[PATCH net 0/2] net/smc: fixes 2021-08-09

[PATCH net 0/2] net/smc: fixes 2021-08-09

[Guvenc Gulce]

Hi Dave & Jakub,
 
please apply the following patch series for smc to netdev's net tree. 
One patch fixes invalid connection counting for links and the other
one fixes an access to an already cleared link.

Thanks,

Guvenc

Guvenc Gulce (1):
  net/smc: Correct smc link connection counter in case of smc client

Karsten Graul (1):
  net/smc: fix wait on already cleared link

 net/smc/af_smc.c   |  2 +-
 net/smc/smc_core.c |  4 ++--
 net/smc/smc_core.h |  4 ++++
 net/smc/smc_llc.c  | 10 ++++------
 net/smc/smc_tx.c   | 18 +++++++++++++++++-
 net/smc/smc_wr.c   | 10 ++++++++++
 6 files changed, 38 insertions(+), 10 deletions(-)

-- 
2.25.1

[PATCH net 1/2] net/smc: fix wait on already cleared link

[Guvenc Gulce]

From: Karsten Graul <kgraul@linux.ibm.com>

There can be a race between the waiters for a tx work request buffer
and the link down processing that finally clears the link. Although
all waiters are woken up before the link is cleared there might be
waiters which did not yet get back control and are still waiting.
This results in an access to a cleared wait queue head.

Fix this by introducing atomic reference counting around the wait calls,
and wait with the link clear processing until all waiters have finished.
Move the work request layer related calls into smc_wr.c and set the
link state to INACTIVE before calling smcr_link_clear() in
smc_llc_srv_add_link().

Fixes: 15e1b99aadfb ("net/smc: no WR buffer wait for terminating link group")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
---
 net/smc/smc_core.h |  2 ++
 net/smc/smc_llc.c  | 10 ++++------
 net/smc/smc_tx.c   | 18 +++++++++++++++++-
 net/smc/smc_wr.c   | 10 ++++++++++
 4 files changed, 33 insertions(+), 7 deletions(-)

diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h
index 6d6fd1397c87..64d86298e4df 100644
--- a/net/smc/smc_core.h
+++ b/net/smc/smc_core.h
@@ -97,6 +97,7 @@ struct smc_link {
 	unsigned long		*wr_tx_mask;	/* bit mask of used indexes */
 	u32			wr_tx_cnt;	/* number of WR send buffers */
 	wait_queue_head_t	wr_tx_wait;	/* wait for free WR send buf */
+	atomic_t		wr_tx_refcnt;	/* tx refs to link */
 
 	struct smc_wr_buf	*wr_rx_bufs;	/* WR recv payload buffers */
 	struct ib_recv_wr	*wr_rx_ibs;	/* WR recv meta data */
@@ -109,6 +110,7 @@ struct smc_link {
 
 	struct ib_reg_wr	wr_reg;		/* WR register memory region */
 	wait_queue_head_t	wr_reg_wait;	/* wait for wr_reg result */
+	atomic_t		wr_reg_refcnt;	/* reg refs to link */
 	enum smc_wr_reg_state	wr_reg_state;	/* state of wr_reg request */
 
 	u8			gid[SMC_GID_SIZE];/* gid matching used vlan id*/
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 273eaf1bfe49..2e7560eba981 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -888,6 +888,7 @@ int smc_llc_cli_add_link(struct smc_link *link, struct smc_llc_qentry *qentry)
 	if (!rc)
 		goto out;
 out_clear_lnk:
+	lnk_new->state = SMC_LNK_INACTIVE;
 	smcr_link_clear(lnk_new, false);
 out_reject:
 	smc_llc_cli_add_link_reject(qentry);
@@ -1184,6 +1185,7 @@ int smc_llc_srv_add_link(struct smc_link *link)
 		goto out_err;
 	return 0;
 out_err:
+	link_new->state = SMC_LNK_INACTIVE;
 	smcr_link_clear(link_new, false);
 	return rc;
 }
@@ -1286,10 +1288,8 @@ static void smc_llc_process_cli_delete_link(struct smc_link_group *lgr)
 	del_llc->reason = 0;
 	smc_llc_send_message(lnk, &qentry->msg); /* response */
 
-	if (smc_link_downing(&lnk_del->state)) {
-		if (smc_switch_conns(lgr, lnk_del, false))
-			smc_wr_tx_wait_no_pending_sends(lnk_del);
-	}
+	if (smc_link_downing(&lnk_del->state))
+		smc_switch_conns(lgr, lnk_del, false);
 	smcr_link_clear(lnk_del, true);
 
 	active_links = smc_llc_active_link_count(lgr);
@@ -1805,8 +1805,6 @@ void smc_llc_link_clear(struct smc_link *link, bool log)
 				    link->smcibdev->ibdev->name, link->ibport);
 	complete(&link->llc_testlink_resp);
 	cancel_delayed_work_sync(&link->llc_testlink_wrk);
-	smc_wr_wakeup_reg_wait(link);
-	smc_wr_wakeup_tx_wait(link);
 }
 
 /* register a new rtoken at the remote peer (for all links) */
diff --git a/net/smc/smc_tx.c b/net/smc/smc_tx.c
index 289025cd545a..c79361dfcdfb 100644
--- a/net/smc/smc_tx.c
+++ b/net/smc/smc_tx.c
@@ -496,7 +496,7 @@ static int smc_tx_rdma_writes(struct smc_connection *conn,
 /* Wakeup sndbuf consumers from any context (IRQ or process)
  * since there is more data to transmit; usable snd_wnd as max transmit
  */
-static int smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
+static int _smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
 {
 	struct smc_cdc_producer_flags *pflags = &conn->local_tx_ctrl.prod_flags;
 	struct smc_link *link = conn->lnk;
@@ -550,6 +550,22 @@ static int smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
 	return rc;
 }
 
+static int smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
+{
+	struct smc_link *link = conn->lnk;
+	int rc = -ENOLINK;
+
+	if (!link)
+		return rc;
+
+	atomic_inc(&link->wr_tx_refcnt);
+	if (smc_link_usable(link))
+		rc = _smcr_tx_sndbuf_nonempty(conn);
+	if (atomic_dec_and_test(&link->wr_tx_refcnt))
+		wake_up_all(&link->wr_tx_wait);
+	return rc;
+}
+
 static int smcd_tx_sndbuf_nonempty(struct smc_connection *conn)
 {
 	struct smc_cdc_producer_flags *pflags = &conn->local_tx_ctrl.prod_flags;
diff --git a/net/smc/smc_wr.c b/net/smc/smc_wr.c
index cbc73a7e4d59..a419e9af36b9 100644
--- a/net/smc/smc_wr.c
+++ b/net/smc/smc_wr.c
@@ -322,9 +322,12 @@ int smc_wr_reg_send(struct smc_link *link, struct ib_mr *mr)
 	if (rc)
 		return rc;
 
+	atomic_inc(&link->wr_reg_refcnt);
 	rc = wait_event_interruptible_timeout(link->wr_reg_wait,
 					      (link->wr_reg_state != POSTED),
 					      SMC_WR_REG_MR_WAIT_TIME);
+	if (atomic_dec_and_test(&link->wr_reg_refcnt))
+		wake_up_all(&link->wr_reg_wait);
 	if (!rc) {
 		/* timeout - terminate link */
 		smcr_link_down_cond_sched(link);
@@ -566,10 +569,15 @@ void smc_wr_free_link(struct smc_link *lnk)
 		return;
 	ibdev = lnk->smcibdev->ibdev;
 
+	smc_wr_wakeup_reg_wait(lnk);
+	smc_wr_wakeup_tx_wait(lnk);
+
 	if (smc_wr_tx_wait_no_pending_sends(lnk))
 		memset(lnk->wr_tx_mask, 0,
 		       BITS_TO_LONGS(SMC_WR_BUF_CNT) *
 						sizeof(*lnk->wr_tx_mask));
+	wait_event(lnk->wr_reg_wait, (!atomic_read(&lnk->wr_reg_refcnt)));
+	wait_event(lnk->wr_tx_wait, (!atomic_read(&lnk->wr_tx_refcnt)));
 
 	if (lnk->wr_rx_dma_addr) {
 		ib_dma_unmap_single(ibdev, lnk->wr_rx_dma_addr,
@@ -728,7 +736,9 @@ int smc_wr_create_link(struct smc_link *lnk)
 	memset(lnk->wr_tx_mask, 0,
 	       BITS_TO_LONGS(SMC_WR_BUF_CNT) * sizeof(*lnk->wr_tx_mask));
 	init_waitqueue_head(&lnk->wr_tx_wait);
+	atomic_set(&lnk->wr_tx_refcnt, 0);
 	init_waitqueue_head(&lnk->wr_reg_wait);
+	atomic_set(&lnk->wr_reg_refcnt, 0);
 	return rc;
 
 dma_unmap:
-- 
2.25.1

[PATCH net 2/2] net/smc: Correct smc link connection counter in case of smc client

[Guvenc Gulce]

SMC clients may be assigned to a different link after the initial
connection between two peers was established. In such a case,
the connection counter was not correctly set.

Update the connection counter correctly when a smc client connection
is assigned to a different smc link.

Fixes: 07d51580ff65 ("net/smc: Add connection counters for links")
Signed-off-by: Guvenc Gulce <guvenc@linux.ibm.com>
Tested-by: Karsten Graul <kgraul@linux.ibm.com>
---
 net/smc/af_smc.c   | 2 +-
 net/smc/smc_core.c | 4 ++--
 net/smc/smc_core.h | 2 ++
 3 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/net/smc/af_smc.c b/net/smc/af_smc.c
index 898389611ae8..c038efc23ce3 100644
--- a/net/smc/af_smc.c
+++ b/net/smc/af_smc.c
@@ -795,7 +795,7 @@ static int smc_connect_rdma(struct smc_sock *smc,
 			reason_code = SMC_CLC_DECL_NOSRVLINK;
 			goto connect_abort;
 		}
-		smc->conn.lnk = link;
+		smc_switch_link_and_count(&smc->conn, link);
 	}
 
 	/* create send buffer and rmb */
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index cd0d7c908b2a..c160ff50c053 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -917,8 +917,8 @@ static int smc_switch_cursor(struct smc_sock *smc, struct smc_cdc_tx_pend *pend,
 	return rc;
 }
 
-static void smc_switch_link_and_count(struct smc_connection *conn,
-				      struct smc_link *to_lnk)
+void smc_switch_link_and_count(struct smc_connection *conn,
+			       struct smc_link *to_lnk)
 {
 	atomic_dec(&conn->lnk->conn_cnt);
 	conn->lnk = to_lnk;
diff --git a/net/smc/smc_core.h b/net/smc/smc_core.h
index 64d86298e4df..c043ecdca5c4 100644
--- a/net/smc/smc_core.h
+++ b/net/smc/smc_core.h
@@ -446,6 +446,8 @@ void smc_core_exit(void);
 int smcr_link_init(struct smc_link_group *lgr, struct smc_link *lnk,
 		   u8 link_idx, struct smc_init_info *ini);
 void smcr_link_clear(struct smc_link *lnk, bool log);
+void smc_switch_link_and_count(struct smc_connection *conn,
+			       struct smc_link *to_lnk);
 int smcr_buf_map_lgr(struct smc_link *lnk);
 int smcr_buf_reg_lgr(struct smc_link *lnk);
 void smcr_lgr_set_type(struct smc_link_group *lgr, enum smc_lgr_type new_type);
-- 
2.25.1

Re: [PATCH net 0/2] net/smc: fixes 2021-08-09

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (refs/heads/master):

On Mon,  9 Aug 2021 11:05:55 +0200 you wrote:
> Hi Dave & Jakub,
> 
> please apply the following patch series for smc to netdev's net tree.
> One patch fixes invalid connection counting for links and the other
> one fixes an access to an already cleared link.
> 
> Thanks,
> 
> [...]

Here is the summary with links:
  - [net,1/2] net/smc: fix wait on already cleared link
    https://git.kernel.org/netdev/net/c/8f3d65c16679
  - [net,2/2] net/smc: Correct smc link connection counter in case of smc client
    https://git.kernel.org/netdev/net/c/64513d269e89

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html