net/smc: fix NULL sndbuf_desc in smc_cdc_tx_handler()
authorD. Wythe <alibuda@linux.alibaba.com>
Wed, 8 Mar 2023 08:17:12 +0000 (16:17 +0800)
committerJakub Kicinski <kuba@kernel.org>
Mon, 13 Mar 2023 23:03:58 +0000 (16:03 -0700)
commit22a825c541d775c1dbe7b2402786025acad6727b
treede9a699ecc429220f9173bb439d3a13b39f44844
parent131db499162274858bdbd7b5323a639da4aab86c
net/smc: fix NULL sndbuf_desc in smc_cdc_tx_handler()

When performing a stress test on SMC-R by rmmod mlx5_ib driver
during the wrk/nginx test, we found that there is a probability
of triggering a panic while terminating all link groups.

This issue dues to the race between smc_smcr_terminate_all()
and smc_buf_create().

smc_smcr_terminate_all

smc_buf_create
/* init */
conn->sndbuf_desc = NULL;
...

__smc_lgr_terminate
smc_conn_kill
smc_close_abort
smc_cdc_get_slot_and_msg_send

__softirqentry_text_start
smc_wr_tx_process_cqe
smc_cdc_tx_handler
READ(conn->sndbuf_desc->len);
/* panic dues to NULL sndbuf_desc */

conn->sndbuf_desc = xxx;

This patch tries to fix the issue by always to check the sndbuf_desc
before send any cdc msg, to make sure that no null pointer is
seen during cqe processing.

Fixes: 0b29ec643613 ("net/smc: immediate termination for SMCR link groups")
Signed-off-by: D. Wythe <alibuda@linux.alibaba.com>
Reviewed-by: Tony Lu <tonylu@linux.alibaba.com>
Reviewed-by: Wenjia Zhang <wenjia@linux.ibm.com>
Link: https://lore.kernel.org/r/1678263432-17329-1-git-send-email-alibuda@linux.alibaba.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
net/smc/smc_cdc.c