We've now implemented code in the pseries platform to use the new PAPR
interface to allow resizing the hash page table (HPT) at runtime.

This patch uses that interface to automatically attempt to resize the HPT
when memory is hot added or removed.  This tries to always keep the HPT at
a reasonable size for our current memory size.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

The hypervisor needs to know a guest is capable of using the HPT resizing
PAPR extension in order to make full advantage of it for memory hotplug.

If the hypervisor knows the guest is HPT resize aware, it can size the
initial HPT based on the initial guest RAM size, relying on the guest to
resize the HPT when more memory is hot-added. Without this, the hypervisor
must size the HPT for the maximum possible guest RAM, which can lead to
a huge waste of space if the guest never actually expends to that maximum
size.

This patch advertises the guest's support for HPT resizing via the
ibm,client-architecture-support OF interface. We use bit 5 of byte 6 of
option vector 5 for this purpose, as defined in the PAPR ACR "HPT
resizing option".

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Reviewed-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

This adds support for using two hypercalls to change the size of the
main hash page table while running as a PAPR guest. For now these
hypercalls are only in experimental qemu versions.

The interface is two part: first H_RESIZE_HPT_PREPARE is used to
allocate and prepare the new hash table. This may be slow, but can be
done asynchronously. Then, H_RESIZE_HPT_COMMIT is used to switch to the
new hash table. This requires that no CPUs be concurrently updating the
HPT, and so must be run under stop_machine().

This also adds a debugfs file which can be used to manually control
HPT resizing or testing purposes.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Paul Mackerras <paulus@samba.org>
[mpe: Rename the debugfs file to "hpt_order"]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

This adds the hypercall numbers and wrapper functions for the hash page
table resizing hypercalls.

These hypercall numbers are defined in the PAPR ACR "HPT resizing
option".

It also adds a new firmware feature flag to track the presence of the
HPT resizing calls.

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

List all the current valid compatible strings for the l2cache binding.
This should stop checkpatch.pl from complaining and will hopefully save
someone from having to debug a typo in their dts.

Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

We don't need asm/xics.h

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Recent versions of OPAL can provide names for the various OPAL interrupts,
so let's use them. This also modernises the code that fetches the
interrupt array to use the helpers provided by the generic code instead
of hand-parsing the property.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[mpe: Free irqs on error, check allocation of names, consolidate error
      handling, whitespace.]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

On some CAPP errors we see console messages that prints unknown HMIs for
which CAPI recovery is in progress. This patch fixes this by printing
correct error info for HMI generated due to CAPP recovery.

Signed-off-by: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
Tested-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

These are common on bare metal machines, so put them in the defconfig.

This adds 216KB to the vmlinux size

Signed-off-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Specifically:
 - CONFIG_BPF_SYSCALL
 - CONFIG_NET_SCHED
 - CONFIG_NET_CLS_BPF
 - CONFIG_NET_CLS_ACT
 - CONFIG_NET_ACT_BPF
 - CONFIG_CGROUP_BPF
 - CONFIG_UPROBE_EVENT

... in pseries, ppc64 and powernv defconfigs.

Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Change the device probe test in the via-cuda.c driver so it will load on
Egret-based machines too. Remove the now redundant via-maciisi.c driver.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

The Egret system controller was the predecessor to the Cuda and the
differences are minor.

On Cuda, byte acknowledgement requires one transition of the TACK
signal; on Egret two are needed. On Cuda, TIP is active low; on Egret
it is active high. And Cuda raises certain interrupts that Egret omits.

Accomodating these differences complicates the Cuda driver slightly
but avoids a lot of duplication (see next patch).

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Initialize data_index where appropriate to improve readability and
assist debugging. This change doesn't affect driver behaviour.

I prefer to see
	current_req->data[data_index++]
in place of
	current_req->data[0]
or
	current_req->data[1]
inasmuchas it becomes obvious what the data_index variable does.

Moreover, the actual value of data_index when examined at any given moment
tells me something about prior events, which did prove helpful.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

The cuda_start() function uses spinlock_irq_save/restore for mutual
exclusion. Let's have cuda_poll() do the same when polling the VIA
interrupt.

The benefit to disabling local irqs when the interrupt is being polled
is that the interrupt handler now has the same timing properties
regardless of whether it is invoked normally or from cuda_poll().

This driver was written back when local irqs remained enabled during
execution of interrupt handlers and cuda_poll() was probably trying
to achieve the same effect by use of enable/disable_irq.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

When a read transaction completes, one of several things will happen:
a new transfer is started by the driver, a new transfer request
is raised by the Cuda (i.e. TREQ asserted), or both happen at once.

When both happen at once, there is a race condition between the TREQ test
in the read_done state and the same test in cuda_start(). Moreover, the
former test uses a stale TREQ value.

Theoretically, this can result in the undesirable outcome that the
interrupt handler completes with the state machine 'idle' when it should
instead start the next transaction.

Avoid this race by calling cuda_start() first and then confirming that it
succeeded. If not, test the current TREQ value before entering the
'reading' state.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

When reading_reply is set, reply_ptr points into an adb_request struct.
Conversely, when reply_ptr instead points into the global cuda_rbuf,
reading_reply must be false.

Unfortunately, this rule can be violated because re-initialization
of reply_ptr and reading_reply presently depends on the TREQ input.

Fix this by re-initializing reply_ptr and reading_reply as soon as they
are known to be invalid.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

If the Cuda driver does not enter the 'read_done' state for some
reason, it may continue in the 'reading' state until the buffer
overflows. Add a bounds check to prevent this.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Introduce some helpers for handling the signalling between VIA and
Cuda. This abstraction will be used to add support for Egret devices,
which utilize slightly different signalling.

Don't invert the sense of the Cuda's active-low signals when storing
them in the 'status' variable. Just assert, negate and test those
signals using the helpers.

The state machine does not need to test its own output signals to
figure out what to do next: the next state depends on the Cuda's TREQ
output. Just call the TREQ_asserted() helper function to test for that.

Similarly, there is no need to store pin directions in the 'status'
variable. That was only useful for debugging messages.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

There is no possibility that current_req can change during execution of
cuda_start(). This can be confirmed by inspection: cuda_lock is always
held whenever cuda_start() is called or current_req is modified.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Add missing log message severity, remove old debug messages and
replace printk() loop with print_hex_dump() call.

Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Without this we will always find the feature disabled.

Fixes: 984d7a1e

 ("powerpc/mm: Fixup kernel read only mapping")
Cc: stable@vger.kernel.org # v4.7+
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Acked-by: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

start,size has the benefit of being easier to search for (start,end
usually gives you the preceeding vector from the one you want, as first
result).

Suggested-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Somewhere along the line, search/replace left some naming garbled,
and untidy alignment (aka. mpe stuffed it up). Might as well fix them
all up now while git blame history doesn't extend too far.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

This adds AUX vectors for the L1I,D, L2 and L3 cache levels
providing for each cache level the size of the cache in bytes
and the geometry (line size and number of ways).

We chose to not use the existing alpha/sh definition which
packs all the information in a single entry per cache level as
it is too restricted to represent some of the geometries used
on POWER.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

All shipping firmware versions have it wrong in the device-tree

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Retrieved from device-tree when available

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

We have two set of identical struct members for the I and D sides
and mostly identical bunches of code to parse the device-tree to
populate them. Instead make a ppc_cache_info structure with one
copy for I and one for D

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

It will be used to calculate the associativity

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

In a number of places we called "cache line size" what is actually
the cache block size, which in the powerpc architecture, means the
effective size to use with cache management instructions (it can
be different from the actual cache line size).

We fix the naming across the board and properly retrieve both
pieces of information when available in the device-tree.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

We don't patch instructions based on the cache lines or block
sizes these days.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

The variables are defined twice in setup_32.c and setup_64.c, do it
once in setup-common.c instead

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

It's an kernel private macro, it doesn't belong there

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Enable support for GCC plugins on powerpc.

Add an additional version check in gcc-plugins-check to advise users to
upgrade to gcc 5.2+ on powerpc to avoid issues with header files (gcc <=
4.6) or missing copies of rs6000-cpus.def (4.8 to 5.1 on 64-bit
targets).

Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Acked-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Commit 38addce8 ("gcc-plugins: Add latent_entropy plugin") excludes
certain powerpc early boot code from the latent entropy plugin by adding
appropriate CFLAGS. It looks like this was supposed to cover
prom_init.o, but ended up saying init.o (which doesn't exist) instead.
Fix the typo.

Fixes: 38addce8

 ("gcc-plugins: Add latent_entropy plugin")
Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

The variable DISABLE_LATENT_ENTROPY_PLUGIN is defined when
CONFIG_PAX_LATENT_ENTROPY is set. This is leftover from the original PaX
version of the plugin code and doesn't actually exist. Change the condition
to depend on CONFIG_GCC_PLUGIN_LATENT_ENTROPY instead.

Fixes: 38addce8

 ("gcc-plugins: Add latent_entropy plugin")
Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Stub out the debugfs functions so that the build doesn't break when
CONFIG_DEBUG_FS=n.

Reported-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

As we add the ability to do DLPAR of additional devices through
the sysfs interface we need to know which devices are supported.
This adds the reporting of supported devices with a comma separated
list reported in the existing /sys/kernel/dlpar.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Extend the existing PRRN infrastructure to perform the actual affinity
updating for cpus and memory in addition to the device tree updating.
For cpus, dynamic affinity updating already appears to exist in the
kernel in the form of arch_update_cpu_topology(). For memory, we must
place a READD operation on the hotplug queue for any phandle included in
the PRRN event that is determined to be an LMB.

Signed-off-by: John Allen <jallen@linux.vnet.ibm.com>
Reviewed-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

Currently, memory must be hot removed and subsequently re-added in order
to dynamically update the affinity of LMBs specified by a PRRN event.
Earlier implementations of the PRRN event handler ran into issues in which
the hot remove would occur successfully, but a hotplug event would be
initiated from another source and grab the hotplug lock preventing the hot
add from occurring. To prevent this situation, this patch introduces the
notion of a hot "readd" action for memory which atomizes a hot remove and
a hot add into a single, serialized operation on the hotplug queue.

Signed-off-by: John Allen <jallen@linux.vnet.ibm.com>
Reviewed-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

When adding and removing LMBs we should make the acquire/release of
the DRC a separate step to allow for a few improvements. First
this will ensure that LMBs removed during a remove by count operation
are all available if a error occurs and we need to add them back. By
first removeing all the LMBs from the kernel before releasing their
DRCs the LMBs are available to add back should an error occur.

Also, this will allow for faster re-add operations of memory for
PRRN event handling since we can skip the unneeded step of having
to release the DRC and the acquire it back.

Signed-off-by: Nathan Fontenot <nfont@linux.vnet.ibm.com>
Signed-off-by: John Allen <jallen@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>