Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jlbec/ocfs2:
  ocfs2: Fix system inodes cache overflow.
  ocfs2: Hold ip_lock when set/clear flags for indexed dir.
  ocfs2: Adjust masklog flag values
  Ocfs2: Teach 'coherency=full' O_DIRECT writes to correctly up_read i_alloc_sem.
  ocfs2/dlm: Migrate lockres with no locks if it has a reference
This commit is contained in:
Linus Torvalds 2010-12-23 16:36:48 -08:00
commit eda4b716ea
8 changed files with 81 additions and 28 deletions

View File

@ -573,11 +573,14 @@ static void ocfs2_dio_end_io(struct kiocb *iocb,
/* this io's submitter should not have unlocked this before we could */
BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
if (ocfs2_iocb_is_sem_locked(iocb)) {
up_read(&inode->i_alloc_sem);
ocfs2_iocb_clear_sem_locked(iocb);
}
ocfs2_iocb_clear_rw_locked(iocb);
level = ocfs2_iocb_rw_locked_level(iocb);
if (!level)
up_read(&inode->i_alloc_sem);
ocfs2_rw_unlock(inode, level);
if (is_async)

View File

@ -68,8 +68,27 @@ static inline void ocfs2_iocb_set_rw_locked(struct kiocb *iocb, int level)
else
clear_bit(1, (unsigned long *)&iocb->private);
}
/*
* Using a named enum representing lock types in terms of #N bit stored in
* iocb->private, which is going to be used for communication bewteen
* ocfs2_dio_end_io() and ocfs2_file_aio_write/read().
*/
enum ocfs2_iocb_lock_bits {
OCFS2_IOCB_RW_LOCK = 0,
OCFS2_IOCB_RW_LOCK_LEVEL,
OCFS2_IOCB_SEM,
OCFS2_IOCB_NUM_LOCKS
};
#define ocfs2_iocb_clear_rw_locked(iocb) \
clear_bit(0, (unsigned long *)&iocb->private)
clear_bit(OCFS2_IOCB_RW_LOCK, (unsigned long *)&iocb->private)
#define ocfs2_iocb_rw_locked_level(iocb) \
test_bit(1, (unsigned long *)&iocb->private)
test_bit(OCFS2_IOCB_RW_LOCK_LEVEL, (unsigned long *)&iocb->private)
#define ocfs2_iocb_set_sem_locked(iocb) \
set_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private)
#define ocfs2_iocb_clear_sem_locked(iocb) \
clear_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private)
#define ocfs2_iocb_is_sem_locked(iocb) \
test_bit(OCFS2_IOCB_SEM, (unsigned long *)&iocb->private)
#endif /* OCFS2_FILE_H */

View File

@ -113,10 +113,11 @@ static struct mlog_attribute mlog_attrs[MLOG_MAX_BITS] = {
define_mask(QUOTA),
define_mask(REFCOUNT),
define_mask(BASTS),
define_mask(RESERVATIONS),
define_mask(CLUSTER),
define_mask(ERROR),
define_mask(NOTICE),
define_mask(KTHREAD),
define_mask(RESERVATIONS),
};
static struct attribute *mlog_attr_ptrs[MLOG_MAX_BITS] = {NULL, };

View File

@ -81,7 +81,7 @@
#include <linux/sched.h>
/* bits that are frequently given and infrequently matched in the low word */
/* NOTE: If you add a flag, you need to also update mlog.c! */
/* NOTE: If you add a flag, you need to also update masklog.c! */
#define ML_ENTRY 0x0000000000000001ULL /* func call entry */
#define ML_EXIT 0x0000000000000002ULL /* func call exit */
#define ML_TCP 0x0000000000000004ULL /* net cluster/tcp.c */
@ -114,13 +114,14 @@
#define ML_XATTR 0x0000000020000000ULL /* ocfs2 extended attributes */
#define ML_QUOTA 0x0000000040000000ULL /* ocfs2 quota operations */
#define ML_REFCOUNT 0x0000000080000000ULL /* refcount tree operations */
#define ML_BASTS 0x0000001000000000ULL /* dlmglue asts and basts */
#define ML_BASTS 0x0000000100000000ULL /* dlmglue asts and basts */
#define ML_RESERVATIONS 0x0000000200000000ULL /* ocfs2 alloc reservations */
#define ML_CLUSTER 0x0000000400000000ULL /* cluster stack */
/* bits that are infrequently given and frequently matched in the high word */
#define ML_ERROR 0x0000000100000000ULL /* sent to KERN_ERR */
#define ML_NOTICE 0x0000000200000000ULL /* setn to KERN_NOTICE */
#define ML_KTHREAD 0x0000000400000000ULL /* kernel thread activity */
#define ML_RESERVATIONS 0x0000000800000000ULL /* ocfs2 alloc reservations */
#define ML_CLUSTER 0x0000001000000000ULL /* cluster stack */
#define ML_ERROR 0x1000000000000000ULL /* sent to KERN_ERR */
#define ML_NOTICE 0x2000000000000000ULL /* setn to KERN_NOTICE */
#define ML_KTHREAD 0x4000000000000000ULL /* kernel thread activity */
#define MLOG_INITIAL_AND_MASK (ML_ERROR|ML_NOTICE)
#define MLOG_INITIAL_NOT_MASK (ML_ENTRY|ML_EXIT)

View File

@ -2461,8 +2461,10 @@ static int ocfs2_dx_dir_attach_index(struct ocfs2_super *osb,
di->i_dx_root = cpu_to_le64(dr_blkno);
spin_lock(&OCFS2_I(dir)->ip_lock);
OCFS2_I(dir)->ip_dyn_features |= OCFS2_INDEXED_DIR_FL;
di->i_dyn_features = cpu_to_le16(OCFS2_I(dir)->ip_dyn_features);
spin_unlock(&OCFS2_I(dir)->ip_lock);
ocfs2_journal_dirty(handle, di_bh);
@ -4466,8 +4468,10 @@ static int ocfs2_dx_dir_remove_index(struct inode *dir,
goto out_commit;
}
spin_lock(&OCFS2_I(dir)->ip_lock);
OCFS2_I(dir)->ip_dyn_features &= ~OCFS2_INDEXED_DIR_FL;
di->i_dyn_features = cpu_to_le16(OCFS2_I(dir)->ip_dyn_features);
spin_unlock(&OCFS2_I(dir)->ip_lock);
di->i_dx_root = cpu_to_le64(0ULL);
ocfs2_journal_dirty(handle, di_bh);

View File

@ -2346,7 +2346,8 @@ static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data)
*/
static int dlm_is_lockres_migrateable(struct dlm_ctxt *dlm,
struct dlm_lock_resource *res,
int *numlocks)
int *numlocks,
int *hasrefs)
{
int ret;
int i;
@ -2356,6 +2357,9 @@ static int dlm_is_lockres_migrateable(struct dlm_ctxt *dlm,
assert_spin_locked(&res->spinlock);
*numlocks = 0;
*hasrefs = 0;
ret = -EINVAL;
if (res->owner == DLM_LOCK_RES_OWNER_UNKNOWN) {
mlog(0, "cannot migrate lockres with unknown owner!\n");
@ -2386,7 +2390,13 @@ static int dlm_is_lockres_migrateable(struct dlm_ctxt *dlm,
}
*numlocks = count;
mlog(0, "migrateable lockres having %d locks\n", *numlocks);
count = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
if (count < O2NM_MAX_NODES)
*hasrefs = 1;
mlog(0, "%s: res %.*s, Migrateable, locks %d, refs %d\n", dlm->name,
res->lockname.len, res->lockname.name, *numlocks, *hasrefs);
leave:
return ret;
@ -2408,7 +2418,7 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
const char *name;
unsigned int namelen;
int mle_added = 0;
int numlocks;
int numlocks, hasrefs;
int wake = 0;
if (!dlm_grab(dlm))
@ -2417,13 +2427,13 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
name = res->lockname.name;
namelen = res->lockname.len;
mlog(0, "migrating %.*s to %u\n", namelen, name, target);
mlog(0, "%s: Migrating %.*s to %u\n", dlm->name, namelen, name, target);
/*
* ensure this lockres is a proper candidate for migration
*/
spin_lock(&res->spinlock);
ret = dlm_is_lockres_migrateable(dlm, res, &numlocks);
ret = dlm_is_lockres_migrateable(dlm, res, &numlocks, &hasrefs);
if (ret < 0) {
spin_unlock(&res->spinlock);
goto leave;
@ -2431,10 +2441,8 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
spin_unlock(&res->spinlock);
/* no work to do */
if (numlocks == 0) {
mlog(0, "no locks were found on this lockres! done!\n");
if (numlocks == 0 && !hasrefs)
goto leave;
}
/*
* preallocate up front
@ -2459,14 +2467,14 @@ static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
* find a node to migrate the lockres to
*/
mlog(0, "picking a migration node\n");
spin_lock(&dlm->spinlock);
/* pick a new node */
if (!test_bit(target, dlm->domain_map) ||
target >= O2NM_MAX_NODES) {
target = dlm_pick_migration_target(dlm, res);
}
mlog(0, "node %u chosen for migration\n", target);
mlog(0, "%s: res %.*s, Node %u chosen for migration\n", dlm->name,
namelen, name, target);
if (target >= O2NM_MAX_NODES ||
!test_bit(target, dlm->domain_map)) {
@ -2667,7 +2675,7 @@ int dlm_empty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
{
int ret;
int lock_dropped = 0;
int numlocks;
int numlocks, hasrefs;
spin_lock(&res->spinlock);
if (res->owner != dlm->node_num) {
@ -2681,8 +2689,8 @@ int dlm_empty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
}
/* No need to migrate a lockres having no locks */
ret = dlm_is_lockres_migrateable(dlm, res, &numlocks);
if (ret >= 0 && numlocks == 0) {
ret = dlm_is_lockres_migrateable(dlm, res, &numlocks, &hasrefs);
if (ret >= 0 && numlocks == 0 && !hasrefs) {
spin_unlock(&res->spinlock);
goto leave;
}
@ -2915,6 +2923,12 @@ static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm,
}
queue++;
}
nodenum = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
if (nodenum < O2NM_MAX_NODES) {
spin_unlock(&res->spinlock);
return nodenum;
}
spin_unlock(&res->spinlock);
mlog(0, "have not found a suitable target yet! checking domain map\n");

View File

@ -2241,11 +2241,15 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
mutex_lock(&inode->i_mutex);
ocfs2_iocb_clear_sem_locked(iocb);
relock:
/* to match setattr's i_mutex -> i_alloc_sem -> rw_lock ordering */
if (direct_io) {
down_read(&inode->i_alloc_sem);
have_alloc_sem = 1;
/* communicate with ocfs2_dio_end_io */
ocfs2_iocb_set_sem_locked(iocb);
}
/*
@ -2382,8 +2386,10 @@ static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
ocfs2_rw_unlock(inode, rw_level);
out_sems:
if (have_alloc_sem)
if (have_alloc_sem) {
up_read(&inode->i_alloc_sem);
ocfs2_iocb_clear_sem_locked(iocb);
}
mutex_unlock(&inode->i_mutex);
@ -2527,6 +2533,8 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
goto bail;
}
ocfs2_iocb_clear_sem_locked(iocb);
/*
* buffered reads protect themselves in ->readpage(). O_DIRECT reads
* need locks to protect pending reads from racing with truncate.
@ -2534,6 +2542,7 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
if (filp->f_flags & O_DIRECT) {
down_read(&inode->i_alloc_sem);
have_alloc_sem = 1;
ocfs2_iocb_set_sem_locked(iocb);
ret = ocfs2_rw_lock(inode, 0);
if (ret < 0) {
@ -2575,8 +2584,10 @@ static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
}
bail:
if (have_alloc_sem)
if (have_alloc_sem) {
up_read(&inode->i_alloc_sem);
ocfs2_iocb_clear_sem_locked(iocb);
}
if (rw_level != -1)
ocfs2_rw_unlock(inode, rw_level);
mlog_exit(ret);

View File

@ -350,7 +350,7 @@ enum {
#define OCFS2_LAST_LOCAL_SYSTEM_INODE LOCAL_GROUP_QUOTA_SYSTEM_INODE
NUM_SYSTEM_INODES
};
#define NUM_GLOBAL_SYSTEM_INODES OCFS2_LAST_GLOBAL_SYSTEM_INODE
#define NUM_GLOBAL_SYSTEM_INODES OCFS2_FIRST_LOCAL_SYSTEM_INODE
#define NUM_LOCAL_SYSTEM_INODES \
(NUM_SYSTEM_INODES - OCFS2_FIRST_LOCAL_SYSTEM_INODE)