*** irdma 6.8 rc fixes ***

Small set of fixes for the irdma driver.

Mike Marciniszyn (1):
RDMA/irdma: Fix KASAN issue with tasklet

Mustafa Ismail (2):
RDMA/irdma: Set the CQ read threshold for GEN 1
RDMA/irdma: Add AE for too many RNRS

Shiraz Saleem (1):
RDMA/irdma: Validate max_send_wr and max_recv_wr

 drivers/infiniband/hw/irdma/defs.h  | 1 +
 drivers/infiniband/hw/irdma/hw.c    | 8 ++++++++
 drivers/infiniband/hw/irdma/verbs.c | 9 +++++----
 3 files changed, 14 insertions(+), 4 deletions(-)

2.42.0

From: Mike Marciniszyn <[email protected]>

KASAN testing revealed the following issue assocated with freeing an IRQ.

[50006.466686] Call Trace:
[50006.466691]  <IRQ>
[50006.489538]  dump_stack+0x5c/0x80
[50006.493475]  print_address_description.constprop.6+0x1a/0x150
[50006.499872]  ? irdma_sc_process_ceq+0x483/0x790 [irdma]
[50006.505742]  ? irdma_sc_process_ceq+0x483/0x790 [irdma]
[50006.511644]  kasan_report.cold.11+0x7f/0x118
[50006.516572]  ? irdma_sc_process_ceq+0x483/0x790 [irdma]
[50006.522473]  irdma_sc_process_ceq+0x483/0x790 [irdma]
[50006.528232]  irdma_process_ceq+0xb2/0x400 [irdma]
[50006.533601]  ? irdma_hw_flush_wqes_callback+0x370/0x370 [irdma]
[50006.540298]  irdma_ceq_dpc+0x44/0x100 [irdma]
[50006.545306]  tasklet_action_common.isra.14+0x148/0x2c0
[50006.551096]  __do_softirq+0x1d0/0xaf8
[50006.555396]  irq_exit_rcu+0x219/0x260
[50006.559670]  irq_exit+0xa/0x20
[50006.563320]  smp_apic_timer_interrupt+0x1bf/0x690
[50006.568645]  apic_timer_interrupt+0xf/0x20
[50006.573341]  </IRQ>

The issue is that a tasklet could be pending on another core racing the delete of the irq.

Fix by insuring any scheduled tasklet is killed after deleting the irq.

---
 drivers/infiniband/hw/irdma/hw.c | 7 +++++++
 1 file changed, 7 insertions(+)

index bd4b2b896444..2f8d18d8be3b 100644
@@ -570,6 +570,13 @@ static void irdma_destroy_irq(struct irdma_pci_f *rf,
 	dev->irq_ops->irdma_dis_irq(dev, msix_vec->idx);
 	irq_update_affinity_hint(msix_vec->irq, NULL);
 	free_irq(msix_vec->irq, dev_id);
+	if (rf == dev_id) {
+		tasklet_kill(&rf->dpc_tasklet);
+	} else {
+		struct irdma_ceq *iwceq = (struct irdma_ceq *)dev_id;
+
+		tasklet_kill(&iwceq->dpc_tasklet);
+	}
 }
 
 /**

2.42.0

From: Shiraz Saleem <[email protected]>

Validate that max_send_wr and max_recv_wr is within the supported range.

---
 drivers/infiniband/hw/irdma/verbs.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

index b5eb8d421988..cb828e3da478 100644
@@ -839,7 +839,9 @@ static int irdma_validate_qp_attrs(struct ib_qp_init_attr *init_attr,
 
 	if (init_attr->cap.max_inline_data > uk_attrs->max_hw_inline ||
 	    init_attr->cap.max_send_sge > uk_attrs->max_hw_wq_frags ||
-	    init_attr->cap.max_recv_sge > uk_attrs->max_hw_wq_frags)
+	    init_attr->cap.max_recv_sge > uk_attrs->max_hw_wq_frags ||
+	    init_attr->cap.max_send_wr > uk_attrs->max_hw_wq_quanta ||
+	    init_attr->cap.max_recv_wr > uk_attrs->max_hw_rq_quanta)
 		return -EINVAL;
 
 	if (rdma_protocol_roce(&iwdev->ibdev, 1)) {

2.42.0

From: Mustafa Ismail <[email protected]>

The CQ shadow read threshold is currently not set for GEN 2. This could cause an invalid CQ overflow condition, so remove the GEN check that exclused GEN 1.

---
 drivers/infiniband/hw/irdma/verbs.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

index cb828e3da478..0b046c061742 100644
@@ -2186,9 +2186,8 @@ static int irdma_create_cq(struct ib_cq *ibcq,
 		info.cq_base_pa = iwcq->kmem.pa;
 	}
 
-	if (dev->hw_attrs.uk_attrs.hw_rev >= IRDMA_GEN_2)
-		info.shadow_read_threshold = min(info.cq_uk_init_info.cq_size / 2,
-						 (u32)IRDMA_MAX_CQ_READ_THRESH);
+	info.shadow_read_threshold = min(info.cq_uk_init_info.cq_size / 2,
+					 (u32)IRDMA_MAX_CQ_READ_THRESH);
 
 	if (irdma_sc_cq_init(cq, &info)) {
 		ibdev_dbg(&iwdev->ibdev, "VERBS: init cq fail\n");

2.42.0

From: Mustafa Ismail <[email protected]>

Add IRDMA_AE_LLP_TOO_MANY_RNRS to the list of AE's processed as an abnormal asyncronous event.

---
 drivers/infiniband/hw/irdma/defs.h | 1 +
 drivers/infiniband/hw/irdma/hw.c   | 1 +
 2 files changed, 2 insertions(+)

index 8fb752f2eda2..2cb4b96db721 100644
@@ -346,6 +346,7 @@ enum irdma_cqp_op_type {
 #define IRDMA_AE_LLP_TOO_MANY_KEEPALIVE_RETRIES				0x050b
 #define IRDMA_AE_LLP_DOUBT_REACHABILITY					0x050c
 #define IRDMA_AE_LLP_CONNECTION_ESTABLISHED				0x050e
+#define IRDMA_AE_LLP_TOO_MANY_RNRS					0x050f
 #define IRDMA_AE_RESOURCE_EXHAUSTION					0x0520
 #define IRDMA_AE_RESET_SENT						0x0601
 #define IRDMA_AE_TERMINATE_SENT						0x0602

index 2f8d18d8be3b..ad50b77282f8 100644
@@ -387,6 +387,7 @@ static void irdma_process_aeq(struct irdma_pci_f *rf)
 		case IRDMA_AE_LLP_TOO_MANY_RETRIES:
 		case IRDMA_AE_LCE_QP_CATASTROPHIC:
 		case IRDMA_AE_LCE_FUNCTION_CATASTROPHIC:
+		case IRDMA_AE_LLP_TOO_MANY_RNRS:
 		case IRDMA_AE_LCE_CQ_CATASTROPHIC:
 		case IRDMA_AE_UDA_XMIT_DGRAM_TOO_LONG:
 		default:

2.42.0

Applied, thanks!

[1/4] RDMA/irdma: Fix KASAN issue with tasklet

      https://git.kernel.org/rdma/rdma/c/bd97cea7b18a0a
[2/4] RDMA/irdma: Validate max_send_wr and max_recv_wr
      https://git.kernel.org/rdma/rdma/c/ee107186bcfd25
[3/4] RDMA/irdma: Set the CQ read threshold for GEN 1
      https://git.kernel.org/rdma/rdma/c/4823a004d8301d
[4/4] RDMA/irdma: Add AE for too many RNRS
      https://git.kernel.org/rdma/rdma/c/772e5fb3884306

Best regards,

Leon Romanovsky <[email protected]>