netdev - [PATCH net-next 1/2] net: add netif_set_real_num

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20210803130527.2411250-2-kuba@kernel.org>
Date:   Tue,  3 Aug 2021 06:05:26 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     davem@...emloft.net
Cc:     netdev@...r.kernel.org, simon.horman@...igine.com,
        alexanderduyck@...com, oss-drivers@...igine.com,
        Jakub Kicinski <kuba@...nel.org>
Subject: [PATCH net-next 1/2] net: add netif_set_real_num_queues() for device reconfig

netif_set_real_num_rx_queues() and netif_set_real_num_tx_queues()
can fail which breaks drivers trying to implement reconfiguration
in a way that can't leave the device half-broken. In other words
those functions are incompatible with prepare/commit approach.

Luckily setting real number of queues can fail only if the number
is increased, meaning that if we order operations correctly we
can guarantee ending up with either new config (success), or
the old one (on error).

Provide a helper implementing such logic so that drivers don't
have to duplicate it.

Signed-off-by: Jakub Kicinski <kuba@...nel.org>
---
 include/linux/netdevice.h |  2 ++
 net/core/dev.c            | 44 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 46 insertions(+)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index cd136499ec59..1b4d4509d04b 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -3916,6 +3916,8 @@ static inline int netif_set_real_num_rx_queues(struct net_device *dev,
 	return 0;
 }
 #endif
+int netif_set_real_num_queues(struct net_device *dev,
+			      unsigned int txq, unsigned int rxq);
 
 static inline struct netdev_rx_queue *
 __netif_get_rx_queue(struct net_device *dev, unsigned int rxq)
diff --git a/net/core/dev.c b/net/core/dev.c
index 4a1401008db9..360cb2f1b1e9 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2973,6 +2973,50 @@ int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq)
 EXPORT_SYMBOL(netif_set_real_num_rx_queues);
 #endif
 
+/**
+ *	netif_set_real_num_queues - set actual number of RX and TX queues used
+ *	@dev: Network device
+ *	@txq: Actual number of TX queues
+ *	@rxq: Actual number of RX queues
+ *
+ *	Set the real number of both TX and RX queues.
+ *	Does nothing if the number of queues is already correct.
+ */
+int netif_set_real_num_queues(struct net_device *dev,
+			      unsigned int txq, unsigned int rxq)
+{
+	unsigned int old_rxq = dev->real_num_rx_queues;
+	int err;
+
+	if (txq < 1 || txq > dev->num_tx_queues ||
+	    rxq < 1 || rxq > dev->num_rx_queues)
+		return -EINVAL;
+
+	/* Start from increases, so the error path only does decreases -
+	 * decreases can't fail.
+	 */
+	if (rxq > dev->real_num_rx_queues) {
+		err = netif_set_real_num_rx_queues(dev, rxq);
+		if (err)
+			return err;
+	}
+	if (txq > dev->real_num_tx_queues) {
+		err = netif_set_real_num_tx_queues(dev, txq);
+		if (err)
+			goto undo_rx;
+	}
+	if (rxq < dev->real_num_rx_queues)
+		WARN_ON(netif_set_real_num_rx_queues(dev, rxq));
+	if (txq < dev->real_num_tx_queues)
+		WARN_ON(netif_set_real_num_tx_queues(dev, txq));
+
+	return 0;
+undo_rx:
+	WARN_ON(netif_set_real_num_rx_queues(dev, old_rxq));
+	return err;
+}
+EXPORT_SYMBOL(netif_set_real_num_queues);
+
 /**
  * netif_get_num_default_rss_queues - default number of RSS queues
  *
-- 
2.31.1