linux-rdma.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Leon Romanovsky <leon@kernel.org>
To: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Doug Ledford <dledford@redhat.com>,
	RDMA mailing list <linux-rdma@vger.kernel.org>,
	Daniel Jurgens <danielj@mellanox.com>,
	Erez Shitrit <erezsh@mellanox.com>,
	Maor Gottlieb <maorg@mellanox.com>,
	Michael Guralnik <michaelgur@mellanox.com>,
	Moni Shoua <monis@mellanox.com>,
	Parav Pandit <parav@mellanox.com>,
	Sean Hefty <sean.hefty@intel.com>,
	Valentine Fatiev <valentinef@mellanox.com>,
	Yishai Hadas <yishaih@mellanox.com>,
	Yonatan Cohen <yonatanc@mellanox.com>,
	Zhu Yanjun <yanjunz@mellanox.com>
Subject: Re: [PATCH rdma-rc 4/9] IB/ipoib: Fix double free of skb in case of multicast traffic in CM mode
Date: Thu, 13 Feb 2020 20:10:12 +0200	[thread overview]
Message-ID: <20200213181012.GG679970@unreal> (raw)
In-Reply-To: <20200213153743.GA19802@ziepe.ca>

On Thu, Feb 13, 2020 at 11:37:43AM -0400, Jason Gunthorpe wrote:
> On Wed, Feb 12, 2020 at 09:26:30AM +0200, Leon Romanovsky wrote:
>
> > diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h
> > index 2aa3457a30ce..c614cb87d09b 100644
> > --- a/drivers/infiniband/ulp/ipoib/ipoib.h
> > +++ b/drivers/infiniband/ulp/ipoib/ipoib.h
> > @@ -379,6 +379,7 @@ struct ipoib_dev_priv {
> >  	struct ipoib_tx_buf *tx_ring;
> >  	unsigned int	     tx_head;
> >  	unsigned int	     tx_tail;
> > +	atomic_t             tx_outstanding;
> >  	struct ib_sge	     tx_sge[MAX_SKB_FRAGS + 1];
> >  	struct ib_ud_wr      tx_wr;
> >  	struct ib_wc	     send_wc[MAX_SEND_CQE];
> > diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
> > index c59e00a0881f..db6aace83fe5 100644
> > --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c
> > +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c
> > @@ -756,7 +756,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
> >  		return;
> >  	}
> >
> > -	if ((priv->tx_head - priv->tx_tail) == ipoib_sendq_size - 1) {
> > +	if (atomic_read(&priv->tx_outstanding) == ipoib_sendq_size - 1) {
> >  		ipoib_dbg(priv, "TX ring 0x%x full, stopping kernel net queue\n",
> >  			  tx->qp->qp_num);
> >  		netif_stop_queue(dev);
> > @@ -786,7 +786,7 @@ void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_
> >  	} else {
> >  		netif_trans_update(dev);
> >  		++tx->tx_head;
> > -		++priv->tx_head;
> > +		atomic_inc(&priv->tx_outstanding);
> >  	}
>
> This use of an atomic is very weird, probably wrong.
>
> Why is it an atomic?  priv->tx_head wasn't an atomic, and every place
> touching tx_outstanding was also touching tx_head.
>
> I assume there is some hidden locking here? Or much more stuff is
> busted up.
>
> In that case, drop the atomic.
>
> However, if the atomic is needed (where/why?) then something has to
> be dealing with the races, and if the write side is fully locked then
> an atomic is the wrong choice, use READ_ONCE/WRITE_ONCE instead

I thought that atomic_t is appropriate here. Valentin wanted to
implement "window" and he needed to ensure that this window is counted
correctly while it doesn't need to be 100% accurate.

Thanks

>
> Jason

  reply	other threads:[~2020-02-13 18:10 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-02-12  7:26 [PATCH rdma-rc 0/9] Fixes for v5.6 Leon Romanovsky
2020-02-12  7:26 ` [PATCH rdma-rc 1/9] RDMA/ucma: Mask QPN to be 24 bits according to IBTA Leon Romanovsky
2020-02-12  7:26 ` [PATCH rdma-rc 2/9] RDMA/core: Fix protection fault in get_pkey_idx_qp_list Leon Romanovsky
2020-02-12  8:01   ` Leon Romanovsky
2020-02-12  8:06   ` Leon Romanovsky
2020-02-12  7:26 ` [PATCH rdma-rc 3/9] Revert "RDMA/cma: Simplify rdma_resolve_addr() error flow" Leon Romanovsky
2020-02-13 13:30   ` Jason Gunthorpe
2020-02-14  3:11     ` Parav Pandit
2020-02-14 14:08       ` Jason Gunthorpe
2020-02-14 14:48         ` Parav Pandit
2020-02-19 20:40   ` Jason Gunthorpe
2020-02-12  7:26 ` [PATCH rdma-rc 4/9] IB/ipoib: Fix double free of skb in case of multicast traffic in CM mode Leon Romanovsky
2020-02-13 15:37   ` Jason Gunthorpe
2020-02-13 18:10     ` Leon Romanovsky [this message]
2020-02-13 18:26       ` Jason Gunthorpe
2020-02-13 18:36         ` Leon Romanovsky
2020-02-13 19:09           ` Jason Gunthorpe
2020-02-12  7:26 ` [PATCH rdma-rc 5/9] RDMA/core: Add missing list deletion on freeing event queue Leon Romanovsky
2020-02-12  7:26 ` [PATCH rdma-rc 6/9] IB/mlx5: Fix async events cleanup flows Leon Romanovsky
2020-02-12  7:26 ` [PATCH rdma-rc 7/9] RDMA/rxe: Fix soft lockup problem due to using tasklets in softirq Leon Romanovsky
2020-02-12  7:26 ` [PATCH rdma-rc 8/9] IB/umad: Fix kernel crash while unloading ib_umad Leon Romanovsky
2020-02-13 14:28   ` Jason Gunthorpe
2020-02-13 18:03     ` Leon Romanovsky
2020-02-12  7:26 ` [PATCH rdma-rc 9/9] RDMA/mlx5: Prevent overflow in mmap offset calculations Leon Romanovsky
2020-02-13 18:03 ` [PATCH rdma-rc 0/9] Fixes for v5.6 Jason Gunthorpe

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200213181012.GG679970@unreal \
    --to=leon@kernel.org \
    --cc=danielj@mellanox.com \
    --cc=dledford@redhat.com \
    --cc=erezsh@mellanox.com \
    --cc=jgg@ziepe.ca \
    --cc=linux-rdma@vger.kernel.org \
    --cc=maorg@mellanox.com \
    --cc=michaelgur@mellanox.com \
    --cc=monis@mellanox.com \
    --cc=parav@mellanox.com \
    --cc=sean.hefty@intel.com \
    --cc=valentinef@mellanox.com \
    --cc=yanjunz@mellanox.com \
    --cc=yishaih@mellanox.com \
    --cc=yonatanc@mellanox.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).