2020-01-07 17:36:35

by Madhuparna Bhowmik

Subject: [PATCH 1/3] infiniband: hw: hfi1: verbs.c: Use built-in RCU list checking

From: Madhuparna Bhowmik <[email protected]>

list_for_each_entry_rcu has built-in RCU and lock checking.
Pass cond argument to list_for_each_entry_rcu.

Signed-off-by: Madhuparna Bhowmik <[email protected]>
---
drivers/infiniband/hw/hfi1/verbs.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 089e201d7550..cab2ff185240 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -515,7 +515,8 @@ static inline void hfi1_handle_packet(struct hfi1_packet *packet,
opa_get_lid(packet->dlid, 9B));
if (!mcast)
goto drop;
- list_for_each_entry_rcu(p, &mcast->qp_list, list) {
+ list_for_each_entry_rcu(p, &mcast->qp_list, list
+ lock_is_held(&(ibp->rvp.lock).dep_map)) {
packet->qp = p->qp;
if (hfi1_do_pkey_check(packet))
goto drop;
--
2.17.1

2020-01-07 17:37:42

by Madhuparna Bhowmik

Subject: [PATCH 2/3] infiniband: hw: qib: qib_verbs: Use built-in RCU list checking

From: Madhuparna Bhowmik <[email protected]>

Use built-in RCU and lock checking for list_for_each_entry_rcu
by passing cond argument to it.

Signed-off-by: Madhuparna Bhowmik <[email protected]>
---
drivers/infiniband/hw/qib/qib_verbs.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index 33778d451b82..4dc7514ce626 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -329,7 +329,8 @@ void qib_ib_rcv(struct qib_ctxtdata *rcd, void *rhdr, void *data, u32 tlen)
if (mcast == NULL)
goto drop;
this_cpu_inc(ibp->pmastats->n_multicast_rcv);
- list_for_each_entry_rcu(p, &mcast->qp_list, list)
+ list_for_each_entry_rcu(p, &mcast->qp_list, list
+ lock_is_held(&(ibp->rvp.lock).dep_map))
qib_qp_rcv(rcd, hdr, 1, data, tlen, p->qp);
/*
* Notify rvt_multicast_detach() if it is waiting for us
--
2.17.1

2020-01-07 17:38:14

by Madhuparna Bhowmik

Subject: [PATCH 3/3] infiniband: sw: rdmavt: mcast.c: Use built-in RCU list checking

From: Madhuparna Bhowmik <[email protected]>

Use built-in RCU and lock-checking for list_for_each_entry_rcu()
by passing the cond argument.

Signed-off-by: Madhuparna Bhowmik <[email protected]>
---
drivers/infiniband/sw/rdmavt/mcast.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/sw/rdmavt/mcast.c b/drivers/infiniband/sw/rdmavt/mcast.c
index dd11c6fcd060..5fce375f22f4 100644
--- a/drivers/infiniband/sw/rdmavt/mcast.c
+++ b/drivers/infiniband/sw/rdmavt/mcast.c
@@ -224,7 +224,8 @@ static int rvt_mcast_add(struct rvt_dev_info *rdi, struct rvt_ibport *ibp,
}

/* Search the QP list to see if this is already there. */
- list_for_each_entry_rcu(p, &tmcast->qp_list, list) {
+ list_for_each_entry_rcu(p, &tmcast->qp_list, list
+ lock_is_held(&(ibp->lock).dep_map)) {
if (p->qp == mqp->qp) {
ret = ESRCH;
goto bail;
--
2.17.1

2020-01-07 18:27:56

by Jason Gunthorpe

Subject: Re: [PATCH 1/3] infiniband: hw: hfi1: verbs.c: Use built-in RCU list checking

On Tue, Jan 07, 2020 at 11:05:08PM +0530, [email protected] wrote:
> From: Madhuparna Bhowmik <[email protected]>
>
> list_for_each_entry_rcu has built-in RCU and lock checking.
> Pass cond argument to list_for_each_entry_rcu.
>
> Signed-off-by: Madhuparna Bhowmik <[email protected]>
> drivers/infiniband/hw/hfi1/verbs.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
> index 089e201d7550..cab2ff185240 100644
> +++ b/drivers/infiniband/hw/hfi1/verbs.c
> @@ -515,7 +515,8 @@ static inline void hfi1_handle_packet(struct hfi1_packet *packet,
> opa_get_lid(packet->dlid, 9B));
> if (!mcast)
> goto drop;
> - list_for_each_entry_rcu(p, &mcast->qp_list, list) {
> + list_for_each_entry_rcu(p, &mcast->qp_list, list
> + lock_is_held(&(ibp->rvp.lock).dep_map)) {

This is missing a ',' and isn't indented properly. Does it even
compile?

The idea seems sound though.

Jason