redonkable/alistair23-linux
redonkable/alistair23-linux
1
0
Fork 0
Code Releases Activity

RDMA/nes: Multiple disconnects cause crash during AE handling

Browse source 
There is a double disconnect during AE processing, causing crashes.
While fixing the crash, also simplify the AE handling code.

Signed-off-by: Faisal Latif <faisal.latif@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This commit is contained in:
Faisal Latif 2010-02-12 19:58:05 +00:00 committed by Roland Dreier
parent 43093b9412
commit 30b172ff8e
1 changed files with 28 additions and 60 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

88
drivers/infiniband/hw/nes/nes_hw.c
View file

@ -3352,8 +3352,6 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
u16 async_event_id; u16 async_event_id;
u8 tcp_state; u8 tcp_state;
u8 iwarp_state; u8 iwarp_state;
int must_disconn = 1;
int must_terminate = 0;
struct ib_event ibevent; struct ib_event ibevent;
nes_debug(NES_DBG_AEQ, "\n"); nes_debug(NES_DBG_AEQ, "\n");
@ -3367,6 +3365,8 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
BUG_ON(!context); BUG_ON(!context);
} }
/* context is nesqp unless async_event_id == CQ ERROR */
nesqp = (struct nes_qp *)(unsigned long)context;
async_event_id = (u16)aeq_info; async_event_id = (u16)aeq_info;
tcp_state = (aeq_info & NES_AEQE_TCP_STATE_MASK) >> NES_AEQE_TCP_STATE_SHIFT; tcp_state = (aeq_info & NES_AEQE_TCP_STATE_MASK) >> NES_AEQE_TCP_STATE_SHIFT;
iwarp_state = (aeq_info & NES_AEQE_IWARP_STATE_MASK) >> NES_AEQE_IWARP_STATE_SHIFT; iwarp_state = (aeq_info & NES_AEQE_IWARP_STATE_MASK) >> NES_AEQE_IWARP_STATE_SHIFT;
@ -3378,8 +3378,6 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
switch (async_event_id) { switch (async_event_id) {
case NES_AEQE_AEID_LLP_FIN_RECEIVED: case NES_AEQE_AEID_LLP_FIN_RECEIVED:
nesqp = (struct nes_qp *)(unsigned long)context;
if (nesqp->term_flags) if (nesqp->term_flags)
return; /* Ignore it, wait for close complete */ return; /* Ignore it, wait for close complete */
@ -3394,79 +3392,48 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
async_event_id, nesqp->last_aeq, tcp_state); async_event_id, nesqp->last_aeq, tcp_state);
} }
if ((tcp_state != NES_AEQE_TCP_STATE_CLOSE_WAIT) || break;
(nesqp->ibqp_state != IB_QPS_RTS)) {
/* FIN Received but tcp state or IB state moved on,
should expect a close complete */
return;
}
case NES_AEQE_AEID_LLP_CLOSE_COMPLETE: case NES_AEQE_AEID_LLP_CLOSE_COMPLETE:
nesqp = (struct nes_qp *)(unsigned long)context;
if (nesqp->term_flags) { if (nesqp->term_flags) {
nes_terminate_done(nesqp, 0); nes_terminate_done(nesqp, 0);
return; return;
} }
spin_lock_irqsave(&nesqp->lock, flags);
nesqp->hw_iwarp_state = NES_AEQE_IWARP_STATE_CLOSING;
spin_unlock_irqrestore(&nesqp->lock, flags);
nes_hw_modify_qp(nesdev, nesqp, NES_CQP_QP_IWARP_STATE_CLOSING, 0, 0);
nes_cm_disconn(nesqp);
break;
case NES_AEQE_AEID_LLP_CONNECTION_RESET:
case NES_AEQE_AEID_RESET_SENT: case NES_AEQE_AEID_RESET_SENT:
nesqp = (struct nes_qp *)(unsigned long)context; tcp_state = NES_AEQE_TCP_STATE_CLOSED;
if (async_event_id == NES_AEQE_AEID_RESET_SENT) {
tcp_state = NES_AEQE_TCP_STATE_CLOSED;
}
spin_lock_irqsave(&nesqp->lock, flags); spin_lock_irqsave(&nesqp->lock, flags);
nesqp->hw_iwarp_state = iwarp_state; nesqp->hw_iwarp_state = iwarp_state;
nesqp->hw_tcp_state = tcp_state; nesqp->hw_tcp_state = tcp_state;
nesqp->last_aeq = async_event_id; nesqp->last_aeq = async_event_id;
nesqp->hte_added = 0;
if ((tcp_state == NES_AEQE_TCP_STATE_CLOSED) ||
(tcp_state == NES_AEQE_TCP_STATE_TIME_WAIT)) {
nesqp->hte_added = 0;
next_iwarp_state = NES_CQP_QP_IWARP_STATE_ERROR | NES_CQP_QP_DEL_HTE;
}
if ((nesqp->ibqp_state == IB_QPS_RTS) &&
((tcp_state == NES_AEQE_TCP_STATE_CLOSE_WAIT) ||
(async_event_id == NES_AEQE_AEID_LLP_CONNECTION_RESET))) {
switch (nesqp->hw_iwarp_state) {
case NES_AEQE_IWARP_STATE_RTS:
next_iwarp_state = NES_CQP_QP_IWARP_STATE_CLOSING;
nesqp->hw_iwarp_state = NES_AEQE_IWARP_STATE_CLOSING;
break;
case NES_AEQE_IWARP_STATE_TERMINATE:
must_disconn = 0; /* terminate path takes care of disconn */
if (nesqp->term_flags == 0)
must_terminate = 1;
break;
}
} else {
if (async_event_id == NES_AEQE_AEID_LLP_FIN_RECEIVED) {
/* FIN Received but ib state not RTS,
close complete will be on its way */
must_disconn = 0;
}
}
spin_unlock_irqrestore(&nesqp->lock, flags); spin_unlock_irqrestore(&nesqp->lock, flags);
next_iwarp_state = NES_CQP_QP_IWARP_STATE_ERROR | NES_CQP_QP_DEL_HTE;
nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0, 0);
nes_cm_disconn(nesqp);
break;
if (must_terminate) case NES_AEQE_AEID_LLP_CONNECTION_RESET:
nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_FATAL); if (atomic_read(&nesqp->close_timer_started))
else if (must_disconn) { return;
if (next_iwarp_state) { spin_lock_irqsave(&nesqp->lock, flags);
nes_debug(NES_DBG_AEQ, "issuing hw modifyqp for QP%u. next state = 0x%08X\n", nesqp->hw_iwarp_state = iwarp_state;
nesqp->hwqp.qp_id, next_iwarp_state); nesqp->hw_tcp_state = tcp_state;
nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0, 0); nesqp->last_aeq = async_event_id;
} spin_unlock_irqrestore(&nesqp->lock, flags);
nes_cm_disconn(nesqp); nes_cm_disconn(nesqp);
}
break; break;
case NES_AEQE_AEID_TERMINATE_SENT: case NES_AEQE_AEID_TERMINATE_SENT:
nesqp = (struct nes_qp *)(unsigned long)context;
nes_terminate_send_fin(nesdev, nesqp, aeqe); nes_terminate_send_fin(nesdev, nesqp, aeqe);
break; break;
case NES_AEQE_AEID_LLP_TERMINATE_RECEIVED: case NES_AEQE_AEID_LLP_TERMINATE_RECEIVED:
nesqp = (struct nes_qp *)(unsigned long)context;
nes_terminate_received(nesdev, nesqp, aeqe); nes_terminate_received(nesdev, nesqp, aeqe);
break; break;
@ -3480,7 +3447,8 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
case NES_AEQE_AEID_DDP_UBE_DDP_MESSAGE_TOO_LONG_FOR_AVAILABLE_BUFFER: case NES_AEQE_AEID_DDP_UBE_DDP_MESSAGE_TOO_LONG_FOR_AVAILABLE_BUFFER:
case NES_AEQE_AEID_AMP_BOUNDS_VIOLATION: case NES_AEQE_AEID_AMP_BOUNDS_VIOLATION:
case NES_AEQE_AEID_AMP_TO_WRAP: case NES_AEQE_AEID_AMP_TO_WRAP:
nesqp = (struct nes_qp *)(unsigned long)context; printk(KERN_ERR PFX "QP[%u] async_event_id=0x%04X IB_EVENT_QP_ACCESS_ERR\n",
nesqp->hwqp.qp_id, async_event_id);
nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_ACCESS_ERR); nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_ACCESS_ERR);
break; break;
@ -3488,7 +3456,6 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
case NES_AEQE_AEID_LLP_SEGMENT_TOO_SMALL: case NES_AEQE_AEID_LLP_SEGMENT_TOO_SMALL:
case NES_AEQE_AEID_DDP_UBE_INVALID_MO: case NES_AEQE_AEID_DDP_UBE_INVALID_MO:
case NES_AEQE_AEID_DDP_UBE_INVALID_QN: case NES_AEQE_AEID_DDP_UBE_INVALID_QN:
nesqp = (struct nes_qp *)(unsigned long)context;
if (iwarp_opcode(nesqp, aeq_info) > IWARP_OPCODE_TERM) { if (iwarp_opcode(nesqp, aeq_info) > IWARP_OPCODE_TERM) {
aeq_info &= 0xffff0000; aeq_info &= 0xffff0000;
aeq_info |= NES_AEQE_AEID_RDMAP_ROE_UNEXPECTED_OPCODE; aeq_info |= NES_AEQE_AEID_RDMAP_ROE_UNEXPECTED_OPCODE;
@ -3530,7 +3497,8 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
case NES_AEQE_AEID_STAG_ZERO_INVALID: case NES_AEQE_AEID_STAG_ZERO_INVALID:
case NES_AEQE_AEID_ROE_INVALID_RDMA_READ_REQUEST: case NES_AEQE_AEID_ROE_INVALID_RDMA_READ_REQUEST:
case NES_AEQE_AEID_ROE_INVALID_RDMA_WRITE_OR_READ_RESP: case NES_AEQE_AEID_ROE_INVALID_RDMA_WRITE_OR_READ_RESP:
nesqp = (struct nes_qp *)(unsigned long)context; printk(KERN_ERR PFX "QP[%u] async_event_id=0x%04X IB_EVENT_QP_FATAL\n",
nesqp->hwqp.qp_id, async_event_id);
nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_FATAL); nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_FATAL);
break; break;