diff options
Diffstat (limited to 'fs/smb/client/smbdirect.c')
| -rw-r--r-- | fs/smb/client/smbdirect.c | 2287 |
1 files changed, 1346 insertions, 941 deletions
diff --git a/fs/smb/client/smbdirect.c b/fs/smb/client/smbdirect.c index b0b7254661e9..788a0670c4a8 100644 --- a/fs/smb/client/smbdirect.c +++ b/fs/smb/client/smbdirect.c @@ -7,37 +7,42 @@ #include <linux/module.h> #include <linux/highmem.h> #include <linux/folio_queue.h> +#define __SMBDIRECT_SOCKET_DISCONNECT(__sc) smbd_disconnect_rdma_connection(__sc) +#include "../common/smbdirect/smbdirect_pdu.h" #include "smbdirect.h" #include "cifs_debug.h" #include "cifsproto.h" #include "smb2proto.h" -static struct smbd_response *get_empty_queue_buffer( - struct smbd_connection *info); -static struct smbd_response *get_receive_buffer( - struct smbd_connection *info); +const struct smbdirect_socket_parameters *smbd_get_parameters(struct smbd_connection *conn) +{ + struct smbdirect_socket *sc = &conn->socket; + + return &sc->parameters; +} + +static struct smbdirect_recv_io *get_receive_buffer( + struct smbdirect_socket *sc); static void put_receive_buffer( - struct smbd_connection *info, - struct smbd_response *response); -static int allocate_receive_buffers(struct smbd_connection *info, int num_buf); -static void destroy_receive_buffers(struct smbd_connection *info); + struct smbdirect_socket *sc, + struct smbdirect_recv_io *response); +static int allocate_receive_buffers(struct smbdirect_socket *sc, int num_buf); +static void destroy_receive_buffers(struct smbdirect_socket *sc); -static void put_empty_packet( - struct smbd_connection *info, struct smbd_response *response); static void enqueue_reassembly( - struct smbd_connection *info, - struct smbd_response *response, int data_length); -static struct smbd_response *_get_first_reassembly( - struct smbd_connection *info); + struct smbdirect_socket *sc, + struct smbdirect_recv_io *response, int data_length); +static struct smbdirect_recv_io *_get_first_reassembly( + struct smbdirect_socket *sc); static int smbd_post_recv( - struct smbd_connection *info, - struct smbd_response *response); + struct smbdirect_socket *sc, + struct smbdirect_recv_io *response); -static int smbd_post_send_empty(struct smbd_connection *info); +static int smbd_post_send_empty(struct smbdirect_socket *sc); -static void destroy_mr_list(struct smbd_connection *info); -static int allocate_mr_list(struct smbd_connection *info); +static void destroy_mr_list(struct smbdirect_socket *sc); +static int allocate_mr_list(struct smbdirect_socket *sc); struct smb_extract_to_rdma { struct ib_sge *sge; @@ -50,9 +55,6 @@ struct smb_extract_to_rdma { static ssize_t smb_extract_iter_to_rdma(struct iov_iter *iter, size_t len, struct smb_extract_to_rdma *rdma); -/* SMBD version number */ -#define SMBD_V1 0x0100 - /* Port numbers for SMBD transport */ #define SMB_PORT 445 #define SMBD_PORT 5445 @@ -63,6 +65,9 @@ static ssize_t smb_extract_iter_to_rdma(struct iov_iter *iter, size_t len, /* SMBD negotiation timeout in seconds */ #define SMBD_NEGOTIATE_TIMEOUT 120 +/* The timeout to wait for a keepalive message from peer in seconds */ +#define KEEPALIVE_RECV_TIMEOUT 5 + /* SMBD minimum receive size and fragmented sized defined in [MS-SMBD] */ #define SMBD_MIN_RECEIVE_SIZE 128 #define SMBD_MIN_FRAGMENTED_SIZE 131072 @@ -161,78 +166,297 @@ do { \ #define log_rdma_mr(level, fmt, args...) \ log_rdma(level, LOG_RDMA_MR, fmt, ##args) +static void smbd_disconnect_wake_up_all(struct smbdirect_socket *sc) +{ + /* + * Wake up all waiters in all wait queues + * in order to notice the broken connection. + */ + wake_up_all(&sc->status_wait); + wake_up_all(&sc->send_io.lcredits.wait_queue); + wake_up_all(&sc->send_io.credits.wait_queue); + wake_up_all(&sc->send_io.pending.dec_wait_queue); + wake_up_all(&sc->send_io.pending.zero_wait_queue); + wake_up_all(&sc->recv_io.reassembly.wait_queue); + wake_up_all(&sc->mr_io.ready.wait_queue); + wake_up_all(&sc->mr_io.cleanup.wait_queue); +} + static void smbd_disconnect_rdma_work(struct work_struct *work) { - struct smbd_connection *info = - container_of(work, struct smbd_connection, disconnect_work); + struct smbdirect_socket *sc = + container_of(work, struct smbdirect_socket, disconnect_work); + + if (sc->first_error == 0) + sc->first_error = -ECONNABORTED; - if (info->transport_status == SMBD_CONNECTED) { - info->transport_status = SMBD_DISCONNECTING; - rdma_disconnect(info->id); + /* + * make sure this and other work is not queued again + * but here we don't block and avoid + * disable[_delayed]_work_sync() + */ + disable_work(&sc->disconnect_work); + disable_work(&sc->recv_io.posted.refill_work); + disable_work(&sc->mr_io.recovery_work); + disable_work(&sc->idle.immediate_work); + disable_delayed_work(&sc->idle.timer_work); + + switch (sc->status) { + case SMBDIRECT_SOCKET_NEGOTIATE_NEEDED: + case SMBDIRECT_SOCKET_NEGOTIATE_RUNNING: + case SMBDIRECT_SOCKET_NEGOTIATE_FAILED: + case SMBDIRECT_SOCKET_CONNECTED: + case SMBDIRECT_SOCKET_ERROR: + sc->status = SMBDIRECT_SOCKET_DISCONNECTING; + rdma_disconnect(sc->rdma.cm_id); + break; + + case SMBDIRECT_SOCKET_CREATED: + case SMBDIRECT_SOCKET_RESOLVE_ADDR_NEEDED: + case SMBDIRECT_SOCKET_RESOLVE_ADDR_RUNNING: + case SMBDIRECT_SOCKET_RESOLVE_ADDR_FAILED: + case SMBDIRECT_SOCKET_RESOLVE_ROUTE_NEEDED: + case SMBDIRECT_SOCKET_RESOLVE_ROUTE_RUNNING: + case SMBDIRECT_SOCKET_RESOLVE_ROUTE_FAILED: + case SMBDIRECT_SOCKET_RDMA_CONNECT_NEEDED: + case SMBDIRECT_SOCKET_RDMA_CONNECT_RUNNING: + case SMBDIRECT_SOCKET_RDMA_CONNECT_FAILED: + /* + * rdma_connect() never reached + * RDMA_CM_EVENT_ESTABLISHED + */ + sc->status = SMBDIRECT_SOCKET_DISCONNECTED; + break; + + case SMBDIRECT_SOCKET_DISCONNECTING: + case SMBDIRECT_SOCKET_DISCONNECTED: + case SMBDIRECT_SOCKET_DESTROYED: + break; } + + /* + * Wake up all waiters in all wait queues + * in order to notice the broken connection. + */ + smbd_disconnect_wake_up_all(sc); } -static void smbd_disconnect_rdma_connection(struct smbd_connection *info) +static void smbd_disconnect_rdma_connection(struct smbdirect_socket *sc) { - queue_work(info->workqueue, &info->disconnect_work); + if (sc->first_error == 0) + sc->first_error = -ECONNABORTED; + + /* + * make sure other work (than disconnect_work) is + * not queued again but here we don't block and avoid + * disable[_delayed]_work_sync() + */ + disable_work(&sc->recv_io.posted.refill_work); + disable_work(&sc->mr_io.recovery_work); + disable_work(&sc->idle.immediate_work); + disable_delayed_work(&sc->idle.timer_work); + + switch (sc->status) { + case SMBDIRECT_SOCKET_RESOLVE_ADDR_FAILED: + case SMBDIRECT_SOCKET_RESOLVE_ROUTE_FAILED: + case SMBDIRECT_SOCKET_RDMA_CONNECT_FAILED: + case SMBDIRECT_SOCKET_NEGOTIATE_FAILED: + case SMBDIRECT_SOCKET_ERROR: + case SMBDIRECT_SOCKET_DISCONNECTING: + case SMBDIRECT_SOCKET_DISCONNECTED: + case SMBDIRECT_SOCKET_DESTROYED: + /* + * Keep the current error status + */ + break; + + case SMBDIRECT_SOCKET_RESOLVE_ADDR_NEEDED: + case SMBDIRECT_SOCKET_RESOLVE_ADDR_RUNNING: + sc->status = SMBDIRECT_SOCKET_RESOLVE_ADDR_FAILED; + break; + + case SMBDIRECT_SOCKET_RESOLVE_ROUTE_NEEDED: + case SMBDIRECT_SOCKET_RESOLVE_ROUTE_RUNNING: + sc->status = SMBDIRECT_SOCKET_RESOLVE_ROUTE_FAILED; + break; + + case SMBDIRECT_SOCKET_RDMA_CONNECT_NEEDED: + case SMBDIRECT_SOCKET_RDMA_CONNECT_RUNNING: + sc->status = SMBDIRECT_SOCKET_RDMA_CONNECT_FAILED; + break; + + case SMBDIRECT_SOCKET_NEGOTIATE_NEEDED: + case SMBDIRECT_SOCKET_NEGOTIATE_RUNNING: + sc->status = SMBDIRECT_SOCKET_NEGOTIATE_FAILED; + break; + + case SMBDIRECT_SOCKET_CREATED: + sc->status = SMBDIRECT_SOCKET_DISCONNECTED; + break; + + case SMBDIRECT_SOCKET_CONNECTED: + sc->status = SMBDIRECT_SOCKET_ERROR; + break; + } + + /* + * Wake up all waiters in all wait queues + * in order to notice the broken connection. + */ + smbd_disconnect_wake_up_all(sc); + + queue_work(sc->workqueue, &sc->disconnect_work); } /* Upcall from RDMA CM */ static int smbd_conn_upcall( struct rdma_cm_id *id, struct rdma_cm_event *event) { - struct smbd_connection *info = id->context; + struct smbdirect_socket *sc = id->context; + struct smbdirect_socket_parameters *sp = &sc->parameters; + const char *event_name = rdma_event_msg(event->event); + u8 peer_initiator_depth; + u8 peer_responder_resources; - log_rdma_event(INFO, "event=%d status=%d\n", - event->event, event->status); + log_rdma_event(INFO, "event=%s status=%d\n", + event_name, event->status); switch (event->event) { case RDMA_CM_EVENT_ADDR_RESOLVED: + if (SMBDIRECT_CHECK_STATUS_DISCONNECT(sc, SMBDIRECT_SOCKET_RESOLVE_ADDR_RUNNING)) + break; + sc->status = SMBDIRECT_SOCKET_RESOLVE_ROUTE_NEEDED; + wake_up(&sc->status_wait); + break; + case RDMA_CM_EVENT_ROUTE_RESOLVED: - info->ri_rc = 0; - complete(&info->ri_done); + if (SMBDIRECT_CHECK_STATUS_DISCONNECT(sc, SMBDIRECT_SOCKET_RESOLVE_ROUTE_RUNNING)) + break; + sc->status = SMBDIRECT_SOCKET_RDMA_CONNECT_NEEDED; + wake_up(&sc->status_wait); break; case RDMA_CM_EVENT_ADDR_ERROR: - info->ri_rc = -EHOSTUNREACH; - complete(&info->ri_done); + log_rdma_event(ERR, "connecting failed event=%s\n", event_name); + sc->status = SMBDIRECT_SOCKET_RESOLVE_ADDR_FAILED; + smbd_disconnect_rdma_work(&sc->disconnect_work); break; case RDMA_CM_EVENT_ROUTE_ERROR: - info->ri_rc = -ENETUNREACH; - complete(&info->ri_done); + log_rdma_event(ERR, "connecting failed event=%s\n", event_name); + sc->status = SMBDIRECT_SOCKET_RESOLVE_ROUTE_FAILED; + smbd_disconnect_rdma_work(&sc->disconnect_work); break; case RDMA_CM_EVENT_ESTABLISHED: - log_rdma_event(INFO, "connected event=%d\n", event->event); - info->transport_status = SMBD_CONNECTED; - wake_up_interruptible(&info->conn_wait); + log_rdma_event(INFO, "connected event=%s\n", event_name); + + /* + * Here we work around an inconsistency between + * iWarp and other devices (at least rxe and irdma using RoCEv2) + */ + if (rdma_protocol_iwarp(id->device, id->port_num)) { + /* + * iWarp devices report the peer's values + * with the perspective of the peer here. + * Tested with siw and irdma (in iwarp mode) + * We need to change to our perspective here, + * so we need to switch the values. + */ + peer_initiator_depth = event->param.conn.responder_resources; + peer_responder_resources = event->param.conn.initiator_depth; + } else { + /* + * Non iWarp devices report the peer's values + * already changed to our perspective here. + * Tested with rxe and irdma (in roce mode). + */ + peer_initiator_depth = event->param.conn.initiator_depth; + peer_responder_resources = event->param.conn.responder_resources; + } + if (rdma_protocol_iwarp(id->device, id->port_num) && + event->param.conn.private_data_len == 8) { + /* + * Legacy clients with only iWarp MPA v1 support + * need a private blob in order to negotiate + * the IRD/ORD values. + */ + const __be32 *ird_ord_hdr = event->param.conn.private_data; + u32 ird32 = be32_to_cpu(ird_ord_hdr[0]); + u32 ord32 = be32_to_cpu(ird_ord_hdr[1]); + + /* + * cifs.ko sends the legacy IRD/ORD negotiation + * event if iWarp MPA v2 was used. + * + * Here we check that the values match and only + * mark the client as legacy if they don't match. + */ + if ((u32)event->param.conn.initiator_depth != ird32 || + (u32)event->param.conn.responder_resources != ord32) { + /* + * There are broken clients (old cifs.ko) + * using little endian and also + * struct rdma_conn_param only uses u8 + * for initiator_depth and responder_resources, + * so we truncate the value to U8_MAX. + * + * smb_direct_accept_client() will then + * do the real negotiation in order to + * select the minimum between client and + * server. + */ + ird32 = min_t(u32, ird32, U8_MAX); + ord32 = min_t(u32, ord32, U8_MAX); + + sc->rdma.legacy_iwarp = true; + peer_initiator_depth = (u8)ird32; + peer_responder_resources = (u8)ord32; + } + } + + /* + * negotiate the value by using the minimum + * between client and server if the client provided + * non 0 values. + */ + if (peer_initiator_depth != 0) + sp->initiator_depth = + min_t(u8, sp->initiator_depth, + peer_initiator_depth); + if (peer_responder_resources != 0) + sp->responder_resources = + min_t(u8, sp->responder_resources, + peer_responder_resources); + + if (SMBDIRECT_CHECK_STATUS_DISCONNECT(sc, SMBDIRECT_SOCKET_RDMA_CONNECT_RUNNING)) + break; + sc->status = SMBDIRECT_SOCKET_NEGOTIATE_NEEDED; + wake_up(&sc->status_wait); break; case RDMA_CM_EVENT_CONNECT_ERROR: case RDMA_CM_EVENT_UNREACHABLE: case RDMA_CM_EVENT_REJECTED: - log_rdma_event(INFO, "connecting failed event=%d\n", event->event); - info->transport_status = SMBD_DISCONNECTED; - wake_up_interruptible(&info->conn_wait); + log_rdma_event(ERR, "connecting failed event=%s\n", event_name); + sc->status = SMBDIRECT_SOCKET_RDMA_CONNECT_FAILED; + smbd_disconnect_rdma_work(&sc->disconnect_work); break; case RDMA_CM_EVENT_DEVICE_REMOVAL: case RDMA_CM_EVENT_DISCONNECTED: /* This happens when we fail the negotiation */ - if (info->transport_status == SMBD_NEGOTIATE_FAILED) { - info->transport_status = SMBD_DISCONNECTED; - wake_up(&info->conn_wait); - break; + if (sc->status == SMBDIRECT_SOCKET_NEGOTIATE_FAILED) { + log_rdma_event(ERR, "event=%s during negotiation\n", event_name); } - info->transport_status = SMBD_DISCONNECTED; - wake_up_interruptible(&info->disconn_wait); - wake_up_interruptible(&info->wait_reassembly_queue); - wake_up_interruptible_all(&info->wait_send_queue); + sc->status = SMBDIRECT_SOCKET_DISCONNECTED; + smbd_disconnect_rdma_work(&sc->disconnect_work); break; default: + log_rdma_event(ERR, "unexpected event=%s status=%d\n", + event_name, event->status); break; } @@ -243,15 +467,15 @@ static int smbd_conn_upcall( static void smbd_qp_async_error_upcall(struct ib_event *event, void *context) { - struct smbd_connection *info = context; + struct smbdirect_socket *sc = context; - log_rdma_event(ERR, "%s on device %s info %p\n", - ib_event_msg(event->event), event->device->name, info); + log_rdma_event(ERR, "%s on device %s socket %p\n", + ib_event_msg(event->event), event->device->name, sc); switch (event->event) { case IB_EVENT_CQ_ERR: case IB_EVENT_QP_FATAL: - smbd_disconnect_rdma_connection(info); + smbd_disconnect_rdma_connection(sc); break; default: @@ -259,12 +483,12 @@ smbd_qp_async_error_upcall(struct ib_event *event, void *context) } } -static inline void *smbd_request_payload(struct smbd_request *request) +static inline void *smbdirect_send_io_payload(struct smbdirect_send_io *request) { return (void *)request->packet; } -static inline void *smbd_response_payload(struct smbd_response *response) +static inline void *smbdirect_recv_io_payload(struct smbdirect_recv_io *response) { return (void *)response->packet; } @@ -273,33 +497,40 @@ static inline void *smbd_response_payload(struct smbd_response *response) static void send_done(struct ib_cq *cq, struct ib_wc *wc) { int i; - struct smbd_request *request = - container_of(wc->wr_cqe, struct smbd_request, cqe); + struct smbdirect_send_io *request = + container_of(wc->wr_cqe, struct smbdirect_send_io, cqe); + struct smbdirect_socket *sc = request->socket; + int lcredits = 0; - log_rdma_send(INFO, "smbd_request 0x%p completed wc->status=%d\n", - request, wc->status); - - if (wc->status != IB_WC_SUCCESS || wc->opcode != IB_WC_SEND) { - log_rdma_send(ERR, "wc->status=%d wc->opcode=%d\n", - wc->status, wc->opcode); - smbd_disconnect_rdma_connection(request->info); - } + log_rdma_send(INFO, "smbdirect_send_io 0x%p completed wc->status=%s\n", + request, ib_wc_status_msg(wc->status)); for (i = 0; i < request->num_sge; i++) - ib_dma_unmap_single(request->info->id->device, + ib_dma_unmap_single(sc->ib.dev, request->sge[i].addr, request->sge[i].length, DMA_TO_DEVICE); + mempool_free(request, sc->send_io.mem.pool); + lcredits += 1; - if (atomic_dec_and_test(&request->info->send_pending)) - wake_up(&request->info->wait_send_pending); + if (wc->status != IB_WC_SUCCESS || wc->opcode != IB_WC_SEND) { + if (wc->status != IB_WC_WR_FLUSH_ERR) + log_rdma_send(ERR, "wc->status=%s wc->opcode=%d\n", + ib_wc_status_msg(wc->status), wc->opcode); + smbd_disconnect_rdma_connection(sc); + return; + } - wake_up(&request->info->wait_post_send); + atomic_add(lcredits, &sc->send_io.lcredits.count); + wake_up(&sc->send_io.lcredits.wait_queue); - mempool_free(request, request->info->request_mempool); + if (atomic_dec_and_test(&sc->send_io.pending.count)) + wake_up(&sc->send_io.pending.zero_wait_queue); + + wake_up(&sc->send_io.pending.dec_wait_queue); } -static void dump_smbd_negotiate_resp(struct smbd_negotiate_resp *resp) +static void dump_smbdirect_negotiate_resp(struct smbdirect_negotiate_resp *resp) { log_rdma_event(INFO, "resp message min_version %u max_version %u negotiated_version %u credits_requested %u credits_granted %u status %u max_readwrite_size %u preferred_send_size %u max_receive_size %u max_fragmented_size %u\n", resp->min_version, resp->max_version, @@ -315,52 +546,52 @@ static void dump_smbd_negotiate_resp(struct smbd_negotiate_resp *resp) * return value: true if negotiation is a success, false if failed */ static bool process_negotiation_response( - struct smbd_response *response, int packet_length) + struct smbdirect_recv_io *response, int packet_length) { - struct smbd_connection *info = response->info; - struct smbd_negotiate_resp *packet = smbd_response_payload(response); + struct smbdirect_socket *sc = response->socket; + struct smbdirect_socket_parameters *sp = &sc->parameters; + struct smbdirect_negotiate_resp *packet = smbdirect_recv_io_payload(response); - if (packet_length < sizeof(struct smbd_negotiate_resp)) { + if (packet_length < sizeof(struct smbdirect_negotiate_resp)) { log_rdma_event(ERR, "error: packet_length=%d\n", packet_length); return false; } - if (le16_to_cpu(packet->negotiated_version) != SMBD_V1) { + if (le16_to_cpu(packet->negotiated_version) != SMBDIRECT_V1) { log_rdma_event(ERR, "error: negotiated_version=%x\n", le16_to_cpu(packet->negotiated_version)); return false; } - info->protocol = le16_to_cpu(packet->negotiated_version); if (packet->credits_requested == 0) { log_rdma_event(ERR, "error: credits_requested==0\n"); return false; } - info->receive_credit_target = le16_to_cpu(packet->credits_requested); + sc->recv_io.credits.target = le16_to_cpu(packet->credits_requested); + sc->recv_io.credits.target = min_t(u16, sc->recv_io.credits.target, sp->recv_credit_max); if (packet->credits_granted == 0) { log_rdma_event(ERR, "error: credits_granted==0\n"); return false; } - atomic_set(&info->send_credits, le16_to_cpu(packet->credits_granted)); - - atomic_set(&info->receive_credits, 0); + atomic_set(&sc->send_io.lcredits.count, sp->send_credit_target); + atomic_set(&sc->send_io.credits.count, le16_to_cpu(packet->credits_granted)); - if (le32_to_cpu(packet->preferred_send_size) > info->max_receive_size) { + if (le32_to_cpu(packet->preferred_send_size) > sp->max_recv_size) { log_rdma_event(ERR, "error: preferred_send_size=%d\n", le32_to_cpu(packet->preferred_send_size)); return false; } - info->max_receive_size = le32_to_cpu(packet->preferred_send_size); + sp->max_recv_size = le32_to_cpu(packet->preferred_send_size); if (le32_to_cpu(packet->max_receive_size) < SMBD_MIN_RECEIVE_SIZE) { log_rdma_event(ERR, "error: max_receive_size=%d\n", le32_to_cpu(packet->max_receive_size)); return false; } - info->max_send_size = min_t(int, info->max_send_size, - le32_to_cpu(packet->max_receive_size)); + sp->max_send_size = min_t(u32, sp->max_send_size, + le32_to_cpu(packet->max_receive_size)); if (le32_to_cpu(packet->max_fragmented_size) < SMBD_MIN_FRAGMENTED_SIZE) { @@ -368,100 +599,82 @@ static bool process_negotiation_response( le32_to_cpu(packet->max_fragmented_size)); return false; } - info->max_fragmented_send_size = + sp->max_fragmented_send_size = le32_to_cpu(packet->max_fragmented_size); - info->rdma_readwrite_threshold = - rdma_readwrite_threshold > info->max_fragmented_send_size ? - info->max_fragmented_send_size : - rdma_readwrite_threshold; - info->max_readwrite_size = min_t(u32, + sp->max_read_write_size = min_t(u32, le32_to_cpu(packet->max_readwrite_size), - info->max_frmr_depth * PAGE_SIZE); - info->max_frmr_depth = info->max_readwrite_size / PAGE_SIZE; + sp->max_frmr_depth * PAGE_SIZE); + sp->max_frmr_depth = sp->max_read_write_size / PAGE_SIZE; + sc->recv_io.expected = SMBDIRECT_EXPECT_DATA_TRANSFER; return true; } static void smbd_post_send_credits(struct work_struct *work) { - int ret = 0; - int use_receive_queue = 1; int rc; - struct smbd_response *response; - struct smbd_connection *info = - container_of(work, struct smbd_connection, - post_send_credits_work); + struct smbdirect_recv_io *response; + struct smbdirect_socket *sc = + container_of(work, struct smbdirect_socket, recv_io.posted.refill_work); - if (info->transport_status != SMBD_CONNECTED) { - wake_up(&info->wait_receive_queues); + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) { return; } - if (info->receive_credit_target > - atomic_read(&info->receive_credits)) { + if (sc->recv_io.credits.target > + atomic_read(&sc->recv_io.credits.count)) { while (true) { - if (use_receive_queue) - response = get_receive_buffer(info); - else - response = get_empty_queue_buffer(info); - if (!response) { - /* now switch to empty packet queue */ - if (use_receive_queue) { - use_receive_queue = 0; - continue; - } else - break; - } + response = get_receive_buffer(sc); + if (!response) + break; - response->type = SMBD_TRANSFER_DATA; response->first_segment = false; - rc = smbd_post_recv(info, response); + rc = smbd_post_recv(sc, response); if (rc) { log_rdma_recv(ERR, "post_recv failed rc=%d\n", rc); - put_receive_buffer(info, response); + put_receive_buffer(sc, response); break; } - ret++; + atomic_inc(&sc->recv_io.posted.count); } } - spin_lock(&info->lock_new_credits_offered); - info->new_credits_offered += ret; - spin_unlock(&info->lock_new_credits_offered); - /* Promptly send an immediate packet as defined in [MS-SMBD] 3.1.1.1 */ - info->send_immediate = true; - if (atomic_read(&info->receive_credits) < - info->receive_credit_target - 1) { - if (info->keep_alive_requested == KEEP_ALIVE_PENDING || - info->send_immediate) { - log_keep_alive(INFO, "send an empty message\n"); - smbd_post_send_empty(info); - } + if (atomic_read(&sc->recv_io.credits.count) < + sc->recv_io.credits.target - 1) { + log_keep_alive(INFO, "schedule send of an empty message\n"); + queue_work(sc->workqueue, &sc->idle.immediate_work); } } /* Called from softirq, when recv is done */ static void recv_done(struct ib_cq *cq, struct ib_wc *wc) { - struct smbd_data_transfer *data_transfer; - struct smbd_response *response = - container_of(wc->wr_cqe, struct smbd_response, cqe); - struct smbd_connection *info = response->info; - int data_length = 0; - - log_rdma_recv(INFO, "response=0x%p type=%d wc status=%d wc opcode %d byte_len=%d pkey_index=%u\n", - response, response->type, wc->status, wc->opcode, + struct smbdirect_data_transfer *data_transfer; + struct smbdirect_recv_io *response = + container_of(wc->wr_cqe, struct smbdirect_recv_io, cqe); + struct smbdirect_socket *sc = response->socket; + struct smbdirect_socket_parameters *sp = &sc->parameters; + u16 old_recv_credit_target; + u32 data_offset = 0; + u32 data_length = 0; + u32 remaining_data_length = 0; + bool negotiate_done = false; + + log_rdma_recv(INFO, + "response=0x%p type=%d wc status=%s wc opcode %d byte_len=%d pkey_index=%u\n", + response, sc->recv_io.expected, + ib_wc_status_msg(wc->status), wc->opcode, wc->byte_len, wc->pkey_index); if (wc->status != IB_WC_SUCCESS || wc->opcode != IB_WC_RECV) { - log_rdma_recv(INFO, "wc->status=%d opcode=%d\n", - wc->status, wc->opcode); - smbd_disconnect_rdma_connection(info); + if (wc->status != IB_WC_WR_FLUSH_ERR) + log_rdma_recv(ERR, "wc->status=%s opcode=%d\n", + ib_wc_status_msg(wc->status), wc->opcode); goto error; } @@ -471,55 +684,81 @@ static void recv_done(struct ib_cq *cq, struct ib_wc *wc) response->sge.length, DMA_FROM_DEVICE); - switch (response->type) { + /* + * Reset timer to the keepalive interval in + * order to trigger our next keepalive message. + */ + sc->idle.keepalive = SMBDIRECT_KEEPALIVE_NONE; + mod_delayed_work(sc->workqueue, &sc->idle.timer_work, + msecs_to_jiffies(sp->keepalive_interval_msec)); + + switch (sc->recv_io.expected) { /* SMBD negotiation response */ - case SMBD_NEGOTIATE_RESP: - dump_smbd_negotiate_resp(smbd_response_payload(response)); - info->full_packet_received = true; - info->negotiate_done = + case SMBDIRECT_EXPECT_NEGOTIATE_REP: + dump_smbdirect_negotiate_resp(smbdirect_recv_io_payload(response)); + sc->recv_io.reassembly.full_packet_received = true; + negotiate_done = process_negotiation_response(response, wc->byte_len); - complete(&info->negotiate_completion); - break; + put_receive_buffer(sc, response); + if (SMBDIRECT_CHECK_STATUS_WARN(sc, SMBDIRECT_SOCKET_NEGOTIATE_RUNNING)) + negotiate_done = false; + if (!negotiate_done) { + sc->status = SMBDIRECT_SOCKET_NEGOTIATE_FAILED; + smbd_disconnect_rdma_connection(sc); + } else { + sc->status = SMBDIRECT_SOCKET_CONNECTED; + wake_up(&sc->status_wait); + } + + return; /* SMBD data transfer packet */ - case SMBD_TRANSFER_DATA: - data_transfer = smbd_response_payload(response); + case SMBDIRECT_EXPECT_DATA_TRANSFER: + data_transfer = smbdirect_recv_io_payload(response); + + if (wc->byte_len < + offsetof(struct smbdirect_data_transfer, padding)) + goto error; + + remaining_data_length = le32_to_cpu(data_transfer->remaining_data_length); + data_offset = le32_to_cpu(data_transfer->data_offset); data_length = le32_to_cpu(data_transfer->data_length); + if (wc->byte_len < data_offset || + (u64)wc->byte_len < (u64)data_offset + data_length) + goto error; + + if (remaining_data_length > sp->max_fragmented_recv_size || + data_length > sp->max_fragmented_recv_size || + (u64)remaining_data_length + (u64)data_length > (u64)sp->max_fragmented_recv_size) + goto error; - /* - * If this is a packet with data playload place the data in - * reassembly queue and wake up the reading thread - */ if (data_length) { - if (info->full_packet_received) + if (sc->recv_io.reassembly.full_packet_received) response->first_segment = true; if (le32_to_cpu(data_transfer->remaining_data_length)) - info->full_packet_received = false; + sc->recv_io.reassembly.full_packet_received = false; else - info->full_packet_received = true; - - enqueue_reassembly( - info, - response, - data_length); - } else - put_empty_packet(info, response); - - if (data_length) - wake_up_interruptible(&info->wait_reassembly_queue); + sc->recv_io.reassembly.full_packet_received = true; + } - atomic_dec(&info->receive_credits); - info->receive_credit_target = + atomic_dec(&sc->recv_io.posted.count); + atomic_dec(&sc->recv_io.credits.count); + old_recv_credit_target = sc->recv_io.credits.target; + sc->recv_io.credits.target = le16_to_cpu(data_transfer->credits_requested); + sc->recv_io.credits.target = + min_t(u16, sc->recv_io.credits.target, sp->recv_credit_max); + sc->recv_io.credits.target = + max_t(u16, sc->recv_io.credits.target, 1); if (le16_to_cpu(data_transfer->credits_granted)) { atomic_add(le16_to_cpu(data_transfer->credits_granted), - &info->send_credits); + &sc->send_io.credits.count); /* * We have new send credits granted from remote peer * If any sender is waiting for credits, unblock it */ - wake_up_interruptible(&info->wait_send_queue); + wake_up(&sc->send_io.credits.wait_queue); } log_incoming(INFO, "data flags %d data_offset %d data_length %d remaining_data_length %d\n", @@ -528,33 +767,53 @@ static void recv_done(struct ib_cq *cq, struct ib_wc *wc) le32_to_cpu(data_transfer->data_length), le32_to_cpu(data_transfer->remaining_data_length)); - /* Send a KEEP_ALIVE response right away if requested */ - info->keep_alive_requested = KEEP_ALIVE_NONE; + /* Send an immediate response right away if requested */ if (le16_to_cpu(data_transfer->flags) & - SMB_DIRECT_RESPONSE_REQUESTED) { - info->keep_alive_requested = KEEP_ALIVE_PENDING; + SMBDIRECT_FLAG_RESPONSE_REQUESTED) { + log_keep_alive(INFO, "schedule send of immediate response\n"); + queue_work(sc->workqueue, &sc->idle.immediate_work); } + /* + * If this is a packet with data playload place the data in + * reassembly queue and wake up the reading thread + */ + if (data_length) { + if (sc->recv_io.credits.target > old_recv_credit_target) + queue_work(sc->workqueue, &sc->recv_io.posted.refill_work); + + enqueue_reassembly(sc, response, data_length); + wake_up(&sc->recv_io.reassembly.wait_queue); + } else + put_receive_buffer(sc, response); + return; - default: - log_rdma_recv(ERR, - "unexpected response type=%d\n", response->type); + case SMBDIRECT_EXPECT_NEGOTIATE_REQ: + /* Only server... */ + break; } + /* + * This is an internal error! + */ + log_rdma_recv(ERR, "unexpected response type=%d\n", sc->recv_io.expected); + WARN_ON_ONCE(sc->recv_io.expected != SMBDIRECT_EXPECT_DATA_TRANSFER); error: - put_receive_buffer(info, response); + put_receive_buffer(sc, response); + smbd_disconnect_rdma_connection(sc); } static struct rdma_cm_id *smbd_create_id( - struct smbd_connection *info, + struct smbdirect_socket *sc, struct sockaddr *dstaddr, int port) { + struct smbdirect_socket_parameters *sp = &sc->parameters; struct rdma_cm_id *id; int rc; __be16 *sport; - id = rdma_create_id(&init_net, smbd_conn_upcall, info, + id = rdma_create_id(&init_net, smbd_conn_upcall, sc, RDMA_PS_TCP, IB_QPT_RC); if (IS_ERR(id)) { rc = PTR_ERR(id); @@ -569,43 +828,57 @@ static struct rdma_cm_id *smbd_create_id( *sport = htons(port); - init_completion(&info->ri_done); - info->ri_rc = -ETIMEDOUT; - + WARN_ON_ONCE(sc->status != SMBDIRECT_SOCKET_RESOLVE_ADDR_NEEDED); + sc->status = SMBDIRECT_SOCKET_RESOLVE_ADDR_RUNNING; rc = rdma_resolve_addr(id, NULL, (struct sockaddr *)dstaddr, - RDMA_RESOLVE_TIMEOUT); + sp->resolve_addr_timeout_msec); if (rc) { log_rdma_event(ERR, "rdma_resolve_addr() failed %i\n", rc); goto out; } - rc = wait_for_completion_interruptible_timeout( - &info->ri_done, msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT)); + rc = wait_event_interruptible_timeout( + sc->status_wait, + sc->status != SMBDIRECT_SOCKET_RESOLVE_ADDR_RUNNING, + msecs_to_jiffies(sp->resolve_addr_timeout_msec)); /* e.g. if interrupted returns -ERESTARTSYS */ if (rc < 0) { log_rdma_event(ERR, "rdma_resolve_addr timeout rc: %i\n", rc); goto out; } - rc = info->ri_rc; - if (rc) { + if (sc->status == SMBDIRECT_SOCKET_RESOLVE_ADDR_RUNNING) { + rc = -ETIMEDOUT; + log_rdma_event(ERR, "rdma_resolve_addr() completed %i\n", rc); + goto out; + } + if (sc->status != SMBDIRECT_SOCKET_RESOLVE_ROUTE_NEEDED) { + rc = -EHOSTUNREACH; log_rdma_event(ERR, "rdma_resolve_addr() completed %i\n", rc); goto out; } - info->ri_rc = -ETIMEDOUT; - rc = rdma_resolve_route(id, RDMA_RESOLVE_TIMEOUT); + WARN_ON_ONCE(sc->status != SMBDIRECT_SOCKET_RESOLVE_ROUTE_NEEDED); + sc->status = SMBDIRECT_SOCKET_RESOLVE_ROUTE_RUNNING; + rc = rdma_resolve_route(id, sp->resolve_route_timeout_msec); if (rc) { log_rdma_event(ERR, "rdma_resolve_route() failed %i\n", rc); goto out; } - rc = wait_for_completion_interruptible_timeout( - &info->ri_done, msecs_to_jiffies(RDMA_RESOLVE_TIMEOUT)); + rc = wait_event_interruptible_timeout( + sc->status_wait, + sc->status != SMBDIRECT_SOCKET_RESOLVE_ROUTE_RUNNING, + msecs_to_jiffies(sp->resolve_route_timeout_msec)); /* e.g. if interrupted returns -ERESTARTSYS */ if (rc < 0) { log_rdma_event(ERR, "rdma_resolve_addr timeout rc: %i\n", rc); goto out; } - rc = info->ri_rc; - if (rc) { + if (sc->status == SMBDIRECT_SOCKET_RESOLVE_ROUTE_RUNNING) { + rc = -ETIMEDOUT; + log_rdma_event(ERR, "rdma_resolve_route() completed %i\n", rc); + goto out; + } + if (sc->status != SMBDIRECT_SOCKET_RDMA_CONNECT_NEEDED) { + rc = -ENETUNREACH; log_rdma_event(ERR, "rdma_resolve_route() completed %i\n", rc); goto out; } @@ -632,44 +905,42 @@ static bool frwr_is_supported(struct ib_device_attr *attrs) } static int smbd_ia_open( - struct smbd_connection *info, + struct smbdirect_socket *sc, struct sockaddr *dstaddr, int port) { + struct smbdirect_socket_parameters *sp = &sc->parameters; int rc; - info->id = smbd_create_id(info, dstaddr, port); - if (IS_ERR(info->id)) { - rc = PTR_ERR(info->id); + WARN_ON_ONCE(sc->status != SMBDIRECT_SOCKET_CREATED); + sc->status = SMBDIRECT_SOCKET_RESOLVE_ADDR_NEEDED; + + sc->rdma.cm_id = smbd_create_id(sc, dstaddr, port); + if (IS_ERR(sc->rdma.cm_id)) { + rc = PTR_ERR(sc->rdma.cm_id); goto out1; } + sc->ib.dev = sc->rdma.cm_id->device; - if (!frwr_is_supported(&info->id->device->attrs)) { + if (!frwr_is_supported(&sc->ib.dev->attrs)) { log_rdma_event(ERR, "Fast Registration Work Requests (FRWR) is not supported\n"); log_rdma_event(ERR, "Device capability flags = %llx max_fast_reg_page_list_len = %u\n", - info->id->device->attrs.device_cap_flags, - info->id->device->attrs.max_fast_reg_page_list_len); + sc->ib.dev->attrs.device_cap_flags, + sc->ib.dev->attrs.max_fast_reg_page_list_len); rc = -EPROTONOSUPPORT; goto out2; } - info->max_frmr_depth = min_t(int, - smbd_max_frmr_depth, - info->id->device->attrs.max_fast_reg_page_list_len); - info->mr_type = IB_MR_TYPE_MEM_REG; - if (info->id->device->attrs.kernel_cap_flags & IBK_SG_GAPS_REG) - info->mr_type = IB_MR_TYPE_SG_GAPS; - - info->pd = ib_alloc_pd(info->id->device, 0); - if (IS_ERR(info->pd)) { - rc = PTR_ERR(info->pd); - log_rdma_event(ERR, "ib_alloc_pd() returned %d\n", rc); - goto out2; - } + sp->max_frmr_depth = min_t(u32, + sp->max_frmr_depth, + sc->ib.dev->attrs.max_fast_reg_page_list_len); + sc->mr_io.type = IB_MR_TYPE_MEM_REG; + if (sc->ib.dev->attrs.kernel_cap_flags & IBK_SG_GAPS_REG) + sc->mr_io.type = IB_MR_TYPE_SG_GAPS; return 0; out2: - rdma_destroy_id(info->id); - info->id = NULL; + rdma_destroy_id(sc->rdma.cm_id); + sc->rdma.cm_id = NULL; out1: return rc; @@ -681,43 +952,44 @@ out1: * After negotiation, the transport is connected and ready for * carrying upper layer SMB payload */ -static int smbd_post_send_negotiate_req(struct smbd_connection *info) +static int smbd_post_send_negotiate_req(struct smbdirect_socket *sc) { + struct smbdirect_socket_parameters *sp = &sc->parameters; struct ib_send_wr send_wr; int rc = -ENOMEM; - struct smbd_request *request; - struct smbd_negotiate_req *packet; + struct smbdirect_send_io *request; + struct smbdirect_negotiate_req *packet; - request = mempool_alloc(info->request_mempool, GFP_KERNEL); + request = mempool_alloc(sc->send_io.mem.pool, GFP_KERNEL); if (!request) return rc; - request->info = info; + request->socket = sc; - packet = smbd_request_payload(request); - packet->min_version = cpu_to_le16(SMBD_V1); - packet->max_version = cpu_to_le16(SMBD_V1); + packet = smbdirect_send_io_payload(request); + packet->min_version = cpu_to_le16(SMBDIRECT_V1); + packet->max_version = cpu_to_le16(SMBDIRECT_V1); packet->reserved = 0; - packet->credits_requested = cpu_to_le16(info->send_credit_target); - packet->preferred_send_size = cpu_to_le32(info->max_send_size); - packet->max_receive_size = cpu_to_le32(info->max_receive_size); + packet->credits_requested = cpu_to_le16(sp->send_credit_target); + packet->preferred_send_size = cpu_to_le32(sp->max_send_size); + packet->max_receive_size = cpu_to_le32(sp->max_recv_size); packet->max_fragmented_size = - cpu_to_le32(info->max_fragmented_recv_size); + cpu_to_le32(sp->max_fragmented_recv_size); request->num_sge = 1; request->sge[0].addr = ib_dma_map_single( - info->id->device, (void *)packet, + sc->ib.dev, (void *)packet, sizeof(*packet), DMA_TO_DEVICE); - if (ib_dma_mapping_error(info->id->device, request->sge[0].addr)) { + if (ib_dma_mapping_error(sc->ib.dev, request->sge[0].addr)) { rc = -EIO; goto dma_mapping_failed; } request->sge[0].length = sizeof(*packet); - request->sge[0].lkey = info->pd->local_dma_lkey; + request->sge[0].lkey = sc->ib.pd->local_dma_lkey; ib_dma_sync_single_for_device( - info->id->device, request->sge[0].addr, + sc->ib.dev, request->sge[0].addr, request->sge[0].length, DMA_TO_DEVICE); request->cqe.done = send_done; @@ -733,21 +1005,21 @@ static int smbd_post_send_negotiate_req(struct smbd_connection *info) request->sge[0].addr, request->sge[0].length, request->sge[0].lkey); - atomic_inc(&info->send_pending); - rc = ib_post_send(info->id->qp, &send_wr, NULL); + atomic_inc(&sc->send_io.pending.count); + rc = ib_post_send(sc->ib.qp, &send_wr, NULL); if (!rc) return 0; /* if we reach here, post send failed */ log_rdma_send(ERR, "ib_post_send failed rc=%d\n", rc); - atomic_dec(&info->send_pending); - ib_dma_unmap_single(info->id->device, request->sge[0].addr, + atomic_dec(&sc->send_io.pending.count); + ib_dma_unmap_single(sc->ib.dev, request->sge[0].addr, request->sge[0].length, DMA_TO_DEVICE); - smbd_disconnect_rdma_connection(info); + smbd_disconnect_rdma_connection(sc); dma_mapping_failed: - mempool_free(request, info->request_mempool); + mempool_free(request, sc->send_io.mem.pool); return rc; } @@ -759,14 +1031,20 @@ dma_mapping_failed: * buffer as possible, and extend the receive credits to remote peer * return value: the new credtis being granted. */ -static int manage_credits_prior_sending(struct smbd_connection *info) +static int manage_credits_prior_sending(struct smbdirect_socket *sc) { int new_credits; - spin_lock(&info->lock_new_credits_offered); - new_credits = info->new_credits_offered; - info->new_credits_offered = 0; - spin_unlock(&info->lock_new_credits_offered); + if (atomic_read(&sc->recv_io.credits.count) >= sc->recv_io.credits.target) + return 0; + + new_credits = atomic_read(&sc->recv_io.posted.count); + if (new_credits == 0) + return 0; + + new_credits -= atomic_read(&sc->recv_io.credits.count); + if (new_credits <= 0) + return 0; return new_credits; } @@ -774,24 +1052,32 @@ static int manage_credits_prior_sending(struct smbd_connection *info) /* * Check if we need to send a KEEP_ALIVE message * The idle connection timer triggers a KEEP_ALIVE message when expires - * SMB_DIRECT_RESPONSE_REQUESTED is set in the message flag to have peer send + * SMBDIRECT_FLAG_RESPONSE_REQUESTED is set in the message flag to have peer send * back a response. * return value: - * 1 if SMB_DIRECT_RESPONSE_REQUESTED needs to be set + * 1 if SMBDIRECT_FLAG_RESPONSE_REQUESTED needs to be set * 0: otherwise */ -static int manage_keep_alive_before_sending(struct smbd_connection *info) +static int manage_keep_alive_before_sending(struct smbdirect_socket *sc) { - if (info->keep_alive_requested == KEEP_ALIVE_PENDING) { - info->keep_alive_requested = KEEP_ALIVE_SENT; + struct smbdirect_socket_parameters *sp = &sc->parameters; + + if (sc->idle.keepalive == SMBDIRECT_KEEPALIVE_PENDING) { + sc->idle.keepalive = SMBDIRECT_KEEPALIVE_SENT; + /* + * Now use the keepalive timeout (instead of keepalive interval) + * in order to wait for a response + */ + mod_delayed_work(sc->workqueue, &sc->idle.timer_work, + msecs_to_jiffies(sp->keepalive_timeout_msec)); return 1; } return 0; } /* Post the send request */ -static int smbd_post_send(struct smbd_connection *info, - struct smbd_request *request) +static int smbd_post_send(struct smbdirect_socket *sc, + struct smbdirect_send_io *request) { struct ib_send_wr send_wr; int rc, i; @@ -801,7 +1087,7 @@ static int smbd_post_send(struct smbd_connection *info, "rdma_request sge[%d] addr=0x%llx length=%u\n", i, request->sge[i].addr, request->sge[i].length); ib_dma_sync_single_for_device( - info->id->device, + sc->ib.dev, request->sge[i].addr, request->sge[i].length, DMA_TO_DEVICE); @@ -816,86 +1102,107 @@ static int smbd_post_send(struct smbd_connection *info, send_wr.opcode = IB_WR_SEND; send_wr.send_flags = IB_SEND_SIGNALED; - rc = ib_post_send(info->id->qp, &send_wr, NULL); + rc = ib_post_send(sc->ib.qp, &send_wr, NULL); if (rc) { log_rdma_send(ERR, "ib_post_send failed rc=%d\n", rc); - smbd_disconnect_rdma_connection(info); + smbd_disconnect_rdma_connection(sc); rc = -EAGAIN; - } else - /* Reset timer for idle connection after packet is sent */ - mod_delayed_work(info->workqueue, &info->idle_timer_work, - info->keep_alive_interval*HZ); + } return rc; } -static int smbd_post_send_iter(struct smbd_connection *info, +static int smbd_post_send_iter(struct smbdirect_socket *sc, struct iov_iter *iter, int *_remaining_data_length) { + struct smbdirect_socket_parameters *sp = &sc->parameters; int i, rc; int header_length; int data_length; - struct smbd_request *request; - struct smbd_data_transfer *packet; + struct smbdirect_send_io *request; + struct smbdirect_data_transfer *packet; int new_credits = 0; -wait_credit: - /* Wait for send credits. A SMBD packet needs one credit */ - rc = wait_event_interruptible(info->wait_send_queue, - atomic_read(&info->send_credits) > 0 || - info->transport_status != SMBD_CONNECTED); +wait_lcredit: + /* Wait for local send credits */ + rc = wait_event_interruptible(sc->send_io.lcredits.wait_queue, + atomic_read(&sc->send_io.lcredits.count) > 0 || + sc->status != SMBDIRECT_SOCKET_CONNECTED); if (rc) - goto err_wait_credit; + goto err_wait_lcredit; - if (info->transport_status != SMBD_CONNECTED) { + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) { log_outgoing(ERR, "disconnected not sending on wait_credit\n"); rc = -EAGAIN; - goto err_wait_credit; + goto err_wait_lcredit; } - if (unlikely(atomic_dec_return(&info->send_credits) < 0)) { - atomic_inc(&info->send_credits); - goto wait_credit; + if (unlikely(atomic_dec_return(&sc->send_io.lcredits.count) < 0)) { + atomic_inc(&sc->send_io.lcredits.count); + goto wait_lcredit; } -wait_send_queue: - wait_event(info->wait_post_send, - atomic_read(&info->send_pending) < info->send_credit_target || - info->transport_status != SMBD_CONNECTED); +wait_credit: + /* Wait for send credits. A SMBD packet needs one credit */ + rc = wait_event_interruptible(sc->send_io.credits.wait_queue, + atomic_read(&sc->send_io.credits.count) > 0 || + sc->status != SMBDIRECT_SOCKET_CONNECTED); + if (rc) + goto err_wait_credit; - if (info->transport_status != SMBD_CONNECTED) { - log_outgoing(ERR, "disconnected not sending on wait_send_queue\n"); + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) { + log_outgoing(ERR, "disconnected not sending on wait_credit\n"); rc = -EAGAIN; - goto err_wait_send_queue; + goto err_wait_credit; } - - if (unlikely(atomic_inc_return(&info->send_pending) > - info->send_credit_target)) { - atomic_dec(&info->send_pending); - goto wait_send_queue; + if (unlikely(atomic_dec_return(&sc->send_io.credits.count) < 0)) { + atomic_inc(&sc->send_io.credits.count); + goto wait_credit; } - request = mempool_alloc(info->request_mempool, GFP_KERNEL); + request = mempool_alloc(sc->send_io.mem.pool, GFP_KERNEL); if (!request) { rc = -ENOMEM; goto err_alloc; } - request->info = info; + request->socket = sc; memset(request->sge, 0, sizeof(request->sge)); + /* Map the packet to DMA */ + header_length = sizeof(struct smbdirect_data_transfer); + /* If this is a packet without payload, don't send padding */ + if (!iter) + header_length = offsetof(struct smbdirect_data_transfer, padding); + + packet = smbdirect_send_io_payload(request); + request->sge[0].addr = ib_dma_map_single(sc->ib.dev, + (void *)packet, + header_length, + DMA_TO_DEVICE); + if (ib_dma_mapping_error(sc->ib.dev, request->sge[0].addr)) { + rc = -EIO; + goto err_dma; + } + + request->sge[0].length = header_length; + request->sge[0].lkey = sc->ib.pd->local_dma_lkey; + request->num_sge = 1; + /* Fill in the data payload to find out how much data we can add */ if (iter) { struct smb_extract_to_rdma extract = { - .nr_sge = 1, - .max_sge = SMBDIRECT_MAX_SEND_SGE, + .nr_sge = request->num_sge, + .max_sge = SMBDIRECT_SEND_IO_MAX_SGE, .sge = request->sge, - .device = info->id->device, - .local_dma_lkey = info->pd->local_dma_lkey, + .device = sc->ib.dev, + .local_dma_lkey = sc->ib.pd->local_dma_lkey, .direction = DMA_TO_DEVICE, }; + size_t payload_len = umin(*_remaining_data_length, + sp->max_send_size - sizeof(*packet)); - rc = smb_extract_iter_to_rdma(iter, *_remaining_data_length, + rc = smb_extract_iter_to_rdma(iter, payload_len, &extract); if (rc < 0) goto err_dma; @@ -904,22 +1211,18 @@ wait_send_queue: *_remaining_data_length -= data_length; } else { data_length = 0; - request->num_sge = 1; } /* Fill in the packet header */ - packet = smbd_request_payload(request); - packet->credits_requested = cpu_to_le16(info->send_credit_target); + packet->credits_requested = cpu_to_le16(sp->send_credit_target); - new_credits = manage_credits_prior_sending(info); - atomic_add(new_credits, &info->receive_credits); + new_credits = manage_credits_prior_sending(sc); + atomic_add(new_credits, &sc->recv_io.credits.count); packet->credits_granted = cpu_to_le16(new_credits); - info->send_immediate = false; - packet->flags = 0; - if (manage_keep_alive_before_sending(info)) - packet->flags |= cpu_to_le16(SMB_DIRECT_RESPONSE_REQUESTED); + if (manage_keep_alive_before_sending(sc)) + packet->flags |= cpu_to_le16(SMBDIRECT_FLAG_RESPONSE_REQUESTED); packet->reserved = 0; if (!data_length) @@ -937,53 +1240,42 @@ wait_send_queue: le32_to_cpu(packet->data_length), le32_to_cpu(packet->remaining_data_length)); - /* Map the packet to DMA */ - header_length = sizeof(struct smbd_data_transfer); - /* If this is a packet without payload, don't send padding */ - if (!data_length) - header_length = offsetof(struct smbd_data_transfer, padding); - - request->sge[0].addr = ib_dma_map_single(info->id->device, - (void *)packet, - header_length, - DMA_TO_DEVICE); - if (ib_dma_mapping_error(info->id->device, request->sge[0].addr)) { - rc = -EIO; - request->sge[0].addr = 0; - goto err_dma; - } - - request->sge[0].length = header_length; - request->sge[0].lkey = info->pd->local_dma_lkey; + /* + * Now that we got a local and a remote credit + * we add us as pending + */ + atomic_inc(&sc->send_io.pending.count); - rc = smbd_post_send(info, request); + rc = smbd_post_send(sc, request); if (!rc) return 0; + if (atomic_dec_and_test(&sc->send_io.pending.count)) + wake_up(&sc->send_io.pending.zero_wait_queue); + + wake_up(&sc->send_io.pending.dec_wait_queue); + err_dma: for (i = 0; i < request->num_sge; i++) if (request->sge[i].addr) - ib_dma_unmap_single(info->id->device, + ib_dma_unmap_single(sc->ib.dev, request->sge[i].addr, request->sge[i].length, DMA_TO_DEVICE); - mempool_free(request, info->request_mempool); + mempool_free(request, sc->send_io.mem.pool); - /* roll back receive credits and credits to be offered */ - spin_lock(&info->lock_new_credits_offered); - info->new_credits_offered += new_credits; - spin_unlock(&info->lock_new_credits_offered); - atomic_sub(new_credits, &info->receive_credits); + /* roll back the granted receive credits */ + atomic_sub(new_credits, &sc->recv_io.credits.count); err_alloc: - if (atomic_dec_and_test(&info->send_pending)) - wake_up(&info->wait_send_pending); - -err_wait_send_queue: - /* roll back send credits and pending */ - atomic_inc(&info->send_credits); + atomic_inc(&sc->send_io.credits.count); + wake_up(&sc->send_io.credits.wait_queue); err_wait_credit: + atomic_inc(&sc->send_io.lcredits.count); + wake_up(&sc->send_io.lcredits.wait_queue); + +err_wait_lcredit: return rc; } @@ -992,12 +1284,33 @@ err_wait_credit: * Empty message is used to extend credits to peer to for keep live * while there is no upper layer payload to send at the time */ -static int smbd_post_send_empty(struct smbd_connection *info) +static int smbd_post_send_empty(struct smbdirect_socket *sc) { int remaining_data_length = 0; - info->count_send_empty++; - return smbd_post_send_iter(info, NULL, &remaining_data_length); + sc->statistics.send_empty++; + return smbd_post_send_iter(sc, NULL, &remaining_data_length); +} + +static int smbd_post_send_full_iter(struct smbdirect_socket *sc, + struct iov_iter *iter, + int *_remaining_data_length) +{ + int rc = 0; + + /* + * smbd_post_send_iter() respects the + * negotiated max_send_size, so we need to + * loop until the full iter is posted + */ + + while (iov_iter_count(iter) > 0) { + rc = smbd_post_send_iter(sc, iter, _remaining_data_length); + if (rc < 0) + break; + } + + return rc; } /* @@ -1006,19 +1319,20 @@ static int smbd_post_send_empty(struct smbd_connection *info) * The interaction is controlled by send/receive credit system */ static int smbd_post_recv( - struct smbd_connection *info, struct smbd_response *response) + struct smbdirect_socket *sc, struct smbdirect_recv_io *response) { + struct smbdirect_socket_parameters *sp = &sc->parameters; struct ib_recv_wr recv_wr; int rc = -EIO; response->sge.addr = ib_dma_map_single( - info->id->device, response->packet, - info->max_receive_size, DMA_FROM_DEVICE); - if (ib_dma_mapping_error(info->id->device, response->sge.addr)) + sc->ib.dev, response->packet, + sp->max_recv_size, DMA_FROM_DEVICE); + if (ib_dma_mapping_error(sc->ib.dev, response->sge.addr)) return rc; - response->sge.length = info->max_receive_size; - response->sge.lkey = info->pd->local_dma_lkey; + response->sge.length = sp->max_recv_size; + response->sge.lkey = sc->ib.pd->local_dma_lkey; response->cqe.done = recv_done; @@ -1027,11 +1341,12 @@ static int smbd_post_recv( recv_wr.sg_list = &response->sge; recv_wr.num_sge = 1; - rc = ib_post_recv(info->id->qp, &recv_wr, NULL); + rc = ib_post_recv(sc->ib.qp, &recv_wr, NULL); if (rc) { - ib_dma_unmap_single(info->id->device, response->sge.addr, + ib_dma_unmap_single(sc->ib.dev, response->sge.addr, response->sge.length, DMA_FROM_DEVICE); - smbd_disconnect_rdma_connection(info); + response->sge.length = 0; + smbd_disconnect_rdma_connection(sc); log_rdma_recv(ERR, "ib_post_recv failed rc=%d\n", rc); } @@ -1039,30 +1354,36 @@ static int smbd_post_recv( } /* Perform SMBD negotiate according to [MS-SMBD] 3.1.5.2 */ -static int smbd_negotiate(struct smbd_connection *info) +static int smbd_negotiate(struct smbdirect_socket *sc) { + struct smbdirect_socket_parameters *sp = &sc->parameters; int rc; - struct smbd_response *response = get_receive_buffer(info); + struct smbdirect_recv_io *response = get_receive_buffer(sc); - response->type = SMBD_NEGOTIATE_RESP; - rc = smbd_post_recv(info, response); + WARN_ON_ONCE(sc->status != SMBDIRECT_SOCKET_NEGOTIATE_NEEDED); + sc->status = SMBDIRECT_SOCKET_NEGOTIATE_RUNNING; + + sc->recv_io.expected = SMBDIRECT_EXPECT_NEGOTIATE_REP; + rc = smbd_post_recv(sc, response); log_rdma_event(INFO, "smbd_post_recv rc=%d iov.addr=0x%llx iov.length=%u iov.lkey=0x%x\n", rc, response->sge.addr, response->sge.length, response->sge.lkey); - if (rc) + if (rc) { + put_receive_buffer(sc, response); return rc; + } - init_completion(&info->negotiate_completion); - info->negotiate_done = false; - rc = smbd_post_send_negotiate_req(info); + rc = smbd_post_send_negotiate_req(sc); if (rc) return rc; - rc = wait_for_completion_interruptible_timeout( - &info->negotiate_completion, SMBD_NEGOTIATE_TIMEOUT * HZ); - log_rdma_event(INFO, "wait_for_completion_timeout rc=%d\n", rc); + rc = wait_event_interruptible_timeout( + sc->status_wait, + sc->status != SMBDIRECT_SOCKET_NEGOTIATE_RUNNING, + msecs_to_jiffies(sp->negotiate_timeout_msec)); + log_rdma_event(INFO, "wait_event_interruptible_timeout rc=%d\n", rc); - if (info->negotiate_done) + if (sc->status == SMBDIRECT_SOCKET_CONNECTED) return 0; if (rc == 0) @@ -1075,17 +1396,6 @@ static int smbd_negotiate(struct smbd_connection *info) return rc; } -static void put_empty_packet( - struct smbd_connection *info, struct smbd_response *response) -{ - spin_lock(&info->empty_packet_queue_lock); - list_add_tail(&response->list, &info->empty_packet_queue); - info->count_empty_packet_queue++; - spin_unlock(&info->empty_packet_queue_lock); - - queue_work(info->workqueue, &info->post_send_credits_work); -} - /* * Implement Connection.FragmentReassemblyBuffer defined in [MS-SMBD] 3.1.1.1 * This is a queue for reassembling upper layer payload and present to upper @@ -1097,13 +1407,15 @@ static void put_empty_packet( * data_length: the size of payload in this packet */ static void enqueue_reassembly( - struct smbd_connection *info, - struct smbd_response *response, + struct smbdirect_socket *sc, + struct smbdirect_recv_io *response, int data_length) { - spin_lock(&info->reassembly_queue_lock); - list_add_tail(&response->list, &info->reassembly_queue); - info->reassembly_queue_length++; + unsigned long flags; + + spin_lock_irqsave(&sc->recv_io.reassembly.lock, flags); + list_add_tail(&response->list, &sc->recv_io.reassembly.list); + sc->recv_io.reassembly.queue_length++; /* * Make sure reassembly_data_length is updated after list and * reassembly_queue_length are updated. On the dequeue side @@ -1111,10 +1423,9 @@ static void enqueue_reassembly( * if reassembly_queue_length and list is up to date */ virt_wmb(); - info->reassembly_data_length += data_length; - spin_unlock(&info->reassembly_queue_lock); - info->count_reassembly_queue++; - info->count_enqueue_reassembly_queue++; + sc->recv_io.reassembly.data_length += data_length; + spin_unlock_irqrestore(&sc->recv_io.reassembly.lock, flags); + sc->statistics.enqueue_reassembly_queue++; } /* @@ -1122,34 +1433,15 @@ static void enqueue_reassembly( * Caller is responsible for locking * return value: the first entry if any, NULL if queue is empty */ -static struct smbd_response *_get_first_reassembly(struct smbd_connection *info) -{ - struct smbd_response *ret = NULL; - - if (!list_empty(&info->reassembly_queue)) { - ret = list_first_entry( - &info->reassembly_queue, - struct smbd_response, list); - } - return ret; -} - -static struct smbd_response *get_empty_queue_buffer( - struct smbd_connection *info) +static struct smbdirect_recv_io *_get_first_reassembly(struct smbdirect_socket *sc) { - struct smbd_response *ret = NULL; - unsigned long flags; + struct smbdirect_recv_io *ret = NULL; - spin_lock_irqsave(&info->empty_packet_queue_lock, flags); - if (!list_empty(&info->empty_packet_queue)) { + if (!list_empty(&sc->recv_io.reassembly.list)) { ret = list_first_entry( - &info->empty_packet_queue, - struct smbd_response, list); - list_del(&ret->list); - info->count_empty_packet_queue--; + &sc->recv_io.reassembly.list, + struct smbdirect_recv_io, list); } - spin_unlock_irqrestore(&info->empty_packet_queue_lock, flags); - return ret; } @@ -1159,21 +1451,20 @@ static struct smbd_response *get_empty_queue_buffer( * pre-allocated in advance. * return value: the receive buffer, NULL if none is available */ -static struct smbd_response *get_receive_buffer(struct smbd_connection *info) +static struct smbdirect_recv_io *get_receive_buffer(struct smbdirect_socket *sc) { - struct smbd_response *ret = NULL; + struct smbdirect_recv_io *ret = NULL; unsigned long flags; - spin_lock_irqsave(&info->receive_queue_lock, flags); - if (!list_empty(&info->receive_queue)) { + spin_lock_irqsave(&sc->recv_io.free.lock, flags); + if (!list_empty(&sc->recv_io.free.list)) { ret = list_first_entry( - &info->receive_queue, - struct smbd_response, list); + &sc->recv_io.free.list, + struct smbdirect_recv_io, list); list_del(&ret->list); - info->count_receive_queue--; - info->count_get_receive_buffer++; + sc->statistics.get_receive_buffer++; } - spin_unlock_irqrestore(&info->receive_queue_lock, flags); + spin_unlock_irqrestore(&sc->recv_io.free.lock, flags); return ret; } @@ -1185,100 +1476,103 @@ static struct smbd_response *get_receive_buffer(struct smbd_connection *info) * receive buffer is returned. */ static void put_receive_buffer( - struct smbd_connection *info, struct smbd_response *response) + struct smbdirect_socket *sc, struct smbdirect_recv_io *response) { unsigned long flags; - ib_dma_unmap_single(info->id->device, response->sge.addr, - response->sge.length, DMA_FROM_DEVICE); + if (likely(response->sge.length != 0)) { + ib_dma_unmap_single(sc->ib.dev, + response->sge.addr, + response->sge.length, + DMA_FROM_DEVICE); + response->sge.length = 0; + } - spin_lock_irqsave(&info->receive_queue_lock, flags); - list_add_tail(&response->list, &info->receive_queue); - info->count_receive_queue++; - info->count_put_receive_buffer++; - spin_unlock_irqrestore(&info->receive_queue_lock, flags); + spin_lock_irqsave(&sc->recv_io.free.lock, flags); + list_add_tail(&response->list, &sc->recv_io.free.list); + sc->statistics.put_receive_buffer++; + spin_unlock_irqrestore(&sc->recv_io.free.lock, flags); - queue_work(info->workqueue, &info->post_send_credits_work); + queue_work(sc->workqueue, &sc->recv_io.posted.refill_work); } /* Preallocate all receive buffer on transport establishment */ -static int allocate_receive_buffers(struct smbd_connection *info, int num_buf) +static int allocate_receive_buffers(struct smbdirect_socket *sc, int num_buf) { + struct smbdirect_recv_io *response; int i; - struct smbd_response *response; - - INIT_LIST_HEAD(&info->reassembly_queue); - spin_lock_init(&info->reassembly_queue_lock); - info->reassembly_data_length = 0; - info->reassembly_queue_length = 0; - - INIT_LIST_HEAD(&info->receive_queue); - spin_lock_init(&info->receive_queue_lock); - info->count_receive_queue = 0; - - INIT_LIST_HEAD(&info->empty_packet_queue); - spin_lock_init(&info->empty_packet_queue_lock); - info->count_empty_packet_queue = 0; - - init_waitqueue_head(&info->wait_receive_queues); for (i = 0; i < num_buf; i++) { - response = mempool_alloc(info->response_mempool, GFP_KERNEL); + response = mempool_alloc(sc->recv_io.mem.pool, GFP_KERNEL); if (!response) goto allocate_failed; - response->info = info; - list_add_tail(&response->list, &info->receive_queue); - info->count_receive_queue++; + response->socket = sc; + response->sge.length = 0; + list_add_tail(&response->list, &sc->recv_io.free.list); } return 0; allocate_failed: - while (!list_empty(&info->receive_queue)) { + while (!list_empty(&sc->recv_io.free.list)) { response = list_first_entry( - &info->receive_queue, - struct smbd_response, list); + &sc->recv_io.free.list, + struct smbdirect_recv_io, list); list_del(&response->list); - info->count_receive_queue--; - mempool_free(response, info->response_mempool); + mempool_free(response, sc->recv_io.mem.pool); } return -ENOMEM; } -static void destroy_receive_buffers(struct smbd_connection *info) +static void destroy_receive_buffers(struct smbdirect_socket *sc) { - struct smbd_response *response; + struct smbdirect_recv_io *response; - while ((response = get_receive_buffer(info))) - mempool_free(response, info->response_mempool); + while ((response = get_receive_buffer(sc))) + mempool_free(response, sc->recv_io.mem.pool); +} - while ((response = get_empty_queue_buffer(info))) - mempool_free(response, info->response_mempool); +static void send_immediate_empty_message(struct work_struct *work) +{ + struct smbdirect_socket *sc = + container_of(work, struct smbdirect_socket, idle.immediate_work); + + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) + return; + + log_keep_alive(INFO, "send an empty message\n"); + smbd_post_send_empty(sc); } /* Implement idle connection timer [MS-SMBD] 3.1.6.2 */ static void idle_connection_timer(struct work_struct *work) { - struct smbd_connection *info = container_of( - work, struct smbd_connection, - idle_timer_work.work); + struct smbdirect_socket *sc = + container_of(work, struct smbdirect_socket, idle.timer_work.work); + struct smbdirect_socket_parameters *sp = &sc->parameters; - if (info->keep_alive_requested != KEEP_ALIVE_NONE) { + if (sc->idle.keepalive != SMBDIRECT_KEEPALIVE_NONE) { log_keep_alive(ERR, - "error status info->keep_alive_requested=%d\n", - info->keep_alive_requested); - smbd_disconnect_rdma_connection(info); + "error status sc->idle.keepalive=%d\n", + sc->idle.keepalive); + smbd_disconnect_rdma_connection(sc); return; } - log_keep_alive(INFO, "about to send an empty idle message\n"); - smbd_post_send_empty(info); + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) + return; - /* Setup the next idle timeout work */ - queue_delayed_work(info->workqueue, &info->idle_timer_work, - info->keep_alive_interval*HZ); + /* + * Now use the keepalive timeout (instead of keepalive interval) + * in order to wait for a response + */ + sc->idle.keepalive = SMBDIRECT_KEEPALIVE_PENDING; + mod_delayed_work(sc->workqueue, &sc->idle.timer_work, + msecs_to_jiffies(sp->keepalive_timeout_msec)); + log_keep_alive(INFO, "schedule send of empty idle message\n"); + queue_work(sc->workqueue, &sc->idle.immediate_work); } /* @@ -1289,87 +1583,87 @@ static void idle_connection_timer(struct work_struct *work) void smbd_destroy(struct TCP_Server_Info *server) { struct smbd_connection *info = server->smbd_conn; - struct smbd_response *response; + struct smbdirect_socket *sc; + struct smbdirect_recv_io *response; unsigned long flags; if (!info) { log_rdma_event(INFO, "rdma session already destroyed\n"); return; } + sc = &info->socket; + + log_rdma_event(INFO, "cancelling and disable disconnect_work\n"); + disable_work_sync(&sc->disconnect_work); log_rdma_event(INFO, "destroying rdma session\n"); - if (info->transport_status != SMBD_DISCONNECTED) { - rdma_disconnect(server->smbd_conn->id); + if (sc->status < SMBDIRECT_SOCKET_DISCONNECTING) + smbd_disconnect_rdma_work(&sc->disconnect_work); + if (sc->status < SMBDIRECT_SOCKET_DISCONNECTED) { log_rdma_event(INFO, "wait for transport being disconnected\n"); - wait_event_interruptible( - info->disconn_wait, - info->transport_status == SMBD_DISCONNECTED); + wait_event(sc->status_wait, sc->status == SMBDIRECT_SOCKET_DISCONNECTED); + log_rdma_event(INFO, "waited for transport being disconnected\n"); } + /* + * Wake up all waiters in all wait queues + * in order to notice the broken connection. + * + * Most likely this was already called via + * smbd_disconnect_rdma_work(), but call it again... + */ + smbd_disconnect_wake_up_all(sc); + + log_rdma_event(INFO, "cancelling recv_io.posted.refill_work\n"); + disable_work_sync(&sc->recv_io.posted.refill_work); + log_rdma_event(INFO, "destroying qp\n"); - ib_drain_qp(info->id->qp); - rdma_destroy_qp(info->id); + ib_drain_qp(sc->ib.qp); + rdma_destroy_qp(sc->rdma.cm_id); + sc->ib.qp = NULL; log_rdma_event(INFO, "cancelling idle timer\n"); - cancel_delayed_work_sync(&info->idle_timer_work); - - log_rdma_event(INFO, "wait for all send posted to IB to finish\n"); - wait_event(info->wait_send_pending, - atomic_read(&info->send_pending) == 0); + disable_delayed_work_sync(&sc->idle.timer_work); + log_rdma_event(INFO, "cancelling send immediate work\n"); + disable_work_sync(&sc->idle.immediate_work); /* It's not possible for upper layer to get to reassembly */ log_rdma_event(INFO, "drain the reassembly queue\n"); do { - spin_lock_irqsave(&info->reassembly_queue_lock, flags); - response = _get_first_reassembly(info); + spin_lock_irqsave(&sc->recv_io.reassembly.lock, flags); + response = _get_first_reassembly(sc); if (response) { list_del(&response->list); spin_unlock_irqrestore( - &info->reassembly_queue_lock, flags); - put_receive_buffer(info, response); + &sc->recv_io.reassembly.lock, flags); + put_receive_buffer(sc, response); } else spin_unlock_irqrestore( - &info->reassembly_queue_lock, flags); + &sc->recv_io.reassembly.lock, flags); } while (response); - info->reassembly_data_length = 0; + sc->recv_io.reassembly.data_length = 0; log_rdma_event(INFO, "free receive buffers\n"); - wait_event(info->wait_receive_queues, - info->count_receive_queue + info->count_empty_packet_queue - == info->receive_credit_max); - destroy_receive_buffers(info); + destroy_receive_buffers(sc); - /* - * For performance reasons, memory registration and deregistration - * are not locked by srv_mutex. It is possible some processes are - * blocked on transport srv_mutex while holding memory registration. - * Release the transport srv_mutex to allow them to hit the failure - * path when sending data, and then release memory registrations. - */ log_rdma_event(INFO, "freeing mr list\n"); - wake_up_interruptible_all(&info->wait_mr); - while (atomic_read(&info->mr_used_count)) { - cifs_server_unlock(server); - msleep(1000); - cifs_server_lock(server); - } - destroy_mr_list(info); + destroy_mr_list(sc); - ib_free_cq(info->send_cq); - ib_free_cq(info->recv_cq); - ib_dealloc_pd(info->pd); - rdma_destroy_id(info->id); + ib_free_cq(sc->ib.send_cq); + ib_free_cq(sc->ib.recv_cq); + ib_dealloc_pd(sc->ib.pd); + rdma_destroy_id(sc->rdma.cm_id); /* free mempools */ - mempool_destroy(info->request_mempool); - kmem_cache_destroy(info->request_cache); + mempool_destroy(sc->send_io.mem.pool); + kmem_cache_destroy(sc->send_io.mem.cache); - mempool_destroy(info->response_mempool); - kmem_cache_destroy(info->response_cache); + mempool_destroy(sc->recv_io.mem.pool); + kmem_cache_destroy(sc->recv_io.mem.cache); - info->transport_status = SMBD_DESTROYED; + sc->status = SMBDIRECT_SOCKET_DESTROYED; - destroy_workqueue(info->workqueue); + destroy_workqueue(sc->workqueue); log_rdma_event(INFO, "rdma session destroyed\n"); kfree(info); server->smbd_conn = NULL; @@ -1392,7 +1686,7 @@ int smbd_reconnect(struct TCP_Server_Info *server) * This is possible if transport is disconnected and we haven't received * notification from RDMA, but upper layer has detected timeout */ - if (server->smbd_conn->transport_status == SMBD_CONNECTED) { + if (server->smbd_conn->socket.status == SMBDIRECT_SOCKET_CONNECTED) { log_rdma_event(INFO, "disconnecting transport\n"); smbd_destroy(server); } @@ -1411,77 +1705,78 @@ create_conn: return -ENOENT; } -static void destroy_caches_and_workqueue(struct smbd_connection *info) +static void destroy_caches(struct smbdirect_socket *sc) { - destroy_receive_buffers(info); - destroy_workqueue(info->workqueue); - mempool_destroy(info->response_mempool); - kmem_cache_destroy(info->response_cache); - mempool_destroy(info->request_mempool); - kmem_cache_destroy(info->request_cache); + destroy_receive_buffers(sc); + mempool_destroy(sc->recv_io.mem.pool); + kmem_cache_destroy(sc->recv_io.mem.cache); + mempool_destroy(sc->send_io.mem.pool); + kmem_cache_destroy(sc->send_io.mem.cache); } #define MAX_NAME_LEN 80 -static int allocate_caches_and_workqueue(struct smbd_connection *info) +static int allocate_caches(struct smbdirect_socket *sc) { + struct smbdirect_socket_parameters *sp = &sc->parameters; char name[MAX_NAME_LEN]; int rc; - scnprintf(name, MAX_NAME_LEN, "smbd_request_%p", info); - info->request_cache = + if (WARN_ON_ONCE(sp->max_recv_size < sizeof(struct smbdirect_data_transfer))) + return -ENOMEM; + + scnprintf(name, MAX_NAME_LEN, "smbdirect_send_io_%p", sc); + sc->send_io.mem.cache = kmem_cache_create( name, - sizeof(struct smbd_request) + - sizeof(struct smbd_data_transfer), + sizeof(struct smbdirect_send_io) + + sizeof(struct smbdirect_data_transfer), 0, SLAB_HWCACHE_ALIGN, NULL); - if (!info->request_cache) + if (!sc->send_io.mem.cache) return -ENOMEM; - info->request_mempool = - mempool_create(info->send_credit_target, mempool_alloc_slab, - mempool_free_slab, info->request_cache); - if (!info->request_mempool) + sc->send_io.mem.pool = + mempool_create(sp->send_credit_target, mempool_alloc_slab, + mempool_free_slab, sc->send_io.mem.cache); + if (!sc->send_io.mem.pool) goto out1; - scnprintf(name, MAX_NAME_LEN, "smbd_response_%p", info); - info->response_cache = - kmem_cache_create( - name, - sizeof(struct smbd_response) + - info->max_receive_size, - 0, SLAB_HWCACHE_ALIGN, NULL); - if (!info->response_cache) + scnprintf(name, MAX_NAME_LEN, "smbdirect_recv_io_%p", sc); + + struct kmem_cache_args response_args = { + .align = __alignof__(struct smbdirect_recv_io), + .useroffset = (offsetof(struct smbdirect_recv_io, packet) + + sizeof(struct smbdirect_data_transfer)), + .usersize = sp->max_recv_size - sizeof(struct smbdirect_data_transfer), + }; + sc->recv_io.mem.cache = + kmem_cache_create(name, + sizeof(struct smbdirect_recv_io) + sp->max_recv_size, + &response_args, SLAB_HWCACHE_ALIGN); + if (!sc->recv_io.mem.cache) goto out2; - info->response_mempool = - mempool_create(info->receive_credit_max, mempool_alloc_slab, - mempool_free_slab, info->response_cache); - if (!info->response_mempool) + sc->recv_io.mem.pool = + mempool_create(sp->recv_credit_max, mempool_alloc_slab, + mempool_free_slab, sc->recv_io.mem.cache); + if (!sc->recv_io.mem.pool) goto out3; - scnprintf(name, MAX_NAME_LEN, "smbd_%p", info); - info->workqueue = create_workqueue(name); - if (!info->workqueue) - goto out4; - - rc = allocate_receive_buffers(info, info->receive_credit_max); + rc = allocate_receive_buffers(sc, sp->recv_credit_max); if (rc) { log_rdma_event(ERR, "failed to allocate receive buffers\n"); - goto out5; + goto out4; } return 0; -out5: - destroy_workqueue(info->workqueue); out4: - mempool_destroy(info->response_mempool); + mempool_destroy(sc->recv_io.mem.pool); out3: - kmem_cache_destroy(info->response_cache); + kmem_cache_destroy(sc->recv_io.mem.cache); out2: - mempool_destroy(info->request_mempool); + mempool_destroy(sc->send_io.mem.pool); out1: - kmem_cache_destroy(info->request_cache); + kmem_cache_destroy(sc->send_io.mem.cache); return -ENOMEM; } @@ -1491,113 +1786,151 @@ static struct smbd_connection *_smbd_get_connection( { int rc; struct smbd_connection *info; + struct smbdirect_socket *sc; + struct smbdirect_socket_parameters *sp; struct rdma_conn_param conn_param; + struct ib_qp_cap qp_cap; struct ib_qp_init_attr qp_attr; struct sockaddr_in *addr_in = (struct sockaddr_in *) dstaddr; struct ib_port_immutable port_immutable; - u32 ird_ord_hdr[2]; + __be32 ird_ord_hdr[2]; + char wq_name[80]; + struct workqueue_struct *workqueue; info = kzalloc(sizeof(struct smbd_connection), GFP_KERNEL); if (!info) return NULL; - - info->transport_status = SMBD_CONNECTING; - rc = smbd_ia_open(info, dstaddr, port); + sc = &info->socket; + scnprintf(wq_name, ARRAY_SIZE(wq_name), "smbd_%p", sc); + workqueue = create_workqueue(wq_name); + if (!workqueue) + goto create_wq_failed; + smbdirect_socket_init(sc); + sc->workqueue = workqueue; + sp = &sc->parameters; + + INIT_WORK(&sc->disconnect_work, smbd_disconnect_rdma_work); + + sp->resolve_addr_timeout_msec = RDMA_RESOLVE_TIMEOUT; + sp->resolve_route_timeout_msec = RDMA_RESOLVE_TIMEOUT; + sp->rdma_connect_timeout_msec = RDMA_RESOLVE_TIMEOUT; + sp->negotiate_timeout_msec = SMBD_NEGOTIATE_TIMEOUT * 1000; + sp->initiator_depth = 1; + sp->responder_resources = SMBD_CM_RESPONDER_RESOURCES; + sp->recv_credit_max = smbd_receive_credit_max; + sp->send_credit_target = smbd_send_credit_target; + sp->max_send_size = smbd_max_send_size; + sp->max_fragmented_recv_size = smbd_max_fragmented_recv_size; + sp->max_recv_size = smbd_max_receive_size; + sp->max_frmr_depth = smbd_max_frmr_depth; + sp->keepalive_interval_msec = smbd_keep_alive_interval * 1000; + sp->keepalive_timeout_msec = KEEPALIVE_RECV_TIMEOUT * 1000; + + rc = smbd_ia_open(sc, dstaddr, port); if (rc) { log_rdma_event(INFO, "smbd_ia_open rc=%d\n", rc); goto create_id_failed; } - if (smbd_send_credit_target > info->id->device->attrs.max_cqe || - smbd_send_credit_target > info->id->device->attrs.max_qp_wr) { + if (sp->send_credit_target > sc->ib.dev->attrs.max_cqe || + sp->send_credit_target > sc->ib.dev->attrs.max_qp_wr) { log_rdma_event(ERR, "consider lowering send_credit_target = %d. Possible CQE overrun, device reporting max_cqe %d max_qp_wr %d\n", - smbd_send_credit_target, - info->id->device->attrs.max_cqe, - info->id->device->attrs.max_qp_wr); + sp->send_credit_target, + sc->ib.dev->attrs.max_cqe, + sc->ib.dev->attrs.max_qp_wr); goto config_failed; } - if (smbd_receive_credit_max > info->id->device->attrs.max_cqe || - smbd_receive_credit_max > info->id->device->attrs.max_qp_wr) { + if (sp->recv_credit_max > sc->ib.dev->attrs.max_cqe || + sp->recv_credit_max > sc->ib.dev->attrs.max_qp_wr) { log_rdma_event(ERR, "consider lowering receive_credit_max = %d. Possible CQE overrun, device reporting max_cqe %d max_qp_wr %d\n", - smbd_receive_credit_max, - info->id->device->attrs.max_cqe, - info->id->device->attrs.max_qp_wr); + sp->recv_credit_max, + sc->ib.dev->attrs.max_cqe, + sc->ib.dev->attrs.max_qp_wr); goto config_failed; } - info->receive_credit_max = smbd_receive_credit_max; - info->send_credit_target = smbd_send_credit_target; - info->max_send_size = smbd_max_send_size; - info->max_fragmented_recv_size = smbd_max_fragmented_recv_size; - info->max_receive_size = smbd_max_receive_size; - info->keep_alive_interval = smbd_keep_alive_interval; - - if (info->id->device->attrs.max_send_sge < SMBDIRECT_MAX_SEND_SGE || - info->id->device->attrs.max_recv_sge < SMBDIRECT_MAX_RECV_SGE) { + if (sc->ib.dev->attrs.max_send_sge < SMBDIRECT_SEND_IO_MAX_SGE || + sc->ib.dev->attrs.max_recv_sge < SMBDIRECT_RECV_IO_MAX_SGE) { log_rdma_event(ERR, "device %.*s max_send_sge/max_recv_sge = %d/%d too small\n", IB_DEVICE_NAME_MAX, - info->id->device->name, - info->id->device->attrs.max_send_sge, - info->id->device->attrs.max_recv_sge); + sc->ib.dev->name, + sc->ib.dev->attrs.max_send_sge, + sc->ib.dev->attrs.max_recv_sge); goto config_failed; } - info->send_cq = NULL; - info->recv_cq = NULL; - info->send_cq = - ib_alloc_cq_any(info->id->device, info, - info->send_credit_target, IB_POLL_SOFTIRQ); - if (IS_ERR(info->send_cq)) { - info->send_cq = NULL; + sp->responder_resources = + min_t(u8, sp->responder_resources, + sc->ib.dev->attrs.max_qp_rd_atom); + log_rdma_mr(INFO, "responder_resources=%d\n", + sp->responder_resources); + + /* + * We use allocate sp->responder_resources * 2 MRs + * and each MR needs WRs for REG and INV, so + * we use '* 4'. + * + * +1 for ib_drain_qp() + */ + memset(&qp_cap, 0, sizeof(qp_cap)); + qp_cap.max_send_wr = sp->send_credit_target + sp->responder_resources * 4 + 1; + qp_cap.max_recv_wr = sp->recv_credit_max + 1; + qp_cap.max_send_sge = SMBDIRECT_SEND_IO_MAX_SGE; + qp_cap.max_recv_sge = SMBDIRECT_RECV_IO_MAX_SGE; + + sc->ib.pd = ib_alloc_pd(sc->ib.dev, 0); + if (IS_ERR(sc->ib.pd)) { + rc = PTR_ERR(sc->ib.pd); + sc->ib.pd = NULL; + log_rdma_event(ERR, "ib_alloc_pd() returned %d\n", rc); + goto alloc_pd_failed; + } + + sc->ib.send_cq = + ib_alloc_cq_any(sc->ib.dev, sc, + qp_cap.max_send_wr, IB_POLL_SOFTIRQ); + if (IS_ERR(sc->ib.send_cq)) { + sc->ib.send_cq = NULL; goto alloc_cq_failed; } - info->recv_cq = - ib_alloc_cq_any(info->id->device, info, - info->receive_credit_max, IB_POLL_SOFTIRQ); - if (IS_ERR(info->recv_cq)) { - info->recv_cq = NULL; + sc->ib.recv_cq = + ib_alloc_cq_any(sc->ib.dev, sc, + qp_cap.max_recv_wr, IB_POLL_SOFTIRQ); + if (IS_ERR(sc->ib.recv_cq)) { + sc->ib.recv_cq = NULL; goto alloc_cq_failed; } memset(&qp_attr, 0, sizeof(qp_attr)); qp_attr.event_handler = smbd_qp_async_error_upcall; - qp_attr.qp_context = info; - qp_attr.cap.max_send_wr = info->send_credit_target; - qp_attr.cap.max_recv_wr = info->receive_credit_max; - qp_attr.cap.max_send_sge = SMBDIRECT_MAX_SEND_SGE; - qp_attr.cap.max_recv_sge = SMBDIRECT_MAX_RECV_SGE; - qp_attr.cap.max_inline_data = 0; + qp_attr.qp_context = sc; + qp_attr.cap = qp_cap; qp_attr.sq_sig_type = IB_SIGNAL_REQ_WR; qp_attr.qp_type = IB_QPT_RC; - qp_attr.send_cq = info->send_cq; - qp_attr.recv_cq = info->recv_cq; + qp_attr.send_cq = sc->ib.send_cq; + qp_attr.recv_cq = sc->ib.recv_cq; qp_attr.port_num = ~0; - rc = rdma_create_qp(info->id, info->pd, &qp_attr); + rc = rdma_create_qp(sc->rdma.cm_id, sc->ib.pd, &qp_attr); if (rc) { log_rdma_event(ERR, "rdma_create_qp failed %i\n", rc); goto create_qp_failed; } + sc->ib.qp = sc->rdma.cm_id->qp; memset(&conn_param, 0, sizeof(conn_param)); - conn_param.initiator_depth = 0; - - conn_param.responder_resources = - min(info->id->device->attrs.max_qp_rd_atom, - SMBD_CM_RESPONDER_RESOURCES); - info->responder_resources = conn_param.responder_resources; - log_rdma_mr(INFO, "responder_resources=%d\n", - info->responder_resources); + conn_param.initiator_depth = sp->initiator_depth; + conn_param.responder_resources = sp->responder_resources; /* Need to send IRD/ORD in private data for iWARP */ - info->id->device->ops.get_port_immutable( - info->id->device, info->id->port_num, &port_immutable); + sc->ib.dev->ops.get_port_immutable( + sc->ib.dev, sc->rdma.cm_id->port_num, &port_immutable); if (port_immutable.core_cap_flags & RDMA_CORE_PORT_IWARP) { - ird_ord_hdr[0] = info->responder_resources; - ird_ord_hdr[1] = 1; + ird_ord_hdr[0] = cpu_to_be32(conn_param.responder_resources); + ird_ord_hdr[1] = cpu_to_be32(conn_param.initiator_depth); conn_param.private_data = ird_ord_hdr; conn_param.private_data_len = sizeof(ird_ord_hdr); } else { @@ -1612,53 +1945,51 @@ static struct smbd_connection *_smbd_get_connection( log_rdma_event(INFO, "connecting to IP %pI4 port %d\n", &addr_in->sin_addr, port); - init_waitqueue_head(&info->conn_wait); - init_waitqueue_head(&info->disconn_wait); - init_waitqueue_head(&info->wait_reassembly_queue); - rc = rdma_connect(info->id, &conn_param); + WARN_ON_ONCE(sc->status != SMBDIRECT_SOCKET_RDMA_CONNECT_NEEDED); + sc->status = SMBDIRECT_SOCKET_RDMA_CONNECT_RUNNING; + rc = rdma_connect(sc->rdma.cm_id, &conn_param); if (rc) { log_rdma_event(ERR, "rdma_connect() failed with %i\n", rc); goto rdma_connect_failed; } - wait_event_interruptible( - info->conn_wait, info->transport_status != SMBD_CONNECTING); + wait_event_interruptible_timeout( + sc->status_wait, + sc->status != SMBDIRECT_SOCKET_RDMA_CONNECT_RUNNING, + msecs_to_jiffies(sp->rdma_connect_timeout_msec)); - if (info->transport_status != SMBD_CONNECTED) { + if (sc->status != SMBDIRECT_SOCKET_NEGOTIATE_NEEDED) { log_rdma_event(ERR, "rdma_connect failed port=%d\n", port); goto rdma_connect_failed; } log_rdma_event(INFO, "rdma_connect connected\n"); - rc = allocate_caches_and_workqueue(info); + rc = allocate_caches(sc); if (rc) { log_rdma_event(ERR, "cache allocation failed\n"); goto allocate_cache_failed; } - init_waitqueue_head(&info->wait_send_queue); - INIT_DELAYED_WORK(&info->idle_timer_work, idle_connection_timer); - queue_delayed_work(info->workqueue, &info->idle_timer_work, - info->keep_alive_interval*HZ); - - init_waitqueue_head(&info->wait_send_pending); - atomic_set(&info->send_pending, 0); - - init_waitqueue_head(&info->wait_post_send); + INIT_WORK(&sc->idle.immediate_work, send_immediate_empty_message); + INIT_DELAYED_WORK(&sc->idle.timer_work, idle_connection_timer); + /* + * start with the negotiate timeout and SMBDIRECT_KEEPALIVE_PENDING + * so that the timer will cause a disconnect. + */ + sc->idle.keepalive = SMBDIRECT_KEEPALIVE_PENDING; + mod_delayed_work(sc->workqueue, &sc->idle.timer_work, + msecs_to_jiffies(sp->negotiate_timeout_msec)); - INIT_WORK(&info->disconnect_work, smbd_disconnect_rdma_work); - INIT_WORK(&info->post_send_credits_work, smbd_post_send_credits); - info->new_credits_offered = 0; - spin_lock_init(&info->lock_new_credits_offered); + INIT_WORK(&sc->recv_io.posted.refill_work, smbd_post_send_credits); - rc = smbd_negotiate(info); + rc = smbd_negotiate(sc); if (rc) { log_rdma_event(ERR, "smbd_negotiate rc=%d\n", rc); goto negotiation_failed; } - rc = allocate_mr_list(info); + rc = allocate_mr_list(sc); if (rc) { log_rdma_mr(ERR, "memory registration allocation failed\n"); goto allocate_mr_failed; @@ -1673,30 +2004,33 @@ allocate_mr_failed: return NULL; negotiation_failed: - cancel_delayed_work_sync(&info->idle_timer_work); - destroy_caches_and_workqueue(info); - info->transport_status = SMBD_NEGOTIATE_FAILED; - init_waitqueue_head(&info->conn_wait); - rdma_disconnect(info->id); - wait_event(info->conn_wait, - info->transport_status == SMBD_DISCONNECTED); + disable_delayed_work_sync(&sc->idle.timer_work); + destroy_caches(sc); + sc->status = SMBDIRECT_SOCKET_NEGOTIATE_FAILED; + rdma_disconnect(sc->rdma.cm_id); + wait_event(sc->status_wait, + sc->status == SMBDIRECT_SOCKET_DISCONNECTED); allocate_cache_failed: rdma_connect_failed: - rdma_destroy_qp(info->id); + rdma_destroy_qp(sc->rdma.cm_id); create_qp_failed: alloc_cq_failed: - if (info->send_cq) - ib_free_cq(info->send_cq); - if (info->recv_cq) - ib_free_cq(info->recv_cq); + if (sc->ib.send_cq) + ib_free_cq(sc->ib.send_cq); + if (sc->ib.recv_cq) + ib_free_cq(sc->ib.recv_cq); + ib_dealloc_pd(sc->ib.pd); + +alloc_pd_failed: config_failed: - ib_dealloc_pd(info->pd); - rdma_destroy_id(info->id); + rdma_destroy_id(sc->rdma.cm_id); create_id_failed: + destroy_workqueue(sc->workqueue); +create_wq_failed: kfree(info); return NULL; } @@ -1705,6 +2039,7 @@ struct smbd_connection *smbd_get_connection( struct TCP_Server_Info *server, struct sockaddr *dstaddr) { struct smbd_connection *ret; + const struct smbdirect_socket_parameters *sp; int port = SMBD_PORT; try_again: @@ -1715,42 +2050,58 @@ try_again: port = SMB_PORT; goto try_again; } + if (!ret) + return NULL; + + sp = &ret->socket.parameters; + + server->rdma_readwrite_threshold = + rdma_readwrite_threshold > sp->max_fragmented_send_size ? + sp->max_fragmented_send_size : + rdma_readwrite_threshold; + return ret; } /* - * Receive data from receive reassembly queue + * Receive data from the transport's receive reassembly queue * All the incoming data packets are placed in reassembly queue - * buf: the buffer to read data into + * iter: the buffer to read data into * size: the length of data to read * return value: actual data read - * Note: this implementation copies the data from reassebmly queue to receive + * + * Note: this implementation copies the data from reassembly queue to receive * buffers used by upper layer. This is not the optimal code path. A better way * to do it is to not have upper layer allocate its receive buffers but rather * borrow the buffer from reassembly queue, and return it after data is * consumed. But this will require more changes to upper layer code, and also * need to consider packet boundaries while they still being reassembled. */ -static int smbd_recv_buf(struct smbd_connection *info, char *buf, - unsigned int size) +int smbd_recv(struct smbd_connection *info, struct msghdr *msg) { - struct smbd_response *response; - struct smbd_data_transfer *data_transfer; + struct smbdirect_socket *sc = &info->socket; + struct smbdirect_recv_io *response; + struct smbdirect_data_transfer *data_transfer; + size_t size = iov_iter_count(&msg->msg_iter); int to_copy, to_read, data_read, offset; u32 data_length, remaining_data_length, data_offset; int rc; + if (WARN_ON_ONCE(iov_iter_rw(&msg->msg_iter) == WRITE)) + return -EINVAL; /* It's a bug in upper layer to get there */ + again: /* * No need to hold the reassembly queue lock all the time as we are * the only one reading from the front of the queue. The transport * may add more entries to the back of the queue at the same time */ - log_read(INFO, "size=%d info->reassembly_data_length=%d\n", size, - info->reassembly_data_length); - if (info->reassembly_data_length >= size) { + log_read(INFO, "size=%zd sc->recv_io.reassembly.data_length=%d\n", size, + sc->recv_io.reassembly.data_length); + if (sc->recv_io.reassembly.data_length >= size) { int queue_length; int queue_removed = 0; + unsigned long flags; /* * Need to make sure reassembly_data_length is read before @@ -1760,13 +2111,13 @@ again: * updated in SOFTIRQ as more data is received */ virt_rmb(); - queue_length = info->reassembly_queue_length; + queue_length = sc->recv_io.reassembly.queue_length; data_read = 0; to_read = size; - offset = info->first_entry_offset; + offset = sc->recv_io.reassembly.first_entry_offset; while (data_read < size) { - response = _get_first_reassembly(info); - data_transfer = smbd_response_payload(response); + response = _get_first_reassembly(sc); + data_transfer = smbdirect_recv_io_payload(response); data_length = le32_to_cpu(data_transfer->data_length); remaining_data_length = le32_to_cpu( @@ -1784,7 +2135,10 @@ again: if (response->first_segment && size == 4) { unsigned int rfc1002_len = data_length + remaining_data_length; - *((__be32 *)buf) = cpu_to_be32(rfc1002_len); + __be32 rfc1002_hdr = cpu_to_be32(rfc1002_len); + if (copy_to_iter(&rfc1002_hdr, sizeof(rfc1002_hdr), + &msg->msg_iter) != sizeof(rfc1002_hdr)) + return -EFAULT; data_read = 4; response->first_segment = false; log_read(INFO, "returning rfc1002 length %d\n", @@ -1793,10 +2147,9 @@ again: } to_copy = min_t(int, data_length - offset, to_read); - memcpy( - buf + data_read, - (char *)data_transfer + data_offset + offset, - to_copy); + if (copy_to_iter((char *)data_transfer + data_offset + offset, + to_copy, &msg->msg_iter) != to_copy) + return -EFAULT; /* move on to the next buffer? */ if (to_copy == data_length - offset) { @@ -1808,16 +2161,15 @@ again: if (queue_length) list_del(&response->list); else { - spin_lock_irq( - &info->reassembly_queue_lock); + spin_lock_irqsave( + &sc->recv_io.reassembly.lock, flags); list_del(&response->list); - spin_unlock_irq( - &info->reassembly_queue_lock); + spin_unlock_irqrestore( + &sc->recv_io.reassembly.lock, flags); } queue_removed++; - info->count_reassembly_queue--; - info->count_dequeue_reassembly_queue++; - put_receive_buffer(info, response); + sc->statistics.dequeue_reassembly_queue++; + put_receive_buffer(sc, response); offset = 0; log_read(INFO, "put_receive_buffer offset=0\n"); } else @@ -1831,29 +2183,29 @@ again: to_read, data_read, offset); } - spin_lock_irq(&info->reassembly_queue_lock); - info->reassembly_data_length -= data_read; - info->reassembly_queue_length -= queue_removed; - spin_unlock_irq(&info->reassembly_queue_lock); + spin_lock_irqsave(&sc->recv_io.reassembly.lock, flags); + sc->recv_io.reassembly.data_length -= data_read; + sc->recv_io.reassembly.queue_length -= queue_removed; + spin_unlock_irqrestore(&sc->recv_io.reassembly.lock, flags); - info->first_entry_offset = offset; + sc->recv_io.reassembly.first_entry_offset = offset; log_read(INFO, "returning to thread data_read=%d reassembly_data_length=%d first_entry_offset=%d\n", - data_read, info->reassembly_data_length, - info->first_entry_offset); + data_read, sc->recv_io.reassembly.data_length, + sc->recv_io.reassembly.first_entry_offset); read_rfc1002_done: return data_read; } log_read(INFO, "wait_event on more data\n"); rc = wait_event_interruptible( - info->wait_reassembly_queue, - info->reassembly_data_length >= size || - info->transport_status != SMBD_CONNECTED); + sc->recv_io.reassembly.wait_queue, + sc->recv_io.reassembly.data_length >= size || + sc->status != SMBDIRECT_SOCKET_CONNECTED); /* Don't return any data if interrupted */ if (rc) return rc; - if (info->transport_status != SMBD_CONNECTED) { + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) { log_read(ERR, "disconnected\n"); return -ECONNABORTED; } @@ -1862,89 +2214,6 @@ read_rfc1002_done: } /* - * Receive a page from receive reassembly queue - * page: the page to read data into - * to_read: the length of data to read - * return value: actual data read - */ -static int smbd_recv_page(struct smbd_connection *info, - struct page *page, unsigned int page_offset, - unsigned int to_read) -{ - int ret; - char *to_address; - void *page_address; - - /* make sure we have the page ready for read */ - ret = wait_event_interruptible( - info->wait_reassembly_queue, - info->reassembly_data_length >= to_read || - info->transport_status != SMBD_CONNECTED); - if (ret) - return ret; - - /* now we can read from reassembly queue and not sleep */ - page_address = kmap_atomic(page); - to_address = (char *) page_address + page_offset; - - log_read(INFO, "reading from page=%p address=%p to_read=%d\n", - page, to_address, to_read); - - ret = smbd_recv_buf(info, to_address, to_read); - kunmap_atomic(page_address); - - return ret; -} - -/* - * Receive data from transport - * msg: a msghdr point to the buffer, can be ITER_KVEC or ITER_BVEC - * return: total bytes read, or 0. SMB Direct will not do partial read. - */ -int smbd_recv(struct smbd_connection *info, struct msghdr *msg) -{ - char *buf; - struct page *page; - unsigned int to_read, page_offset; - int rc; - - if (iov_iter_rw(&msg->msg_iter) == WRITE) { - /* It's a bug in upper layer to get there */ - cifs_dbg(VFS, "Invalid msg iter dir %u\n", - iov_iter_rw(&msg->msg_iter)); - rc = -EINVAL; - goto out; - } - - switch (iov_iter_type(&msg->msg_iter)) { - case ITER_KVEC: - buf = msg->msg_iter.kvec->iov_base; - to_read = msg->msg_iter.kvec->iov_len; - rc = smbd_recv_buf(info, buf, to_read); - break; - - case ITER_BVEC: - page = msg->msg_iter.bvec->bv_page; - page_offset = msg->msg_iter.bvec->bv_offset; - to_read = msg->msg_iter.bvec->bv_len; - rc = smbd_recv_page(info, page, page_offset, to_read); - break; - - default: - /* It's a bug in upper layer to get there */ - cifs_dbg(VFS, "Invalid msg type %d\n", - iov_iter_type(&msg->msg_iter)); - rc = -EINVAL; - } - -out: - /* SMBDirect will read it all or nothing */ - if (rc > 0) - msg->msg_iter.count = 0; - return rc; -} - -/* * Send data to transport * Each rqst is transported as a SMBDirect payload * rqst: the data to write @@ -1954,12 +2223,14 @@ int smbd_send(struct TCP_Server_Info *server, int num_rqst, struct smb_rqst *rqst_array) { struct smbd_connection *info = server->smbd_conn; + struct smbdirect_socket *sc = &info->socket; + struct smbdirect_socket_parameters *sp = &sc->parameters; struct smb_rqst *rqst; struct iov_iter iter; unsigned int remaining_data_length, klen; int rc, i, rqst_idx; - if (info->transport_status != SMBD_CONNECTED) + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) return -EAGAIN; /* @@ -1971,10 +2242,10 @@ int smbd_send(struct TCP_Server_Info *server, for (i = 0; i < num_rqst; i++) remaining_data_length += smb_rqst_len(server, &rqst_array[i]); - if (unlikely(remaining_data_length > info->max_fragmented_send_size)) { + if (unlikely(remaining_data_length > sp->max_fragmented_send_size)) { /* assertion: payload never exceeds negotiated maximum */ log_write(ERR, "payload size %d > max size %d\n", - remaining_data_length, info->max_fragmented_send_size); + remaining_data_length, sp->max_fragmented_send_size); return -EINVAL; } @@ -2000,14 +2271,14 @@ int smbd_send(struct TCP_Server_Info *server, klen += rqst->rq_iov[i].iov_len; iov_iter_kvec(&iter, ITER_SOURCE, rqst->rq_iov, rqst->rq_nvec, klen); - rc = smbd_post_send_iter(info, &iter, &remaining_data_length); + rc = smbd_post_send_full_iter(sc, &iter, &remaining_data_length); if (rc < 0) break; if (iov_iter_count(&rqst->rq_iter) > 0) { /* And then the data pages if there are any */ - rc = smbd_post_send_iter(info, &rqst->rq_iter, - &remaining_data_length); + rc = smbd_post_send_full_iter(sc, &rqst->rq_iter, + &remaining_data_length); if (rc < 0) break; } @@ -2021,22 +2292,25 @@ int smbd_send(struct TCP_Server_Info *server, * that means all the I/Os have been out and we are good to return */ - wait_event(info->wait_send_pending, - atomic_read(&info->send_pending) == 0); + wait_event(sc->send_io.pending.zero_wait_queue, + atomic_read(&sc->send_io.pending.count) == 0 || + sc->status != SMBDIRECT_SOCKET_CONNECTED); + + if (sc->status != SMBDIRECT_SOCKET_CONNECTED && rc == 0) + rc = -EAGAIN; return rc; } static void register_mr_done(struct ib_cq *cq, struct ib_wc *wc) { - struct smbd_mr *mr; - struct ib_cqe *cqe; + struct smbdirect_mr_io *mr = + container_of(wc->wr_cqe, struct smbdirect_mr_io, cqe); + struct smbdirect_socket *sc = mr->socket; if (wc->status) { log_rdma_mr(ERR, "status=%d\n", wc->status); - cqe = wc->wr_cqe; - mr = container_of(cqe, struct smbd_mr, cqe); - smbd_disconnect_rdma_connection(mr->conn); + smbd_disconnect_rdma_connection(sc); } } @@ -2051,13 +2325,14 @@ static void register_mr_done(struct ib_cq *cq, struct ib_wc *wc) */ static void smbd_mr_recovery_work(struct work_struct *work) { - struct smbd_connection *info = - container_of(work, struct smbd_connection, mr_recovery_work); - struct smbd_mr *smbdirect_mr; + struct smbdirect_socket *sc = + container_of(work, struct smbdirect_socket, mr_io.recovery_work); + struct smbdirect_socket_parameters *sp = &sc->parameters; + struct smbdirect_mr_io *smbdirect_mr; int rc; - list_for_each_entry(smbdirect_mr, &info->mr_list, list) { - if (smbdirect_mr->state == MR_ERROR) { + list_for_each_entry(smbdirect_mr, &sc->mr_io.all.list, list) { + if (smbdirect_mr->state == SMBDIRECT_MR_ERROR) { /* recover this MR entry */ rc = ib_dereg_mr(smbdirect_mr->mr); @@ -2065,25 +2340,25 @@ static void smbd_mr_recovery_work(struct work_struct *work) log_rdma_mr(ERR, "ib_dereg_mr failed rc=%x\n", rc); - smbd_disconnect_rdma_connection(info); + smbd_disconnect_rdma_connection(sc); continue; } smbdirect_mr->mr = ib_alloc_mr( - info->pd, info->mr_type, - info->max_frmr_depth); + sc->ib.pd, sc->mr_io.type, + sp->max_frmr_depth); if (IS_ERR(smbdirect_mr->mr)) { log_rdma_mr(ERR, "ib_alloc_mr failed mr_type=%x max_frmr_depth=%x\n", - info->mr_type, - info->max_frmr_depth); - smbd_disconnect_rdma_connection(info); + sc->mr_io.type, + sp->max_frmr_depth); + smbd_disconnect_rdma_connection(sc); continue; } } else /* This MR is being used, don't recover it */ continue; - smbdirect_mr->state = MR_READY; + smbdirect_mr->state = SMBDIRECT_MR_READY; /* smbdirect_mr->state is updated by this function * and is read and updated by I/O issuing CPUs trying @@ -2092,23 +2367,89 @@ static void smbd_mr_recovery_work(struct work_struct *work) * value is updated before waking up any calls to * get_mr() from the I/O issuing CPUs */ - if (atomic_inc_return(&info->mr_ready_count) == 1) - wake_up_interruptible(&info->wait_mr); + if (atomic_inc_return(&sc->mr_io.ready.count) == 1) + wake_up(&sc->mr_io.ready.wait_queue); } } -static void destroy_mr_list(struct smbd_connection *info) +static void smbd_mr_disable_locked(struct smbdirect_mr_io *mr) { - struct smbd_mr *mr, *tmp; + struct smbdirect_socket *sc = mr->socket; + + lockdep_assert_held(&mr->mutex); + + if (mr->state == SMBDIRECT_MR_DISABLED) + return; - cancel_work_sync(&info->mr_recovery_work); - list_for_each_entry_safe(mr, tmp, &info->mr_list, list) { - if (mr->state == MR_INVALIDATED) - ib_dma_unmap_sg(info->id->device, mr->sgt.sgl, - mr->sgt.nents, mr->dir); + if (mr->mr) ib_dereg_mr(mr->mr); - kfree(mr->sgt.sgl); - kfree(mr); + if (mr->sgt.nents) + ib_dma_unmap_sg(sc->ib.dev, mr->sgt.sgl, mr->sgt.nents, mr->dir); + kfree(mr->sgt.sgl); + + mr->mr = NULL; + mr->sgt.sgl = NULL; + mr->sgt.nents = 0; + + mr->state = SMBDIRECT_MR_DISABLED; +} + +static void smbd_mr_free_locked(struct kref *kref) +{ + struct smbdirect_mr_io *mr = + container_of(kref, struct smbdirect_mr_io, kref); + + lockdep_assert_held(&mr->mutex); + + /* + * smbd_mr_disable_locked() should already be called! + */ + if (WARN_ON_ONCE(mr->state != SMBDIRECT_MR_DISABLED)) + smbd_mr_disable_locked(mr); + + mutex_unlock(&mr->mutex); + mutex_destroy(&mr->mutex); + kfree(mr); +} + +static void destroy_mr_list(struct smbdirect_socket *sc) +{ + struct smbdirect_mr_io *mr, *tmp; + LIST_HEAD(all_list); + unsigned long flags; + + disable_work_sync(&sc->mr_io.recovery_work); + + spin_lock_irqsave(&sc->mr_io.all.lock, flags); + list_splice_tail_init(&sc->mr_io.all.list, &all_list); + spin_unlock_irqrestore(&sc->mr_io.all.lock, flags); + + list_for_each_entry_safe(mr, tmp, &all_list, list) { + mutex_lock(&mr->mutex); + + smbd_mr_disable_locked(mr); + list_del(&mr->list); + mr->socket = NULL; + + /* + * No kref_put_mutex() as it's already locked. + * + * If smbd_mr_free_locked() is called + * and the mutex is unlocked and mr is gone, + * in that case kref_put() returned 1. + * + * If kref_put() returned 0 we know that + * smbd_mr_free_locked() didn't + * run. Not by us nor by anyone else, as we + * still hold the mutex, so we need to unlock. + * + * If the mr is still registered it will + * be dangling (detached from the connection + * waiting for smbd_deregister_mr() to be + * called in order to free the memory. + */ + if (!kref_put(&mr->kref, smbd_mr_free_locked)) + mutex_unlock(&mr->mutex); } } @@ -2119,56 +2460,66 @@ static void destroy_mr_list(struct smbd_connection *info) * Recovery is done in smbd_mr_recovery_work. The content of list entry changes * as MRs are used and recovered for I/O, but the list links will not change */ -static int allocate_mr_list(struct smbd_connection *info) +static int allocate_mr_list(struct smbdirect_socket *sc) { - int i; - struct smbd_mr *smbdirect_mr, *tmp; - - INIT_LIST_HEAD(&info->mr_list); - init_waitqueue_head(&info->wait_mr); - spin_lock_init(&info->mr_list_lock); - atomic_set(&info->mr_ready_count, 0); - atomic_set(&info->mr_used_count, 0); - init_waitqueue_head(&info->wait_for_mr_cleanup); - INIT_WORK(&info->mr_recovery_work, smbd_mr_recovery_work); + struct smbdirect_socket_parameters *sp = &sc->parameters; + struct smbdirect_mr_io *mr; + int ret; + u32 i; + + if (sp->responder_resources == 0) { + log_rdma_mr(ERR, "responder_resources negotiated as 0\n"); + return -EINVAL; + } + /* Allocate more MRs (2x) than hardware responder_resources */ - for (i = 0; i < info->responder_resources * 2; i++) { - smbdirect_mr = kzalloc(sizeof(*smbdirect_mr), GFP_KERNEL); - if (!smbdirect_mr) - goto cleanup_entries; - smbdirect_mr->mr = ib_alloc_mr(info->pd, info->mr_type, - info->max_frmr_depth); - if (IS_ERR(smbdirect_mr->mr)) { + for (i = 0; i < sp->responder_resources * 2; i++) { + mr = kzalloc(sizeof(*mr), GFP_KERNEL); + if (!mr) { + ret = -ENOMEM; + goto kzalloc_mr_failed; + } + + kref_init(&mr->kref); + mutex_init(&mr->mutex); + + mr->mr = ib_alloc_mr(sc->ib.pd, + sc->mr_io.type, + sp->max_frmr_depth); + if (IS_ERR(mr->mr)) { + ret = PTR_ERR(mr->mr); log_rdma_mr(ERR, "ib_alloc_mr failed mr_type=%x max_frmr_depth=%x\n", - info->mr_type, info->max_frmr_depth); - goto out; + sc->mr_io.type, sp->max_frmr_depth); + goto ib_alloc_mr_failed; } - smbdirect_mr->sgt.sgl = kcalloc(info->max_frmr_depth, - sizeof(struct scatterlist), - GFP_KERNEL); - if (!smbdirect_mr->sgt.sgl) { + + mr->sgt.sgl = kcalloc(sp->max_frmr_depth, + sizeof(struct scatterlist), + GFP_KERNEL); + if (!mr->sgt.sgl) { + ret = -ENOMEM; log_rdma_mr(ERR, "failed to allocate sgl\n"); - ib_dereg_mr(smbdirect_mr->mr); - goto out; + goto kcalloc_sgl_failed; } - smbdirect_mr->state = MR_READY; - smbdirect_mr->conn = info; + mr->state = SMBDIRECT_MR_READY; + mr->socket = sc; - list_add_tail(&smbdirect_mr->list, &info->mr_list); - atomic_inc(&info->mr_ready_count); + list_add_tail(&mr->list, &sc->mr_io.all.list); + atomic_inc(&sc->mr_io.ready.count); } + + INIT_WORK(&sc->mr_io.recovery_work, smbd_mr_recovery_work); + return 0; -out: - kfree(smbdirect_mr); -cleanup_entries: - list_for_each_entry_safe(smbdirect_mr, tmp, &info->mr_list, list) { - list_del(&smbdirect_mr->list); - ib_dereg_mr(smbdirect_mr->mr); - kfree(smbdirect_mr->sgt.sgl); - kfree(smbdirect_mr); - } - return -ENOMEM; +kcalloc_sgl_failed: + ib_dereg_mr(mr->mr); +ib_alloc_mr_failed: + mutex_destroy(&mr->mutex); + kfree(mr); +kzalloc_mr_failed: + destroy_mr_list(sc); + return ret; } /* @@ -2179,37 +2530,38 @@ cleanup_entries: * issuing I/O trying to get MR at the same time, mr_list_lock is used to * protect this situation. */ -static struct smbd_mr *get_mr(struct smbd_connection *info) +static struct smbdirect_mr_io *get_mr(struct smbdirect_socket *sc) { - struct smbd_mr *ret; + struct smbdirect_mr_io *ret; + unsigned long flags; int rc; again: - rc = wait_event_interruptible(info->wait_mr, - atomic_read(&info->mr_ready_count) || - info->transport_status != SMBD_CONNECTED); + rc = wait_event_interruptible(sc->mr_io.ready.wait_queue, + atomic_read(&sc->mr_io.ready.count) || + sc->status != SMBDIRECT_SOCKET_CONNECTED); if (rc) { log_rdma_mr(ERR, "wait_event_interruptible rc=%x\n", rc); return NULL; } - if (info->transport_status != SMBD_CONNECTED) { - log_rdma_mr(ERR, "info->transport_status=%x\n", - info->transport_status); + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) { + log_rdma_mr(ERR, "sc->status=%x\n", sc->status); return NULL; } - spin_lock(&info->mr_list_lock); - list_for_each_entry(ret, &info->mr_list, list) { - if (ret->state == MR_READY) { - ret->state = MR_REGISTERED; - spin_unlock(&info->mr_list_lock); - atomic_dec(&info->mr_ready_count); - atomic_inc(&info->mr_used_count); + spin_lock_irqsave(&sc->mr_io.all.lock, flags); + list_for_each_entry(ret, &sc->mr_io.all.list, list) { + if (ret->state == SMBDIRECT_MR_READY) { + ret->state = SMBDIRECT_MR_REGISTERED; + kref_get(&ret->kref); + spin_unlock_irqrestore(&sc->mr_io.all.lock, flags); + atomic_dec(&sc->mr_io.ready.count); + atomic_inc(&sc->mr_io.used.count); return ret; } } - spin_unlock(&info->mr_list_lock); + spin_unlock_irqrestore(&sc->mr_io.all.lock, flags); /* * It is possible that we could fail to get MR because other processes may * try to acquire a MR at the same time. If this is the case, retry it. @@ -2220,8 +2572,7 @@ again: /* * Transcribe the pages from an iterator into an MR scatterlist. */ -static int smbd_iter_to_mr(struct smbd_connection *info, - struct iov_iter *iter, +static int smbd_iter_to_mr(struct iov_iter *iter, struct sg_table *sgt, unsigned int max_sg) { @@ -2243,66 +2594,65 @@ static int smbd_iter_to_mr(struct smbd_connection *info, * need_invalidate: true if this MR needs to be locally invalidated after I/O * return value: the MR registered, NULL if failed. */ -struct smbd_mr *smbd_register_mr(struct smbd_connection *info, +struct smbdirect_mr_io *smbd_register_mr(struct smbd_connection *info, struct iov_iter *iter, bool writing, bool need_invalidate) { - struct smbd_mr *smbdirect_mr; + struct smbdirect_socket *sc = &info->socket; + struct smbdirect_socket_parameters *sp = &sc->parameters; + struct smbdirect_mr_io *mr; int rc, num_pages; - enum dma_data_direction dir; struct ib_reg_wr *reg_wr; - num_pages = iov_iter_npages(iter, info->max_frmr_depth + 1); - if (num_pages > info->max_frmr_depth) { + num_pages = iov_iter_npages(iter, sp->max_frmr_depth + 1); + if (num_pages > sp->max_frmr_depth) { log_rdma_mr(ERR, "num_pages=%d max_frmr_depth=%d\n", - num_pages, info->max_frmr_depth); + num_pages, sp->max_frmr_depth); WARN_ON_ONCE(1); return NULL; } - smbdirect_mr = get_mr(info); - if (!smbdirect_mr) { + mr = get_mr(sc); + if (!mr) { log_rdma_mr(ERR, "get_mr returning NULL\n"); return NULL; } - dir = writing ? DMA_FROM_DEVICE : DMA_TO_DEVICE; - smbdirect_mr->dir = dir; - smbdirect_mr->need_invalidate = need_invalidate; - smbdirect_mr->sgt.nents = 0; - smbdirect_mr->sgt.orig_nents = 0; + mutex_lock(&mr->mutex); + + mr->dir = writing ? DMA_FROM_DEVICE : DMA_TO_DEVICE; + mr->need_invalidate = need_invalidate; + mr->sgt.nents = 0; + mr->sgt.orig_nents = 0; log_rdma_mr(INFO, "num_pages=0x%x count=0x%zx depth=%u\n", - num_pages, iov_iter_count(iter), info->max_frmr_depth); - smbd_iter_to_mr(info, iter, &smbdirect_mr->sgt, info->max_frmr_depth); + num_pages, iov_iter_count(iter), sp->max_frmr_depth); + smbd_iter_to_mr(iter, &mr->sgt, sp->max_frmr_depth); - rc = ib_dma_map_sg(info->id->device, smbdirect_mr->sgt.sgl, - smbdirect_mr->sgt.nents, dir); + rc = ib_dma_map_sg(sc->ib.dev, mr->sgt.sgl, mr->sgt.nents, mr->dir); if (!rc) { log_rdma_mr(ERR, "ib_dma_map_sg num_pages=%x dir=%x rc=%x\n", - num_pages, dir, rc); + num_pages, mr->dir, rc); goto dma_map_error; } - rc = ib_map_mr_sg(smbdirect_mr->mr, smbdirect_mr->sgt.sgl, - smbdirect_mr->sgt.nents, NULL, PAGE_SIZE); - if (rc != smbdirect_mr->sgt.nents) { + rc = ib_map_mr_sg(mr->mr, mr->sgt.sgl, mr->sgt.nents, NULL, PAGE_SIZE); + if (rc != mr->sgt.nents) { log_rdma_mr(ERR, - "ib_map_mr_sg failed rc = %d nents = %x\n", - rc, smbdirect_mr->sgt.nents); + "ib_map_mr_sg failed rc = %d nents = %x\n", + rc, mr->sgt.nents); goto map_mr_error; } - ib_update_fast_reg_key(smbdirect_mr->mr, - ib_inc_rkey(smbdirect_mr->mr->rkey)); - reg_wr = &smbdirect_mr->wr; + ib_update_fast_reg_key(mr->mr, ib_inc_rkey(mr->mr->rkey)); + reg_wr = &mr->wr; reg_wr->wr.opcode = IB_WR_REG_MR; - smbdirect_mr->cqe.done = register_mr_done; - reg_wr->wr.wr_cqe = &smbdirect_mr->cqe; + mr->cqe.done = register_mr_done; + reg_wr->wr.wr_cqe = &mr->cqe; reg_wr->wr.num_sge = 0; reg_wr->wr.send_flags = IB_SEND_SIGNALED; - reg_wr->mr = smbdirect_mr->mr; - reg_wr->key = smbdirect_mr->mr->rkey; + reg_wr->mr = mr->mr; + reg_wr->key = mr->mr->rkey; reg_wr->access = writing ? IB_ACCESS_REMOTE_WRITE | IB_ACCESS_LOCAL_WRITE : IB_ACCESS_REMOTE_READ; @@ -2312,39 +2662,66 @@ struct smbd_mr *smbd_register_mr(struct smbd_connection *info, * on IB_WR_REG_MR. Hardware enforces a barrier and order of execution * on the next ib_post_send when we actually send I/O to remote peer */ - rc = ib_post_send(info->id->qp, ®_wr->wr, NULL); - if (!rc) - return smbdirect_mr; + rc = ib_post_send(sc->ib.qp, ®_wr->wr, NULL); + if (!rc) { + /* + * get_mr() gave us a reference + * via kref_get(&mr->kref), we keep that and let + * the caller use smbd_deregister_mr() + * to remove it again. + */ + mutex_unlock(&mr->mutex); + return mr; + } log_rdma_mr(ERR, "ib_post_send failed rc=%x reg_wr->key=%x\n", rc, reg_wr->key); - /* If all failed, attempt to recover this MR by setting it MR_ERROR*/ + /* If all failed, attempt to recover this MR by setting it SMBDIRECT_MR_ERROR*/ map_mr_error: - ib_dma_unmap_sg(info->id->device, smbdirect_mr->sgt.sgl, - smbdirect_mr->sgt.nents, smbdirect_mr->dir); + ib_dma_unmap_sg(sc->ib.dev, mr->sgt.sgl, mr->sgt.nents, mr->dir); dma_map_error: - smbdirect_mr->state = MR_ERROR; - if (atomic_dec_and_test(&info->mr_used_count)) - wake_up(&info->wait_for_mr_cleanup); + mr->sgt.nents = 0; + mr->state = SMBDIRECT_MR_ERROR; + if (atomic_dec_and_test(&sc->mr_io.used.count)) + wake_up(&sc->mr_io.cleanup.wait_queue); - smbd_disconnect_rdma_connection(info); + smbd_disconnect_rdma_connection(sc); + + /* + * get_mr() gave us a reference + * via kref_get(&mr->kref), we need to remove it again + * on error. + * + * No kref_put_mutex() as it's already locked. + * + * If smbd_mr_free_locked() is called + * and the mutex is unlocked and mr is gone, + * in that case kref_put() returned 1. + * + * If kref_put() returned 0 we know that + * smbd_mr_free_locked() didn't + * run. Not by us nor by anyone else, as we + * still hold the mutex, so we need to unlock. + */ + if (!kref_put(&mr->kref, smbd_mr_free_locked)) + mutex_unlock(&mr->mutex); return NULL; } static void local_inv_done(struct ib_cq *cq, struct ib_wc *wc) { - struct smbd_mr *smbdirect_mr; + struct smbdirect_mr_io *smbdirect_mr; struct ib_cqe *cqe; cqe = wc->wr_cqe; - smbdirect_mr = container_of(cqe, struct smbd_mr, cqe); - smbdirect_mr->state = MR_INVALIDATED; + smbdirect_mr = container_of(cqe, struct smbdirect_mr_io, cqe); + smbdirect_mr->state = SMBDIRECT_MR_INVALIDATED; if (wc->status != IB_WC_SUCCESS) { log_rdma_mr(ERR, "invalidate failed status=%x\n", wc->status); - smbdirect_mr->state = MR_ERROR; + smbdirect_mr->state = SMBDIRECT_MR_ERROR; } complete(&smbdirect_mr->invalidate_done); } @@ -2355,58 +2732,85 @@ static void local_inv_done(struct ib_cq *cq, struct ib_wc *wc) * and we have to locally invalidate the buffer to prevent data is being * modified by remote peer after upper layer consumes it */ -int smbd_deregister_mr(struct smbd_mr *smbdirect_mr) +void smbd_deregister_mr(struct smbdirect_mr_io *mr) { - struct ib_send_wr *wr; - struct smbd_connection *info = smbdirect_mr->conn; - int rc = 0; + struct smbdirect_socket *sc = mr->socket; + + mutex_lock(&mr->mutex); + if (mr->state == SMBDIRECT_MR_DISABLED) + goto put_kref; + + if (sc->status != SMBDIRECT_SOCKET_CONNECTED) { + smbd_mr_disable_locked(mr); + goto put_kref; + } + + if (mr->need_invalidate) { + struct ib_send_wr *wr = &mr->inv_wr; + int rc; - if (smbdirect_mr->need_invalidate) { /* Need to finish local invalidation before returning */ - wr = &smbdirect_mr->inv_wr; wr->opcode = IB_WR_LOCAL_INV; - smbdirect_mr->cqe.done = local_inv_done; - wr->wr_cqe = &smbdirect_mr->cqe; + mr->cqe.done = local_inv_done; + wr->wr_cqe = &mr->cqe; wr->num_sge = 0; - wr->ex.invalidate_rkey = smbdirect_mr->mr->rkey; + wr->ex.invalidate_rkey = mr->mr->rkey; wr->send_flags = IB_SEND_SIGNALED; - init_completion(&smbdirect_mr->invalidate_done); - rc = ib_post_send(info->id->qp, wr, NULL); + init_completion(&mr->invalidate_done); + rc = ib_post_send(sc->ib.qp, wr, NULL); if (rc) { log_rdma_mr(ERR, "ib_post_send failed rc=%x\n", rc); - smbd_disconnect_rdma_connection(info); + smbd_mr_disable_locked(mr); + smbd_disconnect_rdma_connection(sc); goto done; } - wait_for_completion(&smbdirect_mr->invalidate_done); - smbdirect_mr->need_invalidate = false; + wait_for_completion(&mr->invalidate_done); + mr->need_invalidate = false; } else /* - * For remote invalidation, just set it to MR_INVALIDATED + * For remote invalidation, just set it to SMBDIRECT_MR_INVALIDATED * and defer to mr_recovery_work to recover the MR for next use */ - smbdirect_mr->state = MR_INVALIDATED; - - if (smbdirect_mr->state == MR_INVALIDATED) { - ib_dma_unmap_sg( - info->id->device, smbdirect_mr->sgt.sgl, - smbdirect_mr->sgt.nents, - smbdirect_mr->dir); - smbdirect_mr->state = MR_READY; - if (atomic_inc_return(&info->mr_ready_count) == 1) - wake_up_interruptible(&info->wait_mr); + mr->state = SMBDIRECT_MR_INVALIDATED; + + if (mr->sgt.nents) { + ib_dma_unmap_sg(sc->ib.dev, mr->sgt.sgl, mr->sgt.nents, mr->dir); + mr->sgt.nents = 0; + } + + if (mr->state == SMBDIRECT_MR_INVALIDATED) { + mr->state = SMBDIRECT_MR_READY; + if (atomic_inc_return(&sc->mr_io.ready.count) == 1) + wake_up(&sc->mr_io.ready.wait_queue); } else /* * Schedule the work to do MR recovery for future I/Os MR * recovery is slow and don't want it to block current I/O */ - queue_work(info->workqueue, &info->mr_recovery_work); + queue_work(sc->workqueue, &sc->mr_io.recovery_work); done: - if (atomic_dec_and_test(&info->mr_used_count)) - wake_up(&info->wait_for_mr_cleanup); + if (atomic_dec_and_test(&sc->mr_io.used.count)) + wake_up(&sc->mr_io.cleanup.wait_queue); - return rc; +put_kref: + /* + * No kref_put_mutex() as it's already locked. + * + * If smbd_mr_free_locked() is called + * and the mutex is unlocked and mr is gone, + * in that case kref_put() returned 1. + * + * If kref_put() returned 0 we know that + * smbd_mr_free_locked() didn't + * run. Not by us nor by anyone else, as we + * still hold the mutex, so we need to unlock + * and keep the mr in SMBDIRECT_MR_READY or + * SMBDIRECT_MR_ERROR state. + */ + if (!kref_put(&mr->kref, smbd_mr_free_locked)) + mutex_unlock(&mr->mutex); } static bool smb_set_sge(struct smb_extract_to_rdma *rdma, @@ -2552,13 +2956,14 @@ static ssize_t smb_extract_folioq_to_rdma(struct iov_iter *iter, size_t fsize = folioq_folio_size(folioq, slot); if (offset < fsize) { - size_t part = umin(maxsize - ret, fsize - offset); + size_t part = umin(maxsize, fsize - offset); if (!smb_set_sge(rdma, folio_page(folio, 0), offset, part)) return -EIO; offset += part; ret += part; + maxsize -= part; } if (offset >= fsize) { @@ -2573,7 +2978,7 @@ static ssize_t smb_extract_folioq_to_rdma(struct iov_iter *iter, slot = 0; } } - } while (rdma->nr_sge < rdma->max_sge || maxsize > 0); + } while (rdma->nr_sge < rdma->max_sge && maxsize > 0); iter->folioq = folioq; iter->folioq_slot = slot; |