From: Karsten Graul kgraul@linux.ibm.com

mainline inclusion from mainline-v5.16-rc1 commit fa0866625543b4d8b3d026e4e0ef5ec25a453920 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

SMC-Dv2 allows users to define EIDs which allows to create separate name spaces enabling users to cluster their SMC-Dv2 connections. Add support for user defined EIDs and extent the generic netlink interface so users can add, remove and dump EIDs.

Signed-off-by: Karsten Graul kgraul@linux.ibm.com Reviewed-by: Guvenc Gulce guvenc@linux.ibm.com Signed-off-by: Guvenc Gulce guvenc@linux.ibm.com Signed-off-by: David S. Miller davem@davemloft.net

Conflicts: net/smc/af_smc.c net/smc/smc.h [The conflict occurs because the commit 194730a9beb5("net/smc: Make SMC statistics network namespace aware") is not merged] Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- include/uapi/linux/smc.h | 15 +++ net/smc/af_smc.c | 34 +++-- net/smc/smc.h | 3 - net/smc/smc_clc.c | 263 +++++++++++++++++++++++++++++++++++++-- net/smc/smc_clc.h | 16 ++- net/smc/smc_core.h | 1 + net/smc/smc_netlink.c | 32 ++++- net/smc/smc_netlink.h | 2 + 8 files changed, 335 insertions(+), 31 deletions(-)

diff --git a/include/uapi/linux/smc.h b/include/uapi/linux/smc.h index 0f7f87c70baf..e3728af2832b 100644 --- a/include/uapi/linux/smc.h +++ b/include/uapi/linux/smc.h @@ -38,6 +38,9 @@ enum { /* SMC PNET Table commands */ #define SMC_GENL_FAMILY_VERSION 1

#define SMC_PCI_ID_STR_LEN 16 /* Max length of pci id string */ +#define SMC_MAX_HOSTNAME_LEN 32 /* Max length of the hostname */ +#define SMC_MAX_UEID 4 /* Max number of user EIDs */ +#define SMC_MAX_EID_LEN 32 /* Max length of an EID */

/* SMC_GENL_FAMILY commands */ enum { @@ -49,6 +52,10 @@ enum { SMC_NETLINK_GET_DEV_SMCR, SMC_NETLINK_GET_STATS, SMC_NETLINK_GET_FBACK_STATS, + SMC_NETLINK_DUMP_UEID, + SMC_NETLINK_ADD_UEID, + SMC_NETLINK_REMOVE_UEID, + SMC_NETLINK_FLUSH_UEID, };

/* SMC_GENL_FAMILY top level attributes */ @@ -242,4 +249,12 @@ enum { __SMC_NLA_FBACK_STATS_MAX, SMC_NLA_FBACK_STATS_MAX = __SMC_NLA_FBACK_STATS_MAX - 1 }; + +/* SMC_NETLINK_UEID attributes */ +enum { + SMC_NLA_EID_TABLE_UNSPEC, + SMC_NLA_EID_TABLE_ENTRY, /* string */ + __SMC_NLA_EID_TABLE_MAX, + SMC_NLA_EID_TABLE_MAX = __SMC_NLA_EID_TABLE_MAX - 1 +}; #endif /* _UAPI_LINUX_SMC_H */ diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index d245562cc30e..705e3e7b8c21 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -1101,7 +1101,7 @@ static int smc_connect_rdma(struct smc_sock *smc, }

reason_code = smc_clc_send_confirm(smc, ini->first_contact_local, - SMC_V1); + SMC_V1, NULL); if (reason_code) goto connect_abort;

@@ -1155,6 +1155,7 @@ static int smc_connect_ism(struct smc_sock *smc, struct smc_clc_msg_accept_confirm *aclc, struct smc_init_info *ini) { + u8 *eid = NULL; int rc = 0;

ini->is_smcd = true; @@ -1190,8 +1191,15 @@ static int smc_connect_ism(struct smc_sock *smc, smc_rx_init(smc); smc_tx_init(smc);

+ if (aclc->hdr.version > SMC_V1) { + struct smc_clc_msg_accept_confirm_v2 *clc_v2 = + (struct smc_clc_msg_accept_confirm_v2 *)aclc; + + eid = clc_v2->eid; + } + rc = smc_clc_send_confirm(smc, ini->first_contact_local, - aclc->hdr.version); + aclc->hdr.version, eid); if (rc) goto connect_abort; mutex_unlock(&smc_server_lgr_pending); @@ -1870,9 +1878,8 @@ static void smc_find_ism_v2_device_serv(struct smc_sock *new_smc, pclc_smcd = smc_get_clc_msg_smcd(pclc); smc_v2_ext = smc_get_clc_v2_ext(pclc); smcd_v2_ext = smc_get_clc_smcd_v2_ext(smc_v2_ext); - if (!smcd_v2_ext || - !smc_v2_ext->hdr.flag.seid) { /* no system EID support for SMCD */ - smc_find_ism_store_rc(SMC_CLC_DECL_NOSEID, ini); + if (!smcd_v2_ext) { + smc_find_ism_store_rc(SMC_CLC_DECL_NOV2DEXT, ini); goto not_found; }

@@ -1892,13 +1899,13 @@ static void smc_find_ism_v2_device_serv(struct smc_sock *new_smc, } mutex_unlock(&smcd_dev_list.mutex);

- if (ini->ism_dev[0]) { - smc_ism_get_system_eid(ini->ism_dev[0], &eid); - if (memcmp(eid, smcd_v2_ext->system_eid, SMC_MAX_EID_LEN)) - goto not_found; - } else { + if (!ini->ism_dev[0]) + goto not_found; + + smc_ism_get_system_eid(ini->ism_dev[0], &eid); + if (!smc_clc_match_eid(ini->negotiated_eid, smc_v2_ext, + smcd_v2_ext->system_eid, eid)) goto not_found; - }

/* separate - outside the smcd_dev_list.lock */ smcd_version = ini->smcd_version; @@ -1916,6 +1923,7 @@ static void smc_find_ism_v2_device_serv(struct smc_sock *new_smc, } /* no V2 ISM device could be initialized */ ini->smcd_version = smcd_version; /* restore original value */ + ini->negotiated_eid[0] = 0;

not_found: ini->smcd_version &= ~SMC_V2; @@ -2133,7 +2141,8 @@ static void smc_listen_work(struct work_struct *work)

/* send SMC Accept CLC message */ rc = smc_clc_send_accept(new_smc, ini->first_contact_local, - ini->smcd_version == SMC_V2 ? SMC_V2 : SMC_V1); + ini->smcd_version == SMC_V2 ? SMC_V2 : SMC_V1, + ini->negotiated_eid); if (rc) goto out_unlock;

@@ -3065,6 +3074,7 @@ static void __exit smc_exit(void) proto_unregister(&smc_proto); smc_pnet_exit(); smc_nl_exit(); + smc_clc_exit(); unregister_pernet_subsys(&smc_net_ops); rcu_barrier(); } diff --git a/net/smc/smc.h b/net/smc/smc.h index 7c50514aaa7e..4fb888366cad 100644 --- a/net/smc/smc.h +++ b/net/smc/smc.h @@ -26,9 +26,6 @@ */ #define SMC_AUTOCORKING_DEFAULT_SIZE 0x10000 /* 64K by default */

-#define SMC_MAX_HOSTNAME_LEN 32 -#define SMC_MAX_EID_LEN 32 - extern struct proto smc_proto; extern struct proto smc_proto6;

diff --git a/net/smc/smc_clc.c b/net/smc/smc_clc.c index 838470f897d0..327f84a4ddc8 100644 --- a/net/smc/smc_clc.c +++ b/net/smc/smc_clc.c @@ -26,6 +26,7 @@ #include "smc_clc.h" #include "smc_ib.h" #include "smc_ism.h" +#include "smc_netlink.h"

#define SMCR_CLC_ACCEPT_CONFIRM_LEN 68 #define SMCD_CLC_ACCEPT_CONFIRM_LEN 48 @@ -39,6 +40,223 @@ static const char SMCD_EYECATCHER[4] = {'\xe2', '\xd4', '\xc3', '\xc4'};

static u8 smc_hostname[SMC_MAX_HOSTNAME_LEN];

+struct smc_clc_eid_table { + rwlock_t lock; + struct list_head list; + u8 ueid_cnt; + u8 seid_enabled; +}; + +static struct smc_clc_eid_table smc_clc_eid_table; + +struct smc_clc_eid_entry { + struct list_head list; + u8 eid[SMC_MAX_EID_LEN]; +}; + +/* The size of a user EID is 32 characters. + * Valid characters should be (single-byte character set) A-Z, 0-9, '.' and '-'. + * Blanks should only be used to pad to the expected size. + * First character must be alphanumeric. + */ +static bool smc_clc_ueid_valid(char *ueid) +{ + char *end = ueid + SMC_MAX_EID_LEN; + + while (--end >= ueid && isspace(*end)) + ; + if (end < ueid) + return false; + if (!isalnum(*ueid) || islower(*ueid)) + return false; + while (ueid <= end) { + if ((!isalnum(*ueid) || islower(*ueid)) && *ueid != '.' && + *ueid != '-') + return false; + ueid++; + } + return true; +} + +static int smc_clc_ueid_add(char *ueid) +{ + struct smc_clc_eid_entry *new_ueid, *tmp_ueid; + int rc; + + if (!smc_clc_ueid_valid(ueid)) + return -EINVAL; + + /* add a new ueid entry to the ueid table if there isn't one */ + new_ueid = kzalloc(sizeof(*new_ueid), GFP_KERNEL); + if (!new_ueid) + return -ENOMEM; + memcpy(new_ueid->eid, ueid, SMC_MAX_EID_LEN); + + write_lock(&smc_clc_eid_table.lock); + if (smc_clc_eid_table.ueid_cnt >= SMC_MAX_UEID) { + rc = -ERANGE; + goto err_out; + } + list_for_each_entry(tmp_ueid, &smc_clc_eid_table.list, list) { + if (!memcmp(tmp_ueid->eid, ueid, SMC_MAX_EID_LEN)) { + rc = -EEXIST; + goto err_out; + } + } + list_add_tail(&new_ueid->list, &smc_clc_eid_table.list); + smc_clc_eid_table.ueid_cnt++; + write_unlock(&smc_clc_eid_table.lock); + return 0; + +err_out: + write_unlock(&smc_clc_eid_table.lock); + kfree(new_ueid); + return rc; +} + +int smc_nl_add_ueid(struct sk_buff *skb, struct genl_info *info) +{ + struct nlattr *nla_ueid = info->attrs[SMC_NLA_EID_TABLE_ENTRY]; + char *ueid; + + if (!nla_ueid || nla_len(nla_ueid) != SMC_MAX_EID_LEN + 1) + return -EINVAL; + ueid = (char *)nla_data(nla_ueid); + + return smc_clc_ueid_add(ueid); +} + +/* remove one or all ueid entries from the table */ +static int smc_clc_ueid_remove(char *ueid) +{ + struct smc_clc_eid_entry *lst_ueid, *tmp_ueid; + int rc = -ENOENT; + + /* remove table entry */ + write_lock(&smc_clc_eid_table.lock); + list_for_each_entry_safe(lst_ueid, tmp_ueid, &smc_clc_eid_table.list, + list) { + if (!ueid || !memcmp(lst_ueid->eid, ueid, SMC_MAX_EID_LEN)) { + list_del(&lst_ueid->list); + smc_clc_eid_table.ueid_cnt--; + kfree(lst_ueid); + rc = 0; + } + } + write_unlock(&smc_clc_eid_table.lock); + return rc; +} + +int smc_nl_remove_ueid(struct sk_buff *skb, struct genl_info *info) +{ + struct nlattr *nla_ueid = info->attrs[SMC_NLA_EID_TABLE_ENTRY]; + char *ueid; + + if (!nla_ueid || nla_len(nla_ueid) != SMC_MAX_EID_LEN + 1) + return -EINVAL; + ueid = (char *)nla_data(nla_ueid); + + return smc_clc_ueid_remove(ueid); +} + +int smc_nl_flush_ueid(struct sk_buff *skb, struct genl_info *info) +{ + smc_clc_ueid_remove(NULL); + return 0; +} + +static int smc_nl_ueid_dumpinfo(struct sk_buff *skb, u32 portid, u32 seq, + u32 flags, char *ueid) +{ + char ueid_str[SMC_MAX_EID_LEN + 1]; + void *hdr; + + hdr = genlmsg_put(skb, portid, seq, &smc_gen_nl_family, + flags, SMC_NETLINK_DUMP_UEID); + if (!hdr) + return -ENOMEM; + snprintf(ueid_str, sizeof(ueid_str), "%s", ueid); + if (nla_put_string(skb, SMC_NLA_EID_TABLE_ENTRY, ueid_str)) { + genlmsg_cancel(skb, hdr); + return -EMSGSIZE; + } + genlmsg_end(skb, hdr); + return 0; +} + +static int _smc_nl_ueid_dump(struct sk_buff *skb, u32 portid, u32 seq, + int start_idx) +{ + struct smc_clc_eid_entry *lst_ueid; + int idx = 0; + + read_lock(&smc_clc_eid_table.lock); + list_for_each_entry(lst_ueid, &smc_clc_eid_table.list, list) { + if (idx++ < start_idx) + continue; + if (smc_nl_ueid_dumpinfo(skb, portid, seq, NLM_F_MULTI, + lst_ueid->eid)) { + --idx; + break; + } + } + read_unlock(&smc_clc_eid_table.lock); + return idx; +} + +int smc_nl_dump_ueid(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct smc_nl_dmp_ctx *cb_ctx = smc_nl_dmp_ctx(cb); + int idx; + + idx = _smc_nl_ueid_dump(skb, NETLINK_CB(cb->skb).portid, + cb->nlh->nlmsg_seq, cb_ctx->pos[0]); + + cb_ctx->pos[0] = idx; + return skb->len; +} + +static bool _smc_clc_match_ueid(u8 *peer_ueid) +{ + struct smc_clc_eid_entry *tmp_ueid; + + list_for_each_entry(tmp_ueid, &smc_clc_eid_table.list, list) { + if (!memcmp(tmp_ueid->eid, peer_ueid, SMC_MAX_EID_LEN)) + return true; + } + return false; +} + +bool smc_clc_match_eid(u8 *negotiated_eid, + struct smc_clc_v2_extension *smc_v2_ext, + u8 *peer_eid, u8 *local_eid) +{ + bool match = false; + int i; + + negotiated_eid[0] = 0; + read_lock(&smc_clc_eid_table.lock); + if (smc_clc_eid_table.seid_enabled && + smc_v2_ext->hdr.flag.seid && + !memcmp(peer_eid, local_eid, SMC_MAX_EID_LEN)) { + memcpy(negotiated_eid, peer_eid, SMC_MAX_EID_LEN); + match = true; + goto out; + } + + for (i = 0; i < smc_v2_ext->hdr.eid_cnt; i++) { + if (_smc_clc_match_ueid(smc_v2_ext->user_eids[i])) { + memcpy(negotiated_eid, smc_v2_ext->user_eids[i], + SMC_MAX_EID_LEN); + match = true; + goto out; + } + } +out: + read_unlock(&smc_clc_eid_table.lock); + return match; +} + /* check arriving CLC proposal */ static bool smc_clc_msg_prop_valid(struct smc_clc_msg_proposal *pclc) { @@ -551,6 +769,7 @@ int smc_clc_send_proposal(struct smc_sock *smc, struct smc_init_info *ini) if (ini->smc_type_v2 == SMC_TYPE_N) { pclc_smcd->v2_ext_offset = 0; } else { + struct smc_clc_eid_entry *ueident; u16 v2_ext_offset; u8 *eid = NULL;

@@ -561,10 +780,19 @@ int smc_clc_send_proposal(struct smc_sock *smc, struct smc_init_info *ini) pclc_prfx->ipv6_prefixes_cnt * sizeof(ipv6_prfx[0]); pclc_smcd->v2_ext_offset = htons(v2_ext_offset); - v2_ext->hdr.eid_cnt = 0; + + read_lock(&smc_clc_eid_table.lock); + v2_ext->hdr.eid_cnt = smc_clc_eid_table.ueid_cnt; + plen += smc_clc_eid_table.ueid_cnt * SMC_MAX_EID_LEN; + i = 0; + list_for_each_entry(ueident, &smc_clc_eid_table.list, list) { + memcpy(v2_ext->user_eids[i++], ueident->eid, + sizeof(ueident->eid)); + } + v2_ext->hdr.flag.seid = smc_clc_eid_table.seid_enabled; + read_unlock(&smc_clc_eid_table.lock); v2_ext->hdr.ism_gid_cnt = ini->ism_offered_cnt; v2_ext->hdr.flag.release = SMC_RELEASE; - v2_ext->hdr.flag.seid = 1; v2_ext->hdr.smcd_v2_ext_offset = htons(sizeof(*v2_ext) - offsetofend(struct smc_clnt_opts_area_hdr, smcd_v2_ext_offset) + @@ -573,7 +801,7 @@ int smc_clc_send_proposal(struct smc_sock *smc, struct smc_init_info *ini) smc_ism_get_system_eid(ini->ism_dev[0], &eid); else smc_ism_get_system_eid(ini->ism_dev[1], &eid); - if (eid) + if (eid && v2_ext->hdr.flag.seid) memcpy(smcd_v2_ext->system_eid, eid, SMC_MAX_EID_LEN); plen += sizeof(*v2_ext) + sizeof(*smcd_v2_ext); if (ini->ism_offered_cnt) { @@ -608,7 +836,8 @@ int smc_clc_send_proposal(struct smc_sock *smc, struct smc_init_info *ini) } if (ini->smc_type_v2 != SMC_TYPE_N) { vec[i].iov_base = v2_ext; - vec[i++].iov_len = sizeof(*v2_ext); + vec[i++].iov_len = sizeof(*v2_ext) + + (v2_ext->hdr.eid_cnt * SMC_MAX_EID_LEN); vec[i].iov_base = smcd_v2_ext; vec[i++].iov_len = sizeof(*smcd_v2_ext); if (ini->ism_offered_cnt) { @@ -636,7 +865,8 @@ int smc_clc_send_proposal(struct smc_sock *smc, struct smc_init_info *ini) /* build and send CLC CONFIRM / ACCEPT message */ static int smc_clc_send_confirm_accept(struct smc_sock *smc, struct smc_clc_msg_accept_confirm_v2 *clc_v2, - int first_contact, u8 version) + int first_contact, u8 version, + u8 *eid) { struct smc_connection *conn = &smc->conn; struct smc_clc_msg_accept_confirm *clc; @@ -664,11 +894,8 @@ static int smc_clc_send_confirm_accept(struct smc_sock *smc, if (version == SMC_V1) { clc->hdr.length = htons(SMCD_CLC_ACCEPT_CONFIRM_LEN); } else { - u8 *eid = NULL; - clc_v2->chid = htons(smc_ism_get_chid(conn->lgr->smcd)); - smc_ism_get_system_eid(conn->lgr->smcd, &eid); - if (eid) + if (eid[0]) memcpy(clc_v2->eid, eid, SMC_MAX_EID_LEN); len = SMCD_CLC_ACCEPT_CONFIRM_LEN_V2; if (first_contact) @@ -735,7 +962,7 @@ static int smc_clc_send_confirm_accept(struct smc_sock *smc,

/* send CLC CONFIRM message across internal TCP socket */ int smc_clc_send_confirm(struct smc_sock *smc, bool clnt_first_contact, - u8 version) + u8 version, u8 *eid) { struct smc_clc_msg_accept_confirm_v2 cclc_v2; int reason_code = 0; @@ -745,7 +972,7 @@ int smc_clc_send_confirm(struct smc_sock *smc, bool clnt_first_contact, memset(&cclc_v2, 0, sizeof(cclc_v2)); cclc_v2.hdr.type = SMC_CLC_CONFIRM; len = smc_clc_send_confirm_accept(smc, &cclc_v2, clnt_first_contact, - version); + version, eid); if (len < ntohs(cclc_v2.hdr.length)) { if (len >= 0) { reason_code = -ENETUNREACH; @@ -760,7 +987,7 @@ int smc_clc_send_confirm(struct smc_sock *smc, bool clnt_first_contact,

/* send CLC ACCEPT message across internal TCP socket */ int smc_clc_send_accept(struct smc_sock *new_smc, bool srv_first_contact, - u8 version) + u8 version, u8 *negotiated_eid) { struct smc_clc_msg_accept_confirm_v2 aclc_v2; int len; @@ -768,7 +995,7 @@ int smc_clc_send_accept(struct smc_sock *new_smc, bool srv_first_contact, memset(&aclc_v2, 0, sizeof(aclc_v2)); aclc_v2.hdr.type = SMC_CLC_ACCEPT; len = smc_clc_send_confirm_accept(new_smc, &aclc_v2, srv_first_contact, - version); + version, negotiated_eid); if (len < ntohs(aclc_v2.hdr.length)) len = len >= 0 ? -EPROTO : -new_smc->clcsock->sk->sk_err;

@@ -788,4 +1015,14 @@ void __init smc_clc_init(void) u = utsname(); memcpy(smc_hostname, u->nodename, min_t(size_t, strlen(u->nodename), sizeof(smc_hostname))); + + INIT_LIST_HEAD(&smc_clc_eid_table.list); + rwlock_init(&smc_clc_eid_table.lock); + smc_clc_eid_table.ueid_cnt = 0; + smc_clc_eid_table.seid_enabled = 1; +} + +void smc_clc_exit(void) +{ + smc_clc_ueid_remove(NULL); } diff --git a/net/smc/smc_clc.h b/net/smc/smc_clc.h index 6f93e27e341b..5981dc37c018 100644 --- a/net/smc/smc_clc.h +++ b/net/smc/smc_clc.h @@ -14,8 +14,10 @@ #define _SMC_CLC_H

#include <rdma/ib_verbs.h> +#include <linux/smc.h>

#include "smc.h" +#include "smc_netlink.h"

#define SMC_CLC_PROPOSAL 0x01 #define SMC_CLC_ACCEPT 0x02 @@ -158,6 +160,7 @@ struct smc_clc_msg_proposal { /* clc proposal message sent by Linux */ } __aligned(4);

#define SMC_CLC_MAX_V6_PREFIX 8 +#define SMC_CLC_MAX_UEID 8

struct smc_clc_msg_proposal_area { struct smc_clc_msg_proposal pclc_base; @@ -165,6 +168,7 @@ struct smc_clc_msg_proposal_area { struct smc_clc_msg_proposal_prefix pclc_prfx; struct smc_clc_ipv6_prefix pclc_prfx_ipv6[SMC_CLC_MAX_V6_PREFIX]; struct smc_clc_v2_extension pclc_v2_ext; + u8 user_eids[SMC_CLC_MAX_UEID][SMC_MAX_EID_LEN]; struct smc_clc_smcd_v2_extension pclc_smcd_v2_ext; struct smc_clc_smcd_gid_chid pclc_gidchids[SMC_MAX_ISM_DEVS]; struct smc_clc_msg_trail pclc_trl; @@ -330,10 +334,18 @@ int smc_clc_wait_msg(struct smc_sock *smc, void *buf, int buflen, int smc_clc_send_decline(struct smc_sock *smc, u32 peer_diag_info, u8 version); int smc_clc_send_proposal(struct smc_sock *smc, struct smc_init_info *ini); int smc_clc_send_confirm(struct smc_sock *smc, bool clnt_first_contact, - u8 version); + u8 version, u8 *eid); int smc_clc_send_accept(struct smc_sock *smc, bool srv_first_contact, - u8 version); + u8 version, u8 *negotiated_eid); void smc_clc_init(void) __init; +void smc_clc_exit(void); void smc_clc_get_hostname(u8 **host); +bool smc_clc_match_eid(u8 *negotiated_eid, + struct smc_clc_v2_extension *smc_v2_ext, + u8 *peer_eid, u8 *local_eid); +int smc_nl_dump_ueid(struct sk_buff *skb, struct netlink_callback *cb); +int smc_nl_add_ueid(struct sk_buff *skb, struct genl_info *info); +int smc_nl_remove_ueid(struct sk_buff *skb, struct genl_info *info); +int smc_nl_flush_ueid(struct sk_buff *skb, struct genl_info *info);

#endif diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 2e4c1a663a9d..6a20c53739ff 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -328,6 +328,7 @@ struct smc_init_info { u8 first_contact_local; unsigned short vlan_id; u32 rc; + u8 negotiated_eid[SMC_MAX_EID_LEN]; /* SMC-R */ struct smc_clc_msg_local *ib_lcl; struct smc_ib_device *ib_dev; diff --git a/net/smc/smc_netlink.c b/net/smc/smc_netlink.c index 6fb6f96c1d17..4548ff2df245 100644 --- a/net/smc/smc_netlink.c +++ b/net/smc/smc_netlink.c @@ -19,11 +19,19 @@ #include "smc_core.h" #include "smc_ism.h" #include "smc_ib.h" +#include "smc_clc.h" #include "smc_stats.h" #include "smc_netlink.h"

-#define SMC_CMD_MAX_ATTR 1 +const struct nla_policy +smc_gen_ueid_policy[SMC_NLA_EID_TABLE_MAX + 1] = { + [SMC_NLA_EID_TABLE_UNSPEC] = { .type = NLA_UNSPEC }, + [SMC_NLA_EID_TABLE_ENTRY] = { .type = NLA_STRING, + .len = SMC_MAX_EID_LEN, + }, +};

+#define SMC_CMD_MAX_ATTR 1 /* SMC_GENL generic netlink operation definition */ static const struct genl_ops smc_gen_nl_ops[] = { { @@ -66,6 +74,28 @@ static const struct genl_ops smc_gen_nl_ops[] = { /* can be retrieved by unprivileged users */ .dumpit = smc_nl_get_fback_stats, }, + { + .cmd = SMC_NETLINK_DUMP_UEID, + /* can be retrieved by unprivileged users */ + .dumpit = smc_nl_dump_ueid, + }, + { + .cmd = SMC_NETLINK_ADD_UEID, + .flags = GENL_ADMIN_PERM, + .doit = smc_nl_add_ueid, + .policy = smc_gen_ueid_policy, + }, + { + .cmd = SMC_NETLINK_REMOVE_UEID, + .flags = GENL_ADMIN_PERM, + .doit = smc_nl_remove_ueid, + .policy = smc_gen_ueid_policy, + }, + { + .cmd = SMC_NETLINK_FLUSH_UEID, + .flags = GENL_ADMIN_PERM, + .doit = smc_nl_flush_ueid, + }, };

static const struct nla_policy smc_gen_nl_policy[2] = { diff --git a/net/smc/smc_netlink.h b/net/smc/smc_netlink.h index 5ce2c0a89ccd..e8c6c3f0e98c 100644 --- a/net/smc/smc_netlink.h +++ b/net/smc/smc_netlink.h @@ -17,6 +17,8 @@

extern struct genl_family smc_gen_nl_family;

+extern const struct nla_policy smc_gen_ueid_policy[]; + struct smc_nl_dmp_ctx { int pos[3]; };

From: Karsten Graul kgraul@linux.ibm.com

mainline inclusion from mainline-v5.16-rc1 commit 3c572145c24e21c24e1cd0fd168011eaba85da8e category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

With SMC-Dv2 users can configure if the static system EID should be used during CLC handshake, or if only user EIDs are allowed. Add generic netlink support to enable and disable the system EID, and to retrieve the system EID and its current enabled state.

Signed-off-by: Karsten Graul kgraul@linux.ibm.com Reviewed-by: Guvenc Gulce guvenc@linux.ibm.com Signed-off-by: Guvenc Gulce guvenc@linux.ibm.com Signed-off-by: David S. Miller davem@davemloft.net --- include/uapi/linux/smc.h | 12 ++++++++ net/smc/smc_clc.c | 62 ++++++++++++++++++++++++++++++++++++++++ net/smc/smc_clc.h | 3 ++ net/smc/smc_netlink.c | 15 ++++++++++ 4 files changed, 92 insertions(+)

diff --git a/include/uapi/linux/smc.h b/include/uapi/linux/smc.h index e3728af2832b..b175bd0165a1 100644 --- a/include/uapi/linux/smc.h +++ b/include/uapi/linux/smc.h @@ -56,6 +56,9 @@ enum { SMC_NETLINK_ADD_UEID, SMC_NETLINK_REMOVE_UEID, SMC_NETLINK_FLUSH_UEID, + SMC_NETLINK_DUMP_SEID, + SMC_NETLINK_ENABLE_SEID, + SMC_NETLINK_DISABLE_SEID, };

/* SMC_GENL_FAMILY top level attributes */ @@ -257,4 +260,13 @@ enum { __SMC_NLA_EID_TABLE_MAX, SMC_NLA_EID_TABLE_MAX = __SMC_NLA_EID_TABLE_MAX - 1 }; + +/* SMC_NETLINK_SEID attributes */ +enum { + SMC_NLA_SEID_UNSPEC, + SMC_NLA_SEID_ENTRY, /* string */ + SMC_NLA_SEID_ENABLED, /* u8 */ + __SMC_NLA_SEID_TABLE_MAX, + SMC_NLA_SEID_TABLE_MAX = __SMC_NLA_SEID_TABLE_MAX - 1 +}; #endif /* _UAPI_LINUX_SMC_H */ diff --git a/net/smc/smc_clc.c b/net/smc/smc_clc.c index 3753000dc246..632100c37aa9 100644 --- a/net/smc/smc_clc.c +++ b/net/smc/smc_clc.c @@ -143,6 +143,10 @@ static int smc_clc_ueid_remove(char *ueid) rc = 0; } } + if (!rc && !smc_clc_eid_table.ueid_cnt) { + smc_clc_eid_table.seid_enabled = 1; + rc = -EAGAIN; /* indicate success and enabling of seid */ + } write_unlock(&smc_clc_eid_table.lock); return rc; } @@ -216,6 +220,64 @@ int smc_nl_dump_ueid(struct sk_buff *skb, struct netlink_callback *cb) return skb->len; }

+int smc_nl_dump_seid(struct sk_buff *skb, struct netlink_callback *cb) +{ + struct smc_nl_dmp_ctx *cb_ctx = smc_nl_dmp_ctx(cb); + char seid_str[SMC_MAX_EID_LEN + 1]; + u8 seid_enabled; + void *hdr; + u8 *seid; + + if (cb_ctx->pos[0]) + return skb->len; + + hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq, + &smc_gen_nl_family, NLM_F_MULTI, + SMC_NETLINK_DUMP_SEID); + if (!hdr) + return -ENOMEM; + if (!smc_ism_is_v2_capable()) + goto end; + + smc_ism_get_system_eid(&seid); + snprintf(seid_str, sizeof(seid_str), "%s", seid); + if (nla_put_string(skb, SMC_NLA_SEID_ENTRY, seid_str)) + goto err; + read_lock(&smc_clc_eid_table.lock); + seid_enabled = smc_clc_eid_table.seid_enabled; + read_unlock(&smc_clc_eid_table.lock); + if (nla_put_u8(skb, SMC_NLA_SEID_ENABLED, seid_enabled)) + goto err; +end: + genlmsg_end(skb, hdr); + cb_ctx->pos[0]++; + return skb->len; +err: + genlmsg_cancel(skb, hdr); + return -EMSGSIZE; +} + +int smc_nl_enable_seid(struct sk_buff *skb, struct genl_info *info) +{ + write_lock(&smc_clc_eid_table.lock); + smc_clc_eid_table.seid_enabled = 1; + write_unlock(&smc_clc_eid_table.lock); + return 0; +} + +int smc_nl_disable_seid(struct sk_buff *skb, struct genl_info *info) +{ + int rc = 0; + + write_lock(&smc_clc_eid_table.lock); + if (!smc_clc_eid_table.ueid_cnt) + rc = -ENOENT; + else + smc_clc_eid_table.seid_enabled = 0; + write_unlock(&smc_clc_eid_table.lock); + return rc; +} + static bool _smc_clc_match_ueid(u8 *peer_ueid) { struct smc_clc_eid_entry *tmp_ueid; diff --git a/net/smc/smc_clc.h b/net/smc/smc_clc.h index 5981dc37c018..af42101df0f7 100644 --- a/net/smc/smc_clc.h +++ b/net/smc/smc_clc.h @@ -347,5 +347,8 @@ int smc_nl_dump_ueid(struct sk_buff *skb, struct netlink_callback *cb); int smc_nl_add_ueid(struct sk_buff *skb, struct genl_info *info); int smc_nl_remove_ueid(struct sk_buff *skb, struct genl_info *info); int smc_nl_flush_ueid(struct sk_buff *skb, struct genl_info *info); +int smc_nl_dump_seid(struct sk_buff *skb, struct netlink_callback *cb); +int smc_nl_enable_seid(struct sk_buff *skb, struct genl_info *info); +int smc_nl_disable_seid(struct sk_buff *skb, struct genl_info *info);

#endif diff --git a/net/smc/smc_netlink.c b/net/smc/smc_netlink.c index 4548ff2df245..f13ab0661ed5 100644 --- a/net/smc/smc_netlink.c +++ b/net/smc/smc_netlink.c @@ -96,6 +96,21 @@ static const struct genl_ops smc_gen_nl_ops[] = { .flags = GENL_ADMIN_PERM, .doit = smc_nl_flush_ueid, }, + { + .cmd = SMC_NETLINK_DUMP_SEID, + /* can be retrieved by unprivileged users */ + .dumpit = smc_nl_dump_seid, + }, + { + .cmd = SMC_NETLINK_ENABLE_SEID, + .flags = GENL_ADMIN_PERM, + .doit = smc_nl_enable_seid, + }, + { + .cmd = SMC_NETLINK_DISABLE_SEID, + .flags = GENL_ADMIN_PERM, + .doit = smc_nl_disable_seid, + }, };

static const struct nla_policy smc_gen_nl_policy[2] = {

From: Karsten Graul kgraul@linux.ibm.com

mainline inclusion from mainline-v5.16-rc1 commit 42042dbbc2ebb926e594b22490374d9343c746ef category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

Prepare the connection establishment with SMC-Rv2. Detect eligible RoCE cards and indicate all supported SMC modes for the connection.

Signed-off-by: Karsten Graul kgraul@linux.ibm.com Signed-off-by: David S. Miller davem@davemloft.net --- net/smc/af_smc.c | 97 +++++++++++++++++++++++++++++----------------- net/smc/smc_clc.c | 11 ++++++ net/smc/smc_clc.h | 12 ++++++ net/smc/smc_core.h | 28 +++++++++++++ 4 files changed, 113 insertions(+), 35 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index 1280e9f5500c..3ca0b8447464 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -874,7 +874,9 @@ static int smc_find_rdma_device(struct smc_sock *smc, struct smc_init_info *ini) * used for the internal TCP socket */ smc_pnet_find_roce_resource(smc->clcsock->sk, ini); - if (!ini->ib_dev) + if (!ini->check_smcrv2 && !ini->ib_dev) + return SMC_CLC_DECL_NOSMCRDEV; + if (ini->check_smcrv2 && !ini->smcrv2.ib_dev_v2) return SMC_CLC_DECL_NOSMCRDEV; return 0; } @@ -958,27 +960,42 @@ static int smc_find_proposal_devices(struct smc_sock *smc, int rc = 0;

/* check if there is an ism device available */ - if (ini->smcd_version & SMC_V1) { - if (smc_find_ism_device(smc, ini) || - smc_connect_ism_vlan_setup(smc, ini)) { - if (ini->smc_type_v1 == SMC_TYPE_B) - ini->smc_type_v1 = SMC_TYPE_R; - else - ini->smc_type_v1 = SMC_TYPE_N; - } /* else ISM V1 is supported for this connection */ - if (smc_find_rdma_device(smc, ini)) { - if (ini->smc_type_v1 == SMC_TYPE_B) - ini->smc_type_v1 = SMC_TYPE_D; - else - ini->smc_type_v1 = SMC_TYPE_N; - } /* else RDMA is supported for this connection */ - } - if (smc_ism_is_v2_capable() && smc_find_ism_v2_device_clnt(smc, ini)) - ini->smc_type_v2 = SMC_TYPE_N; + if (!(ini->smcd_version & SMC_V1) || + smc_find_ism_device(smc, ini) || + smc_connect_ism_vlan_setup(smc, ini)) + ini->smcd_version &= ~SMC_V1; + /* else ISM V1 is supported for this connection */ + + /* check if there is an rdma device available */ + if (!(ini->smcr_version & SMC_V1) || + smc_find_rdma_device(smc, ini)) + ini->smcr_version &= ~SMC_V1; + /* else RDMA is supported for this connection */ + + ini->smc_type_v1 = smc_indicated_type(ini->smcd_version & SMC_V1, + ini->smcr_version & SMC_V1); + + /* check if there is an ism v2 device available */ + if (!(ini->smcd_version & SMC_V2) || + !smc_ism_is_v2_capable() || + smc_find_ism_v2_device_clnt(smc, ini)) + ini->smcd_version &= ~SMC_V2; + + /* check if there is an rdma v2 device available */ + ini->check_smcrv2 = true; + ini->smcrv2.saddr = smc->clcsock->sk->sk_rcv_saddr; + if (!(ini->smcr_version & SMC_V2) || + smc->clcsock->sk->sk_family != AF_INET || + !smc_clc_ueid_count() || + smc_find_rdma_device(smc, ini)) + ini->smcr_version &= ~SMC_V2; + ini->check_smcrv2 = false; + + ini->smc_type_v2 = smc_indicated_type(ini->smcd_version & SMC_V2, + ini->smcr_version & SMC_V2);

/* if neither ISM nor RDMA are supported, fallback */ - if (!smcr_indicated(ini->smc_type_v1) && - ini->smc_type_v1 == SMC_TYPE_N && ini->smc_type_v2 == SMC_TYPE_N) + if (ini->smc_type_v1 == SMC_TYPE_N && ini->smc_type_v2 == SMC_TYPE_N) rc = SMC_CLC_DECL_NOSMCDEV;

return rc; @@ -1222,17 +1239,24 @@ static int smc_connect_ism(struct smc_sock *smc, static int smc_connect_check_aclc(struct smc_init_info *ini, struct smc_clc_msg_accept_confirm *aclc) { - if ((aclc->hdr.typev1 == SMC_TYPE_R && - !smcr_indicated(ini->smc_type_v1)) || - (aclc->hdr.typev1 == SMC_TYPE_D && - ((!smcd_indicated(ini->smc_type_v1) && - !smcd_indicated(ini->smc_type_v2)) || - (aclc->hdr.version == SMC_V1 && - !smcd_indicated(ini->smc_type_v1)) || - (aclc->hdr.version == SMC_V2 && - !smcd_indicated(ini->smc_type_v2))))) + if (aclc->hdr.typev1 != SMC_TYPE_R && + aclc->hdr.typev1 != SMC_TYPE_D) return SMC_CLC_DECL_MODEUNSUPP;

+ if (aclc->hdr.version >= SMC_V2) { + if ((aclc->hdr.typev1 == SMC_TYPE_R && + !smcr_indicated(ini->smc_type_v2)) || + (aclc->hdr.typev1 == SMC_TYPE_D && + !smcd_indicated(ini->smc_type_v2))) + return SMC_CLC_DECL_MODEUNSUPP; + } else { + if ((aclc->hdr.typev1 == SMC_TYPE_R && + !smcr_indicated(ini->smc_type_v1)) || + (aclc->hdr.typev1 == SMC_TYPE_D && + !smcd_indicated(ini->smc_type_v1))) + return SMC_CLC_DECL_MODEUNSUPP; + } + return 0; }

@@ -1263,14 +1287,15 @@ static int __smc_connect(struct smc_sock *smc) return smc_connect_decline_fallback(smc, SMC_CLC_DECL_MEM, version);

- ini->smcd_version = SMC_V1; - ini->smcd_version |= smc_ism_is_v2_capable() ? SMC_V2 : 0; + ini->smcd_version = SMC_V1 | SMC_V2; + ini->smcr_version = SMC_V1 | SMC_V2; ini->smc_type_v1 = SMC_TYPE_B; - ini->smc_type_v2 = smc_ism_is_v2_capable() ? SMC_TYPE_D : SMC_TYPE_N; + ini->smc_type_v2 = SMC_TYPE_B;

/* get vlan id from IP device */ if (smc_vlan_by_tcpsk(smc->clcsock, ini)) { ini->smcd_version &= ~SMC_V1; + ini->smcr_version = 0; ini->smc_type_v1 = SMC_TYPE_N; if (!ini->smcd_version) { rc = SMC_CLC_DECL_GETVLANERR; @@ -1298,15 +1323,17 @@ static int __smc_connect(struct smc_sock *smc) /* check if smc modes and versions of CLC proposal and accept match */ rc = smc_connect_check_aclc(ini, aclc); version = aclc->hdr.version == SMC_V1 ? SMC_V1 : SMC_V2; - ini->smcd_version = version; if (rc) goto vlan_cleanup;

/* depending on previous steps, connect using rdma or ism */ - if (aclc->hdr.typev1 == SMC_TYPE_R) + if (aclc->hdr.typev1 == SMC_TYPE_R) { + ini->smcr_version = version; rc = smc_connect_rdma(smc, aclc, ini); - else if (aclc->hdr.typev1 == SMC_TYPE_D) + } else if (aclc->hdr.typev1 == SMC_TYPE_D) { + ini->smcd_version = version; rc = smc_connect_ism(smc, aclc, ini); + } if (rc) goto vlan_cleanup;

diff --git a/net/smc/smc_clc.c b/net/smc/smc_clc.c index 632100c37aa9..a73cf97856a2 100644 --- a/net/smc/smc_clc.c +++ b/net/smc/smc_clc.c @@ -114,6 +114,17 @@ static int smc_clc_ueid_add(char *ueid) return rc; }

+int smc_clc_ueid_count(void) +{ + int count; + + read_lock(&smc_clc_eid_table.lock); + count = smc_clc_eid_table.ueid_cnt; + read_unlock(&smc_clc_eid_table.lock); + + return count; +} + int smc_nl_add_ueid(struct sk_buff *skb, struct genl_info *info) { struct nlattr *nla_ueid = info->attrs[SMC_NLA_EID_TABLE_ENTRY]; diff --git a/net/smc/smc_clc.h b/net/smc/smc_clc.h index af42101df0f7..765aa549ab0d 100644 --- a/net/smc/smc_clc.h +++ b/net/smc/smc_clc.h @@ -282,6 +282,17 @@ static inline bool smcd_indicated(int smc_type) return smc_type == SMC_TYPE_D || smc_type == SMC_TYPE_B; }

+static inline u8 smc_indicated_type(int is_smcd, int is_smcr) +{ + if (is_smcd && is_smcr) + return SMC_TYPE_B; + if (is_smcd) + return SMC_TYPE_D; + if (is_smcr) + return SMC_TYPE_R; + return SMC_TYPE_N; +} + /* get SMC-D info from proposal message */ static inline struct smc_clc_msg_smcd * smc_get_clc_msg_smcd(struct smc_clc_msg_proposal *prop) @@ -343,6 +354,7 @@ void smc_clc_get_hostname(u8 **host); bool smc_clc_match_eid(u8 *negotiated_eid, struct smc_clc_v2_extension *smc_v2_ext, u8 *peer_eid, u8 *local_eid); +int smc_clc_ueid_count(void); int smc_nl_dump_ueid(struct sk_buff *skb, struct netlink_callback *cb); int smc_nl_add_ueid(struct sk_buff *skb, struct genl_info *info); int smc_nl_remove_ueid(struct sk_buff *skb, struct genl_info *info); diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 6a20c53739ff..4b7f2c3bf9cb 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -320,6 +320,31 @@ struct smc_link_group {

struct smc_clc_msg_local;

+#define GID_LIST_SIZE 2 + +struct smc_gidlist { + u8 len; + u8 list[GID_LIST_SIZE][SMC_GID_SIZE]; +}; + +struct smc_init_info_smcrv2 { + /* Input fields */ + __be32 saddr; + struct sock *clc_sk; + __be32 daddr; + + /* Output fields when saddr is set */ + struct smc_ib_device *ib_dev_v2; + u8 ib_port_v2; + u8 ib_gid_v2[SMC_GID_SIZE]; + + /* Additional output fields when clc_sk and daddr is set as well */ + u8 uses_gateway; + u8 nexthop_mac[ETH_ALEN]; + + struct smc_gidlist gidlist; +}; + struct smc_init_info { u8 is_smcd; u8 smc_type_v1; @@ -331,10 +356,13 @@ struct smc_init_info { u8 negotiated_eid[SMC_MAX_EID_LEN]; /* SMC-R */ struct smc_clc_msg_local *ib_lcl; + u8 smcr_version; + u8 check_smcrv2; struct smc_ib_device *ib_dev; u8 ib_gid[SMC_GID_SIZE]; u8 ib_port; u32 ib_clcqpn; + struct smc_init_info_smcrv2 smcrv2; /* SMC-D */ u64 ism_peer_gid[SMC_MAX_ISM_DEVS + 1]; struct smcd_dev *ism_dev[SMC_MAX_ISM_DEVS + 1];

From: Karsten Graul kgraul@linux.ibm.com

mainline inclusion from mainline-v5.16-rc1 commit e49300a6bf6218c835403545e9356141a6340181 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

Implement the server side of the SMC-Rv2 processing. Process incoming CLC messages, find eligible devices and check for a valid route to the remote peer.

Signed-off-by: Karsten Graul kgraul@linux.ibm.com Signed-off-by: David S. Miller davem@davemloft.net

Conflicts: net/smc/smc_core.c [The conflict occurs because the commit ddc992866f13("net/smc: Add link counters for IB device ports") is not merged] Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- net/smc/af_smc.c | 165 ++++++++++++++++++++++++++++++++------------- net/smc/smc_clc.h | 1 + net/smc/smc_core.c | 61 ++++++++++++----- net/smc/smc_core.h | 1 - 4 files changed, 164 insertions(+), 64 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index e31509819a97..c1fc8d0a27c1 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -1849,33 +1849,48 @@ static int smc_listen_v2_check(struct smc_sock *new_smc,

ini->smc_type_v1 = pclc->hdr.typev1; ini->smc_type_v2 = pclc->hdr.typev2; - ini->smcd_version = ini->smc_type_v1 != SMC_TYPE_N ? SMC_V1 : 0; - if (pclc->hdr.version > SMC_V1) - ini->smcd_version |= - ini->smc_type_v2 != SMC_TYPE_N ? SMC_V2 : 0; - if (!(ini->smcd_version & SMC_V2)) { + ini->smcd_version = smcd_indicated(ini->smc_type_v1) ? SMC_V1 : 0; + ini->smcr_version = smcr_indicated(ini->smc_type_v1) ? SMC_V1 : 0; + if (pclc->hdr.version > SMC_V1) { + if (smcd_indicated(ini->smc_type_v2)) + ini->smcd_version |= SMC_V2; + if (smcr_indicated(ini->smc_type_v2)) + ini->smcr_version |= SMC_V2; + } + if (!(ini->smcd_version & SMC_V2) && !(ini->smcr_version & SMC_V2)) { rc = SMC_CLC_DECL_PEERNOSMC; goto out; } - if (!smc_ism_is_v2_capable()) { - ini->smcd_version &= ~SMC_V2; - rc = SMC_CLC_DECL_NOISM2SUPP; - goto out; - } pclc_v2_ext = smc_get_clc_v2_ext(pclc); if (!pclc_v2_ext) { ini->smcd_version &= ~SMC_V2; + ini->smcr_version &= ~SMC_V2; rc = SMC_CLC_DECL_NOV2EXT; goto out; } pclc_smcd_v2_ext = smc_get_clc_smcd_v2_ext(pclc_v2_ext); - if (!pclc_smcd_v2_ext) { - ini->smcd_version &= ~SMC_V2; - rc = SMC_CLC_DECL_NOV2DEXT; + if (ini->smcd_version & SMC_V2) { + if (!smc_ism_is_v2_capable()) { + ini->smcd_version &= ~SMC_V2; + rc = SMC_CLC_DECL_NOISM2SUPP; + } else if (!pclc_smcd_v2_ext) { + ini->smcd_version &= ~SMC_V2; + rc = SMC_CLC_DECL_NOV2DEXT; + } else if (!pclc_v2_ext->hdr.eid_cnt && + !pclc_v2_ext->hdr.flag.seid) { + ini->smcd_version &= ~SMC_V2; + rc = SMC_CLC_DECL_NOUEID; + } + } + if (ini->smcr_version & SMC_V2) { + if (!pclc_v2_ext->hdr.eid_cnt) { + ini->smcr_version &= ~SMC_V2; + rc = SMC_CLC_DECL_NOUEID; + } }

out: - if (!ini->smcd_version) + if (!ini->smcd_version && !ini->smcr_version) return rc;

return 0; @@ -1998,10 +2013,6 @@ static void smc_find_ism_v2_device_serv(struct smc_sock *new_smc, pclc_smcd = smc_get_clc_msg_smcd(pclc); smc_v2_ext = smc_get_clc_v2_ext(pclc); smcd_v2_ext = smc_get_clc_smcd_v2_ext(smc_v2_ext); - if (!smcd_v2_ext) { - smc_find_ism_store_rc(SMC_CLC_DECL_NOV2DEXT, ini); - goto not_found; - }

mutex_lock(&smcd_dev_list.mutex); if (pclc_smcd->ism.chid) @@ -2019,8 +2030,10 @@ static void smc_find_ism_v2_device_serv(struct smc_sock *new_smc, } mutex_unlock(&smcd_dev_list.mutex);

- if (!ini->ism_dev[0]) + if (!ini->ism_dev[0]) { + smc_find_ism_store_rc(SMC_CLC_DECL_NOSMCD2DEV, ini); goto not_found; + }

smc_ism_get_system_eid(&eid); if (!smc_clc_match_eid(ini->negotiated_eid, smc_v2_ext, @@ -2073,6 +2086,7 @@ static void smc_find_ism_v1_device_serv(struct smc_sock *new_smc,

not_found: smc_find_ism_store_rc(rc, ini); + ini->smcd_version &= ~SMC_V1; ini->ism_dev[0] = NULL; ini->is_smcd = false; } @@ -2096,24 +2110,69 @@ static int smc_listen_rdma_reg(struct smc_sock *new_smc, bool local_first) return 0; }

+static void smc_find_rdma_v2_device_serv(struct smc_sock *new_smc, + struct smc_clc_msg_proposal *pclc, + struct smc_init_info *ini) +{ + struct smc_clc_v2_extension *smc_v2_ext; + u8 smcr_version; + int rc; + + if (!(ini->smcr_version & SMC_V2) || !smcr_indicated(ini->smc_type_v2)) + goto not_found; + + smc_v2_ext = smc_get_clc_v2_ext(pclc); + if (!smc_clc_match_eid(ini->negotiated_eid, smc_v2_ext, NULL, NULL)) + goto not_found; + + /* prepare RDMA check */ + memcpy(ini->peer_systemid, pclc->lcl.id_for_peer, SMC_SYSTEMID_LEN); + memcpy(ini->peer_gid, smc_v2_ext->roce, SMC_GID_SIZE); + memcpy(ini->peer_mac, pclc->lcl.mac, ETH_ALEN); + ini->check_smcrv2 = true; + ini->smcrv2.clc_sk = new_smc->clcsock->sk; + ini->smcrv2.saddr = new_smc->clcsock->sk->sk_rcv_saddr; + ini->smcrv2.daddr = smc_ib_gid_to_ipv4(smc_v2_ext->roce); + rc = smc_find_rdma_device(new_smc, ini); + if (rc) { + smc_find_ism_store_rc(rc, ini); + goto not_found; + } + if (!ini->smcrv2.uses_gateway) + memcpy(ini->smcrv2.nexthop_mac, pclc->lcl.mac, ETH_ALEN); + + smcr_version = ini->smcr_version; + ini->smcr_version = SMC_V2; + rc = smc_listen_rdma_init(new_smc, ini); + if (!rc) + rc = smc_listen_rdma_reg(new_smc, ini->first_contact_local); + if (!rc) + return; + ini->smcr_version = smcr_version; + smc_find_ism_store_rc(rc, ini); + +not_found: + ini->smcr_version &= ~SMC_V2; + ini->check_smcrv2 = false; +} + static int smc_find_rdma_v1_device_serv(struct smc_sock *new_smc, struct smc_clc_msg_proposal *pclc, struct smc_init_info *ini) { int rc;

- if (!smcr_indicated(ini->smc_type_v1)) + if (!(ini->smcr_version & SMC_V1) || !smcr_indicated(ini->smc_type_v1)) return SMC_CLC_DECL_NOSMCDEV;

/* prepare RDMA check */ - ini->ib_lcl = &pclc->lcl; + memcpy(ini->peer_systemid, pclc->lcl.id_for_peer, SMC_SYSTEMID_LEN); + memcpy(ini->peer_gid, pclc->lcl.gid, SMC_GID_SIZE); + memcpy(ini->peer_mac, pclc->lcl.mac, ETH_ALEN); rc = smc_find_rdma_device(new_smc, ini); if (rc) { /* no RDMA device found */ - if (ini->smc_type_v1 == SMC_TYPE_B) - /* neither ISM nor RDMA device found */ - rc = SMC_CLC_DECL_NOSMCDEV; - return rc; + return SMC_CLC_DECL_NOSMCDEV; } rc = smc_listen_rdma_init(new_smc, ini); if (rc) @@ -2126,51 +2185,60 @@ static int smc_listen_find_device(struct smc_sock *new_smc, struct smc_clc_msg_proposal *pclc, struct smc_init_info *ini) { - int rc; + int prfx_rc;

/* check for ISM device matching V2 proposed device */ smc_find_ism_v2_device_serv(new_smc, pclc, ini); if (ini->ism_dev[0]) return 0;

- if (!(ini->smcd_version & SMC_V1)) - return ini->rc ?: SMC_CLC_DECL_NOSMCD2DEV; - - /* check for matching IP prefix and subnet length */ - rc = smc_listen_prfx_check(new_smc, pclc); - if (rc) - return ini->rc ?: rc; + /* check for matching IP prefix and subnet length (V1) */ + prfx_rc = smc_listen_prfx_check(new_smc, pclc); + if (prfx_rc) + smc_find_ism_store_rc(prfx_rc, ini);

/* get vlan id from IP device */ if (smc_vlan_by_tcpsk(new_smc->clcsock, ini)) return ini->rc ?: SMC_CLC_DECL_GETVLANERR;

/* check for ISM device matching V1 proposed device */ - smc_find_ism_v1_device_serv(new_smc, pclc, ini); + if (!prfx_rc) + smc_find_ism_v1_device_serv(new_smc, pclc, ini); if (ini->ism_dev[0]) return 0;

- if (pclc->hdr.typev1 == SMC_TYPE_D) + if (!smcr_indicated(pclc->hdr.typev1) && + !smcr_indicated(pclc->hdr.typev2)) /* skip RDMA and decline */ return ini->rc ?: SMC_CLC_DECL_NOSMCDDEV;

- /* check if RDMA is available */ - rc = smc_find_rdma_v1_device_serv(new_smc, pclc, ini); - smc_find_ism_store_rc(rc, ini); + /* check if RDMA V2 is available */ + smc_find_rdma_v2_device_serv(new_smc, pclc, ini); + if (ini->smcrv2.ib_dev_v2) + return 0;

- return (!rc) ? 0 : ini->rc; + /* check if RDMA V1 is available */ + if (!prfx_rc) { + int rc; + + rc = smc_find_rdma_v1_device_serv(new_smc, pclc, ini); + smc_find_ism_store_rc(rc, ini); + return (!rc) ? 0 : ini->rc; + } + return SMC_CLC_DECL_NOSMCDEV; }

/* listen worker: finish RDMA setup */ static int smc_listen_rdma_finish(struct smc_sock *new_smc, struct smc_clc_msg_accept_confirm *cclc, - bool local_first) + bool local_first, + struct smc_init_info *ini) { struct smc_link *link = new_smc->conn.lnk; int reason_code = 0;

if (local_first) - smc_link_save_peer_info(link, cclc, NULL); + smc_link_save_peer_info(link, cclc, ini);

if (smc_rmb_rtoken_handling(&new_smc->conn, link, cclc)) return SMC_CLC_DECL_ERR_RTOK; @@ -2191,12 +2259,13 @@ static void smc_listen_work(struct work_struct *work) { struct smc_sock *new_smc = container_of(work, struct smc_sock, smc_listen_work); - u8 version = smc_ism_is_v2_capable() ? SMC_V2 : SMC_V1; struct socket *newclcsock = new_smc->clcsock; struct smc_clc_msg_accept_confirm *cclc; struct smc_clc_msg_proposal_area *buf; struct smc_clc_msg_proposal *pclc; struct smc_init_info *ini = NULL; + u8 proposal_version = SMC_V1; + u8 accept_version; int rc = 0;

if (new_smc->listen_smc->sk.sk_state != SMC_LISTEN) @@ -2230,7 +2299,9 @@ static void smc_listen_work(struct work_struct *work) SMC_CLC_PROPOSAL, CLC_WAIT_TIME); if (rc) goto out_decl; - version = pclc->hdr.version == SMC_V1 ? SMC_V1 : version; + + if (pclc->hdr.version > SMC_V1) + proposal_version = SMC_V2;

/* IPSec connections opt out of SMC optimizations */ if (using_ipsec(new_smc)) { @@ -2260,9 +2331,9 @@ static void smc_listen_work(struct work_struct *work) goto out_unlock;

/* send SMC Accept CLC message */ + accept_version = ini->is_smcd ? ini->smcd_version : ini->smcr_version; rc = smc_clc_send_accept(new_smc, ini->first_contact_local, - ini->smcd_version == SMC_V2 ? SMC_V2 : SMC_V1, - ini->negotiated_eid); + accept_version, ini->negotiated_eid); if (rc) goto out_unlock;

@@ -2284,7 +2355,7 @@ static void smc_listen_work(struct work_struct *work) /* finish worker */ if (!ini->is_smcd) { rc = smc_listen_rdma_finish(new_smc, cclc, - ini->first_contact_local); + ini->first_contact_local, ini); if (rc) goto out_unlock; mutex_unlock(&smc_server_lgr_pending); @@ -2298,7 +2369,7 @@ static void smc_listen_work(struct work_struct *work) mutex_unlock(&smc_server_lgr_pending); out_decl: smc_listen_decline(new_smc, rc, ini ? ini->first_contact_local : 0, - version); + proposal_version); out_free: kfree(ini); kfree(buf); diff --git a/net/smc/smc_clc.h b/net/smc/smc_clc.h index 998eba0c924f..036e696cf1db 100644 --- a/net/smc/smc_clc.h +++ b/net/smc/smc_clc.h @@ -44,6 +44,7 @@ #define SMC_CLC_DECL_NOV2DEXT 0x03030005 /* peer sent no clc SMC-Dv2 ext. */ #define SMC_CLC_DECL_NOSEID 0x03030006 /* peer sent no SEID */ #define SMC_CLC_DECL_NOSMCD2DEV 0x03030007 /* no SMC-Dv2 device found */ +#define SMC_CLC_DECL_NOUEID 0x03030008 /* peer sent no UEID */ #define SMC_CLC_DECL_MODEUNSUPP 0x03040000 /* smc modes do not match (R or D)*/ #define SMC_CLC_DECL_RMBE_EC 0x03050000 /* peer has eyecatcher in RMBE */ #define SMC_CLC_DECL_OPTUNSUPP 0x03060000 /* fastopen sockopt not supported */ diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 3509387c76d3..0814aa713845 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -686,11 +686,21 @@ static void smcr_copy_dev_info_to_link(struct smc_link *link) int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk, u8 link_idx, struct smc_init_info *ini) { + struct smc_ib_device *smcibdev; u8 rndvec[3]; int rc;

+ if (lgr->smc_version == SMC_V2) { + lnk->smcibdev = ini->smcrv2.ib_dev_v2; + lnk->ibport = ini->smcrv2.ib_port_v2; + } else { + lnk->smcibdev = ini->ib_dev; + lnk->ibport = ini->ib_port; + } + get_device(&ini->ib_dev->ibdev->dev); atomic_inc(&ini->ib_dev->lnk_cnt); + lnk->path_mtu = lnk->smcibdev->pattr[lnk->ibport - 1].active_mtu; refcount_set(&lnk->refcnt, 1); /* link refcnt is set to 1 */ lnk->clearing = 0; lnk->link_id = smcr_next_link_id(lgr); @@ -698,16 +708,13 @@ int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk, smc_lgr_hold(lgr); /* lgr_put in smcr_link_clear() */ lnk->link_idx = link_idx; lnk->wr_rx_id_compl = 0; - lnk->smcibdev = ini->ib_dev; - lnk->ibport = ini->ib_port; smc_ibdev_cnt_inc(lnk); smcr_copy_dev_info_to_link(lnk); - lnk->path_mtu = ini->ib_dev->pattr[ini->ib_port - 1].active_mtu; atomic_set(&lnk->conn_cnt, 0); smc_llc_link_set_uid(lnk); INIT_WORK(&lnk->link_down_wrk, smc_link_down_work); - if (!ini->ib_dev->initialized) { - rc = (int)smc_ib_setup_per_ibdev(ini->ib_dev); + if (!lnk->smcibdev->initialized) { + rc = (int)smc_ib_setup_per_ibdev(lnk->smcibdev); if (rc) goto out; } @@ -746,11 +753,12 @@ int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk, smc_llc_link_clear(lnk, false); out: smc_ibdev_cnt_dec(lnk); - put_device(&ini->ib_dev->ibdev->dev); + put_device(&lnk->smcibdev->ibdev->dev); + smcibdev = lnk->smcibdev; memset(lnk, 0, sizeof(struct smc_link)); lnk->state = SMC_LNK_UNUSED; - if (!atomic_dec_return(&ini->ib_dev->lnk_cnt)) - wake_up(&ini->ib_dev->lnks_deleted); + if (!atomic_dec_return(&smcibdev->lnk_cnt)) + wake_up(&smcibdev->lnks_deleted); return rc; }

@@ -815,10 +823,25 @@ static int smc_lgr_create(struct smc_sock *smc, struct smc_init_info *ini) atomic_inc(&ini->ism_dev[ini->ism_selected]->lgr_cnt); } else { /* SMC-R specific settings */ + struct smc_ib_device *ibdev; + int ibport; + lgr->role = smc->listen_smc ? SMC_SERV : SMC_CLNT; - memcpy(lgr->peer_systemid, ini->ib_lcl->id_for_peer, + lgr->smc_version = ini->smcr_version; + memcpy(lgr->peer_systemid, ini->peer_systemid, SMC_SYSTEMID_LEN); - memcpy(lgr->pnet_id, ini->ib_dev->pnetid[ini->ib_port - 1], + if (lgr->smc_version == SMC_V2) { + ibdev = ini->smcrv2.ib_dev_v2; + ibport = ini->smcrv2.ib_port_v2; + lgr->saddr = ini->smcrv2.saddr; + lgr->uses_gateway = ini->smcrv2.uses_gateway; + memcpy(lgr->nexthop_mac, ini->smcrv2.nexthop_mac, + ETH_ALEN); + } else { + ibdev = ini->ib_dev; + ibport = ini->ib_port; + } + memcpy(lgr->pnet_id, ibdev->pnetid[ibport - 1], SMC_MAX_PNETID_LEN); smc_llc_lgr_init(lgr, smc);

@@ -1740,13 +1763,15 @@ int smc_vlan_by_tcpsk(struct socket *clcsock, struct smc_init_info *ini) return rc; }

-static bool smcr_lgr_match(struct smc_link_group *lgr, - struct smc_clc_msg_local *lcl, +static bool smcr_lgr_match(struct smc_link_group *lgr, u8 smcr_version, + u8 peer_systemid[], + u8 peer_gid[], + u8 peer_mac_v1[], enum smc_lgr_role role, u32 clcqpn) { int i;

- if (memcmp(lgr->peer_systemid, lcl->id_for_peer, SMC_SYSTEMID_LEN) || + if (memcmp(lgr->peer_systemid, peer_systemid, SMC_SYSTEMID_LEN) || lgr->role != role) return false;

@@ -1754,8 +1779,9 @@ static bool smcr_lgr_match(struct smc_link_group *lgr, if (!smc_link_active(&lgr->lnk[i])) continue; if ((lgr->role == SMC_SERV || lgr->lnk[i].peer_qpn == clcqpn) && - !memcmp(lgr->lnk[i].peer_gid, &lcl->gid, SMC_GID_SIZE) && - !memcmp(lgr->lnk[i].peer_mac, lcl->mac, sizeof(lcl->mac))) + !memcmp(lgr->lnk[i].peer_gid, peer_gid, SMC_GID_SIZE) && + (smcr_version == SMC_V2 || + !memcmp(lgr->lnk[i].peer_mac, peer_mac_v1, ETH_ALEN))) return true; } return false; @@ -1794,7 +1820,10 @@ int smc_conn_create(struct smc_sock *smc, struct smc_init_info *ini) if ((ini->is_smcd ? smcd_lgr_match(lgr, ini->ism_dev[ini->ism_selected], ini->ism_peer_gid[ini->ism_selected]) : - smcr_lgr_match(lgr, ini->ib_lcl, role, ini->ib_clcqpn)) && + smcr_lgr_match(lgr, ini->smcr_version, + ini->peer_systemid, + ini->peer_gid, ini->peer_mac, role, + ini->ib_clcqpn)) && !lgr->sync_err && (ini->smcd_version == SMC_V2 || lgr->vlan_id == ini->vlan_id) && diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 34e7a20d5613..004d227521b3 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -358,7 +358,6 @@ struct smc_init_info { u32 rc; u8 negotiated_eid[SMC_MAX_EID_LEN]; /* SMC-R */ - struct smc_clc_msg_local *ib_lcl; u8 smcr_version; u8 check_smcrv2; u8 peer_gid[SMC_GID_SIZE];

From: Karsten Graul kgraul@linux.ibm.com

mainline inclusion from mainline-v5.16-rc1 commit 24fb68111d4509524b483b2577f1b20a24f5fdfd category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

In smc_ib.c, scan for RoCE devices that support UDP encapsulation. Find an eligible device and check that there is a route to the remote peer.

Signed-off-by: Karsten Graul kgraul@linux.ibm.com Signed-off-by: David S. Miller davem@davemloft.net --- net/smc/smc_core.c | 4 ++- net/smc/smc_ib.c | 74 +++++++++++++++++++++++++++++++++++++++------- net/smc/smc_ib.h | 3 +- net/smc/smc_pnet.c | 41 +++++++++++++++---------- 4 files changed, 94 insertions(+), 28 deletions(-)

diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 0814aa713845..ba63f1120cf0 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -722,7 +722,9 @@ int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk, lnk->psn_initial = rndvec[0] + (rndvec[1] << 8) + (rndvec[2] << 16); rc = smc_ib_determine_gid(lnk->smcibdev, lnk->ibport, - ini->vlan_id, lnk->gid, &lnk->sgid_index); + ini->vlan_id, lnk->gid, &lnk->sgid_index, + lgr->smc_version == SMC_V2 ? + &ini->smcrv2 : NULL); if (rc) goto out; rc = smc_llc_link_init(lnk); diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c index e6e6c0ea1c8a..0c213b5d6f87 100644 --- a/net/smc/smc_ib.c +++ b/net/smc/smc_ib.c @@ -17,6 +17,7 @@ #include <linux/scatterlist.h> #include <linux/wait.h> #include <linux/mutex.h> +#include <linux/inetdevice.h> #include <rdma/ib_verbs.h> #include <rdma/ib_cache.h>

@@ -62,16 +63,23 @@ static int smc_ib_modify_qp_rtr(struct smc_link *lnk) IB_QP_STATE | IB_QP_AV | IB_QP_PATH_MTU | IB_QP_DEST_QPN | IB_QP_RQ_PSN | IB_QP_MAX_DEST_RD_ATOMIC | IB_QP_MIN_RNR_TIMER; struct ib_qp_attr qp_attr; + u8 hop_lim = 1;

memset(&qp_attr, 0, sizeof(qp_attr)); qp_attr.qp_state = IB_QPS_RTR; qp_attr.path_mtu = min(lnk->path_mtu, lnk->peer_mtu); qp_attr.ah_attr.type = RDMA_AH_ATTR_TYPE_ROCE; rdma_ah_set_port_num(&qp_attr.ah_attr, lnk->ibport); - rdma_ah_set_grh(&qp_attr.ah_attr, NULL, 0, lnk->sgid_index, 1, 0); + if (lnk->lgr->smc_version == SMC_V2 && lnk->lgr->uses_gateway) + hop_lim = IPV6_DEFAULT_HOPLIMIT; + rdma_ah_set_grh(&qp_attr.ah_attr, NULL, 0, lnk->sgid_index, hop_lim, 0); rdma_ah_set_dgid_raw(&qp_attr.ah_attr, lnk->peer_gid); - memcpy(&qp_attr.ah_attr.roce.dmac, lnk->peer_mac, - sizeof(lnk->peer_mac)); + if (lnk->lgr->smc_version == SMC_V2 && lnk->lgr->uses_gateway) + memcpy(&qp_attr.ah_attr.roce.dmac, lnk->lgr->nexthop_mac, + sizeof(lnk->lgr->nexthop_mac)); + else + memcpy(&qp_attr.ah_attr.roce.dmac, lnk->peer_mac, + sizeof(lnk->peer_mac)); qp_attr.dest_qp_num = lnk->peer_qpn; qp_attr.rq_psn = lnk->peer_psn; /* starting receive packet seq # */ qp_attr.max_dest_rd_atomic = 1; /* max # of resources for incoming @@ -210,9 +218,54 @@ int smc_ib_find_route(__be32 saddr, __be32 daddr, return -ENOENT; }

+static int smc_ib_determine_gid_rcu(const struct net_device *ndev, + const struct ib_gid_attr *attr, + u8 gid[], u8 *sgid_index, + struct smc_init_info_smcrv2 *smcrv2) +{ + if (!smcrv2 && attr->gid_type == IB_GID_TYPE_ROCE) { + if (gid) + memcpy(gid, &attr->gid, SMC_GID_SIZE); + if (sgid_index) + *sgid_index = attr->index; + return 0; + } + if (smcrv2 && attr->gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP && + smc_ib_gid_to_ipv4((u8 *)&attr->gid) != cpu_to_be32(INADDR_NONE)) { + struct in_device *in_dev = __in_dev_get_rcu(ndev); + const struct in_ifaddr *ifa; + bool subnet_match = false; + + if (!in_dev) + goto out; + in_dev_for_each_ifa_rcu(ifa, in_dev) { + if (!inet_ifa_match(smcrv2->saddr, ifa)) + continue; + subnet_match = true; + break; + } + if (!subnet_match) + goto out; + if (smcrv2->daddr && smc_ib_find_route(smcrv2->saddr, + smcrv2->daddr, + smcrv2->nexthop_mac, + &smcrv2->uses_gateway)) + goto out; + + if (gid) + memcpy(gid, &attr->gid, SMC_GID_SIZE); + if (sgid_index) + *sgid_index = attr->index; + return 0; + } +out: + return -ENODEV; +} + /* determine the gid for an ib-device port and vlan id */ int smc_ib_determine_gid(struct smc_ib_device *smcibdev, u8 ibport, - unsigned short vlan_id, u8 gid[], u8 *sgid_index) + unsigned short vlan_id, u8 gid[], u8 *sgid_index, + struct smc_init_info_smcrv2 *smcrv2) { const struct ib_gid_attr *attr; const struct net_device *ndev; @@ -229,13 +282,12 @@ int smc_ib_determine_gid(struct smc_ib_device *smcibdev, u8 ibport, ((!vlan_id && !is_vlan_dev(ndev)) || (vlan_id && is_vlan_dev(ndev) && vlan_dev_vlan_id(ndev) == vlan_id))) { - rcu_read_unlock(); - if (gid) - memcpy(gid, &attr->gid, SMC_GID_SIZE); - if (sgid_index) - *sgid_index = attr->index; - rdma_put_gid_attr(attr); - return 0; + if (!smc_ib_determine_gid_rcu(ndev, attr, gid, + sgid_index, smcrv2)) { + rcu_read_unlock(); + rdma_put_gid_attr(attr); + return 0; + } } rcu_read_unlock(); rdma_put_gid_attr(attr); diff --git a/net/smc/smc_ib.h b/net/smc/smc_ib.h index 32d2f6d0fcca..fdb54871b52f 100644 --- a/net/smc/smc_ib.h +++ b/net/smc/smc_ib.h @@ -104,7 +104,8 @@ void smc_ib_sync_sg_for_device(struct smc_link *lnk, struct smc_buf_desc *buf_slot, enum dma_data_direction data_direction); int smc_ib_determine_gid(struct smc_ib_device *smcibdev, u8 ibport, - unsigned short vlan_id, u8 gid[], u8 *sgid_index); + unsigned short vlan_id, u8 gid[], u8 *sgid_index, + struct smc_init_info_smcrv2 *smcrv2); int smc_ib_find_route(__be32 saddr, __be32 daddr, u8 nexthop_mac[], u8 *uses_gateway); bool smc_ib_is_valid_local_systemid(void); diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c index bf8c662a6116..b3c54c7409c3 100644 --- a/net/smc/smc_pnet.c +++ b/net/smc/smc_pnet.c @@ -965,6 +965,26 @@ static int smc_pnet_find_ndev_pnetid_by_table(struct net_device *ndev, return rc; }

+static int smc_pnet_determine_gid(struct smc_ib_device *ibdev, int i, + struct smc_init_info *ini) +{ + if (!ini->check_smcrv2 && + !smc_ib_determine_gid(ibdev, i, ini->vlan_id, ini->ib_gid, NULL, + NULL)) { + ini->ib_dev = ibdev; + ini->ib_port = i; + return 0; + } + if (ini->check_smcrv2 && + !smc_ib_determine_gid(ibdev, i, ini->vlan_id, ini->smcrv2.ib_gid_v2, + NULL, &ini->smcrv2)) { + ini->smcrv2.ib_dev_v2 = ibdev; + ini->smcrv2.ib_port_v2 = i; + return 0; + } + return -ENODEV; +} + /* find a roce device for the given pnetid */ static void _smc_pnet_find_roce_by_pnetid(u8 *pnet_id, struct smc_init_info *ini, @@ -973,7 +993,6 @@ static void _smc_pnet_find_roce_by_pnetid(u8 *pnet_id, struct smc_ib_device *ibdev; int i;

- ini->ib_dev = NULL; mutex_lock(&smc_ib_devices.mutex); list_for_each_entry(ibdev, &smc_ib_devices.list, list) { if (ibdev == known_dev) @@ -983,12 +1002,9 @@ static void _smc_pnet_find_roce_by_pnetid(u8 *pnet_id, continue; if (smc_pnet_match(ibdev->pnetid[i - 1], pnet_id) && smc_ib_port_active(ibdev, i) && - !test_bit(i - 1, ibdev->ports_going_away) && - !smc_ib_determine_gid(ibdev, i, ini->vlan_id, - ini->ib_gid, NULL)) { - ini->ib_dev = ibdev; - ini->ib_port = i; - goto out; + !test_bit(i - 1, ibdev->ports_going_away)) { + if (!smc_pnet_determine_gid(ibdev, i, ini)) + goto out; } } } @@ -1028,12 +1044,9 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, dev_put(ndev); if (netdev == ndev && smc_ib_port_active(ibdev, i) && - !test_bit(i - 1, ibdev->ports_going_away) && - !smc_ib_determine_gid(ibdev, i, ini->vlan_id, - ini->ib_gid, NULL)) { - ini->ib_dev = ibdev; - ini->ib_port = i; - break; + !test_bit(i - 1, ibdev->ports_going_away)) { + if (!smc_pnet_determine_gid(ibdev, i, ini)) + break; } } } @@ -1095,8 +1108,6 @@ void smc_pnet_find_roce_resource(struct sock *sk, struct smc_init_info *ini) { struct dst_entry *dst = sk_dst_get(sk);

- ini->ib_dev = NULL; - ini->ib_port = 0; if (!dst) goto out; if (!dst->dev)

From: Karsten Graul kgraul@linux.ibm.com

mainline inclusion from mainline-v5.16-rc1 commit b4ba4652b3f8b7c9bbb5786f8acf4724bdab2196 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

Add support for large v2 LLC control messages in smc_llc.c. The new large work request buffer allows to combine control messages into one packet that had to be spread over several packets before. Add handling of the new v2 LLC messages.

Signed-off-by: Karsten Graul kgraul@linux.ibm.com Signed-off-by: David S. Miller davem@davemloft.net

Conflicts: net/smc/smc_llc.c [The conflict occurs because the commit f7ffd1b2110f("net/smc: Allow virtually contiguous sndbufs or RMBs for SMC-R") is merged] Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- net/smc/af_smc.c | 8 +- net/smc/smc.h | 20 +- net/smc/smc_core.h | 1 + net/smc/smc_llc.c | 556 ++++++++++++++++++++++++++++++++++++--------- net/smc/smc_llc.h | 12 +- net/smc/smc_wr.c | 43 ++++ net/smc/smc_wr.h | 6 + 7 files changed, 531 insertions(+), 115 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index c1fc8d0a27c1..cd362f04aa5f 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -1052,9 +1052,9 @@ static int smc_connect_clc(struct smc_sock *smc, SMC_CLC_ACCEPT, CLC_WAIT_TIME); }

-static void smc_fill_gid_list(struct smc_link_group *lgr, - struct smc_gidlist *gidlist, - struct smc_ib_device *known_dev, u8 *known_gid) +void smc_fill_gid_list(struct smc_link_group *lgr, + struct smc_gidlist *gidlist, + struct smc_ib_device *known_dev, u8 *known_gid) { struct smc_init_info *alt_ini = NULL;

@@ -1767,7 +1767,7 @@ static int smcr_serv_conf_first_link(struct smc_sock *smc) smcr_lgr_set_type(link->lgr, SMC_LGR_SINGLE);

/* initial contact - try to establish second link */ - smc_llc_srv_add_link(link); + smc_llc_srv_add_link(link, NULL); return 0; }

diff --git a/net/smc/smc.h b/net/smc/smc.h index 4fb888366cad..fb94ab6636d1 100644 --- a/net/smc/smc.h +++ b/net/smc/smc.h @@ -53,7 +53,20 @@ enum smc_state { /* possible states of an SMC socket */ struct smc_link_group;

struct smc_wr_rx_hdr { /* common prefix part of LLC and CDC to demultiplex */ - u8 type; + union { + u8 type; +#if defined(__BIG_ENDIAN_BITFIELD) + struct { + u8 llc_version:4, + llc_type:4; + }; +#elif defined(__LITTLE_ENDIAN_BITFIELD) + struct { + u8 llc_type:4, + llc_version:4; + }; +#endif + }; } __aligned(1);

struct smc_cdc_conn_state_flags { @@ -350,7 +363,12 @@ static inline bool using_ipsec(struct smc_sock *smc) } #endif

+struct smc_gidlist; + struct sock *smc_accept_dequeue(struct sock *parent, struct socket *new_sock); void smc_close_non_accepted(struct sock *sk); +void smc_fill_gid_list(struct smc_link_group *lgr, + struct smc_gidlist *gidlist, + struct smc_ib_device *known_dev, u8 *known_gid);

#endif /* __SMC_H */ diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h index 39494db9cc03..0caa4f1d2931 100644 --- a/net/smc/smc_core.h +++ b/net/smc/smc_core.h @@ -234,6 +234,7 @@ enum smc_llc_flowtype { SMC_LLC_FLOW_NONE = 0, SMC_LLC_FLOW_ADD_LINK = 2, SMC_LLC_FLOW_DEL_LINK = 4, + SMC_LLC_FLOW_REQ_ADD_LINK = 5, SMC_LLC_FLOW_RKEY = 6, };

diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c index 050cbaad2ae9..166713cb17d3 100644 --- a/net/smc/smc_llc.c +++ b/net/smc/smc_llc.c @@ -23,16 +23,24 @@

struct smc_llc_hdr { struct smc_wr_rx_hdr common; - u8 length; /* 44 */ -#if defined(__BIG_ENDIAN_BITFIELD) - u8 reserved:4, - add_link_rej_rsn:4; + union { + struct { + u8 length; /* 44 */ + #if defined(__BIG_ENDIAN_BITFIELD) + u8 reserved:4, + add_link_rej_rsn:4; #elif defined(__LITTLE_ENDIAN_BITFIELD) - u8 add_link_rej_rsn:4, - reserved:4; + u8 add_link_rej_rsn:4, + reserved:4; #endif + }; + u16 length_v2; /* 44 - 8192*/ + }; u8 flags; -}; +} __packed; /* format defined in + * IBM Shared Memory Communications Version 2 + * (https://www.ibm.com/support/pages/node/6326337) + */

#define SMC_LLC_FLAG_NO_RMBE_EYEC 0x03

@@ -76,6 +84,32 @@ struct smc_llc_msg_add_link_cont_rt { __be64 rmb_vaddr_new; };

+struct smc_llc_msg_add_link_v2_ext { +#if defined(__BIG_ENDIAN_BITFIELD) + u8 v2_direct : 1, + reserved : 7; +#elif defined(__LITTLE_ENDIAN_BITFIELD) + u8 reserved : 7, + v2_direct : 1; +#endif + u8 reserved2; + u8 client_target_gid[SMC_GID_SIZE]; + u8 reserved3[8]; + u16 num_rkeys; + struct smc_llc_msg_add_link_cont_rt rt[]; +} __packed; /* format defined in + * IBM Shared Memory Communications Version 2 + * (https://www.ibm.com/support/pages/node/6326337) + */ + +struct smc_llc_msg_req_add_link_v2 { + struct smc_llc_hdr hd; + u8 reserved[20]; + u8 gid_cnt; + u8 reserved2[3]; + u8 gid[][SMC_GID_SIZE]; +}; + #define SMC_LLC_RKEYS_PER_CONT_MSG 2

struct smc_llc_msg_add_link_cont { /* type 0x03 */ @@ -114,7 +148,8 @@ struct smc_rmb_rtoken { __be64 rmb_vaddr; } __packed; /* format defined in RFC7609 */

-#define SMC_LLC_RKEYS_PER_MSG 3 +#define SMC_LLC_RKEYS_PER_MSG 3 +#define SMC_LLC_RKEYS_PER_MSG_V2 255

struct smc_llc_msg_confirm_rkey { /* type 0x06 */ struct smc_llc_hdr hd; @@ -135,9 +170,18 @@ struct smc_llc_msg_delete_rkey { /* type 0x09 */ u8 reserved2[4]; };

+struct smc_llc_msg_delete_rkey_v2 { /* type 0x29 */ + struct smc_llc_hdr hd; + u8 num_rkeys; + u8 num_inval_rkeys; + u8 reserved[2]; + __be32 rkey[]; +}; + union smc_llc_msg { struct smc_llc_msg_confirm_link confirm_link; struct smc_llc_msg_add_link add_link; + struct smc_llc_msg_req_add_link_v2 req_add_link; struct smc_llc_msg_add_link_cont add_link_cont; struct smc_llc_msg_del_link delete_link;

@@ -189,7 +233,7 @@ static inline void smc_llc_flow_qentry_set(struct smc_llc_flow *flow, static void smc_llc_flow_parallel(struct smc_link_group *lgr, u8 flow_type, struct smc_llc_qentry *qentry) { - u8 msg_type = qentry->msg.raw.hdr.common.type; + u8 msg_type = qentry->msg.raw.hdr.common.llc_type;

if ((msg_type == SMC_LLC_ADD_LINK || msg_type == SMC_LLC_DELETE_LINK) && flow_type != msg_type && !lgr->delayed_event) { @@ -219,7 +263,7 @@ static bool smc_llc_flow_start(struct smc_llc_flow *flow, spin_unlock_bh(&lgr->llc_flow_lock); return false; } - switch (qentry->msg.raw.hdr.common.type) { + switch (qentry->msg.raw.hdr.common.llc_type) { case SMC_LLC_ADD_LINK: flow->type = SMC_LLC_FLOW_ADD_LINK; break; @@ -306,7 +350,7 @@ struct smc_llc_qentry *smc_llc_wait(struct smc_link_group *lgr, smc_llc_flow_qentry_del(flow); goto out; } - rcv_msg = flow->qentry->msg.raw.hdr.common.type; + rcv_msg = flow->qentry->msg.raw.hdr.common.llc_type; if (exp_msg && rcv_msg != exp_msg) { if (exp_msg == SMC_LLC_ADD_LINK && rcv_msg == SMC_LLC_DELETE_LINK) { @@ -374,6 +418,30 @@ static int smc_llc_add_pending_send(struct smc_link *link, return 0; }

+static int smc_llc_add_pending_send_v2(struct smc_link *link, + struct smc_wr_v2_buf **wr_buf, + struct smc_wr_tx_pend_priv **pend) +{ + int rc; + + rc = smc_wr_tx_get_v2_slot(link, smc_llc_tx_handler, wr_buf, pend); + if (rc < 0) + return rc; + return 0; +} + +static void smc_llc_init_msg_hdr(struct smc_llc_hdr *hdr, + struct smc_link_group *lgr, size_t len) +{ + if (lgr->smc_version == SMC_V2) { + hdr->common.llc_version = SMC_V2; + hdr->length_v2 = len; + } else { + hdr->common.llc_version = 0; + hdr->length = len; + } +} + /* high-level API to send LLC confirm link */ int smc_llc_send_confirm_link(struct smc_link *link, enum smc_llc_reqresp reqresp) @@ -390,8 +458,8 @@ int smc_llc_send_confirm_link(struct smc_link *link, goto put_out; confllc = (struct smc_llc_msg_confirm_link *)wr_buf; memset(confllc, 0, sizeof(*confllc)); - confllc->hd.common.type = SMC_LLC_CONFIRM_LINK; - confllc->hd.length = sizeof(struct smc_llc_msg_confirm_link); + confllc->hd.common.llc_type = SMC_LLC_CONFIRM_LINK; + smc_llc_init_msg_hdr(&confllc->hd, link->lgr, sizeof(*confllc)); confllc->hd.flags |= SMC_LLC_FLAG_NO_RMBE_EYEC; if (reqresp == SMC_LLC_RESP) confllc->hd.flags |= SMC_LLC_FLAG_RESP; @@ -426,8 +494,8 @@ static int smc_llc_send_confirm_rkey(struct smc_link *send_link, goto put_out; rkeyllc = (struct smc_llc_msg_confirm_rkey *)wr_buf; memset(rkeyllc, 0, sizeof(*rkeyllc)); - rkeyllc->hd.common.type = SMC_LLC_CONFIRM_RKEY; - rkeyllc->hd.length = sizeof(struct smc_llc_msg_confirm_rkey); + rkeyllc->hd.common.llc_type = SMC_LLC_CONFIRM_RKEY; + smc_llc_init_msg_hdr(&rkeyllc->hd, send_link->lgr, sizeof(*rkeyllc));

rtok_ix = 1; for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) { @@ -474,8 +542,8 @@ static int smc_llc_send_delete_rkey(struct smc_link *link, goto put_out; rkeyllc = (struct smc_llc_msg_delete_rkey *)wr_buf; memset(rkeyllc, 0, sizeof(*rkeyllc)); - rkeyllc->hd.common.type = SMC_LLC_DELETE_RKEY; - rkeyllc->hd.length = sizeof(struct smc_llc_msg_delete_rkey); + rkeyllc->hd.common.llc_type = SMC_LLC_DELETE_RKEY; + smc_llc_init_msg_hdr(&rkeyllc->hd, link->lgr, sizeof(*rkeyllc)); rkeyllc->num_rkeys = 1; rkeyllc->rkey[0] = htonl(rmb_desc->mr[link->link_idx]->rkey); /* send llc message */ @@ -485,26 +553,116 @@ static int smc_llc_send_delete_rkey(struct smc_link *link, return rc; }

+/* return first buffer from any of the next buf lists */ +static struct smc_buf_desc *_smc_llc_get_next_rmb(struct smc_link_group *lgr, + int *buf_lst) +{ + struct smc_buf_desc *buf_pos; + + while (*buf_lst < SMC_RMBE_SIZES) { + buf_pos = list_first_entry_or_null(&lgr->rmbs[*buf_lst], + struct smc_buf_desc, list); + if (buf_pos) + return buf_pos; + (*buf_lst)++; + } + return NULL; +} + +/* return next rmb from buffer lists */ +static struct smc_buf_desc *smc_llc_get_next_rmb(struct smc_link_group *lgr, + int *buf_lst, + struct smc_buf_desc *buf_pos) +{ + struct smc_buf_desc *buf_next; + + if (!buf_pos || list_is_last(&buf_pos->list, &lgr->rmbs[*buf_lst])) { + (*buf_lst)++; + return _smc_llc_get_next_rmb(lgr, buf_lst); + } + buf_next = list_next_entry(buf_pos, list); + return buf_next; +} + +static struct smc_buf_desc *smc_llc_get_first_rmb(struct smc_link_group *lgr, + int *buf_lst) +{ + *buf_lst = 0; + return smc_llc_get_next_rmb(lgr, buf_lst, NULL); +} + +static int smc_llc_fill_ext_v2(struct smc_llc_msg_add_link_v2_ext *ext, + struct smc_link *link, struct smc_link *link_new) +{ + struct smc_link_group *lgr = link->lgr; + struct smc_buf_desc *buf_pos; + int prim_lnk_idx, lnk_idx, i; + struct smc_buf_desc *rmb; + int len = sizeof(*ext); + int buf_lst; + + ext->v2_direct = !lgr->uses_gateway; + memcpy(ext->client_target_gid, link_new->gid, SMC_GID_SIZE); + + prim_lnk_idx = link->link_idx; + lnk_idx = link_new->link_idx; + down_write(&lgr->rmbs_lock); + ext->num_rkeys = lgr->conns_num; + if (!ext->num_rkeys) + goto out; + buf_pos = smc_llc_get_first_rmb(lgr, &buf_lst); + for (i = 0; i < ext->num_rkeys; i++) { + if (!buf_pos) + break; + rmb = buf_pos; + ext->rt[i].rmb_key = htonl(rmb->mr[prim_lnk_idx]->rkey); + ext->rt[i].rmb_key_new = htonl(rmb->mr[lnk_idx]->rkey); + ext->rt[i].rmb_vaddr_new = + cpu_to_be64((u64)sg_dma_address(rmb->sgt[lnk_idx].sgl)); + buf_pos = smc_llc_get_next_rmb(lgr, &buf_lst, buf_pos); + while (buf_pos && !(buf_pos)->used) + buf_pos = smc_llc_get_next_rmb(lgr, &buf_lst, buf_pos); + } + len += i * sizeof(ext->rt[0]); +out: + up_write(&lgr->rmbs_lock); + return len; +} + /* send ADD LINK request or response */ int smc_llc_send_add_link(struct smc_link *link, u8 mac[], u8 gid[], struct smc_link *link_new, enum smc_llc_reqresp reqresp) { + struct smc_llc_msg_add_link_v2_ext *ext = NULL; struct smc_llc_msg_add_link *addllc; struct smc_wr_tx_pend_priv *pend; - struct smc_wr_buf *wr_buf; + int len = sizeof(*addllc); int rc;

if (!smc_wr_tx_link_hold(link)) return -ENOLINK; - rc = smc_llc_add_pending_send(link, &wr_buf, &pend); - if (rc) - goto put_out; - addllc = (struct smc_llc_msg_add_link *)wr_buf; + if (link->lgr->smc_version == SMC_V2) { + struct smc_wr_v2_buf *wr_buf; + + rc = smc_llc_add_pending_send_v2(link, &wr_buf, &pend); + if (rc) + goto put_out; + addllc = (struct smc_llc_msg_add_link *)wr_buf; + ext = (struct smc_llc_msg_add_link_v2_ext *) + &wr_buf->raw[sizeof(*addllc)]; + memset(ext, 0, SMC_WR_TX_SIZE); + } else { + struct smc_wr_buf *wr_buf; + + rc = smc_llc_add_pending_send(link, &wr_buf, &pend); + if (rc) + goto put_out; + addllc = (struct smc_llc_msg_add_link *)wr_buf; + }

memset(addllc, 0, sizeof(*addllc)); - addllc->hd.common.type = SMC_LLC_ADD_LINK; - addllc->hd.length = sizeof(struct smc_llc_msg_add_link); + addllc->hd.common.llc_type = SMC_LLC_ADD_LINK; if (reqresp == SMC_LLC_RESP) addllc->hd.flags |= SMC_LLC_FLAG_RESP; memcpy(addllc->sender_mac, mac, ETH_ALEN); @@ -519,8 +677,14 @@ int smc_llc_send_add_link(struct smc_link *link, u8 mac[], u8 gid[], addllc->qp_mtu = min(link_new->path_mtu, link_new->peer_mtu); } + if (ext && link_new) + len += smc_llc_fill_ext_v2(ext, link, link_new); + smc_llc_init_msg_hdr(&addllc->hd, link->lgr, len); /* send llc message */ - rc = smc_wr_tx_send(link, pend); + if (link->lgr->smc_version == SMC_V2) + rc = smc_wr_tx_v2_send(link, pend, len); + else + rc = smc_wr_tx_send(link, pend); put_out: smc_wr_tx_link_put(link); return rc; @@ -544,8 +708,8 @@ int smc_llc_send_delete_link(struct smc_link *link, u8 link_del_id, delllc = (struct smc_llc_msg_del_link *)wr_buf;

memset(delllc, 0, sizeof(*delllc)); - delllc->hd.common.type = SMC_LLC_DELETE_LINK; - delllc->hd.length = sizeof(struct smc_llc_msg_del_link); + delllc->hd.common.llc_type = SMC_LLC_DELETE_LINK; + smc_llc_init_msg_hdr(&delllc->hd, link->lgr, sizeof(*delllc)); if (reqresp == SMC_LLC_RESP) delllc->hd.flags |= SMC_LLC_FLAG_RESP; if (orderly) @@ -577,8 +741,8 @@ static int smc_llc_send_test_link(struct smc_link *link, u8 user_data[16]) goto put_out; testllc = (struct smc_llc_msg_test_link *)wr_buf; memset(testllc, 0, sizeof(*testllc)); - testllc->hd.common.type = SMC_LLC_TEST_LINK; - testllc->hd.length = sizeof(struct smc_llc_msg_test_link); + testllc->hd.common.llc_type = SMC_LLC_TEST_LINK; + smc_llc_init_msg_hdr(&testllc->hd, link->lgr, sizeof(*testllc)); memcpy(testllc->user_data, user_data, sizeof(testllc->user_data)); /* send llc message */ rc = smc_wr_tx_send(link, pend); @@ -654,44 +818,6 @@ static int smc_llc_alloc_alt_link(struct smc_link_group *lgr, return -EMLINK; }

-/* return first buffer from any of the next buf lists */ -static struct smc_buf_desc *_smc_llc_get_next_rmb(struct smc_link_group *lgr, - int *buf_lst) -{ - struct smc_buf_desc *buf_pos; - - while (*buf_lst < SMC_RMBE_SIZES) { - buf_pos = list_first_entry_or_null(&lgr->rmbs[*buf_lst], - struct smc_buf_desc, list); - if (buf_pos) - return buf_pos; - (*buf_lst)++; - } - return NULL; -} - -/* return next rmb from buffer lists */ -static struct smc_buf_desc *smc_llc_get_next_rmb(struct smc_link_group *lgr, - int *buf_lst, - struct smc_buf_desc *buf_pos) -{ - struct smc_buf_desc *buf_next; - - if (!buf_pos || list_is_last(&buf_pos->list, &lgr->rmbs[*buf_lst])) { - (*buf_lst)++; - return _smc_llc_get_next_rmb(lgr, buf_lst); - } - buf_next = list_next_entry(buf_pos, list); - return buf_next; -} - -static struct smc_buf_desc *smc_llc_get_first_rmb(struct smc_link_group *lgr, - int *buf_lst) -{ - *buf_lst = 0; - return smc_llc_get_next_rmb(lgr, buf_lst, NULL); -} - /* send one add_link_continue msg */ static int smc_llc_add_link_cont(struct smc_link *link, struct smc_link *link_new, u8 *num_rkeys_todo, @@ -738,7 +864,7 @@ static int smc_llc_add_link_cont(struct smc_link *link, (*num_rkeys_todo)--; *buf_pos = smc_llc_get_next_rmb(lgr, buf_lst, *buf_pos); } - addc_llc->hd.common.type = SMC_LLC_ADD_LINK_CONT; + addc_llc->hd.common.llc_type = SMC_LLC_ADD_LINK_CONT; addc_llc->hd.length = sizeof(struct smc_llc_msg_add_link_cont); if (lgr->role == SMC_CLNT) addc_llc->hd.flags |= SMC_LLC_FLAG_RESP; @@ -797,6 +923,8 @@ static int smc_llc_cli_add_link_reject(struct smc_llc_qentry *qentry) qentry->msg.raw.hdr.flags |= SMC_LLC_FLAG_RESP; qentry->msg.raw.hdr.flags |= SMC_LLC_FLAG_ADD_LNK_REJ; qentry->msg.raw.hdr.add_link_rej_rsn = SMC_LLC_REJ_RSN_NO_ALT_PATH; + smc_llc_init_msg_hdr(&qentry->msg.raw.hdr, qentry->link->lgr, + sizeof(qentry->msg)); return smc_llc_send_message(qentry->link, &qentry->msg); }

@@ -817,7 +945,7 @@ static int smc_llc_cli_conf_link(struct smc_link *link, SMC_LLC_DEL_LOST_PATH); return -ENOLINK; } - if (qentry->msg.raw.hdr.common.type != SMC_LLC_CONFIRM_LINK) { + if (qentry->msg.raw.hdr.common.llc_type != SMC_LLC_CONFIRM_LINK) { /* received DELETE_LINK instead */ qentry->msg.raw.hdr.flags |= SMC_LLC_FLAG_RESP; smc_llc_send_message(link, &qentry->msg); @@ -858,6 +986,26 @@ static int smc_llc_cli_conf_link(struct smc_link *link, return 0; }

+static void smc_llc_save_add_link_rkeys(struct smc_link *link, + struct smc_link *link_new) +{ + struct smc_llc_msg_add_link_v2_ext *ext; + struct smc_link_group *lgr = link->lgr; + int max, i; + + ext = (struct smc_llc_msg_add_link_v2_ext *)((u8 *)lgr->wr_rx_buf_v2 + + SMC_WR_TX_SIZE); + max = min_t(u8, ext->num_rkeys, SMC_LLC_RKEYS_PER_MSG_V2); + down_write(&lgr->rmbs_lock); + for (i = 0; i < max; i++) { + smc_rtoken_set(lgr, link->link_idx, link_new->link_idx, + ext->rt[i].rmb_key, + ext->rt[i].rmb_vaddr_new, + ext->rt[i].rmb_key_new); + } + up_write(&lgr->rmbs_lock); +} + static void smc_llc_save_add_link_info(struct smc_link *link, struct smc_llc_msg_add_link *add_llc) { @@ -888,14 +1036,24 @@ int smc_llc_cli_add_link(struct smc_link *link, struct smc_llc_qentry *qentry) }

ini->vlan_id = lgr->vlan_id; + if (lgr->smc_version == SMC_V2) { + ini->check_smcrv2 = true; + ini->smcrv2.saddr = lgr->saddr; + ini->smcrv2.daddr = smc_ib_gid_to_ipv4(llc->sender_gid); + } smc_pnet_find_alt_roce(lgr, ini, link->smcibdev); if (!memcmp(llc->sender_gid, link->peer_gid, SMC_GID_SIZE) && - !memcmp(llc->sender_mac, link->peer_mac, ETH_ALEN)) { - if (!ini->ib_dev) + (lgr->smc_version == SMC_V2 || + !memcmp(llc->sender_mac, link->peer_mac, ETH_ALEN))) { + if (!ini->ib_dev && !ini->smcrv2.ib_dev_v2) goto out_reject; lgr_new_t = SMC_LGR_ASYMMETRIC_PEER; } - if (!ini->ib_dev) { + if (lgr->smc_version == SMC_V2 && !ini->smcrv2.ib_dev_v2) { + lgr_new_t = SMC_LGR_ASYMMETRIC_LOCAL; + ini->smcrv2.ib_dev_v2 = link->smcibdev; + ini->smcrv2.ib_port_v2 = link->ibport; + } else if (lgr->smc_version < SMC_V2 && !ini->ib_dev) { lgr_new_t = SMC_LGR_ASYMMETRIC_LOCAL; ini->ib_dev = link->smcibdev; ini->ib_port = link->ibport; @@ -920,14 +1078,18 @@ int smc_llc_cli_add_link(struct smc_link *link, struct smc_llc_qentry *qentry) goto out_clear_lnk;

rc = smc_llc_send_add_link(link, - lnk_new->smcibdev->mac[ini->ib_port - 1], + lnk_new->smcibdev->mac[lnk_new->ibport - 1], lnk_new->gid, lnk_new, SMC_LLC_RESP); if (rc) goto out_clear_lnk; - rc = smc_llc_cli_rkey_exchange(link, lnk_new); - if (rc) { - rc = 0; - goto out_clear_lnk; + if (lgr->smc_version == SMC_V2) { + smc_llc_save_add_link_rkeys(link, lnk_new); + } else { + rc = smc_llc_cli_rkey_exchange(link, lnk_new); + if (rc) { + rc = 0; + goto out_clear_lnk; + } } rc = smc_llc_cli_conf_link(link, ini, lnk_new, lgr_new_t); if (!rc) @@ -943,6 +1105,44 @@ int smc_llc_cli_add_link(struct smc_link *link, struct smc_llc_qentry *qentry) return rc; }

+static void smc_llc_send_request_add_link(struct smc_link *link) +{ + struct smc_llc_msg_req_add_link_v2 *llc; + struct smc_wr_tx_pend_priv *pend; + struct smc_wr_v2_buf *wr_buf; + struct smc_gidlist gidlist; + int rc, len, i; + + if (!smc_wr_tx_link_hold(link)) + return; + if (link->lgr->type == SMC_LGR_SYMMETRIC || + link->lgr->type == SMC_LGR_ASYMMETRIC_PEER) + goto put_out; + + smc_fill_gid_list(link->lgr, &gidlist, link->smcibdev, link->gid); + if (gidlist.len <= 1) + goto put_out; + + rc = smc_llc_add_pending_send_v2(link, &wr_buf, &pend); + if (rc) + goto put_out; + llc = (struct smc_llc_msg_req_add_link_v2 *)wr_buf; + memset(llc, 0, SMC_WR_TX_SIZE); + + llc->hd.common.llc_type = SMC_LLC_REQ_ADD_LINK; + for (i = 0; i < gidlist.len; i++) + memcpy(llc->gid[i], gidlist.list[i], sizeof(gidlist.list[0])); + llc->gid_cnt = gidlist.len; + len = sizeof(*llc) + (gidlist.len * sizeof(gidlist.list[0])); + smc_llc_init_msg_hdr(&llc->hd, link->lgr, len); + rc = smc_wr_tx_v2_send(link, pend, len); + if (!rc) + /* set REQ_ADD_LINK flow and wait for response from peer */ + link->lgr->llc_flow_lcl.type = SMC_LLC_FLOW_REQ_ADD_LINK; +put_out: + smc_wr_tx_link_put(link); +} + /* as an SMC client, invite server to start the add_link processing */ static void smc_llc_cli_add_link_invite(struct smc_link *link, struct smc_llc_qentry *qentry) @@ -950,6 +1150,11 @@ static void smc_llc_cli_add_link_invite(struct smc_link *link, struct smc_link_group *lgr = smc_get_lgr(link); struct smc_init_info *ini = NULL;

+ if (lgr->smc_version == SMC_V2) { + smc_llc_send_request_add_link(link); + goto out; + } + if (lgr->type == SMC_LGR_SYMMETRIC || lgr->type == SMC_LGR_ASYMMETRIC_PEER) goto out; @@ -982,7 +1187,7 @@ static bool smc_llc_is_empty_llc_message(union smc_llc_msg *llc)

static bool smc_llc_is_local_add_link(union smc_llc_msg *llc) { - if (llc->raw.hdr.common.type == SMC_LLC_ADD_LINK && + if (llc->raw.hdr.common.llc_type == SMC_LLC_ADD_LINK && smc_llc_is_empty_llc_message(llc)) return true; return false; @@ -1149,7 +1354,7 @@ static int smc_llc_srv_conf_link(struct smc_link *link, /* receive CONFIRM LINK response over the RoCE fabric */ qentry = smc_llc_wait(lgr, link, SMC_LLC_WAIT_FIRST_TIME, 0); if (!qentry || - qentry->msg.raw.hdr.common.type != SMC_LLC_CONFIRM_LINK) { + qentry->msg.raw.hdr.common.llc_type != SMC_LLC_CONFIRM_LINK) { /* send DELETE LINK */ smc_llc_send_delete_link(link, link_new->link_id, SMC_LLC_REQ, false, SMC_LLC_DEL_LOST_PATH); @@ -1168,24 +1373,55 @@ static int smc_llc_srv_conf_link(struct smc_link *link, return 0; }

-int smc_llc_srv_add_link(struct smc_link *link) +static void smc_llc_send_req_add_link_response(struct smc_llc_qentry *qentry) +{ + qentry->msg.raw.hdr.flags |= SMC_LLC_FLAG_RESP; + smc_llc_init_msg_hdr(&qentry->msg.raw.hdr, qentry->link->lgr, + sizeof(qentry->msg)); + memset(&qentry->msg.raw.data, 0, sizeof(qentry->msg.raw.data)); + smc_llc_send_message(qentry->link, &qentry->msg); +} + +int smc_llc_srv_add_link(struct smc_link *link, + struct smc_llc_qentry *req_qentry) { enum smc_lgr_type lgr_new_t = SMC_LGR_SYMMETRIC; struct smc_link_group *lgr = link->lgr; struct smc_llc_msg_add_link *add_llc; struct smc_llc_qentry *qentry = NULL; + bool send_req_add_link_resp = false; struct smc_link *link_new = NULL; - struct smc_init_info *ini; + struct smc_init_info *ini = NULL; int lnk_idx, rc = 0;

+ if (req_qentry && + req_qentry->msg.raw.hdr.common.llc_type == SMC_LLC_REQ_ADD_LINK) + send_req_add_link_resp = true; + ini = kzalloc(sizeof(*ini), GFP_KERNEL); - if (!ini) - return -ENOMEM; + if (!ini) { + rc = -ENOMEM; + goto out; + }

/* ignore client add link recommendation, start new flow */ ini->vlan_id = lgr->vlan_id; + if (lgr->smc_version == SMC_V2) { + ini->check_smcrv2 = true; + ini->smcrv2.saddr = lgr->saddr; + if (send_req_add_link_resp) { + struct smc_llc_msg_req_add_link_v2 *req_add = + &req_qentry->msg.req_add_link; + + ini->smcrv2.daddr = smc_ib_gid_to_ipv4(req_add->gid[0]); + } + } smc_pnet_find_alt_roce(lgr, ini, link->smcibdev); - if (!ini->ib_dev) { + if (lgr->smc_version == SMC_V2 && !ini->smcrv2.ib_dev_v2) { + lgr_new_t = SMC_LGR_ASYMMETRIC_LOCAL; + ini->smcrv2.ib_dev_v2 = link->smcibdev; + ini->smcrv2.ib_port_v2 = link->ibport; + } else if (lgr->smc_version < SMC_V2 && !ini->ib_dev) { lgr_new_t = SMC_LGR_ASYMMETRIC_LOCAL; ini->ib_dev = link->smcibdev; ini->ib_port = link->ibport; @@ -1200,11 +1436,17 @@ int smc_llc_srv_add_link(struct smc_link *link) if (rc) goto out; link_new = &lgr->lnk[lnk_idx]; + + rc = smcr_buf_map_lgr(link_new); + if (rc) + goto out_err; + rc = smc_llc_send_add_link(link, - link_new->smcibdev->mac[ini->ib_port - 1], + link_new->smcibdev->mac[link_new->ibport-1], link_new->gid, link_new, SMC_LLC_REQ); if (rc) goto out_err; + send_req_add_link_resp = false; /* receive ADD LINK response over the RoCE fabric */ qentry = smc_llc_wait(lgr, link, SMC_LLC_WAIT_TIME, SMC_LLC_ADD_LINK); if (!qentry) { @@ -1219,24 +1461,26 @@ int smc_llc_srv_add_link(struct smc_link *link) } if (lgr->type == SMC_LGR_SINGLE && (!memcmp(add_llc->sender_gid, link->peer_gid, SMC_GID_SIZE) && - !memcmp(add_llc->sender_mac, link->peer_mac, ETH_ALEN))) { + (lgr->smc_version == SMC_V2 || + !memcmp(add_llc->sender_mac, link->peer_mac, ETH_ALEN)))) { lgr_new_t = SMC_LGR_ASYMMETRIC_PEER; } smc_llc_save_add_link_info(link_new, add_llc); smc_llc_flow_qentry_del(&lgr->llc_flow_lcl);

rc = smc_ib_ready_link(link_new); - if (rc) - goto out_err; - rc = smcr_buf_map_lgr(link_new); if (rc) goto out_err; rc = smcr_buf_reg_lgr(link_new); if (rc) goto out_err; - rc = smc_llc_srv_rkey_exchange(link, link_new); - if (rc) - goto out_err; + if (lgr->smc_version == SMC_V2) { + smc_llc_save_add_link_rkeys(link, link_new); + } else { + rc = smc_llc_srv_rkey_exchange(link, link_new); + if (rc) + goto out_err; + } rc = smc_llc_srv_conf_link(link, link_new, lgr_new_t); if (rc) goto out_err; @@ -1249,23 +1493,27 @@ int smc_llc_srv_add_link(struct smc_link *link) } out: kfree(ini); + if (send_req_add_link_resp) + smc_llc_send_req_add_link_response(req_qentry); return rc; }

static void smc_llc_process_srv_add_link(struct smc_link_group *lgr) { struct smc_link *link = lgr->llc_flow_lcl.qentry->link; + struct smc_llc_qentry *qentry; int rc;

- smc_llc_flow_qentry_del(&lgr->llc_flow_lcl); + qentry = smc_llc_flow_qentry_clr(&lgr->llc_flow_lcl);

down_write(&lgr->llc_conf_lock); - rc = smc_llc_srv_add_link(link); + rc = smc_llc_srv_add_link(link, qentry); if (!rc && lgr->type == SMC_LGR_SYMMETRIC) { /* delete any asymmetric link */ smc_llc_delete_asym_link(lgr); } up_write(&lgr->llc_conf_lock); + kfree(qentry); }

/* enqueue a local add_link req to trigger a new add_link flow */ @@ -1273,8 +1521,8 @@ void smc_llc_add_link_local(struct smc_link *link) { struct smc_llc_msg_add_link add_llc = {};

- add_llc.hd.length = sizeof(add_llc); - add_llc.hd.common.type = SMC_LLC_ADD_LINK; + add_llc.hd.common.llc_type = SMC_LLC_ADD_LINK; + smc_llc_init_msg_hdr(&add_llc.hd, link->lgr, sizeof(add_llc)); /* no dev and port needed */ smc_llc_enqueue(link, (union smc_llc_msg *)&add_llc); } @@ -1296,7 +1544,8 @@ static void smc_llc_add_link_work(struct work_struct *work) else smc_llc_process_srv_add_link(lgr); out: - smc_llc_flow_stop(lgr, &lgr->llc_flow_lcl); + if (lgr->llc_flow_lcl.type != SMC_LLC_FLOW_REQ_ADD_LINK) + smc_llc_flow_stop(lgr, &lgr->llc_flow_lcl); }

/* enqueue a local del_link msg to trigger a new del_link flow, @@ -1306,8 +1555,8 @@ void smc_llc_srv_delete_link_local(struct smc_link *link, u8 del_link_id) { struct smc_llc_msg_del_link del_llc = {};

- del_llc.hd.length = sizeof(del_llc); - del_llc.hd.common.type = SMC_LLC_DELETE_LINK; + del_llc.hd.common.llc_type = SMC_LLC_DELETE_LINK; + smc_llc_init_msg_hdr(&del_llc.hd, link->lgr, sizeof(del_llc)); del_llc.link_num = del_link_id; del_llc.reason = htonl(SMC_LLC_DEL_LOST_PATH); del_llc.hd.flags |= SMC_LLC_FLAG_DEL_LINK_ORDERLY; @@ -1377,8 +1626,8 @@ void smc_llc_send_link_delete_all(struct smc_link_group *lgr, bool ord, u32 rsn) struct smc_llc_msg_del_link delllc = {}; int i;

- delllc.hd.common.type = SMC_LLC_DELETE_LINK; - delllc.hd.length = sizeof(delllc); + delllc.hd.common.llc_type = SMC_LLC_DELETE_LINK; + smc_llc_init_msg_hdr(&delllc.hd, lgr, sizeof(delllc)); if (ord) delllc.hd.flags |= SMC_LLC_FLAG_DEL_LINK_ORDERLY; delllc.hd.flags |= SMC_LLC_FLAG_DEL_LINK_ALL; @@ -1494,6 +1743,8 @@ static void smc_llc_rmt_conf_rkey(struct smc_link_group *lgr) link = qentry->link;

num_entries = llc->rtoken[0].num_rkeys; + if (num_entries > SMC_LLC_RKEYS_PER_MSG) + goto out_err; /* first rkey entry is for receiving link */ rk_idx = smc_rtoken_add(link, llc->rtoken[0].rmb_vaddr, @@ -1512,6 +1763,7 @@ static void smc_llc_rmt_conf_rkey(struct smc_link_group *lgr) llc->hd.flags |= SMC_LLC_FLAG_RKEY_RETRY; out: llc->hd.flags |= SMC_LLC_FLAG_RESP; + smc_llc_init_msg_hdr(&llc->hd, link->lgr, sizeof(*llc)); smc_llc_send_message(link, &qentry->msg); smc_llc_flow_qentry_del(&lgr->llc_flow_rmt); } @@ -1529,6 +1781,28 @@ static void smc_llc_rmt_delete_rkey(struct smc_link_group *lgr) llc = &qentry->msg.delete_rkey; link = qentry->link;

+ if (lgr->smc_version == SMC_V2) { + struct smc_llc_msg_delete_rkey_v2 *llcv2; + + memcpy(lgr->wr_rx_buf_v2, llc, sizeof(*llc)); + llcv2 = (struct smc_llc_msg_delete_rkey_v2 *)lgr->wr_rx_buf_v2; + llcv2->num_inval_rkeys = 0; + + max = min_t(u8, llcv2->num_rkeys, SMC_LLC_RKEYS_PER_MSG_V2); + for (i = 0; i < max; i++) { + if (smc_rtoken_delete(link, llcv2->rkey[i])) + llcv2->num_inval_rkeys++; + } + memset(&llc->rkey[0], 0, sizeof(llc->rkey)); + memset(&llc->reserved2, 0, sizeof(llc->reserved2)); + smc_llc_init_msg_hdr(&llc->hd, link->lgr, sizeof(*llc)); + if (llcv2->num_inval_rkeys) { + llc->hd.flags |= SMC_LLC_FLAG_RKEY_NEG; + llc->err_mask = llcv2->num_inval_rkeys; + } + goto finish; + } + max = min_t(u8, llc->num_rkeys, SMC_LLC_DEL_RKEY_MAX); for (i = 0; i < max; i++) { if (smc_rtoken_delete(link, llc->rkey[i])) @@ -1538,6 +1812,7 @@ static void smc_llc_rmt_delete_rkey(struct smc_link_group *lgr) llc->hd.flags |= SMC_LLC_FLAG_RKEY_NEG; llc->err_mask = err_mask; } +finish: llc->hd.flags |= SMC_LLC_FLAG_RESP; smc_llc_send_message(link, &qentry->msg); smc_llc_flow_qentry_del(&lgr->llc_flow_rmt); @@ -1573,7 +1848,7 @@ static void smc_llc_event_handler(struct smc_llc_qentry *qentry) if (!smc_link_usable(link)) goto out;

- switch (llc->raw.hdr.common.type) { + switch (llc->raw.hdr.common.llc_type) { case SMC_LLC_TEST_LINK: llc->test_link.hd.flags |= SMC_LLC_FLAG_RESP; smc_llc_send_message(link, llc); @@ -1598,8 +1873,18 @@ static void smc_llc_event_handler(struct smc_llc_qentry *qentry) smc_llc_flow_qentry_set(&lgr->llc_flow_lcl, qentry); wake_up(&lgr->llc_msg_waiter); - } else if (smc_llc_flow_start(&lgr->llc_flow_lcl, - qentry)) { + return; + } + if (lgr->llc_flow_lcl.type == + SMC_LLC_FLOW_REQ_ADD_LINK) { + /* server started add_link processing */ + lgr->llc_flow_lcl.type = SMC_LLC_FLOW_ADD_LINK; + smc_llc_flow_qentry_set(&lgr->llc_flow_lcl, + qentry); + schedule_work(&lgr->llc_add_link_work); + return; + } + if (smc_llc_flow_start(&lgr->llc_flow_lcl, qentry)) { schedule_work(&lgr->llc_add_link_work); } } else if (smc_llc_flow_start(&lgr->llc_flow_lcl, qentry)) { @@ -1647,6 +1932,23 @@ static void smc_llc_event_handler(struct smc_llc_qentry *qentry) smc_llc_flow_stop(lgr, &lgr->llc_flow_rmt); } return; + case SMC_LLC_REQ_ADD_LINK: + /* handle response here, smc_llc_flow_stop() cannot be called + * in tasklet context + */ + if (lgr->role == SMC_CLNT && + lgr->llc_flow_lcl.type == SMC_LLC_FLOW_REQ_ADD_LINK && + (llc->raw.hdr.flags & SMC_LLC_FLAG_RESP)) { + smc_llc_flow_stop(link->lgr, &lgr->llc_flow_lcl); + } else if (lgr->role == SMC_SERV) { + if (smc_llc_flow_start(&lgr->llc_flow_lcl, qentry)) { + /* as smc server, handle client suggestion */ + lgr->llc_flow_lcl.type = SMC_LLC_FLOW_ADD_LINK; + schedule_work(&lgr->llc_add_link_work); + } + return; + } + break; default: smc_llc_protocol_violation(lgr, llc->raw.hdr.common.type); break; @@ -1690,7 +1992,7 @@ static void smc_llc_rx_response(struct smc_link *link, { enum smc_llc_flowtype flowtype = link->lgr->llc_flow_lcl.type; struct smc_llc_flow *flow = &link->lgr->llc_flow_lcl; - u8 llc_type = qentry->msg.raw.hdr.common.type; + u8 llc_type = qentry->msg.raw.hdr.common.llc_type;

switch (llc_type) { case SMC_LLC_TEST_LINK: @@ -1716,7 +2018,8 @@ static void smc_llc_rx_response(struct smc_link *link, /* not used because max links is 3 */ break; default: - smc_llc_protocol_violation(link->lgr, llc_type); + smc_llc_protocol_violation(link->lgr, + qentry->msg.raw.hdr.common.type); break; } kfree(qentry); @@ -1741,7 +2044,8 @@ static void smc_llc_enqueue(struct smc_link *link, union smc_llc_msg *llc) memcpy(&qentry->msg, llc, sizeof(union smc_llc_msg));

/* process responses immediately */ - if (llc->raw.hdr.flags & SMC_LLC_FLAG_RESP) { + if ((llc->raw.hdr.flags & SMC_LLC_FLAG_RESP) && + llc->raw.hdr.common.llc_type != SMC_LLC_REQ_ADD_LINK) { smc_llc_rx_response(link, qentry); return; } @@ -1761,8 +2065,13 @@ static void smc_llc_rx_handler(struct ib_wc *wc, void *buf)

if (wc->byte_len < sizeof(*llc)) return; /* short message */ - if (llc->raw.hdr.length != sizeof(*llc)) - return; /* invalid message */ + if (!llc->raw.hdr.common.llc_version) { + if (llc->raw.hdr.length != sizeof(*llc)) + return; /* invalid message */ + } else { + if (llc->raw.hdr.length_v2 < sizeof(*llc)) + return; /* invalid message */ + }

smc_llc_enqueue(link, llc); } @@ -1981,6 +2290,35 @@ static struct smc_wr_rx_handler smc_llc_rx_handlers[] = { .handler = smc_llc_rx_handler, .type = SMC_LLC_DELETE_RKEY }, + /* V2 types */ + { + .handler = smc_llc_rx_handler, + .type = SMC_LLC_CONFIRM_LINK_V2 + }, + { + .handler = smc_llc_rx_handler, + .type = SMC_LLC_TEST_LINK_V2 + }, + { + .handler = smc_llc_rx_handler, + .type = SMC_LLC_ADD_LINK_V2 + }, + { + .handler = smc_llc_rx_handler, + .type = SMC_LLC_DELETE_LINK_V2 + }, + { + .handler = smc_llc_rx_handler, + .type = SMC_LLC_REQ_ADD_LINK_V2 + }, + { + .handler = smc_llc_rx_handler, + .type = SMC_LLC_CONFIRM_RKEY_V2 + }, + { + .handler = smc_llc_rx_handler, + .type = SMC_LLC_DELETE_RKEY_V2 + }, { .handler = NULL, } diff --git a/net/smc/smc_llc.h b/net/smc/smc_llc.h index cc00a2ec4e92..4404e52b3346 100644 --- a/net/smc/smc_llc.h +++ b/net/smc/smc_llc.h @@ -30,10 +30,19 @@ enum smc_llc_msg_type { SMC_LLC_ADD_LINK = 0x02, SMC_LLC_ADD_LINK_CONT = 0x03, SMC_LLC_DELETE_LINK = 0x04, + SMC_LLC_REQ_ADD_LINK = 0x05, SMC_LLC_CONFIRM_RKEY = 0x06, SMC_LLC_TEST_LINK = 0x07, SMC_LLC_CONFIRM_RKEY_CONT = 0x08, SMC_LLC_DELETE_RKEY = 0x09, + /* V2 types */ + SMC_LLC_CONFIRM_LINK_V2 = 0x21, + SMC_LLC_ADD_LINK_V2 = 0x22, + SMC_LLC_DELETE_LINK_V2 = 0x24, + SMC_LLC_REQ_ADD_LINK_V2 = 0x25, + SMC_LLC_CONFIRM_RKEY_V2 = 0x26, + SMC_LLC_TEST_LINK_V2 = 0x27, + SMC_LLC_DELETE_RKEY_V2 = 0x29, };

#define smc_link_downing(state) \ @@ -102,7 +111,8 @@ void smc_llc_flow_qentry_del(struct smc_llc_flow *flow); void smc_llc_send_link_delete_all(struct smc_link_group *lgr, bool ord, u32 rsn); int smc_llc_cli_add_link(struct smc_link *link, struct smc_llc_qentry *qentry); -int smc_llc_srv_add_link(struct smc_link *link); +int smc_llc_srv_add_link(struct smc_link *link, + struct smc_llc_qentry *req_qentry); void smc_llc_add_link_local(struct smc_link *link); int smc_llc_init(void) __init;

diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c index 8d5a658baff5..f5dbea5b6b79 100644 --- a/net/smc/smc_wr.c +++ b/net/smc/smc_wr.c @@ -244,6 +244,33 @@ int smc_wr_tx_get_free_slot(struct smc_link *link, return 0; }

+int smc_wr_tx_get_v2_slot(struct smc_link *link, + smc_wr_tx_handler handler, + struct smc_wr_v2_buf **wr_buf, + struct smc_wr_tx_pend_priv **wr_pend_priv) +{ + struct smc_wr_tx_pend *wr_pend; + struct ib_send_wr *wr_ib; + u64 wr_id; + + if (link->wr_tx_v2_pend->idx == link->wr_tx_cnt) + return -EBUSY; + + *wr_buf = NULL; + *wr_pend_priv = NULL; + wr_id = smc_wr_tx_get_next_wr_id(link); + wr_pend = link->wr_tx_v2_pend; + wr_pend->wr_id = wr_id; + wr_pend->handler = handler; + wr_pend->link = link; + wr_pend->idx = link->wr_tx_cnt; + wr_ib = link->wr_tx_v2_ib; + wr_ib->wr_id = wr_id; + *wr_buf = link->lgr->wr_tx_buf_v2; + *wr_pend_priv = &wr_pend->priv; + return 0; +} + int smc_wr_tx_put_slot(struct smc_link *link, struct smc_wr_tx_pend_priv *wr_pend_priv) { @@ -293,6 +320,22 @@ int smc_wr_tx_send(struct smc_link *link, struct smc_wr_tx_pend_priv *priv) return rc; }

+int smc_wr_tx_v2_send(struct smc_link *link, struct smc_wr_tx_pend_priv *priv, + int len) +{ + int rc; + + link->wr_tx_v2_ib->sg_list[0].length = len; + ib_req_notify_cq(link->smcibdev->roce_cq_send, + IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS); + rc = ib_post_send(link->roce_qp, link->wr_tx_v2_ib, NULL); + if (rc) { + smc_wr_tx_put_slot(link, priv); + smcr_link_down_cond_sched(link); + } + return rc; +} + /* Send prepared WR slot via ib_post_send and wait for send completion * notification. * @priv: pointer to smc_wr_tx_pend_priv identifying prepared message buffer diff --git a/net/smc/smc_wr.h b/net/smc/smc_wr.h index d8b67318b3c1..abf3be7f8bec 100644 --- a/net/smc/smc_wr.h +++ b/net/smc/smc_wr.h @@ -117,10 +117,16 @@ int smc_wr_tx_get_free_slot(struct smc_link *link, smc_wr_tx_handler handler, struct smc_wr_buf **wr_buf, struct smc_rdma_wr **wrs, struct smc_wr_tx_pend_priv **wr_pend_priv); +int smc_wr_tx_get_v2_slot(struct smc_link *link, + smc_wr_tx_handler handler, + struct smc_wr_v2_buf **wr_buf, + struct smc_wr_tx_pend_priv **wr_pend_priv); int smc_wr_tx_put_slot(struct smc_link *link, struct smc_wr_tx_pend_priv *wr_pend_priv); int smc_wr_tx_send(struct smc_link *link, struct smc_wr_tx_pend_priv *wr_pend_priv); +int smc_wr_tx_v2_send(struct smc_link *link, + struct smc_wr_tx_pend_priv *priv, int len); int smc_wr_tx_send_wait(struct smc_link *link, struct smc_wr_tx_pend_priv *priv, unsigned long timeout); void smc_wr_tx_cq_handler(struct ib_cq *ib_cq, void *cq_context);

From: Karsten Graul kgraul@linux.ibm.com

mainline inclusion from mainline-v5.16-rc1 commit 29397e34c76b57ce596493008176cf374f1cb867 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

With SMC-Rv2 the GID is an IP address that can be deleted from the device. When an IB_EVENT_GID_CHANGE event is provided then iterate over all active links and check if their GID is still defined. Otherwise stop the affected link.

Signed-off-by: Karsten Graul kgraul@linux.ibm.com Signed-off-by: David S. Miller davem@davemloft.net --- net/smc/smc_ib.c | 53 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 53 insertions(+)

diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c index 0125280eb147..9701ec43f0b1 100644 --- a/net/smc/smc_ib.c +++ b/net/smc/smc_ib.c @@ -295,6 +295,58 @@ int smc_ib_determine_gid(struct smc_ib_device *smcibdev, u8 ibport, return -ENODEV; }

+/* check if gid is still defined on smcibdev */ +static bool smc_ib_check_link_gid(u8 gid[SMC_GID_SIZE], bool smcrv2, + struct smc_ib_device *smcibdev, u8 ibport) +{ + const struct ib_gid_attr *attr; + bool rc = false; + int i; + + for (i = 0; !rc && i < smcibdev->pattr[ibport - 1].gid_tbl_len; i++) { + attr = rdma_get_gid_attr(smcibdev->ibdev, ibport, i); + if (IS_ERR(attr)) + continue; + + rcu_read_lock(); + if ((!smcrv2 && attr->gid_type == IB_GID_TYPE_ROCE) || + (smcrv2 && attr->gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP && + !(ipv6_addr_type((const struct in6_addr *)&attr->gid) + & IPV6_ADDR_LINKLOCAL))) + if (!memcmp(gid, &attr->gid, SMC_GID_SIZE)) + rc = true; + rcu_read_unlock(); + rdma_put_gid_attr(attr); + } + return rc; +} + +/* check all links if the gid is still defined on smcibdev */ +static void smc_ib_gid_check(struct smc_ib_device *smcibdev, u8 ibport) +{ + struct smc_link_group *lgr; + int i; + + spin_lock_bh(&smc_lgr_list.lock); + list_for_each_entry(lgr, &smc_lgr_list.list, list) { + if (strncmp(smcibdev->pnetid[ibport - 1], lgr->pnet_id, + SMC_MAX_PNETID_LEN)) + continue; /* lgr is not affected */ + if (list_empty(&lgr->list)) + continue; + for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) { + if (lgr->lnk[i].state == SMC_LNK_UNUSED || + lgr->lnk[i].smcibdev != smcibdev) + continue; + if (!smc_ib_check_link_gid(lgr->lnk[i].gid, + lgr->smc_version == SMC_V2, + smcibdev, ibport)) + smcr_port_err(smcibdev, ibport); + } + } + spin_unlock_bh(&smc_lgr_list.lock); +} + static int smc_ib_remember_port_attr(struct smc_ib_device *smcibdev, u8 ibport) { int rc; @@ -333,6 +385,7 @@ static void smc_ib_port_event_work(struct work_struct *work) } else { clear_bit(port_idx, smcibdev->ports_going_away); smcr_port_add(smcibdev, port_idx + 1); + smc_ib_gid_check(smcibdev, port_idx + 1); } } }

From: liuyacan liuyacan@corp.netease.com

mainline inclusion from mainline-v5.19-rc1 commit 8c3b8dc5cc9bf6d273ebe18b16e2d6882bcfb36d category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

In the process of checking whether RDMAv2 is available, the current implementation first sets ini->smcrv2.ib_dev_v2, and then allocates smc buf desc, but the latter may fail. Unfortunately, the caller will only check the former. In this case, a NULL pointer reference will occur in smc_clc_send_confirm_accept() when accessing conn->rmb_desc.

This patch does two things: 1. Use the return code to determine whether V2 is available. 2. If the return code is NODEV, continue to check whether V1 is available.

Fixes: e49300a6bf62 ("net/smc: add listen processing for SMC-Rv2") Signed-off-by: liuyacan liuyacan@corp.netease.com Signed-off-by: David S. Miller davem@davemloft.net Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- net/smc/af_smc.c | 44 +++++++++++++++++++++++++++----------------- 1 file changed, 27 insertions(+), 17 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index cd362f04aa5f..6359f2acabdf 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -2110,13 +2110,13 @@ static int smc_listen_rdma_reg(struct smc_sock *new_smc, bool local_first) return 0; }

-static void smc_find_rdma_v2_device_serv(struct smc_sock *new_smc, - struct smc_clc_msg_proposal *pclc, - struct smc_init_info *ini) +static int smc_find_rdma_v2_device_serv(struct smc_sock *new_smc, + struct smc_clc_msg_proposal *pclc, + struct smc_init_info *ini) { struct smc_clc_v2_extension *smc_v2_ext; u8 smcr_version; - int rc; + int rc = 0;

if (!(ini->smcr_version & SMC_V2) || !smcr_indicated(ini->smc_type_v2)) goto not_found; @@ -2134,26 +2134,31 @@ static void smc_find_rdma_v2_device_serv(struct smc_sock *new_smc, ini->smcrv2.saddr = new_smc->clcsock->sk->sk_rcv_saddr; ini->smcrv2.daddr = smc_ib_gid_to_ipv4(smc_v2_ext->roce); rc = smc_find_rdma_device(new_smc, ini); - if (rc) { - smc_find_ism_store_rc(rc, ini); + if (rc) goto not_found; - } + if (!ini->smcrv2.uses_gateway) memcpy(ini->smcrv2.nexthop_mac, pclc->lcl.mac, ETH_ALEN);

smcr_version = ini->smcr_version; ini->smcr_version = SMC_V2; rc = smc_listen_rdma_init(new_smc, ini); - if (!rc) - rc = smc_listen_rdma_reg(new_smc, ini->first_contact_local); - if (!rc) - return; - ini->smcr_version = smcr_version; - smc_find_ism_store_rc(rc, ini); + if (rc) { + ini->smcr_version = smcr_version; + goto not_found; + } + rc = smc_listen_rdma_reg(new_smc, ini->first_contact_local); + if (rc) { + ini->smcr_version = smcr_version; + goto not_found; + } + return 0;

not_found: + rc = rc ?: SMC_CLC_DECL_NOSMCDEV; ini->smcr_version &= ~SMC_V2; ini->check_smcrv2 = false; + return rc; }

static int smc_find_rdma_v1_device_serv(struct smc_sock *new_smc, @@ -2186,6 +2191,7 @@ static int smc_listen_find_device(struct smc_sock *new_smc, struct smc_init_info *ini) { int prfx_rc; + int rc;

/* check for ISM device matching V2 proposed device */ smc_find_ism_v2_device_serv(new_smc, pclc, ini); @@ -2213,14 +2219,18 @@ static int smc_listen_find_device(struct smc_sock *new_smc, return ini->rc ?: SMC_CLC_DECL_NOSMCDDEV;

/* check if RDMA V2 is available */ - smc_find_rdma_v2_device_serv(new_smc, pclc, ini); - if (ini->smcrv2.ib_dev_v2) + rc = smc_find_rdma_v2_device_serv(new_smc, pclc, ini); + if (!rc) return 0;

+ /* skip V1 check if V2 is unavailable for non-Device reason */ + if (rc != SMC_CLC_DECL_NOSMCDEV && + rc != SMC_CLC_DECL_NOSMCRDEV && + rc != SMC_CLC_DECL_NOSMCDDEV) + return rc; + /* check if RDMA V1 is available */ if (!prfx_rc) { - int rc; - rc = smc_find_rdma_v1_device_serv(new_smc, pclc, ini); smc_find_ism_store_rc(rc, ini); return (!rc) ? 0 : ini->rc;

From: liuyacan liuyacan@corp.netease.com

mainline inclusion from mainline-v5.19-rc1 commit b3b1a17538d3ef6a9667b2271216fd16d7678ab5 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

In the process of checking whether RDMAv2 is available, the current implementation first sets ini->smcrv2.ib_dev_v2, and then allocates smc buf desc and register rmb, but the latter may fail. In this case, the pointer should be reset.

Fixes: e49300a6bf62 ("net/smc: add listen processing for SMC-Rv2") Signed-off-by: liuyacan liuyacan@corp.netease.com Reviewed-by: Karsten Graul kgraul@linux.ibm.com Link: https://lore.kernel.org/r/20220525085408.812273-1-liuyacan@corp.netease.com Signed-off-by: Jakub Kicinski kuba@kernel.org Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- net/smc/af_smc.c | 1 + 1 file changed, 1 insertion(+)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index cd362f04aa5f..ce8624702407 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -2153,6 +2153,7 @@ static void smc_find_rdma_v2_device_serv(struct smc_sock *new_smc,

not_found: ini->smcr_version &= ~SMC_V2; + ini->smcrv2.ib_dev_v2 = NULL; ini->check_smcrv2 = false; }

From: Wen Gu guwen@linux.alibaba.com

mainline inclusion from mainline-v6.4-rc4 commit 35112271672ae98f45df7875244a4e33aa215e31 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

We found a crash when using SMCRv2 with 2 Mellanox ConnectX-4. It can be reproduced by:

- smc_run nginx - smc_run wrk -t 32 -c 500 -d 30 http://<ip>:<port>

BUG: kernel NULL pointer dereference, address: 0000000000000014 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 8000000108713067 P4D 8000000108713067 PUD 151127067 PMD 0 Oops: 0000 [#1] PREEMPT SMP PTI CPU: 4 PID: 2441 Comm: kworker/4:249 Kdump: loaded Tainted: G W E 6.4.0-rc1+ #42 Workqueue: smc_hs_wq smc_listen_work [smc] RIP: 0010:smc_clc_send_confirm_accept+0x284/0x580 [smc] RSP: 0018:ffffb8294b2d7c78 EFLAGS: 00010a06 RAX: ffff8f1873238880 RBX: ffffb8294b2d7dc8 RCX: 0000000000000000 RDX: 00000000000000b4 RSI: 0000000000000001 RDI: 0000000000b40c00 RBP: ffffb8294b2d7db8 R08: ffff8f1815c5860c R09: 0000000000000000 R10: 0000000000000400 R11: 0000000000000000 R12: ffff8f1846f56180 R13: ffff8f1815c5860c R14: 0000000000000001 R15: 0000000000000001 FS: 0000000000000000(0000) GS:ffff8f1aefd00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000014 CR3: 00000001027a0001 CR4: 00000000003706e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: <TASK> ? mlx5_ib_map_mr_sg+0xa1/0xd0 [mlx5_ib] ? smcr_buf_map_link+0x24b/0x290 [smc] ? __smc_buf_create+0x4ee/0x9b0 [smc] smc_clc_send_accept+0x4c/0xb0 [smc] smc_listen_work+0x346/0x650 [smc] ? __schedule+0x279/0x820 process_one_work+0x1e5/0x3f0 worker_thread+0x4d/0x2f0 ? __pfx_worker_thread+0x10/0x10 kthread+0xe5/0x120 ? __pfx_kthread+0x10/0x10 ret_from_fork+0x2c/0x50 </TASK>

During the CLC handshake, server sequentially tries available SMCRv2 and SMCRv1 devices in smc_listen_work().

If an SMCRv2 device is found. SMCv2 based link group and link will be assigned to the connection. Then assumed that some buffer assignment errors happen later in the CLC handshake, such as RMB registration failure, server will give up SMCRv2 and try SMCRv1 device instead. But the resources assigned to the connection won't be reset.

When server tries SMCRv1 device, the connection creation process will be executed again. Since conn->lnk has been assigned when trying SMCRv2, it will not be set to the correct SMCRv1 link in smcr_lgr_conn_assign_link(). So in such situation, conn->lgr points to correct SMCRv1 link group but conn->lnk points to the SMCRv2 link mistakenly.

Then in smc_clc_send_confirm_accept(), conn->rmb_desc->mr[link->link_idx] will be accessed. Since the link->link_idx is not correct, the related MR may not have been initialized, so crash happens.

| Try SMCRv2 device first | |-> conn->lgr: assign existed SMCRv2 link group; | |-> conn->link: assign existed SMCRv2 link (link_idx may be 1 in SMC_LGR_SYMMETRIC); | |-> sndbuf & RMB creation fails, quit; | | Try SMCRv1 device then | |-> conn->lgr: create SMCRv1 link group and assign; | |-> conn->link: keep SMCRv2 link mistakenly; | |-> sndbuf & RMB creation succeed, only RMB->mr[link_idx = 0] | initialized. | | Then smc_clc_send_confirm_accept() accesses | conn->rmb_desc->mr[conn->link->link_idx, which is 1], then crash. v

This patch tries to fix this by cleaning conn->lnk before assigning link. In addition, it is better to reset the connection and clean the resources assigned if trying SMCRv2 failed in buffer creation or registration.

Fixes: e49300a6bf62 ("net/smc: add listen processing for SMC-Rv2") Link: https://lore.kernel.org/r/20220523055056.2078994-1-liuyacan@corp.netease.com... Signed-off-by: Wen Gu guwen@linux.alibaba.com Reviewed-by: Tony Lu tonylu@linux.alibaba.com Signed-off-by: David S. Miller davem@davemloft.net

Conflicts: net/smc/af_smc.c [The conflict occurs because the commit 8cf3f3e42374("net/smc: use helper smc_conn_abort() in listen processing") is not merged] Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- net/smc/af_smc.c | 9 +++++++-- net/smc/smc_core.c | 1 + 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index ce8624702407..f0f5e7381227 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -1924,8 +1924,10 @@ static int smc_listen_rdma_init(struct smc_sock *new_smc, return rc;

/* create send buffer and rmb */ - if (smc_buf_create(new_smc, false)) + if (smc_buf_create(new_smc, false)) { + smc_connect_abort(new_smc, ini->first_contact_local); return SMC_CLC_DECL_MEM; + }

return 0; } @@ -2144,8 +2146,11 @@ static void smc_find_rdma_v2_device_serv(struct smc_sock *new_smc, smcr_version = ini->smcr_version; ini->smcr_version = SMC_V2; rc = smc_listen_rdma_init(new_smc, ini); - if (!rc) + if (!rc) { rc = smc_listen_rdma_reg(new_smc, ini->first_contact_local); + if (rc) + smc_connect_abort(new_smc, ini->first_contact_local); + } if (!rc) return; ini->smcr_version = smcr_version; diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c index 3a5b6b05f05d..d057ffc294a9 100644 --- a/net/smc/smc_core.c +++ b/net/smc/smc_core.c @@ -127,6 +127,7 @@ static int smcr_lgr_conn_assign_link(struct smc_connection *conn, bool first) int i, j;

/* do link balancing */ + conn->lnk = NULL; /* reset conn->lnk first */ for (i = 0; i < SMC_LINKS_PER_LGR_MAX; i++) { struct smc_link *lnk = &conn->lgr->lnk[i];

From: Wen Gu guwen@linux.alibaba.com

mainline inclusion from mainline-v6.4-rc5 commit 71c6aa0305e3d2365d3bfd0134b4025d9e7ba388 category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

We encountered a crash when using SMCRv2. It is caused by a logical error in smc_llc_fill_ext_v2().

BUG: kernel NULL pointer dereference, address: 0000000000000014 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 0 P4D 0 Oops: 0000 [#1] PREEMPT SMP PTI CPU: 7 PID: 453 Comm: kworker/7:4 Kdump: loaded Tainted: G W E 6.4.0-rc3+ #44 Workqueue: events smc_llc_add_link_work [smc] RIP: 0010:smc_llc_fill_ext_v2+0x117/0x280 [smc] RSP: 0018:ffffacb5c064bd88 EFLAGS: 00010282 RAX: ffff9a6bc1c3c02c RBX: ffff9a6be3558000 RCX: 0000000000000000 RDX: 0000000000000002 RSI: 0000000000000002 RDI: 000000000000000a RBP: ffffacb5c064bdb8 R08: 0000000000000040 R09: 000000000000000c R10: ffff9a6bc0910300 R11: 0000000000000002 R12: 0000000000000000 R13: 0000000000000002 R14: ffff9a6bc1c3c02c R15: ffff9a6be3558250 FS: 0000000000000000(0000) GS:ffff9a6eefdc0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000014 CR3: 000000010b078003 CR4: 00000000003706e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: <TASK> smc_llc_send_add_link+0x1ae/0x2f0 [smc] smc_llc_srv_add_link+0x2c9/0x5a0 [smc] ? cc_mkenc+0x40/0x60 smc_llc_add_link_work+0xb8/0x140 [smc] process_one_work+0x1e5/0x3f0 worker_thread+0x4d/0x2f0 ? __pfx_worker_thread+0x10/0x10 kthread+0xe5/0x120 ? __pfx_kthread+0x10/0x10 ret_from_fork+0x2c/0x50 </TASK>

When an alernate RNIC is available in system, SMC will try to add a new link based on the RNIC for resilience. All the RMBs in use will be mapped to the new link. Then the RMBs' MRs corresponding to the new link will be filled into SMCRv2 LLC ADD LINK messages.

However, smc_llc_fill_ext_v2() mistakenly accesses to unused RMBs which haven't been mapped to the new link and have no valid MRs, thus causing a crash. So this patch fixes the logic.

Fixes: b4ba4652b3f8 ("net/smc: extend LLC layer for SMC-Rv2") Signed-off-by: Wen Gu guwen@linux.alibaba.com Signed-off-by: Paolo Abeni pabeni@redhat.com

Conflicts: net/smc/smc_llc.c [The conflict occurs because the commit b8d199451c99("net/smc: Allow virtually contiguous sndbufs or RMBs for SMC-R") is merged] Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- net/smc/smc_llc.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c index 36611a96a760..984bcc57a00f 100644 --- a/net/smc/smc_llc.c +++ b/net/smc/smc_llc.c @@ -615,16 +615,17 @@ static int smc_llc_fill_ext_v2(struct smc_llc_msg_add_link_v2_ext *ext, goto out; buf_pos = smc_llc_get_first_rmb(lgr, &buf_lst); for (i = 0; i < ext->num_rkeys; i++) { + while (buf_pos && !(buf_pos)->used) + buf_pos = smc_llc_get_next_rmb(lgr, &buf_lst, buf_pos); if (!buf_pos) break; rmb = buf_pos; ext->rt[i].rmb_key = htonl(rmb->mr[prim_lnk_idx]->rkey); ext->rt[i].rmb_key_new = htonl(rmb->mr[lnk_idx]->rkey); - ext->rt[i].rmb_vaddr_new = + ext->rt[i].rmb_vaddr_new = rmb->is_vm ? + cpu_to_be64((uintptr_t)rmb->cpu_addr) : cpu_to_be64((u64)sg_dma_address(rmb->sgt[lnk_idx].sgl)); buf_pos = smc_llc_get_next_rmb(lgr, &buf_lst, buf_pos); - while (buf_pos && !(buf_pos)->used) - buf_pos = smc_llc_get_next_rmb(lgr, &buf_lst, buf_pos); } len += i * sizeof(ext->rt[0]); out:

From: Dust Li dust.li@linux.alibaba.com

mainline inclusion from mainline-v6.6-rc7 commit 4abbd2e3c1db671fa1286390f1310aec78386f1d category: feature bugzilla: https://gitee.com/openeuler/kernel/issues/IAGKJU

Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?i...

-------------------------------------------

In the smc_listen_work(), if smc_listen_prfx_check() failed, the real reason: SMC_CLC_DECL_DIFFPREFIX was dropped, and SMC_CLC_DECL_NOSMCDEV was returned.

Althrough this is also kind of SMC_CLC_DECL_NOSMCDEV, but return the real reason is much friendly for debugging.

Fixes: e49300a6bf62 ("net/smc: add listen processing for SMC-Rv2") Signed-off-by: Dust Li dust.li@linux.alibaba.com Reviewed-by: Alexandra Winter wintera@linux.ibm.com Reviewed-by: Wenjia Zhang wenjia@linux.ibm.com Link: https://lore.kernel.org/r/20231012123729.29307-1-dust.li@linux.alibaba.com Signed-off-by: Jakub Kicinski kuba@kernel.org Signed-off-by: Zhengchao Shao shaozhengchao@huawei.com --- net/smc/af_smc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c index 36aceab08e41..5b45c5799d8e 100644 --- a/net/smc/af_smc.c +++ b/net/smc/af_smc.c @@ -2232,7 +2232,7 @@ static int smc_listen_find_device(struct smc_sock *new_smc, smc_find_ism_store_rc(rc, ini); return (!rc) ? 0 : ini->rc; } - return SMC_CLC_DECL_NOSMCDEV; + return prfx_rc; }

/* listen worker: finish RDMA setup */