From 9c3da0991754d480328eeaa2b90cb231a1cea9b6 Mon Sep 17 00:00:00 2001 From: Harvey Harrison Date: Sat, 17 Jan 2009 17:11:57 -0800 Subject: IB: Remove __constant_{endian} uses The base versions handle constant folding just fine, use them directly. The replacements are OK in the include/ files as they are not exported to userspace so we don't need the __ prefixed versions. This patch does not affect code generation at all. Signed-off-by: Harvey Harrison Signed-off-by: Roland Dreier --- drivers/infiniband/core/cm.c | 15 ++--- drivers/infiniband/core/cm_msgs.h | 22 +++---- drivers/infiniband/core/mad_rmpp.c | 2 +- drivers/infiniband/hw/cxgb3/iwch_qp.c | 4 +- drivers/infiniband/hw/ehca/ehca_sqp.c | 8 +-- drivers/infiniband/hw/ipath/ipath_eeprom.c | 4 +- drivers/infiniband/hw/ipath/ipath_mad.c | 95 +++++++++++++-------------- drivers/infiniband/hw/ipath/ipath_rc.c | 2 +- drivers/infiniband/hw/ipath/ipath_sdma.c | 4 +- drivers/infiniband/hw/ipath/ipath_uc.c | 2 +- drivers/infiniband/hw/ipath/ipath_ud.c | 4 +- drivers/infiniband/hw/ipath/ipath_user_sdma.c | 6 +- drivers/infiniband/hw/ipath/ipath_verbs.c | 2 +- drivers/infiniband/hw/ipath/ipath_verbs.h | 10 +-- drivers/infiniband/hw/mlx4/qp.c | 22 +++---- 15 files changed, 100 insertions(+), 102 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index f1e82a92e61..5130fc55b8e 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -927,8 +927,7 @@ int ib_cm_listen(struct ib_cm_id *cm_id, __be64 service_id, __be64 service_mask, unsigned long flags; int ret = 0; - service_mask = service_mask ? service_mask : - __constant_cpu_to_be64(~0ULL); + service_mask = service_mask ? service_mask : ~cpu_to_be64(0); service_id &= service_mask; if ((service_id & IB_SERVICE_ID_AGN_MASK) == IB_CM_ASSIGN_SERVICE_ID && (service_id != IB_CM_ASSIGN_SERVICE_ID)) @@ -954,7 +953,7 @@ int ib_cm_listen(struct ib_cm_id *cm_id, __be64 service_id, __be64 service_mask, spin_lock_irqsave(&cm.lock, flags); if (service_id == IB_CM_ASSIGN_SERVICE_ID) { cm_id->service_id = cpu_to_be64(cm.listen_service_id++); - cm_id->service_mask = __constant_cpu_to_be64(~0ULL); + cm_id->service_mask = ~cpu_to_be64(0); } else { cm_id->service_id = service_id; cm_id->service_mask = service_mask; @@ -1134,7 +1133,7 @@ int ib_send_cm_req(struct ib_cm_id *cm_id, goto error1; } cm_id->service_id = param->service_id; - cm_id->service_mask = __constant_cpu_to_be64(~0ULL); + cm_id->service_mask = ~cpu_to_be64(0); cm_id_priv->timeout_ms = cm_convert_to_ms( param->primary_path->packet_life_time) * 2 + cm_convert_to_ms( @@ -1545,7 +1544,7 @@ static int cm_req_handler(struct cm_work *work) cm_id_priv->id.cm_handler = listen_cm_id_priv->id.cm_handler; cm_id_priv->id.context = listen_cm_id_priv->id.context; cm_id_priv->id.service_id = req_msg->service_id; - cm_id_priv->id.service_mask = __constant_cpu_to_be64(~0ULL); + cm_id_priv->id.service_mask = ~cpu_to_be64(0); cm_process_routed_req(req_msg, work->mad_recv_wc->wc); cm_format_paths_from_req(req_msg, &work->path[0], &work->path[1]); @@ -2898,7 +2897,7 @@ int ib_send_cm_sidr_req(struct ib_cm_id *cm_id, goto out; cm_id->service_id = param->service_id; - cm_id->service_mask = __constant_cpu_to_be64(~0ULL); + cm_id->service_mask = ~cpu_to_be64(0); cm_id_priv->timeout_ms = param->timeout_ms; cm_id_priv->max_cm_retries = param->max_cm_retries; ret = cm_alloc_msg(cm_id_priv, &msg); @@ -2992,7 +2991,7 @@ static int cm_sidr_req_handler(struct cm_work *work) cm_id_priv->id.cm_handler = cur_cm_id_priv->id.cm_handler; cm_id_priv->id.context = cur_cm_id_priv->id.context; cm_id_priv->id.service_id = sidr_req_msg->service_id; - cm_id_priv->id.service_mask = __constant_cpu_to_be64(~0ULL); + cm_id_priv->id.service_mask = ~cpu_to_be64(0); cm_format_sidr_req_event(work, &cur_cm_id_priv->id); cm_process_work(cm_id_priv, work); @@ -3789,7 +3788,7 @@ static int __init ib_cm_init(void) rwlock_init(&cm.device_lock); spin_lock_init(&cm.lock); cm.listen_service_table = RB_ROOT; - cm.listen_service_id = __constant_be64_to_cpu(IB_CM_ASSIGN_SERVICE_ID); + cm.listen_service_id = be64_to_cpu(IB_CM_ASSIGN_SERVICE_ID); cm.remote_id_table = RB_ROOT; cm.remote_qp_table = RB_ROOT; cm.remote_sidr_table = RB_ROOT; diff --git a/drivers/infiniband/core/cm_msgs.h b/drivers/infiniband/core/cm_msgs.h index aec9c7af825..7e63c08f697 100644 --- a/drivers/infiniband/core/cm_msgs.h +++ b/drivers/infiniband/core/cm_msgs.h @@ -44,17 +44,17 @@ #define IB_CM_CLASS_VERSION 2 /* IB specification 1.2 */ -#define CM_REQ_ATTR_ID __constant_htons(0x0010) -#define CM_MRA_ATTR_ID __constant_htons(0x0011) -#define CM_REJ_ATTR_ID __constant_htons(0x0012) -#define CM_REP_ATTR_ID __constant_htons(0x0013) -#define CM_RTU_ATTR_ID __constant_htons(0x0014) -#define CM_DREQ_ATTR_ID __constant_htons(0x0015) -#define CM_DREP_ATTR_ID __constant_htons(0x0016) -#define CM_SIDR_REQ_ATTR_ID __constant_htons(0x0017) -#define CM_SIDR_REP_ATTR_ID __constant_htons(0x0018) -#define CM_LAP_ATTR_ID __constant_htons(0x0019) -#define CM_APR_ATTR_ID __constant_htons(0x001A) +#define CM_REQ_ATTR_ID cpu_to_be16(0x0010) +#define CM_MRA_ATTR_ID cpu_to_be16(0x0011) +#define CM_REJ_ATTR_ID cpu_to_be16(0x0012) +#define CM_REP_ATTR_ID cpu_to_be16(0x0013) +#define CM_RTU_ATTR_ID cpu_to_be16(0x0014) +#define CM_DREQ_ATTR_ID cpu_to_be16(0x0015) +#define CM_DREP_ATTR_ID cpu_to_be16(0x0016) +#define CM_SIDR_REQ_ATTR_ID cpu_to_be16(0x0017) +#define CM_SIDR_REP_ATTR_ID cpu_to_be16(0x0018) +#define CM_LAP_ATTR_ID cpu_to_be16(0x0019) +#define CM_APR_ATTR_ID cpu_to_be16(0x001A) enum cm_msg_sequence { CM_MSG_SEQUENCE_REQ, diff --git a/drivers/infiniband/core/mad_rmpp.c b/drivers/infiniband/core/mad_rmpp.c index 3af2b84cd83..57a3c6f947b 100644 --- a/drivers/infiniband/core/mad_rmpp.c +++ b/drivers/infiniband/core/mad_rmpp.c @@ -735,7 +735,7 @@ process_rmpp_data(struct ib_mad_agent_private *agent, goto bad; } - if (rmpp_hdr->seg_num == __constant_htonl(1)) { + if (rmpp_hdr->seg_num == cpu_to_be32(1)) { if (!(ib_get_rmpp_flags(rmpp_hdr) & IB_MGMT_RMPP_FLAG_FIRST)) { rmpp_status = IB_MGMT_RMPP_STATUS_BAD_SEG; goto bad; diff --git a/drivers/infiniband/hw/cxgb3/iwch_qp.c b/drivers/infiniband/hw/cxgb3/iwch_qp.c index 19661b2f040..48e2b0bcabd 100644 --- a/drivers/infiniband/hw/cxgb3/iwch_qp.c +++ b/drivers/infiniband/hw/cxgb3/iwch_qp.c @@ -99,8 +99,8 @@ static int build_rdma_write(union t3_wr *wqe, struct ib_send_wr *wr, if (wr->opcode == IB_WR_RDMA_WRITE_WITH_IMM) { plen = 4; wqe->write.sgl[0].stag = wr->ex.imm_data; - wqe->write.sgl[0].len = __constant_cpu_to_be32(0); - wqe->write.num_sgle = __constant_cpu_to_be32(0); + wqe->write.sgl[0].len = cpu_to_be32(0); + wqe->write.num_sgle = cpu_to_be32(0); *flit_cnt = 6; } else { plen = 0; diff --git a/drivers/infiniband/hw/ehca/ehca_sqp.c b/drivers/infiniband/hw/ehca/ehca_sqp.c index 44447aaa550..c568b28f4e2 100644 --- a/drivers/infiniband/hw/ehca/ehca_sqp.c +++ b/drivers/infiniband/hw/ehca/ehca_sqp.c @@ -46,11 +46,11 @@ #include "ehca_iverbs.h" #include "hcp_if.h" -#define IB_MAD_STATUS_REDIRECT __constant_htons(0x0002) -#define IB_MAD_STATUS_UNSUP_VERSION __constant_htons(0x0004) -#define IB_MAD_STATUS_UNSUP_METHOD __constant_htons(0x0008) +#define IB_MAD_STATUS_REDIRECT cpu_to_be16(0x0002) +#define IB_MAD_STATUS_UNSUP_VERSION cpu_to_be16(0x0004) +#define IB_MAD_STATUS_UNSUP_METHOD cpu_to_be16(0x0008) -#define IB_PMA_CLASS_PORT_INFO __constant_htons(0x0001) +#define IB_PMA_CLASS_PORT_INFO cpu_to_be16(0x0001) /** * ehca_define_sqp - Defines special queue pair 1 (GSI QP). When special queue diff --git a/drivers/infiniband/hw/ipath/ipath_eeprom.c b/drivers/infiniband/hw/ipath/ipath_eeprom.c index dc37277f1c8..fc7181985e8 100644 --- a/drivers/infiniband/hw/ipath/ipath_eeprom.c +++ b/drivers/infiniband/hw/ipath/ipath_eeprom.c @@ -772,8 +772,8 @@ void ipath_get_eeprom_info(struct ipath_devdata *dd) "0x%x, not 0x%x\n", csum, ifp->if_csum); goto done; } - if (*(__be64 *) ifp->if_guid == 0ULL || - *(__be64 *) ifp->if_guid == __constant_cpu_to_be64(-1LL)) { + if (*(__be64 *) ifp->if_guid == cpu_to_be64(0) || + *(__be64 *) ifp->if_guid == ~cpu_to_be64(0)) { ipath_dev_err(dd, "Invalid GUID %llx from flash; " "ignoring\n", *(unsigned long long *) ifp->if_guid); diff --git a/drivers/infiniband/hw/ipath/ipath_mad.c b/drivers/infiniband/hw/ipath/ipath_mad.c index 17a12319747..16a702d4601 100644 --- a/drivers/infiniband/hw/ipath/ipath_mad.c +++ b/drivers/infiniband/hw/ipath/ipath_mad.c @@ -37,10 +37,10 @@ #include "ipath_verbs.h" #include "ipath_common.h" -#define IB_SMP_UNSUP_VERSION __constant_htons(0x0004) -#define IB_SMP_UNSUP_METHOD __constant_htons(0x0008) -#define IB_SMP_UNSUP_METH_ATTR __constant_htons(0x000C) -#define IB_SMP_INVALID_FIELD __constant_htons(0x001C) +#define IB_SMP_UNSUP_VERSION cpu_to_be16(0x0004) +#define IB_SMP_UNSUP_METHOD cpu_to_be16(0x0008) +#define IB_SMP_UNSUP_METH_ATTR cpu_to_be16(0x000C) +#define IB_SMP_INVALID_FIELD cpu_to_be16(0x001C) static int reply(struct ib_smp *smp) { @@ -789,12 +789,12 @@ static int recv_subn_set_pkeytable(struct ib_smp *smp, return recv_subn_get_pkeytable(smp, ibdev); } -#define IB_PMA_CLASS_PORT_INFO __constant_htons(0x0001) -#define IB_PMA_PORT_SAMPLES_CONTROL __constant_htons(0x0010) -#define IB_PMA_PORT_SAMPLES_RESULT __constant_htons(0x0011) -#define IB_PMA_PORT_COUNTERS __constant_htons(0x0012) -#define IB_PMA_PORT_COUNTERS_EXT __constant_htons(0x001D) -#define IB_PMA_PORT_SAMPLES_RESULT_EXT __constant_htons(0x001E) +#define IB_PMA_CLASS_PORT_INFO cpu_to_be16(0x0001) +#define IB_PMA_PORT_SAMPLES_CONTROL cpu_to_be16(0x0010) +#define IB_PMA_PORT_SAMPLES_RESULT cpu_to_be16(0x0011) +#define IB_PMA_PORT_COUNTERS cpu_to_be16(0x0012) +#define IB_PMA_PORT_COUNTERS_EXT cpu_to_be16(0x001D) +#define IB_PMA_PORT_SAMPLES_RESULT_EXT cpu_to_be16(0x001E) struct ib_perf { u8 base_version; @@ -884,19 +884,19 @@ struct ib_pma_portcounters { __be32 port_rcv_packets; } __attribute__ ((packed)); -#define IB_PMA_SEL_SYMBOL_ERROR __constant_htons(0x0001) -#define IB_PMA_SEL_LINK_ERROR_RECOVERY __constant_htons(0x0002) -#define IB_PMA_SEL_LINK_DOWNED __constant_htons(0x0004) -#define IB_PMA_SEL_PORT_RCV_ERRORS __constant_htons(0x0008) -#define IB_PMA_SEL_PORT_RCV_REMPHYS_ERRORS __constant_htons(0x0010) -#define IB_PMA_SEL_PORT_XMIT_DISCARDS __constant_htons(0x0040) -#define IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS __constant_htons(0x0200) -#define IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS __constant_htons(0x0400) -#define IB_PMA_SEL_PORT_VL15_DROPPED __constant_htons(0x0800) -#define IB_PMA_SEL_PORT_XMIT_DATA __constant_htons(0x1000) -#define IB_PMA_SEL_PORT_RCV_DATA __constant_htons(0x2000) -#define IB_PMA_SEL_PORT_XMIT_PACKETS __constant_htons(0x4000) -#define IB_PMA_SEL_PORT_RCV_PACKETS __constant_htons(0x8000) +#define IB_PMA_SEL_SYMBOL_ERROR cpu_to_be16(0x0001) +#define IB_PMA_SEL_LINK_ERROR_RECOVERY cpu_to_be16(0x0002) +#define IB_PMA_SEL_LINK_DOWNED cpu_to_be16(0x0004) +#define IB_PMA_SEL_PORT_RCV_ERRORS cpu_to_be16(0x0008) +#define IB_PMA_SEL_PORT_RCV_REMPHYS_ERRORS cpu_to_be16(0x0010) +#define IB_PMA_SEL_PORT_XMIT_DISCARDS cpu_to_be16(0x0040) +#define IB_PMA_SEL_LOCAL_LINK_INTEGRITY_ERRORS cpu_to_be16(0x0200) +#define IB_PMA_SEL_EXCESSIVE_BUFFER_OVERRUNS cpu_to_be16(0x0400) +#define IB_PMA_SEL_PORT_VL15_DROPPED cpu_to_be16(0x0800) +#define IB_PMA_SEL_PORT_XMIT_DATA cpu_to_be16(0x1000) +#define IB_PMA_SEL_PORT_RCV_DATA cpu_to_be16(0x2000) +#define IB_PMA_SEL_PORT_XMIT_PACKETS cpu_to_be16(0x4000) +#define IB_PMA_SEL_PORT_RCV_PACKETS cpu_to_be16(0x8000) struct ib_pma_portcounters_ext { u8 reserved; @@ -913,14 +913,14 @@ struct ib_pma_portcounters_ext { __be64 port_multicast_rcv_packets; } __attribute__ ((packed)); -#define IB_PMA_SELX_PORT_XMIT_DATA __constant_htons(0x0001) -#define IB_PMA_SELX_PORT_RCV_DATA __constant_htons(0x0002) -#define IB_PMA_SELX_PORT_XMIT_PACKETS __constant_htons(0x0004) -#define IB_PMA_SELX_PORT_RCV_PACKETS __constant_htons(0x0008) -#define IB_PMA_SELX_PORT_UNI_XMIT_PACKETS __constant_htons(0x0010) -#define IB_PMA_SELX_PORT_UNI_RCV_PACKETS __constant_htons(0x0020) -#define IB_PMA_SELX_PORT_MULTI_XMIT_PACKETS __constant_htons(0x0040) -#define IB_PMA_SELX_PORT_MULTI_RCV_PACKETS __constant_htons(0x0080) +#define IB_PMA_SELX_PORT_XMIT_DATA cpu_to_be16(0x0001) +#define IB_PMA_SELX_PORT_RCV_DATA cpu_to_be16(0x0002) +#define IB_PMA_SELX_PORT_XMIT_PACKETS cpu_to_be16(0x0004) +#define IB_PMA_SELX_PORT_RCV_PACKETS cpu_to_be16(0x0008) +#define IB_PMA_SELX_PORT_UNI_XMIT_PACKETS cpu_to_be16(0x0010) +#define IB_PMA_SELX_PORT_UNI_RCV_PACKETS cpu_to_be16(0x0020) +#define IB_PMA_SELX_PORT_MULTI_XMIT_PACKETS cpu_to_be16(0x0040) +#define IB_PMA_SELX_PORT_MULTI_RCV_PACKETS cpu_to_be16(0x0080) static int recv_pma_get_classportinfo(struct ib_perf *pmp) { @@ -933,7 +933,7 @@ static int recv_pma_get_classportinfo(struct ib_perf *pmp) pmp->status |= IB_SMP_INVALID_FIELD; /* Indicate AllPortSelect is valid (only one port anyway) */ - p->cap_mask = __constant_cpu_to_be16(1 << 8); + p->cap_mask = cpu_to_be16(1 << 8); p->base_version = 1; p->class_version = 1; /* @@ -951,12 +951,11 @@ static int recv_pma_get_classportinfo(struct ib_perf *pmp) * We support 5 counters which only count the mandatory quantities. */ #define COUNTER_MASK(q, n) (q << ((9 - n) * 3)) -#define COUNTER_MASK0_9 \ - __constant_cpu_to_be32(COUNTER_MASK(1, 0) | \ - COUNTER_MASK(1, 1) | \ - COUNTER_MASK(1, 2) | \ - COUNTER_MASK(1, 3) | \ - COUNTER_MASK(1, 4)) +#define COUNTER_MASK0_9 cpu_to_be32(COUNTER_MASK(1, 0) | \ + COUNTER_MASK(1, 1) | \ + COUNTER_MASK(1, 2) | \ + COUNTER_MASK(1, 3) | \ + COUNTER_MASK(1, 4)) static int recv_pma_get_portsamplescontrol(struct ib_perf *pmp, struct ib_device *ibdev, u8 port) @@ -1137,7 +1136,7 @@ static int recv_pma_get_portsamplesresult_ext(struct ib_perf *pmp, status = dev->pma_sample_status; p->sample_status = cpu_to_be16(status); /* 64 bits */ - p->extended_width = __constant_cpu_to_be32(0x80000000); + p->extended_width = cpu_to_be32(0x80000000); for (i = 0; i < ARRAY_SIZE(dev->pma_counter_select); i++) p->counter[i] = (status != IB_PMA_SAMPLE_STATUS_DONE) ? 0 : cpu_to_be64( @@ -1185,7 +1184,7 @@ static int recv_pma_get_portcounters(struct ib_perf *pmp, pmp->status |= IB_SMP_INVALID_FIELD; if (cntrs.symbol_error_counter > 0xFFFFUL) - p->symbol_error_counter = __constant_cpu_to_be16(0xFFFF); + p->symbol_error_counter = cpu_to_be16(0xFFFF); else p->symbol_error_counter = cpu_to_be16((u16)cntrs.symbol_error_counter); @@ -1199,17 +1198,17 @@ static int recv_pma_get_portcounters(struct ib_perf *pmp, else p->link_downed_counter = (u8)cntrs.link_downed_counter; if (cntrs.port_rcv_errors > 0xFFFFUL) - p->port_rcv_errors = __constant_cpu_to_be16(0xFFFF); + p->port_rcv_errors = cpu_to_be16(0xFFFF); else p->port_rcv_errors = cpu_to_be16((u16) cntrs.port_rcv_errors); if (cntrs.port_rcv_remphys_errors > 0xFFFFUL) - p->port_rcv_remphys_errors = __constant_cpu_to_be16(0xFFFF); + p->port_rcv_remphys_errors = cpu_to_be16(0xFFFF); else p->port_rcv_remphys_errors = cpu_to_be16((u16)cntrs.port_rcv_remphys_errors); if (cntrs.port_xmit_discards > 0xFFFFUL) - p->port_xmit_discards = __constant_cpu_to_be16(0xFFFF); + p->port_xmit_discards = cpu_to_be16(0xFFFF); else p->port_xmit_discards = cpu_to_be16((u16)cntrs.port_xmit_discards); @@ -1220,24 +1219,24 @@ static int recv_pma_get_portcounters(struct ib_perf *pmp, p->lli_ebor_errors = (cntrs.local_link_integrity_errors << 4) | cntrs.excessive_buffer_overrun_errors; if (cntrs.vl15_dropped > 0xFFFFUL) - p->vl15_dropped = __constant_cpu_to_be16(0xFFFF); + p->vl15_dropped = cpu_to_be16(0xFFFF); else p->vl15_dropped = cpu_to_be16((u16)cntrs.vl15_dropped); if (cntrs.port_xmit_data > 0xFFFFFFFFUL) - p->port_xmit_data = __constant_cpu_to_be32(0xFFFFFFFF); + p->port_xmit_data = cpu_to_be32(0xFFFFFFFF); else p->port_xmit_data = cpu_to_be32((u32)cntrs.port_xmit_data); if (cntrs.port_rcv_data > 0xFFFFFFFFUL) - p->port_rcv_data = __constant_cpu_to_be32(0xFFFFFFFF); + p->port_rcv_data = cpu_to_be32(0xFFFFFFFF); else p->port_rcv_data = cpu_to_be32((u32)cntrs.port_rcv_data); if (cntrs.port_xmit_packets > 0xFFFFFFFFUL) - p->port_xmit_packets = __constant_cpu_to_be32(0xFFFFFFFF); + p->port_xmit_packets = cpu_to_be32(0xFFFFFFFF); else p->port_xmit_packets = cpu_to_be32((u32)cntrs.port_xmit_packets); if (cntrs.port_rcv_packets > 0xFFFFFFFFUL) - p->port_rcv_packets = __constant_cpu_to_be32(0xFFFFFFFF); + p->port_rcv_packets = cpu_to_be32(0xFFFFFFFF); else p->port_rcv_packets = cpu_to_be32((u32) cntrs.port_rcv_packets); diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c index 9170710b950..79b3dbc9717 100644 --- a/drivers/infiniband/hw/ipath/ipath_rc.c +++ b/drivers/infiniband/hw/ipath/ipath_rc.c @@ -1744,7 +1744,7 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, /* Signal completion event if the solicited bit is set. */ ipath_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, (ohdr->bth[0] & - __constant_cpu_to_be32(1 << 23)) != 0); + cpu_to_be32(1 << 23)) != 0); break; case OP(RDMA_WRITE_FIRST): diff --git a/drivers/infiniband/hw/ipath/ipath_sdma.c b/drivers/infiniband/hw/ipath/ipath_sdma.c index 8e255adf5d9..4b069859085 100644 --- a/drivers/infiniband/hw/ipath/ipath_sdma.c +++ b/drivers/infiniband/hw/ipath/ipath_sdma.c @@ -781,10 +781,10 @@ retry: descqp = &dd->ipath_sdma_descq[dd->ipath_sdma_descq_cnt].qw[0]; descqp -= 2; /* SDmaLastDesc */ - descqp[0] |= __constant_cpu_to_le64(1ULL << 11); + descqp[0] |= cpu_to_le64(1ULL << 11); if (tx->txreq.flags & IPATH_SDMA_TXREQ_F_INTREQ) { /* SDmaIntReq */ - descqp[0] |= __constant_cpu_to_le64(1ULL << 15); + descqp[0] |= cpu_to_le64(1ULL << 15); } /* Commit writes to memory and advance the tail on the chip */ diff --git a/drivers/infiniband/hw/ipath/ipath_uc.c b/drivers/infiniband/hw/ipath/ipath_uc.c index 82cc588b8bf..22e60998f1a 100644 --- a/drivers/infiniband/hw/ipath/ipath_uc.c +++ b/drivers/infiniband/hw/ipath/ipath_uc.c @@ -419,7 +419,7 @@ void ipath_uc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, /* Signal completion event if the solicited bit is set. */ ipath_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, (ohdr->bth[0] & - __constant_cpu_to_be32(1 << 23)) != 0); + cpu_to_be32(1 << 23)) != 0); break; case OP(RDMA_WRITE_FIRST): diff --git a/drivers/infiniband/hw/ipath/ipath_ud.c b/drivers/infiniband/hw/ipath/ipath_ud.c index 91c74cc797a..6076cb61bf6 100644 --- a/drivers/infiniband/hw/ipath/ipath_ud.c +++ b/drivers/infiniband/hw/ipath/ipath_ud.c @@ -370,7 +370,7 @@ int ipath_make_ud_req(struct ipath_qp *qp) */ ohdr->bth[1] = ah_attr->dlid >= IPATH_MULTICAST_LID_BASE && ah_attr->dlid != IPATH_PERMISSIVE_LID ? - __constant_cpu_to_be32(IPATH_MULTICAST_QPN) : + cpu_to_be32(IPATH_MULTICAST_QPN) : cpu_to_be32(wqe->wr.wr.ud.remote_qpn); ohdr->bth[2] = cpu_to_be32(qp->s_next_psn++ & IPATH_PSN_MASK); /* @@ -573,7 +573,7 @@ void ipath_ud_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, /* Signal completion event if the solicited bit is set. */ ipath_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, (ohdr->bth[0] & - __constant_cpu_to_be32(1 << 23)) != 0); + cpu_to_be32(1 << 23)) != 0); bail:; } diff --git a/drivers/infiniband/hw/ipath/ipath_user_sdma.c b/drivers/infiniband/hw/ipath/ipath_user_sdma.c index 82d9a0b5ca2..7bff4b9baa0 100644 --- a/drivers/infiniband/hw/ipath/ipath_user_sdma.c +++ b/drivers/infiniband/hw/ipath/ipath_user_sdma.c @@ -667,13 +667,13 @@ static inline __le64 ipath_sdma_make_desc0(struct ipath_devdata *dd, static inline __le64 ipath_sdma_make_first_desc0(__le64 descq) { - return descq | __constant_cpu_to_le64(1ULL << 12); + return descq | cpu_to_le64(1ULL << 12); } static inline __le64 ipath_sdma_make_last_desc0(__le64 descq) { /* last */ /* dma head */ - return descq | __constant_cpu_to_le64(1ULL << 11 | 1ULL << 13); + return descq | cpu_to_le64(1ULL << 11 | 1ULL << 13); } static inline __le64 ipath_sdma_make_desc1(u64 addr) @@ -763,7 +763,7 @@ static int ipath_user_sdma_push_pkts(struct ipath_devdata *dd, if (ofs >= IPATH_SMALLBUF_DWORDS) { for (i = 0; i < pkt->naddr; i++) { dd->ipath_sdma_descq[dtail].qw[0] |= - __constant_cpu_to_le64(1ULL << 14); + cpu_to_le64(1ULL << 14); if (++dtail == dd->ipath_sdma_descq_cnt) dtail = 0; } diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c index cdf0e6abd34..9289ab4b0ae 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs.c +++ b/drivers/infiniband/hw/ipath/ipath_verbs.c @@ -1585,7 +1585,7 @@ static int ipath_query_port(struct ib_device *ibdev, u64 ibcstat; memset(props, 0, sizeof(*props)); - props->lid = lid ? lid : __constant_be16_to_cpu(IB_LID_PERMISSIVE); + props->lid = lid ? lid : be16_to_cpu(IB_LID_PERMISSIVE); props->lmc = dd->ipath_lmc; props->sm_lid = dev->sm_lid; props->sm_sl = dev->sm_sl; diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h index 11e3f613df9..ae6cff4abff 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs.h +++ b/drivers/infiniband/hw/ipath/ipath_verbs.h @@ -86,11 +86,11 @@ #define IB_PMA_SAMPLE_STATUS_RUNNING 0x02 /* Mandatory IB performance counter select values. */ -#define IB_PMA_PORT_XMIT_DATA __constant_htons(0x0001) -#define IB_PMA_PORT_RCV_DATA __constant_htons(0x0002) -#define IB_PMA_PORT_XMIT_PKTS __constant_htons(0x0003) -#define IB_PMA_PORT_RCV_PKTS __constant_htons(0x0004) -#define IB_PMA_PORT_XMIT_WAIT __constant_htons(0x0005) +#define IB_PMA_PORT_XMIT_DATA cpu_to_be16(0x0001) +#define IB_PMA_PORT_RCV_DATA cpu_to_be16(0x0002) +#define IB_PMA_PORT_XMIT_PKTS cpu_to_be16(0x0003) +#define IB_PMA_PORT_RCV_PKTS cpu_to_be16(0x0004) +#define IB_PMA_PORT_XMIT_WAIT cpu_to_be16(0x0005) struct ib_reth { __be64 vaddr; diff --git a/drivers/infiniband/hw/mlx4/qp.c b/drivers/infiniband/hw/mlx4/qp.c index a91cb4c3fa5..f385a24d31d 100644 --- a/drivers/infiniband/hw/mlx4/qp.c +++ b/drivers/infiniband/hw/mlx4/qp.c @@ -71,17 +71,17 @@ enum { }; static const __be32 mlx4_ib_opcode[] = { - [IB_WR_SEND] = __constant_cpu_to_be32(MLX4_OPCODE_SEND), - [IB_WR_LSO] = __constant_cpu_to_be32(MLX4_OPCODE_LSO), - [IB_WR_SEND_WITH_IMM] = __constant_cpu_to_be32(MLX4_OPCODE_SEND_IMM), - [IB_WR_RDMA_WRITE] = __constant_cpu_to_be32(MLX4_OPCODE_RDMA_WRITE), - [IB_WR_RDMA_WRITE_WITH_IMM] = __constant_cpu_to_be32(MLX4_OPCODE_RDMA_WRITE_IMM), - [IB_WR_RDMA_READ] = __constant_cpu_to_be32(MLX4_OPCODE_RDMA_READ), - [IB_WR_ATOMIC_CMP_AND_SWP] = __constant_cpu_to_be32(MLX4_OPCODE_ATOMIC_CS), - [IB_WR_ATOMIC_FETCH_AND_ADD] = __constant_cpu_to_be32(MLX4_OPCODE_ATOMIC_FA), - [IB_WR_SEND_WITH_INV] = __constant_cpu_to_be32(MLX4_OPCODE_SEND_INVAL), - [IB_WR_LOCAL_INV] = __constant_cpu_to_be32(MLX4_OPCODE_LOCAL_INVAL), - [IB_WR_FAST_REG_MR] = __constant_cpu_to_be32(MLX4_OPCODE_FMR), + [IB_WR_SEND] = cpu_to_be32(MLX4_OPCODE_SEND), + [IB_WR_LSO] = cpu_to_be32(MLX4_OPCODE_LSO), + [IB_WR_SEND_WITH_IMM] = cpu_to_be32(MLX4_OPCODE_SEND_IMM), + [IB_WR_RDMA_WRITE] = cpu_to_be32(MLX4_OPCODE_RDMA_WRITE), + [IB_WR_RDMA_WRITE_WITH_IMM] = cpu_to_be32(MLX4_OPCODE_RDMA_WRITE_IMM), + [IB_WR_RDMA_READ] = cpu_to_be32(MLX4_OPCODE_RDMA_READ), + [IB_WR_ATOMIC_CMP_AND_SWP] = cpu_to_be32(MLX4_OPCODE_ATOMIC_CS), + [IB_WR_ATOMIC_FETCH_AND_ADD] = cpu_to_be32(MLX4_OPCODE_ATOMIC_FA), + [IB_WR_SEND_WITH_INV] = cpu_to_be32(MLX4_OPCODE_SEND_INVAL), + [IB_WR_LOCAL_INV] = cpu_to_be32(MLX4_OPCODE_LOCAL_INVAL), + [IB_WR_FAST_REG_MR] = cpu_to_be32(MLX4_OPCODE_FMR), }; static struct mlx4_ib_sqp *to_msqp(struct mlx4_ib_qp *mqp) -- cgit v1.2.3 From f0f6f346a1edaec23b990c25f53478669e56fa70 Mon Sep 17 00:00:00 2001 From: Moni Shoua Date: Wed, 28 Jan 2009 14:54:35 -0800 Subject: IB/mlx4: Fix dispatch of IB_EVENT_LID_CHANGE event When snooping a PortInfo MAD, its client_reregister bit is checked. If the bit is ON then a CLIENT_REREGISTER event is dispatched, otherwise a LID_CHANGE event is dispatched. This way of decision ignores the cases where the MAD changes the LID along with an instruction to reregister (so a necessary LID_CHANGE event won't be dispatched) or the MAD is neither of these (and an unnecessary LID_CHANGE event will be dispatched). This causes problems at least with IPoIB, which will do a "light" flush on reregister, rather than the "heavy" flush required due to a LID change. Fix this by dispatching a CLIENT_REREGISTER event if the client_reregister bit is set, but also compare the LID in the MAD to the current LID. If and only if they are not identical then a LID_CHANGE event is dispatched. Signed-off-by: Moni Shoua Signed-off-by: Jack Morgenstein Signed-off-by: Yossi Etigin Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/mad.c | 27 ++++++++++++++++++++------- 1 file changed, 20 insertions(+), 7 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/mad.c b/drivers/infiniband/hw/mlx4/mad.c index 606f1e2ef28..19e68ab6616 100644 --- a/drivers/infiniband/hw/mlx4/mad.c +++ b/drivers/infiniband/hw/mlx4/mad.c @@ -147,7 +147,8 @@ static void update_sm_ah(struct mlx4_ib_dev *dev, u8 port_num, u16 lid, u8 sl) * Snoop SM MADs for port info and P_Key table sets, so we can * synthesize LID change and P_Key change events. */ -static void smp_snoop(struct ib_device *ibdev, u8 port_num, struct ib_mad *mad) +static void smp_snoop(struct ib_device *ibdev, u8 port_num, struct ib_mad *mad, + u16 prev_lid) { struct ib_event event; @@ -157,6 +158,7 @@ static void smp_snoop(struct ib_device *ibdev, u8 port_num, struct ib_mad *mad) if (mad->mad_hdr.attr_id == IB_SMP_ATTR_PORT_INFO) { struct ib_port_info *pinfo = (struct ib_port_info *) ((struct ib_smp *) mad)->data; + u16 lid = be16_to_cpu(pinfo->lid); update_sm_ah(to_mdev(ibdev), port_num, be16_to_cpu(pinfo->sm_lid), @@ -165,12 +167,15 @@ static void smp_snoop(struct ib_device *ibdev, u8 port_num, struct ib_mad *mad) event.device = ibdev; event.element.port_num = port_num; - if (pinfo->clientrereg_resv_subnetto & 0x80) + if (pinfo->clientrereg_resv_subnetto & 0x80) { event.event = IB_EVENT_CLIENT_REREGISTER; - else - event.event = IB_EVENT_LID_CHANGE; + ib_dispatch_event(&event); + } - ib_dispatch_event(&event); + if (prev_lid != lid) { + event.event = IB_EVENT_LID_CHANGE; + ib_dispatch_event(&event); + } } if (mad->mad_hdr.attr_id == IB_SMP_ATTR_PKEY_TABLE) { @@ -228,8 +233,9 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num, struct ib_wc *in_wc, struct ib_grh *in_grh, struct ib_mad *in_mad, struct ib_mad *out_mad) { - u16 slid; + u16 slid, prev_lid = 0; int err; + struct ib_port_attr pattr; slid = in_wc ? in_wc->slid : be16_to_cpu(IB_LID_PERMISSIVE); @@ -263,6 +269,13 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num, } else return IB_MAD_RESULT_SUCCESS; + if ((in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED || + in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) && + in_mad->mad_hdr.method == IB_MGMT_METHOD_SET && + in_mad->mad_hdr.attr_id == IB_SMP_ATTR_PORT_INFO && + !ib_query_port(ibdev, port_num, &pattr)) + prev_lid = pattr.lid; + err = mlx4_MAD_IFC(to_mdev(ibdev), mad_flags & IB_MAD_IGNORE_MKEY, mad_flags & IB_MAD_IGNORE_BKEY, @@ -271,7 +284,7 @@ int mlx4_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num, return IB_MAD_RESULT_FAILURE; if (!out_mad->mad_hdr.status) { - smp_snoop(ibdev, port_num, in_mad); + smp_snoop(ibdev, port_num, in_mad, prev_lid); node_desc_override(ibdev, out_mad); } -- cgit v1.2.3 From 270b8b85134c299799dddec624ceeb5671330131 Mon Sep 17 00:00:00 2001 From: Moni Shoua Date: Wed, 28 Jan 2009 15:15:56 -0800 Subject: IB/mthca: Fix dispatch of IB_EVENT_LID_CHANGE event When snooping a PortInfo MAD, its client_reregister bit is checked. If the bit is ON then a CLIENT_REREGISTER event is dispatched, otherwise a LID_CHANGE event is dispatched. This way of decision ignores the cases where the MAD changes the LID along with an instruction to reregister (so a necessary LID_CHANGE event won't be dispatched) or the MAD is neither of these (and an unnecessary LID_CHANGE event will be dispatched). This causes problems at least with IPoIB, which will do a "light" flush on reregister, rather than the "heavy" flush required due to a LID change. Fix this by dispatching a CLIENT_REREGISTER event if the client_reregister bit is set, but also compare the LID in the MAD to the current LID. If and only if they are not identical then a LID_CHANGE event is dispatched. Signed-off-by: Moni Shoua Signed-off-by: Jack Morgenstein Signed-off-by: Yossi Etigin Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_mad.c | 25 +++++++++++++++++++------ 1 file changed, 19 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c index 640449582ab..5648659ff0b 100644 --- a/drivers/infiniband/hw/mthca/mthca_mad.c +++ b/drivers/infiniband/hw/mthca/mthca_mad.c @@ -104,7 +104,8 @@ static void update_sm_ah(struct mthca_dev *dev, */ static void smp_snoop(struct ib_device *ibdev, u8 port_num, - struct ib_mad *mad) + struct ib_mad *mad, + u16 prev_lid) { struct ib_event event; @@ -114,6 +115,7 @@ static void smp_snoop(struct ib_device *ibdev, if (mad->mad_hdr.attr_id == IB_SMP_ATTR_PORT_INFO) { struct ib_port_info *pinfo = (struct ib_port_info *) ((struct ib_smp *) mad)->data; + u16 lid = be16_to_cpu(pinfo->lid); mthca_update_rate(to_mdev(ibdev), port_num); update_sm_ah(to_mdev(ibdev), port_num, @@ -123,12 +125,15 @@ static void smp_snoop(struct ib_device *ibdev, event.device = ibdev; event.element.port_num = port_num; - if (pinfo->clientrereg_resv_subnetto & 0x80) + if (pinfo->clientrereg_resv_subnetto & 0x80) { event.event = IB_EVENT_CLIENT_REREGISTER; - else - event.event = IB_EVENT_LID_CHANGE; + ib_dispatch_event(&event); + } - ib_dispatch_event(&event); + if (prev_lid != lid) { + event.event = IB_EVENT_LID_CHANGE; + ib_dispatch_event(&event); + } } if (mad->mad_hdr.attr_id == IB_SMP_ATTR_PKEY_TABLE) { @@ -196,6 +201,8 @@ int mthca_process_mad(struct ib_device *ibdev, int err; u8 status; u16 slid = in_wc ? in_wc->slid : be16_to_cpu(IB_LID_PERMISSIVE); + u16 prev_lid = 0; + struct ib_port_attr pattr; /* Forward locally generated traps to the SM */ if (in_mad->mad_hdr.method == IB_MGMT_METHOD_TRAP && @@ -233,6 +240,12 @@ int mthca_process_mad(struct ib_device *ibdev, return IB_MAD_RESULT_SUCCESS; } else return IB_MAD_RESULT_SUCCESS; + if ((in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_LID_ROUTED || + in_mad->mad_hdr.mgmt_class == IB_MGMT_CLASS_SUBN_DIRECTED_ROUTE) && + in_mad->mad_hdr.method == IB_MGMT_METHOD_SET && + in_mad->mad_hdr.attr_id == IB_SMP_ATTR_PORT_INFO && + !ib_query_port(ibdev, port_num, &pattr)) + prev_lid = pattr.lid; err = mthca_MAD_IFC(to_mdev(ibdev), mad_flags & IB_MAD_IGNORE_MKEY, @@ -252,7 +265,7 @@ int mthca_process_mad(struct ib_device *ibdev, } if (!out_mad->mad_hdr.status) { - smp_snoop(ibdev, port_num, in_mad); + smp_snoop(ibdev, port_num, in_mad, prev_lid); node_desc_override(ibdev, out_mad); } -- cgit v1.2.3 From 71d98b4628ee869d62814f6d8607d76cab4b9ec5 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Tue, 17 Feb 2009 14:51:47 -0800 Subject: IPoIB: In unicast_arp_send(), only free newly-created paths If path_rec_start() returns error, call path_free() only if the path was newly-created. If we free an existing path whose valid flag was zero, (but do not detach it from the list) we cause corruption of the path list (of which it is a member), and get a kernel crash. The simplest solution is to not free an existing path -- just leave it in the list as-is (i.e., with its valid flag cleared). Thanks to Yossi Etigin of Voltaire for identifying the problem flow which caused the kernel crash. Signed-off-by: Jack Morgenstein Signed-off-by: Moni Shua Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/ipoib_main.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 0bd2a4ff084..353c13b91e8 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -660,8 +660,12 @@ static void unicast_arp_send(struct sk_buff *skb, struct net_device *dev, path = __path_find(dev, phdr->hwaddr + 4); if (!path || !path->valid) { - if (!path) + int new_path = 0; + + if (!path) { path = path_rec_create(dev, phdr->hwaddr + 4); + new_path = 1; + } if (path) { /* put pseudoheader back on for next time */ skb_push(skb, sizeof *phdr); @@ -669,7 +673,8 @@ static void unicast_arp_send(struct sk_buff *skb, struct net_device *dev, if (!path->query && path_rec_start(dev, path)) { spin_unlock_irqrestore(&priv->lock, flags); - path_free(dev, path); + if (new_path) + path_free(dev, path); return; } else __path_add(dev, path); -- cgit v1.2.3 From 71c4512201575c7cc008b364e2e2c75cc7085c26 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Sun, 22 Feb 2009 20:04:34 -0800 Subject: IB/ipath: Fix memory leak in init_shadow_tids() error path If the second vmalloc() fails, the wrong pointer is pased to vfree(), so the first vmalloc() ends up getting leaked. This was spotted by the Coverity checker (CID 2709). Signed-off-by: Roland Dreier --- drivers/infiniband/hw/ipath/ipath_init_chip.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ipath/ipath_init_chip.c b/drivers/infiniband/hw/ipath/ipath_init_chip.c index 64aeefbd2a5..077879c0bdb 100644 --- a/drivers/infiniband/hw/ipath/ipath_init_chip.c +++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c @@ -455,7 +455,7 @@ static void init_shadow_tids(struct ipath_devdata *dd) if (!addrs) { ipath_dev_err(dd, "failed to allocate shadow dma handle " "array, no expected sends!\n"); - vfree(dd->ipath_pageshadow); + vfree(pages); dd->ipath_pageshadow = NULL; return; } -- cgit v1.2.3 From e538052746d570c874650a24eed89fca6e4c93dc Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Sun, 22 Feb 2009 20:14:37 -0800 Subject: IB/ipath: Really run work in ipath_release_user_pages_on_close() ipath_release_user_pages_on_close() just allocated a structure to schedule work with but just returned (leaking the structure) rather than actually doing schedule_work(). Fix the logic to what was intended. This was spotted by the Coverity checker (CID 2700). Signed-off-by: Roland Dreier --- drivers/infiniband/hw/ipath/ipath_user_pages.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/ipath/ipath_user_pages.c b/drivers/infiniband/hw/ipath/ipath_user_pages.c index 0190edc8044..855911e7396 100644 --- a/drivers/infiniband/hw/ipath/ipath_user_pages.c +++ b/drivers/infiniband/hw/ipath/ipath_user_pages.c @@ -209,20 +209,20 @@ void ipath_release_user_pages_on_close(struct page **p, size_t num_pages) mm = get_task_mm(current); if (!mm) - goto bail; + return; work = kmalloc(sizeof(*work), GFP_KERNEL); if (!work) goto bail_mm; - goto bail; - INIT_WORK(&work->work, user_pages_account); work->mm = mm; work->num_pages = num_pages; + schedule_work(&work->work); + return; + bail_mm: mmput(mm); -bail: return; } -- cgit v1.2.3 From 9206dff15705267c19f8fed391c4fb95975540a3 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Wed, 25 Feb 2009 13:27:46 -0800 Subject: IB: Remove sysfs files before unregistering device Move the ib_device_unregister_sysfs() call from ib_dealloc_device() to ib_unregister_device(). The old code allows device unregister to proceed even if some sysfs files are open, which leaves a window where userspace can open a file before a device is removed but then end up reading the file after the device is removed, which leads to various kernel crashes either because the device data structure is freed or because the low-level driver code is gone after module removal. By not returning from ib_unregister_device() until after all sysfs entries are removed, we make sure that data structures and/or module code is not freed until after all sysfs access is done. Reported-by: Jack Morgenstein Signed-off-by: Roland Dreier --- drivers/infiniband/core/device.c | 4 +++- drivers/infiniband/core/sysfs.c | 3 +++ 2 files changed, 6 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/device.c b/drivers/infiniband/core/device.c index 7913b804311..d1fba415333 100644 --- a/drivers/infiniband/core/device.c +++ b/drivers/infiniband/core/device.c @@ -193,7 +193,7 @@ void ib_dealloc_device(struct ib_device *device) BUG_ON(device->reg_state != IB_DEV_UNREGISTERED); - ib_device_unregister_sysfs(device); + kobject_put(&device->dev.kobj); } EXPORT_SYMBOL(ib_dealloc_device); @@ -348,6 +348,8 @@ void ib_unregister_device(struct ib_device *device) mutex_unlock(&device_mutex); + ib_device_unregister_sysfs(device); + spin_lock_irqsave(&device->client_data_lock, flags); list_for_each_entry_safe(context, tmp, &device->client_data_list, list) kfree(context); diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c index b43f7d3682d..5270aeb56e9 100644 --- a/drivers/infiniband/core/sysfs.c +++ b/drivers/infiniband/core/sysfs.c @@ -848,6 +848,9 @@ void ib_device_unregister_sysfs(struct ib_device *device) struct kobject *p, *t; struct ib_port *port; + /* Hold kobject until ib_dealloc_device() */ + kobject_get(&device->dev.kobj); + list_for_each_entry_safe(p, t, &device->port_list, entry) { list_del(&p->entry); port = container_of(p, struct ib_port, kobj); -- cgit v1.2.3 From 1aedb7721f05461f777fdee25b50d8a168c425ed Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Fri, 27 Feb 2009 10:30:46 -0800 Subject: IB/iser: Remove hard setting of path MTU Remove hard setting of the IB MTU used by iSER's RC queue-pair to 1K, as this was done due to inter-op issues with an old iser target which is not used any more. Signed-off-by: Or Gerlitz Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/iser/iser_verbs.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index 319b188145b..ea9e1556e0d 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -401,13 +401,6 @@ static void iser_route_handler(struct rdma_cm_id *cma_id) if (ret) goto failure; - iser_dbg("path.mtu is %d setting it to %d\n", - cma_id->route.path_rec->mtu, IB_MTU_1024); - - /* we must set the MTU to 1024 as this is what the target is assuming */ - if (cma_id->route.path_rec->mtu > IB_MTU_1024) - cma_id->route.path_rec->mtu = IB_MTU_1024; - memset(&conn_param, 0, sizeof conn_param); conn_param.responder_resources = 4; conn_param.initiator_depth = 1; -- cgit v1.2.3 From 1d9bc6d648ece77ffb41c5a577eab81fac5ad4de Mon Sep 17 00:00:00 2001 From: Ralph Campbell Date: Fri, 27 Feb 2009 10:34:30 -0800 Subject: IB/mad: Fix null pointer dereference in local_completions() handle_outgoing_dr_smp() can queue a struct ib_mad_local_private *local on the mad_agent_priv->local_work work queue with local->mad_priv == NULL if device->process_mad() returns IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_REPLY and (!ib_response_mad(&mad_priv->mad.mad) || !mad_agent_priv->agent.recv_handler). In this case, local_completions() will be called with local->mad_priv == NULL. The code does check for this case and skips calling recv_mad_agent->agent.recv_handler() but recv == 0 so kmem_cache_free() is called with a NULL pointer. Also, since recv isn't reinitialized each time through the loop, it can cause a memory leak if recv should have been zero. Signed-off-by: Ralph Campbell --- drivers/infiniband/core/mad.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index 5c54fc2350b..735ad4ea10f 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -2356,7 +2356,7 @@ static void local_completions(struct work_struct *work) struct ib_mad_local_private *local; struct ib_mad_agent_private *recv_mad_agent; unsigned long flags; - int recv = 0; + int free_mad; struct ib_wc wc; struct ib_mad_send_wc mad_send_wc; @@ -2370,14 +2370,15 @@ static void local_completions(struct work_struct *work) completion_list); list_del(&local->completion_list); spin_unlock_irqrestore(&mad_agent_priv->lock, flags); + free_mad = 0; if (local->mad_priv) { recv_mad_agent = local->recv_mad_agent; if (!recv_mad_agent) { printk(KERN_ERR PFX "No receive MAD agent for local completion\n"); + free_mad = 1; goto local_send_completion; } - recv = 1; /* * Defined behavior is to complete response * before request @@ -2422,7 +2423,7 @@ local_send_completion: spin_lock_irqsave(&mad_agent_priv->lock, flags); atomic_dec(&mad_agent_priv->refcount); - if (!recv) + if (free_mad) kmem_cache_free(ib_mad_cache, local->mad_priv); kfree(local); } -- cgit v1.2.3 From d9620a4c82c61a91c9313f80ba951c902573c028 Mon Sep 17 00:00:00 2001 From: Ralph Campbell Date: Fri, 27 Feb 2009 14:44:32 -0800 Subject: IB/mad: initialize mad_agent_priv before putting on lists There is a potential race in ib_register_mad_agent() where the struct ib_mad_agent_private is not fully initialized before it is added to the list of agents per IB port. This means the ib_mad_agent_private could be seen before the refcount, spin locks, and linked lists are initialized. The fix is to initialize the structure earlier. Signed-off-by: Ralph Campbell Signed-off-by: Roland Dreier --- drivers/infiniband/core/mad.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index 735ad4ea10f..dbcd285405e 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -301,6 +301,16 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, mad_agent_priv->agent.context = context; mad_agent_priv->agent.qp = port_priv->qp_info[qpn].qp; mad_agent_priv->agent.port_num = port_num; + spin_lock_init(&mad_agent_priv->lock); + INIT_LIST_HEAD(&mad_agent_priv->send_list); + INIT_LIST_HEAD(&mad_agent_priv->wait_list); + INIT_LIST_HEAD(&mad_agent_priv->done_list); + INIT_LIST_HEAD(&mad_agent_priv->rmpp_list); + INIT_DELAYED_WORK(&mad_agent_priv->timed_work, timeout_sends); + INIT_LIST_HEAD(&mad_agent_priv->local_list); + INIT_WORK(&mad_agent_priv->local_work, local_completions); + atomic_set(&mad_agent_priv->refcount, 1); + init_completion(&mad_agent_priv->comp); spin_lock_irqsave(&port_priv->reg_lock, flags); mad_agent_priv->agent.hi_tid = ++ib_mad_client_id; @@ -350,17 +360,6 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, list_add_tail(&mad_agent_priv->agent_list, &port_priv->agent_list); spin_unlock_irqrestore(&port_priv->reg_lock, flags); - spin_lock_init(&mad_agent_priv->lock); - INIT_LIST_HEAD(&mad_agent_priv->send_list); - INIT_LIST_HEAD(&mad_agent_priv->wait_list); - INIT_LIST_HEAD(&mad_agent_priv->done_list); - INIT_LIST_HEAD(&mad_agent_priv->rmpp_list); - INIT_DELAYED_WORK(&mad_agent_priv->timed_work, timeout_sends); - INIT_LIST_HEAD(&mad_agent_priv->local_list); - INIT_WORK(&mad_agent_priv->local_work, local_completions); - atomic_set(&mad_agent_priv->refcount, 1); - init_completion(&mad_agent_priv->comp); - return &mad_agent_priv->agent; error4: -- cgit v1.2.3 From 4780c1953f9bef07365b13af01ae4e8238ecd3de Mon Sep 17 00:00:00 2001 From: Ralph Campbell Date: Tue, 3 Mar 2009 14:22:17 -0800 Subject: IB/mad: Fix ib_post_send_mad() returning 0 with no generate send comp If ib_post_send_mad() returns 0, the API guarantees that there will be a callback to send_buf->mad_agent->send_handler() so that the sender can call ib_free_send_mad(). Otherwise, the ib_mad_send_buf will be leaked and the mad_agent reference count will never go to zero and the IB device module cannot be unloaded. The above can happen without this patch if process_mad() returns (IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED). If process_mad() returns IB_MAD_RESULT_SUCCESS and there is no agent registered to receive the mad being sent, handle_outgoing_dr_smp() returns zero which causes a MAD packet which is at the end of the directed route to be incorrectly sent on the wire but doesn't cause a hang since the HCA generates a send completion. Signed-off-by: Ralph Campbell Signed-off-by: Roland Dreier --- drivers/infiniband/core/mad.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index dbcd285405e..de922a04ca2 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -742,9 +742,7 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, break; case IB_MAD_RESULT_SUCCESS | IB_MAD_RESULT_CONSUMED: kmem_cache_free(ib_mad_cache, mad_priv); - kfree(local); - ret = 1; - goto out; + break; case IB_MAD_RESULT_SUCCESS: /* Treat like an incoming receive MAD */ port_priv = ib_get_mad_port(mad_agent_priv->agent.device, @@ -755,10 +753,12 @@ static int handle_outgoing_dr_smp(struct ib_mad_agent_private *mad_agent_priv, &mad_priv->mad.mad); } if (!port_priv || !recv_mad_agent) { + /* + * No receiving agent so drop packet and + * generate send completion. + */ kmem_cache_free(ib_mad_cache, mad_priv); - kfree(local); - ret = 0; - goto out; + break; } local->mad_priv = mad_priv; local->recv_mad_agent = recv_mad_agent; -- cgit v1.2.3 From 6b708b3dde0ab3a10a0eea7774c1d6482f32f587 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Tue, 3 Mar 2009 14:30:01 -0800 Subject: IB/sa_query: Fix AH leak due to update_sm_ah() race Our testing uncovered a race condition in ib_sa_event(): spin_lock_irqsave(&port->ah_lock, flags); if (port->sm_ah) kref_put(&port->sm_ah->ref, free_sm_ah); port->sm_ah = NULL; spin_unlock_irqrestore(&port->ah_lock, flags); schedule_work(&sa_dev->port[event->element.port_num - sa_dev->start_port].update_task); If two events occur back-to-back (e.g., client-reregister and LID change), both may pass the spinlock-protected code above before the scheduled work updates the port->sm_ah handle. Then if the scheduled work ends up running twice, the second operation will then find a non-NULL port->sm_ah, and will simply overwrite it in update_sm_ah -- resulting in an AH leak. Signed-off-by: Jack Morgenstein Signed-off-by: Roland Dreier --- drivers/infiniband/core/sa_query.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index 7863a50d56f..1865049e80f 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -395,6 +395,8 @@ static void update_sm_ah(struct work_struct *work) } spin_lock_irq(&port->ah_lock); + if (port->sm_ah) + kref_put(&port->sm_ah->ref, free_sm_ah); port->sm_ah = new_ah; spin_unlock_irq(&port->ah_lock); -- cgit v1.2.3 From 6432f366842c78d17b2df974c68f39cb5794f2c2 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Wed, 4 Mar 2009 15:22:39 -0800 Subject: IB: Remove useless ibdev_is_alive() tests from sysfs code Some attribute show functions test ibdev_is_alive() to make sure that it's OK to access device state. However, the sysfs attributes will not be registered until the device is fully initialized, and they'll be unregistered before anything is torn down, so ibdev_is_alive() doesn't do anything useful. Remove it. Signed-off-by: Roland Dreier --- drivers/infiniband/core/sysfs.c | 16 ---------------- 1 file changed, 16 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c index 5270aeb56e9..5c04cfb54cb 100644 --- a/drivers/infiniband/core/sysfs.c +++ b/drivers/infiniband/core/sysfs.c @@ -66,11 +66,6 @@ struct port_table_attribute { int index; }; -static inline int ibdev_is_alive(const struct ib_device *dev) -{ - return dev->reg_state == IB_DEV_REGISTERED; -} - static ssize_t port_attr_show(struct kobject *kobj, struct attribute *attr, char *buf) { @@ -80,8 +75,6 @@ static ssize_t port_attr_show(struct kobject *kobj, if (!port_attr->show) return -EIO; - if (!ibdev_is_alive(p->ibdev)) - return -ENODEV; return port_attr->show(p, port_attr, buf); } @@ -562,9 +555,6 @@ static ssize_t show_node_type(struct device *device, { struct ib_device *dev = container_of(device, struct ib_device, dev); - if (!ibdev_is_alive(dev)) - return -ENODEV; - switch (dev->node_type) { case RDMA_NODE_IB_CA: return sprintf(buf, "%d: CA\n", dev->node_type); case RDMA_NODE_RNIC: return sprintf(buf, "%d: RNIC\n", dev->node_type); @@ -581,9 +571,6 @@ static ssize_t show_sys_image_guid(struct device *device, struct ib_device_attr attr; ssize_t ret; - if (!ibdev_is_alive(dev)) - return -ENODEV; - ret = ib_query_device(dev, &attr); if (ret) return ret; @@ -600,9 +587,6 @@ static ssize_t show_node_guid(struct device *device, { struct ib_device *dev = container_of(device, struct ib_device, dev); - if (!ibdev_is_alive(dev)) - return -ENODEV; - return sprintf(buf, "%04x:%04x:%04x:%04x\n", be16_to_cpu(((__be16 *) &dev->node_guid)[0]), be16_to_cpu(((__be16 *) &dev->node_guid)[1]), -- cgit v1.2.3 From dae5d13a7e6efcf6e0e00c0febb530b894fa13a3 Mon Sep 17 00:00:00 2001 From: Don Wood Date: Fri, 6 Mar 2009 15:12:09 -0800 Subject: RDMA/nes: Account for freed PBL after HW operation Fix occurrences where the software PBL counts were changed before the hardware was updated. This bug allowed another thread to overallocate the hardware resources. Add proper PBL accounting in case nes_reg_mr() fails. Signed-off-by: Don Wood Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_verbs.c | 70 +++++++++++++++++++++-------------- 1 file changed, 42 insertions(+), 28 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index 4fdb72454f9..e66681349c9 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -551,6 +551,7 @@ static int nes_dealloc_fmr(struct ib_fmr *ibfmr) struct nes_device *nesdev = nesvnic->nesdev; struct nes_adapter *nesadapter = nesdev->nesadapter; int i = 0; + int rc; /* free the resources */ if (nesfmr->leaf_pbl_cnt == 0) { @@ -572,7 +573,9 @@ static int nes_dealloc_fmr(struct ib_fmr *ibfmr) nesmr->ibmw.rkey = ibfmr->rkey; nesmr->ibmw.uobject = NULL; - if (nesfmr->nesmr.pbls_used != 0) { + rc = nes_dealloc_mw(&nesmr->ibmw); + + if ((rc == 0) && (nesfmr->nesmr.pbls_used != 0)) { spin_lock_irqsave(&nesadapter->pbl_lock, flags); if (nesfmr->nesmr.pbl_4k) { nesadapter->free_4kpbl += nesfmr->nesmr.pbls_used; @@ -584,7 +587,7 @@ static int nes_dealloc_fmr(struct ib_fmr *ibfmr) spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); } - return nes_dealloc_mw(&nesmr->ibmw); + return rc; } @@ -1993,7 +1996,16 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, stag, ret, cqp_request->major_code, cqp_request->minor_code); major_code = cqp_request->major_code; nes_put_cqp_request(nesdev, cqp_request); - + if ((!ret || major_code) && pbl_count != 0) { + spin_lock_irqsave(&nesadapter->pbl_lock, flags); + if (pbl_count > 1) + nesadapter->free_4kpbl += pbl_count+1; + else if (residual_page_count > 32) + nesadapter->free_4kpbl += pbl_count; + else + nesadapter->free_256pbl += pbl_count; + spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + } if (!ret) return -ETIME; else if (major_code) @@ -2607,24 +2619,6 @@ static int nes_dereg_mr(struct ib_mr *ib_mr) cqp_request->waiting = 1; cqp_wqe = &cqp_request->cqp_wqe; - spin_lock_irqsave(&nesadapter->pbl_lock, flags); - if (nesmr->pbls_used != 0) { - if (nesmr->pbl_4k) { - nesadapter->free_4kpbl += nesmr->pbls_used; - if (nesadapter->free_4kpbl > nesadapter->max_4kpbl) { - printk(KERN_ERR PFX "free 4KB PBLs(%u) has exceeded the max(%u)\n", - nesadapter->free_4kpbl, nesadapter->max_4kpbl); - } - } else { - nesadapter->free_256pbl += nesmr->pbls_used; - if (nesadapter->free_256pbl > nesadapter->max_256pbl) { - printk(KERN_ERR PFX "free 256B PBLs(%u) has exceeded the max(%u)\n", - nesadapter->free_256pbl, nesadapter->max_256pbl); - } - } - } - - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); nes_fill_init_cqp_wqe(cqp_wqe, nesdev); set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_WQE_OPCODE_IDX, NES_CQP_DEALLOCATE_STAG | NES_CQP_STAG_VA_TO | @@ -2642,11 +2636,6 @@ static int nes_dereg_mr(struct ib_mr *ib_mr) " CQP Major:Minor codes = 0x%04X:0x%04X\n", ib_mr->rkey, ret, cqp_request->major_code, cqp_request->minor_code); - nes_free_resource(nesadapter, nesadapter->allocated_mrs, - (ib_mr->rkey & 0x0fffff00) >> 8); - - kfree(nesmr); - major_code = cqp_request->major_code; minor_code = cqp_request->minor_code; @@ -2662,8 +2651,33 @@ static int nes_dereg_mr(struct ib_mr *ib_mr) " to destroy STag, ib_mr=%p, rkey = 0x%08X\n", major_code, minor_code, ib_mr, ib_mr->rkey); return -EIO; - } else - return 0; + } + + if (nesmr->pbls_used != 0) { + spin_lock_irqsave(&nesadapter->pbl_lock, flags); + if (nesmr->pbl_4k) { + nesadapter->free_4kpbl += nesmr->pbls_used; + if (nesadapter->free_4kpbl > nesadapter->max_4kpbl) + printk(KERN_ERR PFX "free 4KB PBLs(%u) has " + "exceeded the max(%u)\n", + nesadapter->free_4kpbl, + nesadapter->max_4kpbl); + } else { + nesadapter->free_256pbl += nesmr->pbls_used; + if (nesadapter->free_256pbl > nesadapter->max_256pbl) + printk(KERN_ERR PFX "free 256B PBLs(%u) has " + "exceeded the max(%u)\n", + nesadapter->free_256pbl, + nesadapter->max_256pbl); + } + spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + } + nes_free_resource(nesadapter, nesadapter->allocated_mrs, + (ib_mr->rkey & 0x0fffff00) >> 8); + + kfree(nesmr); + + return 0; } -- cgit v1.2.3 From cd6853d3eb453aee2574521b7ce2cd5a45492a59 Mon Sep 17 00:00:00 2001 From: Chien Tung Date: Fri, 6 Mar 2009 15:12:10 -0800 Subject: RDMA/nes: Update copyright to new legal entity and year Update copyright to the new legal entity, Intel-NE, Inc., an Intel company. Update copyright for the new year. Signed-off-by: Chien Tung Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes.c | 2 +- drivers/infiniband/hw/nes/nes.h | 2 +- drivers/infiniband/hw/nes/nes_cm.c | 2 +- drivers/infiniband/hw/nes/nes_cm.h | 2 +- drivers/infiniband/hw/nes/nes_context.h | 2 +- drivers/infiniband/hw/nes/nes_hw.c | 2 +- drivers/infiniband/hw/nes/nes_hw.h | 2 +- drivers/infiniband/hw/nes/nes_nic.c | 2 +- drivers/infiniband/hw/nes/nes_user.h | 2 +- drivers/infiniband/hw/nes/nes_utils.c | 2 +- drivers/infiniband/hw/nes/nes_verbs.c | 2 +- drivers/infiniband/hw/nes/nes_verbs.h | 2 +- 12 files changed, 12 insertions(+), 12 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes.c b/drivers/infiniband/hw/nes/nes.c index b9611ade9ea..ca599767ffb 100644 --- a/drivers/infiniband/hw/nes/nes.c +++ b/drivers/infiniband/hw/nes/nes.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved. * * This software is available to you under a choice of one of two diff --git a/drivers/infiniband/hw/nes/nes.h b/drivers/infiniband/hw/nes/nes.h index 13a5bb1a7bc..04b12ad2339 100644 --- a/drivers/infiniband/hw/nes/nes.h +++ b/drivers/infiniband/hw/nes/nes.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved. * * This software is available to you under a choice of one of two diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c index a01b4488208..6f42ab6bd02 100644 --- a/drivers/infiniband/hw/nes/nes_cm.c +++ b/drivers/infiniband/hw/nes/nes_cm.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_cm.h b/drivers/infiniband/hw/nes/nes_cm.h index fafa35042eb..4ab2bebf12a 100644 --- a/drivers/infiniband/hw/nes/nes_cm.h +++ b/drivers/infiniband/hw/nes/nes_cm.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_context.h b/drivers/infiniband/hw/nes/nes_context.h index da9daba8e66..0fb8d81d9a6 100644 --- a/drivers/infiniband/hw/nes/nes_context.h +++ b/drivers/infiniband/hw/nes/nes_context.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index 5d139db1b77..cb4a5f32f2f 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h index bc0b4de0445..6f8712d7a2c 100644 --- a/drivers/infiniband/hw/nes/nes_hw.h +++ b/drivers/infiniband/hw/nes/nes_hw.h @@ -1,5 +1,5 @@ /* -* Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. +* Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_nic.c b/drivers/infiniband/hw/nes/nes_nic.c index 57a47cf7e51..8e1d073913e 100644 --- a/drivers/infiniband/hw/nes/nes_nic.c +++ b/drivers/infiniband/hw/nes/nes_nic.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_user.h b/drivers/infiniband/hw/nes/nes_user.h index e64306bce80..cc90c14b49e 100644 --- a/drivers/infiniband/hw/nes/nes_user.h +++ b/drivers/infiniband/hw/nes/nes_user.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * Copyright (c) 2005 Topspin Communications. All rights reserved. * Copyright (c) 2005 Cisco Systems. All rights reserved. * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved. diff --git a/drivers/infiniband/hw/nes/nes_utils.c b/drivers/infiniband/hw/nes/nes_utils.c index 6f3bc1b6bf2..a282031d15c 100644 --- a/drivers/infiniband/hw/nes/nes_utils.c +++ b/drivers/infiniband/hw/nes/nes_utils.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index e66681349c9..b42b17ac771 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * * This software is available to you under a choice of one of two * licenses. You may choose to be licensed under the terms of the GNU diff --git a/drivers/infiniband/hw/nes/nes_verbs.h b/drivers/infiniband/hw/nes/nes_verbs.h index 6c6b4da5184..da3c368f1ef 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.h +++ b/drivers/infiniband/hw/nes/nes_verbs.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2006 - 2008 NetEffect, Inc. All rights reserved. + * Copyright (c) 2006 - 2009 Intel-NE, Inc. All rights reserved. * Copyright (c) 2005 Open Grid Computing, Inc. All rights reserved. * * This software is available to you under a choice of one of two -- cgit v1.2.3 From b9c367e7e694502085f21d9d2686645798080aa3 Mon Sep 17 00:00:00 2001 From: Chien Tung Date: Fri, 6 Mar 2009 15:12:10 -0800 Subject: RDMA/nes: Report correct vendor_id and vendor_part_id ibv_devinfo displays 0 for vendor_id and vendor_part_id. Fill in OUI and device_id for those two fields. Signed-off-by: Chien Tung Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_hw.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index cb4a5f32f2f..3a72a06f1df 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -254,6 +254,7 @@ struct nes_adapter *nes_init_adapter(struct nes_device *nesdev, u8 hw_rev) { u32 adapter_size; u32 arp_table_size; u16 vendor_id; + u16 device_id; u8 OneG_Mode; u8 func_index; @@ -356,6 +357,13 @@ struct nes_adapter *nes_init_adapter(struct nes_device *nesdev, u8 hw_rev) { return NULL; } + nesadapter->vendor_id = (((u32) nesadapter->mac_addr_high) << 8) | + (nesadapter->mac_addr_low >> 24); + + pci_bus_read_config_word(nesdev->pcidev->bus, nesdev->pcidev->devfn, + PCI_DEVICE_ID, &device_id); + nesadapter->vendor_part_id = device_id; + if (nes_init_serdes(nesdev, hw_rev, port_count, nesadapter, OneG_Mode)) { kfree(nesadapter); -- cgit v1.2.3 From 7b14ab0b438f20c7d6599985b036bb2864fd2524 Mon Sep 17 00:00:00 2001 From: Chien Tung Date: Fri, 6 Mar 2009 15:12:11 -0800 Subject: RDMA/nes: Fix tmp_addr compilation warning In find_node(), tmp_addr causes an "unused variable" warning when INFINIBAND_NES_DEBUG is not defined. It's only used in a nes_debug() and the print does not make sense. So take out the whole thing. Reported-by: Manish Katiyar Signed-off-by: Chien Tung Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_cm.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c index 6f42ab6bd02..bd918df3d22 100644 --- a/drivers/infiniband/hw/nes/nes_cm.c +++ b/drivers/infiniband/hw/nes/nes_cm.c @@ -778,14 +778,10 @@ static struct nes_cm_node *find_node(struct nes_cm_core *cm_core, unsigned long flags; struct list_head *hte; struct nes_cm_node *cm_node; - __be32 tmp_addr = cpu_to_be32(loc_addr); /* get a handle on the hte */ hte = &cm_core->connected_nodes; - nes_debug(NES_DBG_CM, "Searching for an owner node: %pI4:%x from core %p->%p\n", - &tmp_addr, loc_port, cm_core, hte); - /* walk list and find cm_node associated with this session ID */ spin_lock_irqsave(&cm_core->ht_lock, flags); list_for_each_entry(cm_node, hte, list) { -- cgit v1.2.3 From fd87778cb99429f5e2e041213a5c9c564bbe7b78 Mon Sep 17 00:00:00 2001 From: Don Wood Date: Fri, 6 Mar 2009 15:12:11 -0800 Subject: RDMA/nes: Inform hardware that asynchronous event has been handled When asynchronous events are processed by software, it is necessary to let the hardware know that software has handled the event. This frees up the entry in the asynchronous event queue. Signed-off-by: Don Wood Signed-off-by: Chien Tung Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_hw.c | 2 ++ drivers/infiniband/hw/nes/nes_hw.h | 1 + 2 files changed, 3 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index 3a72a06f1df..1c5e946ce22 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -2269,6 +2269,8 @@ static void nes_process_aeq(struct nes_device *nesdev, struct nes_hw_aeq *aeq) if (++head >= aeq_size) head = 0; + + nes_write32(nesdev->regs + NES_AEQ_ALLOC, 1 << 16); } while (1); aeq->aeq_head = head; diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h index 6f8712d7a2c..bf7ecfa5f97 100644 --- a/drivers/infiniband/hw/nes/nes_hw.h +++ b/drivers/infiniband/hw/nes/nes_hw.h @@ -61,6 +61,7 @@ enum pci_regs { NES_CQ_ACK = 0x0034, NES_WQE_ALLOC = 0x0040, NES_CQE_ALLOC = 0x0044, + NES_AEQ_ALLOC = 0x0048 }; enum indexed_regs { -- cgit v1.2.3 From 2869975cfbd58dc6591d8c3ba1f171e7f758be28 Mon Sep 17 00:00:00 2001 From: Faisal Latif Date: Fri, 6 Mar 2009 15:12:11 -0800 Subject: RDMA/nes: Remove LLTX NETIF_F_LLTX is deprecated. Remove private TX locking from the driver and remove the NETIF_F_LLTX feature flag. This also fixes a warning in some configs that comes from doing skb_linearize() call in the hard_start_xmit method with IRQs disabled (if HIGHMEM is enabled, skb_linearize() may end up enabling BHs, which is a no-no if hard IRQs are disabled in that context). By getting rid of LLTX, we do not disable IRQs when skb_linearize() is called. Remove the sq_lock as it is not needed for non-LLTX. Fix ethtool not to show the counter for sq_lock. Reported-by: aluno3@poczta.onet.pl Signed-off-by: Faisal Latif Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_hw.c | 5 +- drivers/infiniband/hw/nes/nes_hw.h | 2 - drivers/infiniband/hw/nes/nes_nic.c | 140 ++++++++++++++++-------------------- 3 files changed, 63 insertions(+), 84 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c index 1c5e946ce22..9a51f25c6ce 100644 --- a/drivers/infiniband/hw/nes/nes_hw.c +++ b/drivers/infiniband/hw/nes/nes_hw.c @@ -1644,7 +1644,6 @@ int nes_init_nic_qp(struct nes_device *nesdev, struct net_device *netdev) nesvnic->post_cqp_request = nes_post_cqp_request; nesvnic->mcrq_mcast_filter = NULL; - spin_lock_init(&nesvnic->nic.sq_lock); spin_lock_init(&nesvnic->nic.rq_lock); /* setup the RQ */ @@ -2632,9 +2631,9 @@ void nes_nic_ce_handler(struct nes_device *nesdev, struct nes_hw_nic_cq *cq) } else break; } - if (skb) - dev_kfree_skb_any(skb); } + if (skb) + dev_kfree_skb_any(skb); nesnic->sq_tail++; nesnic->sq_tail &= nesnic->sq_size-1; if (sq_cqes > 128) { diff --git a/drivers/infiniband/hw/nes/nes_hw.h b/drivers/infiniband/hw/nes/nes_hw.h index bf7ecfa5f97..f41a8710d2a 100644 --- a/drivers/infiniband/hw/nes/nes_hw.h +++ b/drivers/infiniband/hw/nes/nes_hw.h @@ -876,7 +876,6 @@ struct nes_hw_nic { u8 replenishing_rq; u8 reserved; - spinlock_t sq_lock; spinlock_t rq_lock; }; @@ -1148,7 +1147,6 @@ struct nes_ib_device; struct nes_vnic { struct nes_ib_device *nesibdev; u64 sq_full; - u64 sq_locked; u64 tso_requests; u64 segmented_tso_requests; u64 linearized_skbs; diff --git a/drivers/infiniband/hw/nes/nes_nic.c b/drivers/infiniband/hw/nes/nes_nic.c index 8e1d073913e..025ed9f7d9c 100644 --- a/drivers/infiniband/hw/nes/nes_nic.c +++ b/drivers/infiniband/hw/nes/nes_nic.c @@ -400,8 +400,7 @@ static int nes_nic_send(struct sk_buff *skb, struct net_device *netdev) if (skb_headlen(skb) == skb->len) { if (skb_headlen(skb) <= NES_FIRST_FRAG_SIZE) { nic_sqe->wqe_words[NES_NIC_SQ_WQE_LENGTH_2_1_IDX] = 0; - nesnic->tx_skb[nesnic->sq_head] = NULL; - dev_kfree_skb(skb); + nesnic->tx_skb[nesnic->sq_head] = skb; } } else { /* Deal with Fragments */ @@ -453,7 +452,6 @@ static int nes_netdev_start_xmit(struct sk_buff *skb, struct net_device *netdev) u32 wqe_count=1; u32 send_rc; struct iphdr *iph; - unsigned long flags; __le16 *wqe_fragment_length; u32 nr_frags; u32 original_first_length; @@ -480,13 +478,6 @@ static int nes_netdev_start_xmit(struct sk_buff *skb, struct net_device *netdev) if (netif_queue_stopped(netdev)) return NETDEV_TX_BUSY; - local_irq_save(flags); - if (!spin_trylock(&nesnic->sq_lock)) { - local_irq_restore(flags); - nesvnic->sq_locked++; - return NETDEV_TX_LOCKED; - } - /* Check if SQ is full */ if ((((nesnic->sq_tail+(nesnic->sq_size*2))-nesnic->sq_head) & (nesnic->sq_size - 1)) == 1) { if (!netif_queue_stopped(netdev)) { @@ -498,7 +489,6 @@ static int nes_netdev_start_xmit(struct sk_buff *skb, struct net_device *netdev) } } nesvnic->sq_full++; - spin_unlock_irqrestore(&nesnic->sq_lock, flags); return NETDEV_TX_BUSY; } @@ -531,7 +521,6 @@ sq_no_longer_full: } } nesvnic->sq_full++; - spin_unlock_irqrestore(&nesnic->sq_lock, flags); nes_debug(NES_DBG_NIC_TX, "%s: HNIC SQ full- TSO request has too many frags!\n", netdev->name); return NETDEV_TX_BUSY; @@ -656,17 +645,13 @@ tso_sq_no_longer_full: skb_set_transport_header(skb, hoffset); skb_set_network_header(skb, nhoffset); send_rc = nes_nic_send(skb, netdev); - if (send_rc != NETDEV_TX_OK) { - spin_unlock_irqrestore(&nesnic->sq_lock, flags); + if (send_rc != NETDEV_TX_OK) return NETDEV_TX_OK; - } } } else { send_rc = nes_nic_send(skb, netdev); - if (send_rc != NETDEV_TX_OK) { - spin_unlock_irqrestore(&nesnic->sq_lock, flags); + if (send_rc != NETDEV_TX_OK) return NETDEV_TX_OK; - } } barrier(); @@ -676,7 +661,6 @@ tso_sq_no_longer_full: (wqe_count << 24) | (1 << 23) | nesvnic->nic.qp_id); netdev->trans_start = jiffies; - spin_unlock_irqrestore(&nesnic->sq_lock, flags); return NETDEV_TX_OK; } @@ -1012,7 +996,6 @@ static const char nes_ethtool_stringset[][ETH_GSTRING_LEN] = { "Pause Frames Received", "Internal Routing Errors", "SQ SW Dropped SKBs", - "SQ Locked", "SQ Full", "Segmented TSO Requests", "Rx Symbol Errors", @@ -1129,16 +1112,17 @@ static void nes_netdev_get_ethtool_stats(struct net_device *netdev, struct nes_device *nesdev = nesvnic->nesdev; u32 nic_count; u32 u32temp; + u32 index = 0; target_ethtool_stats->n_stats = NES_ETHTOOL_STAT_COUNT; - target_stat_values[0] = nesvnic->nesdev->link_status_interrupts; - target_stat_values[1] = nesvnic->linearized_skbs; - target_stat_values[2] = nesvnic->tso_requests; + target_stat_values[index] = nesvnic->nesdev->link_status_interrupts; + target_stat_values[++index] = nesvnic->linearized_skbs; + target_stat_values[++index] = nesvnic->tso_requests; u32temp = nes_read_indexed(nesdev, NES_IDX_MAC_TX_PAUSE_FRAMES + (nesvnic->nesdev->mac_index*0x200)); nesvnic->nesdev->mac_pause_frames_sent += u32temp; - target_stat_values[3] = nesvnic->nesdev->mac_pause_frames_sent; + target_stat_values[++index] = nesvnic->nesdev->mac_pause_frames_sent; u32temp = nes_read_indexed(nesdev, NES_IDX_MAC_RX_PAUSE_FRAMES + (nesvnic->nesdev->mac_index*0x200)); @@ -1209,60 +1193,59 @@ static void nes_netdev_get_ethtool_stats(struct net_device *netdev, nesvnic->endnode_ipv4_tcp_retransmits += u32temp; } - target_stat_values[4] = nesvnic->nesdev->mac_pause_frames_received; - target_stat_values[5] = nesdev->nesadapter->nic_rx_eth_route_err; - target_stat_values[6] = nesvnic->tx_sw_dropped; - target_stat_values[7] = nesvnic->sq_locked; - target_stat_values[8] = nesvnic->sq_full; - target_stat_values[9] = nesvnic->segmented_tso_requests; - target_stat_values[10] = nesvnic->nesdev->mac_rx_symbol_err_frames; - target_stat_values[11] = nesvnic->nesdev->mac_rx_jabber_frames; - target_stat_values[12] = nesvnic->nesdev->mac_rx_oversized_frames; - target_stat_values[13] = nesvnic->nesdev->mac_rx_short_frames; - target_stat_values[14] = nesvnic->endnode_nstat_rx_discard; - target_stat_values[15] = nesvnic->endnode_nstat_rx_octets; - target_stat_values[16] = nesvnic->endnode_nstat_rx_frames; - target_stat_values[17] = nesvnic->endnode_nstat_tx_octets; - target_stat_values[18] = nesvnic->endnode_nstat_tx_frames; - target_stat_values[19] = mh_detected; - target_stat_values[20] = mh_pauses_sent; - target_stat_values[21] = nesvnic->endnode_ipv4_tcp_retransmits; - target_stat_values[22] = atomic_read(&cm_connects); - target_stat_values[23] = atomic_read(&cm_accepts); - target_stat_values[24] = atomic_read(&cm_disconnects); - target_stat_values[25] = atomic_read(&cm_connecteds); - target_stat_values[26] = atomic_read(&cm_connect_reqs); - target_stat_values[27] = atomic_read(&cm_rejects); - target_stat_values[28] = atomic_read(&mod_qp_timouts); - target_stat_values[29] = atomic_read(&qps_created); - target_stat_values[30] = atomic_read(&sw_qps_destroyed); - target_stat_values[31] = atomic_read(&qps_destroyed); - target_stat_values[32] = atomic_read(&cm_closes); - target_stat_values[33] = cm_packets_sent; - target_stat_values[34] = cm_packets_bounced; - target_stat_values[35] = cm_packets_created; - target_stat_values[36] = cm_packets_received; - target_stat_values[37] = cm_packets_dropped; - target_stat_values[38] = cm_packets_retrans; - target_stat_values[39] = cm_listens_created; - target_stat_values[40] = cm_listens_destroyed; - target_stat_values[41] = cm_backlog_drops; - target_stat_values[42] = atomic_read(&cm_loopbacks); - target_stat_values[43] = atomic_read(&cm_nodes_created); - target_stat_values[44] = atomic_read(&cm_nodes_destroyed); - target_stat_values[45] = atomic_read(&cm_accel_dropped_pkts); - target_stat_values[46] = atomic_read(&cm_resets_recvd); - target_stat_values[47] = int_mod_timer_init; - target_stat_values[48] = int_mod_cq_depth_1; - target_stat_values[49] = int_mod_cq_depth_4; - target_stat_values[50] = int_mod_cq_depth_16; - target_stat_values[51] = int_mod_cq_depth_24; - target_stat_values[52] = int_mod_cq_depth_32; - target_stat_values[53] = int_mod_cq_depth_128; - target_stat_values[54] = int_mod_cq_depth_256; - target_stat_values[55] = nesvnic->lro_mgr.stats.aggregated; - target_stat_values[56] = nesvnic->lro_mgr.stats.flushed; - target_stat_values[57] = nesvnic->lro_mgr.stats.no_desc; + target_stat_values[++index] = nesvnic->nesdev->mac_pause_frames_received; + target_stat_values[++index] = nesdev->nesadapter->nic_rx_eth_route_err; + target_stat_values[++index] = nesvnic->tx_sw_dropped; + target_stat_values[++index] = nesvnic->sq_full; + target_stat_values[++index] = nesvnic->segmented_tso_requests; + target_stat_values[++index] = nesvnic->nesdev->mac_rx_symbol_err_frames; + target_stat_values[++index] = nesvnic->nesdev->mac_rx_jabber_frames; + target_stat_values[++index] = nesvnic->nesdev->mac_rx_oversized_frames; + target_stat_values[++index] = nesvnic->nesdev->mac_rx_short_frames; + target_stat_values[++index] = nesvnic->endnode_nstat_rx_discard; + target_stat_values[++index] = nesvnic->endnode_nstat_rx_octets; + target_stat_values[++index] = nesvnic->endnode_nstat_rx_frames; + target_stat_values[++index] = nesvnic->endnode_nstat_tx_octets; + target_stat_values[++index] = nesvnic->endnode_nstat_tx_frames; + target_stat_values[++index] = mh_detected; + target_stat_values[++index] = mh_pauses_sent; + target_stat_values[++index] = nesvnic->endnode_ipv4_tcp_retransmits; + target_stat_values[++index] = atomic_read(&cm_connects); + target_stat_values[++index] = atomic_read(&cm_accepts); + target_stat_values[++index] = atomic_read(&cm_disconnects); + target_stat_values[++index] = atomic_read(&cm_connecteds); + target_stat_values[++index] = atomic_read(&cm_connect_reqs); + target_stat_values[++index] = atomic_read(&cm_rejects); + target_stat_values[++index] = atomic_read(&mod_qp_timouts); + target_stat_values[++index] = atomic_read(&qps_created); + target_stat_values[++index] = atomic_read(&sw_qps_destroyed); + target_stat_values[++index] = atomic_read(&qps_destroyed); + target_stat_values[++index] = atomic_read(&cm_closes); + target_stat_values[++index] = cm_packets_sent; + target_stat_values[++index] = cm_packets_bounced; + target_stat_values[++index] = cm_packets_created; + target_stat_values[++index] = cm_packets_received; + target_stat_values[++index] = cm_packets_dropped; + target_stat_values[++index] = cm_packets_retrans; + target_stat_values[++index] = cm_listens_created; + target_stat_values[++index] = cm_listens_destroyed; + target_stat_values[++index] = cm_backlog_drops; + target_stat_values[++index] = atomic_read(&cm_loopbacks); + target_stat_values[++index] = atomic_read(&cm_nodes_created); + target_stat_values[++index] = atomic_read(&cm_nodes_destroyed); + target_stat_values[++index] = atomic_read(&cm_accel_dropped_pkts); + target_stat_values[++index] = atomic_read(&cm_resets_recvd); + target_stat_values[++index] = int_mod_timer_init; + target_stat_values[++index] = int_mod_cq_depth_1; + target_stat_values[++index] = int_mod_cq_depth_4; + target_stat_values[++index] = int_mod_cq_depth_16; + target_stat_values[++index] = int_mod_cq_depth_24; + target_stat_values[++index] = int_mod_cq_depth_32; + target_stat_values[++index] = int_mod_cq_depth_128; + target_stat_values[++index] = int_mod_cq_depth_256; + target_stat_values[++index] = nesvnic->lro_mgr.stats.aggregated; + target_stat_values[++index] = nesvnic->lro_mgr.stats.flushed; + target_stat_values[++index] = nesvnic->lro_mgr.stats.no_desc; } @@ -1616,7 +1599,6 @@ struct net_device *nes_netdev_init(struct nes_device *nesdev, nes_debug(NES_DBG_INIT, "Enabling VLAN Insert/Delete.\n"); netdev->features |= NETIF_F_HW_VLAN_TX | NETIF_F_HW_VLAN_RX; netdev->vlan_rx_register = nes_netdev_vlan_rx_register; - netdev->features |= NETIF_F_LLTX; /* Fill in the port structure */ nesvnic->netdev = netdev; -- cgit v1.2.3 From 0145f341a951b998d6d0fa38992a42d2a90b5bab Mon Sep 17 00:00:00 2001 From: Don Wood Date: Fri, 6 Mar 2009 15:15:00 -0800 Subject: RDMA/nes: Improve use of PBLs Two level 256 byte PBLs was not implemented so the driver could report out of memory when in fact there were PBLs still available. This solution prefers to use 4KB PBLs over two level 256B PBLs until the number of 4KB PBLs falls below a threshold. At this point the 4KB PBL structure is converted to use 256B PBLs which prevents the driver from running out of 4KB PBLs too quickly. Signed-off-by: Don Wood Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_verbs.c | 191 ++++++++++++++++++++++++---------- 1 file changed, 134 insertions(+), 57 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_verbs.c b/drivers/infiniband/hw/nes/nes_verbs.c index b42b17ac771..96d953540a2 100644 --- a/drivers/infiniband/hw/nes/nes_verbs.c +++ b/drivers/infiniband/hw/nes/nes_verbs.c @@ -1887,21 +1887,75 @@ static int nes_destroy_cq(struct ib_cq *ib_cq) return ret; } +/** + * root_256 + */ +static u32 root_256(struct nes_device *nesdev, + struct nes_root_vpbl *root_vpbl, + struct nes_root_vpbl *new_root, + u16 pbl_count_4k, + u16 pbl_count_256) +{ + u64 leaf_pbl; + int i, j, k; + + if (pbl_count_4k == 1) { + new_root->pbl_vbase = pci_alloc_consistent(nesdev->pcidev, + 512, &new_root->pbl_pbase); + + if (new_root->pbl_vbase == NULL) + return 0; + + leaf_pbl = (u64)root_vpbl->pbl_pbase; + for (i = 0; i < 16; i++) { + new_root->pbl_vbase[i].pa_low = + cpu_to_le32((u32)leaf_pbl); + new_root->pbl_vbase[i].pa_high = + cpu_to_le32((u32)((((u64)leaf_pbl) >> 32))); + leaf_pbl += 256; + } + } else { + for (i = 3; i >= 0; i--) { + j = i * 16; + root_vpbl->pbl_vbase[j] = root_vpbl->pbl_vbase[i]; + leaf_pbl = le32_to_cpu(root_vpbl->pbl_vbase[j].pa_low) + + (((u64)le32_to_cpu(root_vpbl->pbl_vbase[j].pa_high)) + << 32); + for (k = 1; k < 16; k++) { + leaf_pbl += 256; + root_vpbl->pbl_vbase[j + k].pa_low = + cpu_to_le32((u32)leaf_pbl); + root_vpbl->pbl_vbase[j + k].pa_high = + cpu_to_le32((u32)((((u64)leaf_pbl) >> 32))); + } + } + } + + return 1; +} + /** * nes_reg_mr */ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, u32 stag, u64 region_length, struct nes_root_vpbl *root_vpbl, - dma_addr_t single_buffer, u16 pbl_count, u16 residual_page_count, - int acc, u64 *iova_start) + dma_addr_t single_buffer, u16 pbl_count_4k, + u16 residual_page_count_4k, int acc, u64 *iova_start, + u16 *actual_pbl_cnt, u8 *used_4k_pbls) { struct nes_hw_cqp_wqe *cqp_wqe; struct nes_cqp_request *cqp_request; unsigned long flags; int ret; struct nes_adapter *nesadapter = nesdev->nesadapter; - /* int count; */ + uint pg_cnt = 0; + u16 pbl_count_256; + u16 pbl_count = 0; + u8 use_256_pbls = 0; + u8 use_4k_pbls = 0; + u16 use_two_level = (pbl_count_4k > 1) ? 1 : 0; + struct nes_root_vpbl new_root = {0, 0, 0}; u32 opcode = 0; u16 major_code; @@ -1914,41 +1968,70 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, cqp_request->waiting = 1; cqp_wqe = &cqp_request->cqp_wqe; - spin_lock_irqsave(&nesadapter->pbl_lock, flags); - /* track PBL resources */ - if (pbl_count != 0) { - if (pbl_count > 1) { - /* Two level PBL */ - if ((pbl_count+1) > nesadapter->free_4kpbl) { - nes_debug(NES_DBG_MR, "Out of 4KB Pbls for two level request.\n"); - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - nes_free_cqp_request(nesdev, cqp_request); - return -ENOMEM; - } else { - nesadapter->free_4kpbl -= pbl_count+1; - } - } else if (residual_page_count > 32) { - if (pbl_count > nesadapter->free_4kpbl) { - nes_debug(NES_DBG_MR, "Out of 4KB Pbls.\n"); - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - nes_free_cqp_request(nesdev, cqp_request); - return -ENOMEM; - } else { - nesadapter->free_4kpbl -= pbl_count; + if (pbl_count_4k) { + spin_lock_irqsave(&nesadapter->pbl_lock, flags); + + pg_cnt = ((pbl_count_4k - 1) * 512) + residual_page_count_4k; + pbl_count_256 = (pg_cnt + 31) / 32; + if (pg_cnt <= 32) { + if (pbl_count_256 <= nesadapter->free_256pbl) + use_256_pbls = 1; + else if (pbl_count_4k <= nesadapter->free_4kpbl) + use_4k_pbls = 1; + } else if (pg_cnt <= 2048) { + if (((pbl_count_4k + use_two_level) <= nesadapter->free_4kpbl) && + (nesadapter->free_4kpbl > (nesadapter->max_4kpbl >> 1))) { + use_4k_pbls = 1; + } else if ((pbl_count_256 + 1) <= nesadapter->free_256pbl) { + use_256_pbls = 1; + use_two_level = 1; + } else if ((pbl_count_4k + use_two_level) <= nesadapter->free_4kpbl) { + use_4k_pbls = 1; } } else { - if (pbl_count > nesadapter->free_256pbl) { - nes_debug(NES_DBG_MR, "Out of 256B Pbls.\n"); - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); - nes_free_cqp_request(nesdev, cqp_request); - return -ENOMEM; - } else { - nesadapter->free_256pbl -= pbl_count; - } + if ((pbl_count_4k + 1) <= nesadapter->free_4kpbl) + use_4k_pbls = 1; + } + + if (use_256_pbls) { + pbl_count = pbl_count_256; + nesadapter->free_256pbl -= pbl_count + use_two_level; + } else if (use_4k_pbls) { + pbl_count = pbl_count_4k; + nesadapter->free_4kpbl -= pbl_count + use_two_level; + } else { + spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + nes_debug(NES_DBG_MR, "Out of Pbls\n"); + nes_free_cqp_request(nesdev, cqp_request); + return -ENOMEM; } + + spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); } - spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + if (use_256_pbls && use_two_level) { + if (root_256(nesdev, root_vpbl, &new_root, pbl_count_4k, pbl_count_256) == 1) { + if (new_root.pbl_pbase != 0) + root_vpbl = &new_root; + } else { + spin_lock_irqsave(&nesadapter->pbl_lock, flags); + nesadapter->free_256pbl += pbl_count_256 + use_two_level; + use_256_pbls = 0; + + if (pbl_count_4k == 1) + use_two_level = 0; + pbl_count = pbl_count_4k; + + if ((pbl_count_4k + use_two_level) <= nesadapter->free_4kpbl) { + nesadapter->free_4kpbl -= pbl_count + use_two_level; + use_4k_pbls = 1; + } + spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); + + if (use_4k_pbls == 0) + return -ENOMEM; + } + } opcode = NES_CQP_REGISTER_STAG | NES_CQP_STAG_RIGHTS_LOCAL_READ | NES_CQP_STAG_VA_TO | NES_CQP_STAG_MR; @@ -1977,10 +2060,9 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, } else { set_wqe_64bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_PA_LOW_IDX, root_vpbl->pbl_pbase); set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_PBL_BLK_COUNT_IDX, pbl_count); - set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_PBL_LEN_IDX, - (((pbl_count - 1) * 4096) + (residual_page_count*8))); + set_wqe_32bit_value(cqp_wqe->wqe_words, NES_CQP_STAG_WQE_PBL_LEN_IDX, (pg_cnt * 8)); - if ((pbl_count > 1) || (residual_page_count > 32)) + if (use_4k_pbls) cqp_wqe->wqe_words[NES_CQP_WQE_OPCODE_IDX] |= cpu_to_le32(NES_CQP_STAG_PBL_BLK_SIZE); } barrier(); @@ -1996,23 +2078,26 @@ static int nes_reg_mr(struct nes_device *nesdev, struct nes_pd *nespd, stag, ret, cqp_request->major_code, cqp_request->minor_code); major_code = cqp_request->major_code; nes_put_cqp_request(nesdev, cqp_request); + if ((!ret || major_code) && pbl_count != 0) { spin_lock_irqsave(&nesadapter->pbl_lock, flags); - if (pbl_count > 1) - nesadapter->free_4kpbl += pbl_count+1; - else if (residual_page_count > 32) - nesadapter->free_4kpbl += pbl_count; - else - nesadapter->free_256pbl += pbl_count; + if (use_256_pbls) + nesadapter->free_256pbl += pbl_count + use_two_level; + else if (use_4k_pbls) + nesadapter->free_4kpbl += pbl_count + use_two_level; spin_unlock_irqrestore(&nesadapter->pbl_lock, flags); } + if (new_root.pbl_pbase) + pci_free_consistent(nesdev->pcidev, 512, new_root.pbl_vbase, + new_root.pbl_pbase); + if (!ret) return -ETIME; else if (major_code) return -EIO; - else - return 0; + *actual_pbl_cnt = pbl_count + use_two_level; + *used_4k_pbls = use_4k_pbls; return 0; } @@ -2177,18 +2262,14 @@ static struct ib_mr *nes_reg_phys_mr(struct ib_pd *ib_pd, pbl_count = root_pbl_index; } ret = nes_reg_mr(nesdev, nespd, stag, region_length, &root_vpbl, - buffer_list[0].addr, pbl_count, (u16)cur_pbl_index, acc, iova_start); + buffer_list[0].addr, pbl_count, (u16)cur_pbl_index, acc, iova_start, + &nesmr->pbls_used, &nesmr->pbl_4k); if (ret == 0) { nesmr->ibmr.rkey = stag; nesmr->ibmr.lkey = stag; nesmr->mode = IWNES_MEMREG_TYPE_MEM; ibmr = &nesmr->ibmr; - nesmr->pbl_4k = ((pbl_count > 1) || (cur_pbl_index > 32)) ? 1 : 0; - nesmr->pbls_used = pbl_count; - if (pbl_count > 1) { - nesmr->pbls_used++; - } } else { kfree(nesmr); ibmr = ERR_PTR(-ENOMEM); @@ -2466,8 +2547,9 @@ static struct ib_mr *nes_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, stag, (unsigned int)iova_start, (unsigned int)region_length, stag_index, (unsigned long long)region->length, pbl_count); - ret = nes_reg_mr( nesdev, nespd, stag, region->length, &root_vpbl, - first_dma_addr, pbl_count, (u16)cur_pbl_index, acc, &iova_start); + ret = nes_reg_mr(nesdev, nespd, stag, region->length, &root_vpbl, + first_dma_addr, pbl_count, (u16)cur_pbl_index, acc, + &iova_start, &nesmr->pbls_used, &nesmr->pbl_4k); nes_debug(NES_DBG_MR, "ret=%d\n", ret); @@ -2476,11 +2558,6 @@ static struct ib_mr *nes_reg_user_mr(struct ib_pd *pd, u64 start, u64 length, nesmr->ibmr.lkey = stag; nesmr->mode = IWNES_MEMREG_TYPE_MEM; ibmr = &nesmr->ibmr; - nesmr->pbl_4k = ((pbl_count > 1) || (cur_pbl_index > 32)) ? 1 : 0; - nesmr->pbls_used = pbl_count; - if (pbl_count > 1) { - nesmr->pbls_used++; - } } else { ib_umem_release(region); kfree(nesmr); -- cgit v1.2.3 From 9d5ab13325d0bb855cf856946c140a68ceb20e32 Mon Sep 17 00:00:00 2001 From: Faisal Latif Date: Fri, 6 Mar 2009 15:15:01 -0800 Subject: RDMA/nes: Handle MPA Reject message properly While doing testing, there are failures as MPA Reject call is not handled. To handle MPA Reject call, following changes are done: *Handle inbound/outbound MPA Reject response message. When nes_reject() is called for pending MPA request reply, send the MPA Reject message to its peer (active side)cm_node. The peer cm_node (active side) will indicate Reject message event for the pending Connect Request. *Handle MPA Reject response message for loopback connections and listener. When MPA Request is rejected, check if it is a loopback connection and if it is then it will send Reject message event to its peer loopback node. Also when destroying listener, check if the cm_nodes for that listener are loopback or not. *Add gracefull connection close with the MPA Reject response message. Send gracefull close (FIN, FIN ACK..) to terminate the cm_nodes. *Some code re-org while making the above changes. Removed recv_list and recv_list_lock from the cm_node structure as there can be only one receive close entry on the timer. Also implemented handle_recv_entry() as receive close entry is processed from both nes_rem_ref_cm_node() as well as nes_cm_timer_tick(). Signed-off-by: Faisal Latif Signed-off-by: Roland Dreier --- drivers/infiniband/hw/nes/nes_cm.c | 580 +++++++++++++++++++++++++------------ drivers/infiniband/hw/nes/nes_cm.h | 10 +- 2 files changed, 399 insertions(+), 191 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/nes/nes_cm.c b/drivers/infiniband/hw/nes/nes_cm.c index bd918df3d22..5327f2bec6b 100644 --- a/drivers/infiniband/hw/nes/nes_cm.c +++ b/drivers/infiniband/hw/nes/nes_cm.c @@ -103,6 +103,7 @@ static int nes_disconnect(struct nes_qp *nesqp, int abrupt); static void nes_disconnect_worker(struct work_struct *work); static int send_mpa_request(struct nes_cm_node *, struct sk_buff *); +static int send_mpa_reject(struct nes_cm_node *); static int send_syn(struct nes_cm_node *, u32, struct sk_buff *); static int send_reset(struct nes_cm_node *, struct sk_buff *); static int send_ack(struct nes_cm_node *cm_node, struct sk_buff *skb); @@ -113,8 +114,7 @@ static void process_packet(struct nes_cm_node *, struct sk_buff *, static void active_open_err(struct nes_cm_node *, struct sk_buff *, int); static void passive_open_err(struct nes_cm_node *, struct sk_buff *, int); static void cleanup_retrans_entry(struct nes_cm_node *); -static void handle_rcv_mpa(struct nes_cm_node *, struct sk_buff *, - enum nes_cm_event_type); +static void handle_rcv_mpa(struct nes_cm_node *, struct sk_buff *); static void free_retrans_entry(struct nes_cm_node *cm_node); static int handle_tcp_options(struct nes_cm_node *cm_node, struct tcphdr *tcph, struct sk_buff *skb, int optionsize, int passive); @@ -124,6 +124,8 @@ static void cm_event_connected(struct nes_cm_event *); static void cm_event_connect_error(struct nes_cm_event *); static void cm_event_reset(struct nes_cm_event *); static void cm_event_mpa_req(struct nes_cm_event *); +static void cm_event_mpa_reject(struct nes_cm_event *); +static void handle_recv_entry(struct nes_cm_node *cm_node, u32 rem_node); static void print_core(struct nes_cm_core *core); @@ -196,7 +198,6 @@ static struct nes_cm_event *create_event(struct nes_cm_node *cm_node, */ static int send_mpa_request(struct nes_cm_node *cm_node, struct sk_buff *skb) { - int ret; if (!skb) { nes_debug(NES_DBG_CM, "skb set to NULL\n"); return -1; @@ -206,11 +207,27 @@ static int send_mpa_request(struct nes_cm_node *cm_node, struct sk_buff *skb) form_cm_frame(skb, cm_node, NULL, 0, &cm_node->mpa_frame, cm_node->mpa_frame_size, SET_ACK); - ret = schedule_nes_timer(cm_node, skb, NES_TIMER_TYPE_SEND, 1, 0); - if (ret < 0) - return ret; + return schedule_nes_timer(cm_node, skb, NES_TIMER_TYPE_SEND, 1, 0); +} - return 0; + + +static int send_mpa_reject(struct nes_cm_node *cm_node) +{ + struct sk_buff *skb = NULL; + + skb = dev_alloc_skb(MAX_CM_BUFFER); + if (!skb) { + nes_debug(NES_DBG_CM, "Failed to get a Free pkt\n"); + return -ENOMEM; + } + + /* send an MPA reject frame */ + form_cm_frame(skb, cm_node, NULL, 0, &cm_node->mpa_frame, + cm_node->mpa_frame_size, SET_ACK | SET_FIN); + + cm_node->state = NES_CM_STATE_FIN_WAIT1; + return schedule_nes_timer(cm_node, skb, NES_TIMER_TYPE_SEND, 1, 0); } @@ -218,14 +235,17 @@ static int send_mpa_request(struct nes_cm_node *cm_node, struct sk_buff *skb) * recv_mpa - process a received TCP pkt, we are expecting an * IETF MPA frame */ -static int parse_mpa(struct nes_cm_node *cm_node, u8 *buffer, u32 len) +static int parse_mpa(struct nes_cm_node *cm_node, u8 *buffer, u32 *type, + u32 len) { struct ietf_mpa_frame *mpa_frame; + *type = NES_MPA_REQUEST_ACCEPT; + /* assume req frame is in tcp data payload */ if (len < sizeof(struct ietf_mpa_frame)) { nes_debug(NES_DBG_CM, "The received ietf buffer was too small (%x)\n", len); - return -1; + return -EINVAL; } mpa_frame = (struct ietf_mpa_frame *)buffer; @@ -234,14 +254,25 @@ static int parse_mpa(struct nes_cm_node *cm_node, u8 *buffer, u32 len) if (cm_node->mpa_frame_size + sizeof(struct ietf_mpa_frame) != len) { nes_debug(NES_DBG_CM, "The received ietf buffer was not right" " complete (%x + %x != %x)\n", - cm_node->mpa_frame_size, (u32)sizeof(struct ietf_mpa_frame), len); - return -1; + cm_node->mpa_frame_size, + (u32)sizeof(struct ietf_mpa_frame), len); + return -EINVAL; + } + /* make sure it does not exceed the max size */ + if (len > MAX_CM_BUFFER) { + nes_debug(NES_DBG_CM, "The received ietf buffer was too large" + " (%x + %x != %x)\n", + cm_node->mpa_frame_size, + (u32)sizeof(struct ietf_mpa_frame), len); + return -EINVAL; } /* copy entire MPA frame to our cm_node's frame */ memcpy(cm_node->mpa_frame_buf, buffer + sizeof(struct ietf_mpa_frame), cm_node->mpa_frame_size); + if (mpa_frame->flags & IETF_MPA_FLAGS_REJECT) + *type = NES_MPA_REQUEST_REJECT; return 0; } @@ -380,7 +411,7 @@ int schedule_nes_timer(struct nes_cm_node *cm_node, struct sk_buff *skb, new_send = kzalloc(sizeof(*new_send), GFP_ATOMIC); if (!new_send) - return -1; + return -ENOMEM; /* new_send->timetosend = currenttime */ new_send->retrycount = NES_DEFAULT_RETRYS; @@ -394,9 +425,11 @@ int schedule_nes_timer(struct nes_cm_node *cm_node, struct sk_buff *skb, if (type == NES_TIMER_TYPE_CLOSE) { new_send->timetosend += (HZ/10); - spin_lock_irqsave(&cm_node->recv_list_lock, flags); - list_add_tail(&new_send->list, &cm_node->recv_list); - spin_unlock_irqrestore(&cm_node->recv_list_lock, flags); + if (cm_node->recv_entry) { + WARN_ON(1); + return -EINVAL; + } + cm_node->recv_entry = new_send; } if (type == NES_TIMER_TYPE_SEND) { @@ -435,24 +468,78 @@ int schedule_nes_timer(struct nes_cm_node *cm_node, struct sk_buff *skb, return ret; } +static void nes_retrans_expired(struct nes_cm_node *cm_node) +{ + switch (cm_node->state) { + case NES_CM_STATE_SYN_RCVD: + case NES_CM_STATE_CLOSING: + rem_ref_cm_node(cm_node->cm_core, cm_node); + break; + case NES_CM_STATE_LAST_ACK: + case NES_CM_STATE_FIN_WAIT1: + case NES_CM_STATE_MPAREJ_RCVD: + send_reset(cm_node, NULL); + break; + default: + create_event(cm_node, NES_CM_EVENT_ABORTED); + } +} + +static void handle_recv_entry(struct nes_cm_node *cm_node, u32 rem_node) +{ + struct nes_timer_entry *recv_entry = cm_node->recv_entry; + struct iw_cm_id *cm_id = cm_node->cm_id; + struct nes_qp *nesqp; + unsigned long qplockflags; + + if (!recv_entry) + return; + nesqp = (struct nes_qp *)recv_entry->skb; + if (nesqp) { + spin_lock_irqsave(&nesqp->lock, qplockflags); + if (nesqp->cm_id) { + nes_debug(NES_DBG_CM, "QP%u: cm_id = %p, " + "refcount = %d: HIT A " + "NES_TIMER_TYPE_CLOSE with something " + "to do!!!\n", nesqp->hwqp.qp_id, cm_id, + atomic_read(&nesqp->refcount)); + nesqp->hw_tcp_state = NES_AEQE_TCP_STATE_CLOSED; + nesqp->last_aeq = NES_AEQE_AEID_RESET_SENT; + nesqp->ibqp_state = IB_QPS_ERR; + spin_unlock_irqrestore(&nesqp->lock, qplockflags); + nes_cm_disconn(nesqp); + } else { + spin_unlock_irqrestore(&nesqp->lock, qplockflags); + nes_debug(NES_DBG_CM, "QP%u: cm_id = %p, " + "refcount = %d: HIT A " + "NES_TIMER_TYPE_CLOSE with nothing " + "to do!!!\n", nesqp->hwqp.qp_id, cm_id, + atomic_read(&nesqp->refcount)); + } + } else if (rem_node) { + /* TIME_WAIT state */ + rem_ref_cm_node(cm_node->cm_core, cm_node); + } + if (cm_node->cm_id) + cm_id->rem_ref(cm_id); + kfree(recv_entry); + cm_node->recv_entry = NULL; +} /** * nes_cm_timer_tick */ static void nes_cm_timer_tick(unsigned long pass) { - unsigned long flags, qplockflags; + unsigned long flags; unsigned long nexttimeout = jiffies + NES_LONG_TIME; - struct iw_cm_id *cm_id; struct nes_cm_node *cm_node; struct nes_timer_entry *send_entry, *recv_entry; - struct list_head *list_core, *list_core_temp; - struct list_head *list_node, *list_node_temp; + struct list_head *list_core_temp; + struct list_head *list_node; struct nes_cm_core *cm_core = g_cm_core; - struct nes_qp *nesqp; u32 settimer = 0; int ret = NETDEV_TX_OK; - enum nes_cm_node_state last_state; struct list_head timer_list; INIT_LIST_HEAD(&timer_list); @@ -461,7 +548,7 @@ static void nes_cm_timer_tick(unsigned long pass) list_for_each_safe(list_node, list_core_temp, &cm_core->connected_nodes) { cm_node = container_of(list_node, struct nes_cm_node, list); - if (!list_empty(&cm_node->recv_list) || (cm_node->send_entry)) { + if ((cm_node->recv_entry) || (cm_node->send_entry)) { add_ref_cm_node(cm_node); list_add(&cm_node->timer_entry, &timer_list); } @@ -471,54 +558,18 @@ static void nes_cm_timer_tick(unsigned long pass) list_for_each_safe(list_node, list_core_temp, &timer_list) { cm_node = container_of(list_node, struct nes_cm_node, timer_entry); - spin_lock_irqsave(&cm_node->recv_list_lock, flags); - list_for_each_safe(list_core, list_node_temp, - &cm_node->recv_list) { - recv_entry = container_of(list_core, - struct nes_timer_entry, list); - if (!recv_entry) - break; + recv_entry = cm_node->recv_entry; + + if (recv_entry) { if (time_after(recv_entry->timetosend, jiffies)) { if (nexttimeout > recv_entry->timetosend || - !settimer) { + !settimer) { nexttimeout = recv_entry->timetosend; settimer = 1; } - continue; - } - list_del(&recv_entry->list); - cm_id = cm_node->cm_id; - spin_unlock_irqrestore(&cm_node->recv_list_lock, flags); - nesqp = (struct nes_qp *)recv_entry->skb; - spin_lock_irqsave(&nesqp->lock, qplockflags); - if (nesqp->cm_id) { - nes_debug(NES_DBG_CM, "QP%u: cm_id = %p, " - "refcount = %d: HIT A " - "NES_TIMER_TYPE_CLOSE with something " - "to do!!!\n", nesqp->hwqp.qp_id, cm_id, - atomic_read(&nesqp->refcount)); - nesqp->hw_tcp_state = NES_AEQE_TCP_STATE_CLOSED; - nesqp->last_aeq = NES_AEQE_AEID_RESET_SENT; - nesqp->ibqp_state = IB_QPS_ERR; - spin_unlock_irqrestore(&nesqp->lock, - qplockflags); - nes_cm_disconn(nesqp); - } else { - spin_unlock_irqrestore(&nesqp->lock, - qplockflags); - nes_debug(NES_DBG_CM, "QP%u: cm_id = %p, " - "refcount = %d: HIT A " - "NES_TIMER_TYPE_CLOSE with nothing " - "to do!!!\n", nesqp->hwqp.qp_id, cm_id, - atomic_read(&nesqp->refcount)); - } - if (cm_id) - cm_id->rem_ref(cm_id); - - kfree(recv_entry); - spin_lock_irqsave(&cm_node->recv_list_lock, flags); + } else + handle_recv_entry(cm_node, 1); } - spin_unlock_irqrestore(&cm_node->recv_list_lock, flags); spin_lock_irqsave(&cm_node->retrans_list_lock, flags); do { @@ -533,12 +584,11 @@ static void nes_cm_timer_tick(unsigned long pass) nexttimeout = send_entry->timetosend; settimer = 1; - break; } } else { free_retrans_entry(cm_node); - break; } + break; } if ((cm_node->state == NES_CM_STATE_TSA) || @@ -550,16 +600,12 @@ static void nes_cm_timer_tick(unsigned long pass) if (!send_entry->retranscount || !send_entry->retrycount) { cm_packets_dropped++; - last_state = cm_node->state; - cm_node->state = NES_CM_STATE_CLOSED; free_retrans_entry(cm_node); + spin_unlock_irqrestore( &cm_node->retrans_list_lock, flags); - if (last_state == NES_CM_STATE_SYN_RCVD) - rem_ref_cm_node(cm_core, cm_node); - else - create_event(cm_node, - NES_CM_EVENT_ABORTED); + nes_retrans_expired(cm_node); + cm_node->state = NES_CM_STATE_CLOSED; spin_lock_irqsave(&cm_node->retrans_list_lock, flags); break; @@ -714,7 +760,7 @@ static int send_reset(struct nes_cm_node *cm_node, struct sk_buff *skb) skb = dev_alloc_skb(MAX_CM_BUFFER); if (!skb) { nes_debug(NES_DBG_CM, "Failed to get a Free pkt\n"); - return -1; + return -ENOMEM; } form_cm_frame(skb, cm_node, NULL, 0, NULL, 0, flags); @@ -871,7 +917,8 @@ static int add_hte_node(struct nes_cm_core *cm_core, struct nes_cm_node *cm_node static int mini_cm_dec_refcnt_listen(struct nes_cm_core *cm_core, struct nes_cm_listener *listener, int free_hanging_nodes) { - int ret = 1; + int ret = -EINVAL; + int err = 0; unsigned long flags; struct list_head *list_pos = NULL; struct list_head *list_temp = NULL; @@ -900,10 +947,60 @@ static int mini_cm_dec_refcnt_listen(struct nes_cm_core *cm_core, list_for_each_safe(list_pos, list_temp, &reset_list) { cm_node = container_of(list_pos, struct nes_cm_node, - reset_entry); - cleanup_retrans_entry(cm_node); - send_reset(cm_node, NULL); - rem_ref_cm_node(cm_node->cm_core, cm_node); + reset_entry); + { + struct nes_cm_node *loopback = cm_node->loopbackpartner; + if (NES_CM_STATE_FIN_WAIT1 <= cm_node->state) { + rem_ref_cm_node(cm_node->cm_core, cm_node); + } else { + if (!loopback) { + cleanup_retrans_entry(cm_node); + err = send_reset(cm_node, NULL); + if (err) { + cm_node->state = + NES_CM_STATE_CLOSED; + WARN_ON(1); + } else { + cm_node->state = + NES_CM_STATE_CLOSED; + rem_ref_cm_node( + cm_node->cm_core, + cm_node); + } + } else { + struct nes_cm_event event; + + event.cm_node = loopback; + event.cm_info.rem_addr = + loopback->rem_addr; + event.cm_info.loc_addr = + loopback->loc_addr; + event.cm_info.rem_port = + loopback->rem_port; + event.cm_info.loc_port = + loopback->loc_port; + event.cm_info.cm_id = loopback->cm_id; + cm_event_connect_error(&event); + loopback->state = NES_CM_STATE_CLOSED; + + event.cm_node = cm_node; + event.cm_info.rem_addr = + cm_node->rem_addr; + event.cm_info.loc_addr = + cm_node->loc_addr; + event.cm_info.rem_port = + cm_node->rem_port; + event.cm_info.loc_port = + cm_node->loc_port; + event.cm_info.cm_id = cm_node->cm_id; + cm_event_reset(&event); + + rem_ref_cm_node(cm_node->cm_core, + cm_node); + + } + } + } } spin_lock_irqsave(&cm_core->listen_list_lock, flags); @@ -964,6 +1061,7 @@ static inline int mini_cm_accelerated(struct nes_cm_core *cm_core, if (cm_node->accept_pend) { BUG_ON(!cm_node->listener); atomic_dec(&cm_node->listener->pend_accepts_cnt); + cm_node->accept_pend = 0; BUG_ON(atomic_read(&cm_node->listener->pend_accepts_cnt) < 0); } @@ -990,7 +1088,7 @@ static int nes_addr_resolve_neigh(struct nes_vnic *nesvnic, u32 dst_ip) memset(&fl, 0, sizeof fl); fl.nl_u.ip4_u.daddr = htonl(dst_ip); if (ip_route_output_key(&init_net, &rt, &fl)) { - printk("%s: ip_route_output_key failed for 0x%08X\n", + printk(KERN_ERR "%s: ip_route_output_key failed for 0x%08X\n", __func__, dst_ip); return rc; } @@ -1053,8 +1151,6 @@ static struct nes_cm_node *make_cm_node(struct nes_cm_core *cm_core, cm_node->cm_id); spin_lock_init(&cm_node->retrans_list_lock); - INIT_LIST_HEAD(&cm_node->recv_list); - spin_lock_init(&cm_node->recv_list_lock); cm_node->loopbackpartner = NULL; atomic_set(&cm_node->ref_count, 1); @@ -1122,10 +1218,7 @@ static int add_ref_cm_node(struct nes_cm_node *cm_node) static int rem_ref_cm_node(struct nes_cm_core *cm_core, struct nes_cm_node *cm_node) { - unsigned long flags, qplockflags; - struct nes_timer_entry *recv_entry; - struct iw_cm_id *cm_id; - struct list_head *list_core, *list_node_temp; + unsigned long flags; struct nes_qp *nesqp; if (!cm_node) @@ -1146,38 +1239,9 @@ static int rem_ref_cm_node(struct nes_cm_core *cm_core, atomic_dec(&cm_node->listener->pend_accepts_cnt); BUG_ON(atomic_read(&cm_node->listener->pend_accepts_cnt) < 0); } - BUG_ON(cm_node->send_entry); - spin_lock_irqsave(&cm_node->recv_list_lock, flags); - list_for_each_safe(list_core, list_node_temp, &cm_node->recv_list) { - recv_entry = container_of(list_core, struct nes_timer_entry, - list); - list_del(&recv_entry->list); - cm_id = cm_node->cm_id; - spin_unlock_irqrestore(&cm_node->recv_list_lock, flags); - nesqp = (struct nes_qp *)recv_entry->skb; - spin_lock_irqsave(&nesqp->lock, qplockflags); - if (nesqp->cm_id) { - nes_debug(NES_DBG_CM, "QP%u: cm_id = %p: HIT A " - "NES_TIMER_TYPE_CLOSE with something to do!\n", - nesqp->hwqp.qp_id, cm_id); - nesqp->hw_tcp_state = NES_AEQE_TCP_STATE_CLOSED; - nesqp->last_aeq = NES_AEQE_AEID_RESET_SENT; - nesqp->ibqp_state = IB_QPS_ERR; - spin_unlock_irqrestore(&nesqp->lock, qplockflags); - nes_cm_disconn(nesqp); - } else { - spin_unlock_irqrestore(&nesqp->lock, qplockflags); - nes_debug(NES_DBG_CM, "QP%u: cm_id = %p: HIT A " - "NES_TIMER_TYPE_CLOSE with nothing to do!\n", - nesqp->hwqp.qp_id, cm_id); - } - cm_id->rem_ref(cm_id); - - kfree(recv_entry); - spin_lock_irqsave(&cm_node->recv_list_lock, flags); - } - spin_unlock_irqrestore(&cm_node->recv_list_lock, flags); - + WARN_ON(cm_node->send_entry); + if (cm_node->recv_entry) + handle_recv_entry(cm_node, 0); if (cm_node->listener) { mini_cm_dec_refcnt_listen(cm_core, cm_node->listener, 0); } else { @@ -1262,8 +1326,7 @@ static void drop_packet(struct sk_buff *skb) dev_kfree_skb_any(skb); } -static void handle_fin_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, - struct tcphdr *tcph) +static void handle_fin_pkt(struct nes_cm_node *cm_node) { nes_debug(NES_DBG_CM, "Received FIN, cm_node = %p, state = %u. " "refcnt=%d\n", cm_node, cm_node->state, @@ -1275,23 +1338,30 @@ static void handle_fin_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, case NES_CM_STATE_SYN_SENT: case NES_CM_STATE_ESTABLISHED: case NES_CM_STATE_MPAREQ_SENT: + case NES_CM_STATE_MPAREJ_RCVD: cm_node->state = NES_CM_STATE_LAST_ACK; - send_fin(cm_node, skb); + send_fin(cm_node, NULL); break; case NES_CM_STATE_FIN_WAIT1: cm_node->state = NES_CM_STATE_CLOSING; - send_ack(cm_node, skb); + send_ack(cm_node, NULL); + /* Wait for ACK as this is simultanous close.. + * After we receive ACK, do not send anything.. + * Just rm the node.. Done.. */ break; case NES_CM_STATE_FIN_WAIT2: cm_node->state = NES_CM_STATE_TIME_WAIT; - send_ack(cm_node, skb); + send_ack(cm_node, NULL); + schedule_nes_timer(cm_node, NULL, NES_TIMER_TYPE_CLOSE, 1, 0); + break; + case NES_CM_STATE_TIME_WAIT: cm_node->state = NES_CM_STATE_CLOSED; + rem_ref_cm_node(cm_node->cm_core, cm_node); break; case NES_CM_STATE_TSA: default: nes_debug(NES_DBG_CM, "Error Rcvd FIN for node-%p state = %d\n", cm_node, cm_node->state); - drop_packet(skb); break; } } @@ -1337,23 +1407,35 @@ static void handle_rst_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, cleanup_retrans_entry(cm_node); drop_packet(skb); break; + case NES_CM_STATE_TIME_WAIT: + cleanup_retrans_entry(cm_node); + cm_node->state = NES_CM_STATE_CLOSED; + rem_ref_cm_node(cm_node->cm_core, cm_node); + drop_packet(skb); + break; + case NES_CM_STATE_FIN_WAIT1: + cleanup_retrans_entry(cm_node); + nes_debug(NES_DBG_CM, "Bad state %s[%u]\n", __func__, __LINE__); default: drop_packet(skb); break; } } -static void handle_rcv_mpa(struct nes_cm_node *cm_node, struct sk_buff *skb, - enum nes_cm_event_type type) + +static void handle_rcv_mpa(struct nes_cm_node *cm_node, struct sk_buff *skb) { - int ret; + int ret = 0; int datasize = skb->len; u8 *dataloc = skb->data; - ret = parse_mpa(cm_node, dataloc, datasize); - if (ret < 0) { + + enum nes_cm_event_type type = NES_CM_EVENT_UNKNOWN; + u32 res_type; + ret = parse_mpa(cm_node, dataloc, &res_type, datasize); + if (ret) { nes_debug(NES_DBG_CM, "didn't like MPA Request\n"); - if (type == NES_CM_EVENT_CONNECTED) { + if (cm_node->state == NES_CM_STATE_MPAREQ_SENT) { nes_debug(NES_DBG_CM, "%s[%u] create abort for " "cm_node=%p listener=%p state=%d\n", __func__, __LINE__, cm_node, cm_node->listener, @@ -1362,18 +1444,38 @@ static void handle_rcv_mpa(struct nes_cm_node *cm_node, struct sk_buff *skb, } else { passive_open_err(cm_node, skb, 1); } - } else { - cleanup_retrans_entry(cm_node); - dev_kfree_skb_any(skb); - if (type == NES_CM_EVENT_CONNECTED) + return; + } + + switch (cm_node->state) { + case NES_CM_STATE_ESTABLISHED: + if (res_type == NES_MPA_REQUEST_REJECT) { + /*BIG problem as we are receiving the MPA.. So should + * not be REJECT.. This is Passive Open.. We can + * only receive it Reject for Active Open...*/ + WARN_ON(1); + } + cm_node->state = NES_CM_STATE_MPAREQ_RCVD; + type = NES_CM_EVENT_MPA_REQ; + atomic_set(&cm_node->passive_state, + NES_PASSIVE_STATE_INDICATED); + break; + case NES_CM_STATE_MPAREQ_SENT: + if (res_type == NES_MPA_REQUEST_REJECT) { + type = NES_CM_EVENT_MPA_REJECT; + cm_node->state = NES_CM_STATE_MPAREJ_RCVD; + } else { + type = NES_CM_EVENT_CONNECTED; cm_node->state = NES_CM_STATE_TSA; - else - atomic_set(&cm_node->passive_state, - NES_PASSIVE_STATE_INDICATED); - create_event(cm_node, type); + } + break; + default: + WARN_ON(1); + break; } - return ; + dev_kfree_skb_any(skb); + create_event(cm_node, type); } static void indicate_pkt_err(struct nes_cm_node *cm_node, struct sk_buff *skb) @@ -1461,8 +1563,6 @@ static void handle_syn_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, break; case NES_CM_STATE_LISTENING: /* Passive OPEN */ - cm_node->accept_pend = 1; - atomic_inc(&cm_node->listener->pend_accepts_cnt); if (atomic_read(&cm_node->listener->pend_accepts_cnt) > cm_node->listener->backlog) { nes_debug(NES_DBG_CM, "drop syn due to backlog " @@ -1480,6 +1580,9 @@ static void handle_syn_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, } cm_node->tcp_cntxt.rcv_nxt = inc_sequence + 1; BUG_ON(cm_node->send_entry); + cm_node->accept_pend = 1; + atomic_inc(&cm_node->listener->pend_accepts_cnt); + cm_node->state = NES_CM_STATE_SYN_RCVD; send_syn(cm_node, 1, skb); break; @@ -1514,6 +1617,7 @@ static void handle_synack_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, inc_sequence = ntohl(tcph->seq); switch (cm_node->state) { case NES_CM_STATE_SYN_SENT: + cleanup_retrans_entry(cm_node); /* active open */ if (check_syn(cm_node, tcph, skb)) return; @@ -1563,10 +1667,7 @@ static void handle_ack_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, u32 rem_seq; int ret; int optionsize; - u32 temp_seq = cm_node->tcp_cntxt.loc_seq_num; - optionsize = (tcph->doff << 2) - sizeof(struct tcphdr); - cm_node->tcp_cntxt.loc_seq_num = ntohl(tcph->ack_seq); if (check_seq(cm_node, tcph, skb)) return; @@ -1576,7 +1677,7 @@ static void handle_ack_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, rem_seq = ntohl(tcph->seq); rem_seq_ack = ntohl(tcph->ack_seq); datasize = skb->len; - + cleanup_retrans_entry(cm_node); switch (cm_node->state) { case NES_CM_STATE_SYN_RCVD: /* Passive OPEN */ @@ -1584,7 +1685,6 @@ static void handle_ack_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, if (ret) break; cm_node->tcp_cntxt.rem_ack_num = ntohl(tcph->ack_seq); - cm_node->tcp_cntxt.loc_seq_num = temp_seq; if (cm_node->tcp_cntxt.rem_ack_num != cm_node->tcp_cntxt.loc_seq_num) { nes_debug(NES_DBG_CM, "rem_ack_num != loc_seq_num\n"); @@ -1593,31 +1693,30 @@ static void handle_ack_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, return; } cm_node->state = NES_CM_STATE_ESTABLISHED; + cleanup_retrans_entry(cm_node); if (datasize) { cm_node->tcp_cntxt.rcv_nxt = inc_sequence + datasize; - cm_node->state = NES_CM_STATE_MPAREQ_RCVD; - handle_rcv_mpa(cm_node, skb, NES_CM_EVENT_MPA_REQ); - } else { /* rcvd ACK only */ + handle_rcv_mpa(cm_node, skb); + } else { /* rcvd ACK only */ dev_kfree_skb_any(skb); cleanup_retrans_entry(cm_node); } break; case NES_CM_STATE_ESTABLISHED: /* Passive OPEN */ - /* We expect mpa frame to be received only */ + cleanup_retrans_entry(cm_node); if (datasize) { cm_node->tcp_cntxt.rcv_nxt = inc_sequence + datasize; - cm_node->state = NES_CM_STATE_MPAREQ_RCVD; - handle_rcv_mpa(cm_node, skb, - NES_CM_EVENT_MPA_REQ); + handle_rcv_mpa(cm_node, skb); } else drop_packet(skb); break; case NES_CM_STATE_MPAREQ_SENT: + cleanup_retrans_entry(cm_node); cm_node->tcp_cntxt.rem_ack_num = ntohl(tcph->ack_seq); if (datasize) { cm_node->tcp_cntxt.rcv_nxt = inc_sequence + datasize; - handle_rcv_mpa(cm_node, skb, NES_CM_EVENT_CONNECTED); + handle_rcv_mpa(cm_node, skb); } else { /* Could be just an ack pkt.. */ cleanup_retrans_entry(cm_node); dev_kfree_skb_any(skb); @@ -1628,13 +1727,24 @@ static void handle_ack_pkt(struct nes_cm_node *cm_node, struct sk_buff *skb, cleanup_retrans_entry(cm_node); send_reset(cm_node, skb); break; + case NES_CM_STATE_LAST_ACK: + cleanup_retrans_entry(cm_node); + cm_node->state = NES_CM_STATE_CLOSED; + cm_node->cm_id->rem_ref(cm_node->cm_id); + case NES_CM_STATE_CLOSING: + cleanup_retrans_entry(cm_node); + rem_ref_cm_node(cm_node->cm_core, cm_node); + drop_packet(skb); + break; case NES_CM_STATE_FIN_WAIT1: + cleanup_retrans_entry(cm_node); + drop_packet(skb); + cm_node->state = NES_CM_STATE_FIN_WAIT2; + break; case NES_CM_STATE_SYN_SENT: case NES_CM_STATE_FIN_WAIT2: case NES_CM_STATE_TSA: case NES_CM_STATE_MPAREQ_RCVD: - case NES_CM_STATE_LAST_ACK: - case NES_CM_STATE_CLOSING: case NES_CM_STATE_UNKNOWN: default: drop_packet(skb); @@ -1744,6 +1854,7 @@ static void process_packet(struct nes_cm_node *cm_node, struct sk_buff *skb, { enum nes_tcpip_pkt_type pkt_type = NES_PKT_TYPE_UNKNOWN; struct tcphdr *tcph = tcp_hdr(skb); + u32 fin_set = 0; skb_pull(skb, ip_hdr(skb)->ihl << 2); nes_debug(NES_DBG_CM, "process_packet: cm_node=%p state =%d syn=%d " @@ -1756,10 +1867,10 @@ static void process_packet(struct nes_cm_node *cm_node, struct sk_buff *skb, pkt_type = NES_PKT_TYPE_SYN; if (tcph->ack) pkt_type = NES_PKT_TYPE_SYNACK; - } else if (tcph->fin) - pkt_type = NES_PKT_TYPE_FIN; - else if (tcph->ack) + } else if (tcph->ack) pkt_type = NES_PKT_TYPE_ACK; + if (tcph->fin) + fin_set = 1; switch (pkt_type) { case NES_PKT_TYPE_SYN: @@ -1770,15 +1881,16 @@ static void process_packet(struct nes_cm_node *cm_node, struct sk_buff *skb, break; case NES_PKT_TYPE_ACK: handle_ack_pkt(cm_node, skb, tcph); + if (fin_set) + handle_fin_pkt(cm_node); break; case NES_PKT_TYPE_RST: handle_rst_pkt(cm_node, skb, tcph); break; - case NES_PKT_TYPE_FIN: - handle_fin_pkt(cm_node, skb, tcph); - break; default: drop_packet(skb); + if (fin_set) + handle_fin_pkt(cm_node); break; } } @@ -1921,7 +2033,7 @@ static struct nes_cm_node *mini_cm_connect(struct nes_cm_core *cm_core, loopbackremotenode->tcp_cntxt.rcv_wscale; loopbackremotenode->tcp_cntxt.snd_wscale = cm_node->tcp_cntxt.rcv_wscale; - + loopbackremotenode->state = NES_CM_STATE_MPAREQ_RCVD; create_event(loopbackremotenode, NES_CM_EVENT_MPA_REQ); } return cm_node; @@ -1976,7 +2088,11 @@ static int mini_cm_reject(struct nes_cm_core *cm_core, struct ietf_mpa_frame *mpa_frame, struct nes_cm_node *cm_node) { int ret = 0; + int err = 0; int passive_state; + struct nes_cm_event event; + struct iw_cm_id *cm_id = cm_node->cm_id; + struct nes_cm_node *loopback = cm_node->loopbackpartner; nes_debug(NES_DBG_CM, "%s cm_node=%p type=%d state=%d\n", __func__, cm_node, cm_node->tcp_cntxt.client, cm_node->state); @@ -1985,12 +2101,38 @@ static int mini_cm_reject(struct nes_cm_core *cm_core, return ret; cleanup_retrans_entry(cm_node); - passive_state = atomic_add_return(1, &cm_node->passive_state); - cm_node->state = NES_CM_STATE_CLOSED; - if (passive_state == NES_SEND_RESET_EVENT) + if (!loopback) { + passive_state = atomic_add_return(1, &cm_node->passive_state); + if (passive_state == NES_SEND_RESET_EVENT) { + cm_node->state = NES_CM_STATE_CLOSED; + rem_ref_cm_node(cm_core, cm_node); + } else { + ret = send_mpa_reject(cm_node); + if (ret) { + cm_node->state = NES_CM_STATE_CLOSED; + err = send_reset(cm_node, NULL); + if (err) + WARN_ON(1); + } else + cm_id->add_ref(cm_id); + } + } else { + cm_node->cm_id = NULL; + event.cm_node = loopback; + event.cm_info.rem_addr = loopback->rem_addr; + event.cm_info.loc_addr = loopback->loc_addr; + event.cm_info.rem_port = loopback->rem_port; + event.cm_info.loc_port = loopback->loc_port; + event.cm_info.cm_id = loopback->cm_id; + cm_event_mpa_reject(&event); rem_ref_cm_node(cm_core, cm_node); - else - ret = send_reset(cm_node, NULL); + loopback->state = NES_CM_STATE_CLOSING; + + cm_id = loopback->cm_id; + rem_ref_cm_node(cm_core, loopback); + cm_id->rem_ref(cm_id); + } + return ret; } @@ -2027,6 +2169,7 @@ static int mini_cm_close(struct nes_cm_core *cm_core, struct nes_cm_node *cm_nod case NES_CM_STATE_CLOSING: ret = -1; break; + case NES_CM_STATE_MPAREJ_RCVD: case NES_CM_STATE_LISTENING: case NES_CM_STATE_UNKNOWN: case NES_CM_STATE_INITED: @@ -2223,15 +2366,15 @@ static int mini_cm_set(struct nes_cm_core *cm_core, u32 type, u32 value) int ret = 0; switch (type) { - case NES_CM_SET_PKT_SIZE: - cm_core->mtu = value; - break; - case NES_CM_SET_FREE_PKT_Q_SIZE: - cm_core->free_tx_pkt_max = value; - break; - default: - /* unknown set option */ - ret = -EINVAL; + case NES_CM_SET_PKT_SIZE: + cm_core->mtu = value; + break; + case NES_CM_SET_FREE_PKT_Q_SIZE: + cm_core->free_tx_pkt_max = value; + break; + default: + /* unknown set option */ + ret = -EINVAL; } return ret; @@ -2621,9 +2764,7 @@ int nes_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param) NES_QPCONTEXT_ORDIRD_WRPDU); } else { nesqp->nesqp_context->ird_ord_sizes |= - cpu_to_le32((NES_QPCONTEXT_ORDIRD_LSMM_PRESENT | - NES_QPCONTEXT_ORDIRD_WRPDU | - NES_QPCONTEXT_ORDIRD_ALSMM)); + cpu_to_le32(NES_QPCONTEXT_ORDIRD_WRPDU); } nesqp->skip_lsmm = 1; @@ -2745,23 +2886,35 @@ int nes_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param) int nes_reject(struct iw_cm_id *cm_id, const void *pdata, u8 pdata_len) { struct nes_cm_node *cm_node; + struct nes_cm_node *loopback; + struct nes_cm_core *cm_core; atomic_inc(&cm_rejects); cm_node = (struct nes_cm_node *) cm_id->provider_data; + loopback = cm_node->loopbackpartner; cm_core = cm_node->cm_core; + cm_node->cm_id = cm_id; cm_node->mpa_frame_size = sizeof(struct ietf_mpa_frame) + pdata_len; + if (cm_node->mpa_frame_size > MAX_CM_BUFFER) + return -EINVAL; + strcpy(&cm_node->mpa_frame.key[0], IEFT_MPA_KEY_REP); - memcpy(&cm_node->mpa_frame.priv_data, pdata, pdata_len); + if (loopback) { + memcpy(&loopback->mpa_frame.priv_data, pdata, pdata_len); + loopback->mpa_frame.priv_data_len = pdata_len; + loopback->mpa_frame_size = sizeof(struct ietf_mpa_frame) + + pdata_len; + } else { + memcpy(&cm_node->mpa_frame.priv_data, pdata, pdata_len); + cm_node->mpa_frame.priv_data_len = cpu_to_be16(pdata_len); + } - cm_node->mpa_frame.priv_data_len = cpu_to_be16(pdata_len); cm_node->mpa_frame.rev = mpa_version; cm_node->mpa_frame.flags = IETF_MPA_FLAGS_CRC | IETF_MPA_FLAGS_REJECT; - cm_core->api->reject(cm_core, &cm_node->mpa_frame, cm_node); - - return 0; + return cm_core->api->reject(cm_core, &cm_node->mpa_frame, cm_node); } @@ -3267,16 +3420,59 @@ static void cm_event_mpa_req(struct nes_cm_event *event) cm_event.local_addr.sin_port = htons(event->cm_info.loc_port); cm_event.local_addr.sin_addr.s_addr = htonl(event->cm_info.loc_addr); + cm_event.remote_addr.sin_family = AF_INET; + cm_event.remote_addr.sin_port = htons(event->cm_info.rem_port); + cm_event.remote_addr.sin_addr.s_addr = htonl(event->cm_info.rem_addr); + cm_event.private_data = cm_node->mpa_frame_buf; + cm_event.private_data_len = (u8) cm_node->mpa_frame_size; + + ret = cm_id->event_handler(cm_id, &cm_event); + if (ret) + printk(KERN_ERR "%s[%u] OFA CM event_handler returned, ret=%d\n", + __func__, __LINE__, ret); + return; +} + + +static void cm_event_mpa_reject(struct nes_cm_event *event) +{ + struct iw_cm_id *cm_id; + struct iw_cm_event cm_event; + struct nes_cm_node *cm_node; + int ret; + + cm_node = event->cm_node; + if (!cm_node) + return; + cm_id = cm_node->cm_id; + + atomic_inc(&cm_connect_reqs); + nes_debug(NES_DBG_CM, "cm_node = %p - cm_id = %p, jiffies = %lu\n", + cm_node, cm_id, jiffies); + + cm_event.event = IW_CM_EVENT_CONNECT_REPLY; + cm_event.status = -ECONNREFUSED; + cm_event.provider_data = cm_id->provider_data; + + cm_event.local_addr.sin_family = AF_INET; + cm_event.local_addr.sin_port = htons(event->cm_info.loc_port); + cm_event.local_addr.sin_addr.s_addr = htonl(event->cm_info.loc_addr); + cm_event.remote_addr.sin_family = AF_INET; cm_event.remote_addr.sin_port = htons(event->cm_info.rem_port); cm_event.remote_addr.sin_addr.s_addr = htonl(event->cm_info.rem_addr); - cm_event.private_data = cm_node->mpa_frame_buf; - cm_event.private_data_len = (u8) cm_node->mpa_frame_size; + cm_event.private_data = cm_node->mpa_frame_buf; + cm_event.private_data_len = (u8) cm_node->mpa_frame_size; + + nes_debug(NES_DBG_CM, "call CM_EVENT_MPA_REJECTED, local_addr=%08x, " + "remove_addr=%08x\n", + cm_event.local_addr.sin_addr.s_addr, + cm_event.remote_addr.sin_addr.s_addr); ret = cm_id->event_handler(cm_id, &cm_event); if (ret) - printk("%s[%u] OFA CM event_handler returned, ret=%d\n", + printk(KERN_ERR "%s[%u] OFA CM event_handler returned, ret=%d\n", __func__, __LINE__, ret); return; @@ -3341,6 +3537,14 @@ static void nes_cm_event_handler(struct work_struct *work) cm_event_connected(event); nes_debug(NES_DBG_CM, "CM Event: CONNECTED\n"); break; + case NES_CM_EVENT_MPA_REJECT: + if ((!event->cm_node->cm_id) || + (event->cm_node->state == NES_CM_STATE_TSA)) + break; + cm_event_mpa_reject(event); + nes_debug(NES_DBG_CM, "CM Event: REJECT\n"); + break; + case NES_CM_EVENT_ABORTED: if ((!event->cm_node->cm_id) || (event->cm_node->state == NES_CM_STATE_TSA)) diff --git a/drivers/infiniband/hw/nes/nes_cm.h b/drivers/infiniband/hw/nes/nes_cm.h index 4ab2bebf12a..d5f778202eb 100644 --- a/drivers/infiniband/hw/nes/nes_cm.h +++ b/drivers/infiniband/hw/nes/nes_cm.h @@ -39,6 +39,9 @@ #define NES_MANAGE_APBVT_DEL 0 #define NES_MANAGE_APBVT_ADD 1 +#define NES_MPA_REQUEST_ACCEPT 1 +#define NES_MPA_REQUEST_REJECT 2 + /* IETF MPA -- defines, enums, structs */ #define IEFT_MPA_KEY_REQ "MPA ID Req Frame" #define IEFT_MPA_KEY_REP "MPA ID Rep Frame" @@ -186,6 +189,7 @@ enum nes_cm_node_state { NES_CM_STATE_ACCEPTING, NES_CM_STATE_MPAREQ_SENT, NES_CM_STATE_MPAREQ_RCVD, + NES_CM_STATE_MPAREJ_RCVD, NES_CM_STATE_TSA, NES_CM_STATE_FIN_WAIT1, NES_CM_STATE_FIN_WAIT2, @@ -278,13 +282,12 @@ struct nes_cm_node { struct nes_timer_entry *send_entry; spinlock_t retrans_list_lock; - struct list_head recv_list; - spinlock_t recv_list_lock; + struct nes_timer_entry *recv_entry; int send_write0; union { struct ietf_mpa_frame mpa_frame; - u8 mpa_frame_buf[NES_CM_DEFAULT_MTU]; + u8 mpa_frame_buf[MAX_CM_BUFFER]; }; u16 mpa_frame_size; struct iw_cm_id *cm_id; @@ -326,6 +329,7 @@ enum nes_cm_event_type { NES_CM_EVENT_MPA_REQ, NES_CM_EVENT_MPA_CONNECT, NES_CM_EVENT_MPA_ACCEPT, + NES_CM_EVENT_MPA_REJECT, NES_CM_EVENT_MPA_ESTABLISHED, NES_CM_EVENT_CONNECTED, NES_CM_EVENT_CLOSED, -- cgit v1.2.3 From a6a47771b113be8e694aedd80f66ea94d05bd8df Mon Sep 17 00:00:00 2001 From: Yevgeny Petrilin Date: Wed, 18 Mar 2009 19:49:54 -0700 Subject: IB/mlx4: Unregister IB device prior to CLOSE PORT command According to the ConnectX programmer's reference manual, all operations should be stopped, all QPs should be torn down and all WQEs flushed before the CLOSE_PORT command is invoked. In some cases reversing the order of operations (as implemented now) could cause a loss of completions. Signed-off-by: Yevgeny Petrilin Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mlx4/main.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 61588bd273b..2ccb9d31771 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -699,11 +699,12 @@ static void mlx4_ib_remove(struct mlx4_dev *dev, void *ibdev_ptr) struct mlx4_ib_dev *ibdev = ibdev_ptr; int p; + mlx4_ib_mad_cleanup(ibdev); + ib_unregister_device(&ibdev->ib_dev); + for (p = 1; p <= ibdev->num_ports; ++p) mlx4_CLOSE_PORT(dev, p); - mlx4_ib_mad_cleanup(ibdev); - ib_unregister_device(&ibdev->ib_dev); iounmap(ibdev->uar_map); mlx4_uar_free(dev, &ibdev->priv_uar); mlx4_pd_free(dev, ibdev->priv_pdn); -- cgit v1.2.3