From: Bob Pearson <rpearsonhpe@gmail.com>
To: jgg@nvidia.com, zyjzyj2000@gmail.com, linux-rdma@vger.kernel.org
Cc: Bob Pearson <rpearsonhpe@gmail.com>
Subject: [PATCH for-next v3 01/10] RDMA/rxe: Make rxe_alloc() take pool lock
Date: Fri, 22 Oct 2021 14:18:16 -0500 [thread overview]
Message-ID: <20211022191824.18307-2-rpearsonhpe@gmail.com> (raw)
In-Reply-To: <20211022191824.18307-1-rpearsonhpe@gmail.com>
In rxe_pool.c there are two separate pool APIs for creating a new object
rxe_alloc() and rxe_alloc_locked(). Currently they are identical except
for GFP_KERNEL vs GFP_ATOMIC. Make rxe_alloc() take the pool lock which
is in line with the other APIs in the library and was the original intent.
Keep kzalloc() outside of the lock to avoid using GFP_ATOMIC.
Make similar change to rxe_add_to_pool. The code checking the pool limit
is potentially racy without a lock. The lock around finishing the pool
element provides a memory barrier before 'publishing' the object.
Signed-off-by: Bob Pearson <rpearsonhpe@gmail.com>
---
v3
Kept rxe_alloc and rxe_alloc_locked separate to allow use of GFP_KERNEL
in rxe_alloc
drivers/infiniband/sw/rxe/rxe_pool.c | 18 +++++++++++++++---
1 file changed, 15 insertions(+), 3 deletions(-)
diff --git a/drivers/infiniband/sw/rxe/rxe_pool.c b/drivers/infiniband/sw/rxe/rxe_pool.c
index 7b4cb46edfd9..8138e459b6c1 100644
--- a/drivers/infiniband/sw/rxe/rxe_pool.c
+++ b/drivers/infiniband/sw/rxe/rxe_pool.c
@@ -356,39 +356,51 @@ void *rxe_alloc(struct rxe_pool *pool)
{
struct rxe_type_info *info = &rxe_type_info[pool->type];
struct rxe_pool_entry *elem;
+ unsigned long flags;
u8 *obj;
+ write_lock_irqsave(&pool->pool_lock, flags);
if (atomic_inc_return(&pool->num_elem) > pool->max_elem)
goto out_cnt;
+ write_unlock_irqrestore(&pool->pool_lock, flags);
obj = kzalloc(info->size, GFP_KERNEL);
- if (!obj)
- goto out_cnt;
+ if (!obj) {
+ atomic_dec(&pool->num_elem);
+ return NULL;
+ }
+ write_lock_irqsave(&pool->pool_lock, flags);
elem = (struct rxe_pool_entry *)(obj + info->elem_offset);
-
elem->pool = pool;
kref_init(&elem->ref_cnt);
+ write_unlock_irqrestore(&pool->pool_lock, flags);
return obj;
out_cnt:
atomic_dec(&pool->num_elem);
+ write_unlock_irqrestore(&pool->pool_lock, flags);
return NULL;
}
int __rxe_add_to_pool(struct rxe_pool *pool, struct rxe_pool_entry *elem)
{
+ unsigned long flags;
+
+ write_lock_irqsave(&pool->pool_lock, flags);
if (atomic_inc_return(&pool->num_elem) > pool->max_elem)
goto out_cnt;
elem->pool = pool;
kref_init(&elem->ref_cnt);
+ write_unlock_irqrestore(&pool->pool_lock, flags);
return 0;
out_cnt:
atomic_dec(&pool->num_elem);
+ write_unlock_irqrestore(&pool->pool_lock, flags);
return -EINVAL;
}
--
2.30.2
next prev parent reply other threads:[~2021-10-22 19:19 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-22 19:18 [PATCH for-next v3 00/10] Correct race conditions in rdma_rxe Bob Pearson
2021-10-22 19:18 ` Bob Pearson [this message]
2021-10-25 17:55 ` [PATCH for-next v3 01/10] RDMA/rxe: Make rxe_alloc() take pool lock Jason Gunthorpe
2021-10-25 19:09 ` Robert Pearson
2021-10-25 19:41 ` Jason Gunthorpe
2021-10-22 19:18 ` [PATCH for-next v3 02/10] RDMA/rxe: Copy setup parameters into rxe_pool Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 03/10] RDMA/rxe: Save object pointer in pool element Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 04/10] RDMA/rxe: Combine rxe_add_index with rxe_alloc Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 05/10] RDMA/rxe: Combine rxe_add_key " Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 06/10] RDMA/rxe: Separate out last rxe_drop_ref Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 07/10] RDMA/rxe: Rewrite rxe_mcast.c Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 08/10] RDMA/rxe: Fix ref error in rxe_av.c Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 09/10] RDMA/rxe: Replace mr by rkey in responder resources Bob Pearson
2021-10-22 19:18 ` [PATCH for-next v3 10/10] RDMA/rxe: Minor cleanup in rxe_pool.c Bob Pearson
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=20211022191824.18307-2-rpearsonhpe@gmail.com \
--to=rpearsonhpe@gmail.com \
--cc=jgg@nvidia.com \
--cc=linux-rdma@vger.kernel.org \
--cc=zyjzyj2000@gmail.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).