qed: Fix race with multiple VFs

A PF syncronizes all IOV activity relating to its VFs
by using a single workqueue handling the work.
The workqueue would reach a bitmask of pending VF events
and act upon each in turn.

Problem is that the indication of a VF message [which sets
the 'vf event' bit for that VF] arrives and is set in
the slowpath attention context, which isn't syncronized with
the processing of the events.
When multiple VFs are present, it's possible that PF would
lose the indication of one of the VF's pending evens, leading
that VF to later timeout.

Instead of adding locks/barriers, simply move from a bitmask
into a per-VF indication inside that VF entry in the PF database.

Signed-off-by: Yuval Mintz <Yuval.Mintz@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 26 insertions, 17 deletions

diff --git a/drivers/net/ethernet/qlogic/qed/qed_sriov.c b/drivers/net/ethernet/qlogic/qed/qed_sriov.c
index 29ed785f1dc2..253c2bbe1e4e 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sriov.c
+++ b/drivers/net/ethernet/qlogic/qed/qed_sriov.c
@@ -3014,8 +3014,7 @@ cleanup:
 		ack_vfs[vfid / 32] |= BIT((vfid % 32));
 		p_hwfn->pf_iov_info->pending_flr[rel_vf_id / 64] &=
 		    ~(1ULL << (rel_vf_id % 64));
-		p_hwfn->pf_iov_info->pending_events[rel_vf_id / 64] &=
-		    ~(1ULL << (rel_vf_id % 64));
+		p_vf->vf_mbx.b_pending_msg = false;
 	}
 
 	return rc;
@@ -3128,11 +3127,20 @@ static void qed_iov_process_mbx_req(struct qed_hwfn *p_hwfn,
 	mbx = &p_vf->vf_mbx;
 
 	/* qed_iov_process_mbx_request */
-	DP_VERBOSE(p_hwfn, QED_MSG_IOV,
-		   "VF[%02x]: Processing mailbox message\n", p_vf->abs_vf_id);
+	if (!mbx->b_pending_msg) {
+		DP_NOTICE(p_hwfn,
+			  "VF[%02x]: Trying to process mailbox message when none is pending\n",
+			  p_vf->abs_vf_id);
+		return;
+	}
+	mbx->b_pending_msg = false;
 
 	mbx->first_tlv = mbx->req_virt->first_tlv;
 
+	DP_VERBOSE(p_hwfn, QED_MSG_IOV,
+		   "VF[%02x]: Processing mailbox message [type %04x]\n",
+		   p_vf->abs_vf_id, mbx->first_tlv.tl.type);
+
 	/* check if tlv type is known */
 	if (qed_iov_tlv_supported(mbx->first_tlv.tl.type) &&
 	    !p_vf->b_malicious) {
@@ -3219,20 +3227,19 @@ static void qed_iov_process_mbx_req(struct qed_hwfn *p_hwfn,
 	}
 }
 
-static void qed_iov_pf_add_pending_events(struct qed_hwfn *p_hwfn, u8 vfid)
+void qed_iov_pf_get_pending_events(struct qed_hwfn *p_hwfn, u64 *events)
 {
-	u64 add_bit = 1ULL << (vfid % 64);
+	int i;
 
-	p_hwfn->pf_iov_info->pending_events[vfid / 64] |= add_bit;
-}
+	memset(events, 0, sizeof(u64) * QED_VF_ARRAY_LENGTH);
 
-static void qed_iov_pf_get_and_clear_pending_events(struct qed_hwfn *p_hwfn,
-						    u64 *events)
-{
-	u64 *p_pending_events = p_hwfn->pf_iov_info->pending_events;
+	qed_for_each_vf(p_hwfn, i) {
+		struct qed_vf_info *p_vf;
 
-	memcpy(events, p_pending_events, sizeof(u64) * QED_VF_ARRAY_LENGTH);
-	memset(p_pending_events, 0, sizeof(u64) * QED_VF_ARRAY_LENGTH);
+		p_vf = &p_hwfn->pf_iov_info->vfs_array[i];
+		if (p_vf->vf_mbx.b_pending_msg)
+			events[i / 64] |= 1ULL << (i % 64);
+	}
 }
 
 static struct qed_vf_info *qed_sriov_get_vf_from_absid(struct qed_hwfn *p_hwfn,
@@ -3266,7 +3273,7 @@ static int qed_sriov_vfpf_msg(struct qed_hwfn *p_hwfn,
 	p_vf->vf_mbx.pending_req = (((u64)vf_msg->hi) << 32) | vf_msg->lo;
 
 	/* Mark the event and schedule the workqueue */
-	qed_iov_pf_add_pending_events(p_hwfn, p_vf->relative_vf_id);
+	p_vf->vf_mbx.b_pending_msg = true;
 	qed_schedule_iov(p_hwfn, QED_IOV_WQ_MSG_FLAG);
 
 	return 0;
@@ -4030,7 +4037,7 @@ static void qed_handle_vf_msg(struct qed_hwfn *hwfn)
 		return;
 	}
 
-	qed_iov_pf_get_and_clear_pending_events(hwfn, events);
+	qed_iov_pf_get_pending_events(hwfn, events);
 
 	DP_VERBOSE(hwfn, QED_MSG_IOV,
 		   "Event mask of VF events: 0x%llx 0x%llx 0x%llx\n",
diff --git a/drivers/net/ethernet/qlogic/qed/qed_sriov.h b/drivers/net/ethernet/qlogic/qed/qed_sriov.h
index fc08cc2da6a7..a89605821522 100644
--- a/drivers/net/ethernet/qlogic/qed/qed_sriov.h
+++ b/drivers/net/ethernet/qlogic/qed/qed_sriov.h
@@ -140,6 +140,9 @@ struct qed_iov_vf_mbx {
 	/* Address in VF where a pending message is located */
 	dma_addr_t pending_req;
 
+	/* Message from VF awaits handling */
+	bool b_pending_msg;
+
 	u8 *offset;
 
 	/* saved VF request header */
@@ -232,7 +235,6 @@ struct qed_vf_info {
  */
 struct qed_pf_iov {
 	struct qed_vf_info vfs_array[MAX_NUM_VFS];
-	u64 pending_events[QED_VF_ARRAY_LENGTH];
 	u64 pending_flr[QED_VF_ARRAY_LENGTH];
 
 	/* Allocate message address continuosuly and split to each VF */