From: Faisal Latif <faisal.latif-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
To: Roland Dreier <rdreier-FYB4Gu1CFyUAvxtiuMwx3w@public.gmane.org>,
-c@vger.kernel.org,
linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: [PATCH] RDMA/nes: multiple disconnect cause crash during AE handling
Date: Fri, 12 Feb 2010 13:58:05 -0600 [thread overview]
Message-ID: <20100212195805.GA73576@flatif-MOBL> (raw)
There is double disconnect during AE processing causing crash. While fixing the
crash, also simplify the AE handling code.
Signed-off-by: Faisal Latif <faisal.latif-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/infiniband/hw/nes/nes_hw.c | 88 +++++++++++------------------------
1 files changed, 28 insertions(+), 60 deletions(-)
diff --git a/drivers/infiniband/hw/nes/nes_hw.c b/drivers/infiniband/hw/nes/nes_hw.c
index b1c2cbb..310cc7c 100644
--- a/drivers/infiniband/hw/nes/nes_hw.c
+++ b/drivers/infiniband/hw/nes/nes_hw.c
@@ -3352,8 +3352,6 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
u16 async_event_id;
u8 tcp_state;
u8 iwarp_state;
- int must_disconn = 1;
- int must_terminate = 0;
struct ib_event ibevent;
nes_debug(NES_DBG_AEQ, "\n");
@@ -3367,6 +3365,8 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
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;
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;
@@ -3378,8 +3378,6 @@ static void nes_process_iwarp_aeqe(struct nes_device *nesdev,
switch (async_event_id) {
case NES_AEQE_AEID_LLP_FIN_RECEIVED:
- nesqp = (struct nes_qp *)(unsigned long)context;
-
if (nesqp->term_flags)
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);
}
- if ((tcp_state != NES_AEQE_TCP_STATE_CLOSE_WAIT) ||
- (nesqp->ibqp_state != IB_QPS_RTS)) {
- /* FIN Received but tcp state or IB state moved on,
- should expect a close complete */
- return;
- }
-
+ break;
case NES_AEQE_AEID_LLP_CLOSE_COMPLETE:
- nesqp = (struct nes_qp *)(unsigned long)context;
if (nesqp->term_flags) {
nes_terminate_done(nesqp, 0);
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:
- nesqp = (struct nes_qp *)(unsigned long)context;
- if (async_event_id == NES_AEQE_AEID_RESET_SENT) {
- tcp_state = NES_AEQE_TCP_STATE_CLOSED;
- }
+ tcp_state = NES_AEQE_TCP_STATE_CLOSED;
spin_lock_irqsave(&nesqp->lock, flags);
nesqp->hw_iwarp_state = iwarp_state;
nesqp->hw_tcp_state = tcp_state;
nesqp->last_aeq = async_event_id;
-
- 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;
- }
- }
+ nesqp->hte_added = 0;
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)
- nes_terminate_connection(nesdev, nesqp, aeqe, IB_EVENT_QP_FATAL);
- else if (must_disconn) {
- if (next_iwarp_state) {
- nes_debug(NES_DBG_AEQ, "issuing hw modifyqp for QP%u. next state = 0x%08X\n",
- nesqp->hwqp.qp_id, next_iwarp_state);
- nes_hw_modify_qp(nesdev, nesqp, next_iwarp_state, 0, 0);
- }
- nes_cm_disconn(nesqp);
- }
+ case NES_AEQE_AEID_LLP_CONNECTION_RESET:
+ if (atomic_read(&nesqp->close_timer_started))
+ return;
+ spin_lock_irqsave(&nesqp->lock, flags);
+ nesqp->hw_iwarp_state = iwarp_state;
+ nesqp->hw_tcp_state = tcp_state;
+ nesqp->last_aeq = async_event_id;
+ spin_unlock_irqrestore(&nesqp->lock, flags);
+ nes_cm_disconn(nesqp);
break;
case NES_AEQE_AEID_TERMINATE_SENT:
- nesqp = (struct nes_qp *)(unsigned long)context;
nes_terminate_send_fin(nesdev, nesqp, aeqe);
break;
case NES_AEQE_AEID_LLP_TERMINATE_RECEIVED:
- nesqp = (struct nes_qp *)(unsigned long)context;
nes_terminate_received(nesdev, nesqp, aeqe);
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_AMP_BOUNDS_VIOLATION:
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);
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_DDP_UBE_INVALID_MO:
case NES_AEQE_AEID_DDP_UBE_INVALID_QN:
- nesqp = (struct nes_qp *)(unsigned long)context;
if (iwarp_opcode(nesqp, aeq_info) > IWARP_OPCODE_TERM) {
aeq_info &= 0xffff0000;
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_ROE_INVALID_RDMA_READ_REQUEST:
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);
break;
--
1.5.3.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next reply other threads:[~2010-02-12 19:58 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-02-12 19:58 Faisal Latif [this message]
2010-02-19 19:41 ` [PATCH] RDMA/nes: multiple disconnect cause crash during AE handling Roland Dreier
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20100212195805.GA73576@flatif-MOBL \
--to=faisal.latif-ral2jqcrhueavxtiumwx3w@public.gmane.org \
--cc=-c@vger.kernel.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=rdreier-FYB4Gu1CFyUAvxtiuMwx3w@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.