/* root_squash for inter-MDS operations */
static int mdt_root_squash(struct mdt_thread_info *info, lnet_nid_t peernid)
{
struct md_ucred *ucred = mdt_ucred(info);
ENTRY;
if (!info->mti_mdt->mdt_squash_uid || ucred->mu_fsuid)
RETURN(0);
if (match_nosquash_list(&info->mti_mdt->mdt_squash_sem,
&info->mti_mdt->mdt_nosquash_nids,
peernid)) {
CDEBUG(D_OTHER, "%s is in nosquash_nids list\n",
libcfs_nid2str(peernid));
RETURN(0);
}
CDEBUG(D_OTHER, "squash req from %s, (%d:%d/%x)=>(%d:%d/%x)\n",
libcfs_nid2str(peernid),
ucred->mu_fsuid, ucred->mu_fsgid, ucred->mu_cap,
info->mti_mdt->mdt_squash_uid, info->mti_mdt->mdt_squash_gid,
0);
ucred->mu_fsuid = info->mti_mdt->mdt_squash_uid;
ucred->mu_fsgid = info->mti_mdt->mdt_squash_gid;
ucred->mu_cap = 0;
ucred->mu_suppgids[0] = -1;
ucred->mu_suppgids[1] = -1;
RETURN(0);
}
void
_kgnilnd_debug_conn(kgn_conn_t *conn, struct libcfs_debug_msg_data *msgdata,
const char *fmt, ...)
{
va_list args;
va_start(args, fmt);
libcfs_debug_vmsg2(msgdata, fmt, args,
" conn@0x%p->%s:%s cq %u, to %ds, "
" RX %d @ %lu/%lus; TX %d @ %lus/%lus; "
" NOOP %lus/%lu/%lus; sched %lus/%lus/%lus ago \n",
conn, conn->gnc_peer ? libcfs_nid2str(conn->gnc_peer->gnp_nid) :
"<?>", kgnilnd_conn_state2str(conn),
conn->gnc_cqid, conn->gnc_timeout,
conn->gnc_rx_seq,
cfs_duration_sec(jiffies - conn->gnc_last_rx),
cfs_duration_sec(jiffies - conn->gnc_last_rx_cq),
conn->gnc_tx_seq,
cfs_duration_sec(jiffies - conn->gnc_last_tx),
cfs_duration_sec(jiffies - conn->gnc_last_tx_cq),
cfs_duration_sec(jiffies - conn->gnc_last_noop_want),
cfs_duration_sec(jiffies - conn->gnc_last_noop_sent),
cfs_duration_sec(jiffies - conn->gnc_last_noop_cq),
cfs_duration_sec(jiffies - conn->gnc_last_sched_ask),
cfs_duration_sec(jiffies - conn->gnc_last_sched_do),
cfs_duration_sec(jiffies - conn->gnc_device->gnd_sched_alive));
va_end(args);
}
static
int do_add_uuid(char * func, char *uuid, lnet_nid_t nid)
{
int rc;
struct lustre_cfg_bufs bufs;
struct lustre_cfg *lcfg;
lustre_cfg_bufs_reset(&bufs, lcfg_devname);
if (uuid != NULL)
lustre_cfg_bufs_set_string(&bufs, 1, uuid);
lcfg = lustre_cfg_new(LCFG_ADD_UUID, &bufs);
if (lcfg == NULL) {
rc = -ENOMEM;
} else {
lcfg->lcfg_nid = nid;
rc = lcfg_ioctl(func, OBD_DEV_ID, lcfg);
lustre_cfg_free(lcfg);
}
if (rc) {
fprintf(stderr, "IOC_PORTAL_ADD_UUID failed: %s\n",
strerror(errno));
return -1;
}
if (uuid != NULL)
printf("Added uuid %s: %s\n", uuid, libcfs_nid2str(nid));
return 0;
}
static
int do_add_uuid(char * func, char *uuid, lnet_nid_t nid)
{
int rc;
struct lustre_cfg_bufs bufs;
struct lustre_cfg *lcfg;
lustre_cfg_bufs_reset(&bufs, lcfg_devname);
if (uuid)
lustre_cfg_bufs_set_string(&bufs, 1, uuid);
lcfg = lustre_cfg_new(LCFG_ADD_UUID, &bufs);
lcfg->lcfg_nid = nid;
/* Poison NAL -- pre 1.4.6 will LASSERT on 0 NAL, this way it
doesn't work without crashing (bz 10130) */
lcfg->lcfg_nal = 0x5a;
#if 0
fprintf(stderr, "adding\tnid: %d\tuuid: %s\n",
lcfg->lcfg_nid, uuid);
#endif
rc = lcfg_ioctl(func, OBD_DEV_ID, lcfg);
lustre_cfg_free(lcfg);
if (rc) {
fprintf(stderr, "IOC_PORTAL_ADD_UUID failed: %s\n",
strerror(errno));
return -1;
}
printf ("Added uuid %s: %s\n", uuid, libcfs_nid2str(nid));
return 0;
}
int ptlrpc_connection_put(struct ptlrpc_connection *conn)
{
int rc = 0;
ENTRY;
if (!conn)
RETURN(rc);
LASSERT(!cfs_hlist_unhashed(&conn->c_hash));
/*
* We do not remove connection from hashtable and
* do not free it even if last caller released ref,
* as we want to have it cached for the case it is
* needed again.
*
* Deallocating it and later creating new connection
* again would be wastful. This way we also avoid
* expensive locking to protect things from get/put
* race when found cached connection is freed by
* ptlrpc_connection_put().
*
* It will be freed later in module unload time,
* when ptlrpc_connection_fini()->lh_exit->conn_exit()
* path is called.
*/
if (cfs_atomic_dec_return(&conn->c_refcount) == 1)
rc = 1;
CDEBUG(D_INFO, "PUT conn=%p refcount %d to %s\n",
conn, cfs_atomic_read(&conn->c_refcount),
libcfs_nid2str(conn->c_peer.nid));
RETURN(rc);
}
char *
libcfs_id2str(lnet_process_id_t id)
{
char *str = libcfs_next_nidstring();
if (id.pid == LNET_PID_ANY) {
snprintf(str, LNET_NIDSTR_SIZE,
"LNET_PID_ANY-%s", libcfs_nid2str(id.nid));
return str;
}
snprintf(str, LNET_NIDSTR_SIZE, "%s%u-%s",
((id.pid & LNET_PID_USERFLAG) != 0) ? "U" : "",
(id.pid & ~LNET_PID_USERFLAG), libcfs_nid2str(id.nid));
return str;
}
void
_kgnilnd_debug_tx(kgn_tx_t *tx, struct libcfs_debug_msg_data *msgdata,
const char *fmt, ...)
{
kgn_tx_ev_id_t *id = &tx->tx_id;
char *nid = "<?>";
va_list args;
if (tx->tx_conn && tx->tx_conn->gnc_peer) {
nid = libcfs_nid2str(tx->tx_conn->gnc_peer->gnp_nid);
}
va_start(args, fmt);
libcfs_debug_vmsg2(msgdata, fmt, args,
" tx@0x%p->%s id "LPX64
"/%u/%d:%d msg %x/%s/%d x%d q %s@%lds->0x%p f %x re %d\n",
tx, nid, id->txe_cookie, id->txe_smsg_id, id->txe_cqid,
id->txe_idx, tx->tx_msg.gnm_type,
kgnilnd_msgtype2str(tx->tx_msg.gnm_type), tx->tx_buftype,
tx->tx_msg.gnm_seq,
kgnilnd_tx_state2str(tx->tx_list_state),
cfs_duration_sec((long)jiffies - tx->tx_qtime), tx->tx_list_p,
tx->tx_state, tx->tx_retrans);
va_end(args);
}
static char *convert_hostnames(char *s1)
{
char *converted, *s2 = 0, *c;
char sep;
int left = MAXNIDSTR;
lnet_nid_t nid;
converted = malloc(left);
if (converted == NULL) {
fprintf(stderr, "out of memory: needed %d bytes\n",
MAXNIDSTR);
return NULL;
}
c = converted;
while ((left > 0) && (*s1 != '/')) {
s2 = strpbrk(s1, ",:");
if (!s2)
goto out_free;
sep = *s2;
*s2 = '\0';
nid = libcfs_str2nid(s1);
*s2 = sep; /* back to original string */
if (nid == LNET_NID_ANY)
goto out_free;
c += snprintf(c, left, "%s%c", libcfs_nid2str(nid), sep);
left = converted + MAXNIDSTR - c;
s1 = s2 + 1;
}
snprintf(c, left, "%s", s1);
return converted;
out_free:
fprintf(stderr, "%s: Can't parse NID '%s'\n", progname, s1);
free(converted);
return NULL;
}
void
lnet_connect_console_error (int rc, lnet_nid_t peer_nid,
__u32 peer_ip, int peer_port)
{
switch (rc) {
/* "normal" errors */
case -ECONNREFUSED:
CDEBUG(D_NETERROR, "Connection to %s at host %u.%u.%u.%u "
"on port %d was refused: "
"check that Lustre is running on that node.\n",
libcfs_nid2str(peer_nid),
HIPQUAD(peer_ip), peer_port);
break;
case -EHOSTUNREACH:
case -ENETUNREACH:
CDEBUG(D_NETERROR, "Connection to %s at host %u.%u.%u.%u "
"was unreachable: the network or that node may "
"be down, or Lustre may be misconfigured.\n",
libcfs_nid2str(peer_nid), HIPQUAD(peer_ip));
break;
case -ETIMEDOUT:
CDEBUG(D_NETERROR, "Connection to %s at host %u.%u.%u.%u on "
"port %d took too long: that node may be hung "
"or experiencing high load.\n",
libcfs_nid2str(peer_nid),
HIPQUAD(peer_ip), peer_port);
break;
case -ECONNRESET:
LCONSOLE_ERROR_MSG(0x11b, "Connection to %s at host %u.%u.%u.%u"
" on port %d was reset: "
"is it running a compatible version of "
"Lustre and is %s one of its NIDs?\n",
libcfs_nid2str(peer_nid),
HIPQUAD(peer_ip), peer_port,
libcfs_nid2str(peer_nid));
break;
case -EPROTO:
LCONSOLE_ERROR_MSG(0x11c, "Protocol error connecting to %s at "
"host %u.%u.%u.%u on port %d: is it running "
"a compatible version of Lustre?\n",
libcfs_nid2str(peer_nid),
HIPQUAD(peer_ip), peer_port);
break;
case -EADDRINUSE:
LCONSOLE_ERROR_MSG(0x11d, "No privileged ports available to "
"connect to %s at host %u.%u.%u.%u on port "
"%d\n", libcfs_nid2str(peer_nid),
HIPQUAD(peer_ip), peer_port);
break;
default:
LCONSOLE_ERROR_MSG(0x11e, "Unexpected error %d connecting to %s"
" at host %u.%u.%u.%u on port %d\n", rc,
libcfs_nid2str(peer_nid),
HIPQUAD(peer_ip), peer_port);
break;
}
}
static char *convert_hostnames(char *s1)
{
char *converted, *s2 = 0, *c, *end, sep;
int left = MAXNIDSTR;
lnet_nid_t nid;
converted = malloc(left);
if (converted == NULL) {
return NULL;
}
end = s1 + strlen(s1);
c = converted;
while ((left > 0) && (s1 < end)) {
s2 = strpbrk(s1, ",:");
if (!s2)
s2 = end;
sep = *s2;
*s2 = '\0';
nid = libcfs_str2nid(s1);
*s2 = sep;
if (nid == LNET_NID_ANY) {
fprintf(stderr, "%s: Cannot resolve hostname '%s'.\n",
progname, s1);
free(converted);
return NULL;
}
if (strncmp(libcfs_nid2str(nid), "127.0.0.1",
strlen("127.0.0.1")) == 0) {
fprintf(stderr, "%s: The NID '%s' resolves to the "
"loopback address '%s'. Lustre requires a "
"non-loopback address.\n",
progname, s1, libcfs_nid2str(nid));
free(converted);
return NULL;
}
c += snprintf(c, left, "%s%c", libcfs_nid2str(nid), sep);
left = converted + MAXNIDSTR - c;
s1 = s2 + 1;
}
return converted;
}
struct ptlrpc_connection *
ptlrpc_connection_addref(struct ptlrpc_connection *conn)
{
atomic_inc(&conn->c_refcount);
CDEBUG(D_INFO, "conn=%p refcount %d to %s\n",
conn, atomic_read(&conn->c_refcount),
libcfs_nid2str(conn->c_peer.nid));
return conn;
}
struct ptlrpc_connection *
ptlrpc_connection_addref(struct ptlrpc_connection *conn)
{
ENTRY;
cfs_atomic_inc(&conn->c_refcount);
CDEBUG(D_INFO, "conn=%p refcount %d to %s\n",
conn, cfs_atomic_read(&conn->c_refcount),
libcfs_nid2str(conn->c_peer.nid));
RETURN(conn);
}
/*
* similar as in ost_connect_check_sptlrpc()
*/
static int mgs_connect_check_sptlrpc(struct ptlrpc_request *req)
{
struct obd_export *exp = req->rq_export;
struct obd_device *obd = exp->exp_obd;
struct fs_db *fsdb;
struct sptlrpc_flavor flvr;
int rc = 0;
if (exp->exp_flvr.sf_rpc == SPTLRPC_FLVR_INVALID) {
rc = mgs_find_or_make_fsdb(obd, MGSSELF_NAME, &fsdb);
if (rc)
return rc;
cfs_mutex_lock(&fsdb->fsdb_mutex);
if (sptlrpc_rule_set_choose(&fsdb->fsdb_srpc_gen,
LUSTRE_SP_MGC, LUSTRE_SP_MGS,
req->rq_peer.nid,
&flvr) == 0) {
/* by defualt allow any flavors */
flvr.sf_rpc = SPTLRPC_FLVR_ANY;
}
cfs_mutex_unlock(&fsdb->fsdb_mutex);
cfs_spin_lock(&exp->exp_lock);
exp->exp_sp_peer = req->rq_sp_from;
exp->exp_flvr = flvr;
if (exp->exp_flvr.sf_rpc != SPTLRPC_FLVR_ANY &&
exp->exp_flvr.sf_rpc != req->rq_flvr.sf_rpc) {
CERROR("invalid rpc flavor %x, expect %x, from %s\n",
req->rq_flvr.sf_rpc, exp->exp_flvr.sf_rpc,
libcfs_nid2str(req->rq_peer.nid));
rc = -EACCES;
}
cfs_spin_unlock(&exp->exp_lock);
} else {
if (exp->exp_sp_peer != req->rq_sp_from) {
CERROR("RPC source %s doesn't match %s\n",
sptlrpc_part2name(req->rq_sp_from),
sptlrpc_part2name(exp->exp_sp_peer));
rc = -EACCES;
} else {
rc = sptlrpc_target_export_check(exp, req);
}
}
return rc;
}
static void print_1_cfg(struct lustre_cfg *lcfg)
{
int i;
if (lcfg->lcfg_nid)
printf("nid=%s("LPX64") ", libcfs_nid2str(lcfg->lcfg_nid),
lcfg->lcfg_nid);
if (lcfg->lcfg_nal)
printf("nal=%d ", lcfg->lcfg_nal);
for (i = 0; i < lcfg->lcfg_bufcount; i++)
printf("%d:%.*s ", i, lcfg->lcfg_buflens[i],
(char*)lustre_cfg_buf(lcfg, i));
return;
}
struct ptlrpc_connection *
ptlrpc_connection_get(struct lnet_process_id peer, lnet_nid_t self,
struct obd_uuid *uuid)
{
struct ptlrpc_connection *conn, *conn2;
ENTRY;
peer.nid = LNetPrimaryNID(peer.nid);
conn = cfs_hash_lookup(conn_hash, &peer);
if (conn)
GOTO(out, conn);
OBD_ALLOC_PTR(conn);
if (!conn)
RETURN(NULL);
conn->c_peer = peer;
conn->c_self = self;
INIT_HLIST_NODE(&conn->c_hash);
atomic_set(&conn->c_refcount, 1);
if (uuid)
obd_str2uuid(&conn->c_remote_uuid, uuid->uuid);
/*
* Add the newly created conn to the hash, on key collision we
* lost a racing addition and must destroy our newly allocated
* connection. The object which exists in the hash will be
* returned and may be compared against out object.
*/
/* In the function below, .hs_keycmp resolves to
* conn_keycmp() */
/* coverity[overrun-buffer-val] */
conn2 = cfs_hash_findadd_unique(conn_hash, &peer, &conn->c_hash);
if (conn != conn2) {
OBD_FREE_PTR(conn);
conn = conn2;
}
EXIT;
out:
CDEBUG(D_INFO, "conn=%p refcount %d to %s\n",
conn, atomic_read(&conn->c_refcount),
libcfs_nid2str(conn->c_peer.nid));
return conn;
}
struct ptlrpc_connection *
ptlrpc_connection_get(lnet_process_id_t peer, lnet_nid_t self,
struct obd_uuid *uuid)
{
struct ptlrpc_connection *conn, *conn2;
conn = cfs_hash_lookup(conn_hash, &peer);
if (conn)
goto out;
conn = kzalloc(sizeof(*conn), GFP_NOFS);
if (!conn)
return NULL;
conn->c_peer = peer;
conn->c_self = self;
INIT_HLIST_NODE(&conn->c_hash);
atomic_set(&conn->c_refcount, 1);
if (uuid)
obd_str2uuid(&conn->c_remote_uuid, uuid->uuid);
/*
* Add the newly created conn to the hash, on key collision we
* lost a racing addition and must destroy our newly allocated
* connection. The object which exists in the has will be
* returned and may be compared against out object.
*/
/* In the function below, .hs_keycmp resolves to
* conn_keycmp()
*/
/* coverity[overrun-buffer-val] */
conn2 = cfs_hash_findadd_unique(conn_hash, &peer, &conn->c_hash);
if (conn != conn2) {
kfree(conn);
conn = conn2;
}
out:
CDEBUG(D_INFO, "conn=%p refcount %d to %s\n",
conn, atomic_read(&conn->c_refcount),
libcfs_nid2str(conn->c_peer.nid));
return conn;
}
static void print_lustre_cfg(struct lustre_cfg *lcfg)
{
int i;
if (!(libcfs_debug & D_OTHER)) /* don't loop on nothing */
return;
CDEBUG(D_OTHER, "lustre_cfg: %p\n", lcfg);
CDEBUG(D_OTHER, "\tlcfg->lcfg_version: %#x\n", lcfg->lcfg_version);
CDEBUG(D_OTHER, "\tlcfg->lcfg_command: %#x\n", lcfg->lcfg_command);
CDEBUG(D_OTHER, "\tlcfg->lcfg_num: %#x\n", lcfg->lcfg_num);
CDEBUG(D_OTHER, "\tlcfg->lcfg_flags: %#x\n", lcfg->lcfg_flags);
CDEBUG(D_OTHER, "\tlcfg->lcfg_nid: %s\n", libcfs_nid2str(lcfg->lcfg_nid));
CDEBUG(D_OTHER, "\tlcfg->lcfg_bufcount: %d\n", lcfg->lcfg_bufcount);
if (lcfg->lcfg_bufcount < LUSTRE_CFG_MAX_BUFCOUNT)
for (i = 0; i < lcfg->lcfg_bufcount; i++)
CDEBUG(D_OTHER, "\tlcfg->lcfg_buflens[%d]: %d\n",
i, lcfg->lcfg_buflens[i]);
}
static void lprocfs_free_client_stats(struct nid_stat *client_stat)
{
CDEBUG(D_CONFIG, "stat %p - data %p/%p\n", client_stat,
client_stat->nid_proc, client_stat->nid_stats);
LASSERTF(atomic_read(&client_stat->nid_exp_ref_count) == 0,
"nid %s:count %d\n", libcfs_nid2str(client_stat->nid),
atomic_read(&client_stat->nid_exp_ref_count));
if (client_stat->nid_proc)
lprocfs_remove(&client_stat->nid_proc);
if (client_stat->nid_stats)
lprocfs_free_stats(&client_stat->nid_stats);
if (client_stat->nid_ldlm_stats)
lprocfs_free_stats(&client_stat->nid_ldlm_stats);
OBD_FREE_PTR(client_stat);
return;
}
int mdt_init_idmap(struct tgt_session_info *tsi)
{
struct ptlrpc_request *req = tgt_ses_req(tsi);
struct mdt_export_data *med = mdt_req2med(req);
struct obd_export *exp = req->rq_export;
char *client = libcfs_nid2str(req->rq_peer.nid);
int rc = 0;
ENTRY;
if (exp_connect_rmtclient(exp)) {
mutex_lock(&med->med_idmap_mutex);
if (!med->med_idmap)
med->med_idmap = lustre_idmap_init();
mutex_unlock(&med->med_idmap_mutex);
if (IS_ERR(med->med_idmap)) {
long err = PTR_ERR(med->med_idmap);
med->med_idmap = NULL;
CERROR("%s: client %s -> target %s "
"failed to init idmap [%ld]!\n",
tgt_name(tsi->tsi_tgt), client,
tgt_name(tsi->tsi_tgt), err);
RETURN(err);
} else if (!med->med_idmap) {
CERROR("%s: client %s -> target %s "
"failed to init(2) idmap!\n",
tgt_name(tsi->tsi_tgt), client,
tgt_name(tsi->tsi_tgt));
RETURN(-ENOMEM);
}
CDEBUG(D_SEC, "%s: client %s -> target %s is remote.\n",
tgt_name(tsi->tsi_tgt), client,
tgt_name(tsi->tsi_tgt));
/* NB, MDS_CONNECT establish root idmap too! */
rc = mdt_handle_idmap(tsi);
}
RETURN(rc);
}
/* Ensure this is not a failover node that is connecting first*/
static int mgs_check_failover_reg(struct mgs_target_info *mti)
{
lnet_nid_t nid;
char *ptr;
int i;
ptr = mti->mti_params;
while (class_find_param(ptr, PARAM_FAILNODE, &ptr) == 0) {
while (class_parse_nid(ptr, &nid, &ptr) == 0) {
for (i = 0; i < mti->mti_nid_count; i++) {
if (nid == mti->mti_nids[i]) {
LCONSOLE_WARN("Denying initial registra"
"tion attempt from nid %s"
", specified as failover"
"\n",libcfs_nid2str(nid));
return -EADDRNOTAVAIL;
}
}
}
}
return 0;
}
int mdt_init_idmap(struct mdt_thread_info *info)
{
struct ptlrpc_request *req = mdt_info_req(info);
struct mdt_export_data *med = mdt_req2med(req);
struct obd_export *exp = req->rq_export;
char *client = libcfs_nid2str(req->rq_peer.nid);
struct obd_device *obd = exp->exp_obd;
int rc = 0;
ENTRY;
if (exp_connect_rmtclient(exp)) {
cfs_down(&med->med_idmap_sem);
if (!med->med_idmap)
med->med_idmap = lustre_idmap_init();
cfs_up(&med->med_idmap_sem);
if (IS_ERR(med->med_idmap)) {
long err = PTR_ERR(med->med_idmap);
med->med_idmap = NULL;
CERROR("client %s -> target %s "
"failed to init idmap [%ld]!\n",
client, obd->obd_name, err);
RETURN(err);
} else if (!med->med_idmap) {
CERROR("client %s -> target %s "
"failed to init(2) idmap!\n",
client, obd->obd_name);
RETURN(-ENOMEM);
}
CDEBUG(D_SEC, "client %s -> target %s is remote.\n",
client, obd->obd_name);
/* NB, MDS_CONNECT establish root idmap too! */
rc = mdt_handle_idmap(info);
}
RETURN(rc);
}
int lprocfs_exp_setup(struct obd_export *exp, lnet_nid_t *nid, int *newnid)
{
struct nid_stat *new_stat, *old_stat;
struct obd_device *obd = NULL;
struct proc_dir_entry *entry;
char *buffer = NULL;
int rc = 0;
ENTRY;
*newnid = 0;
if (!exp || !exp->exp_obd || !exp->exp_obd->obd_proc_exports_entry ||
!exp->exp_obd->obd_nid_stats_hash)
RETURN(-EINVAL);
/* not test against zero because eric say:
* You may only test nid against another nid, or LNET_NID_ANY.
* Anything else is nonsense.*/
if (!nid || *nid == LNET_NID_ANY)
RETURN(0);
spin_lock(&exp->exp_lock);
if (exp->exp_nid_stats != NULL) {
spin_unlock(&exp->exp_lock);
RETURN(-EALREADY);
}
spin_unlock(&exp->exp_lock);
obd = exp->exp_obd;
CDEBUG(D_CONFIG, "using hash %p\n", obd->obd_nid_stats_hash);
OBD_ALLOC_PTR(new_stat);
if (new_stat == NULL)
RETURN(-ENOMEM);
new_stat->nid = *nid;
new_stat->nid_obd = exp->exp_obd;
/* we need set default refcount to 1 to balance obd_disconnect */
atomic_set(&new_stat->nid_exp_ref_count, 1);
old_stat = cfs_hash_findadd_unique(obd->obd_nid_stats_hash,
nid, &new_stat->nid_hash);
CDEBUG(D_INFO, "Found stats %p for nid %s - ref %d\n",
old_stat, libcfs_nid2str(*nid),
atomic_read(&new_stat->nid_exp_ref_count));
/* Return -EALREADY here so that we know that the /proc
* entry already has been created */
if (old_stat != new_stat) {
spin_lock(&exp->exp_lock);
if (exp->exp_nid_stats) {
LASSERT(exp->exp_nid_stats == old_stat);
nidstat_putref(exp->exp_nid_stats);
}
exp->exp_nid_stats = old_stat;
spin_unlock(&exp->exp_lock);
GOTO(destroy_new, rc = -EALREADY);
}
/* not found - create */
OBD_ALLOC(buffer, LNET_NIDSTR_SIZE);
if (buffer == NULL)
GOTO(destroy_new, rc = -ENOMEM);
memcpy(buffer, libcfs_nid2str(*nid), LNET_NIDSTR_SIZE);
#ifndef HAVE_ONLY_PROCFS_SEQ
new_stat->nid_proc = lprocfs_register(buffer,
obd->obd_proc_exports_entry,
NULL, NULL);
#else
new_stat->nid_proc = lprocfs_seq_register(buffer,
obd->obd_proc_exports_entry,
NULL, NULL);
#endif
OBD_FREE(buffer, LNET_NIDSTR_SIZE);
if (IS_ERR(new_stat->nid_proc)) {
rc = PTR_ERR(new_stat->nid_proc);
new_stat->nid_proc = NULL;
CERROR("%s: cannot create proc entry for export %s: rc = %d\n",
obd->obd_name, libcfs_nid2str(*nid), rc);
GOTO(destroy_new_ns, rc);
}
entry = lprocfs_add_simple(new_stat->nid_proc, "uuid", new_stat,
&lprocfs_exp_uuid_fops);
if (IS_ERR(entry)) {
CWARN("Error adding the NID stats file\n");
rc = PTR_ERR(entry);
GOTO(destroy_new_ns, rc);
}
entry = lprocfs_add_simple(new_stat->nid_proc, "hash", new_stat,
&lprocfs_exp_hash_fops);
if (IS_ERR(entry)) {
CWARN("Error adding the hash file\n");
rc = PTR_ERR(entry);
GOTO(destroy_new_ns, rc);
}
spin_lock(&exp->exp_lock);
exp->exp_nid_stats = new_stat;
spin_unlock(&exp->exp_lock);
*newnid = 1;
/* protect competitive add to list, not need locking on destroy */
spin_lock(&obd->obd_nid_lock);
list_add(&new_stat->nid_list, &obd->obd_nid_stats);
spin_unlock(&obd->obd_nid_lock);
RETURN(rc);
destroy_new_ns:
if (new_stat->nid_proc != NULL)
lprocfs_remove(&new_stat->nid_proc);
cfs_hash_del(obd->obd_nid_stats_hash, nid, &new_stat->nid_hash);
destroy_new:
nidstat_putref(new_stat);
OBD_FREE_PTR(new_stat);
RETURN(rc);
}
/** Set up a mgc obd to process startup logs
*
* \param sb [in] super block of the mgc obd
*
* \retval 0 success, otherwise error code
*/
int lustre_start_mgc(struct super_block *sb)
{
struct obd_connect_data *data = NULL;
struct lustre_sb_info *lsi = s2lsi(sb);
struct obd_device *obd;
struct obd_export *exp;
struct obd_uuid *uuid;
class_uuid_t uuidc;
lnet_nid_t nid;
char *mgcname = NULL, *niduuid = NULL, *mgssec = NULL;
char *ptr;
int recov_bk;
int rc = 0, i = 0, j, len;
LASSERT(lsi->lsi_lmd);
/* Find the first non-lo MGS nid for our MGC name */
if (IS_SERVER(lsi)) {
/* mount -o mgsnode=nid */
ptr = lsi->lsi_lmd->lmd_mgs;
if (lsi->lsi_lmd->lmd_mgs &&
(class_parse_nid(lsi->lsi_lmd->lmd_mgs, &nid, &ptr) == 0)) {
i++;
} else if (IS_MGS(lsi)) {
lnet_process_id_t id;
while ((rc = LNetGetId(i++, &id)) != -ENOENT) {
if (LNET_NETTYP(LNET_NIDNET(id.nid)) == LOLND)
continue;
nid = id.nid;
i++;
break;
}
}
} else { /* client */
/* Use nids from mount line: uml1,1@elan:uml2,2@elan:/lustre */
ptr = lsi->lsi_lmd->lmd_dev;
if (class_parse_nid(ptr, &nid, &ptr) == 0)
i++;
}
if (i == 0) {
CERROR("No valid MGS nids found.\n");
return -EINVAL;
}
mutex_lock(&mgc_start_lock);
len = strlen(LUSTRE_MGC_OBDNAME) + strlen(libcfs_nid2str(nid)) + 1;
OBD_ALLOC(mgcname, len);
OBD_ALLOC(niduuid, len + 2);
if (!mgcname || !niduuid)
GOTO(out_free, rc = -ENOMEM);
sprintf(mgcname, "%s%s", LUSTRE_MGC_OBDNAME, libcfs_nid2str(nid));
mgssec = lsi->lsi_lmd->lmd_mgssec ? lsi->lsi_lmd->lmd_mgssec : "";
OBD_ALLOC_PTR(data);
if (data == NULL)
GOTO(out_free, rc = -ENOMEM);
obd = class_name2obd(mgcname);
if (obd && !obd->obd_stopping) {
rc = obd_set_info_async(NULL, obd->obd_self_export,
strlen(KEY_MGSSEC), KEY_MGSSEC,
strlen(mgssec), mgssec, NULL);
if (rc)
GOTO(out_free, rc);
/* Re-using an existing MGC */
atomic_inc(&obd->u.cli.cl_mgc_refcount);
/* IR compatibility check, only for clients */
if (lmd_is_client(lsi->lsi_lmd)) {
int has_ir;
int vallen = sizeof(*data);
__u32 *flags = &lsi->lsi_lmd->lmd_flags;
rc = obd_get_info(NULL, obd->obd_self_export,
strlen(KEY_CONN_DATA), KEY_CONN_DATA,
&vallen, data, NULL);
LASSERT(rc == 0);
has_ir = OCD_HAS_FLAG(data, IMP_RECOV);
if (has_ir ^ !(*flags & LMD_FLG_NOIR)) {
/* LMD_FLG_NOIR is for test purpose only */
LCONSOLE_WARN(
"Trying to mount a client with IR setting "
"not compatible with current mgc. "
"Force to use current mgc setting that is "
"IR %s.\n",
has_ir ? "enabled" : "disabled");
if (has_ir)
*flags &= ~LMD_FLG_NOIR;
else
*flags |= LMD_FLG_NOIR;
}
}
recov_bk = 0;
/* If we are restarting the MGS, don't try to keep the MGC's
old connection, or registration will fail. */
if (IS_MGS(lsi)) {
CDEBUG(D_MOUNT, "New MGS with live MGC\n");
recov_bk = 1;
}
/* Try all connections, but only once (again).
We don't want to block another target from starting
(using its local copy of the log), but we do want to connect
if at all possible. */
recov_bk++;
CDEBUG(D_MOUNT, "%s: Set MGC reconnect %d\n", mgcname,recov_bk);
rc = obd_set_info_async(NULL, obd->obd_self_export,
sizeof(KEY_INIT_RECOV_BACKUP),
KEY_INIT_RECOV_BACKUP,
sizeof(recov_bk), &recov_bk, NULL);
GOTO(out, rc = 0);
}
CDEBUG(D_MOUNT, "Start MGC '%s'\n", mgcname);
/* Add the primary nids for the MGS */
i = 0;
sprintf(niduuid, "%s_%x", mgcname, i);
if (IS_SERVER(lsi)) {
ptr = lsi->lsi_lmd->lmd_mgs;
if (IS_MGS(lsi)) {
/* Use local nids (including LO) */
lnet_process_id_t id;
while ((rc = LNetGetId(i++, &id)) != -ENOENT) {
rc = do_lcfg(mgcname, id.nid,
LCFG_ADD_UUID, niduuid, 0,0,0);
}
} else {
/* Use mgsnode= nids */
/* mount -o mgsnode=nid */
if (lsi->lsi_lmd->lmd_mgs) {
ptr = lsi->lsi_lmd->lmd_mgs;
} else if (class_find_param(ptr, PARAM_MGSNODE,
&ptr) != 0) {
CERROR("No MGS nids given.\n");
GOTO(out_free, rc = -EINVAL);
}
while (class_parse_nid(ptr, &nid, &ptr) == 0) {
rc = do_lcfg(mgcname, nid,
LCFG_ADD_UUID, niduuid, 0,0,0);
i++;
}
}
} else { /* client */
/* Use nids from mount line: uml1,1@elan:uml2,2@elan:/lustre */
ptr = lsi->lsi_lmd->lmd_dev;
while (class_parse_nid(ptr, &nid, &ptr) == 0) {
rc = do_lcfg(mgcname, nid,
LCFG_ADD_UUID, niduuid, 0,0,0);
i++;
/* Stop at the first failover nid */
if (*ptr == ':')
break;
}
}
if (i == 0) {
CERROR("No valid MGS nids found.\n");
GOTO(out_free, rc = -EINVAL);
}
lsi->lsi_lmd->lmd_mgs_failnodes = 1;
/* Random uuid for MGC allows easier reconnects */
OBD_ALLOC_PTR(uuid);
ll_generate_random_uuid(uuidc);
class_uuid_unparse(uuidc, uuid);
/* Start the MGC */
rc = lustre_start_simple(mgcname, LUSTRE_MGC_NAME,
(char *)uuid->uuid, LUSTRE_MGS_OBDNAME,
niduuid, 0, 0);
OBD_FREE_PTR(uuid);
if (rc)
GOTO(out_free, rc);
/* Add any failover MGS nids */
i = 1;
while (ptr && ((*ptr == ':' ||
class_find_param(ptr, PARAM_MGSNODE, &ptr) == 0))) {
/* New failover node */
sprintf(niduuid, "%s_%x", mgcname, i);
j = 0;
while (class_parse_nid_quiet(ptr, &nid, &ptr) == 0) {
j++;
rc = do_lcfg(mgcname, nid,
LCFG_ADD_UUID, niduuid, 0,0,0);
if (*ptr == ':')
break;
}
if (j > 0) {
rc = do_lcfg(mgcname, 0, LCFG_ADD_CONN,
niduuid, 0, 0, 0);
i++;
} else {
/* at ":/fsname" */
break;
}
}
lsi->lsi_lmd->lmd_mgs_failnodes = i;
obd = class_name2obd(mgcname);
if (!obd) {
CERROR("Can't find mgcobd %s\n", mgcname);
GOTO(out_free, rc = -ENOTCONN);
}
rc = obd_set_info_async(NULL, obd->obd_self_export,
strlen(KEY_MGSSEC), KEY_MGSSEC,
strlen(mgssec), mgssec, NULL);
if (rc)
GOTO(out_free, rc);
/* Keep a refcount of servers/clients who started with "mount",
so we know when we can get rid of the mgc. */
atomic_set(&obd->u.cli.cl_mgc_refcount, 1);
/* Try all connections, but only once. */
recov_bk = 1;
rc = obd_set_info_async(NULL, obd->obd_self_export,
sizeof(KEY_INIT_RECOV_BACKUP),
KEY_INIT_RECOV_BACKUP,
sizeof(recov_bk), &recov_bk, NULL);
if (rc)
/* nonfatal */
CWARN("can't set %s %d\n", KEY_INIT_RECOV_BACKUP, rc);
/* We connect to the MGS at setup, and don't disconnect until cleanup */
data->ocd_connect_flags = OBD_CONNECT_VERSION | OBD_CONNECT_AT |
OBD_CONNECT_FULL20 | OBD_CONNECT_IMP_RECOV |
OBD_CONNECT_LVB_TYPE;
#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 2, 50, 0)
data->ocd_connect_flags |= OBD_CONNECT_MNE_SWAB;
#else
#warning "LU-1644: Remove old OBD_CONNECT_MNE_SWAB fixup and imp_need_mne_swab"
#endif
if (lmd_is_client(lsi->lsi_lmd) &&
lsi->lsi_lmd->lmd_flags & LMD_FLG_NOIR)
data->ocd_connect_flags &= ~OBD_CONNECT_IMP_RECOV;
data->ocd_version = LUSTRE_VERSION_CODE;
rc = obd_connect(NULL, &exp, obd, &(obd->obd_uuid), data, NULL);
if (rc) {
CERROR("connect failed %d\n", rc);
GOTO(out, rc);
}
obd->u.cli.cl_mgc_mgsexp = exp;
out:
/* Keep the mgc info in the sb. Note that many lsi's can point
to the same mgc.*/
lsi->lsi_mgc = obd;
out_free:
mutex_unlock(&mgc_start_lock);
if (data)
OBD_FREE_PTR(data);
if (mgcname)
OBD_FREE(mgcname, len);
if (niduuid)
OBD_FREE(niduuid, len + 2);
return rc;
}
void
lnet_finalize (__unusedx lnet_ni_t *ni, lnet_msg_t *msg, int status)
{
#ifdef __KERNEL__
int i;
int my_slot;
#endif
lnet_libmd_t *md;
LASSERT (!in_interrupt ());
if (msg == NULL)
return;
#if 0
CDEBUG(D_WARNING, "%s msg->%s Flags:%s%s%s%s%s%s%s%s%s%s%s txp %s rxp %s\n",
lnet_msgtyp2str(msg->msg_type), libcfs_id2str(msg->msg_target),
msg->msg_target_is_router ? "t" : "",
msg->msg_routing ? "X" : "",
msg->msg_ack ? "A" : "",
msg->msg_sending ? "S" : "",
msg->msg_receiving ? "R" : "",
msg->msg_delayed ? "d" : "",
msg->msg_txcredit ? "C" : "",
msg->msg_peertxcredit ? "c" : "",
msg->msg_rtrcredit ? "F" : "",
msg->msg_peerrtrcredit ? "f" : "",
msg->msg_onactivelist ? "!" : "",
msg->msg_txpeer == NULL ? "<none>" : libcfs_nid2str(msg->msg_txpeer->lp_nid),
msg->msg_rxpeer == NULL ? "<none>" : libcfs_nid2str(msg->msg_rxpeer->lp_nid));
#endif
LNET_LOCK();
LASSERT (msg->msg_onactivelist);
msg->msg_ev.status = status;
md = msg->msg_md;
if (md != NULL) {
int unlink;
/* Now it's safe to drop my caller's ref */
md->md_refcount--;
LASSERT (md->md_refcount >= 0);
unlink = lnet_md_unlinkable(md);
msg->msg_ev.unlinked = unlink;
if (md->md_eq != NULL)
lnet_enq_event_locked(md->md_eq, &msg->msg_ev);
if (unlink)
lnet_md_unlink(md);
msg->msg_md = NULL;
}
list_add_tail (&msg->msg_list, &the_lnet.ln_finalizeq);
/* Recursion breaker. Don't complete the message here if I am (or
* enough other threads are) already completing messages */
#ifdef __KERNEL__
my_slot = -1;
for (i = 0; i < the_lnet.ln_nfinalizers; i++) {
if (the_lnet.ln_finalizers[i] == cfs_current())
goto out;
if (my_slot < 0 && the_lnet.ln_finalizers[i] == NULL)
my_slot = i;
}
if (my_slot < 0)
goto out;
the_lnet.ln_finalizers[my_slot] = cfs_current();
#else
if (the_lnet.ln_finalizing)
goto out;
the_lnet.ln_finalizing = 1;
#endif
while (!list_empty(&the_lnet.ln_finalizeq)) {
msg = list_entry(the_lnet.ln_finalizeq.next,
lnet_msg_t, msg_list);
list_del(&msg->msg_list);
/* NB drops and regains the lnet lock if it actually does
* anything, so my finalizing friends can chomp along too */
lnet_complete_msg_locked(msg);
}
#ifdef __KERNEL__
the_lnet.ln_finalizers[my_slot] = NULL;
#else
the_lnet.ln_finalizing = 0;
#endif
out:
LNET_UNLOCK();
}
int
lnet_accept(socket_t *sock, __u32 magic)
{
lnet_acceptor_connreq_t cr;
__u32 peer_ip;
int peer_port;
int rc;
int flip;
lnet_ni_t *ni;
char *str;
LASSERT(sizeof(cr) <= 16); /* not too big for the stack */
rc = libcfs_sock_getaddr(sock, 1, &peer_ip, &peer_port);
LASSERT(rc == 0); /* we succeeded before */
if (!lnet_accept_magic(magic, LNET_PROTO_ACCEPTOR_MAGIC)) {
if (lnet_accept_magic(magic, LNET_PROTO_MAGIC)) {
/* future version compatibility!
* When LNET unifies protocols over all LNDs, the first
* thing sent will be a version query. I send back
* LNET_PROTO_ACCEPTOR_MAGIC to tell her I'm "old" */
memset(&cr, 0, sizeof(cr));
cr.acr_magic = LNET_PROTO_ACCEPTOR_MAGIC;
cr.acr_version = LNET_PROTO_ACCEPTOR_VERSION;
rc = libcfs_sock_write(sock, &cr, sizeof(cr),
accept_timeout);
if (rc != 0)
CERROR("Error sending magic+version in response to LNET magic from %pI4h: %d\n",
&peer_ip, rc);
return -EPROTO;
}
if (magic == le32_to_cpu(LNET_PROTO_TCP_MAGIC))
str = "'old' socknal/tcpnal";
else if (lnet_accept_magic(magic, LNET_PROTO_RA_MAGIC))
str = "'old' ranal";
else
str = "unrecognised";
LCONSOLE_ERROR_MSG(0x11f, "Refusing connection from %pI4h magic %08x: %s acceptor protocol\n",
&peer_ip, magic, str);
return -EPROTO;
}
flip = (magic != LNET_PROTO_ACCEPTOR_MAGIC);
rc = libcfs_sock_read(sock, &cr.acr_version,
sizeof(cr.acr_version),
accept_timeout);
if (rc != 0) {
CERROR("Error %d reading connection request version from %pI4h\n",
rc, &peer_ip);
return -EIO;
}
if (flip)
__swab32s(&cr.acr_version);
if (cr.acr_version != LNET_PROTO_ACCEPTOR_VERSION) {
/* future version compatibility!
* An acceptor-specific protocol rev will first send a version
* query. I send back my current version to tell her I'm
* "old". */
int peer_version = cr.acr_version;
memset(&cr, 0, sizeof(cr));
cr.acr_magic = LNET_PROTO_ACCEPTOR_MAGIC;
cr.acr_version = LNET_PROTO_ACCEPTOR_VERSION;
rc = libcfs_sock_write(sock, &cr, sizeof(cr),
accept_timeout);
if (rc != 0)
CERROR("Error sending magic+version in response to version %d from %pI4h: %d\n",
peer_version, &peer_ip, rc);
return -EPROTO;
}
rc = libcfs_sock_read(sock, &cr.acr_nid,
sizeof(cr) -
offsetof(lnet_acceptor_connreq_t, acr_nid),
accept_timeout);
if (rc != 0) {
CERROR("Error %d reading connection request from %pI4h\n",
rc, &peer_ip);
return -EIO;
}
if (flip)
__swab64s(&cr.acr_nid);
ni = lnet_net2ni(LNET_NIDNET(cr.acr_nid));
if (ni == NULL || /* no matching net */
ni->ni_nid != cr.acr_nid) { /* right NET, wrong NID! */
if (ni != NULL)
lnet_ni_decref(ni);
LCONSOLE_ERROR_MSG(0x120, "Refusing connection from %pI4h for %s: No matching NI\n",
&peer_ip, libcfs_nid2str(cr.acr_nid));
return -EPERM;
}
if (ni->ni_lnd->lnd_accept == NULL) {
/* This catches a request for the loopback LND */
lnet_ni_decref(ni);
LCONSOLE_ERROR_MSG(0x121, "Refusing connection from %pI4h for %s: NI doesn not accept IP connections\n",
&peer_ip, libcfs_nid2str(cr.acr_nid));
return -EPERM;
}
CDEBUG(D_NET, "Accept %s from %pI4h\n",
libcfs_nid2str(cr.acr_nid), &peer_ip);
rc = ni->ni_lnd->lnd_accept(ni, sock);
lnet_ni_decref(ni);
return rc;
}
int lustre_lnet_show_peer_credits(int seq_no, struct cYAML **show_rc,
struct cYAML **err_rc)
{
struct lnet_ioctl_peer peer_info;
int rc = LUSTRE_CFG_RC_OUT_OF_MEM, ncpt = 0, i = 0, j = 0;
int l_errno = 0;
struct cYAML *root = NULL, *peer = NULL, *first_seq = NULL,
*peer_root = NULL;
char err_str[LNET_MAX_STR_LEN];
bool ncpt_set = false;
snprintf(err_str, sizeof(err_str),
"\"out of memory\"");
/* create struct cYAML root object */
root = cYAML_create_object(NULL, NULL);
if (root == NULL)
goto out;
peer_root = cYAML_create_seq(root, "peer");
if (peer_root == NULL)
goto out;
do {
for (i = 0;; i++) {
LIBCFS_IOC_INIT_V2(peer_info, pr_hdr);
peer_info.pr_count = i;
peer_info.pr_lnd_u.pr_peer_credits.cr_ncpt = j;
rc = l_ioctl(LNET_DEV_ID,
IOC_LIBCFS_GET_PEER_INFO, &peer_info);
if (rc != 0) {
l_errno = errno;
break;
}
if (ncpt_set != 0) {
ncpt = peer_info.pr_lnd_u.pr_peer_credits.
cr_ncpt;
ncpt_set = true;
}
peer = cYAML_create_seq_item(peer_root);
if (peer == NULL)
goto out;
if (first_seq == NULL)
first_seq = peer;
if (cYAML_create_string(peer, "nid",
libcfs_nid2str
(peer_info.pr_nid)) == NULL)
goto out;
if (cYAML_create_string(peer, "state",
peer_info.pr_lnd_u.
pr_peer_credits.
cr_aliveness) ==
NULL)
goto out;
if (cYAML_create_number(peer, "refcount",
peer_info.pr_lnd_u.
pr_peer_credits.
cr_refcount) == NULL)
goto out;
if (cYAML_create_number(peer, "max_ni_tx_credits",
peer_info.pr_lnd_u.
pr_peer_credits.
cr_ni_peer_tx_credits)
== NULL)
goto out;
if (cYAML_create_number(peer, "available_tx_credits",
peer_info.pr_lnd_u.
pr_peer_credits.
cr_peer_tx_credits)
== NULL)
goto out;
if (cYAML_create_number(peer, "available_rtr_credits",
peer_info.pr_lnd_u.
pr_peer_credits.
cr_peer_rtr_credits)
== NULL)
goto out;
if (cYAML_create_number(peer, "min_rtr_credits",
peer_info.pr_lnd_u.
pr_peer_credits.
cr_peer_min_rtr_credits)
== NULL)
goto out;
if (cYAML_create_number(peer, "tx_q_num_of_buf",
peer_info.pr_lnd_u.
pr_peer_credits.
cr_peer_tx_qnob)
== NULL)
goto out;
}
if (l_errno != ENOENT) {
snprintf(err_str,
sizeof(err_str),
"\"cannot get peer information: %s\"",
strerror(l_errno));
rc = -l_errno;
goto out;
}
j++;
} while (j < ncpt);
/* print output iff show_rc is not provided */
if (show_rc == NULL)
cYAML_print_tree(root);
snprintf(err_str, sizeof(err_str), "\"success\"");
rc = LUSTRE_CFG_RC_NO_ERR;
out:
if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR) {
cYAML_free_tree(root);
} else if (show_rc != NULL && *show_rc != NULL) {
struct cYAML *show_node;
/* find the peer node, if one doesn't exist then
* insert one. Otherwise add to the one there
*/
show_node = cYAML_get_object_item(*show_rc,
"peer_credits");
if (show_node != NULL && cYAML_is_sequence(show_node)) {
cYAML_insert_child(show_node, first_seq);
free(peer_root);
free(root);
} else if (show_node == NULL) {
cYAML_insert_sibling((*show_rc)->cy_child,
peer_root);
free(root);
} else {
cYAML_free_tree(root);
}
} else {
*show_rc = root;
}
cYAML_build_error(rc, seq_no, SHOW_CMD, "peer_credits", err_str,
err_rc);
return rc;
}
int mdt_check_ucred(struct mdt_thread_info *info)
{
struct ptlrpc_request *req = mdt_info_req(info);
struct mdt_device *mdt = info->mti_mdt;
struct ptlrpc_user_desc *pud = req->rq_user_desc;
struct md_ucred *ucred = mdt_ucred(info);
struct md_identity *identity = NULL;
lnet_nid_t peernid = req->rq_peer.nid;
__u32 perm = 0;
__u32 remote = exp_connect_rmtclient(info->mti_exp);
int setuid;
int setgid;
int rc = 0;
ENTRY;
if ((ucred->mu_valid == UCRED_OLD) || (ucred->mu_valid == UCRED_NEW))
RETURN(0);
if (!req->rq_auth_gss || req->rq_auth_usr_mdt || !req->rq_user_desc)
RETURN(0);
/* sanity check: if we use strong authentication, we expect the
* uid which client claimed is true */
if (remote) {
if (req->rq_auth_mapped_uid == INVALID_UID) {
CDEBUG(D_SEC, "remote user not mapped, deny access!\n");
RETURN(-EACCES);
}
if (ptlrpc_user_desc_do_idmap(req, pud))
RETURN(-EACCES);
if (req->rq_auth_mapped_uid != pud->pud_uid) {
CDEBUG(D_SEC, "remote client %s: auth/mapped uid %u/%u "
"while client claims %u:%u/%u:%u\n",
libcfs_nid2str(peernid), req->rq_auth_uid,
req->rq_auth_mapped_uid,
pud->pud_uid, pud->pud_gid,
pud->pud_fsuid, pud->pud_fsgid);
RETURN(-EACCES);
}
} else {
if (req->rq_auth_uid != pud->pud_uid) {
CDEBUG(D_SEC, "local client %s: auth uid %u "
"while client claims %u:%u/%u:%u\n",
libcfs_nid2str(peernid), req->rq_auth_uid,
pud->pud_uid, pud->pud_gid,
pud->pud_fsuid, pud->pud_fsgid);
RETURN(-EACCES);
}
}
if (is_identity_get_disabled(mdt->mdt_identity_cache)) {
if (remote) {
CDEBUG(D_SEC, "remote client must run with identity_get "
"enabled!\n");
RETURN(-EACCES);
}
RETURN(0);
}
identity = mdt_identity_get(mdt->mdt_identity_cache, pud->pud_uid);
if (IS_ERR(identity)) {
if (unlikely(PTR_ERR(identity) == -EREMCHG &&
!remote)) {
RETURN(0);
} else {
CDEBUG(D_SEC, "Deny access without identity: uid %u\n",
pud->pud_uid);
RETURN(-EACCES);
}
}
perm = mdt_identity_get_perm(identity, remote, peernid);
/* find out the setuid/setgid attempt */
setuid = (pud->pud_uid != pud->pud_fsuid);
setgid = (pud->pud_gid != pud->pud_fsgid ||
pud->pud_gid != identity->mi_gid);
/* check permission of setuid */
if (setuid && !(perm & CFS_SETUID_PERM)) {
CDEBUG(D_SEC, "mdt blocked setuid attempt (%u -> %u) from %s\n",
pud->pud_uid, pud->pud_fsuid, libcfs_nid2str(peernid));
GOTO(out, rc = -EACCES);
}
/* check permission of setgid */
if (setgid && !(perm & CFS_SETGID_PERM)) {
CDEBUG(D_SEC, "mdt blocked setgid attempt (%u:%u/%u:%u -> %u) "
"from %s\n", pud->pud_uid, pud->pud_gid,
pud->pud_fsuid, pud->pud_fsgid, identity->mi_gid,
libcfs_nid2str(peernid));
GOTO(out, rc = -EACCES);
}
EXIT;
out:
mdt_identity_put(mdt->mdt_identity_cache, identity);
return rc;
}
static int new_init_ucred(struct mdt_thread_info *info, ucred_init_type_t type,
void *buf)
{
struct ptlrpc_request *req = mdt_info_req(info);
struct mdt_device *mdt = info->mti_mdt;
struct ptlrpc_user_desc *pud = req->rq_user_desc;
struct md_ucred *ucred = mdt_ucred(info);
lnet_nid_t peernid = req->rq_peer.nid;
__u32 perm = 0;
__u32 remote = exp_connect_rmtclient(info->mti_exp);
int setuid;
int setgid;
int rc = 0;
ENTRY;
LASSERT(req->rq_auth_gss);
LASSERT(!req->rq_auth_usr_mdt);
LASSERT(req->rq_user_desc);
ucred->mu_valid = UCRED_INVALID;
ucred->mu_o_uid = pud->pud_uid;
ucred->mu_o_gid = pud->pud_gid;
ucred->mu_o_fsuid = pud->pud_fsuid;
ucred->mu_o_fsgid = pud->pud_fsgid;
if (type == BODY_INIT) {
struct mdt_body *body = (struct mdt_body *)buf;
ucred->mu_suppgids[0] = body->suppgid;
ucred->mu_suppgids[1] = -1;
}
/* sanity check: we expect the uid which client claimed is true */
if (remote) {
if (req->rq_auth_mapped_uid == INVALID_UID) {
CDEBUG(D_SEC, "remote user not mapped, deny access!\n");
RETURN(-EACCES);
}
if (ptlrpc_user_desc_do_idmap(req, pud))
RETURN(-EACCES);
if (req->rq_auth_mapped_uid != pud->pud_uid) {
CDEBUG(D_SEC, "remote client %s: auth/mapped uid %u/%u "
"while client claims %u:%u/%u:%u\n",
libcfs_nid2str(peernid), req->rq_auth_uid,
req->rq_auth_mapped_uid,
pud->pud_uid, pud->pud_gid,
pud->pud_fsuid, pud->pud_fsgid);
RETURN(-EACCES);
}
} else {
if (req->rq_auth_uid != pud->pud_uid) {
CDEBUG(D_SEC, "local client %s: auth uid %u "
"while client claims %u:%u/%u:%u\n",
libcfs_nid2str(peernid), req->rq_auth_uid,
pud->pud_uid, pud->pud_gid,
pud->pud_fsuid, pud->pud_fsgid);
RETURN(-EACCES);
}
}
if (is_identity_get_disabled(mdt->mdt_identity_cache)) {
if (remote) {
CDEBUG(D_SEC, "remote client must run with identity_get "
"enabled!\n");
RETURN(-EACCES);
} else {
ucred->mu_identity = NULL;
perm = CFS_SETUID_PERM | CFS_SETGID_PERM |
CFS_SETGRP_PERM;
}
} else {
struct md_identity *identity;
identity = mdt_identity_get(mdt->mdt_identity_cache,
pud->pud_uid);
if (IS_ERR(identity)) {
if (unlikely(PTR_ERR(identity) == -EREMCHG &&
!remote)) {
ucred->mu_identity = NULL;
perm = CFS_SETUID_PERM | CFS_SETGID_PERM |
CFS_SETGRP_PERM;
} else {
CDEBUG(D_SEC, "Deny access without identity: uid %u\n",
pud->pud_uid);
RETURN(-EACCES);
}
} else {
ucred->mu_identity = identity;
perm = mdt_identity_get_perm(ucred->mu_identity,
remote, peernid);
}
}
/* find out the setuid/setgid attempt */
setuid = (pud->pud_uid != pud->pud_fsuid);
setgid = ((pud->pud_gid != pud->pud_fsgid) ||
(ucred->mu_identity &&
(pud->pud_gid != ucred->mu_identity->mi_gid)));
/* check permission of setuid */
if (setuid && !(perm & CFS_SETUID_PERM)) {
CDEBUG(D_SEC, "mdt blocked setuid attempt (%u -> %u) from %s\n",
pud->pud_uid, pud->pud_fsuid, libcfs_nid2str(peernid));
GOTO(out, rc = -EACCES);
}
/* check permission of setgid */
if (setgid && !(perm & CFS_SETGID_PERM)) {
CDEBUG(D_SEC, "mdt blocked setgid attempt (%u:%u/%u:%u -> %u) "
"from %s\n", pud->pud_uid, pud->pud_gid,
pud->pud_fsuid, pud->pud_fsgid,
ucred->mu_identity->mi_gid, libcfs_nid2str(peernid));
GOTO(out, rc = -EACCES);
}
/*
* NB: remote client not allowed to setgroups anyway.
*/
if (!remote && perm & CFS_SETGRP_PERM) {
if (pud->pud_ngroups) {
/* setgroups for local client */
ucred->mu_ginfo = cfs_groups_alloc(pud->pud_ngroups);
if (!ucred->mu_ginfo) {
CERROR("failed to alloc %d groups\n",
pud->pud_ngroups);
GOTO(out, rc = -ENOMEM);
}
lustre_groups_from_list(ucred->mu_ginfo,
pud->pud_groups);
lustre_groups_sort(ucred->mu_ginfo);
} else {
ucred->mu_ginfo = NULL;
}
} else {
ucred->mu_suppgids[0] = -1;
ucred->mu_suppgids[1] = -1;
ucred->mu_ginfo = NULL;
}
ucred->mu_uid = pud->pud_uid;
ucred->mu_gid = pud->pud_gid;
ucred->mu_fsuid = pud->pud_fsuid;
ucred->mu_fsgid = pud->pud_fsgid;
/* process root_squash here. */
mdt_root_squash(info, peernid);
/* remove fs privilege for non-root user. */
if (ucred->mu_fsuid)
ucred->mu_cap = pud->pud_cap & ~CFS_CAP_FS_MASK;
else
ucred->mu_cap = pud->pud_cap;
if (remote && !(perm & CFS_RMTOWN_PERM))
ucred->mu_cap &= ~(CFS_CAP_SYS_RESOURCE_MASK |
CFS_CAP_CHOWN_MASK);
ucred->mu_valid = UCRED_NEW;
EXIT;
out:
if (rc) {
if (ucred->mu_ginfo) {
cfs_put_group_info(ucred->mu_ginfo);
ucred->mu_ginfo = NULL;
}
if (ucred->mu_identity) {
mdt_identity_put(mdt->mdt_identity_cache,
ucred->mu_identity);
ucred->mu_identity = NULL;
}
}
return rc;
}
int lustre_lnet_show_route(char *nw, char *gw, int hops, int prio, int detail,
int seq_no, struct cYAML **show_rc,
struct cYAML **err_rc)
{
struct lnet_ioctl_config_data data;
lnet_nid_t gateway_nid;
int rc = LUSTRE_CFG_RC_OUT_OF_MEM;
int l_errno = 0;
__u32 net = LNET_NIDNET(LNET_NID_ANY);
int i;
struct cYAML *root = NULL, *route = NULL, *item = NULL;
struct cYAML *first_seq = NULL;
char err_str[LNET_MAX_STR_LEN];
bool exist = false;
snprintf(err_str, sizeof(err_str),
"\"out of memory\"");
if (nw != NULL) {
net = libcfs_str2net(nw);
if (net == LNET_NIDNET(LNET_NID_ANY)) {
snprintf(err_str,
sizeof(err_str),
"\"cannot parse net '%s'\"", nw);
rc = LUSTRE_CFG_RC_BAD_PARAM;
goto out;
}
if (LNET_NETTYP(net) == CIBLND ||
LNET_NETTYP(net) == OPENIBLND ||
LNET_NETTYP(net) == IIBLND ||
LNET_NETTYP(net) == VIBLND) {
snprintf(err_str,
sizeof(err_str),
"\"obsolete LNet type '%s'\"",
libcfs_lnd2str(net));
rc = LUSTRE_CFG_RC_BAD_PARAM;
goto out;
}
} else {
/* show all routes without filtering on net */
net = LNET_NIDNET(LNET_NID_ANY);
}
if (gw != NULL) {
gateway_nid = libcfs_str2nid(gw);
if (gateway_nid == LNET_NID_ANY) {
snprintf(err_str,
sizeof(err_str),
"\"cannot parse gateway NID '%s'\"", gw);
rc = LUSTRE_CFG_RC_BAD_PARAM;
goto out;
}
} else
/* show all routes with out filtering on gateway */
gateway_nid = LNET_NID_ANY;
if ((hops < 1 && hops != -1) || hops > 255) {
snprintf(err_str,
sizeof(err_str),
"\"invalid hop count %d, must be between 0 and 256\"",
hops);
rc = LUSTRE_CFG_RC_OUT_OF_RANGE_PARAM;
goto out;
}
/* create struct cYAML root object */
root = cYAML_create_object(NULL, NULL);
if (root == NULL)
goto out;
route = cYAML_create_seq(root, "route");
if (route == NULL)
goto out;
for (i = 0;; i++) {
LIBCFS_IOC_INIT_V2(data, cfg_hdr);
data.cfg_count = i;
rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_ROUTE, &data);
if (rc != 0) {
l_errno = errno;
break;
}
/* filter on provided data */
if (net != LNET_NIDNET(LNET_NID_ANY) &&
net != data.cfg_net)
continue;
if (gateway_nid != LNET_NID_ANY &&
gateway_nid != data.cfg_nid)
continue;
if (hops != -1 &&
hops != data.cfg_config_u.cfg_route.rtr_hop)
continue;
if (prio != -1 &&
prio != data.cfg_config_u.cfg_route.rtr_priority)
continue;
/* default rc to -1 incase we hit the goto */
rc = -1;
exist = true;
item = cYAML_create_seq_item(route);
if (item == NULL)
goto out;
if (first_seq == NULL)
first_seq = item;
if (cYAML_create_string(item, "net",
libcfs_net2str(data.cfg_net)) == NULL)
goto out;
if (cYAML_create_string(item, "gateway",
libcfs_nid2str(data.cfg_nid)) == NULL)
goto out;
if (detail) {
if (cYAML_create_number(item, "hop",
data.cfg_config_u.cfg_route.
rtr_hop) ==
NULL)
goto out;
if (cYAML_create_number(item, "priority",
data.cfg_config_u.
cfg_route.rtr_priority) == NULL)
goto out;
if (cYAML_create_string(item, "state",
data.cfg_config_u.cfg_route.
rtr_flags ?
"up" : "down") == NULL)
goto out;
}
}
/* print output iff show_rc is not provided */
if (show_rc == NULL)
cYAML_print_tree(root);
if (l_errno != ENOENT) {
snprintf(err_str,
sizeof(err_str),
"\"cannot get routes: %s\"",
strerror(l_errno));
rc = -l_errno;
goto out;
} else
rc = LUSTRE_CFG_RC_NO_ERR;
snprintf(err_str, sizeof(err_str), "\"success\"");
out:
if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR || !exist) {
cYAML_free_tree(root);
} else if (show_rc != NULL && *show_rc != NULL) {
struct cYAML *show_node;
/* find the route node, if one doesn't exist then
* insert one. Otherwise add to the one there
*/
show_node = cYAML_get_object_item(*show_rc, "route");
if (show_node != NULL && cYAML_is_sequence(show_node)) {
cYAML_insert_child(show_node, first_seq);
free(route);
free(root);
} else if (show_node == NULL) {
cYAML_insert_sibling((*show_rc)->cy_child,
route);
free(root);
} else {
cYAML_free_tree(root);
}
} else {
*show_rc = root;
}
cYAML_build_error(rc, seq_no, SHOW_CMD, "route", err_str, err_rc);
return rc;
}
int lustre_lnet_show_net(char *nw, int detail, int seq_no,
struct cYAML **show_rc, struct cYAML **err_rc)
{
char *buf;
struct lnet_ioctl_config_lnd_tunables *lnd_cfg;
struct lnet_ioctl_config_data *data;
struct lnet_ioctl_net_config *net_config;
__u32 net = LNET_NIDNET(LNET_NID_ANY);
int rc = LUSTRE_CFG_RC_OUT_OF_MEM, i, j;
int l_errno = 0;
struct cYAML *root = NULL, *tunables = NULL, *net_node = NULL,
*interfaces = NULL, *item = NULL, *first_seq = NULL;
int str_buf_len = LNET_MAX_SHOW_NUM_CPT * 2;
char str_buf[str_buf_len];
char *pos;
char err_str[LNET_MAX_STR_LEN];
bool exist = false;
size_t buf_len;
snprintf(err_str, sizeof(err_str), "\"out of memory\"");
buf_len = sizeof(*data) + sizeof(*net_config) + sizeof(*lnd_cfg);
buf = calloc(1, buf_len);
if (buf == NULL)
goto out;
data = (struct lnet_ioctl_config_data *)buf;
if (nw != NULL) {
net = libcfs_str2net(nw);
if (net == LNET_NIDNET(LNET_NID_ANY)) {
snprintf(err_str,
sizeof(err_str),
"\"cannot parse net '%s'\"", nw);
rc = LUSTRE_CFG_RC_BAD_PARAM;
goto out;
}
}
root = cYAML_create_object(NULL, NULL);
if (root == NULL)
goto out;
net_node = cYAML_create_seq(root, "net");
if (net_node == NULL)
goto out;
for (i = 0;; i++) {
pos = str_buf;
memset(buf, 0, buf_len);
LIBCFS_IOC_INIT_V2(*data, cfg_hdr);
/*
* set the ioc_len to the proper value since INIT assumes
* size of data
*/
data->cfg_hdr.ioc_len = buf_len;
data->cfg_count = i;
rc = l_ioctl(LNET_DEV_ID, IOC_LIBCFS_GET_NET, data);
if (rc != 0) {
l_errno = errno;
break;
}
/* filter on provided data */
if (net != LNET_NIDNET(LNET_NID_ANY) &&
net != LNET_NIDNET(data->cfg_nid))
continue;
/* default rc to -1 in case we hit the goto */
rc = -1;
exist = true;
net_config = (struct lnet_ioctl_net_config *)data->cfg_bulk;
/* create the tree to be printed. */
item = cYAML_create_seq_item(net_node);
if (item == NULL)
goto out;
if (first_seq == NULL)
first_seq = item;
if (cYAML_create_string(item, "net",
libcfs_net2str(
LNET_NIDNET(data->cfg_nid)))
== NULL)
goto out;
if (cYAML_create_string(item, "nid",
libcfs_nid2str(data->cfg_nid)) == NULL)
goto out;
if (cYAML_create_string(item, "status",
(net_config->ni_status ==
LNET_NI_STATUS_UP) ?
"up" : "down") == NULL)
goto out;
/* don't add interfaces unless there is at least one
* interface */
if (strlen(net_config->ni_interfaces[0]) > 0) {
interfaces = cYAML_create_object(item, "interfaces");
if (interfaces == NULL)
goto out;
for (j = 0; j < LNET_MAX_INTERFACES; j++) {
if (lustre_interface_show_net(interfaces, j,
detail, data,
net_config) < 0)
goto out;
}
}
if (detail) {
char *limit;
tunables = cYAML_create_object(item, "tunables");
if (tunables == NULL)
goto out;
if (cYAML_create_number(tunables, "peer_timeout",
data->cfg_config_u.cfg_net.
net_peer_timeout) == NULL)
goto out;
if (cYAML_create_number(tunables, "peer_credits",
data->cfg_config_u.cfg_net.
net_peer_tx_credits) == NULL)
goto out;
if (cYAML_create_number(tunables,
"peer_buffer_credits",
data->cfg_config_u.cfg_net.
net_peer_rtr_credits) == NULL)
goto out;
if (cYAML_create_number(tunables, "credits",
data->cfg_config_u.cfg_net.
net_max_tx_credits) == NULL)
goto out;
/* out put the CPTs in the format: "[x,x,x,...]" */
limit = str_buf + str_buf_len - 3;
pos += snprintf(pos, limit - pos, "\"[");
for (j = 0 ; data->cfg_ncpts > 1 &&
j < data->cfg_ncpts &&
pos < limit; j++) {
pos += snprintf(pos, limit - pos,
"%d", net_config->ni_cpts[j]);
if ((j + 1) < data->cfg_ncpts)
pos += snprintf(pos, limit - pos, ",");
}
pos += snprintf(pos, 3, "]\"");
if (data->cfg_ncpts > 1 &&
cYAML_create_string(tunables, "CPT",
str_buf) == NULL)
goto out;
}
}
/* Print out the net information only if show_rc is not provided */
if (show_rc == NULL)
cYAML_print_tree(root);
if (l_errno != ENOENT) {
snprintf(err_str,
sizeof(err_str),
"\"cannot get networks: %s\"",
strerror(l_errno));
rc = -l_errno;
goto out;
} else
rc = LUSTRE_CFG_RC_NO_ERR;
snprintf(err_str, sizeof(err_str), "\"success\"");
out:
if (show_rc == NULL || rc != LUSTRE_CFG_RC_NO_ERR || !exist) {
cYAML_free_tree(root);
} else if (show_rc != NULL && *show_rc != NULL) {
struct cYAML *show_node;
/* find the net node, if one doesn't exist
* then insert one. Otherwise add to the one there
*/
show_node = cYAML_get_object_item(*show_rc, "net");
if (show_node != NULL && cYAML_is_sequence(show_node)) {
cYAML_insert_child(show_node, first_seq);
free(net_node);
free(root);
} else if (show_node == NULL) {
cYAML_insert_sibling((*show_rc)->cy_child,
net_node);
free(root);
} else {
cYAML_free_tree(root);
}
} else {
*show_rc = root;
}
cYAML_build_error(rc, seq_no, SHOW_CMD, "net", err_str, err_rc);
return rc;
}
