hwmon: (xgene) Fix use after free bug in xgene_hwmon_remove due to race condition

[ Upstream commit 9e728fdb4526ea1b4679b4fefa1bc551b414fedc ]

In xgene_hwmon_probe, &ctx->workq is bound with xgene_hwmon_evt_work.
Then it will be started.

If we remove the driver which will call xgene_hwmon_remove to clean up,
there may be unfinished work.

The possible sequence is as follows:

Fix it by finishing the work before cleanup in xgene_hwmon_remove.

CPU0                  CPU1

                    |xgene_hwmon_evt_work
xgene_hwmon_remove   |
kfifo_free(&ctx->async_msg_fifo);|
                    |
                    |kfifo_out_spinlocked
                    |//use &ctx->async_msg_fifo
Fixes: d99fa72f98cb ("hwmon: (xgene) Fix crash when alarm occurs before driver probe")
Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
Link: https://lore.kernel.org/r/20230310084007.1403388-1-zyytlz.wz@163.com
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

diff --git a/drivers/hwmon/xgene-hwmon.c b/drivers/hwmon/xgene-hwmon.c
index 5cde837bfd094c57a3cf719b0fe0e4c2645dce8e..d1abea49f01bef60223746099d5cdc339e8277dd 100644 (file)
--- a/drivers/hwmon/xgene-hwmon.c
+++ b/drivers/hwmon/xgene-hwmon.c
@@ -761,6 +761,7 @@ static int xgene_hwmon_remove(struct platform_device *pdev)
 {
        struct xgene_hwmon_dev *ctx = platform_get_drvdata(pdev);
 
+       cancel_work_sync(&ctx->workq);
        hwmon_device_unregister(ctx->hwmon_dev);
        kfifo_free(&ctx->async_msg_fifo);
        if (acpi_disabled)