2022년 1월 4일 (화) 오전 9:45, Namjae Jeon <linkinjeon@xxxxxxxxxx>님이 작성:
>
> 2022-01-03 9:02 GMT+09:00, Hyunchul Lee <hyc.lee@xxxxxxxxx>:
> > if CONFIG_LOCKDEP is enabled, the following
> > kernel warning message is generated because
> > rdma_accept() is called not under CM(Connection
> > Manager) handler.
> >
> > [ 63.211405 ] WARNING: CPU: 1 PID: 345 at
> > drivers/infiniband/core/cma.c:4405 rdma_accept+0x17a/0x350
> > [ 63.212080 ] RIP: 0010:rdma_accept+0x17a/0x350
> > ...
> > [ 63.214036 ] Call Trace:
> > [ 63.214098 ] <TASK>
> > [ 63.214185 ] smb_direct_accept_client+0xb4/0x170 [ksmbd]
> > [ 63.214412 ] smb_direct_prepare+0x322/0x8c0 [ksmbd]
> > [ 63.214555 ] ? rcu_read_lock_sched_held+0x3a/0x70
> > [ 63.214700 ] ksmbd_conn_handler_loop+0x63/0x270 [ksmbd]
> > [ 63.214826 ] ? ksmbd_conn_alive+0x80/0x80 [ksmbd]
> > [ 63.214952 ] kthread+0x171/0x1a0
> > [ 63.215039 ] ? set_kthread_struct+0x40/0x40
> > [ 63.215128 ] ret_from_fork+0x22/0x30
> I could not understand why lockdep trigger this warning.
> Could you elaborate more ?
>
rdma_accept checks whether the handler_mutex is held. if not,
this warning is generated. And CM holds the handler_mutex before
ksmbd's handler callback is called.
> >
> > To avoid this, move creating a queue pair and accepting
> > a client from transport_ops->prepare() to
> > smb_direct_handle_connect_request().
> >
> > Signed-off-by: Hyunchul Lee <hyc.lee@xxxxxxxxx>
> > ---
> > fs/ksmbd/transport_rdma.c | 97 +++++++++++++++++++++++----------------
> > 1 file changed, 57 insertions(+), 40 deletions(-)
> >
> > diff --git a/fs/ksmbd/transport_rdma.c b/fs/ksmbd/transport_rdma.c
> > index f89b64e27836..b37e4b9580ae 100644
> > --- a/fs/ksmbd/transport_rdma.c
> > +++ b/fs/ksmbd/transport_rdma.c
> > @@ -568,6 +568,7 @@ static void recv_done(struct ib_cq *cq, struct ib_wc
> > *wc)
> > }
> > t->negotiation_requested = true;
> > t->full_packet_received = true;
> > + enqueue_reassembly(t, recvmsg, 0);
> Is this a fix related to this patch?
Yes, this is required to receive the negotiation request
from a client.
Before this patch, posting a buffer and waiting for
the request is done in a function. Because
we have the reference of the buffer, we don't need to
append it into the queue.
> > wake_up_interruptible(&t->wait_status);
> > break;
> > case SMB_DIRECT_MSG_DATA_TRANSFER: {
> > @@ -1594,19 +1595,13 @@ static int smb_direct_accept_client(struct
> > smb_direct_transport *t)
> > pr_err("error at rdma_accept: %d\n", ret);
> > return ret;
> > }
> > -
> > - wait_event_interruptible(t->wait_status,
> > - t->status != SMB_DIRECT_CS_NEW);
> > - if (t->status != SMB_DIRECT_CS_CONNECTED)
> > - return -ENOTCONN;
> > return 0;
> > }
> >
> > -static int smb_direct_negotiate(struct smb_direct_transport *t)
> > +static int smb_direct_prepare_negotiation(struct smb_direct_transport *t)
> > {
> > int ret;
> > struct smb_direct_recvmsg *recvmsg;
> > - struct smb_direct_negotiate_req *req;
> >
> > recvmsg = get_free_recvmsg(t);
> > if (!recvmsg)
> > @@ -1616,44 +1611,20 @@ static int smb_direct_negotiate(struct
> > smb_direct_transport *t)
> > ret = smb_direct_post_recv(t, recvmsg);
> > if (ret) {
> > pr_err("Can't post recv: %d\n", ret);
> > - goto out;
> > + goto out_err;
> > }
> >
> > t->negotiation_requested = false;
> > ret = smb_direct_accept_client(t);
> > if (ret) {
> > pr_err("Can't accept client\n");
> > - goto out;
> > + goto out_err;
> > }
> >
> > smb_direct_post_recv_credits(&t->post_recv_credits_work.work);
> > -
> > - ksmbd_debug(RDMA, "Waiting for SMB_DIRECT negotiate request\n");
> > - ret = wait_event_interruptible_timeout(t->wait_status,
> > - t->negotiation_requested ||
> > - t->status == SMB_DIRECT_CS_DISCONNECTED,
> > - SMB_DIRECT_NEGOTIATE_TIMEOUT * HZ);
> > - if (ret <= 0 || t->status == SMB_DIRECT_CS_DISCONNECTED) {
> > - ret = ret < 0 ? ret : -ETIMEDOUT;
> > - goto out;
> > - }
> > -
> > - ret = smb_direct_check_recvmsg(recvmsg);
> > - if (ret == -ECONNABORTED)
> > - goto out;
> > -
> > - req = (struct smb_direct_negotiate_req *)recvmsg->packet;
> > - t->max_recv_size = min_t(int, t->max_recv_size,
> > - le32_to_cpu(req->preferred_send_size));
> > - t->max_send_size = min_t(int, t->max_send_size,
> > - le32_to_cpu(req->max_receive_size));
> > - t->max_fragmented_send_size =
> > - le32_to_cpu(req->max_fragmented_size);
> > -
> > - ret = smb_direct_send_negotiate_response(t, ret);
> > -out:
> > - if (recvmsg)
> > - put_recvmsg(t, recvmsg);
> > + return 0;
> > +out_err:
> > + put_recvmsg(t, recvmsg);
> > return ret;
> > }
> >
> > @@ -1890,6 +1861,47 @@ static int smb_direct_create_qpair(struct
> > smb_direct_transport *t,
> > static int smb_direct_prepare(struct ksmbd_transport *t)
> > {
> > struct smb_direct_transport *st = smb_trans_direct_transfort(t);
> > + struct smb_direct_recvmsg *recvmsg;
> > + struct smb_direct_negotiate_req *req;
> > + int ret;
> > +
> > + ksmbd_debug(RDMA, "Waiting for SMB_DIRECT negotiate request\n");
> > + ret = wait_event_interruptible_timeout(st->wait_status,
> > + st->negotiation_requested ||
> > + st->status == SMB_DIRECT_CS_DISCONNECTED,
> > + SMB_DIRECT_NEGOTIATE_TIMEOUT * HZ);
> > + if (ret <= 0 || st->status == SMB_DIRECT_CS_DISCONNECTED)
> > + return ret < 0 ? ret : -ETIMEDOUT;
> > +
> > + recvmsg = get_first_reassembly(st);
> > + if (!recvmsg)
> > + return -ECONNABORTED;
> > +
> > + ret = smb_direct_check_recvmsg(recvmsg);
> > + if (ret == -ECONNABORTED)
> > + goto out;
> > +
> > + req = (struct smb_direct_negotiate_req *)recvmsg->packet;
> > + st->max_recv_size = min_t(int, st->max_recv_size,
> > + le32_to_cpu(req->preferred_send_size));
> > + st->max_send_size = min_t(int, st->max_send_size,
> > + le32_to_cpu(req->max_receive_size));
> > + st->max_fragmented_send_size =
> > + le32_to_cpu(req->max_fragmented_size);
> > +
> > + ret = smb_direct_send_negotiate_response(st, ret);
> > +out:
> > + spin_lock_irq(&st->reassembly_queue_lock);
> > + st->reassembly_queue_length--;
> > + list_del(&recvmsg->list);
> > + spin_unlock_irq(&st->reassembly_queue_lock);
> > + put_recvmsg(st, recvmsg);
> > +
> > + return ret;
> > +}
> > +
> > +static int smb_direct_connect(struct smb_direct_transport *st)
> > +{
> > int ret;
> > struct ib_qp_cap qp_cap;
> >
> > @@ -1911,13 +1923,11 @@ static int smb_direct_prepare(struct ksmbd_transport
> > *t)
> > return ret;
> > }
> >
> > - ret = smb_direct_negotiate(st);
> > + ret = smb_direct_prepare_negotiation(st);
> > if (ret) {
> > pr_err("Can't negotiate: %d\n", ret);
> > return ret;
> > }
> > -
> > - st->status = SMB_DIRECT_CS_CONNECTED;
> > return 0;
> > }
> >
> > @@ -1933,6 +1943,7 @@ static bool rdma_frwr_is_supported(struct
> > ib_device_attr *attrs)
> > static int smb_direct_handle_connect_request(struct rdma_cm_id *new_cm_id)
> > {
> > struct smb_direct_transport *t;
> > + int ret;
> >
> > if (!rdma_frwr_is_supported(&new_cm_id->device->attrs)) {
> > ksmbd_debug(RDMA,
> > @@ -1945,11 +1956,17 @@ static int smb_direct_handle_connect_request(struct
> > rdma_cm_id *new_cm_id)
> > if (!t)
> > return -ENOMEM;
> >
> > + ret = smb_direct_connect(t);
> > + if (ret) {
> > + free_transport(t);
> > + return ret;
> I think that it is better to use goto statement to out after freeing
> transport structure.
Okay, I will change it.
> > + }
> > +
> > KSMBD_TRANS(t)->handler = kthread_run(ksmbd_conn_handler_loop,
> > KSMBD_TRANS(t)->conn, "ksmbd:r%u",
> > smb_direct_port);
> > if (IS_ERR(KSMBD_TRANS(t)->handler)) {
> > - int ret = PTR_ERR(KSMBD_TRANS(t)->handler);
> > + ret = PTR_ERR(KSMBD_TRANS(t)->handler);
> >
> > pr_err("Can't start thread\n");
> > free_transport(t);
> > --
> > 2.25.1
> >
> >
--
Thanks,
Hyunchul
