netdev - Re: [PATCH rdma-next 1/1] RDMA/mana_ib: Set correct device into ib

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID:
 <PAXPR83MB0559F4678E73B0091A8ADFBBB4D62@PAXPR83MB0559.EURPRD83.prod.outlook.com>
Date: Wed, 26 Jun 2024 09:05:05 +0000
From: Konstantin Taranov <kotaranov@...rosoft.com>
To: Leon Romanovsky <leon@...nel.org>, Konstantin Taranov
	<kotaranov@...ux.microsoft.com>
CC: Wei Hu <weh@...rosoft.com>, "sharmaajay@...rosoft.com"
	<sharmaajay@...rosoft.com>, Long Li <longli@...rosoft.com>, "jgg@...pe.ca"
	<jgg@...pe.ca>, "linux-rdma@...r.kernel.org" <linux-rdma@...r.kernel.org>,
	linux-netdev <netdev@...r.kernel.org>
Subject: Re: [PATCH rdma-next 1/1] RDMA/mana_ib: Set correct device into ib

> > When mc->ports[0] is not slave, use it in the set_netdev.
> > When mana is used in netvsc, the stored net devices in mana are slaves
> > and GIDs should be taken from their master devices.
> > In the baremetal case, the mc->ports devices will not be slaves.
> 
> I wonder, why do you have "... | IFF_SLAVE" in __netvsc_vf_setup() in a first
> place? Isn't IFF_SLAVE is supposed to be set by bond driver?
> 

I guess it is just a valid use of the IFF_SLAVE bit. In the bond case it is also set
as a BOND netdev. The IFF_SLAVE helps to show users that another master
netdev should be used for networking. But I am not an expert in netvsc.

Actually, another alternative solution for mana_ib is always set the slave device,
but in the GID mgmt code we need the following patch. The problem is that it may require 
testing/confirmation from other ib providers as in the worst case some GIDs will not be listed.

diff --git a/drivers/infiniband/core/roce_gid_mgmt.c b/drivers/infiniband/core/roce_gid_mgmt.c
index d5131b3ba8ab..0f20b4e2d1c2 100644
--- a/drivers/infiniband/core/roce_gid_mgmt.c
+++ b/drivers/infiniband/core/roce_gid_mgmt.c
@@ -141,6 +141,8 @@ static enum bonding_slave_state is_eth_active_slave_of_bonding_rcu(struct net_de
        return BONDING_SLAVE_STATE_NA;
 }

+#define netdev_is_slave(dev)   (((dev)->flags & IFF_SLAVE) == IFF_SLAVE)
+
 #define REQUIRED_BOND_STATES           (BONDING_SLAVE_STATE_ACTIVE |   \
                                         BONDING_SLAVE_STATE_NA)
 static bool
@@ -157,11 +159,14 @@ is_eth_port_of_netdev_filter(struct ib_device *ib_dev, u32 port,
        real_dev = rdma_vlan_dev_real_dev(cookie);
        if (!real_dev)
                real_dev = cookie;
-
+       /*
+        * When rdma netdevice is used in netvsc, the master netdevice should
+        * be considered for GIDs. Therefore, ignore slave rdma netdevices.
+        */
        res = ((rdma_is_upper_dev_rcu(rdma_ndev, cookie) &&
               (is_eth_active_slave_of_bonding_rcu(rdma_ndev, real_dev) &
                REQUIRED_BOND_STATES)) ||
-              real_dev == rdma_ndev);
+              (real_dev == rdma_ndev && !netdev_is_slave(real_dev)));

        rcu_read_unlock();
        return res;
@@ -211,12 +216,14 @@ is_ndev_for_default_gid_filter(struct ib_device *ib_dev, u32 port,

        /*
         * When rdma netdevice is used in bonding, bonding master netdevice
-        * should be considered for default GIDs. Therefore, ignore slave rdma
-        * netdevices when bonding is considered.
+        * should be considered for default GIDs.
+        * When rdma netdevice is used in netvsc, the master netdevice should
+        * be considered for defauld GIDs. Therefore, ignore slave rdma
+        * netdevices.
         * Additionally when event(cookie) netdevice is bond master device,
         * make sure that it the upper netdevice of rdma netdevice.
         */
-       res = ((cookie_ndev == rdma_ndev && !netif_is_bond_slave(rdma_ndev)) ||
+       res = ((cookie_ndev == rdma_ndev && !netdev_is_slave(rdma_ndev)) ||
               (netif_is_bond_master(cookie_ndev) &&
                rdma_is_upper_dev_rcu(rdma_ndev, cookie_ndev)));

> > +#define mana_ndev_is_slave(dev)   (((dev)->flags & IFF_SLAVE) ==
> IFF_SLAVE)
> 
> There is no need in macro for one line of code and there is no need in "==",
> as the result will be boolean.
> 

Sure, can address in v2. I just saw a similar macro in another kernel file.