2019-01-11 05:09:52

by Jia-Ju Bai

[permalink] [raw]
Subject: [BUG] net: huawei: hinic: a possible sleep-in-atomic-context bug in msg_to_mgmt_async

The driver may sleep in an interrupt handler.
The function call path (from bottom to top) in the directory
"drivers/net/ethernet/huawei/hinic/" in Linux-4.17 is:

[FUNC] down
hinic_hw_mgmt.c, 324: down in msg_to_mgmt_async
hinic_hw_mgmt.c, 408: msg_to_mgmt_async in mgmt_recv_msg_handler
hinic_hw_mgmt.c, 464:mgmt_recv_msg_handler in recv_mgmt_msg_handler
hinic_hw_mgmt.c, 484: recv_mgmt_msg_handler in mgmt_msg_aeqe_handler
hinic_hw_eqs.c, 264: [FUNC_PTR]mgmt_msg_aeqe_handler in aeq_irq_handler
hinic_hw_eqs.c, 355: aeq_irq_handler in eq_irq_handler
hinic_hw_eqs.c, 383: eq_irq_handler in ceq_tasklet

Note that [FUNC_PTR] means a function pointer call.

This bug is found by my static analysis tool (DSAC-2) and checked by my
manual code review.

I do not know how to correctly fix this bug, so I just report it.
A possible way may be to replace up() and down() with spin_lock() and
spin_unlock().


Best wishes,
Jia-Ju Bai