[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20221002014047.23066-1-duoming@zju.edu.cn>
Date: Sun, 2 Oct 2022 09:40:47 +0800
From: Duoming Zhou <duoming@....edu.cn>
To: linux-kernel@...r.kernel.org
Cc: martin.petersen@...cle.com, kuba@...nel.org, davem@...emloft.net,
andrii@...nel.org, gregkh@...uxfoundation.org,
linux-scsi@...r.kernel.org, target-devel@...r.kernel.org,
Duoming Zhou <duoming@....edu.cn>
Subject: [PATCH] scsi: target: iscsi: cxgbit: fix sleep-in-atomic-context bug in cxgbit_abort_conn
The function iscsit_handle_time2retain_timeout() is a timer handler that
runs in an atomic context, but it calls "alloc_skb(0, GFP_KERNEL | ...)"
that may sleep. As a result, the sleep-in-atomic-context bug will happen.
The process is shown below:
iscsit_handle_time2retain_timeout()
iscsit_close_session()
iscsit_free_connection_recovery_entries()
iscsit_free_cmd()
__iscsit_free_cmd()
cxgbit_unmap_cmd()
cxgbit_abort_conn()
alloc_skb(0, GFP_KERNEL | ...) //may sleep
This patch changes the gfp_t parameter of alloc_skb() from GFP_KERNEL to
GFP_ATOMIC in order to mitigate the bug.
Fixes: 1ae01724ae92 ("cxgbit: Abort the TCP connection in case of data out timeout")
Signed-off-by: Duoming Zhou <duoming@....edu.cn>
---
drivers/target/iscsi/cxgbit/cxgbit_cm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/target/iscsi/cxgbit/cxgbit_cm.c b/drivers/target/iscsi/cxgbit/cxgbit_cm.c
index 3336d2b78bf..eb3da6d2c62 100644
--- a/drivers/target/iscsi/cxgbit/cxgbit_cm.c
+++ b/drivers/target/iscsi/cxgbit/cxgbit_cm.c
@@ -697,7 +697,7 @@ __cxgbit_abort_conn(struct cxgbit_sock *csk, struct sk_buff *skb)
void cxgbit_abort_conn(struct cxgbit_sock *csk)
{
- struct sk_buff *skb = alloc_skb(0, GFP_KERNEL | __GFP_NOFAIL);
+ struct sk_buff *skb = alloc_skb(0, GFP_ATOMIC | __GFP_NOFAIL);
cxgbit_get_csk(csk);
cxgbit_init_wr_wait(&csk->com.wr_wait);
--
2.17.1
Powered by blists - more mailing lists