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

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

[Karsten Graul]

Commit 8f3d65c16679 ("net/smc: fix wait on already cleared link")
introduced link refcounting to avoid waits on already cleared links.
This patch extents and improves the refcounting to cover all
remaining possible cases for this kind of error situation.

Fixes: 15e1b99aadfb ("net/smc: no WR buffer wait for terminating link group")
Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
---
 net/smc/smc_cdc.c  |  7 +++++-
 net/smc/smc_core.c | 20 ++++++++-------
 net/smc/smc_llc.c  | 63 +++++++++++++++++++++++++++++++++++-----------
 net/smc/smc_tx.c   | 22 ++++------------
 net/smc/smc_wr.h   | 14 +++++++++++
 5 files changed, 85 insertions(+), 41 deletions(-)

diff --git a/net/smc/smc_cdc.c b/net/smc/smc_cdc.c
index f23f558054a7..99acd337ba90 100644
--- a/net/smc/smc_cdc.c
+++ b/net/smc/smc_cdc.c
@@ -150,9 +150,11 @@ static int smcr_cdc_get_slot_and_msg_send(struct smc_connection *conn)
 
 again:
 	link = conn->lnk;
+	if (!smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, NULL, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 
 	spin_lock_bh(&conn->send_lock);
 	if (link != conn->lnk) {
@@ -160,6 +162,7 @@ static int smcr_cdc_get_slot_and_msg_send(struct smc_connection *conn)
 		spin_unlock_bh(&conn->send_lock);
 		smc_wr_tx_put_slot(link,
 				   (struct smc_wr_tx_pend_priv *)pend);
+		smc_wr_tx_link_put(link);
 		if (again)
 			return -ENOLINK;
 		again = true;
@@ -167,6 +170,8 @@ static int smcr_cdc_get_slot_and_msg_send(struct smc_connection *conn)
 	}
 	rc = smc_cdc_msg_send(conn, wr_buf, pend);
 	spin_unlock_bh(&conn->send_lock);
+put_out:
+	smc_wr_tx_link_put(link);
 	return rc;
 }
 
diff --git a/net/smc/smc_core.c b/net/smc/smc_core.c
index 8280c938be80..d2206743dc71 100644
--- a/net/smc/smc_core.c
+++ b/net/smc/smc_core.c
@@ -949,7 +949,7 @@ struct smc_link *smc_switch_conns(struct smc_link_group *lgr,
 		to_lnk = &lgr->lnk[i];
 		break;
 	}
-	if (!to_lnk) {
+	if (!to_lnk || !smc_wr_tx_link_hold(to_lnk)) {
 		smc_lgr_terminate_sched(lgr);
 		return NULL;
 	}
@@ -981,24 +981,26 @@ struct smc_link *smc_switch_conns(struct smc_link_group *lgr,
 		read_unlock_bh(&lgr->conns_lock);
 		/* pre-fetch buffer outside of send_lock, might sleep */
 		rc = smc_cdc_get_free_slot(conn, to_lnk, &wr_buf, NULL, &pend);
-		if (rc) {
-			smcr_link_down_cond_sched(to_lnk);
-			return NULL;
-		}
+		if (rc)
+			goto err_out;
 		/* avoid race with smcr_tx_sndbuf_nonempty() */
 		spin_lock_bh(&conn->send_lock);
 		smc_switch_link_and_count(conn, to_lnk);
 		rc = smc_switch_cursor(smc, pend, wr_buf);
 		spin_unlock_bh(&conn->send_lock);
 		sock_put(&smc->sk);
-		if (rc) {
-			smcr_link_down_cond_sched(to_lnk);
-			return NULL;
-		}
+		if (rc)
+			goto err_out;
 		goto again;
 	}
 	read_unlock_bh(&lgr->conns_lock);
+	smc_wr_tx_link_put(to_lnk);
 	return to_lnk;
+
+err_out:
+	smcr_link_down_cond_sched(to_lnk);
+	smc_wr_tx_link_put(to_lnk);
+	return NULL;
 }
 
 static void smcr_buf_unuse(struct smc_buf_desc *rmb_desc,
diff --git a/net/smc/smc_llc.c b/net/smc/smc_llc.c
index 2e7560eba981..72f4b72eb175 100644
--- a/net/smc/smc_llc.c
+++ b/net/smc/smc_llc.c
@@ -383,9 +383,11 @@ int smc_llc_send_confirm_link(struct smc_link *link,
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
+	if (!smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	confllc = (struct smc_llc_msg_confirm_link *)wr_buf;
 	memset(confllc, 0, sizeof(*confllc));
 	confllc->hd.common.type = SMC_LLC_CONFIRM_LINK;
@@ -402,6 +404,8 @@ int smc_llc_send_confirm_link(struct smc_link *link,
 	confllc->max_links = SMC_LLC_ADD_LNK_MAX_LINKS;
 	/* send llc message */
 	rc = smc_wr_tx_send(link, pend);
+put_out:
+	smc_wr_tx_link_put(link);
 	return rc;
 }
 
@@ -415,9 +419,11 @@ static int smc_llc_send_confirm_rkey(struct smc_link *send_link,
 	struct smc_link *link;
 	int i, rc, rtok_ix;
 
+	if (!smc_wr_tx_link_hold(send_link))
+		return -ENOLINK;
 	rc = smc_llc_add_pending_send(send_link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	rkeyllc = (struct smc_llc_msg_confirm_rkey *)wr_buf;
 	memset(rkeyllc, 0, sizeof(*rkeyllc));
 	rkeyllc->hd.common.type = SMC_LLC_CONFIRM_RKEY;
@@ -444,6 +450,8 @@ static int smc_llc_send_confirm_rkey(struct smc_link *send_link,
 		(u64)sg_dma_address(rmb_desc->sgt[send_link->link_idx].sgl));
 	/* send llc message */
 	rc = smc_wr_tx_send(send_link, pend);
+put_out:
+	smc_wr_tx_link_put(send_link);
 	return rc;
 }
 
@@ -456,9 +464,11 @@ static int smc_llc_send_delete_rkey(struct smc_link *link,
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
+	if (!smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	rkeyllc = (struct smc_llc_msg_delete_rkey *)wr_buf;
 	memset(rkeyllc, 0, sizeof(*rkeyllc));
 	rkeyllc->hd.common.type = SMC_LLC_DELETE_RKEY;
@@ -467,6 +477,8 @@ static int smc_llc_send_delete_rkey(struct smc_link *link,
 	rkeyllc->rkey[0] = htonl(rmb_desc->mr_rx[link->link_idx]->rkey);
 	/* send llc message */
 	rc = smc_wr_tx_send(link, pend);
+put_out:
+	smc_wr_tx_link_put(link);
 	return rc;
 }
 
@@ -480,9 +492,11 @@ int smc_llc_send_add_link(struct smc_link *link, u8 mac[], u8 gid[],
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
+	if (!smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	addllc = (struct smc_llc_msg_add_link *)wr_buf;
 
 	memset(addllc, 0, sizeof(*addllc));
@@ -504,6 +518,8 @@ int smc_llc_send_add_link(struct smc_link *link, u8 mac[], u8 gid[],
 	}
 	/* send llc message */
 	rc = smc_wr_tx_send(link, pend);
+put_out:
+	smc_wr_tx_link_put(link);
 	return rc;
 }
 
@@ -517,9 +533,11 @@ int smc_llc_send_delete_link(struct smc_link *link, u8 link_del_id,
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
+	if (!smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	delllc = (struct smc_llc_msg_del_link *)wr_buf;
 
 	memset(delllc, 0, sizeof(*delllc));
@@ -536,6 +554,8 @@ int smc_llc_send_delete_link(struct smc_link *link, u8 link_del_id,
 	delllc->reason = htonl(reason);
 	/* send llc message */
 	rc = smc_wr_tx_send(link, pend);
+put_out:
+	smc_wr_tx_link_put(link);
 	return rc;
 }
 
@@ -547,9 +567,11 @@ static int smc_llc_send_test_link(struct smc_link *link, u8 user_data[16])
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
+	if (!smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	testllc = (struct smc_llc_msg_test_link *)wr_buf;
 	memset(testllc, 0, sizeof(*testllc));
 	testllc->hd.common.type = SMC_LLC_TEST_LINK;
@@ -557,6 +579,8 @@ static int smc_llc_send_test_link(struct smc_link *link, u8 user_data[16])
 	memcpy(testllc->user_data, user_data, sizeof(testllc->user_data));
 	/* send llc message */
 	rc = smc_wr_tx_send(link, pend);
+put_out:
+	smc_wr_tx_link_put(link);
 	return rc;
 }
 
@@ -567,13 +591,16 @@ static int smc_llc_send_message(struct smc_link *link, void *llcbuf)
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
-	if (!smc_link_usable(link))
+	if (!smc_wr_tx_link_hold(link))
 		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	memcpy(wr_buf, llcbuf, sizeof(union smc_llc_msg));
-	return smc_wr_tx_send(link, pend);
+	rc = smc_wr_tx_send(link, pend);
+put_out:
+	smc_wr_tx_link_put(link);
+	return rc;
 }
 
 /* schedule an llc send on link, may wait for buffers,
@@ -586,13 +613,16 @@ static int smc_llc_send_message_wait(struct smc_link *link, void *llcbuf)
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
-	if (!smc_link_usable(link))
+	if (!smc_wr_tx_link_hold(link))
 		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	memcpy(wr_buf, llcbuf, sizeof(union smc_llc_msg));
-	return smc_wr_tx_send_wait(link, pend, SMC_LLC_WAIT_TIME);
+	rc = smc_wr_tx_send_wait(link, pend, SMC_LLC_WAIT_TIME);
+put_out:
+	smc_wr_tx_link_put(link);
+	return rc;
 }
 
 /********************************* receive ***********************************/
@@ -672,9 +702,11 @@ static int smc_llc_add_link_cont(struct smc_link *link,
 	struct smc_buf_desc *rmb;
 	u8 n;
 
+	if (!smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_llc_add_pending_send(link, &wr_buf, &pend);
 	if (rc)
-		return rc;
+		goto put_out;
 	addc_llc = (struct smc_llc_msg_add_link_cont *)wr_buf;
 	memset(addc_llc, 0, sizeof(*addc_llc));
 
@@ -706,7 +738,10 @@ static int smc_llc_add_link_cont(struct smc_link *link,
 	addc_llc->hd.length = sizeof(struct smc_llc_msg_add_link_cont);
 	if (lgr->role == SMC_CLNT)
 		addc_llc->hd.flags |= SMC_LLC_FLAG_RESP;
-	return smc_wr_tx_send(link, pend);
+	rc = smc_wr_tx_send(link, pend);
+put_out:
+	smc_wr_tx_link_put(link);
+	return rc;
 }
 
 static int smc_llc_cli_rkey_exchange(struct smc_link *link,
diff --git a/net/smc/smc_tx.c b/net/smc/smc_tx.c
index c79361dfcdfb..738a4a99c827 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;
@@ -505,8 +505,11 @@ static int _smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
 	struct smc_wr_buf *wr_buf;
 	int rc;
 
+	if (!link || !smc_wr_tx_link_hold(link))
+		return -ENOLINK;
 	rc = smc_cdc_get_free_slot(conn, link, &wr_buf, &wr_rdma_buf, &pend);
 	if (rc < 0) {
+		smc_wr_tx_link_put(link);
 		if (rc == -EBUSY) {
 			struct smc_sock *smc =
 				container_of(conn, struct smc_sock, conn);
@@ -547,22 +550,7 @@ static int _smcr_tx_sndbuf_nonempty(struct smc_connection *conn)
 
 out_unlock:
 	spin_unlock_bh(&conn->send_lock);
-	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);
+	smc_wr_tx_link_put(link);
 	return rc;
 }
 
diff --git a/net/smc/smc_wr.h b/net/smc/smc_wr.h
index 423b8709f1c9..2bc626f230a5 100644
--- a/net/smc/smc_wr.h
+++ b/net/smc/smc_wr.h
@@ -60,6 +60,20 @@ static inline void smc_wr_tx_set_wr_id(atomic_long_t *wr_tx_id, long val)
 	atomic_long_set(wr_tx_id, val);
 }
 
+static inline bool smc_wr_tx_link_hold(struct smc_link *link)
+{
+	if (!smc_link_usable(link))
+		return false;
+	atomic_inc(&link->wr_tx_refcnt);
+	return true;
+}
+
+static inline void smc_wr_tx_link_put(struct smc_link *link)
+{
+	if (atomic_dec_and_test(&link->wr_tx_refcnt))
+		wake_up_all(&link->wr_tx_wait);
+}
+
 static inline void smc_wr_wakeup_tx_wait(struct smc_link *lnk)
 {
 	wake_up_all(&lnk->wr_tx_wait);
-- 
2.25.1

Re: [PATCH net] net/smc: improved fix wait on already cleared link

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (master)
by David S. Miller <davem@davemloft.net>:

On Thu,  7 Oct 2021 16:14:40 +0200 you wrote:
> Commit 8f3d65c16679 ("net/smc: fix wait on already cleared link")
> introduced link refcounting to avoid waits on already cleared links.
> This patch extents and improves the refcounting to cover all
> remaining possible cases for this kind of error situation.
> 
> Fixes: 15e1b99aadfb ("net/smc: no WR buffer wait for terminating link group")
> Signed-off-by: Karsten Graul <kgraul@linux.ibm.com>
> 
> [...]

Here is the summary with links:
  - [net] net/smc: improved fix wait on already cleared link
    https://git.kernel.org/netdev/net/c/95f7f3e7dc6b

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