All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] scsi: libsas: fix error handling in sas_register_phys()
@ 2022-11-04  6:46 Yang Yingliang
  0 siblings, 0 replies; only message in thread
From: Yang Yingliang @ 2022-11-04  6:46 UTC (permalink / raw)
  To: linux-scsi; +Cc: Yang Yingliang, James E.J. Bottomley, Martin K. Petersen

If sas_phy_alloc() returns error in sas_register_phys(), the phys that
have been added are not deleted, so the memory of them are leaked, also,
this leads the list of phy_attr_cont is not empty, it tiggers a BUG while
calling sas_release_transport() in hisi_sas_exit() when removing module.

kernel BUG at ./include/linux/transport_class.h:92!
CPU: 8 PID: 38014 Comm: rmmod Kdump: loaded Not tainted 6.1.0-rc1+ #176
Hardware name: Huawei TaiShan 2280 /BC11SPCD, BIOS 1.58 10/24/2018
pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
pc : sas_release_transport+0x78/0x84 [scsi_transport_sas]
lr : sas_release_transport+0x2c/0x84 [scsi_transport_sas]
Call trace:
 sas_release_transport+0x78/0x84 [scsi_transport_sas]
 hisi_sas_exit+0x1c/0x9a8 [hisi_sas_main]
 __arm64_sys_delete_module+0x19c/0x358

Fix this by deleting the phys that have been added if sas_phy_alloc()
returns error.

Besides, if sas_phy_add() fails in sas_register_phys(), the phy->dev
is not added to the klist_children of shost_gendev, so the phy can not
be delete in sas_remove_children(), the phy and name memory allocated
in sas_phy_alloc() are leaked.

Fix this by checking and handling return value of sas_phy_add() in
sas_register_phys(), call sas_phy_free() in the error path.

Fixes: 2908d778ab3e ("[SCSI] aic94xx: new driver")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
---
 drivers/scsi/libsas/sas_phy.c | 19 ++++++++++++++++---
 1 file changed, 16 insertions(+), 3 deletions(-)

diff --git a/drivers/scsi/libsas/sas_phy.c b/drivers/scsi/libsas/sas_phy.c
index a0d592d11dfb..84853576eb78 100644
--- a/drivers/scsi/libsas/sas_phy.c
+++ b/drivers/scsi/libsas/sas_phy.c
@@ -115,6 +115,7 @@ static void sas_phye_shutdown(struct work_struct *work)
 
 int sas_register_phys(struct sas_ha_struct *sas_ha)
 {
+	int ret;
 	int i;
 
 	/* Now register the phys. */
@@ -132,8 +133,10 @@ int sas_register_phys(struct sas_ha_struct *sas_ha)
 		phy->frame_rcvd_size = 0;
 
 		phy->phy = sas_phy_alloc(&sas_ha->core.shost->shost_gendev, i);
-		if (!phy->phy)
-			return -ENOMEM;
+		if (!phy->phy) {
+			ret = -ENOMEM;
+			goto err_out;
+		}
 
 		phy->phy->identify.initiator_port_protocols =
 			phy->iproto;
@@ -146,10 +149,20 @@ int sas_register_phys(struct sas_ha_struct *sas_ha)
 		phy->phy->maximum_linkrate = SAS_LINK_RATE_UNKNOWN;
 		phy->phy->negotiated_linkrate = SAS_LINK_RATE_UNKNOWN;
 
-		sas_phy_add(phy->phy);
+		ret = sas_phy_add(phy->phy);
+		if (ret) {
+			sas_phy_free(phy->phy);
+			goto err_out;
+		}
 	}
 
 	return 0;
+
+err_out:
+	while (i--)
+		sas_phy_delete(sas_ha->sas_phy[i]->phy);
+
+	return ret;
 }
 
 const work_func_t sas_phy_event_fns[PHY_NUM_EVENTS] = {
-- 
2.25.1


^ permalink raw reply related	[flat|nested] only message in thread

only message in thread, other threads:[~2022-11-04  6:48 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-11-04  6:46 [PATCH] scsi: libsas: fix error handling in sas_register_phys() Yang Yingliang

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.