Re: [PATCH] RDMA/hns: prevent undefined behavior in hns_roce_set_user_sq_size()

Re: [PATCH] RDMA/hns: prevent undefined behavior in hns_roce_set_user_sq_size()

[Dan Carpenter]

This one still needs to be applied.

regards,
dan carpenter

On Sat, Jun 08, 2019 at 12:25:14PM +0300, Dan Carpenter wrote:
> The "ucmd->log_sq_bb_count" variable is a user controlled variable in
> the 0-255 range.  If we shift more than then number of bits in an int
> then it's undefined behavior (it shift wraps).  It turns out this
> doesn't cause any real issues at runtime, but it's good to check anyway.
> 
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> ---
>  drivers/infiniband/hw/hns/hns_roce_qp.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c
> index 8db2817a249e..006b3e7f4ed5 100644
> --- a/drivers/infiniband/hw/hns/hns_roce_qp.c
> +++ b/drivers/infiniband/hw/hns/hns_roce_qp.c
> @@ -342,7 +342,8 @@ static int hns_roce_set_user_sq_size(struct hns_roce_dev *hr_dev,
>  	u32 max_cnt;
>  
>  	/* Sanity check SQ size before proceeding */
> -	if ((u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes ||
> +	if (ucmd->log_sq_bb_count > 31 ||
> +	    (u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes ||
>  	     ucmd->log_sq_stride > max_sq_stride ||
>  	     ucmd->log_sq_stride < HNS_ROCE_IB_MIN_SQ_STRIDE) {
>  		dev_err(hr_dev->dev, "check SQ size error!\n");
> -- 
> 2.20.1

Re: [PATCH] RDMA/hns: prevent undefined behavior in hns_roce_set_user_sq_size()

[Jason Gunthorpe]

On Mon, Oct 07, 2019 at 03:18:22PM +0300, Dan Carpenter wrote:
> This one still needs to be applied.
> 
> regards,
> dan carpenter

Weird, it is marked changes requested in patchworks. An email must
have been lost??

I think I probably wanted to say that:

> >  	/* Sanity check SQ size before proceeding */
> > -	if ((u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes ||
> > +	if (ucmd->log_sq_bb_count > 31 ||
> > +	    (u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes
> > ||

Overall should probably be coded using check_shl_overflow(), as this
later shift:

	hr_qp->sq.wqe_cnt = 1 << ucmd->log_sq_bb_count;

Is storing it into an int and hardwring '31' because it magically
matches that lower shift is pretty fragile.

More like this?

diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c
index bec48f2593f405..6aa27d6ea3a600 100644
--- a/drivers/infiniband/hw/hns/hns_roce_qp.c
+++ b/drivers/infiniband/hw/hns/hns_roce_qp.c
@@ -332,9 +332,8 @@ static int check_sq_size_with_integrity(struct hns_roce_dev *hr_dev,
 	u8 max_sq_stride = ilog2(roundup_sq_stride);
 
 	/* Sanity check SQ size before proceeding */
-	if ((u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes ||
-	     ucmd->log_sq_stride > max_sq_stride ||
-	     ucmd->log_sq_stride < HNS_ROCE_IB_MIN_SQ_STRIDE) {
+	if (ucmd->log_sq_stride > max_sq_stride ||
+	    ucmd->log_sq_stride < HNS_ROCE_IB_MIN_SQ_STRIDE) {
 		ibdev_err(&hr_dev->ib_dev, "check SQ size error!\n");
 		return -EINVAL;
 	}
@@ -358,13 +357,16 @@ static int hns_roce_set_user_sq_size(struct hns_roce_dev *hr_dev,
 	u32 max_cnt;
 	int ret;
 
+	if (check_shl_overflow(1, ucmd->log_sq_bb_count, &hr_qp->sq.wqe_cnt) ||
+	    hr_qp->sq.wqe_cnt > hr_dev->caps.max_wqes)
+		return -EINVAL;
+
 	ret = check_sq_size_with_integrity(hr_dev, cap, ucmd);
 	if (ret) {
 		ibdev_err(&hr_dev->ib_dev, "Sanity check sq size failed\n");
 		return ret;
 	}
 
-	hr_qp->sq.wqe_cnt = 1 << ucmd->log_sq_bb_count;
 	hr_qp->sq.wqe_shift = ucmd->log_sq_stride;
 
 	max_cnt = max(1U, cap->max_send_sge);

Re: [PATCH] RDMA/hns: prevent undefined behavior in hns_roce_set_user_sq_size()

[Dan Carpenter]

On Thu, Oct 24, 2019 at 01:37:49PM -0300, Jason Gunthorpe wrote:
> On Mon, Oct 07, 2019 at 03:18:22PM +0300, Dan Carpenter wrote:
> > This one still needs to be applied.
> > 
> > regards,
> > dan carpenter
> 
> Weird, it is marked changes requested in patchworks. An email must
> have been lost??
>

Maybe you replied to a different thread?

> I think I probably wanted to say that:
> 
> > >  	/* Sanity check SQ size before proceeding */
> > > -	if ((u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes ||
> > > +	if (ucmd->log_sq_bb_count > 31 ||
> > > +	    (u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes
> > > ||
> 
> Overall should probably be coded using check_shl_overflow(), as this
> later shift:
> 
> 	hr_qp->sq.wqe_cnt = 1 << ucmd->log_sq_bb_count;
> 
> Is storing it into an int and hardwring '31' because it magically
> matches that lower shift is pretty fragile.
> 
> More like this?
> 

Yeah.  I like your patch.  I'd feel silly claiming authorship though.
I'm willing to, because it's nice, but probably you should just give me
Reported-by credit instead.

Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>

regards,
dan carpenter

Re: [PATCH] RDMA/hns: prevent undefined behavior in hns_roce_set_user_sq_size()

[Jason Gunthorpe]

On Thu, Oct 24, 2019 at 09:20:39PM +0300, Dan Carpenter wrote:
> On Thu, Oct 24, 2019 at 01:37:49PM -0300, Jason Gunthorpe wrote:
> > On Mon, Oct 07, 2019 at 03:18:22PM +0300, Dan Carpenter wrote:
> > > This one still needs to be applied.
> > > 
> > > regards,
> > > dan carpenter
> > 
> > Weird, it is marked changes requested in patchworks. An email must
> > have been lost??
> >
> 
> Maybe you replied to a different thread?
> 
> > I think I probably wanted to say that:
> > 
> > > >  	/* Sanity check SQ size before proceeding */
> > > > -	if ((u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes ||
> > > > +	if (ucmd->log_sq_bb_count > 31 ||
> > > > +	    (u32)(1 << ucmd->log_sq_bb_count) > hr_dev->caps.max_wqes
> > > > ||
> > 
> > Overall should probably be coded using check_shl_overflow(), as this
> > later shift:
> > 
> > 	hr_qp->sq.wqe_cnt = 1 << ucmd->log_sq_bb_count;
> > 
> > Is storing it into an int and hardwring '31' because it magically
> > matches that lower shift is pretty fragile.
> > 
> > More like this?
> > 
> 
> Yeah.  I like your patch.  I'd feel silly claiming authorship though.
> I'm willing to, because it's nice, but probably you should just give me
> Reported-by credit instead.
> 
> Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>

Okay applied to for-next

Jason