linux-rdma.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Leon Romanovsky <leon@kernel.org>
To: Doug Ledford <dledford@redhat.com>, Jason Gunthorpe <jgg@mellanox.com>
Cc: linux-rdma@vger.kernel.org, Sean Hefty <sean.hefty@intel.com>
Subject: [PATCH rdma-next 01/15] RDMA/cm: Fix ordering of xa_alloc_cyclic() in ib_create_cm_id()
Date: Tue, 10 Mar 2020 11:25:31 +0200	[thread overview]
Message-ID: <20200310092545.251365-2-leon@kernel.org> (raw)
In-Reply-To: <20200310092545.251365-1-leon@kernel.org>

From: Jason Gunthorpe <jgg@mellanox.com>

xa_alloc_cyclic() is a SMP release to be paired with some later acquire
during xa_load() as part of cm_acquire_id().

As such, xa_alloc_cyclic() must be done after the cm_id is fully
initialized, in particular, it absolutely must be after the
refcount_set(), otherwise the refcount_inc() in cm_acquire_id() may not
see the set.

As there are several cases where a reader will be able to use the
id.local_id after cm_acquire_id in the IB_CM_IDLE state there needs to be
an unfortunate split into a NULL allocate and a finalizing xa_store.

Fixes: a977049dacde ("[PATCH] IB: Add the kernel CM implementation")
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
---
 drivers/infiniband/core/cm.c | 27 +++++++++++----------------
 1 file changed, 11 insertions(+), 16 deletions(-)

diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c
index c9bd9589bd5a..b1fccbf6ebd8 100644
--- a/drivers/infiniband/core/cm.c
+++ b/drivers/infiniband/core/cm.c
@@ -575,18 +575,6 @@ static int cm_init_av_by_path(struct sa_path_rec *path,
 	return 0;
 }

-static int cm_alloc_id(struct cm_id_private *cm_id_priv)
-{
-	int err;
-	u32 id;
-
-	err = xa_alloc_cyclic_irq(&cm.local_id_table, &id, cm_id_priv,
-			xa_limit_32b, &cm.local_id_next, GFP_KERNEL);
-
-	cm_id_priv->id.local_id = (__force __be32)id ^ cm.random_id_operand;
-	return err;
-}
-
 static u32 cm_local_id(__be32 local_id)
 {
 	return (__force u32) (local_id ^ cm.random_id_operand);
@@ -828,6 +816,7 @@ struct ib_cm_id *ib_create_cm_id(struct ib_device *device,
 				 void *context)
 {
 	struct cm_id_private *cm_id_priv;
+	u32 id;
 	int ret;

 	cm_id_priv = kzalloc(sizeof *cm_id_priv, GFP_KERNEL);
@@ -839,9 +828,6 @@ struct ib_cm_id *ib_create_cm_id(struct ib_device *device,
 	cm_id_priv->id.cm_handler = cm_handler;
 	cm_id_priv->id.context = context;
 	cm_id_priv->id.remote_cm_qpn = 1;
-	ret = cm_alloc_id(cm_id_priv);
-	if (ret)
-		goto error;

 	spin_lock_init(&cm_id_priv->lock);
 	init_completion(&cm_id_priv->comp);
@@ -850,11 +836,20 @@ struct ib_cm_id *ib_create_cm_id(struct ib_device *device,
 	INIT_LIST_HEAD(&cm_id_priv->altr_list);
 	atomic_set(&cm_id_priv->work_count, -1);
 	refcount_set(&cm_id_priv->refcount, 1);
+
+	ret = xa_alloc_cyclic_irq(&cm.local_id_table, &id, NULL, xa_limit_32b,
+				  &cm.local_id_next, GFP_KERNEL);
+	if (ret)
+		goto error;
+	cm_id_priv->id.local_id = (__force __be32)id ^ cm.random_id_operand;
+	xa_store_irq(&cm.local_id_table, cm_local_id(cm_id_priv->id.local_id),
+		     cm_id_priv, GFP_KERNEL);
+
 	return &cm_id_priv->id;

 error:
 	kfree(cm_id_priv);
-	return ERR_PTR(-ENOMEM);
+	return ERR_PTR(ret);
 }
 EXPORT_SYMBOL(ib_create_cm_id);

--
2.24.1


  reply	other threads:[~2020-03-10  9:25 UTC|newest]

Thread overview: 17+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-10  9:25 [PATCH rdma-next 00/15] Fix locking around cm_id.state in the ib_cm Leon Romanovsky
2020-03-10  9:25 ` Leon Romanovsky [this message]
2020-03-10  9:25 ` [PATCH rdma-next 02/15] RDMA/cm: Fix checking for allowed duplicate listens Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 03/15] RDMA/cm: Remove a race freeing timewait_info Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 04/15] RDMA/cm: Make the destroy_id flow more robust Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 05/15] RDMA/cm: Simplify establishing a listen cm_id Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 06/15] RDMA/cm: Read id.state under lock when doing pr_debug() Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 07/15] RDMA/cm: Make it clear that there is no concurrency in cm_sidr_req_handler() Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 08/15] RDMA/cm: Make it clearer how concurrency works in cm_req_handler() Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 09/15] RDMA/cm: Add missing locking around id.state in cm_dup_req_handler Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 10/15] RDMA/cm: Add some lockdep assertions for cm_id_priv->lock Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 11/15] RDMA/cm: Allow ib_send_cm_dreq() to be done under lock Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 12/15] RDMA/cm: Allow ib_send_cm_drep() " Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 13/15] RDMA/cm: Allow ib_send_cm_rej() " Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 14/15] RDMA/cm: Allow ib_send_cm_sidr_rep() " Leon Romanovsky
2020-03-10  9:25 ` [PATCH rdma-next 15/15] RDMA/cm: Make sure the cm_id is in the IB_CM_IDLE state in destroy Leon Romanovsky
2020-03-17 20:15 ` [PATCH rdma-next 00/15] Fix locking around cm_id.state in the ib_cm 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=20200310092545.251365-2-leon@kernel.org \
    --to=leon@kernel.org \
    --cc=dledford@redhat.com \
    --cc=jgg@mellanox.com \
    --cc=linux-rdma@vger.kernel.org \
    --cc=sean.hefty@intel.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).