[PATCH net 0/7] drivers: net: Fix 64-bit statistics seqcount init

[PATCH net 0/7] drivers: net: Fix 64-bit statistics seqcount init

[Florian Fainelli]

Hi all,

This patch series fixes a bunch of drivers to have their 64-bit statistics
seqcount cookie be initialized correctly. Most of these drivers (except b44,
gtp) are probably used on 64-bit only hosts and so the lockdep splat might have
never been seen.

Florian Fainelli (7):
  b44: Initialize 64-bit stats seqcount
  i40e: Initialize 64-bit statistics TX ring seqcount
  ixgbe: Initialize 64-bit stats seqcounts
  nfp: Initialize RX and TX ring 64-bit stats seqcounts
  gtp: Initialize 64-bit per-cpu stats correctly
  netvsc: Initialize 64-bit stats seqcount
  ipvlan: Fix 64-bit statistics seqcount initialization

 drivers/net/ethernet/broadcom/b44.c                 | 1 +
 drivers/net/ethernet/intel/i40e/i40e_txrx.c         | 2 ++
 drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c   | 4 ++++
 drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 2 ++
 drivers/net/gtp.c                                   | 2 +-
 drivers/net/hyperv/netvsc.c                         | 2 ++
 drivers/net/ipvlan/ipvlan_main.c                    | 2 +-
 7 files changed, 13 insertions(+), 2 deletions(-)

-- 
2.9.3

[PATCH net 1/7] b44: Initialize 64-bit stats seqcount

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that.

Fixes: eeda8585522b ("b44: add 64 bit stats")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/ethernet/broadcom/b44.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/ethernet/broadcom/b44.c b/drivers/net/ethernet/broadcom/b44.c
index f411936b744c..a1125d10c825 100644
--- a/drivers/net/ethernet/broadcom/b44.c
+++ b/drivers/net/ethernet/broadcom/b44.c
@@ -2368,6 +2368,7 @@ static int b44_init_one(struct ssb_device *sdev,
 	bp->msg_enable = netif_msg_init(b44_debug, B44_DEF_MSG_ENABLE);
 
 	spin_lock_init(&bp->lock);
+	u64_stats_init(&bp->hw_stats.syncp);
 
 	bp->rx_pending = B44_DEF_RX_RING_PENDING;
 	bp->tx_pending = B44_DEF_TX_RING_PENDING;
-- 
2.9.3

[PATCH net 2/7] i40e: Initialize 64-bit statistics TX ring seqcount

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that.

Fixes: 980e9b118642 ("i40e: Add support for 64 bit netstats")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/ethernet/intel/i40e/i40e_txrx.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index b936febc315a..2194960d5855 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -1113,6 +1113,8 @@ int i40e_setup_tx_descriptors(struct i40e_ring *tx_ring)
 	if (!tx_ring->tx_bi)
 		goto err;
 
+	u64_stats_init(&tx_ring->syncp);
+
 	/* round up to nearest 4K */
 	tx_ring->size = tx_ring->count * sizeof(struct i40e_tx_desc);
 	/* add u32 for head writeback, align after this takes care of
-- 
2.9.3

[PATCH net 3/7] ixgbe: Initialize 64-bit stats seqcounts

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that.

Fixes: 4197aa7bb818 ("ixgbevf: provide 64 bit statistics")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
index 084c53582793..032f8ac06357 100644
--- a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
+++ b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
@@ -2988,6 +2988,8 @@ int ixgbevf_setup_tx_resources(struct ixgbevf_ring *tx_ring)
 	if (!tx_ring->tx_buffer_info)
 		goto err;
 
+	u64_stats_init(&tx_ring->syncp);
+
 	/* round up to nearest 4K */
 	tx_ring->size = tx_ring->count * sizeof(union ixgbe_adv_tx_desc);
 	tx_ring->size = ALIGN(tx_ring->size, 4096);
@@ -3046,6 +3048,8 @@ int ixgbevf_setup_rx_resources(struct ixgbevf_ring *rx_ring)
 	if (!rx_ring->rx_buffer_info)
 		goto err;
 
+	u64_stats_init(&rx_ring->syncp);
+
 	/* Round up to nearest 4K */
 	rx_ring->size = rx_ring->count * sizeof(union ixgbe_adv_rx_desc);
 	rx_ring->size = ALIGN(rx_ring->size, 4096);
-- 
2.9.3

[PATCH net 4/7] nfp: Initialize RX and TX ring 64-bit stats seqcounts

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that.

Fixes: 4c3523623dc0 ("net: add driver for Netronome NFP4000/NFP6000 NIC VFs")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
index 18750ff0ede6..4631ca8b8eb2 100644
--- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
+++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c
@@ -513,6 +513,7 @@ nfp_net_tx_ring_init(struct nfp_net_tx_ring *tx_ring,
 	tx_ring->idx = idx;
 	tx_ring->r_vec = r_vec;
 	tx_ring->is_xdp = is_xdp;
+	u64_stats_init(&tx_ring->r_vec->tx_sync);
 
 	tx_ring->qcidx = tx_ring->idx * nn->stride_tx;
 	tx_ring->qcp_q = nn->tx_bar + NFP_QCP_QUEUE_OFF(tx_ring->qcidx);
@@ -532,6 +533,7 @@ nfp_net_rx_ring_init(struct nfp_net_rx_ring *rx_ring,
 
 	rx_ring->idx = idx;
 	rx_ring->r_vec = r_vec;
+	u64_stats_init(&rx_ring->r_vec->rx_sync);
 
 	rx_ring->fl_qcidx = rx_ring->idx * nn->stride_rx;
 	rx_ring->qcp_fl = nn->rx_bar + NFP_QCP_QUEUE_OFF(rx_ring->fl_qcidx);
-- 
2.9.3

[PATCH net 5/7] gtp: Initialize 64-bit per-cpu stats correctly

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that by using netdev_alloc_pcpu_stats() instead of an open coded
allocation.

Fixes: 459aa660eb1d ("gtp: add initial driver for datapath of GPRS Tunneling Protocol (GTP-U)")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/gtp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/gtp.c b/drivers/net/gtp.c
index 1542e837fdfa..f38e32a7ec9c 100644
--- a/drivers/net/gtp.c
+++ b/drivers/net/gtp.c
@@ -364,7 +364,7 @@ static int gtp_dev_init(struct net_device *dev)
 
 	gtp->dev = dev;
 
-	dev->tstats = alloc_percpu(struct pcpu_sw_netstats);
+	dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
 	if (!dev->tstats)
 		return -ENOMEM;
 
-- 
2.9.3

[PATCH net 6/7] netvsc: Initialize 64-bit stats seqcount

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that. In commit 6c80f3fc2398 ("netvsc: report per-channel stats in
ethtool statistics") netdev_alloc_pcpu_stats() was removed in favor of
open-coding the 64-bits statistics, except that u64_stats_init() was
missed.

Fixes: 6c80f3fc2398 ("netvsc: report per-channel stats in ethtool statistics")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/hyperv/netvsc.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 0a9167dd72fb..96f90c75d1b7 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -1302,6 +1302,8 @@ int netvsc_device_add(struct hv_device *device,
 		struct netvsc_channel *nvchan = &net_device->chan_table[i];
 
 		nvchan->channel = device->channel;
+		u64_stats_init(&nvchan->tx_stats.syncp);
+		u64_stats_init(&nvchan->rx_stats.syncp);
 	}
 
 	/* Enable NAPI handler before init callbacks */
-- 
2.9.3

[PATCH net 6/7] netvsc: Initialize 64-bit stats seqcount

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that. In commit 6c80f3fc2398 ("netvsc: report per-channel stats in
ethtool statistics") netdev_alloc_pcpu_stats() was removed in favor of
open-coding the 64-bits statistics, except that u64_stats_init() was
missed.

Fixes: 6c80f3fc2398 ("netvsc: report per-channel stats in ethtool statistics")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/hyperv/netvsc.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
index 0a9167dd72fb..96f90c75d1b7 100644
--- a/drivers/net/hyperv/netvsc.c
+++ b/drivers/net/hyperv/netvsc.c
@@ -1302,6 +1302,8 @@ int netvsc_device_add(struct hv_device *device,
 		struct netvsc_channel *nvchan = &net_device->chan_table[i];
 
 		nvchan->channel = device->channel;
+		u64_stats_init(&nvchan->tx_stats.syncp);
+		u64_stats_init(&nvchan->rx_stats.syncp);
 	}
 
 	/* Enable NAPI handler before init callbacks */
-- 
2.9.3

[PATCH net 7/7] ipvlan: Fix 64-bit statistics seqcount initialization

[Florian Fainelli]

On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
lockdep splat indicating this seqcount is not correctly initialized, fix
that by using the proper helper function: netdev_alloc_pcpu_stats().

Fixes: 2ad7bf363841 ("ipvlan: Initial check-in of the IPVLAN driver.")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
---
 drivers/net/ipvlan/ipvlan_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ipvlan/ipvlan_main.c b/drivers/net/ipvlan/ipvlan_main.c
index f37e3c1fd4e7..8dab74a81303 100644
--- a/drivers/net/ipvlan/ipvlan_main.c
+++ b/drivers/net/ipvlan/ipvlan_main.c
@@ -192,7 +192,7 @@ static int ipvlan_init(struct net_device *dev)
 
 	netdev_lockdep_set_classes(dev);
 
-	ipvlan->pcpu_stats = alloc_percpu(struct ipvl_pcpu_stats);
+	ipvlan->pcpu_stats = netdev_alloc_pcpu_stats(struct ipvl_pcpu_stats);
 	if (!ipvlan->pcpu_stats)
 		return -ENOMEM;
 
-- 
2.9.3

Re: [PATCH net 4/7] nfp: Initialize RX and TX ring 64-bit stats seqcounts

[Simon Horman]

On Tue, Aug 01, 2017 at 12:11:09PM -0700, Florian Fainelli wrote:
> On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
> lockdep splat indicating this seqcount is not correctly initialized, fix
> that.
> 
> Fixes: 4c3523623dc0 ("net: add driver for Netronome NFP4000/NFP6000 NIC VFs")
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>

Reviewed-by: Simon Horman <simon.horman@netronome.com>

FYI, Jakub is on leave this week.

Re: [PATCH net 1/7] b44: Initialize 64-bit stats seqcount

[Michael Chan]

On Tue, Aug 1, 2017 at 12:11 PM, Florian Fainelli <f.fainelli@gmail.com> wrote:
> On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
> lockdep splat indicating this seqcount is not correctly initialized, fix
> that.
>
> Fixes: eeda8585522b ("b44: add 64 bit stats")
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>

Acked-by: Michael Chan <michael.chan@broadcom.com>

Thanks.

Re: [PATCH net 6/7] netvsc: Initialize 64-bit stats seqcount

[Stephen Hemminger]

On Tue,  1 Aug 2017 12:11:12 -0700
Florian Fainelli <f.fainelli@gmail.com> wrote:

> On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
> lockdep splat indicating this seqcount is not correctly initialized, fix
> that. In commit 6c80f3fc2398 ("netvsc: report per-channel stats in
> ethtool statistics") netdev_alloc_pcpu_stats() was removed in favor of
> open-coding the 64-bits statistics, except that u64_stats_init() was
> missed.
> 
> Fixes: 6c80f3fc2398 ("netvsc: report per-channel stats in ethtool statistics")
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
> ---
>  drivers/net/hyperv/netvsc.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/net/hyperv/netvsc.c b/drivers/net/hyperv/netvsc.c
> index 0a9167dd72fb..96f90c75d1b7 100644
> --- a/drivers/net/hyperv/netvsc.c
> +++ b/drivers/net/hyperv/netvsc.c
> @@ -1302,6 +1302,8 @@ int netvsc_device_add(struct hv_device *device,
>  		struct netvsc_channel *nvchan = &net_device->chan_table[i];
>  
>  		nvchan->channel = device->channel;
> +		u64_stats_init(&nvchan->tx_stats.syncp);
> +		u64_stats_init(&nvchan->rx_stats.syncp);
>  	}
>  
>  	/* Enable NAPI handler before init callbacks */


Looks good, thanks. 32 bit guests are still supported but rarely tested.

Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>

Re: [PATCH net 0/7] drivers: net: Fix 64-bit statistics seqcount init

[David Miller]

From: Florian Fainelli <f.fainelli@gmail.com>
Date: Tue,  1 Aug 2017 12:11:05 -0700

> This patch series fixes a bunch of drivers to have their 64-bit statistics
> seqcount cookie be initialized correctly. Most of these drivers (except b44,
> gtp) are probably used on 64-bit only hosts and so the lockdep splat might have
> never been seen.

Series applied, thanks.

Re: [PATCH net 5/7] gtp: Initialize 64-bit per-cpu stats correctly

[Pablo Neira Ayuso]

On Tue, Aug 01, 2017 at 12:11:10PM -0700, Florian Fainelli wrote:
> On 32-bit hosts and with CONFIG_DEBUG_LOCK_ALLOC we should be seeing a
> lockdep splat indicating this seqcount is not correctly initialized, fix
> that by using netdev_alloc_pcpu_stats() instead of an open coded
> allocation.
> 
> Fixes: 459aa660eb1d ("gtp: add initial driver for datapath of GPRS Tunneling Protocol (GTP-U)")
> Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>

Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>

Thanks!