sg: fix dxferp in from_to case

This patch is in response to this report:
http://www.spinics.net/lists/linux-scsi/msg93456.html

One of the strange things that the original sg driver did was let the user provide both a data-out buffer (it followed the sg_header+cdb) _and_ specify a reply length greater than zero. What happened was that the user data-out buffer was copied into some kernel buffers and then the mid level was told a read type operation would take place with the data from the device overwriting the same kernel buffers. The user would then read those kernel buffers back into the user space.

From what I can tell, the above action was broken by a change in 2008 and syzkaller found that out recently.

ChangeLog:
make sure that a user space pointer is passed through when data follows the sg_header structure and command. Fix the abnormal case when a non-zero reply_len is also given.

index 5e82067..ae7d9bd 100644
@@ -652,7 +652,8 @@ sg_write(struct file *filp, const char __user *buf, size_t count, loff_t * ppos)
 	else
 		hp->dxfer_direction = (mxsize > 0) ? SG_DXFER_FROM_DEV : SG_DXFER_NONE;
 	hp->dxfer_len = mxsize;
-	if (hp->dxfer_direction == SG_DXFER_TO_DEV)
+	if ((hp->dxfer_direction == SG_DXFER_TO_DEV) ||
+	    (hp->dxfer_direction == SG_DXFER_TO_FROM_DEV))
 		hp->dxferp = (char __user *)buf + cmd_size;
 	else
 		hp->dxferp = NULL;

Doug> This patch is in response to this report:
Doug> http://www.spinics.net/lists/linux-scsi/msg93456.html

Doug> One of the strange things that the original sg driver did was let Doug> the user provide both a data-out buffer (it followed the Doug> sg_header+cdb) _and_ specify a reply length greater than Doug> zero. What happened was that the user data-out buffer was copied Doug> into some kernel buffers and then the mid level was told a read Doug> type operation would take place with the data from the device Doug> overwriting the same kernel buffers. The user would then read Doug> those kernel buffers back into the user space.

Doug> From what I can tell, the above action was broken by a change in Doug> 2008 and syzkaller found that out recently.

Doug> ChangeLog: make sure that a user space pointer is passed
Doug> through when data follows the sg_header structure and command. Doug> Fix the abnormal case when a non-zero reply_len is also given.

Somebody please review.

Martin K. Petersen	Oracle Linux Engineering

This looks correct to me. hp->dxferp used to be set unconditionally, but commit fad7f01e changed it to only be set in the SG_DXFER_TO_DEV case.

Doug> Make sure that a user space pointer is passed through when data Doug> follows the sg_header structure and command. Fix the abnormal Doug> case when a non-zero reply_len is also given.

Applied to 4.5/scsi-fixes.

Martin K. Petersen	Oracle Linux Engineering