[PATCH] scsi_remove_device locking

[Brian King <brking@us.ibm.com>]

[-- Attachment #1: Type: text/plain, Size: 64 bytes --]


-- 
Brian King
eServer Storage I/O
IBM Linux Technology Center

[-- Attachment #2: scsi_remove_device_locking.patch --]
[-- Type: text/plain, Size: 1467 bytes --]


The following patch fixes an oops I was seeing on a machine with
misconfigured scsi cables, but could feasibly happen in other paths.
The oops was occurring because scsi_remove_device was getting
called for a device before scsi_add_device had fully completed.
This resulted in sysfs_remove_link being called with a NULL dentry.

Signed-off-by: Brian King <brking@us.ibm.com>
---

 linux-2.6.7-bjking1/drivers/scsi/scsi_sysfs.c |    8 +++++++-
 1 files changed, 7 insertions(+), 1 deletion(-)

diff -puN drivers/scsi/scsi_sysfs.c~scsi_remove_device_locking drivers/scsi/scsi_sysfs.c
--- linux-2.6.7/drivers/scsi/scsi_sysfs.c~scsi_remove_device_locking	2004-07-02 11:34:25.000000000 -0500
+++ linux-2.6.7-bjking1/drivers/scsi/scsi_sysfs.c	2004-07-02 11:37:00.000000000 -0500
@@ -524,8 +524,13 @@ int scsi_sysfs_add_sdev(struct scsi_devi
  **/
 void scsi_remove_device(struct scsi_device *sdev)
 {
-	if (scsi_device_set_state(sdev, SDEV_CANCEL) != 0)
+	struct Scsi_Host *shost = sdev->host;
+	down(&shost->scan_mutex);
+
+	if (scsi_device_set_state(sdev, SDEV_CANCEL) != 0) {
+		up(&shost->scan_mutex);
 		return;
+	}
 
 	class_device_unregister(&sdev->sdev_classdev);
 	if (sdev->transport_classdev.class)
@@ -537,6 +542,7 @@ void scsi_remove_device(struct scsi_devi
 	if (sdev->host->transportt->cleanup)
 		sdev->host->transportt->cleanup(sdev);
 	put_device(&sdev->sdev_gendev);
+	up(&shost->scan_mutex);
 }
 
 int scsi_register_driver(struct device_driver *drv)
_