| Message ID | 20241025072356.56093-1-wenjia@linux.ibm.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
| Series | [net] net/smc: Fix lookup of netdev by using ib_device_get_netdev() | expand |
On Fri, 25 Oct 2024 09:23:55 +0200 Wenjia Zhang <wenjia@linux.ibm.com> wrote: > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > alternative to get_netdev") introduced an API ib_device_get_netdev. > The SMC-R variant of the SMC protocol continued to use the old API > ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > device driver. Thus, using ib_device_set_netdev() now became mandatory. > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") > Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") > Reported-by: Aswin K <aswin@linux.ibm.com> > Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> > Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> Reviewed-by: Halil Pasic <pasic@linux.ibm.com>
On Fri, Oct 25, 2024 at 09:23:55AM +0200, Wenjia Zhang wrote: > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > alternative to get_netdev") introduced an API ib_device_get_netdev. > The SMC-R variant of the SMC protocol continued to use the old API > ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > device driver. Thus, using ib_device_set_netdev() now became mandatory. > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") > Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") > Reported-by: Aswin K <aswin@linux.ibm.com> > Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> > Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> Reviewed-by: Simon Horman <horms@kernel.org>
On 2024-10-25 09:23:55, Wenjia Zhang wrote: >Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an >alternative to get_netdev") introduced an API ib_device_get_netdev. >The SMC-R variant of the SMC protocol continued to use the old API >ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b >("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the >get_netdev callback from mlx5_ib_dev_common_roce_ops, calling >ib_device_ops.get_netdev didn't work any more at least by using a mlx5 >device driver. Thus, using ib_device_set_netdev() now became mandatory. > >Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > >Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") >Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") >Reported-by: Aswin K <aswin@linux.ibm.com> >Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> >Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> Reviewed-by: Dust Li <dust.li@linux.alibaba.com> >--- > net/smc/smc_ib.c | 8 ++------ > net/smc/smc_pnet.c | 4 +--- > 2 files changed, 3 insertions(+), 9 deletions(-) > >diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c >index 9297dc20bfe2..9c563cdbea90 100644 >--- a/net/smc/smc_ib.c >+++ b/net/smc/smc_ib.c >@@ -899,9 +899,7 @@ static void smc_copy_netdev_ifindex(struct smc_ib_device *smcibdev, int port) > struct ib_device *ibdev = smcibdev->ibdev; > struct net_device *ndev; > >- if (!ibdev->ops.get_netdev) >- return; >- ndev = ibdev->ops.get_netdev(ibdev, port + 1); >+ ndev = ib_device_get_netdev(ibdev, port + 1); > if (ndev) { > smcibdev->ndev_ifidx[port] = ndev->ifindex; > dev_put(ndev); >@@ -921,9 +919,7 @@ void smc_ib_ndev_change(struct net_device *ndev, unsigned long event) > port_cnt = smcibdev->ibdev->phys_port_cnt; > for (i = 0; i < min_t(size_t, port_cnt, SMC_MAX_PORTS); i++) { > libdev = smcibdev->ibdev; >- if (!libdev->ops.get_netdev) >- continue; >- lndev = libdev->ops.get_netdev(libdev, i + 1); >+ lndev = ib_device_get_netdev(libdev, i + 1); > dev_put(lndev); > if (lndev != ndev) > continue; >diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c >index 1dd362326c0a..8566937c8903 100644 >--- a/net/smc/smc_pnet.c >+++ b/net/smc/smc_pnet.c >@@ -1054,9 +1054,7 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, > for (i = 1; i <= SMC_MAX_PORTS; i++) { > if (!rdma_is_port_valid(ibdev->ibdev, i)) > continue; >- if (!ibdev->ibdev->ops.get_netdev) >- continue; >- ndev = ibdev->ibdev->ops.get_netdev(ibdev->ibdev, i); >+ ndev = ib_device_get_netdev(ibdev->ibdev, i); > if (!ndev) > continue; > dev_put(ndev); >-- >2.43.0 >
On 2024/10/25 15:23, Wenjia Zhang wrote: > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > alternative to get_netdev") introduced an API ib_device_get_netdev. > The SMC-R variant of the SMC protocol continued to use the old API > ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > device driver. Thus, using ib_device_set_netdev() now became mandatory. > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") > Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") > Reported-by: Aswin K <aswin@linux.ibm.com> > Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> > Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> LGTM! Reviewed-by: Wen Gu <guwen@linux.alibaba.com> > --- > net/smc/smc_ib.c | 8 ++------ > net/smc/smc_pnet.c | 4 +--- > 2 files changed, 3 insertions(+), 9 deletions(-) > > diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c > index 9297dc20bfe2..9c563cdbea90 100644 > --- a/net/smc/smc_ib.c > +++ b/net/smc/smc_ib.c > @@ -899,9 +899,7 @@ static void smc_copy_netdev_ifindex(struct smc_ib_device *smcibdev, int port) > struct ib_device *ibdev = smcibdev->ibdev; > struct net_device *ndev; > > - if (!ibdev->ops.get_netdev) > - return; > - ndev = ibdev->ops.get_netdev(ibdev, port + 1); > + ndev = ib_device_get_netdev(ibdev, port + 1); > if (ndev) { > smcibdev->ndev_ifidx[port] = ndev->ifindex; > dev_put(ndev); > @@ -921,9 +919,7 @@ void smc_ib_ndev_change(struct net_device *ndev, unsigned long event) > port_cnt = smcibdev->ibdev->phys_port_cnt; > for (i = 0; i < min_t(size_t, port_cnt, SMC_MAX_PORTS); i++) { > libdev = smcibdev->ibdev; > - if (!libdev->ops.get_netdev) > - continue; > - lndev = libdev->ops.get_netdev(libdev, i + 1); > + lndev = ib_device_get_netdev(libdev, i + 1); > dev_put(lndev); > if (lndev != ndev) > continue; > diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c > index 1dd362326c0a..8566937c8903 100644 > --- a/net/smc/smc_pnet.c > +++ b/net/smc/smc_pnet.c > @@ -1054,9 +1054,7 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, > for (i = 1; i <= SMC_MAX_PORTS; i++) { > if (!rdma_is_port_valid(ibdev->ibdev, i)) > continue; > - if (!ibdev->ibdev->ops.get_netdev) > - continue; > - ndev = ibdev->ibdev->ops.get_netdev(ibdev->ibdev, i); > + ndev = ib_device_get_netdev(ibdev->ibdev, i); > if (!ndev) > continue; > dev_put(ndev);
在 2024/10/25 9:23, Wenjia Zhang 写道: > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > alternative to get_netdev") introduced an API ib_device_get_netdev. > The SMC-R variant of the SMC protocol continued to use the old API > ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling Thanks a lot. Reviewed-by: Zhu Yanjun <yanjun.zhu@linux.dev> Because the commit 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removes the get_netdev callback from mlx5_ib_dev_common_roce_ops, in mlx4, get_netdev is still in mlx4_ib_dev_ops. So the following commit will follow mlx5 to remove get_netdev from mlx4 driver. From a59f2e01428640a332a51b8d910ec166704aa441 Mon Sep 17 00:00:00 2001 From: Zhu Yanjun <yanjun.zhu@linux.dev> Date: Sun, 27 Oct 2024 20:21:27 +0100 Subject: [PATCH 1/1] RDMA/mlx4: Use IB get_netdev functions and remove get_netdev callback In the commit 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the get_netdev callback from mlx5_ib_dev_common_roce_ops, in mlx4, get_netdev callback should also be removed. Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev> --- compile successfully only --- drivers/infiniband/hw/mlx4/main.c | 35 ------------------------------- 1 file changed, 35 deletions(-) diff --git a/drivers/infiniband/hw/mlx4/main.c b/drivers/infiniband/hw/mlx4/main.c index 529db874d67c..cf34d92de7b1 100644 --- a/drivers/infiniband/hw/mlx4/main.c +++ b/drivers/infiniband/hw/mlx4/main.c @@ -123,40 +123,6 @@ static int num_ib_ports(struct mlx4_dev *dev) return ib_ports; } -static struct net_device *mlx4_ib_get_netdev(struct ib_device *device, - u32 port_num) -{ - struct mlx4_ib_dev *ibdev = to_mdev(device); - struct net_device *dev, *ret = NULL; - - rcu_read_lock(); - for_each_netdev_rcu(&init_net, dev) { - if (dev->dev.parent != ibdev->ib_dev.dev.parent || - dev->dev_port + 1 != port_num) - continue; - - if (mlx4_is_bonded(ibdev->dev)) { - struct net_device *upper; - - upper = netdev_master_upper_dev_get_rcu(dev); - if (upper) { - struct net_device *active; - - active = bond_option_active_slave_get_rcu(netdev_priv(upper)); - if (active) - dev = active; - } - } - - dev_hold(dev); - ret = dev; - break; - } - - rcu_read_unlock(); - return ret; -} - static int mlx4_ib_update_gids_v1(struct gid_entry *gids, struct mlx4_ib_dev *ibdev, u32 port_num) @@ -2544,7 +2510,6 @@ static const struct ib_device_ops mlx4_ib_dev_ops = { .get_dev_fw_str = get_fw_ver_str, .get_dma_mr = mlx4_ib_get_dma_mr, .get_link_layer = mlx4_ib_port_link_layer, - .get_netdev = mlx4_ib_get_netdev, .get_port_immutable = mlx4_port_immutable, .map_mr_sg = mlx4_ib_map_mr_sg, .mmap = mlx4_ib_mmap, -- 2.34.1 Zhu Yanjun > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > device driver. Thus, using ib_device_set_netdev() now became mandatory. > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") > Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") > Reported-by: Aswin K <aswin@linux.ibm.com> > Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> > Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> > --- > net/smc/smc_ib.c | 8 ++------ > net/smc/smc_pnet.c | 4 +--- > 2 files changed, 3 insertions(+), 9 deletions(-) > > diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c > index 9297dc20bfe2..9c563cdbea90 100644 > --- a/net/smc/smc_ib.c > +++ b/net/smc/smc_ib.c > @@ -899,9 +899,7 @@ static void smc_copy_netdev_ifindex(struct smc_ib_device *smcibdev, int port) > struct ib_device *ibdev = smcibdev->ibdev; > struct net_device *ndev; > > - if (!ibdev->ops.get_netdev) > - return; > - ndev = ibdev->ops.get_netdev(ibdev, port + 1); > + ndev = ib_device_get_netdev(ibdev, port + 1); > if (ndev) { > smcibdev->ndev_ifidx[port] = ndev->ifindex; > dev_put(ndev); > @@ -921,9 +919,7 @@ void smc_ib_ndev_change(struct net_device *ndev, unsigned long event) > port_cnt = smcibdev->ibdev->phys_port_cnt; > for (i = 0; i < min_t(size_t, port_cnt, SMC_MAX_PORTS); i++) { > libdev = smcibdev->ibdev; > - if (!libdev->ops.get_netdev) > - continue; > - lndev = libdev->ops.get_netdev(libdev, i + 1); > + lndev = ib_device_get_netdev(libdev, i + 1); > dev_put(lndev); > if (lndev != ndev) > continue; > diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c > index 1dd362326c0a..8566937c8903 100644 > --- a/net/smc/smc_pnet.c > +++ b/net/smc/smc_pnet.c > @@ -1054,9 +1054,7 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, > for (i = 1; i <= SMC_MAX_PORTS; i++) { > if (!rdma_is_port_valid(ibdev->ibdev, i)) > continue; > - if (!ibdev->ibdev->ops.get_netdev) > - continue; > - ndev = ibdev->ibdev->ops.get_netdev(ibdev->ibdev, i); > + ndev = ib_device_get_netdev(ibdev->ibdev, i); > if (!ndev) > continue; > dev_put(ndev);
On Fri, Oct 25, 2024 at 09:23:55AM +0200, Wenjia Zhang wrote: > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > alternative to get_netdev") introduced an API ib_device_get_netdev. > The SMC-R variant of the SMC protocol continued to use the old API > ib_device_ops.get_netdev() to lookup netdev. I would say that calls to ibdev ops from ULPs was never been right thing to do. The ib_device_set_netdev() was introduced for the drivers. So the whole commit message is not accurate and better to be rewritten. > As this commit 8d159eb2117b > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > device driver. It is not a correct statement too. All modern drivers (for last 5 years) don't have that .get_netdev() ops, so it is not mlx5 specific, but another justification to say that SMC-R was doing it wrong. > Thus, using ib_device_set_netdev() now became mandatory. ib_device_set_netdev() is mandatory for the drivers, it is nothing to do with ULPs. > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). It is too late for me to do proper review for today, but I would say that it is worth to pay attention to multiple dev_put() calls in the functions around the ib_device_get_netdev(). > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") > Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") It is not related to this change Fixes line. > Reported-by: Aswin K <aswin@linux.ibm.com> > Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> > Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> > --- > net/smc/smc_ib.c | 8 ++------ > net/smc/smc_pnet.c | 4 +--- > 2 files changed, 3 insertions(+), 9 deletions(-) > > diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c > index 9297dc20bfe2..9c563cdbea90 100644 > --- a/net/smc/smc_ib.c > +++ b/net/smc/smc_ib.c > @@ -899,9 +899,7 @@ static void smc_copy_netdev_ifindex(struct smc_ib_device *smcibdev, int port) > struct ib_device *ibdev = smcibdev->ibdev; > struct net_device *ndev; > > - if (!ibdev->ops.get_netdev) > - return; > - ndev = ibdev->ops.get_netdev(ibdev, port + 1); > + ndev = ib_device_get_netdev(ibdev, port + 1); > if (ndev) { > smcibdev->ndev_ifidx[port] = ndev->ifindex; > dev_put(ndev); > @@ -921,9 +919,7 @@ void smc_ib_ndev_change(struct net_device *ndev, unsigned long event) > port_cnt = smcibdev->ibdev->phys_port_cnt; > for (i = 0; i < min_t(size_t, port_cnt, SMC_MAX_PORTS); i++) { > libdev = smcibdev->ibdev; > - if (!libdev->ops.get_netdev) > - continue; > - lndev = libdev->ops.get_netdev(libdev, i + 1); > + lndev = ib_device_get_netdev(libdev, i + 1); > dev_put(lndev); > if (lndev != ndev) > continue; > diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c > index 1dd362326c0a..8566937c8903 100644 > --- a/net/smc/smc_pnet.c > +++ b/net/smc/smc_pnet.c > @@ -1054,9 +1054,7 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, > for (i = 1; i <= SMC_MAX_PORTS; i++) { > if (!rdma_is_port_valid(ibdev->ibdev, i)) > continue; > - if (!ibdev->ibdev->ops.get_netdev) > - continue; > - ndev = ibdev->ibdev->ops.get_netdev(ibdev->ibdev, i); > + ndev = ib_device_get_netdev(ibdev->ibdev, i); > if (!ndev) > continue; > dev_put(ndev); > -- > 2.43.0 > >
On Sun, Oct 27, 2024 at 10:18:57PM +0200, Leon Romanovsky wrote: > On Fri, Oct 25, 2024 at 09:23:55AM +0200, Wenjia Zhang wrote: > > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > > alternative to get_netdev") introduced an API ib_device_get_netdev. > > The SMC-R variant of the SMC protocol continued to use the old API > > ib_device_ops.get_netdev() to lookup netdev. > > I would say that calls to ibdev ops from ULPs was never been right > thing to do. The ib_device_set_netdev() was introduced for the drivers. > > So the whole commit message is not accurate and better to be rewritten. > > > As this commit 8d159eb2117b > > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling > > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > > device driver. > > It is not a correct statement too. All modern drivers (for last 5 years) > don't have that .get_netdev() ops, so it is not mlx5 specific, but another > justification to say that SMC-R was doing it wrong. > > > Thus, using ib_device_set_netdev() now became mandatory. > > ib_device_set_netdev() is mandatory for the drivers, it is nothing to do > with ULPs. > > > > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > > It is too late for me to do proper review for today, but I would say > that it is worth to pay attention to multiple dev_put() calls in the > functions around the ib_device_get_netdev(). > > > > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") Honestly, this patch in Fixes line doesn't look right to me. It pokes inside of ib_device to get netdev index. For example call to smc_ib_ndev_change() will return completely unpredictable results, due to races. It is bad that RDMA ML wasn't even CCed back then, we would say NAK to this patch. https://lore.kernel.org/netdev/20201201192049.53517-6-kgraul@linux.ibm.com/ Thanks
On 10/25/24 3:23 PM, Wenjia Zhang wrote: > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > alternative to get_netdev") introduced an API ib_device_get_netdev. > The SMC-R variant of the SMC protocol continued to use the old API > ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > device driver. Thus, using ib_device_set_netdev() now became mandatory. > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") > Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") > Reported-by: Aswin K <aswin@linux.ibm.com> > Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> > Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> > --- > net/smc/smc_ib.c | 8 ++------ > net/smc/smc_pnet.c | 4 +--- > 2 files changed, 3 insertions(+), 9 deletions(-) > > diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c > index 9297dc20bfe2..9c563cdbea90 100644 > --- a/net/smc/smc_ib.c > +++ b/net/smc/smc_ib.c > @@ -899,9 +899,7 @@ static void smc_copy_netdev_ifindex(struct smc_ib_device *smcibdev, int port) > struct ib_device *ibdev = smcibdev->ibdev; > struct net_device *ndev; > > - if (!ibdev->ops.get_netdev) > - return; > - ndev = ibdev->ops.get_netdev(ibdev, port + 1); > + ndev = ib_device_get_netdev(ibdev, port + 1); > if (ndev) { > smcibdev->ndev_ifidx[port] = ndev->ifindex; > dev_put(ndev); > @@ -921,9 +919,7 @@ void smc_ib_ndev_change(struct net_device *ndev, unsigned long event) > port_cnt = smcibdev->ibdev->phys_port_cnt; > for (i = 0; i < min_t(size_t, port_cnt, SMC_MAX_PORTS); i++) { > libdev = smcibdev->ibdev; > - if (!libdev->ops.get_netdev) > - continue; > - lndev = libdev->ops.get_netdev(libdev, i + 1); > + lndev = ib_device_get_netdev(libdev, i + 1); > dev_put(lndev); > if (lndev != ndev) > continue; > diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c > index 1dd362326c0a..8566937c8903 100644 > --- a/net/smc/smc_pnet.c > +++ b/net/smc/smc_pnet.c > @@ -1054,9 +1054,7 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, > for (i = 1; i <= SMC_MAX_PORTS; i++) { > if (!rdma_is_port_valid(ibdev->ibdev, i)) > continue; > - if (!ibdev->ibdev->ops.get_netdev) > - continue; > - ndev = ibdev->ibdev->ops.get_netdev(ibdev->ibdev, i); > + ndev = ib_device_get_netdev(ibdev->ibdev, i); > if (!ndev) > continue; > dev_put(ndev); Reviewed-by: D. Wythe <alibuda@linux.alibaba.com>
On 10/25/24 09:23, Wenjia Zhang wrote: > Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an > alternative to get_netdev") introduced an API ib_device_get_netdev. > The SMC-R variant of the SMC protocol continued to use the old API > ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b > ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the > get_netdev callback from mlx5_ib_dev_common_roce_ops, calling > ib_device_ops.get_netdev didn't work any more at least by using a mlx5 > device driver. Thus, using ib_device_set_netdev() now became mandatory. > > Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). > > Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") > Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") > Reported-by: Aswin K <aswin@linux.ibm.com> > Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> > Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> Please adjust the commit message as per Leon suggestion. You can retain all the ack collected so far. Thanks, Paolo
diff --git a/net/smc/smc_ib.c b/net/smc/smc_ib.c index 9297dc20bfe2..9c563cdbea90 100644 --- a/net/smc/smc_ib.c +++ b/net/smc/smc_ib.c @@ -899,9 +899,7 @@ static void smc_copy_netdev_ifindex(struct smc_ib_device *smcibdev, int port) struct ib_device *ibdev = smcibdev->ibdev; struct net_device *ndev; - if (!ibdev->ops.get_netdev) - return; - ndev = ibdev->ops.get_netdev(ibdev, port + 1); + ndev = ib_device_get_netdev(ibdev, port + 1); if (ndev) { smcibdev->ndev_ifidx[port] = ndev->ifindex; dev_put(ndev); @@ -921,9 +919,7 @@ void smc_ib_ndev_change(struct net_device *ndev, unsigned long event) port_cnt = smcibdev->ibdev->phys_port_cnt; for (i = 0; i < min_t(size_t, port_cnt, SMC_MAX_PORTS); i++) { libdev = smcibdev->ibdev; - if (!libdev->ops.get_netdev) - continue; - lndev = libdev->ops.get_netdev(libdev, i + 1); + lndev = ib_device_get_netdev(libdev, i + 1); dev_put(lndev); if (lndev != ndev) continue; diff --git a/net/smc/smc_pnet.c b/net/smc/smc_pnet.c index 1dd362326c0a..8566937c8903 100644 --- a/net/smc/smc_pnet.c +++ b/net/smc/smc_pnet.c @@ -1054,9 +1054,7 @@ static void smc_pnet_find_rdma_dev(struct net_device *netdev, for (i = 1; i <= SMC_MAX_PORTS; i++) { if (!rdma_is_port_valid(ibdev->ibdev, i)) continue; - if (!ibdev->ibdev->ops.get_netdev) - continue; - ndev = ibdev->ibdev->ops.get_netdev(ibdev->ibdev, i); + ndev = ib_device_get_netdev(ibdev->ibdev, i); if (!ndev) continue; dev_put(ndev);
Commit c2261dd76b54 ("RDMA/device: Add ib_device_set_netdev() as an alternative to get_netdev") introduced an API ib_device_get_netdev. The SMC-R variant of the SMC protocol continued to use the old API ib_device_ops.get_netdev() to lookup netdev. As this commit 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") removed the get_netdev callback from mlx5_ib_dev_common_roce_ops, calling ib_device_ops.get_netdev didn't work any more at least by using a mlx5 device driver. Thus, using ib_device_set_netdev() now became mandatory. Replace ib_device_ops.get_netdev() with ib_device_get_netdev(). Fixes: 54903572c23c ("net/smc: allow pnetid-less configuration") Fixes: 8d159eb2117b ("RDMA/mlx5: Use IB set_netdev and get_netdev functions") Reported-by: Aswin K <aswin@linux.ibm.com> Reviewed-by: Gerd Bayer <gbayer@linux.ibm.com> Signed-off-by: Wenjia Zhang <wenjia@linux.ibm.com> --- net/smc/smc_ib.c | 8 ++------ net/smc/smc_pnet.c | 4 +--- 2 files changed, 3 insertions(+), 9 deletions(-)