sg_write()/bsg_write() is not fit to be called under KERNEL_DS

Both damn things interpret userland pointers embedded into the payload;
worse, they are actually traversing those.  Leaving aside the bad
API design, this is very much _not_ safe to call with KERNEL_DS.
Bail out early if that happens.

Cc: [email protected]
Signed-off-by: Al Viro <[email protected]>

Author: Al Viro

block/bsg.c

@@ -655,6 +655,9 @@ bsg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
 
 	dprintk("%s: write %Zd bytes\n", bd->name, count);
 
+	if (unlikely(segment_eq(get_fs(), KERNEL_DS)))
+		return -EINVAL;
+
 	bsg_set_block(bd, file);
 
 	bytes_written = 0;

drivers/scsi/sg.c

@@ -581,6 +581,9 @@ sg_write(struct file *filp, const char __user *buf, size_t count, loff_t * ppos)
 	sg_io_hdr_t *hp;
 	unsigned char cmnd[SG_MAX_CDB_SIZE];
 
+	if (unlikely(segment_eq(get_fs(), KERNEL_DS)))
+		return -EINVAL;
+
 	if ((!(sfp = (Sg_fd *) filp->private_data)) || (!(sdp = sfp->parentdp)))
 		return -ENXIO;
 	SCSI_LOG_TIMEOUT(3, sg_printk(KERN_INFO, sdp,