Re: [PATCH] clk: provide public clk_is_enabled function

[Mike Turquette <mturquette@linaro.org>]

Quoting Sebastian Hesselbarth (2013-10-06 12:42:01)
> On 10/06/2013 06:30 PM, Andrew Lunn wrote:
> > On Sun, Oct 06, 2013 at 11:06:09AM +0200, Gerhard Sittig wrote:
> >> On Sat, Oct 05, 2013 at 22:42 +0200, Andrew Lunn wrote:
> >>>
> >>> On Sat, Oct 05, 2013 at 10:24:30PM +0200, Uwe Kleine-König wrote:
> >>>> On Fri, Oct 04, 2013 at 12:08:30PM +0200, Sebastian Hesselbarth wrote:
> >>>>> To determine if a clk has been previously enabled, provide a public
> >>>>> clk_is_enabled function. This is especially helpful to check the state
> >>>>> of clk-gate without actually changing the state of the gate.
> >>>> I wonder what you want to do with the return value.
> >>>>
> >>>> When doing
> >>>>
> >>>>    if (clk_is_enabled(someclk))
> >>>>            do_something();
> >>>>
> >>>> you cannot in general know if the clock is still on when you start to
> >>>> do_something.
> >>>
> >>> Hi Uwe
> >>>
> >>> At least in the use case Sebastian needs it for, we don't need an "in
> >>> general" solution. It is used early boot time to see if the boot
> >>> loader left the clock running.
> >>
> >> Wait, unless I'm missing something, the clk_is_enabled() call
> >> _won't_ determine whether the clock is enabled in hardware
> >> (whether the boot loader created or left this condition), instead
> >> it only determines whether clk_enable() was called previously and
> >> thus the clock _shall_ be enabled.
> >
> > Nope, you are wrong.
> >
> > static int clk_gate_is_enabled(struct clk_hw *hw)
> > {
> >       u32 reg;
> >       struct clk_gate *gate = to_clk_gate(hw);
> >
> >       reg = clk_readl(gate->reg);
> >
> >       /* if a set bit disables this clk, flip it before masking */
> >       if (gate->flags & CLK_GATE_SET_TO_DISABLE)
> >          reg ^= BIT(gate->bit_idx);
> >
> >          reg &= BIT(gate->bit_idx);
> >
> >          return reg ? 1 : 0;
> > }
> >
> > It reads the hardware state.
> >
> >> AFAIK the kernel's CCF support is "self contained" and does not
> >> consider any data or state that was "inherited" from boot staged
> >> before the kernel.  That's why the "disable unused" step disables
> >> everything that wasn't acquired _in the kernel_ regardless of
> >> what the boot loader may have done or what is enabled at reset.
> >
> > Not quite true. It uses the is_enabled(), which gets the real hardware
> > state, to turn off clocks which are unused but on. It will not turn
> > off clock which are already off. So it is inheriting some state from
> > the boot loader, in that it knows if the bootloader turned it
> > on. However this is not propagated into prepare/enable status.
> >
> >>> The other user of the clock is the
> >>> ethernet driver, which we know cannot change it yet, because driver
> >>> probing has not started yet.
> >>
> >> I understand that the situation here is, that the ethernet driver
> >> hasn't probed yet, but the clock driver did.  You are in early setup
> >> code and want to (check and) fetch data from the hardware which the
> >> ethernet driver later needs.
> >
> > Nearly, but not quite. If there is an enabled DT node for the device,
> > and that node does not have a valid local-mac-address property in the
> > node, the bootloader should of programmed the MAC address into the
> > device. If it has done that, the clock should be running, because as
> > soon as you turn the clock off, it forgets the MAC address. Thus, if
> > we find an enabled device in DT, without a valid local-mac-address,
> > and the clock is off, we have a bootloader bug, which we want to
> > report.
> >
> >> What's wrong with an explicit enable/disable around the data
> >> acquisition?
> >
> > It avoids the CPU locking hard, but will not get us a valid MAC
> > address, which is the point of the exercise.
> 
> While I agree to all Andew explained above, I somehow have the strong
> feeling that an clk_is_enabled will just be abused where possible. We
> already have two ppl complaining about it - even though a quick look at
> clk/core.c should have cleared out most of it.
> 
> Maybe we should just enable the clock, get the (possibly bogus) MAC
> and disable it again. We loose one possible FW_BUG report and overwrite
> an invalid local-mac-address property with another invalid one.

Firstly, I'm OK with adding a new clk_is_enabled API for The Right
Reasons, if we can figure out what those reasons are. A major concern is
the lack of locks/barriers around that call that create a critical
section where the enabled state is guaranteed not to change. Those locks
are not exported to drivers nor do I want them to be.

Secondly, this specific Ethernet/MAC address issue seems like another
case of "the driver should be calling clk_get and clk_prepare_enable but
for some reason it doesn't". This seems to be a common pattern and I'm
not sure why. Does calling clk_enable on your clock which has been
already enabled by the bootloader cause issues for you? If not then it
is better to just call clk_enable and have the clock framework book
keeping in-sync with the hardware state.

Is it possible in your case to only detect the invalid MAC address
without sniffing the state of the clock hardware? Isn't it valid to
report a bootloader bug after only looking at the MAC address and not
the clock enabled state?

Regards,
Mike

> 
> Sebastian