static void gdlm_recovery_result(struct gfs2_sbd *sdp, unsigned int jid,
unsigned int result)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
return;
if (jid == ls->ls_jid)
return;
spin_lock(&ls->ls_recover_spin);
if (test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
spin_unlock(&ls->ls_recover_spin);
return;
}
if (ls->ls_recover_size < jid + 1) {
fs_err(sdp, "recovery_result jid %d short size %d",
jid, ls->ls_recover_size);
spin_unlock(&ls->ls_recover_spin);
return;
}
fs_info(sdp, "recover jid %d result %s\n", jid,
result == LM_RD_GAVEUP ? "busy" : "success");
ls->ls_recover_result[jid] = result;
if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags))
queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work,
result == LM_RD_GAVEUP ? HZ : 0);
spin_unlock(&ls->ls_recover_spin);
}
static int control_first_done(struct gfs2_sbd *sdp)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
char lvb_bits[GDLM_LVB_SIZE];
uint32_t start_gen, block_gen;
int error;
restart:
spin_lock(&ls->ls_recover_spin);
start_gen = ls->ls_recover_start;
block_gen = ls->ls_recover_block;
if (test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags) ||
!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
/* sanity check, should not happen */
fs_err(sdp, "control_first_done start %u block %u flags %lx\n",
start_gen, block_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
control_unlock(sdp);
return -1;
}
if (start_gen == block_gen) {
/*
* Wait for the end of a dlm recovery cycle to switch from
* first mounter recovery. We can ignore any recover_slot
* callbacks between the recover_prep and next recover_done
* because we are still the first mounter and any failed nodes
* have not fully mounted, so they don't need recovery.
*/
spin_unlock(&ls->ls_recover_spin);
fs_info(sdp, "control_first_done wait gen %u\n", start_gen);
wait_on_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY,
dlm_recovery_wait, TASK_UNINTERRUPTIBLE);
goto restart;
}
clear_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
set_bit(DFL_FIRST_MOUNT_DONE, &ls->ls_recover_flags);
memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
spin_unlock(&ls->ls_recover_spin);
memset(lvb_bits, 0, sizeof(lvb_bits));
control_lvb_write(ls, start_gen, lvb_bits);
error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT);
if (error)
fs_err(sdp, "control_first_done mounted PR error %d\n", error);
error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
if (error)
fs_err(sdp, "control_first_done control NL error %d\n", error);
return error;
}
static int control_first_done(struct gfs2_sbd *sdp)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
char lvb_bits[GDLM_LVB_SIZE];
uint32_t start_gen, block_gen;
int error;
restart:
spin_lock(&ls->ls_recover_spin);
start_gen = ls->ls_recover_start;
block_gen = ls->ls_recover_block;
if (test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags) ||
!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
/* */
fs_err(sdp, "control_first_done start %u block %u flags %lx\n",
start_gen, block_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
control_unlock(sdp);
return -1;
}
if (start_gen == block_gen) {
/*
*/
spin_unlock(&ls->ls_recover_spin);
fs_info(sdp, "control_first_done wait gen %u\n", start_gen);
wait_on_bit(&ls->ls_recover_flags, DFL_DLM_RECOVERY,
dlm_recovery_wait, TASK_UNINTERRUPTIBLE);
goto restart;
}
clear_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
set_bit(DFL_FIRST_MOUNT_DONE, &ls->ls_recover_flags);
memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
spin_unlock(&ls->ls_recover_spin);
memset(lvb_bits, 0, sizeof(lvb_bits));
control_lvb_write(ls, start_gen, lvb_bits);
error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT);
if (error)
fs_err(sdp, "control_first_done mounted PR error %d\n", error);
error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
if (error)
fs_err(sdp, "control_first_done control NL error %d\n", error);
return error;
}
static int gdlm_mount(struct gfs2_sbd *sdp, const char *fsname)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
int error;
if (fsname == NULL) {
fs_info(sdp, "no fsname found\n");
return -EINVAL;
}
error = dlm_new_lockspace(fsname, strlen(fsname), &ls->ls_dlm,
DLM_LSFL_FS | DLM_LSFL_NEWEXCL |
(ls->ls_nodir ? DLM_LSFL_NODIR : 0),
GDLM_LVB_SIZE);
if (error)
printk(KERN_ERR "dlm_new_lockspace error %d", error);
return error;
}
static void gdlm_recover_slot(void *arg, struct dlm_slot *slot)
{
struct gfs2_sbd *sdp = arg;
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
int jid = slot->slot - 1;
spin_lock(&ls->ls_recover_spin);
if (ls->ls_recover_size < jid + 1) {
fs_err(sdp, "recover_slot jid %d gen %u short size %d",
jid, ls->ls_recover_block, ls->ls_recover_size);
spin_unlock(&ls->ls_recover_spin);
return;
}
if (ls->ls_recover_submit[jid]) {
fs_info(sdp, "recover_slot jid %d gen %u prev %u",
jid, ls->ls_recover_block, ls->ls_recover_submit[jid]);
}
ls->ls_recover_submit[jid] = ls->ls_recover_block;
spin_unlock(&ls->ls_recover_spin);
}
static void gdlm_recovery_result(struct gfs2_sbd *sdp, unsigned int jid,
unsigned int result)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
if (test_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags))
return;
/* don't care about the recovery of own journal during mount */
if (jid == ls->ls_jid)
return;
spin_lock(&ls->ls_recover_spin);
if (test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
spin_unlock(&ls->ls_recover_spin);
return;
}
if (ls->ls_recover_size < jid + 1) {
fs_err(sdp, "recovery_result jid %d short size %d",
jid, ls->ls_recover_size);
spin_unlock(&ls->ls_recover_spin);
return;
}
fs_info(sdp, "recover jid %d result %s\n", jid,
result == LM_RD_GAVEUP ? "busy" : "success");
ls->ls_recover_result[jid] = result;
/* GAVEUP means another node is recovering the journal; delay our
next attempt to recover it, to give the other node a chance to
finish before trying again */
if (!test_bit(DFL_UNMOUNT, &ls->ls_recover_flags))
queue_delayed_work(gfs2_control_wq, &sdp->sd_control_work,
result == LM_RD_GAVEUP ? HZ : 0);
spin_unlock(&ls->ls_recover_spin);
}
int gfs2_mount_args(struct gfs2_sbd *sdp, char *data_arg, int remount)
{
struct gfs2_args *args = &sdp->sd_args;
char *data = data_arg;
char *options, *o, *v;
int error = 0;
if (!remount) {
/* If someone preloaded options, use those instead */
spin_lock(&gfs2_sys_margs_lock);
if (gfs2_sys_margs) {
data = gfs2_sys_margs;
gfs2_sys_margs = NULL;
}
spin_unlock(&gfs2_sys_margs_lock);
/* Set some defaults */
args->ar_num_glockd = GFS2_GLOCKD_DEFAULT;
args->ar_quota = GFS2_QUOTA_DEFAULT;
args->ar_data = GFS2_DATA_DEFAULT;
}
/* Split the options into tokens with the "," character and
process them */
for (options = data; (o = strsep(&options, ",")); ) {
int token, option;
substring_t tmp[MAX_OPT_ARGS];
if (!*o)
continue;
token = match_token(o, tokens, tmp);
switch (token) {
case Opt_lockproto:
v = match_strdup(&tmp[0]);
if (!v) {
fs_info(sdp, "no memory for lockproto\n");
error = -ENOMEM;
goto out_error;
}
if (remount && strcmp(v, args->ar_lockproto)) {
kfree(v);
goto cant_remount;
}
strncpy(args->ar_lockproto, v, GFS2_LOCKNAME_LEN);
args->ar_lockproto[GFS2_LOCKNAME_LEN - 1] = 0;
kfree(v);
break;
case Opt_locktable:
v = match_strdup(&tmp[0]);
if (!v) {
fs_info(sdp, "no memory for locktable\n");
error = -ENOMEM;
goto out_error;
}
if (remount && strcmp(v, args->ar_locktable)) {
kfree(v);
goto cant_remount;
}
strncpy(args->ar_locktable, v, GFS2_LOCKNAME_LEN);
args->ar_locktable[GFS2_LOCKNAME_LEN - 1] = 0;
kfree(v);
break;
case Opt_hostdata:
v = match_strdup(&tmp[0]);
if (!v) {
fs_info(sdp, "no memory for hostdata\n");
error = -ENOMEM;
goto out_error;
}
if (remount && strcmp(v, args->ar_hostdata)) {
kfree(v);
goto cant_remount;
}
strncpy(args->ar_hostdata, v, GFS2_LOCKNAME_LEN);
args->ar_hostdata[GFS2_LOCKNAME_LEN - 1] = 0;
kfree(v);
break;
case Opt_spectator:
if (remount && !args->ar_spectator)
goto cant_remount;
args->ar_spectator = 1;
sdp->sd_vfs->s_flags |= MS_RDONLY;
break;
case Opt_ignore_local_fs:
if (remount && !args->ar_ignore_local_fs)
goto cant_remount;
args->ar_ignore_local_fs = 1;
break;
case Opt_localflocks:
if (remount && !args->ar_localflocks)
goto cant_remount;
args->ar_localflocks = 1;
break;
case Opt_localcaching:
if (remount && !args->ar_localcaching)
goto cant_remount;
args->ar_localcaching = 1;
break;
case Opt_debug:
args->ar_debug = 1;
break;
case Opt_nodebug:
args->ar_debug = 0;
break;
case Opt_upgrade:
if (remount && !args->ar_upgrade)
goto cant_remount;
args->ar_upgrade = 1;
break;
case Opt_num_glockd:
if ((error = match_int(&tmp[0], &option))) {
fs_info(sdp, "problem getting num_glockd\n");
goto out_error;
}
if (remount && option != args->ar_num_glockd)
goto cant_remount;
if (!option || option > GFS2_GLOCKD_MAX) {
fs_info(sdp, "0 < num_glockd <= %u (not %u)\n",
GFS2_GLOCKD_MAX, option);
error = -EINVAL;
goto out_error;
}
args->ar_num_glockd = option;
break;
case Opt_acl:
args->ar_posix_acl = 1;
sdp->sd_vfs->s_flags |= MS_POSIXACL;
break;
case Opt_noacl:
args->ar_posix_acl = 0;
sdp->sd_vfs->s_flags &= ~MS_POSIXACL;
break;
case Opt_quota_off:
args->ar_quota = GFS2_QUOTA_OFF;
break;
case Opt_quota_account:
args->ar_quota = GFS2_QUOTA_ACCOUNT;
break;
case Opt_quota_on:
args->ar_quota = GFS2_QUOTA_ON;
break;
case Opt_suiddir:
args->ar_suiddir = 1;
break;
case Opt_nosuiddir:
args->ar_suiddir = 0;
break;
case Opt_data_writeback:
args->ar_data = GFS2_DATA_WRITEBACK;
break;
case Opt_data_ordered:
args->ar_data = GFS2_DATA_ORDERED;
break;
case Opt_err:
default:
fs_info(sdp, "unknown option: %s\n", o);
error = -EINVAL;
goto out_error;
}
}
out_error:
if (error)
fs_info(sdp, "invalid mount option(s)\n");
if (data != data_arg)
kfree(data);
return error;
cant_remount:
fs_info(sdp, "can't remount with option %s\n", o);
return -EINVAL;
}
static int control_mount(struct gfs2_sbd *sdp)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
uint32_t start_gen, block_gen, mount_gen, lvb_gen;
int mounted_mode;
int retries = 0;
int error;
memset(&ls->ls_mounted_lksb, 0, sizeof(struct dlm_lksb));
memset(&ls->ls_control_lksb, 0, sizeof(struct dlm_lksb));
memset(&ls->ls_control_lvb, 0, GDLM_LVB_SIZE);
ls->ls_control_lksb.sb_lvbptr = ls->ls_control_lvb;
init_completion(&ls->ls_sync_wait);
set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_VALBLK);
if (error) {
fs_err(sdp, "control_mount control_lock NL error %d\n", error);
return error;
}
error = mounted_lock(sdp, DLM_LOCK_NL, 0);
if (error) {
fs_err(sdp, "control_mount mounted_lock NL error %d\n", error);
control_unlock(sdp);
return error;
}
mounted_mode = DLM_LOCK_NL;
restart:
if (retries++ && signal_pending(current)) {
error = -EINTR;
goto fail;
}
/*
* We always start with both locks in NL. control_lock is
* demoted to NL below so we don't need to do it here.
*/
if (mounted_mode != DLM_LOCK_NL) {
error = mounted_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
if (error)
goto fail;
mounted_mode = DLM_LOCK_NL;
}
/*
* Other nodes need to do some work in dlm recovery and gfs2_control
* before the recover_done and control_lock will be ready for us below.
* A delay here is not required but often avoids having to retry.
*/
msleep_interruptible(500);
/*
* Acquire control_lock in EX and mounted_lock in either EX or PR.
* control_lock lvb keeps track of any pending journal recoveries.
* mounted_lock indicates if any other nodes have the fs mounted.
*/
error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE|DLM_LKF_VALBLK);
if (error == -EAGAIN) {
goto restart;
} else if (error) {
fs_err(sdp, "control_mount control_lock EX error %d\n", error);
goto fail;
}
error = mounted_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
if (!error) {
mounted_mode = DLM_LOCK_EX;
goto locks_done;
} else if (error != -EAGAIN) {
fs_err(sdp, "control_mount mounted_lock EX error %d\n", error);
goto fail;
}
error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
if (!error) {
mounted_mode = DLM_LOCK_PR;
goto locks_done;
} else {
/* not even -EAGAIN should happen here */
fs_err(sdp, "control_mount mounted_lock PR error %d\n", error);
goto fail;
}
locks_done:
/*
* If we got both locks above in EX, then we're the first mounter.
* If not, then we need to wait for the control_lock lvb to be
* updated by other mounted nodes to reflect our mount generation.
*
* In simple first mounter cases, first mounter will see zero lvb_gen,
* but in cases where all existing nodes leave/fail before mounting
* nodes finish control_mount, then all nodes will be mounting and
* lvb_gen will be non-zero.
*/
control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits);
if (lvb_gen == 0xFFFFFFFF) {
/* special value to force mount attempts to fail */
fs_err(sdp, "control_mount control_lock disabled\n");
error = -EINVAL;
goto fail;
}
if (mounted_mode == DLM_LOCK_EX) {
/* first mounter, keep both EX while doing first recovery */
spin_lock(&ls->ls_recover_spin);
clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
set_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
fs_info(sdp, "first mounter control generation %u\n", lvb_gen);
return 0;
}
error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
if (error)
goto fail;
/*
* We are not first mounter, now we need to wait for the control_lock
* lvb generation to be >= the generation from our first recover_done
* and all lvb bits to be clear (no pending journal recoveries.)
*/
if (!all_jid_bits_clear(ls->ls_lvb_bits)) {
/* journals need recovery, wait until all are clear */
fs_info(sdp, "control_mount wait for journal recovery\n");
goto restart;
}
spin_lock(&ls->ls_recover_spin);
block_gen = ls->ls_recover_block;
start_gen = ls->ls_recover_start;
mount_gen = ls->ls_recover_mount;
if (lvb_gen < mount_gen) {
/* wait for mounted nodes to update control_lock lvb to our
generation, which might include new recovery bits set */
fs_info(sdp, "control_mount wait1 block %u start %u mount %u "
"lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
lvb_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
goto restart;
}
if (lvb_gen != start_gen) {
/* wait for mounted nodes to update control_lock lvb to the
latest recovery generation */
fs_info(sdp, "control_mount wait2 block %u start %u mount %u "
"lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
lvb_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
goto restart;
}
if (block_gen == start_gen) {
/* dlm recovery in progress, wait for it to finish */
fs_info(sdp, "control_mount wait3 block %u start %u mount %u "
"lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
lvb_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
goto restart;
}
clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
spin_unlock(&ls->ls_recover_spin);
return 0;
fail:
mounted_unlock(sdp);
control_unlock(sdp);
return error;
}
static void gfs2_control_func(struct work_struct *work)
{
struct gfs2_sbd *sdp = container_of(work, struct gfs2_sbd, sd_control_work.work);
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
uint32_t block_gen, start_gen, lvb_gen, flags;
int recover_set = 0;
int write_lvb = 0;
int recover_size;
int i, error;
spin_lock(&ls->ls_recover_spin);
/*
* No MOUNT_DONE means we're still mounting; control_mount()
* will set this flag, after which this thread will take over
* all further clearing of BLOCK_LOCKS.
*
* FIRST_MOUNT means this node is doing first mounter recovery,
* for which recovery control is handled by
* control_mount()/control_first_done(), not this thread.
*/
if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
spin_unlock(&ls->ls_recover_spin);
return;
}
block_gen = ls->ls_recover_block;
start_gen = ls->ls_recover_start;
spin_unlock(&ls->ls_recover_spin);
/*
* Equal block_gen and start_gen implies we are between
* recover_prep and recover_done callbacks, which means
* dlm recovery is in progress and dlm locking is blocked.
* There's no point trying to do any work until recover_done.
*/
if (block_gen == start_gen)
return;
/*
* Propagate recover_submit[] and recover_result[] to lvb:
* dlm_recoverd adds to recover_submit[] jids needing recovery
* gfs2_recover adds to recover_result[] journal recovery results
*
* set lvb bit for jids in recover_submit[] if the lvb has not
* yet been updated for the generation of the failure
*
* clear lvb bit for jids in recover_result[] if the result of
* the journal recovery is SUCCESS
*/
error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
if (error) {
fs_err(sdp, "control lock EX error %d\n", error);
return;
}
control_lvb_read(ls, &lvb_gen, ls->ls_lvb_bits);
spin_lock(&ls->ls_recover_spin);
if (block_gen != ls->ls_recover_block ||
start_gen != ls->ls_recover_start) {
fs_info(sdp, "recover generation %u block1 %u %u\n",
start_gen, block_gen, ls->ls_recover_block);
spin_unlock(&ls->ls_recover_spin);
control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
return;
}
recover_size = ls->ls_recover_size;
if (lvb_gen <= start_gen) {
/*
* Clear lvb bits for jids we've successfully recovered.
* Because all nodes attempt to recover failed journals,
* a journal can be recovered multiple times successfully
* in succession. Only the first will really do recovery,
* the others find it clean, but still report a successful
* recovery. So, another node may have already recovered
* the jid and cleared the lvb bit for it.
*/
for (i = 0; i < recover_size; i++) {
if (ls->ls_recover_result[i] != LM_RD_SUCCESS)
continue;
ls->ls_recover_result[i] = 0;
if (!test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET))
continue;
__clear_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET);
write_lvb = 1;
}
}
if (lvb_gen == start_gen) {
/*
* Failed slots before start_gen are already set in lvb.
*/
for (i = 0; i < recover_size; i++) {
if (!ls->ls_recover_submit[i])
continue;
if (ls->ls_recover_submit[i] < lvb_gen)
ls->ls_recover_submit[i] = 0;
}
} else if (lvb_gen < start_gen) {
/*
* Failed slots before start_gen are not yet set in lvb.
*/
for (i = 0; i < recover_size; i++) {
if (!ls->ls_recover_submit[i])
continue;
if (ls->ls_recover_submit[i] < start_gen) {
ls->ls_recover_submit[i] = 0;
__set_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET);
}
}
/* even if there are no bits to set, we need to write the
latest generation to the lvb */
write_lvb = 1;
} else {
/*
* we should be getting a recover_done() for lvb_gen soon
*/
}
spin_unlock(&ls->ls_recover_spin);
if (write_lvb) {
control_lvb_write(ls, start_gen, ls->ls_lvb_bits);
flags = DLM_LKF_CONVERT | DLM_LKF_VALBLK;
} else {
flags = DLM_LKF_CONVERT;
}
error = control_lock(sdp, DLM_LOCK_NL, flags);
if (error) {
fs_err(sdp, "control lock NL error %d\n", error);
return;
}
/*
* Everyone will see jid bits set in the lvb, run gfs2_recover_set(),
* and clear a jid bit in the lvb if the recovery is a success.
* Eventually all journals will be recovered, all jid bits will
* be cleared in the lvb, and everyone will clear BLOCK_LOCKS.
*/
for (i = 0; i < recover_size; i++) {
if (test_bit_le(i, ls->ls_lvb_bits + JID_BITMAP_OFFSET)) {
fs_info(sdp, "recover generation %u jid %d\n",
start_gen, i);
gfs2_recover_set(sdp, i);
recover_set++;
}
}
if (recover_set)
return;
/*
* No more jid bits set in lvb, all recovery is done, unblock locks
* (unless a new recover_prep callback has occured blocking locks
* again while working above)
*/
spin_lock(&ls->ls_recover_spin);
if (ls->ls_recover_block == block_gen &&
ls->ls_recover_start == start_gen) {
clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
fs_info(sdp, "recover generation %u done\n", start_gen);
gfs2_glock_thaw(sdp);
} else {
fs_info(sdp, "recover generation %u block2 %u %u\n",
start_gen, block_gen, ls->ls_recover_block);
spin_unlock(&ls->ls_recover_spin);
}
}
static int gdlm_mount(struct gfs2_sbd *sdp, const char *table)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
char cluster[GFS2_LOCKNAME_LEN];
const char *fsname;
uint32_t flags;
int error, ops_result;
/*
* initialize everything
*/
INIT_DELAYED_WORK(&sdp->sd_control_work, gfs2_control_func);
spin_lock_init(&ls->ls_recover_spin);
ls->ls_recover_flags = 0;
ls->ls_recover_mount = 0;
ls->ls_recover_start = 0;
ls->ls_recover_block = 0;
ls->ls_recover_size = 0;
ls->ls_recover_submit = NULL;
ls->ls_recover_result = NULL;
ls->ls_lvb_bits = NULL;
error = set_recover_size(sdp, NULL, 0);
if (error)
goto fail;
/*
* prepare dlm_new_lockspace args
*/
fsname = strchr(table, ':');
if (!fsname) {
fs_info(sdp, "no fsname found\n");
error = -EINVAL;
goto fail_free;
}
memset(cluster, 0, sizeof(cluster));
memcpy(cluster, table, strlen(table) - strlen(fsname));
fsname++;
flags = DLM_LSFL_FS | DLM_LSFL_NEWEXCL;
/*
* create/join lockspace
*/
error = dlm_new_lockspace(fsname, cluster, flags, GDLM_LVB_SIZE,
&gdlm_lockspace_ops, sdp, &ops_result,
&ls->ls_dlm);
if (error) {
fs_err(sdp, "dlm_new_lockspace error %d\n", error);
goto fail_free;
}
if (ops_result < 0) {
/*
* dlm does not support ops callbacks,
* old dlm_controld/gfs_controld are used, try without ops.
*/
fs_info(sdp, "dlm lockspace ops not used\n");
free_recover_size(ls);
set_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags);
return 0;
}
if (!test_bit(SDF_NOJOURNALID, &sdp->sd_flags)) {
fs_err(sdp, "dlm lockspace ops disallow jid preset\n");
error = -EINVAL;
goto fail_release;
}
/*
* control_mount() uses control_lock to determine first mounter,
* and for later mounts, waits for any recoveries to be cleared.
*/
error = control_mount(sdp);
if (error) {
fs_err(sdp, "mount control error %d\n", error);
goto fail_release;
}
ls->ls_first = !!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
clear_bit(SDF_NOJOURNALID, &sdp->sd_flags);
smp_mb__after_atomic();
wake_up_bit(&sdp->sd_flags, SDF_NOJOURNALID);
return 0;
fail_release:
dlm_release_lockspace(ls->ls_dlm, 2);
fail_free:
free_recover_size(ls);
fail:
return error;
}
/*
* This is the function called from the main polling loop.
*
* It calls command functions to get strings of data, and sends them to the server.
*
*/
int collect_and_send_metrics(int cycle) {
int retval;
char* command;
StringInfoData commands;
pgstat_report_activity(STATE_RUNNING, "Collecting metrics");
initStringInfo(&commands);
/*
* Populate first cycle command data. These are executed on the first cycle
* of a restart. The bgworker restars every N cycles, as listed at the bottom
* of the main loop in pgsampler.c.
*/
if (cycle == 0) {
command = restart_gucs();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("restart_gucs", command);
pfree(command);
command = relation_info();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_class", command);
pfree(command);
command = database_list();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("databases", command);
pfree(command);
command = column_info();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("columns", command);
pfree(command);
command = index_info();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("indexes", command);
pfree(command);
command = column_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_column", command);
pfree(command);
command = db_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_database", command);
pfree(command);
}
/* HEARTBEAT */
if (cycle % heartbeat_seconds == 0) {
command = heartbeat();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("heartbeats", command);
pfree(command);
}
/* SYSTEM INFO */
if (cycle % system_seconds == 0) {
command = system_info();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_system", command);
pfree(command);
command = fs_info();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_filesystem", command);
pfree(command);
}
/* */
if (cycle % activity_seconds == 0) {
command = activity_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_activity", command);
pfree(command);
}
if (cycle % replication_seconds == 0) {
command = replication_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_replication", command);
pfree(command);
}
/* */
if (cycle % bgwriter_seconds == 0) {
command = bgwriter_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_bgwriter", command);
pfree(command);
}
if (cycle % guc_seconds == 0) {
command = transient_gucs();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("transient_gucs", command);
pfree(command);
}
if (cycle % statements_seconds == 0) {
command = stat_statements();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_statements", command);
pfree(command);
}
/* */
if (cycle % relation_seconds == 0) {
command = table_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_table", command);
pfree(command);
command = index_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_index", command);
pfree(command);
command = table_io_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("statio_user_tables", command);
pfree(command);
command = index_io_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("statio_user_indexes", command);
pfree(command);
command = function_stats();
appendStringInfoString(&commands, command);
if (strcmp(output_mode, "csv") == 0)
write_to_csv("stat_function", command);
pfree(command);
}
/* Send / Write metrics based on output_mode */
if (strcmp(output_mode, "network") == 0) {
pgstat_report_activity(STATE_RUNNING, "Sending metrics to antenna");
retval = send_data(commands.data);
if (retval == NO_DATA_SENT) { //close socket and retry establishing connection and sending data.
// elog(LOG, "reseting..."); //just a note to say reseting socket
if (sockfd != 0)
shutdown(sockfd, SHUT_RDWR);
sockfd = 0;
retval = send_data(commands.data); // we ignore success or failure here. drops data if fails.
}
}
return 0;
}
int mon_ls(int argc, char **argv){
fs_info();
return 0;
}
static int gdlm_mount(struct gfs2_sbd *sdp, const char *table)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
char cluster[GFS2_LOCKNAME_LEN];
const char *fsname;
uint32_t flags;
int error, ops_result;
INIT_DELAYED_WORK(&sdp->sd_control_work, gfs2_control_func);
spin_lock_init(&ls->ls_recover_spin);
ls->ls_recover_flags = 0;
ls->ls_recover_mount = 0;
ls->ls_recover_start = 0;
ls->ls_recover_block = 0;
ls->ls_recover_size = 0;
ls->ls_recover_submit = NULL;
ls->ls_recover_result = NULL;
error = set_recover_size(sdp, NULL, 0);
if (error)
goto fail;
fsname = strchr(table, ':');
if (!fsname) {
fs_info(sdp, "no fsname found\n");
error = -EINVAL;
goto fail_free;
}
memset(cluster, 0, sizeof(cluster));
memcpy(cluster, table, strlen(table) - strlen(fsname));
fsname++;
flags = DLM_LSFL_FS | DLM_LSFL_NEWEXCL;
if (ls->ls_nodir)
flags |= DLM_LSFL_NODIR;
error = dlm_new_lockspace(fsname, cluster, flags, GDLM_LVB_SIZE,
&gdlm_lockspace_ops, sdp, &ops_result,
&ls->ls_dlm);
if (error) {
fs_err(sdp, "dlm_new_lockspace error %d\n", error);
goto fail_free;
}
if (ops_result < 0) {
fs_info(sdp, "dlm lockspace ops not used\n");
free_recover_size(ls);
set_bit(DFL_NO_DLM_OPS, &ls->ls_recover_flags);
return 0;
}
if (!test_bit(SDF_NOJOURNALID, &sdp->sd_flags)) {
fs_err(sdp, "dlm lockspace ops disallow jid preset\n");
error = -EINVAL;
goto fail_release;
}
error = control_mount(sdp);
if (error) {
fs_err(sdp, "mount control error %d\n", error);
goto fail_release;
}
ls->ls_first = !!test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
clear_bit(SDF_NOJOURNALID, &sdp->sd_flags);
smp_mb__after_clear_bit();
wake_up_bit(&sdp->sd_flags, SDF_NOJOURNALID);
return 0;
fail_release:
dlm_release_lockspace(ls->ls_dlm, 2);
fail_free:
free_recover_size(ls);
fail:
return error;
}
static int control_mount(struct gfs2_sbd *sdp)
{
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
char lvb_bits[GDLM_LVB_SIZE];
uint32_t start_gen, block_gen, mount_gen, lvb_gen;
int mounted_mode;
int retries = 0;
int error;
memset(&ls->ls_mounted_lksb, 0, sizeof(struct dlm_lksb));
memset(&ls->ls_control_lksb, 0, sizeof(struct dlm_lksb));
memset(&ls->ls_control_lvb, 0, GDLM_LVB_SIZE);
ls->ls_control_lksb.sb_lvbptr = ls->ls_control_lvb;
init_completion(&ls->ls_sync_wait);
set_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_VALBLK);
if (error) {
fs_err(sdp, "control_mount control_lock NL error %d\n", error);
return error;
}
error = mounted_lock(sdp, DLM_LOCK_NL, 0);
if (error) {
fs_err(sdp, "control_mount mounted_lock NL error %d\n", error);
control_unlock(sdp);
return error;
}
mounted_mode = DLM_LOCK_NL;
restart:
if (retries++ && signal_pending(current)) {
error = -EINTR;
goto fail;
}
if (mounted_mode != DLM_LOCK_NL) {
error = mounted_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
if (error)
goto fail;
mounted_mode = DLM_LOCK_NL;
}
msleep_interruptible(500);
error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE|DLM_LKF_VALBLK);
if (error == -EAGAIN) {
goto restart;
} else if (error) {
fs_err(sdp, "control_mount control_lock EX error %d\n", error);
goto fail;
}
error = mounted_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
if (!error) {
mounted_mode = DLM_LOCK_EX;
goto locks_done;
} else if (error != -EAGAIN) {
fs_err(sdp, "control_mount mounted_lock EX error %d\n", error);
goto fail;
}
error = mounted_lock(sdp, DLM_LOCK_PR, DLM_LKF_CONVERT|DLM_LKF_NOQUEUE);
if (!error) {
mounted_mode = DLM_LOCK_PR;
goto locks_done;
} else {
fs_err(sdp, "control_mount mounted_lock PR error %d\n", error);
goto fail;
}
locks_done:
control_lvb_read(ls, &lvb_gen, lvb_bits);
if (lvb_gen == 0xFFFFFFFF) {
fs_err(sdp, "control_mount control_lock disabled\n");
error = -EINVAL;
goto fail;
}
if (mounted_mode == DLM_LOCK_EX) {
spin_lock(&ls->ls_recover_spin);
clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
set_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
fs_info(sdp, "first mounter control generation %u\n", lvb_gen);
return 0;
}
error = control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
if (error)
goto fail;
if (!all_jid_bits_clear(lvb_bits)) {
fs_info(sdp, "control_mount wait for journal recovery\n");
goto restart;
}
spin_lock(&ls->ls_recover_spin);
block_gen = ls->ls_recover_block;
start_gen = ls->ls_recover_start;
mount_gen = ls->ls_recover_mount;
if (lvb_gen < mount_gen) {
fs_info(sdp, "control_mount wait1 block %u start %u mount %u "
"lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
lvb_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
goto restart;
}
if (lvb_gen != start_gen) {
fs_info(sdp, "control_mount wait2 block %u start %u mount %u "
"lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
lvb_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
goto restart;
}
if (block_gen == start_gen) {
fs_info(sdp, "control_mount wait3 block %u start %u mount %u "
"lvb %u flags %lx\n", block_gen, start_gen, mount_gen,
lvb_gen, ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
goto restart;
}
clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
set_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags);
memset(ls->ls_recover_submit, 0, ls->ls_recover_size*sizeof(uint32_t));
memset(ls->ls_recover_result, 0, ls->ls_recover_size*sizeof(uint32_t));
spin_unlock(&ls->ls_recover_spin);
return 0;
fail:
mounted_unlock(sdp);
control_unlock(sdp);
return error;
}
static void gfs2_control_func(struct work_struct *work)
{
struct gfs2_sbd *sdp = container_of(work, struct gfs2_sbd, sd_control_work.work);
struct lm_lockstruct *ls = &sdp->sd_lockstruct;
char lvb_bits[GDLM_LVB_SIZE];
uint32_t block_gen, start_gen, lvb_gen, flags;
int recover_set = 0;
int write_lvb = 0;
int recover_size;
int i, error;
spin_lock(&ls->ls_recover_spin);
if (!test_bit(DFL_MOUNT_DONE, &ls->ls_recover_flags) ||
test_bit(DFL_FIRST_MOUNT, &ls->ls_recover_flags)) {
spin_unlock(&ls->ls_recover_spin);
return;
}
block_gen = ls->ls_recover_block;
start_gen = ls->ls_recover_start;
spin_unlock(&ls->ls_recover_spin);
if (block_gen == start_gen)
return;
error = control_lock(sdp, DLM_LOCK_EX, DLM_LKF_CONVERT|DLM_LKF_VALBLK);
if (error) {
fs_err(sdp, "control lock EX error %d\n", error);
return;
}
control_lvb_read(ls, &lvb_gen, lvb_bits);
spin_lock(&ls->ls_recover_spin);
if (block_gen != ls->ls_recover_block ||
start_gen != ls->ls_recover_start) {
fs_info(sdp, "recover generation %u block1 %u %u\n",
start_gen, block_gen, ls->ls_recover_block);
spin_unlock(&ls->ls_recover_spin);
control_lock(sdp, DLM_LOCK_NL, DLM_LKF_CONVERT);
return;
}
recover_size = ls->ls_recover_size;
if (lvb_gen <= start_gen) {
for (i = 0; i < recover_size; i++) {
if (ls->ls_recover_result[i] != LM_RD_SUCCESS)
continue;
ls->ls_recover_result[i] = 0;
if (!test_bit_le(i, lvb_bits + JID_BITMAP_OFFSET))
continue;
__clear_bit_le(i, lvb_bits + JID_BITMAP_OFFSET);
write_lvb = 1;
}
}
if (lvb_gen == start_gen) {
for (i = 0; i < recover_size; i++) {
if (!ls->ls_recover_submit[i])
continue;
if (ls->ls_recover_submit[i] < lvb_gen)
ls->ls_recover_submit[i] = 0;
}
} else if (lvb_gen < start_gen) {
for (i = 0; i < recover_size; i++) {
if (!ls->ls_recover_submit[i])
continue;
if (ls->ls_recover_submit[i] < start_gen) {
ls->ls_recover_submit[i] = 0;
__set_bit_le(i, lvb_bits + JID_BITMAP_OFFSET);
}
}
write_lvb = 1;
} else {
}
spin_unlock(&ls->ls_recover_spin);
if (write_lvb) {
control_lvb_write(ls, start_gen, lvb_bits);
flags = DLM_LKF_CONVERT | DLM_LKF_VALBLK;
} else {
flags = DLM_LKF_CONVERT;
}
error = control_lock(sdp, DLM_LOCK_NL, flags);
if (error) {
fs_err(sdp, "control lock NL error %d\n", error);
return;
}
for (i = 0; i < recover_size; i++) {
if (test_bit_le(i, lvb_bits + JID_BITMAP_OFFSET)) {
fs_info(sdp, "recover generation %u jid %d\n",
start_gen, i);
gfs2_recover_set(sdp, i);
recover_set++;
}
}
if (recover_set)
return;
spin_lock(&ls->ls_recover_spin);
if (ls->ls_recover_block == block_gen &&
ls->ls_recover_start == start_gen) {
clear_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags);
spin_unlock(&ls->ls_recover_spin);
fs_info(sdp, "recover generation %u done\n", start_gen);
gfs2_glock_thaw(sdp);
} else {
fs_info(sdp, "recover generation %u block2 %u %u\n",
start_gen, block_gen, ls->ls_recover_block);
spin_unlock(&ls->ls_recover_spin);
}
}