Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep dependencies.

[Kevin Hilman <khilman@deeprootsystems.com>]

"Gopinath, Thara" <thara@ti.com> writes:

>>>-----Original Message-----
>>>From: Kevin Hilman [mailto:khilman@deeprootsystems.com]
>>>Sent: Thursday, December 10, 2009 10:36 PM
>>>To: Gopinath, Thara
>>>Cc: linux-omap@vger.kernel.org; Paul Walmsley
>>>Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the sleep
>>>dependencies.
>>>
>>>"Gopinath, Thara" <thara@ti.com> writes:
>>>
>>>>>>-----Original Message-----
>>>>>>From: Kevin Hilman [mailto:khilman@deeprootsystems.com]
>>>>>>Sent: Thursday, December 10, 2009 4:39 AM
>>>>>>To: Gopinath, Thara
>>>>>>Cc: linux-omap@vger.kernel.org; Paul Walmsley
>>>>>>Subject: Re: [PATCH] OMAP3: hwmod: check for clkdomain pointer before accesing it to change the
>>>sleep
>>>>>>dependencies.
>>>>>>
>>>>>>Thara Gopinath <thara@ti.com> writes:
>>>>>>
>>>>>>> Some clock nodes like wdt1_fck, sr1_fck, sr2_fck etc do not have a
>>>>>>> clock domain associated . For such nodes accessing the clock domain pointers
>>>>>>> in _add_initiator_dep and _del_initiator_dep, will lead to null pointer
>>>>>>> defreferencing crash. Adding support in these functions to check for
>>>>>>> existence of clkdm pointer before trying to acess it. Even if tomorrow
>>>>>>> we correct all the clock nodes to have an associated clock domain, checking
>>>>>>> for the existence of the pointer is a good programming practice.
>>>>>>>
>>>>>>> Signed-off-by: Thara Gopinath <thara@ti.com>
>>>>>>> Cc: Paul Walmsley <paul@pwsan.com>
>>>>>>> ---
>>>>>>> This patch depends on http://patchwork.kernel.org/patch/63383/
>>>>>>
>>>>>>I think clocks without associated clockdomains should be handled with
>>>>>>a flag instead of a check for a NULL pointer.
>>>>>>
>>>>>>Your current approach will silently fail if someone forgets to add
>>>>>>a clockdomain to a clock that should have one.
>>>>
>>>> Are you suggesting adding a flag to the clock node in case it does
>>>> not have an associated clockdomain?
>>>
>>>Yes, or to the hwmod.
>>>
>>>> I am using your tree origin/pm-wip/omap_device branch. On this I see
>>>> a commit (e89a95db7095d998991c564a756a33ee0ec722c5) in which a
>>>> warning is printed in omap2_init_clk_clkdm if there is no
>>>> clockdomain associated with a clock node.
>>>
>>>This is to check for the case of a missing clkdm when there should be
>>>one.  You're trying to solve the problem of when there is no
>>>clockdomain for a given clock.
>>>
>>>> My idea was this warning plus the check so as to avoid the system
>>>> crash would suffice.
>>>
>>>I still think this should be flagged explicitly in the clock/hwmod,
>>>then check using the flag instead of of using NULL pointer check.
>>>
>>>For readability, it makes it more clear that this is checking for some
>>>exceptions and not simply doing error checking.
>
> Agreed. We could add a flag and base the check on the flag. But
> still if some clock node does not have a clkdomain and forgets to
> put this flag, the system will still crash. 

Yes, but with a clear message as to why.

> So have have both the checks ??

I don't think so, because the check you added will allow the system to
continue in a known non-working state.

Kevin