Enabling CONFIG_RCU_NOREF_DEBUG gives the following splat on the
first ingress IPv4 packet:
1 noref entities escaped an RCU section, nesting 259, leaked noref list
ffff8edcefb1dc00
------------[ cut here ]------------
WARNING: CPU: 0 PID: 0 at kernel/rcu/noref_debug.c:87
__rcu_check_noref+0xf8/0x100
Modules linked in: intel_rapl sb_edac x86_pkg_temp_thermal
intel_powerclamp coretemp kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul
ghash_clmulni_intel pcbc aesni_intel crypto_simd glue_helper cryptd mei_me
ipmi_ssif sg mei mxm_wmi iTCO_wdt iTCO_vendor_support dcdbas lpc_ich ipmi_si
pcspkr ipmi_devintf ipmi_msghandler shpchp wmi acpi_power_meter nfsd
auth_rpcgss nfs_acl lockd grace sunrpc ip_tables xfs libcrc32c sd_mod mgag200
drm_kms_helper syscopyarea sysfillrect sysimgblt fb_sys_fops ttm igb drm ixgbe
mdio ahci ptp crc32c_intel i2c_algo_bit libahci pps_core i2c_core libata dca
dm_mirror dm_region_hash dm_log dm_mod
CPU: 0 PID: 0 Comm: swapper/0 Not tainted 4.14.0-rc1.noref_3+ #1609
Hardware name: Dell Inc. PowerEdge R730/072T6D, BIOS 2.4.3 01/17/2017
task: ffffffffae019500 task.stack: ffffffffae000000
RIP: 0010:__rcu_check_noref+0x7b/0xd0
RSP: 0018:ffff900afbe03b30 EFLAGS: 00010246
RAX: 0000000000000034 RBX: ffff900afbfd2500 RCX: 0000000000000000
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000202
RBP: ffff900afbe03b48 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000001 R11: 00000000c388883e R12: 0000000000000103
R13: 000000001d24100a R14: ffff900af13e0000 R15: 0000000000000000
FS: 0000000000000000(0000) GS:ffff900afbe00000(0000)
knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00005589ce876398 CR3: 0000001ff52f9005 CR4: 00000000003606f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
<IRQ>
ip_route_input_noref+0xa6/0x150
? ip_route_input_noref+0x5/0x150
ip_rcv_finish+0x78/0x5e0
ip_rcv+0x2a7/0x540
? packet_rcv+0x52/0x450
__netif_receive_skb_core+0x3b9/0xe10
? netif_receive_skb_internal+0x40/0x390
__netif_receive_skb+0x18/0x60
netif_receive_skb_internal+0x8d/0x390
? netif_receive_skb_internal+0x40/0x390
napi_gro_receive+0x15c/0x1f0
igb_clean_rx_irq+0x36d/0x7f0 [igb]
igb_poll+0x303/0x780 [igb]
? save_stack_trace+0x1b/0x20
? __lock_acquire+0xcf2/0x11c0
? net_rx_action+0xb4/0x520
net_rx_action+0x27d/0x520
__do_softirq+0xd1/0x4f5
irq_exit+0xfb/0x110
do_IRQ+0x67/0x120
common_interrupt+0xa7/0xa7
</IRQ>
RIP: 0010:cpuidle_enter_state+0xd0/0x360
RSP: 0018:ffffffffae003df8 EFLAGS: 00000202 ORIG_RAX: ffffffffffffff79
RAX: ffffffffae019500 RBX: ffffd7d67c604400 RCX: 0000000000000000
RDX: ffffffffae019500 RSI: 0000000000000001 RDI: ffffffffae019500
RBP: ffffffffae003e30 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000001 R11: 0000000000000018 R12: 0000000000000004
R13: 0000000000000000 R14: ffffd7d67c604400 R15: 00000009c8eafd50
? cpuidle_enter_state+0xc9/0x360
cpuidle_enter+0x17/0x20
call_cpuidle+0x23/0x40
do_idle+0x183/0x200
cpu_startup_entry+0x73/0x80
rest_init+0xc3/0xd0
start_kernel+0x4f7/0x518
? set_init_arg+0x5a/0x5a
x86_64_start_reservations+0x24/0x26
x86_64_start_kernel+0x6f/0x72
secondary_startup_64+0xa5/0xa5
Code: f6 75 07 5b 41 5c 41 5d 5d c3 80 3d eb e4 ff 00 00 75 1a 44 89 e2
48 c7 c7 88 54 e7 ad 31 c0 c6 05 d6 e4 ff 00 01 e8 28 af fe ff <0f> ff 41 bd 07
00 00 00 48 8b 33 48 85 f6 74 06 44 39 63 10 74
The rcu protection in ip_route_input_noref() is unneeded and
misleading: the caller still needs to acquire and retain the rcu
lock until the skb - carrying a noref dst on successful return -
is either dropped or the relevant dst is forced to a ref-counted
version.
This change just drops the unneeded lock.
Signed-off-by: Paolo Abeni <pab...@redhat.com>
---
net/ipv4/route.c | 7 +------
1 file changed, 1 insertion(+), 6 deletions(-)
diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index 94d4cd2d5ea4..5a6ca1f16d3f 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -2069,14 +2069,9 @@ int ip_route_input_noref(struct sk_buff *skb, __be32
daddr, __be32 saddr,
u8 tos, struct net_device *dev)
{
struct fib_result res;
- int err;
tos &= IPTOS_RT_MASK;
- rcu_read_lock();
- err = ip_route_input_rcu(skb, daddr, saddr, tos, dev, &res);
- rcu_read_unlock();
-
- return err;
+ return ip_route_input_rcu(skb, daddr, saddr, tos, dev, &res);
}
EXPORT_SYMBOL(ip_route_input_noref);
--
2.13.6
