net: caif: Fix a sleep-in-atomic bug in cfpkt_create_pfx
authorJia-Ju Bai <baijiaju1990@163.com>
Sat, 10 Jun 2017 08:49:39 +0000 (16:49 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 5 Jul 2017 12:37:14 +0000 (14:37 +0200)
commit7de53eed6fda3354ef8f2d3748d0ecfef88e0475
treeaf18e90a680104a325806598ea305a8644c7789d
parent030a77d2f904a3a2cdf9690744d06d075fa2f512
net: caif: Fix a sleep-in-atomic bug in cfpkt_create_pfx

[ Upstream commit f146e872eb12ebbe92d8e583b2637e0741440db3 ]

The kernel may sleep under a rcu read lock in cfpkt_create_pfx, and the
function call path is:
cfcnfg_linkup_rsp (acquire the lock by rcu_read_lock)
  cfctrl_linkdown_req
    cfpkt_create
      cfpkt_create_pfx
        alloc_skb(GFP_KERNEL) --> may sleep
cfserl_receive (acquire the lock by rcu_read_lock)
  cfpkt_split
    cfpkt_create_pfx
      alloc_skb(GFP_KERNEL) --> may sleep

There is "in_interrupt" in cfpkt_create_pfx to decide use "GFP_KERNEL" or
"GFP_ATOMIC". In this situation, "GFP_KERNEL" is used because the function
is called under a rcu read lock, instead in interrupt.

To fix it, only "GFP_ATOMIC" is used in cfpkt_create_pfx.

Signed-off-by: Jia-Ju Bai <baijiaju1990@163.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/caif/cfpkt_skbuff.c