diff options
| author | Florian Fainelli <f.fainelli@gmail.com> | 2019-02-02 05:20:52 +0100 |
|---|---|---|
| committer | David S. Miller <davem@davemloft.net> | 2019-02-03 21:11:12 +0100 |
| commit | 9fb20801dab46238706267896df1b3938d977129 (patch) | |
| tree | 9d20357c867ffca07822a10b5e69489c1b72ac7a /net/bridge | |
| parent | net: devlink: report cell size of shared buffers (diff) | |
| download | linux-9fb20801dab46238706267896df1b3938d977129.tar.xz linux-9fb20801dab46238706267896df1b3938d977129.zip | |
net: Fix ip_mc_{dec,inc}_group allocation context
After 4effd28c1245 ("bridge: join all-snoopers multicast address"), I
started seeing the following sleep in atomic warnings:
[ 26.763893] BUG: sleeping function called from invalid context at mm/slab.h:421
[ 26.771425] in_atomic(): 1, irqs_disabled(): 0, pid: 1658, name: sh
[ 26.777855] INFO: lockdep is turned off.
[ 26.781916] CPU: 0 PID: 1658 Comm: sh Not tainted 5.0.0-rc4 #20
[ 26.787943] Hardware name: BCM97278SV (DT)
[ 26.792118] Call trace:
[ 26.794645] dump_backtrace+0x0/0x170
[ 26.798391] show_stack+0x24/0x30
[ 26.801787] dump_stack+0xa4/0xe4
[ 26.805182] ___might_sleep+0x208/0x218
[ 26.809102] __might_sleep+0x78/0x88
[ 26.812762] kmem_cache_alloc_trace+0x64/0x28c
[ 26.817301] igmp_group_dropped+0x150/0x230
[ 26.821573] ip_mc_dec_group+0x1b0/0x1f8
[ 26.825585] br_ip4_multicast_leave_snoopers.isra.11+0x174/0x190
[ 26.831704] br_multicast_toggle+0x78/0xcc
[ 26.835887] store_bridge_parm+0xc4/0xfc
[ 26.839894] multicast_snooping_store+0x3c/0x4c
[ 26.844517] dev_attr_store+0x44/0x5c
[ 26.848262] sysfs_kf_write+0x50/0x68
[ 26.852006] kernfs_fop_write+0x14c/0x1b4
[ 26.856102] __vfs_write+0x60/0x190
[ 26.859668] vfs_write+0xc8/0x168
[ 26.863059] ksys_write+0x70/0xc8
[ 26.866449] __arm64_sys_write+0x24/0x30
[ 26.870458] el0_svc_common+0xa0/0x11c
[ 26.874291] el0_svc_handler+0x38/0x70
[ 26.878120] el0_svc+0x8/0xc
while toggling the bridge's multicast_snooping attribute dynamically.
Pass a gfp_t down to igmpv3_add_delrec(), introduce
__igmp_group_dropped() and introduce __ip_mc_dec_group() to take a gfp_t
argument.
Similarly introduce ____ip_mc_inc_group() and __ip_mc_inc_group() to
allow caller to specify gfp_t.
IPv6 part of the patch appears fine.
Fixes: 4effd28c1245 ("bridge: join all-snoopers multicast address")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/bridge')
| -rw-r--r-- | net/bridge/br_multicast.c | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/net/bridge/br_multicast.c b/net/bridge/br_multicast.c index 780757b7a82f..1fb885a33c66 100644 --- a/net/bridge/br_multicast.c +++ b/net/bridge/br_multicast.c @@ -1841,7 +1841,7 @@ static void br_ip4_multicast_join_snoopers(struct net_bridge *br) if (!in_dev) return; - ip_mc_inc_group(in_dev, htonl(INADDR_ALLSNOOPERS_GROUP)); + __ip_mc_inc_group(in_dev, htonl(INADDR_ALLSNOOPERS_GROUP), GFP_ATOMIC); in_dev_put(in_dev); } @@ -1872,7 +1872,7 @@ static void br_ip4_multicast_leave_snoopers(struct net_bridge *br) if (WARN_ON(!in_dev)) return; - ip_mc_dec_group(in_dev, htonl(INADDR_ALLSNOOPERS_GROUP)); + __ip_mc_dec_group(in_dev, htonl(INADDR_ALLSNOOPERS_GROUP), GFP_ATOMIC); in_dev_put(in_dev); } |