statfs, rc);
out_statfs:
op_data->op_mds = 0;
op_data->op_data = data;
+ /* If the file is being opened after mknod() (normally due to NFS)
+ * try to use the default stripe data from parent directory for
+ * allocating OST objects. Try to pass the parent FID to MDS. */
+ if (opc == LUSTRE_OPC_CREATE && i1 == i2 && S_ISREG(i2->i_mode) &&
+ !ll_i2info(i2)->lli_has_smd) {
+ struct ll_inode_info *lli = ll_i2info(i2);
+
+ cfs_spin_lock(&lli->lli_lock);
+ if (likely(!lli->lli_has_smd && !fid_is_zero(&lli->lli_pfid)))
+ op_data->op_fid1 = lli->lli_pfid;
+ cfs_spin_unlock(&lli->lli_lock);
+ /** We ignore parent's capability temporary. */
+ }
+
return op_data;
}
if (sbi->ll_flags & LL_SBI_LAZYSTATFS)
seq_puts(seq, ",lazystatfs");
+ if (sbi->ll_flags & LL_SBI_USER_FID2PATH)
+ seq_puts(seq, ",user_fid2path");
+
RETURN(0);
}
+
+/**
+ * Get obd name by cmd, and copy out to user space
+ */
+int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg)
+{
+ struct ll_sb_info *sbi = ll_i2sbi(inode);
+ struct obd_device *obd;
+ ENTRY;
+
+ if (cmd == OBD_IOC_GETDTNAME)
+ obd = class_exp2obd(sbi->ll_dt_exp);
+ else if (cmd == OBD_IOC_GETMDNAME)
+ obd = class_exp2obd(sbi->ll_md_exp);
+ else
+ RETURN(-EINVAL);
+
+ if (!obd)
+ RETURN(-ENOENT);
+
+ if (cfs_copy_to_user((void *)arg, obd->obd_name,
+ strlen(obd->obd_name) + 1))
+ RETURN(-EFAULT);
+
+ RETURN(0);
+}
+
+/**
+ * Get lustre file system name by \a sbi. If \a buf is provided(non-NULL), the
+ * fsname will be returned in this buffer; otherwise, a static buffer will be
+ * used to store the fsname and returned to caller.
+ */
+char *ll_get_fsname(struct super_block *sb, char *buf, int buflen)
+{
+ static char fsname_static[MTI_NAME_MAXLEN];
+ struct lustre_sb_info *lsi = s2lsi(sb);
+ char *ptr;
+ int len;
+
+ if (buf == NULL) {
+ /* this means the caller wants to use static buffer
+ * and it doesn't care about race. Usually this is
+ * in error reporting path */
+ buf = fsname_static;
+ buflen = sizeof(fsname_static);
+ }
+
+ len = strlen(lsi->lsi_lmd->lmd_profile);
+ ptr = strrchr(lsi->lsi_lmd->lmd_profile, '-');
+ if (ptr && (strcmp(ptr, "-client") == 0))
+ len -= 7;
+
+ if (unlikely(len >= buflen))
+ len = buflen - 1;
+ strncpy(buf, lsi->lsi_lmd->lmd_profile, len);
+ buf[len] = '\0';
+
+ return buf;
+}
+
+static char* ll_d_path(struct dentry *dentry, char *buf, int bufsize)
+{
+ char *path = NULL;
+
+#ifdef HAVE_FS_STRUCT_USE_PATH
+ struct path p;
+
+ p.dentry = dentry;
+ p.mnt = current->fs->root.mnt;
+ path_get(&p);
+ path = d_path(&p, buf, bufsize);
+ path_put(&p);
+#else
+ path = d_path(dentry, current->fs->rootmnt, buf, bufsize);
+#endif
+
+ return path;
+}
+
+void ll_dirty_page_discard_warn(cfs_page_t *page, int ioret)
+{
+ char *buf, *path = NULL;
+ struct dentry *dentry = NULL;
+ struct ccc_object *obj = cl_inode2ccc(page->mapping->host);
+
+ buf = (char *)__get_free_page(GFP_KERNEL);
+ if (buf != NULL) {
+ dentry = d_find_alias(page->mapping->host);
+ if (dentry != NULL)
+ path = ll_d_path(dentry, buf, PAGE_SIZE);
+ }
+
+ CWARN("%s: dirty page discard: %s/fid: "DFID"/%s may get corrupted "
+ "(rc %d)\n", ll_get_fsname(page->mapping->host->i_sb, NULL, 0),
+ s2lsi(page->mapping->host->i_sb)->lsi_lmd->lmd_dev,
+ PFID(&obj->cob_header.coh_lu.loh_fid),
+ (path && !IS_ERR(path)) ? path : "", ioret);
+
+ if (dentry != NULL)
+ dput(dentry);
+
+ if (buf != NULL)
+ free_page((unsigned long)buf);
+}