[SCSI] bsg: correct fault if queue object removed while dev_t open

This patch corrects an issue in bsg that results in a general protection
fault if an LLD is removed while an application is using an open file
handle to a bsg device, and the application issues an ioctl. The fault
occurs because the class_dev is NULL, having been cleared in
bsg_unregister_queue() when the driver was removed.  With this
patch, a check is made for the class_dev, and the application
will receive ENXIO if the related object is gone.

Signed-off-by: Carl Lajeunesse <carl.lajeunesse@emulex.com>
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>

diff --git a/block/bsg.c b/block/bsg.c
index f20d6a7..0c8b64a 100644 (file)
--- a/block/bsg.c
+++ b/block/bsg.c
@@ -250,6 +250,14 @@ bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr, fmode_t has_write_perm,
        int ret, rw;
        unsigned int dxfer_len;
        void *dxferp = NULL;
+       struct bsg_class_device *bcd = &q->bsg_dev;
+
+       /* if the LLD has been removed then the bsg_unregister_queue will
+        * eventually be called and the class_dev was freed, so we can no
+        * longer use this request_queue. Return no such address.
+        */
+       if (!bcd->class_dev)
+               return ERR_PTR(-ENXIO);
 
        dprintk("map hdr %llx/%u %llx/%u\n", (unsigned long long) hdr->dout_xferp,
                hdr->dout_xfer_len, (unsigned long long) hdr->din_xferp,