firmware: stratix10-rsu: Fix NULL deref on rsu_send_msg() timeout in probe
Subsystems (2)
Dinh Nguyen
maintainerPATCH 1/1 firmware: stratix10-rsu: Fix NULL deref on rsu_send_msg() timeout in probe
rsu_send_msg() can return -ETIMEDOUT when wait_for_completion_interruptible_timeout() fires while the SMC call is still pending. In stratix10_rsu_probe(), the error paths for COMMAND_RSU_DCMF_VERSION, COMMAND_RSU_DCMF_STATUS, COMMAND_RSU_MAX_RETRY and COMMAND_RSU_GET_SPT_TABLE call stratix10_svc_free_channel() - which sets chan->scl to NULL - but then fall through and queue the next request on the same channel. The next svc kthread that runs will dereference pdata->chan->scl in its receive callback path, triggering a NULL pointer dereference identical to the one fixed by commit c45f7263100c ("firmware: stratix10-rsu: Fix NULL pointer dereference when RSU is disabled") for the COMMAND_RSU_STATUS path.
Apply the same cleanup pattern to the remaining failure paths: remove the async client, free the channel, and return early so no further messages are queued on a channel whose scl has been cleared.
This bug was spotted when Sashiko reviewed another patch:
https://sashiko.dev/#/patchset/cover.1779248894.git.tze.yee.ng%40altera.com
--- drivers/firmware/stratix10-rsu.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-)
drivers/firmware/stratix10-rsu.c +9 -2
@@ -778,32 +778,39 @@ static int stratix10_rsu_probe(struct platform_device *pdev) 0, rsu_dcmf_version_callback);if (ret) {dev_err (dev,"Error, getting DCMF version %i\n" , ret); +stratix10_svc_remove_async_client (priv->chan);stratix10_svc_free_channel (priv->chan); +return ret; } ret =rsu_send_msg (priv,COMMAND_RSU_DCMF_STATUS ,0 , rsu_dcmf_status_callback);if (ret) {dev_err (dev,"Error, getting DCMF status %i\n" , ret); +stratix10_svc_remove_async_client (priv->chan);stratix10_svc_free_channel (priv->chan); +return ret; } ret =rsu_send_msg (priv,COMMAND_RSU_MAX_RETRY ,0 , rsu_max_retry_callback);if (ret) {dev_err (dev,"Error, getting RSU max retry %i\n" , ret); +stratix10_svc_remove_async_client (priv->chan);stratix10_svc_free_channel (priv->chan); +return ret; } - ret =rsu_send_async_msg (dev, priv,COMMAND_RSU_GET_SPT_TABLE ,0 , rsu_async_get_spt_table_callback);if (ret) {dev_err (dev,"Error, getting SPT table %i\n" , ret); +stratix10_svc_remove_async_client (priv->chan);stratix10_svc_free_channel (priv->chan); +return ret; } -return ret; +return 0 ; }static void stratix10_rsu_remove (struct platform_device *pdev)
Signature · 1 line
2.42.0.411.g813d9a9188
…
Quoted 3 lines
Apply the same cleanup pattern to the remaining failure paths: remove the async client, free the channel, and return early so no further messages are queued on a channel whose scl has been cleared.
…
How do you think about to avoid a bit of duplicate source code in the implementation of the function “stratix10_rsu_probe”?
https://elixir.bootlin.com/linux/v7.1-rc4/source/drivers/firmware/stratix10-rsu.c#L715-L807
Regards,
Markus
Dinh Nguyen
maintainerQuoted 9 lines · Markus Elfring
…
Quoted 3 lines
Apply the same cleanup pattern to the remaining failure paths: remove the async client, free the channel, and return early so no further messages are queued on a channel whose scl has been cleared.
…
How do you think about to avoid a bit of duplicate source code in the implementation of the function “stratix10_rsu_probe”?
https://elixir.bootlin.com/linux/v7.1-rc4/source/drivers/firmware/stratix10-rsu.c#L715-L807
Excellent suggestion.
Thanks,
Dinh