On 2019/5/18 0:18, Corey Minyard wrote:
> On Fri, May 17, 2019 at 06:12:44PM +0800, Kefeng Wang wrote:
>> If ipmi_si_platform_init()->platform_driver_register() fails,
>> platform_driver_unregister() called unconditionally will trigger
>> following warning,
>>
>> ipmi_platform: Unable to register driver: -12
>> ------------[ cut here ]------------
>> Unexpected driver unregister!
>> WARNING: CPU: 1 PID: 7210 at drivers/base/driver.c:193
>> driver_unregister+0x60/0x70 drivers/base/driver.c:193
>>
>> Fix it by adding platform_registered variable, only unregister platform
>> driver when it is already successfully registered.
> This is good, but have you found out why the driver was unable to
> register in the first place? That really shouldn't happen.
Yes, it is hard to happen in real scene, the test robot found this issue
when FAULT_INJECTION enabled.
>
> This patch is queued for 5.3.
>
> -corey
>
>> Reported-by: Hulk Robot <hul...@huawei.com>
>> Signed-off-by: Kefeng Wang <wangkefeng.w...@huawei.com>
>> ---
>> drivers/char/ipmi/ipmi_si_platform.c | 6 +++++-
>> 1 file changed, 5 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/char/ipmi/ipmi_si_platform.c
>> b/drivers/char/ipmi/ipmi_si_platform.c
>> index f2a91c4d8cab..0cd849675d99 100644
>> --- a/drivers/char/ipmi/ipmi_si_platform.c
>> +++ b/drivers/char/ipmi/ipmi_si_platform.c
>> @@ -19,6 +19,7 @@
>> #include "ipmi_si.h"
>> #include "ipmi_dmi.h"
>>
>> +static bool platform_registered;
>> static bool si_tryplatform = true;
>> #ifdef CONFIG_ACPI
>> static bool si_tryacpi = true;
>> @@ -469,9 +470,12 @@ void ipmi_si_platform_init(void)
>> int rv = platform_driver_register(&ipmi_platform_driver);
>> if (rv)
>> pr_err("Unable to register driver: %d\n", rv);
>> + else
>> + platform_registered = true;
>> }
>>
>> void ipmi_si_platform_shutdown(void)
>> {
>> - platform_driver_unregister(&ipmi_platform_driver);
>> + if (platform_registered)
>> + platform_driver_unregister(&ipmi_platform_driver);
>> }
>> --
>> 2.20.1
>>
> .
>
