lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date:   Fri, 19 Feb 2021 11:42:48 +0100
From:   Hans de Goede <hdegoede@...hat.com>
To:     "Limonciello, Mario" <Mario.Limonciello@...l.com>,
        "mgross@...ux.intel.com" <mgross@...ux.intel.com>
Cc:     LKML <linux-kernel@...r.kernel.org>,
        "platform-driver-x86@...r.kernel.org" 
        <platform-driver-x86@...r.kernel.org>,
        "Bharathi, Divya" <Divya.Bharathi@...l.com>,
        Alexander Naumann <alexandernaumann@....de>
Subject: Re: [PATCH] platform/x86: dell-wmi-sysman: correct an initialization
 failure

Hi,

On 2/19/21 12:26 AM, Limonciello, Mario wrote:
> 
> 
>> -----Original Message-----
>> From: mark gross <mgross@...ux.intel.com>
>> Sent: Thursday, February 18, 2021 16:49
>> To: Limonciello, Mario
>> Cc: Hans De Goede; Mark Gross; LKML; platform-driver-x86@...r.kernel.org;
>> Bharathi, Divya; Alexander Naumann
>> Subject: Re: [PATCH] platform/x86: dell-wmi-sysman: correct an initialization
>> failure
>>
>>
>> [EXTERNAL EMAIL]
>>
>> On Thu, Feb 18, 2021 at 01:17:23PM -0600, Mario Limonciello wrote:
>>> On Dell systems that don't support this interface the module is
>>> mistakingly returning error code "0", when it should be returning
>>> -ENODEV.  Correct a logic error to guarantee the correct return code.
>>>
>>> Cc: Divya Bharathi <Divya_Bharathi@...l.com>
>>> Reported-by: Alexander Naumann <alexandernaumann@....de>
>>> Signed-off-by: Mario Limonciello <mario.limonciello@...l.com>
>>> ---
>>>  drivers/platform/x86/dell-wmi-sysman/biosattr-interface.c     | 4 +++-
>>>  drivers/platform/x86/dell-wmi-sysman/passwordattr-interface.c | 4 +++-
>>>  drivers/platform/x86/dell-wmi-sysman/sysman.c                 | 4 ++--
>>>  3 files changed, 8 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/platform/x86/dell-wmi-sysman/biosattr-interface.c
>> b/drivers/platform/x86/dell-wmi-sysman/biosattr-interface.c
>>> index f95d8ddace5a..8d59f81f9db4 100644
>>> --- a/drivers/platform/x86/dell-wmi-sysman/biosattr-interface.c
>>> +++ b/drivers/platform/x86/dell-wmi-sysman/biosattr-interface.c
>>> @@ -175,7 +175,9 @@ static struct wmi_driver bios_attr_set_interface_driver
>> = {
>>>
>>>  int init_bios_attr_set_interface(void)
>>>  {
>>> -	return wmi_driver_register(&bios_attr_set_interface_driver);
>>> +	int ret = wmi_driver_register(&bios_attr_set_interface_driver);
>> I have to ask if the propper fix should be in wmi_driver_register
> 
> Do you mean something like this?
> 
> diff --git a/drivers/platform/x86/wmi.c b/drivers/platform/x86/wmi.c
> index c669676ea8e8..89d04c5e3ab9 100644
> --- a/drivers/platform/x86/wmi.c
> +++ b/drivers/platform/x86/wmi.c
> @@ -1415,6 +1415,11 @@ static int acpi_wmi_probe(struct platform_device *device)
>  int __must_check __wmi_driver_register(struct wmi_driver *driver,
>                                        struct module *owner)
>  {
> +       const struct wmi_device_id *id = driver->id_table;
> +
> +       if (!wmi_has_guid(id->guid_string))
> +               return -ENODEV;
> +
>         driver->driver.owner = owner;
>         driver->driver.bus = &wmi_bus_type;
> 

No, drivers should be able to register before the GUID shows up. I know that
the GUID showing up later will likely never happen with WMI, but having a match
check like this in the driver_register function is highly unusual and would
be different from what all other busses do.

But your initial fix here is wrong too, because it does call wmi_driver_register,
which succeeds and then makes sysman_init() exit with -ENODEV.

Returning -ENODEV from sysman_init() is what we want, this causes the entire
insmod to be aborted, without logging an error (because of -ENODEV) so the
code will not be taking up memory.

This means that the memory into which the module was loaded before the kernel
calls sysman_init() will be free-ed and now the *still* registered WMI driver
entry will point to that free-ed memory, which is not good (TM).

So instead init_bios_attr_set_interface() should become something like this:

int init_bios_attr_set_interface(void)
{
	int ret;

	ret = wmi_driver_register(&bios_attr_set_interface_driver);
	if (ret)
		return ret;

	if (!wmi_priv.bios_attr_wdev) {
		wmi_driver_unregister(&bios_attr_set_interface_driver);
		return -ENODEV;
	}

	return 0;
}

And the same for the init_bios_attr_pass_interface() function.

This follows the standard kernel pattern that a function should always
undo any things / resource-allocations it has done on error before
exiting with an error.

Regards,

Hans

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ