[PATCH] nvmet-tcp: fix page fragment cache leak in error path

[PATCH] nvmet-tcp: fix page fragment cache leak in error path

[Geliang Tang]

From: Geliang Tang <tanggeliang@kylinos.cn>

In nvmet_tcp_alloc_queue(), when a connection is closed during the
allocation process (e.g., nvmet_tcp_set_queue_sock() returns -ENOTCONN),
the error handling jumps to out_free_queue without draining the page
fragment cache.

Although nvmet_tcp_free_cmd() is called in some error paths to release
individual page fragments, the underlying page cache reference held by
queue->pf_cache is never released. This results in a page leak each time
a connection fails during allocation, which could lead to memory
exhaustion over time if connections are repeatedly opened and closed.

Fix this by calling page_frag_cache_drain() before freeing the queue
structure in the out_free_queue label.

Fixes: 872d26a391da ("nvmet-tcp: add NVMe over TCP target driver")
Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn>
---
 drivers/nvme/target/tcp.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 20f150d17a96..8b8ec533e352 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -2002,6 +2002,7 @@ static void nvmet_tcp_alloc_queue(struct nvmet_tcp_port *port,
 out_sock:
 	fput(queue->sock->file);
 out_free_queue:
+	page_frag_cache_drain(&queue->pf_cache);
 	kfree(queue);
 out_release:
 	pr_err("failed to allocate queue, error %d\n", ret);
-- 
2.53.0

Re: [PATCH] nvmet-tcp: fix page fragment cache leak in error path

[Christoph Hellwig]

On Mon, May 25, 2026 at 05:50:19PM +0800, Geliang Tang wrote:
> From: Geliang Tang <tanggeliang@kylinos.cn>
> 
> In nvmet_tcp_alloc_queue(), when a connection is closed during the
> allocation process (e.g., nvmet_tcp_set_queue_sock() returns -ENOTCONN),
> the error handling jumps to out_free_queue without draining the page
> fragment cache.

nvmet_tcp_set_queue_sock jumps to out_destroy_sq.

> Although nvmet_tcp_free_cmd() is called in some error paths to release
> individual page fragments, the underlying page cache reference held by
> queue->pf_cache is never released. This results in a page leak each time
> a connection fails during allocation, which could lead to memory
> exhaustion over time if connections are repeatedly opened and closed.
> 
> Fix this by calling page_frag_cache_drain() before freeing the queue
> structure in the out_free_queue label.

The first allocation using pf_cache is the call to nvmet_tcp_alloc_cmd
after ida_alloc.  So afaik this drain should at the out_ida_remove
label.  And given how obscure this is, maybe a comment would not
hurt either?

Otherwise this looks good.