[PATCH] scsi: lpfc: Fix a sleep-in-atomic-context bug in lpfc_unreg_rpi()

From: Jia-Ju Bai
Date: Fri Sep 14 2018 - 23:42:00 EST

The driver may sleep in an interrupt handler.
The function call path (from bottom to top) in Linux-4.17 is:

[FUNC] mempool_alloc(GFP_KERNEL)
drivers/scsi/lpfc/lpfc_hbadisc.c, 4738:
mempool_alloc in lpfc_unreg_rpi
drivers/scsi/lpfc/lpfc_els.c, 2825:
lpfc_unreg_rpi in lpfc_issue_els_logo
drivers/scsi/lpfc/lpfc_els.c, 9488:
lpfc_issue_els_logo in lpfc_sli_abts_recover_port
drivers/scsi/lpfc/lpfc_sli.c, 9806:
lpfc_sli_abts_recover_port in lpfc_sli_abts_err_handler
drivers/scsi/lpfc/lpfc_sli.c, 9914:
lpfc_sli_abts_err_handler in lpfc_sli_async_event_handler
drivers/scsi/lpfc/lpfc_sli.c, 2755:
[FUNC_PTR]lpfc_sli_async_event_handler in lpfc_sli_process_unsol_iocb
drivers/scsi/lpfc/lpfc_sli.c, 3359:
lpfc_sli_process_unsol_iocb in lpfc_sli_handle_fast_ring_event
drivers/scsi/lpfc/lpfc_sli.c, 12383:
lpfc_sli_handle_fast_ring_event in lpfc_sli_fp_intr_handler
(interrupt handler)

Note that [FUNC_PTR] means a function pointer call is used.

To fix this bug, GFP_KERNEL is replaced with GFP_ATOMIC.

This bug is found by my static analysis tool DSAC.

Signed-off-by: Jia-Ju Bai <baijiaju1990@xxxxxxxxx>
drivers/scsi/lpfc/lpfc_hbadisc.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/scsi/lpfc/lpfc_hbadisc.c b/drivers/scsi/lpfc/lpfc_hbadisc.c
index eb71877f12f8..b10122e13d02 100644
--- a/drivers/scsi/lpfc/lpfc_hbadisc.c
+++ b/drivers/scsi/lpfc/lpfc_hbadisc.c
@@ -4730,7 +4730,7 @@ lpfc_unreg_rpi(struct lpfc_vport *vport, struct lpfc_nodelist *ndlp)
"did x%x\n",
ndlp->nlp_rpi, ndlp->nlp_flag,
- mbox = mempool_alloc(phba->mbox_mem_pool, GFP_KERNEL);
+ mbox = mempool_alloc(phba->mbox_mem_pool, GFP_ATOMIC);
if (mbox) {
/* SLI4 ports require the physical rpi value. */
rpi = ndlp->nlp_rpi;