Re: [libata:upstream 7/11] drivers/ata/libata-zpodd.c:83 zpodd_wake_dev() warn: variable dereferenced before check 'ata_dev' (see line 80)

[Jeff Garzik <jgarzik@pobox.com>]

On 01/22/2013 03:45 AM, Aaron Lu wrote:
> On 01/22/2013 04:38 PM, Dan Carpenter wrote:
>> On Tue, Jan 22, 2013 at 04:32:06PM +0800, Aaron Lu wrote:
>>> Hi Dan,
>>>
>>> Thanks for the remind.
>>>
>>> On 01/22/2013 02:25 PM, Dan Carpenter wrote:
>>>> Hi Aaron,
>>>>
>>>> FYI, there are new smatch warnings show up in
>>>>
>>>> tree:   git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev.git upstream
>>>> head:   7e15e9be37eb834aaaca69030064ac97eaf5df2f
>>>> commit: f064a20dded807448669426c9bfb7d03aba5659c [7/11] libata: move acpi notification code to zpodd
>>>>
>>>> New smatch warnings:
>>>> drivers/ata/libata-zpodd.c:83 zpodd_wake_dev() warn: variable dereferenced before check 'ata_dev' (see line 80)
>>>
>>> The ata_dev variable shouldn't ever be NULL here.
>>>
>>>>
>>>> git remote add libata git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev.git
>>>> git remote update libata
>>>> git checkout f064a20dded807448669426c9bfb7d03aba5659c
>>>> vim +/ata_dev +83 drivers/ata/libata-zpodd.c
>>>>
>>>> afe75951 Aaron Lu 2013-01-15  74  	return acpi_device_can_poweroff(acpi_dev);
>>>> afe75951 Aaron Lu 2013-01-15  75  }
>>>> afe75951 Aaron Lu 2013-01-15  76
>>>> f064a20d Aaron Lu 2013-01-15  77  static void zpodd_wake_dev(acpi_handle handle, u32 event, void *context)
>>>> f064a20d Aaron Lu 2013-01-15  78  {
>>>> f064a20d Aaron Lu 2013-01-15  79  	struct ata_device *ata_dev = context;
>>>> f064a20d Aaron Lu 2013-01-15 @80  	struct zpodd *zpodd = ata_dev->zpodd;
>>>> f064a20d Aaron Lu 2013-01-15  81  	struct device *dev = &ata_dev->sdev->sdev_gendev;
>>>> f064a20d Aaron Lu 2013-01-15  82
>>>> f064a20d Aaron Lu 2013-01-15 @83  	if (event == ACPI_NOTIFY_DEVICE_WAKE && ata_dev &&
>>>
>>> If I remove the check for ata_dev here, will the smatch warning go away?
>>> Or should I add a check above? I suppose we need to do something to the
>>> warning, so that it won't pop up in the future, so please feel free to
>>> suggest, thanks!
>>
>> Removing the check would silence the warning.
>
> OK, thanks.
>
> Hi Jeff,
> Should I prepare a new patch to fix this or update the original patch?

If it's in libata-dev.git#upstream, then a patch on top of all the 
others, which addresses the issue.

	Jeff