net: nfc: nci: fix a possible sleep-in-atomic-context bug in nci_uart_tty_receive()
authorJia-Ju Bai <baijiaju1990@gmail.com>
Wed, 18 Dec 2019 09:21:55 +0000 (17:21 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 31 Dec 2019 15:41:23 +0000 (16:41 +0100)
commit65084e27a170642f4bbd39b5677d425d332b72e3
tree5e908a20e8a98c0d30dea4df7df33c92ae578a8a
parente32473a7f768070f42286ecf8199a6287af936c3
net: nfc: nci: fix a possible sleep-in-atomic-context bug in nci_uart_tty_receive()

[ Upstream commit b7ac893652cafadcf669f78452329727e4e255cc ]

The kernel may sleep while holding a spinlock.
The function call path (from bottom to top) in Linux 4.19 is:

net/nfc/nci/uart.c, 349:
nci_skb_alloc in nci_uart_default_recv_buf
net/nfc/nci/uart.c, 255:
(FUNC_PTR)nci_uart_default_recv_buf in nci_uart_tty_receive
net/nfc/nci/uart.c, 254:
spin_lock in nci_uart_tty_receive

nci_skb_alloc(GFP_KERNEL) can sleep at runtime.
(FUNC_PTR) means a function pointer is called.

To fix this bug, GFP_KERNEL is replaced with GFP_ATOMIC for
nci_skb_alloc().

This bug is found by a static analysis tool STCheck written by myself.

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