linux-stable/net/dsa/master.c
Marc Zyngier c8101f7729 net: dsa: Fix lockdep false positive splat
Creating a macvtap on a DSA-backed interface results in the following
splat when lockdep is enabled:

[   19.638080] IPv6: ADDRCONF(NETDEV_CHANGE): lan0: link becomes ready
[   23.041198] device lan0 entered promiscuous mode
[   23.043445] device eth0 entered promiscuous mode
[   23.049255]
[   23.049557] ============================================
[   23.055021] WARNING: possible recursive locking detected
[   23.060490] 5.0.0-rc3-00013-g56c857a1b8d3 #118 Not tainted
[   23.066132] --------------------------------------------
[   23.071598] ip/2861 is trying to acquire lock:
[   23.076171] 00000000f61990cb (_xmit_ETHER){+...}, at: dev_set_rx_mode+0x1c/0x38
[   23.083693]
[   23.083693] but task is already holding lock:
[   23.089696] 00000000ecf0c3b4 (_xmit_ETHER){+...}, at: dev_uc_add+0x24/0x70
[   23.096774]
[   23.096774] other info that might help us debug this:
[   23.103494]  Possible unsafe locking scenario:
[   23.103494]
[   23.109584]        CPU0
[   23.112093]        ----
[   23.114601]   lock(_xmit_ETHER);
[   23.117917]   lock(_xmit_ETHER);
[   23.121233]
[   23.121233]  *** DEADLOCK ***
[   23.121233]
[   23.127325]  May be due to missing lock nesting notation
[   23.127325]
[   23.134315] 2 locks held by ip/2861:
[   23.137987]  #0: 000000003b766c72 (rtnl_mutex){+.+.}, at: rtnetlink_rcv_msg+0x338/0x4e0
[   23.146231]  #1: 00000000ecf0c3b4 (_xmit_ETHER){+...}, at: dev_uc_add+0x24/0x70
[   23.153757]
[   23.153757] stack backtrace:
[   23.158243] CPU: 0 PID: 2861 Comm: ip Not tainted 5.0.0-rc3-00013-g56c857a1b8d3 #118
[   23.166212] Hardware name: Globalscale Marvell ESPRESSOBin Board (DT)
[   23.172843] Call trace:
[   23.175358]  dump_backtrace+0x0/0x188
[   23.179116]  show_stack+0x14/0x20
[   23.182524]  dump_stack+0xb4/0xec
[   23.185928]  __lock_acquire+0x123c/0x1860
[   23.190048]  lock_acquire+0xc8/0x248
[   23.193724]  _raw_spin_lock_bh+0x40/0x58
[   23.197755]  dev_set_rx_mode+0x1c/0x38
[   23.201607]  dev_set_promiscuity+0x3c/0x50
[   23.205820]  dsa_slave_change_rx_flags+0x5c/0x70
[   23.210567]  __dev_set_promiscuity+0x148/0x1e0
[   23.215136]  __dev_set_rx_mode+0x74/0x98
[   23.219167]  dev_uc_add+0x54/0x70
[   23.222575]  macvlan_open+0x170/0x1d0
[   23.226336]  __dev_open+0xe0/0x160
[   23.229830]  __dev_change_flags+0x16c/0x1b8
[   23.234132]  dev_change_flags+0x20/0x60
[   23.238074]  do_setlink+0x2d0/0xc50
[   23.241658]  __rtnl_newlink+0x5f8/0x6e8
[   23.245601]  rtnl_newlink+0x50/0x78
[   23.249184]  rtnetlink_rcv_msg+0x360/0x4e0
[   23.253397]  netlink_rcv_skb+0xe8/0x130
[   23.257338]  rtnetlink_rcv+0x14/0x20
[   23.261012]  netlink_unicast+0x190/0x210
[   23.265043]  netlink_sendmsg+0x288/0x350
[   23.269075]  sock_sendmsg+0x18/0x30
[   23.272659]  ___sys_sendmsg+0x29c/0x2c8
[   23.276602]  __sys_sendmsg+0x60/0xb8
[   23.280276]  __arm64_sys_sendmsg+0x1c/0x28
[   23.284488]  el0_svc_common+0xd8/0x138
[   23.288340]  el0_svc_handler+0x24/0x80
[   23.292192]  el0_svc+0x8/0xc

This looks fairly harmless (no actual deadlock occurs), and is
fixed in a similar way to c6894dec8e ("bridge: fix lockdep
addr_list_lock false positive splat") by putting the addr_list_lock
in its own lockdep class.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Reviewed-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2019-02-04 18:30:43 -08:00

250 lines
6.5 KiB
C

/*
* Handling of a master device, switching frames via its switch fabric CPU port
*
* Copyright (c) 2017 Savoir-faire Linux Inc.
* Vivien Didelot <vivien.didelot@savoirfairelinux.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*/
#include "dsa_priv.h"
static void dsa_master_get_ethtool_stats(struct net_device *dev,
struct ethtool_stats *stats,
uint64_t *data)
{
struct dsa_port *cpu_dp = dev->dsa_ptr;
const struct ethtool_ops *ops = cpu_dp->orig_ethtool_ops;
struct dsa_switch *ds = cpu_dp->ds;
int port = cpu_dp->index;
int count = 0;
if (ops->get_sset_count && ops->get_ethtool_stats) {
count = ops->get_sset_count(dev, ETH_SS_STATS);
ops->get_ethtool_stats(dev, stats, data);
}
if (ds->ops->get_ethtool_stats)
ds->ops->get_ethtool_stats(ds, port, data + count);
}
static void dsa_master_get_ethtool_phy_stats(struct net_device *dev,
struct ethtool_stats *stats,
uint64_t *data)
{
struct dsa_port *cpu_dp = dev->dsa_ptr;
const struct ethtool_ops *ops = cpu_dp->orig_ethtool_ops;
struct dsa_switch *ds = cpu_dp->ds;
int port = cpu_dp->index;
int count = 0;
if (dev->phydev && !ops->get_ethtool_phy_stats) {
count = phy_ethtool_get_sset_count(dev->phydev);
if (count >= 0)
phy_ethtool_get_stats(dev->phydev, stats, data);
} else if (ops->get_sset_count && ops->get_ethtool_phy_stats) {
count = ops->get_sset_count(dev, ETH_SS_PHY_STATS);
ops->get_ethtool_phy_stats(dev, stats, data);
}
if (count < 0)
count = 0;
if (ds->ops->get_ethtool_phy_stats)
ds->ops->get_ethtool_phy_stats(ds, port, data + count);
}
static int dsa_master_get_sset_count(struct net_device *dev, int sset)
{
struct dsa_port *cpu_dp = dev->dsa_ptr;
const struct ethtool_ops *ops = cpu_dp->orig_ethtool_ops;
struct dsa_switch *ds = cpu_dp->ds;
int count = 0;
if (sset == ETH_SS_PHY_STATS && dev->phydev &&
!ops->get_ethtool_phy_stats)
count = phy_ethtool_get_sset_count(dev->phydev);
else if (ops->get_sset_count)
count = ops->get_sset_count(dev, sset);
if (count < 0)
count = 0;
if (ds->ops->get_sset_count)
count += ds->ops->get_sset_count(ds, cpu_dp->index, sset);
return count;
}
static void dsa_master_get_strings(struct net_device *dev, uint32_t stringset,
uint8_t *data)
{
struct dsa_port *cpu_dp = dev->dsa_ptr;
const struct ethtool_ops *ops = cpu_dp->orig_ethtool_ops;
struct dsa_switch *ds = cpu_dp->ds;
int port = cpu_dp->index;
int len = ETH_GSTRING_LEN;
int mcount = 0, count;
unsigned int i;
uint8_t pfx[4];
uint8_t *ndata;
snprintf(pfx, sizeof(pfx), "p%.2d", port);
/* We do not want to be NULL-terminated, since this is a prefix */
pfx[sizeof(pfx) - 1] = '_';
if (stringset == ETH_SS_PHY_STATS && dev->phydev &&
!ops->get_ethtool_phy_stats) {
mcount = phy_ethtool_get_sset_count(dev->phydev);
if (mcount < 0)
mcount = 0;
else
phy_ethtool_get_strings(dev->phydev, data);
} else if (ops->get_sset_count && ops->get_strings) {
mcount = ops->get_sset_count(dev, stringset);
if (mcount < 0)
mcount = 0;
ops->get_strings(dev, stringset, data);
}
if (ds->ops->get_strings) {
ndata = data + mcount * len;
/* This function copies ETH_GSTRINGS_LEN bytes, we will mangle
* the output after to prepend our CPU port prefix we
* constructed earlier
*/
ds->ops->get_strings(ds, port, stringset, ndata);
count = ds->ops->get_sset_count(ds, port, stringset);
for (i = 0; i < count; i++) {
memmove(ndata + (i * len + sizeof(pfx)),
ndata + i * len, len - sizeof(pfx));
memcpy(ndata + i * len, pfx, sizeof(pfx));
}
}
}
static int dsa_master_ethtool_setup(struct net_device *dev)
{
struct dsa_port *cpu_dp = dev->dsa_ptr;
struct dsa_switch *ds = cpu_dp->ds;
struct ethtool_ops *ops;
ops = devm_kzalloc(ds->dev, sizeof(*ops), GFP_KERNEL);
if (!ops)
return -ENOMEM;
cpu_dp->orig_ethtool_ops = dev->ethtool_ops;
if (cpu_dp->orig_ethtool_ops)
memcpy(ops, cpu_dp->orig_ethtool_ops, sizeof(*ops));
ops->get_sset_count = dsa_master_get_sset_count;
ops->get_ethtool_stats = dsa_master_get_ethtool_stats;
ops->get_strings = dsa_master_get_strings;
ops->get_ethtool_phy_stats = dsa_master_get_ethtool_phy_stats;
dev->ethtool_ops = ops;
return 0;
}
static void dsa_master_ethtool_teardown(struct net_device *dev)
{
struct dsa_port *cpu_dp = dev->dsa_ptr;
dev->ethtool_ops = cpu_dp->orig_ethtool_ops;
cpu_dp->orig_ethtool_ops = NULL;
}
static ssize_t tagging_show(struct device *d, struct device_attribute *attr,
char *buf)
{
struct net_device *dev = to_net_dev(d);
struct dsa_port *cpu_dp = dev->dsa_ptr;
return sprintf(buf, "%s\n",
dsa_tag_protocol_to_str(cpu_dp->tag_ops));
}
static DEVICE_ATTR_RO(tagging);
static struct attribute *dsa_slave_attrs[] = {
&dev_attr_tagging.attr,
NULL
};
static const struct attribute_group dsa_group = {
.name = "dsa",
.attrs = dsa_slave_attrs,
};
static void dsa_master_set_mtu(struct net_device *dev, struct dsa_port *cpu_dp)
{
unsigned int mtu = ETH_DATA_LEN + cpu_dp->tag_ops->overhead;
int err;
rtnl_lock();
if (mtu <= dev->max_mtu) {
err = dev_set_mtu(dev, mtu);
if (err)
netdev_dbg(dev, "Unable to set MTU to include for DSA overheads\n");
}
rtnl_unlock();
}
static void dsa_master_reset_mtu(struct net_device *dev)
{
int err;
rtnl_lock();
err = dev_set_mtu(dev, ETH_DATA_LEN);
if (err)
netdev_dbg(dev,
"Unable to reset MTU to exclude DSA overheads\n");
rtnl_unlock();
}
static struct lock_class_key dsa_master_addr_list_lock_key;
int dsa_master_setup(struct net_device *dev, struct dsa_port *cpu_dp)
{
int ret;
dsa_master_set_mtu(dev, cpu_dp);
/* If we use a tagging format that doesn't have an ethertype
* field, make sure that all packets from this point on get
* sent to the tag format's receive function.
*/
wmb();
dev->dsa_ptr = cpu_dp;
lockdep_set_class(&dev->addr_list_lock,
&dsa_master_addr_list_lock_key);
ret = dsa_master_ethtool_setup(dev);
if (ret)
return ret;
ret = sysfs_create_group(&dev->dev.kobj, &dsa_group);
if (ret)
dsa_master_ethtool_teardown(dev);
return ret;
}
void dsa_master_teardown(struct net_device *dev)
{
sysfs_remove_group(&dev->dev.kobj, &dsa_group);
dsa_master_ethtool_teardown(dev);
dsa_master_reset_mtu(dev);
dev->dsa_ptr = NULL;
/* If we used a tagging format that doesn't have an ethertype
* field, make sure that all packets from this point get sent
* without the tag and go through the regular receive path.
*/
wmb();
}