Introduce an init_time callback to the at91_init_soc structure to be able to
tweak the init_time machine callback on a per-soc basis, instead of having to
rely on a global one.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Conflicts:
	arch/arm/mach-at91/setup.c

The timer driver is using some global variables to define some variables it has
to use in most of its functions, like the base address.

Use some container_of calls to have a single dynamic (and local) variable to
hold this content.

The only exception is in the !DT case, where the call chain to
at91sam926x_ioremap_pit and then at91sam926x_pit_init as init_time makes it
hard for the moment to pass the physical address of the timer.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

We can use the generic request_irq now to register a timer interrupt handler,
instead of the more complex setup_irq.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

All the panic messages hardcode the same prefix. Define the pr_fmt macro to
unify its definition.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

The current probe code is, whenever it fails to retrieve a resource, either
panicing, silently returning, or returning and displaying an error.

Make the code consistent by always panicing whenever it fails to grab any
resource.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

The current AT91 DT boards have a completely generic init_time definition.
Remove them from the machine declaration.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Conflicts:
	arch/arm/mach-at91/board-dt-sam9.c
	arch/arm/mach-at91/board-dt-sama5.c

The PIT timer driver until now had a single probe function, disregarding wether
it was probed through DT or in the old-style way. This code later on was
calling some DT function to retrieve the proper values for its base address,
interrupts and clocks.

While this was working, it was preventing the usage of CLOCKSOURCE_OF_DECLARE,
and the two different probe path were not as clearly separated as they could
be.

Rework the probe path to take this into account, and switch to
CLOCKSOURCE_OF_DECLARE.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Until now, the machines, even when CONFIG_OF was enabled, were calling
at91sam926x_ioremap_pit to try to map the PIT address using the defined
physical address.

Obviously, with DT, it's not appropriate anymore, and some code was added to
the function to deal with this case.

Unfortunately, this code was conditionned on CONFIG_OF, which can be enabled,
even though no DT was actually used, which would result in such a case, to this
code being executed, without any reason.

Moreover, the logic that was here before to bail out of the function just check
in the DT to see if the PIT node is there, which is the case in all our DTSI.

All this can be made much more straightforward just by using
of_have_populated_dt to bail out.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Until now, the pit_cycle computation was dividing the rate by HZ, rounding to
the closest integer, but without using the appropriate macro.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

The generic.h header file doesn't have any safe guards against multiple
inclusion. It only worked so far because all the symbols defined in it were
extern, but this is a rather fragile assumption.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Replace all masks and bits definitions by matching calls to BIT and GENMASK.
While we're at it, also fix a few style issues.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Acked-by: NDaniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

After the for_each_matching_node loop, we end up with a null value for np. Then,
of_match_node() is not matching anything and we can't register the standby
function and "ramc no standby function available" is printed.

Fix that by selecting the first available standby function. For now,
at91_pm_set_standby doesn't support multiple different standby functions and no
existing SoCs have different RAM controllers.
Signed-off-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

These headers used to provide an "API" to access the rstc and shdwc registers.
Now that no-one uses this API anymore, we can safely remove those.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Now that there's no user left for the global variables holding the reset and
shutdown controllers base address, we can remove these variables and their
associated mapping function.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Both the reset and poweroff drivers are now implementing what the
show_reset_status function used to do. Hence, we can remove this rather hackish
function that prevents us from doing further cleanup.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Now that all the SoC have been converted to use the new poweroff driver, we can
remove the poweroff code defined in mach-at91.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Register the poweroff driver in the old-style SoC definition so that the driver
is loaded and provide a shutdown hook.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Now that the poweroff code is a driver of its own, remove the DT probing in
mach-at91 and let the usual DT code do its job.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Now that the transition is over and that we probe our reset driver in every
case, we can remove the legacy code from the machine directory.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Make every board call the register_devices callback so that the devices
declared by the SoC are registered.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Register the reset device in the right SoCs so that the reset driver is
actually probed even in the old-style probing case.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Some core devices should be registered by the SoC itself rather than by every
board using this SoC. Introduce a register_devices callback that should be
called during the init_machine in order to do that.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Now that the reset code is a driver of its own, just let the usual DT probing
mecanism do its job, and remove the code entirely in this case.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Adapt the ramc mapping code to handle multiple ram controllers in the DT.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Most of the printed messages are using the "AT91:" prefix in the setup.c file,
but not all of them. Moreover, those who add it hardcode it directly in the
message, while the pr_fmt macro makes it easier for us to support such a case.
Signed-off-by: NMaxime Ripard <maxime.ripard@free-electrons.com>
Acked-by: NAlexandre Belloni <alexandre.belloni@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Remove all the material related to AIC5 support: this interrupt controller
driver is now implemented in drivers/irqchip/atmel-aic.c.
Signed-off-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Remove selection of OLD_IRQ_AT91 when selecting dt boards.
Select ATMEL_AIC_IRQ for sama5 SoCs (a kernel compiled for this SoC will
always use ATMEL_AIC_IRQ driver).
Select ATMEL_AIC_IRQ for at91rm9200 and at91sam9 SoCs only if OLD_IRQ_AT91
is not selected (which means we are compiling a pure DT kernel, without
any legacy board support).

Remove specific irq init code in all dt board files: this init procedure
is automatically handled in of_irq_init which is called by the arm irq core
code and is in charge of calling the appropriate aic init functions.
Signed-off-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Enclose at91_aic_xx calls in IS_ENABLED(CONFIG_OLD_IRQ_AT91) blocks in
order to prepare migration to the new AIC driver.

In the new AIC driver the suspend/resume functions are called by the
generic irq framework and are no longer needed in the PM specific code.

Moreover, the new AIC driver no longer exposes the at91_aic_base variable
which is used by the at91_aic_read functions.
Signed-off-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Introduce the OLD_IRQ_AT91 Kconfig option to prepare migration to the
new AIC driver.

Select this option for all at91 SoCs and all available boards so that we
can later move DT enabled boards to the new irq driver and keep the old
implementation when legacy boards are selected.
Signed-off-by: NBoris BREZILLON <boris.brezillon@free-electrons.com>
Signed-off-by: NNicolas Ferre <nicolas.ferre@atmel.com>

Power efficiency improves on Baytrail (Intel Atom Processor E3000)
when Linux disables C6 auto-demotion.

Based on work by Srinidhi Kasagar <srinidhi.kasagar@intel.com>.
Signed-off-by: NLen Brown <len.brown@intel.com>
Cc: x86@kernel.org

Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Perform a pci_claim_resource() on all valid resources discovered
during the OF device tree scan.

Based almost entirely upon the PCI OF bus probing code which does
the same thing there.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It seems that when a PCI Express bridge is not in use and has no devices
behind it, the ranges property is bogus.  Specifically the size property
is of the form [0xffffffff:...], and if you add this size to the resource
start address the 64-bit calculation will overflow.

Just check specifically for this size value signature and skip them.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Dump the various aspects of the PCI bridge probed at boot time, most
importantly the bridge number ranges, and the ranges property.

This helps diagnose PCI resource issues and other problems by giving
ofpci_debug=1 on the boot command line.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add tracepoint to track hugepage invalidate. This help us
in debugging difficult to track bugs.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

On ppc64 we support 4K hash pte with 64K page size. That requires
us to track the hash pte slot information on a per 4k basis. We do that
by storing the slot details in the second half of pte page. The pte bit
_PAGE_COMBO is used to indicate whether the second half need to be
looked while building real_pte. We need to use read memory barrier while
doing that so that load of hidx is not reordered w.r.t _PAGE_COMBO
check. On the store side we already do a lwsync in __hash_page_4K

CC: <stable@vger.kernel.org>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

We would get wrong results in compiler recomputed old_pmd. Avoid
that by using ACCESS_ONCE

CC: <stable@vger.kernel.org>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

As per ISA, for 4k base page size we compare 14..65 bits of VA specified
with the entry_VA in tlb. That implies we need to make sure we do a
tlbie with all the possible 4k va we used to access the 16MB hugepage.
With 64k base page size we compare 14..57 bits of VA. Hence we cannot
ignore the lower 24 bits of va while tlbie .We also cannot tlb
invalidate a 16MB entry with just one tlbie instruction because
we don't track which va was used to instantiate the tlb entry.

CC: <stable@vger.kernel.org>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

If we changed base page size of the segment, either via sub_page_protect
or via remap_4k_pfn, we do a demote_segment which doesn't flush the hash
table entries. We do a lazy hash page table flush for all mapped pages
in the demoted segment. This happens when we handle hash page fault for
these pages.

We use _PAGE_COMBO bit along with _PAGE_HASHPTE to indicate whether a
pte is backed by 4K hash pte. If we find _PAGE_COMBO not set on the pte,
that implies that we could possibly have older 64K hash pte entries in
the hash page table and we need to invalidate those entries.

Use _PAGE_COMBO to determine the page size with which we should
invalidate the hash table entries on unmap.

CC: <stable@vger.kernel.org>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

If we changed base page size of the segment, either via sub_page_protect
or via remap_4k_pfn, we do a demote_segment which doesn't flush the hash
table entries. We do a lazy hash page table flush for all mapped pages
in the demoted segment. This happens when we handle hash page fault
for these pages.

We use _PAGE_COMBO bit along with _PAGE_HASHPTE to indicate whether a
pte is backed by 4K hash pte. If we find _PAGE_COMBO not set on the pte,
that implies that we could possibly have older 64K hash pte entries in
the hash page table and we need to invalidate those entries.

Handle this correctly for 16M pages

CC: <stable@vger.kernel.org>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>