lkml.org 
[lkml]   [2020]   [Dec]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH 5.10 207/717] RDMA/cma: Fix deadlock on &lock in rdma_cma_listen_on_all() error unwind
Date
From: Jason Gunthorpe <jgg@nvidia.com>

[ Upstream commit dd37d2f59eb839d51b988f6668ce5f0d533b23fd ]

rdma_detroy_id() cannot be called under &lock - we must instead keep the
error'd ID around until &lock can be released, then destroy it.

This is complicated by the usual way listen IDs are destroyed through
cma_process_remove() which can run at any time and will asynchronously
destroy the same ID.

Remove the ID from visiblity of cma_process_remove() before going down the
destroy path outside the locking.

Fixes: c80a0c52d85c ("RDMA/cma: Add missing error handling of listen_id")
Link: https://lore.kernel.org/r/20201118133756.GK244516@ziepe.ca
Reported-by: syzbot+1bc48bf7f78253f664a9@syzkaller.appspotmail.com
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
drivers/infiniband/core/cma.c | 25 ++++++++++++++++++-------
1 file changed, 18 insertions(+), 7 deletions(-)

diff --git a/drivers/infiniband/core/cma.c b/drivers/infiniband/core/cma.c
index 4585f654f8836..c06c87a4dc5e7 100644
--- a/drivers/infiniband/core/cma.c
+++ b/drivers/infiniband/core/cma.c
@@ -2496,7 +2496,8 @@ static int cma_listen_handler(struct rdma_cm_id *id,
}

static int cma_listen_on_dev(struct rdma_id_private *id_priv,
- struct cma_device *cma_dev)
+ struct cma_device *cma_dev,
+ struct rdma_id_private **to_destroy)
{
struct rdma_id_private *dev_id_priv;
struct net *net = id_priv->id.route.addr.dev_addr.net;
@@ -2504,6 +2505,7 @@ static int cma_listen_on_dev(struct rdma_id_private *id_priv,

lockdep_assert_held(&lock);

+ *to_destroy = NULL;
if (cma_family(id_priv) == AF_IB && !rdma_cap_ib_cm(cma_dev->device, 1))
return 0;

@@ -2518,7 +2520,6 @@ static int cma_listen_on_dev(struct rdma_id_private *id_priv,
rdma_addr_size(cma_src_addr(id_priv)));

_cma_attach_to_dev(dev_id_priv, cma_dev);
- list_add_tail(&dev_id_priv->listen_list, &id_priv->listen_list);
cma_id_get(id_priv);
dev_id_priv->internal_id = 1;
dev_id_priv->afonly = id_priv->afonly;
@@ -2528,25 +2529,31 @@ static int cma_listen_on_dev(struct rdma_id_private *id_priv,
ret = rdma_listen(&dev_id_priv->id, id_priv->backlog);
if (ret)
goto err_listen;
+ list_add_tail(&dev_id_priv->listen_list, &id_priv->listen_list);
return 0;
err_listen:
- list_del(&id_priv->listen_list);
+ /* Caller must destroy this after releasing lock */
+ *to_destroy = dev_id_priv;
dev_warn(&cma_dev->device->dev, "RDMA CMA: %s, error %d\n", __func__, ret);
- rdma_destroy_id(&dev_id_priv->id);
return ret;
}

static int cma_listen_on_all(struct rdma_id_private *id_priv)
{
+ struct rdma_id_private *to_destroy;
struct cma_device *cma_dev;
int ret;

mutex_lock(&lock);
list_add_tail(&id_priv->list, &listen_any_list);
list_for_each_entry(cma_dev, &dev_list, list) {
- ret = cma_listen_on_dev(id_priv, cma_dev);
- if (ret)
+ ret = cma_listen_on_dev(id_priv, cma_dev, &to_destroy);
+ if (ret) {
+ /* Prevent racing with cma_process_remove() */
+ if (to_destroy)
+ list_del_init(&to_destroy->list);
goto err_listen;
+ }
}
mutex_unlock(&lock);
return 0;
@@ -2554,6 +2561,8 @@ static int cma_listen_on_all(struct rdma_id_private *id_priv)
err_listen:
list_del(&id_priv->list);
mutex_unlock(&lock);
+ if (to_destroy)
+ rdma_destroy_id(&to_destroy->id);
return ret;
}

@@ -4855,6 +4864,7 @@ static void cma_process_remove(struct cma_device *cma_dev)

static int cma_add_one(struct ib_device *device)
{
+ struct rdma_id_private *to_destroy;
struct cma_device *cma_dev;
struct rdma_id_private *id_priv;
unsigned int i;
@@ -4902,7 +4912,7 @@ static int cma_add_one(struct ib_device *device)
mutex_lock(&lock);
list_add_tail(&cma_dev->list, &dev_list);
list_for_each_entry(id_priv, &listen_any_list, list) {
- ret = cma_listen_on_dev(id_priv, cma_dev);
+ ret = cma_listen_on_dev(id_priv, cma_dev, &to_destroy);
if (ret)
goto free_listen;
}
@@ -4915,6 +4925,7 @@ free_listen:
list_del(&cma_dev->list);
mutex_unlock(&lock);

+ /* cma_process_remove() will delete to_destroy */
cma_process_remove(cma_dev);
kfree(cma_dev->default_roce_tos);
free_gid_type:
--
2.27.0


\
 
 \ /
  Last update: 2020-12-28 15:10    [W:1.552 / U:0.808 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site