In GID_PT mode with lpfc_ns_query=1, a race condition between iterating
the vport->fc_nodes list in lpfc_rscn_recovery_check and cleanup of an
ndlp can trigger a crash while processing the RSCN of another initiator
from the same zone.
During iteration of the vport->fc_nodes list, an ndlp is cleaned up and
released. lpfc_dequeue_node is called from lpfc_cleanup_node leading to
a bad ndlp dereference in lpfc_rscn_recovery_check.
Change list_for_each_entry to list_for_each_entry_safe in
lpfc_rscn_recovery_check to protect against removal of an initiator ndlp,
while walking the vport->fc_nodes list.
Co-developed-by: Justin Tee <justin.tee@xxxxxxxxxxxx>
Signed-off-by: Justin Tee <justin.tee@xxxxxxxxxxxx>
Signed-off-by: James Smart <jsmart2021@xxxxxxxxx>
---
drivers/scsi/lpfc/lpfc_els.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/scsi/lpfc/lpfc_els.c b/drivers/scsi/lpfc/lpfc_els.c
index 583a287b2d0c..3671e0f8e041 100644
--- a/drivers/scsi/lpfc/lpfc_els.c
+++ b/drivers/scsi/lpfc/lpfc_els.c
@@ -7708,10 +7708,10 @@ lpfc_rscn_payload_check(struct lpfc_vport *vport, uint32_t did)
static int
lpfc_rscn_recovery_check(struct lpfc_vport *vport)
{
- struct lpfc_nodelist *ndlp = NULL;
+ struct lpfc_nodelist *ndlp = NULL, *n;
/* Move all affected nodes by pending RSCNs to NPR state. */
- list_for_each_entry(ndlp, &vport->fc_nodes, nlp_listp) {
+ list_for_each_entry_safe(ndlp, n, &vport->fc_nodes, nlp_listp) {
if ((ndlp->nlp_state == NLP_STE_UNUSED_NODE) ||
!lpfc_rscn_payload_check(vport, ndlp->nlp_DID))
continue;
--
2.26.2
