/* rc=0 means ok
1 means update
<0 means error */
static int mgs_check_target(struct obd_device *obd, struct mgs_target_info *mti)
{
int rc;
ENTRY;
rc = mgs_check_index(obd, mti);
if (rc == 0) {
LCONSOLE_ERROR_MSG(0x13b, "%s claims to have registered, but "
"this MGS does not know about it, preventing "
"registration.\n", mti->mti_svname);
rc = -ENOENT;
} else if (rc == -1) {
LCONSOLE_ERROR_MSG(0x13c, "Client log %s-client has "
"disappeared! Regenerating all logs.\n",
mti->mti_fsname);
mti->mti_flags |= LDD_F_WRITECONF;
rc = 1;
} else {
/* Index is correctly marked as used */
/* If the logs don't contain the mti_nids then add
them as failover nids */
rc = mgs_check_failnid(obd, mti);
}
RETURN(rc);
}
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;
}
}
/** Parse mount line options
* e.g. mount -v -t lustre -o abort_recov uml1:uml2:/lustre-client /mnt/lustre
* dev is passed as device=uml1:/lustre by mount.lustre
*/
static int lmd_parse(char *options, struct lustre_mount_data *lmd)
{
char *s1, *s2, *devname = NULL;
struct lustre_mount_data *raw = (struct lustre_mount_data *)options;
int rc = 0;
LASSERT(lmd);
if (!options) {
LCONSOLE_ERROR_MSG(0x162, "Missing mount data: check that "
"/sbin/mount.lustre is installed.\n");
return -EINVAL;
}
/* Options should be a string - try to detect old lmd data */
if ((raw->lmd_magic & 0xffffff00) == (LMD_MAGIC & 0xffffff00)) {
LCONSOLE_ERROR_MSG(0x163, "You're using an old version of "
"/sbin/mount.lustre. Please install "
"version %s\n", LUSTRE_VERSION_STRING);
return -EINVAL;
}
lmd->lmd_magic = LMD_MAGIC;
OBD_ALLOC(lmd->lmd_params, 4096);
if (lmd->lmd_params == NULL)
return -ENOMEM;
lmd->lmd_params[0] = '\0';
/* Set default flags here */
s1 = options;
while (*s1) {
int clear = 0;
int time_min = OBD_RECOVERY_TIME_MIN;
/* Skip whitespace and extra commas */
while (*s1 == ' ' || *s1 == ',')
s1++;
/* Client options are parsed in ll_options: eg. flock,
user_xattr, acl */
/* Parse non-ldiskfs options here. Rather than modifying
ldiskfs, we just zero these out here */
if (strncmp(s1, "abort_recov", 11) == 0) {
lmd->lmd_flags |= LMD_FLG_ABORT_RECOV;
clear++;
} else if (strncmp(s1, "recovery_time_soft=", 19) == 0) {
lmd->lmd_recovery_time_soft = max_t(int,
simple_strtoul(s1 + 19, NULL, 10), time_min);
clear++;
} else if (strncmp(s1, "recovery_time_hard=", 19) == 0) {
/* Ghastly hack to allow LNET to inter-operate with portals.
* NET type 0 becomes an alias for whatever local network we have, and
* this assignment here means we can parse and print its NIDs */
LASSERT (nf != NULL);
LASSERT (nf0->nf_type < 0);
nf0->nf_name = "zero";//nf->nf_name;
nf0->nf_modname = nf->nf_modname;
nf0->nf_addr2str = nf->nf_addr2str;
nf0->nf_str2addr = nf->nf_str2addr;
mb();
nf0->nf_type = 0;
}
EXPORT_SYMBOL(libcfs_isknown_lnd);
EXPORT_SYMBOL(libcfs_lnd2modname);
EXPORT_SYMBOL(libcfs_lnd2str);
EXPORT_SYMBOL(libcfs_str2lnd);
EXPORT_SYMBOL(libcfs_net2str);
EXPORT_SYMBOL(libcfs_nid2str);
EXPORT_SYMBOL(libcfs_str2net);
EXPORT_SYMBOL(libcfs_str2nid);
EXPORT_SYMBOL(libcfs_id2str);
EXPORT_SYMBOL(libcfs_str2anynid);
EXPORT_SYMBOL(libcfs_setnet0alias);
#else /* __KERNEL__ */
void
libcfs_setnet0alias(__unusedx int lnd)
{
LCONSOLE_ERROR_MSG(0x125, "Liblustre cannot interoperate with old "
"Portals.\nportals_compatibility must be set to "
"'none'.\n");
}
static void
lnet_syntax(char *name, char *str, int offset, int width)
{
static char dots[LNET_SINGLE_TEXTBUF_NOB];
static char dashes[LNET_SINGLE_TEXTBUF_NOB];
memset(dots, '.', sizeof(dots));
dots[sizeof(dots)-1] = 0;
memset(dashes, '-', sizeof(dashes));
dashes[sizeof(dashes)-1] = 0;
LCONSOLE_ERROR_MSG(0x10f, "Error parsing '%s=\"%s\"'\n", name, str);
LCONSOLE_ERROR_MSG(0x110, "here...........%.*s..%.*s|%.*s|\n",
(int)strlen(name), dots, offset, dots,
(width < 1) ? 0 : width - 1, dashes);
}
/** Get a config log from the MGS and process it.
* This func is called for both clients and servers.
* Continue to process new statements appended to the logs
* (whenever the config lock is revoked) until lustre_end_log
* is called.
* @param sb The superblock is used by the MGC to write to the local copy of
* the config log
* @param logname The name of the llog to replicate from the MGS
* @param cfg Since the same mgc may be used to follow multiple config logs
* (e.g. ost1, ost2, client), the config_llog_instance keeps the state for
* this log, and is added to the mgc's list of logs to follow.
*/
int lustre_process_log(struct super_block *sb, char *logname,
struct config_llog_instance *cfg)
{
struct lustre_cfg *lcfg;
struct lustre_cfg_bufs *bufs;
struct lustre_sb_info *lsi = s2lsi(sb);
struct obd_device *mgc = lsi->lsi_mgc;
int rc;
LASSERT(mgc);
LASSERT(cfg);
OBD_ALLOC_PTR(bufs);
if (bufs == NULL)
return -ENOMEM;
/* mgc_process_config */
lustre_cfg_bufs_reset(bufs, mgc->obd_name);
lustre_cfg_bufs_set_string(bufs, 1, logname);
lustre_cfg_bufs_set(bufs, 2, cfg, sizeof(*cfg));
lustre_cfg_bufs_set(bufs, 3, &sb, sizeof(sb));
lcfg = lustre_cfg_new(LCFG_LOG_START, bufs);
rc = obd_process_config(mgc, sizeof(*lcfg), lcfg);
lustre_cfg_free(lcfg);
OBD_FREE_PTR(bufs);
if (rc == -EINVAL)
LCONSOLE_ERROR_MSG(0x15b, "%s: The configuration from log '%s'"
"failed from the MGS (%d). Make sure this "
"client and the MGS are running compatible "
"versions of Lustre.\n",
mgc->obd_name, logname, rc);
if (rc)
LCONSOLE_ERROR_MSG(0x15c, "%s: The configuration from log '%s' "
"failed (%d). This may be the result of "
"communication errors between this node and "
"the MGS, a bad configuration, or other "
"errors. See the syslog for more "
"information.\n", mgc->obd_name, logname,
rc);
/* class_obd_list(); */
return rc;
}
static inline int
accept2secure(const char *acc, long *sec)
{
if (!strcmp(acc, "secure")) {
*sec = 1;
return 1;
} else if (!strcmp(acc, "all")) {
*sec = 0;
return 1;
} else if (!strcmp(acc, "none")) {
return 0;
} else {
LCONSOLE_ERROR_MSG(0x124, "Can't parse 'accept=\"%s\"'\n",
acc);
return -EINVAL;
}
}
/* Collect multiple values for mgsnid specifiers */
static int lmd_parse_mgs(struct lustre_mount_data *lmd, char **ptr)
{
lnet_nid_t nid;
char *tail = *ptr;
char *mgsnid;
int length;
int oldlen = 0;
/* Find end of nidlist */
while (class_parse_nid_quiet(tail, &nid, &tail) == 0)
;
length = tail - *ptr;
if (length == 0) {
LCONSOLE_ERROR_MSG(0x159, "Can't parse NID '%s'\n", *ptr);
return -EINVAL;
}
if (lmd->lmd_mgs != NULL)
oldlen = strlen(lmd->lmd_mgs) + 1;
mgsnid = kzalloc(oldlen + length + 1, GFP_NOFS);
if (!mgsnid)
return -ENOMEM;
if (lmd->lmd_mgs != NULL) {
/* Multiple mgsnid= are taken to mean failover locations */
memcpy(mgsnid, lmd->lmd_mgs, oldlen);
mgsnid[oldlen - 1] = ':';
kfree(lmd->lmd_mgs);
}
memcpy(mgsnid + oldlen, *ptr, length);
mgsnid[oldlen + length] = '\0';
lmd->lmd_mgs = mgsnid;
*ptr = tail;
return 0;
}
/**
* Match client and OST server connection feature flags.
*
* Compute the compatibility flags for a connection request based on
* features mutually supported by client and server.
*
* The obd_export::exp_connect_data.ocd_connect_flags field in \a exp
* must not be updated here, otherwise a partially initialized value may
* be exposed. After the connection request is successfully processed,
* the top-level tgt_connect() request handler atomically updates the export
* connect flags from the obd_connect_data::ocd_connect_flags field of the
* reply. \see tgt_connect().
*
* \param[in] env execution environment
* \param[in] exp the obd_export associated with this
* client/target pair
* \param[in] data stores data for this connect request
* \param[in] new_connection is this connection new or not
*
* \retval 0 if success
* \retval -EPROTO client and server feature requirements are
* incompatible
* \retval -EBADF OST index in connect request doesn't match
* real OST index
*/
static int ofd_parse_connect_data(const struct lu_env *env,
struct obd_export *exp,
struct obd_connect_data *data,
bool new_connection)
{
struct ofd_device *ofd = ofd_exp(exp);
struct filter_export_data *fed = &exp->exp_filter_data;
if (!data)
RETURN(0);
CDEBUG(D_RPCTRACE, "%s: cli %s/%p ocd_connect_flags: "LPX64
" ocd_version: %x ocd_grant: %d ocd_index: %u"
" ocd_group %u\n",
exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
data->ocd_connect_flags, data->ocd_version,
data->ocd_grant, data->ocd_index, data->ocd_group);
if (fed->fed_group != 0 && fed->fed_group != data->ocd_group) {
CWARN("!!! This export (nid %s) used object group %d "
"earlier; now it's trying to use group %d! This could "
"be a bug in the MDS. Please report to "
"https://jira.hpdd.intel.com/\n",
obd_export_nid2str(exp), fed->fed_group,
data->ocd_group);
RETURN(-EPROTO);
}
fed->fed_group = data->ocd_group;
data->ocd_connect_flags &= OST_CONNECT_SUPPORTED;
data->ocd_version = LUSTRE_VERSION_CODE;
/* Kindly make sure the SKIP_ORPHAN flag is from MDS. */
if (data->ocd_connect_flags & OBD_CONNECT_MDS)
CDEBUG(D_HA, "%s: Received MDS connection for group %u\n",
exp->exp_obd->obd_name, data->ocd_group);
else if (data->ocd_connect_flags & OBD_CONNECT_SKIP_ORPHAN)
RETURN(-EPROTO);
if (ofd_grant_param_supp(exp)) {
exp->exp_filter_data.fed_pagesize = data->ocd_blocksize;
/* ocd_{blocksize,inodespace} are log2 values */
data->ocd_blocksize = ofd->ofd_blockbits;
data->ocd_inodespace = ofd->ofd_dt_conf.ddp_inodespace;
/* ocd_grant_extent is in 1K blocks */
data->ocd_grant_extent = ofd->ofd_dt_conf.ddp_grant_frag >> 10;
}
if (data->ocd_connect_flags & OBD_CONNECT_GRANT)
data->ocd_grant = ofd_grant_connect(env, exp, data->ocd_grant,
new_connection);
if (data->ocd_connect_flags & OBD_CONNECT_INDEX) {
struct lr_server_data *lsd = &ofd->ofd_lut.lut_lsd;
int index = lsd->lsd_osd_index;
if (index != data->ocd_index) {
LCONSOLE_ERROR_MSG(0x136, "Connection from %s to index"
" %u doesn't match actual OST index"
" %u in last_rcvd file, bad "
"configuration?\n",
obd_export_nid2str(exp), index,
data->ocd_index);
RETURN(-EBADF);
}
if (!(lsd->lsd_feature_compat & OBD_COMPAT_OST)) {
/* this will only happen on the first connect */
lsd->lsd_feature_compat |= OBD_COMPAT_OST;
/* sync is not needed here as lut_client_add will
* set exp_need_sync flag */
tgt_server_data_update(env, &ofd->ofd_lut, 0);
}
}
if (OBD_FAIL_CHECK(OBD_FAIL_OST_BRW_SIZE)) {
data->ocd_brw_size = 65536;
} else if (data->ocd_connect_flags & OBD_CONNECT_BRW_SIZE) {
data->ocd_brw_size = min(data->ocd_brw_size,
(__u32)DT_MAX_BRW_SIZE);
if (data->ocd_brw_size == 0) {
CERROR("%s: cli %s/%p ocd_connect_flags: "LPX64
" ocd_version: %x ocd_grant: %d ocd_index: %u "
"ocd_brw_size is unexpectedly zero, "
"network data corruption?"
"Refusing connection of this client\n",
exp->exp_obd->obd_name,
exp->exp_client_uuid.uuid,
exp, data->ocd_connect_flags, data->ocd_version,
data->ocd_grant, data->ocd_index);
RETURN(-EPROTO);
}
}
if (data->ocd_connect_flags & OBD_CONNECT_CKSUM) {
__u32 cksum_types = data->ocd_cksum_types;
/* The client set in ocd_cksum_types the checksum types it
* supports. We have to mask off the algorithms that we don't
* support */
data->ocd_cksum_types &= cksum_types_supported_server();
if (unlikely(data->ocd_cksum_types == 0)) {
CERROR("%s: Connect with checksum support but no "
"ocd_cksum_types is set\n",
exp->exp_obd->obd_name);
RETURN(-EPROTO);
}
CDEBUG(D_RPCTRACE, "%s: cli %s supports cksum type %x, return "
"%x\n", exp->exp_obd->obd_name, obd_export_nid2str(exp),
cksum_types, data->ocd_cksum_types);
} else {
/* This client does not support OBD_CONNECT_CKSUM
* fall back to CRC32 */
CDEBUG(D_RPCTRACE, "%s: cli %s does not support "
"OBD_CONNECT_CKSUM, CRC32 will be used\n",
exp->exp_obd->obd_name, obd_export_nid2str(exp));
}
if (data->ocd_connect_flags & OBD_CONNECT_MAXBYTES)
data->ocd_maxbytes = ofd->ofd_dt_conf.ddp_maxbytes;
if (OCD_HAS_FLAG(data, PINGLESS)) {
if (ptlrpc_pinger_suppress_pings()) {
spin_lock(&exp->exp_obd->obd_dev_lock);
list_del_init(&exp->exp_obd_chain_timed);
spin_unlock(&exp->exp_obd->obd_dev_lock);
} else {
data->ocd_connect_flags &= ~OBD_CONNECT_PINGLESS;
}
}
RETURN(0);
}
int
lnet_acceptor(void *arg)
{
socket_t *newsock;
int rc;
__u32 magic;
__u32 peer_ip;
int peer_port;
int secure = (int)((long_ptr_t)arg);
LASSERT(lnet_acceptor_state.pta_sock == NULL);
cfs_block_allsigs();
rc = libcfs_sock_listen(&lnet_acceptor_state.pta_sock,
0, accept_port, accept_backlog);
if (rc != 0) {
if (rc == -EADDRINUSE)
LCONSOLE_ERROR_MSG(0x122, "Can't start acceptor on port %d: port already in use\n",
accept_port);
else
LCONSOLE_ERROR_MSG(0x123, "Can't start acceptor on port %d: unexpected error %d\n",
accept_port, rc);
lnet_acceptor_state.pta_sock = NULL;
} else {
LCONSOLE(0, "Accept %s, port %d\n", accept_type, accept_port);
}
/* set init status and unblock parent */
lnet_acceptor_state.pta_shutdown = rc;
complete(&lnet_acceptor_state.pta_signal);
if (rc != 0)
return rc;
while (!lnet_acceptor_state.pta_shutdown) {
rc = libcfs_sock_accept(&newsock, lnet_acceptor_state.pta_sock);
if (rc != 0) {
if (rc != -EAGAIN) {
CWARN("Accept error %d: pausing...\n", rc);
cfs_pause(cfs_time_seconds(1));
}
continue;
}
/* maybe we're waken up with libcfs_sock_abort_accept() */
if (lnet_acceptor_state.pta_shutdown) {
libcfs_sock_release(newsock);
break;
}
rc = libcfs_sock_getaddr(newsock, 1, &peer_ip, &peer_port);
if (rc != 0) {
CERROR("Can't determine new connection's address\n");
goto failed;
}
if (secure && peer_port > LNET_ACCEPTOR_MAX_RESERVED_PORT) {
CERROR("Refusing connection from %pI4h: insecure port %d\n",
&peer_ip, peer_port);
goto failed;
}
rc = libcfs_sock_read(newsock, &magic, sizeof(magic),
accept_timeout);
if (rc != 0) {
CERROR("Error %d reading connection request from %pI4h\n",
rc, &peer_ip);
goto failed;
}
rc = lnet_accept(newsock, magic);
if (rc != 0)
goto failed;
continue;
failed:
libcfs_sock_release(newsock);
}
libcfs_sock_release(lnet_acceptor_state.pta_sock);
lnet_acceptor_state.pta_sock = NULL;
CDEBUG(D_NET, "Acceptor stopping\n");
/* unblock lnet_acceptor_stop() */
complete(&lnet_acceptor_state.pta_signal);
return 0;
}
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;
}
