Re: sg_io HARDENED_USERCOPY_PAGESPAN trace

[Dave Jones]

On Fri, Dec 30, 2016 at 05:37:12AM -0800, Christoph Hellwig wrote:
 > On Thu, Dec 29, 2016 at 10:43:51AM -0500, Dave Jones wrote:
 > > On Wed, Dec 28, 2016 at 11:56:42PM -0800, Christoph Hellwig wrote:
 > >  > On Wed, Dec 28, 2016 at 04:40:16PM -0500, Dave Jones wrote:
 > >  > >  sg_io+0x113/0x470
 > >  > 
 > >  > Can you resolve that to a source line using a gdb?
 > > 
 > > It's the copy_from_user in an inlined copy of blk_fill_sghdr_rq.
 > 
 > That must be this line right at the beginning of blk_fill_sghdr_rq
 > 
 > 	if (copy_from_user(rq->cmd, hdr->cmdp, hdr->cmd_len))
 > 		return -EFAULT;
 > 
 > We're copying the SCSI CDB from the userspace pointer inside the hdr
 > we copied earlier into the request.
 > 
 > req->cmd is set to req->__cmd which is a u8 array with 16 members in
 > struct request by default, but if hdr->cmd_len is bigger than BLK_MAX_CDB
 > (16) we do a separate allocation for it in the caller:
 > 
 > 	if (hdr->cmd_len > BLK_MAX_CDB) {
 > 		rq->cmd = kzalloc(hdr->cmd_len, GFP_KERNEL);
 > 		if (!rq->cmd)
 > 			goto out_put_request;
 > 	}
 > 
 > so I'm not really sure what the problem here could be.

I threw this debug printk into the pagespan code to see what exactly
it was complaining about..

ptr:ffff88042614cff8 end:ffff88042614d003 n:c

so it was copying 12 bytes that spanned two pages.
>From my reading of the config option help text, this thing is
complaining that wasn't allocated with __GFP_COMP maybe ?

	Dave