scsi: enclosure: Fix stale device oops with hot replug

Doing an add/remove/add on a SCSI device in an enclosure leads to an oops
caused by poisoned values in the enclosure device list pointers.  The
reason is because we are keeping the enclosure device across the enclosed
device add/remove/add but the current code is doing a
device_add/device_del/device_add on it.  This is the wrong thing to do in
sysfs, so fix it by not doing a device_del on the enclosure device simply
because of a hot remove of the drive in the slot.

[mkp: added missing email addresses]

Fixes: 43d8eb9cfd0a ("[SCSI] ses: add support for enclosure component hot removal")
Link: https://lore.kernel.org/r/1578532892.3852.10.camel@HansenPartnership.com
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Reported-by: Luo Jiaxing <luojiaxing@huawei.com>
Tested-by: John Garry <john.garry@huawei.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

diff --git a/drivers/misc/enclosure.c b/drivers/misc/enclosure.c
index 6d27ccfe068021127efd5452d7ff872859bcbe5f..3c2d405bc79b951f88a618289bfc8d47e16fac52 100644 (file)
--- a/drivers/misc/enclosure.c
+++ b/drivers/misc/enclosure.c
@@ -406,10 +406,9 @@ int enclosure_remove_device(struct enclosure_device *edev, struct device *dev)
                cdev = &edev->component[i];
                if (cdev->dev == dev) {
                        enclosure_remove_links(cdev);
-                       device_del(&cdev->cdev);
                        put_device(dev);
                        cdev->dev = NULL;
-                       return device_add(&cdev->cdev);
+                       return 0;
                }
        }
        return -ENODEV;