Commit 08bc2dbc authored by Adrian Bunk's avatar Adrian Bunk Committed by Steven Whitehouse

[GFS2] [-mm patch] fs/gfs2/: possible cleanups

This patch contains the following possible cleanups:
- make needlessly global code static
- #if 0 unused functions
- remove the following global function that was both unused and
  unimplemented:
  - super.c: gfs2_do_upgrade()
Signed-off-by: default avatarAdrian Bunk <bunk@stusta.de>
Signed-off-by: default avatarSteven Whitehouse <swhiteho@redhat.com>
parent c56b39cd
......@@ -59,7 +59,7 @@ struct strip_mine {
*
* Returns: errno
*/
#if 0
int gfs2_unstuffer_sync(struct gfs2_inode *ip, struct buffer_head *dibh,
uint64_t block, void *private)
{
......@@ -77,6 +77,7 @@ int gfs2_unstuffer_sync(struct gfs2_inode *ip, struct buffer_head *dibh,
return error;
}
#endif /* 0 */
/**
* gfs2_unstuff_dinode - Unstuff a dinode when the data has grown too big
......
......@@ -13,8 +13,6 @@
typedef int (*gfs2_unstuffer_t) (struct gfs2_inode * ip,
struct buffer_head * dibh, uint64_t block,
void *private);
int gfs2_unstuffer_sync(struct gfs2_inode *ip, struct buffer_head *dibh,
uint64_t block, void *private);
int gfs2_unstuff_dinode(struct gfs2_inode *ip, gfs2_unstuffer_t unstuffer,
void *private);
......
......@@ -669,10 +669,10 @@ static void dirent_del(struct gfs2_inode *dip, struct buffer_head *bh,
* Takes a dent from which to grab space as an argument. Returns the
* newly created dent.
*/
struct gfs2_dirent *gfs2_init_dirent(struct inode *inode,
struct gfs2_dirent *dent,
const struct qstr *name,
struct buffer_head *bh)
static struct gfs2_dirent *gfs2_init_dirent(struct inode *inode,
struct gfs2_dirent *dent,
const struct qstr *name,
struct buffer_head *bh)
{
struct gfs2_inode *ip = inode->u.generic_ip;
struct gfs2_dirent *ndent;
......
......@@ -167,7 +167,7 @@ static int system_eo_remove(struct gfs2_inode *ip, struct gfs2_ea_request *er)
return gfs2_ea_remove_i(ip, er);
}
struct gfs2_eattr_operations gfs2_user_eaops = {
static struct gfs2_eattr_operations gfs2_user_eaops = {
.eo_get = user_eo_get,
.eo_set = user_eo_set,
.eo_remove = user_eo_remove,
......
......@@ -21,7 +21,6 @@ struct gfs2_eattr_operations {
unsigned int gfs2_ea_name2type(const char *name, char **truncated_name);
extern struct gfs2_eattr_operations gfs2_user_eaops;
extern struct gfs2_eattr_operations gfs2_system_eaops;
extern struct gfs2_eattr_operations *gfs2_ea_ops[];
......
......@@ -358,6 +358,7 @@ static int ea_remove_unstuffed(struct gfs2_inode *ip, struct buffer_head *bh,
return error;
}
#if 0
static int gfs2_ea_repack_i(struct gfs2_inode *ip)
{
......@@ -382,6 +383,8 @@ int gfs2_ea_repack(struct gfs2_inode *ip)
return error;
}
#endif /* 0 */
struct ea_list {
struct gfs2_ea_request *ei_er;
unsigned int ei_size;
......
......@@ -61,8 +61,6 @@ struct gfs2_ea_location {
struct gfs2_ea_header *el_prev;
};
int gfs2_ea_repack(struct gfs2_inode *ip);
int gfs2_ea_get_i(struct gfs2_inode *ip, struct gfs2_ea_request *er);
int gfs2_ea_set_i(struct gfs2_inode *ip, struct gfs2_ea_request *er);
int gfs2_ea_remove_i(struct gfs2_inode *ip, struct gfs2_ea_request *er);
......
......@@ -47,6 +47,8 @@ struct greedy {
typedef void (*glock_examiner) (struct gfs2_glock * gl);
static int gfs2_dump_lockstate(struct gfs2_sbd *sdp);
/**
* relaxed_state_ok - is a requested lock compatible with the current lock mode?
* @actual: the current state of the lock
......@@ -228,8 +230,8 @@ static struct gfs2_glock *search_bucket(struct gfs2_gl_hash_bucket *bucket,
* Returns: NULL, or the struct gfs2_glock with the requested number
*/
struct gfs2_glock *gfs2_glock_find(struct gfs2_sbd *sdp,
struct lm_lockname *name)
static struct gfs2_glock *gfs2_glock_find(struct gfs2_sbd *sdp,
struct lm_lockname *name)
{
struct gfs2_gl_hash_bucket *bucket = &sdp->sd_gl_hash[gl_hash(name)];
struct gfs2_glock *gl;
......@@ -421,8 +423,9 @@ void gfs2_holder_uninit(struct gfs2_holder *gh)
* Returns: the holder structure, NULL on ENOMEM
*/
struct gfs2_holder *gfs2_holder_get(struct gfs2_glock *gl, unsigned int state,
int flags, gfp_t gfp_flags)
static struct gfs2_holder *gfs2_holder_get(struct gfs2_glock *gl,
unsigned int state,
int flags, gfp_t gfp_flags)
{
struct gfs2_holder *gh;
......@@ -442,7 +445,7 @@ struct gfs2_holder *gfs2_holder_get(struct gfs2_glock *gl, unsigned int state,
*
*/
void gfs2_holder_put(struct gfs2_holder *gh)
static void gfs2_holder_put(struct gfs2_holder *gh)
{
gfs2_holder_uninit(gh);
kfree(gh);
......@@ -674,7 +677,7 @@ void gfs2_glmutex_lock(struct gfs2_glock *gl)
* Returns: 1 if the glock is acquired
*/
int gfs2_glmutex_trylock(struct gfs2_glock *gl)
static int gfs2_glmutex_trylock(struct gfs2_glock *gl)
{
int acquired = 1;
......@@ -1301,7 +1304,8 @@ void gfs2_glock_dq(struct gfs2_holder *gh)
*
*/
void gfs2_glock_prefetch(struct gfs2_glock *gl, unsigned int state, int flags)
static void gfs2_glock_prefetch(struct gfs2_glock *gl, unsigned int state,
int flags)
{
struct gfs2_glock_operations *glops = gl->gl_ops;
......@@ -1329,7 +1333,7 @@ void gfs2_glock_prefetch(struct gfs2_glock *gl, unsigned int state, int flags)
* @gl: the glock
*
*/
#if 0
void gfs2_glock_force_drop(struct gfs2_glock *gl)
{
struct gfs2_holder gh;
......@@ -1345,6 +1349,7 @@ void gfs2_glock_force_drop(struct gfs2_glock *gl)
wait_for_completion(&gh.gh_wait);
gfs2_holder_uninit(&gh);
}
#endif /* 0 */
static void greedy_work(void *data)
{
......@@ -1697,6 +1702,7 @@ void gfs2_lvb_unhold(struct gfs2_glock *gl)
gfs2_glock_put(gl);
}
#if 0
void gfs2_lvb_sync(struct gfs2_glock *gl)
{
gfs2_glmutex_lock(gl);
......@@ -1707,6 +1713,7 @@ void gfs2_lvb_sync(struct gfs2_glock *gl)
gfs2_glmutex_unlock(gl);
}
#endif /* 0 */
static void blocking_cb(struct gfs2_sbd *sdp, struct lm_lockname *name,
unsigned int state)
......@@ -2307,7 +2314,7 @@ static int dump_glock(struct gfs2_glock *gl)
*
*/
int gfs2_dump_lockstate(struct gfs2_sbd *sdp)
static int gfs2_dump_lockstate(struct gfs2_sbd *sdp)
{
struct gfs2_gl_hash_bucket *bucket;
struct gfs2_glock *gl;
......
......@@ -73,8 +73,6 @@ static inline int gfs2_glock_is_blocking(struct gfs2_glock *gl)
return ret;
}
struct gfs2_glock *gfs2_glock_find(struct gfs2_sbd *sdp,
struct lm_lockname *name);
int gfs2_glock_get(struct gfs2_sbd *sdp,
uint64_t number, struct gfs2_glock_operations *glops,
int create, struct gfs2_glock **glp);
......@@ -85,15 +83,11 @@ void gfs2_holder_init(struct gfs2_glock *gl, unsigned int state, unsigned flags,
void gfs2_holder_reinit(unsigned int state, unsigned flags,
struct gfs2_holder *gh);
void gfs2_holder_uninit(struct gfs2_holder *gh);
struct gfs2_holder *gfs2_holder_get(struct gfs2_glock *gl, unsigned int state,
int flags, gfp_t gfp_flags);
void gfs2_holder_put(struct gfs2_holder *gh);
void gfs2_glock_xmote_th(struct gfs2_glock *gl, unsigned int state, int flags);
void gfs2_glock_drop_th(struct gfs2_glock *gl);
void gfs2_glmutex_lock(struct gfs2_glock *gl);
int gfs2_glmutex_trylock(struct gfs2_glock *gl);
void gfs2_glmutex_unlock(struct gfs2_glock *gl);
int gfs2_glock_nq(struct gfs2_holder *gh);
......@@ -101,9 +95,6 @@ int gfs2_glock_poll(struct gfs2_holder *gh);
int gfs2_glock_wait(struct gfs2_holder *gh);
void gfs2_glock_dq(struct gfs2_holder *gh);
void gfs2_glock_prefetch(struct gfs2_glock *gl, unsigned int state, int flags);
void gfs2_glock_force_drop(struct gfs2_glock *gl);
int gfs2_glock_be_greedy(struct gfs2_glock *gl, unsigned int time);
void gfs2_glock_dq_uninit(struct gfs2_holder *gh);
......@@ -148,7 +139,6 @@ static inline int gfs2_glock_nq_init(struct gfs2_glock *gl,
int gfs2_lvb_hold(struct gfs2_glock *gl);
void gfs2_lvb_unhold(struct gfs2_glock *gl);
void gfs2_lvb_sync(struct gfs2_glock *gl);
void gfs2_glock_cb(lm_fsdata_t *fsdata, unsigned int type, void *data);
......@@ -161,6 +151,4 @@ void gfs2_reclaim_glock(struct gfs2_sbd *sdp);
void gfs2_scand_internal(struct gfs2_sbd *sdp);
void gfs2_gl_hash_clear(struct gfs2_sbd *sdp, int wait);
int gfs2_dump_lockstate(struct gfs2_sbd *sdp);
#endif /* __GLOCK_DOT_H__ */
......@@ -188,11 +188,13 @@ void gfs2_lm_unhold_lvb(struct gfs2_sbd *sdp, lm_lock_t *lock, char *lvb)
sdp->sd_lockstruct.ls_ops->lm_unhold_lvb(lock, lvb);
}
#if 0
void gfs2_lm_sync_lvb(struct gfs2_sbd *sdp, lm_lock_t *lock, char *lvb)
{
if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
sdp->sd_lockstruct.ls_ops->lm_sync_lvb(lock, lvb);
}
#endif /* 0 */
int gfs2_lm_plock_get(struct gfs2_sbd *sdp, struct lm_lockname *name,
struct file *file, struct file_lock *fl)
......
......@@ -26,7 +26,6 @@ unsigned int gfs2_lm_unlock(struct gfs2_sbd *sdp, lm_lock_t *lock,
void gfs2_lm_cancel(struct gfs2_sbd *sdp, lm_lock_t *lock);
int gfs2_lm_hold_lvb(struct gfs2_sbd *sdp, lm_lock_t *lock, char **lvbp);
void gfs2_lm_unhold_lvb(struct gfs2_sbd *sdp, lm_lock_t *lock, char *lvb);
void gfs2_lm_sync_lvb(struct gfs2_sbd *sdp, lm_lock_t *lock, char *lvb);
int gfs2_lm_plock_get(struct gfs2_sbd *sdp,
struct lm_lockname *name,
struct file *file, struct file_lock *fl);
......
......@@ -158,8 +158,8 @@ static inline void make_strname(struct lm_lockname *lockname,
str->namelen = GDLM_STRNAME_BYTES;
}
int gdlm_create_lp(struct gdlm_ls *ls, struct lm_lockname *name,
struct gdlm_lock **lpp)
static int gdlm_create_lp(struct gdlm_ls *ls, struct lm_lockname *name,
struct gdlm_lock **lpp)
{
struct gdlm_lock *lp;
......@@ -276,7 +276,7 @@ unsigned int gdlm_do_lock(struct gdlm_lock *lp)
return LM_OUT_ASYNC;
}
unsigned int gdlm_do_unlock(struct gdlm_lock *lp)
static unsigned int gdlm_do_unlock(struct gdlm_lock *lp)
{
struct gdlm_ls *ls = lp->ls;
unsigned int lkf = 0;
......@@ -378,7 +378,7 @@ void gdlm_cancel(lm_lock_t *lock)
clear_bit(LFL_DLM_CANCEL, &lp->flags);
}
int gdlm_add_lvb(struct gdlm_lock *lp)
static int gdlm_add_lvb(struct gdlm_lock *lp)
{
char *lvb;
......@@ -391,7 +391,7 @@ int gdlm_add_lvb(struct gdlm_lock *lp)
return 0;
}
void gdlm_del_lvb(struct gdlm_lock *lp)
static void gdlm_del_lvb(struct gdlm_lock *lp)
{
kfree(lp->lvb);
lp->lvb = NULL;
......
......@@ -162,12 +162,8 @@ int16_t gdlm_make_lmstate(int16_t);
void gdlm_queue_delayed(struct gdlm_lock *);
void gdlm_submit_delayed(struct gdlm_ls *);
int gdlm_release_all_locks(struct gdlm_ls *);
int gdlm_create_lp(struct gdlm_ls *, struct lm_lockname *, struct gdlm_lock **);
void gdlm_delete_lp(struct gdlm_lock *);
int gdlm_add_lvb(struct gdlm_lock *);
void gdlm_del_lvb(struct gdlm_lock *);
unsigned int gdlm_do_lock(struct gdlm_lock *);
unsigned int gdlm_do_unlock(struct gdlm_lock *);
int gdlm_get_lock(lm_lockspace_t *, struct lm_lockname *, lm_lock_t **);
void gdlm_put_lock(lm_lock_t *);
......
......@@ -16,7 +16,7 @@ extern int gdlm_drop_period;
extern struct lm_lockops gdlm_ops;
int __init init_lock_dlm(void)
static int __init init_lock_dlm(void)
{
int error;
......@@ -48,7 +48,7 @@ int __init init_lock_dlm(void)
return 0;
}
void __exit exit_lock_dlm(void)
static void __exit exit_lock_dlm(void)
{
gdlm_plock_exit();
gdlm_sysfs_exit();
......
......@@ -21,7 +21,7 @@ struct nolock_lockspace {
unsigned int nl_lvb_size;
};
struct lm_lockops nolock_ops;
static struct lm_lockops nolock_ops;
static int nolock_mount(char *table_name, char *host_data,
lm_callback_t cb, lm_fsdata_t *fsdata,
......@@ -208,7 +208,7 @@ static void nolock_recovery_done(lm_lockspace_t *lockspace, unsigned int jid,
{
}
struct lm_lockops nolock_ops = {
static struct lm_lockops nolock_ops = {
.lm_proto_name = "lock_nolock",
.lm_mount = nolock_mount,
.lm_others_may_mount = nolock_others_may_mount,
......@@ -229,7 +229,7 @@ struct lm_lockops nolock_ops = {
.lm_owner = THIS_MODULE,
};
int __init init_nolock(void)
static int __init init_nolock(void)
{
int error;
......@@ -245,7 +245,7 @@ int __init init_nolock(void)
return 0;
}
void __exit exit_nolock(void)
static void __exit exit_nolock(void)
{
gfs_unregister_lockproto(&nolock_ops);
}
......
......@@ -43,6 +43,7 @@ void gfs2_quota_lvb_out(struct gfs2_quota_lvb *qb, char *lvb)
str->qb_value = cpu_to_be64(qb->qb_value);
}
#if 0
void gfs2_quota_lvb_print(struct gfs2_quota_lvb *qb)
{
pv(qb, qb_magic, "%u");
......@@ -50,4 +51,5 @@ void gfs2_quota_lvb_print(struct gfs2_quota_lvb *qb)
pv(qb, qb_warn, "%llu");
pv(qb, qb_value, "%lld");
}
#endif /* 0 */
......@@ -14,7 +14,6 @@
void gfs2_quota_lvb_in(struct gfs2_quota_lvb *qb, char *lvb);
void gfs2_quota_lvb_out(struct gfs2_quota_lvb *qb, char *lvb);
void gfs2_quota_lvb_print(struct gfs2_quota_lvb *qb);
#endif /* __LVB_DOT_H__ */
......@@ -28,7 +28,7 @@
* @count: the number of bytes
*
*/
#if 0
static void print_array(char *title, char *buf, int count)
{
int x;
......@@ -42,6 +42,7 @@ static void print_array(char *title, char *buf, int count)
if (x % 16)
printk("\n");
}
#endif /* 0 */
/*
* gfs2_xxx_in - read in an xxx struct
......@@ -72,7 +73,7 @@ void gfs2_inum_out(const struct gfs2_inum *no, char *buf)
str->no_addr = cpu_to_be64(no->no_addr);
}
void gfs2_inum_print(struct gfs2_inum *no)
static void gfs2_inum_print(struct gfs2_inum *no)
{
pv(no, no_formal_ino, "%llu");
pv(no, no_addr, "%llu");
......@@ -96,7 +97,7 @@ static void gfs2_meta_header_out(struct gfs2_meta_header *mh, char *buf)
str->mh_format = cpu_to_be32(mh->mh_format);
}
void gfs2_meta_header_print(struct gfs2_meta_header *mh)
static void gfs2_meta_header_print(struct gfs2_meta_header *mh)
{
pv(mh, mh_magic, "0x%.8X");
pv(mh, mh_type, "%u");
......@@ -121,6 +122,7 @@ void gfs2_sb_in(struct gfs2_sb *sb, char *buf)
memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
}
#if 0
void gfs2_sb_print(struct gfs2_sb *sb)
{
gfs2_meta_header_print(&sb->sb_header);
......@@ -136,6 +138,7 @@ void gfs2_sb_print(struct gfs2_sb *sb)
pv(sb, sb_lockproto, "%s");
pv(sb, sb_locktable, "%s");
}
#endif /* 0 */
void gfs2_rindex_in(struct gfs2_rindex *ri, char *buf)
{
......@@ -149,6 +152,7 @@ void gfs2_rindex_in(struct gfs2_rindex *ri, char *buf)
}
#if 0
void gfs2_rindex_out(struct gfs2_rindex *ri, char *buf)
{
struct gfs2_rindex *str = (struct gfs2_rindex *)buf;
......@@ -163,6 +167,8 @@ void gfs2_rindex_out(struct gfs2_rindex *ri, char *buf)
memset(str->ri_reserved, 0, sizeof(str->ri_reserved));
}
#endif /* 0 */
void gfs2_rindex_print(struct gfs2_rindex *ri)
{
pv(ri, ri_addr, "%llu");
......@@ -196,6 +202,7 @@ void gfs2_rgrp_out(struct gfs2_rgrp *rg, char *buf)
memset(&str->rg_reserved, 0, sizeof(str->rg_reserved));
}
#if 0
void gfs2_rgrp_print(struct gfs2_rgrp *rg)
{
gfs2_meta_header_print(&rg->rg_header);
......@@ -205,6 +212,7 @@ void gfs2_rgrp_print(struct gfs2_rgrp *rg)
pa(rg, rg_reserved, 36);
}
#endif /* 0 */
void gfs2_quota_in(struct gfs2_quota *qu, char *buf)
{
......@@ -215,6 +223,8 @@ void gfs2_quota_in(struct gfs2_quota *qu, char *buf)
qu->qu_value = be64_to_cpu(str->qu_value);
}
#if 0
void gfs2_quota_out(struct gfs2_quota *qu, char *buf)
{
struct gfs2_quota *str = (struct gfs2_quota *)buf;
......@@ -231,6 +241,8 @@ void gfs2_quota_print(struct gfs2_quota *qu)
pv(qu, qu_value, "%lld");
}
#endif /* 0 */
void gfs2_dinode_in(struct gfs2_dinode *di, char *buf)
{
struct gfs2_dinode *str = (struct gfs2_dinode *)buf;
......@@ -327,6 +339,8 @@ void gfs2_dinode_print(struct gfs2_dinode *di)
pv(di, di_eattr, "%llu");
}
#if 0
void gfs2_dirent_print(struct gfs2_dirent *de, char *name)
{
char buf[GFS2_FNAMESIZE + 1];
......@@ -394,6 +408,8 @@ void gfs2_ea_header_print(struct gfs2_ea_header *ea, char *name)
printk(KERN_INFO " name = %s\n", buf);
}
#endif /* 0 */
void gfs2_log_header_in(struct gfs2_log_header *lh, char *buf)
{
struct gfs2_log_header *str = (struct gfs2_log_header *)buf;
......@@ -406,6 +422,8 @@ void gfs2_log_header_in(struct gfs2_log_header *lh, char *buf)
lh->lh_hash = be32_to_cpu(str->lh_hash);
}
#if 0
void gfs2_log_header_print(struct gfs2_log_header *lh)
{
gfs2_meta_header_print(&lh->lh_header);
......@@ -427,6 +445,8 @@ void gfs2_log_descriptor_print(struct gfs2_log_descriptor *ld)
pa(ld, ld_reserved, 32);
}
#endif /* 0 */
void gfs2_inum_range_in(struct gfs2_inum_range *ir, char *buf)
{
struct gfs2_inum_range *str = (struct gfs2_inum_range *)buf;
......@@ -443,11 +463,13 @@ void gfs2_inum_range_out(struct gfs2_inum_range *ir, char *buf)
str->ir_length = cpu_to_be64(ir->ir_length);
}
#if 0
void gfs2_inum_range_print(struct gfs2_inum_range *ir)
{
pv(ir, ir_start, "%llu");
pv(ir, ir_length, "%llu");
}
#endif /* 0 */
void gfs2_statfs_change_in(struct gfs2_statfs_change *sc, char *buf)
{
......@@ -467,12 +489,14 @@ void gfs2_statfs_change_out(struct gfs2_statfs_change *sc, char *buf)
str->sc_dinodes = cpu_to_be64(sc->sc_dinodes);
}
#if 0
void gfs2_statfs_change_print(struct gfs2_statfs_change *sc)
{
pv(sc, sc_total, "%lld");
pv(sc, sc_free, "%lld");
pv(sc, sc_dinodes, "%lld");
}
#endif /* 0 */
void gfs2_unlinked_tag_in(struct gfs2_unlinked_tag *ut, char *buf)
{
......@@ -491,12 +515,16 @@ void gfs2_unlinked_tag_out(struct gfs2_unlinked_tag *ut, char *buf)
str->__pad = 0;
}
#if 0
void gfs2_unlinked_tag_print(struct gfs2_unlinked_tag *ut)
{
gfs2_inum_print(&ut->ut_inum);
pv(ut, ut_flags, "%u");
}
#endif /* 0 */
void gfs2_quota_change_in(struct gfs2_quota_change *qc, char *buf)
{
struct gfs2_quota_change *str = (struct gfs2_quota_change *)buf;
......@@ -506,6 +534,8 @@ void gfs2_quota_change_in(struct gfs2_quota_change *qc, char *buf)
qc->qc_id = be32_to_cpu(str->qc_id);
}
#if 0
void gfs2_quota_change_print(struct gfs2_quota_change *qc)
{
pv(qc, qc_change, "%lld");
......@@ -513,5 +543,5 @@ void gfs2_quota_change_print(struct gfs2_quota_change *qc)
pv(qc, qc_id, "%u");
}
#endif /* 0 */
......@@ -1086,6 +1086,7 @@ int gfs2_quota_refresh(struct gfs2_sbd *sdp, int user, uint32_t id)
return error;
}
#if 0
int gfs2_quota_read(struct gfs2_sbd *sdp, int user, uint32_t id,
struct gfs2_quota *q)
{
......@@ -1121,6 +1122,7 @@ int gfs2_quota_read(struct gfs2_sbd *sdp, int user, uint32_t id,
return error;
}
#endif /* 0 */
int gfs2_quota_init(struct gfs2_sbd *sdp)
{
......
......@@ -24,8 +24,6 @@ void gfs2_quota_change(struct gfs2_inode *ip, int64_t change,
int gfs2_quota_sync(struct gfs2_sbd *sdp);
int gfs2_quota_refresh(struct gfs2_sbd *sdp, int user, uint32_t id);
int gfs2_quota_read(struct gfs2_sbd *sdp, int user, uint32_t id,
struct gfs2_quota *q);
int gfs2_quota_init(struct gfs2_sbd *sdp);
void gfs2_quota_scan(struct gfs2_sbd *sdp);
......
......@@ -264,11 +264,6 @@ int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent)
return 0;
}
int gfs2_do_upgrade(struct gfs2_sbd *sdp, struct gfs2_glock *sb_gl)
{
return 0;
}
/**
* gfs2_jindex_hold - Grab a lock on the jindex
* @sdp: The GFS2 superblock
......@@ -837,7 +832,8 @@ struct lfcc {
* Returns: errno
*/
int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp, struct gfs2_holder *t_gh)
static int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp,
struct gfs2_holder *t_gh)
{
struct gfs2_inode *ip;
struct gfs2_holder ji_gh;
......
......@@ -14,7 +14,6 @@ void gfs2_tune_init(struct gfs2_tune *gt);
int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb *sb, int silent);
int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent);
int gfs2_do_upgrade(struct gfs2_sbd *sdp, struct gfs2_glock *gl_sb);
static inline unsigned int gfs2_jindex_size(struct gfs2_sbd *sdp)
{
......@@ -46,7 +45,6 @@ int gfs2_statfs_sync(struct gfs2_sbd *sdp);
int gfs2_statfs_i(struct gfs2_sbd *sdp, struct gfs2_statfs_change *sc);
int gfs2_statfs_slow(struct gfs2_sbd *sdp, struct gfs2_statfs_change *sc);
int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp, struct gfs2_holder *t_gh);
int gfs2_freeze_fs(struct gfs2_sbd *sdp);
void gfs2_unfreeze_fs(struct gfs2_sbd *sdp);
......