platform/kernel/linux-starfive.git
4 years agoMIPS: Loongson64: Remove not used pci.c
Tiezhu Yang [Thu, 28 May 2020 11:29:11 +0000 (19:29 +0800)]
MIPS: Loongson64: Remove not used pci.c

After commit 6423e59a64e7 ("MIPS: Loongson64: Switch to generic PCI
driver"), arch/mips/loongson64/pci.c is not used any more, remove it.

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: Define PCI_IOBASE
Jiaxun Yang [Wed, 27 May 2020 13:17:21 +0000 (21:17 +0800)]
MIPS: Loongson64: Define PCI_IOBASE

PCI_IOBASE is used to create VM maps for PCI I/O ports, it is
required by generic PCI drivers to make memory mapped I/O range
work.

To deal with legacy drivers that have fixed I/O ports range we
reserved 0x10000 in PCI_IOBASE, should be enough for i8259 i8042
stuff.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: CPU_LOONGSON2EF need software to maintain cache consistency
Lichao Liu [Thu, 28 May 2020 01:10:31 +0000 (09:10 +0800)]
MIPS: CPU_LOONGSON2EF need software to maintain cache consistency

CPU_LOONGSON2EF need software to maintain cache consistency,
so modify the 'cpu_needs_post_dma_flush' function to return true
when the cpu type is CPU_LOONGSON2EF.

Cc: stable@vger.kernel.org
Signed-off-by: Lichao Liu <liulichao@loongson.cn>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: DTS: Fix build errors used with various configs
Tiezhu Yang [Thu, 28 May 2020 01:59:01 +0000 (09:59 +0800)]
MIPS: DTS: Fix build errors used with various configs

If CONFIG_MIPS_MALTA is not set but CONFIG_LEGACY_BOARD_SEAD3 is set,
the subdir arch/mips/boot/dts/mti will not be built, so the sead3.dts
which depends on CONFIG_LEGACY_BOARD_SEAD3 in this subdir is also not
built, and then there exists the following build error, fix it.

  LD      .tmp_vmlinux.kallsyms1
arch/mips/generic/board-sead3.o:(.mips.machines.init+0x4): undefined reference to `__dtb_sead3_begin'
Makefile:1106: recipe for target 'vmlinux' failed
make: *** [vmlinux] Error 1

Additionally, add CONFIG_FIT_IMAGE_FDT_BOSTON check for subdir img to
fix the following build error when CONFIG_MACH_PISTACHIO is not set but
CONFIG_FIT_IMAGE_FDT_BOSTON is set.

FATAL ERROR: Couldn't open "boot/dts/img/boston.dtb": No such file or directory

Reported-by: kbuild test robot <lkp@intel.com>
Reported-by: Guenter Roeck <linux@roeck-us.net>
Fixes: 41528ba6afe6 ("MIPS: DTS: Only build subdir of current platform")
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: select NO_EXCEPT_FILL
Jiaxun Yang [Wed, 27 May 2020 06:34:34 +0000 (14:34 +0800)]
MIPS: Loongson64: select NO_EXCEPT_FILL

Loongson64 load kernel at 0x82000000 and allocate exception vectors
by ebase. So we don't need to reserve space for exception vectors
at head of kernel.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Fix IRQ tracing when call handle_fpe() and handle_msa_fpe()
YuanJunQing [Wed, 27 May 2020 06:11:30 +0000 (14:11 +0800)]
MIPS: Fix IRQ tracing when call handle_fpe() and handle_msa_fpe()

Register "a1" is unsaved in this function,
 when CONFIG_TRACE_IRQFLAGS is enabled,
 the TRACE_IRQS_OFF macro will call trace_hardirqs_off(),
 and this may change register "a1".
 The changed register "a1" as argument will be send
 to do_fpe() and do_msa_fpe().

Signed-off-by: YuanJunQing <yuanjunqing66@163.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: mm: add page valid judgement in function pte_modify
Bibo Mao [Wed, 27 May 2020 02:25:20 +0000 (10:25 +0800)]
MIPS: mm: add page valid judgement in function pte_modify

If original PTE has _PAGE_ACCESSED bit set, and new pte has no
_PAGE_NO_READ bit set, we can add _PAGE_SILENT_READ bit to enable
page valid bit.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomm/memory.c: Add memory read privilege on page fault handling
Bibo Mao [Wed, 27 May 2020 02:25:19 +0000 (10:25 +0800)]
mm/memory.c: Add memory read privilege on page fault handling

Here add pte_sw_mkyoung function to make page readable on MIPS
platform during page fault handling. This patch improves page
fault latency about 10% on my MIPS machine with lmbench
lat_pagefault case.

It is noop function on other arches, there is no negative
influence on those architectures.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomm/memory.c: Update local TLB if PTE entry exists
Bibo Mao [Wed, 27 May 2020 02:25:18 +0000 (10:25 +0800)]
mm/memory.c: Update local TLB if PTE entry exists

If two threads concurrently fault at the same page, the thread that
won the race updates the PTE and its local TLB. For now, the other
thread gives up, simply does nothing, and continues.

It could happen that this second thread triggers another fault, whereby
it only updates its local TLB while handling the fault. Instead of
triggering another fault, let's directly update the local TLB of the
second thread. Function update_mmu_tlb is used here to update local
TLB on the second thread, and it is defined as empty on other arches.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Do not flush tlb page when updating PTE entry
Bibo Mao [Wed, 27 May 2020 02:25:17 +0000 (10:25 +0800)]
MIPS: Do not flush tlb page when updating PTE entry

It is not necessary to flush tlb page on all CPUs if suitable PTE
entry exists already during page fault handling, just updating
TLB is fine.

Here redefine flush_tlb_fix_spurious_fault as empty on MIPS system.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: Default to a generic board
Paul Cercueil [Mon, 25 May 2020 17:41:55 +0000 (19:41 +0200)]
MIPS: ingenic: Default to a generic board

Having a generic board option makes it possible to create a kernel that
will run on various Ingenic SoCs, as long as the right devicetree is
provided.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: Add support for GCW Zero prototype
Paul Cercueil [Mon, 25 May 2020 17:41:54 +0000 (19:41 +0200)]
MIPS: ingenic: Add support for GCW Zero prototype

Add support for the GCW Zero prototype. The only (?) difference is that
it only has 256 MiB of RAM, compared to the 512 MiB of RAM of the retail
device.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: DTS: Add memory info of GCW Zero
Paul Cercueil [Mon, 25 May 2020 17:41:53 +0000 (19:41 +0200)]
MIPS: ingenic: DTS: Add memory info of GCW Zero

Add memory info of the GCW Zero in its devicetree. The bootloader
generally provides this information, but since it is fixed to 512 MiB,
it doesn't hurt to have it in devicetree. It allows the kernel to boot
without any parameter passed as argument.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: Switch to generic PCI driver
Jiaxun Yang [Tue, 26 May 2020 09:21:16 +0000 (17:21 +0800)]
MIPS: Loongson64: Switch to generic PCI driver

We can now enable generic PCI driver in Kconfig, and remove legacy
PCI driver code.

Radeon vbios quirk is moved to the platform folder to fit the
new structure.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: DTS: Loongson64: Add PCI Controller Node
Jiaxun Yang [Tue, 26 May 2020 09:21:15 +0000 (17:21 +0800)]
MIPS: DTS: Loongson64: Add PCI Controller Node

Add PCI Host controller node for Loongson64 with RS780E PCH dts.
Note that PCI interrupts are probed via legacy way, as different
machine have different interrupt arrangement, we can't cover all
of them in dt.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agodt-bindings: Document Loongson PCI Host Controller
Jiaxun Yang [Tue, 26 May 2020 09:21:14 +0000 (17:21 +0800)]
dt-bindings: Document Loongson PCI Host Controller

PCI host controller found on Loongson PCHs and SoCs.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoPCI: Add Loongson PCI Controller support
Jiaxun Yang [Tue, 26 May 2020 09:21:13 +0000 (17:21 +0800)]
PCI: Add Loongson PCI Controller support

This controller can be found on Loongson-2K SoC, Loongson-3
systems with RS780E/LS7A PCH.

The RS780E part of code was previously located at
arch/mips/pci/ops-loongson3.c and now it can use generic PCI
driver implementation.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoPCI: Don't disable decoding when mmio_always_on is set
Jiaxun Yang [Tue, 26 May 2020 09:21:12 +0000 (17:21 +0800)]
PCI: Don't disable decoding when mmio_always_on is set

Don't disable MEM/IO decoding when a device have both non_compliant_bars
and mmio_always_on.

That would allow us quirk devices with junk in BARs but can't disable
their decoding.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Acked-by: Bjorn Helgaas <helgaas@kernel.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: BCM63xx: fix 6328 boot selection bit
Álvaro Fernández Rojas [Tue, 26 May 2020 11:03:24 +0000 (13:03 +0200)]
MIPS: BCM63xx: fix 6328 boot selection bit

MISC_STRAP_BUS_BOOT_SEL_SHIFT is 18 according to Broadcom's GPL source code.

Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
Acked-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: BCM63XX: fix BCM6358 GPIO count
Daniel González Cabanelas [Sun, 24 May 2020 18:59:53 +0000 (20:59 +0200)]
MIPS: BCM63XX: fix BCM6358 GPIO count

The BCM6358 SoC has only 38 available GPIOs. Fix it.

Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: tools: Fix resource leak in elf-entry.c
Kaige Li [Thu, 14 May 2020 12:59:41 +0000 (20:59 +0800)]
MIPS: tools: Fix resource leak in elf-entry.c

There is a file descriptor resource leak in elf-entry.c, fix this
by adding fclose() before return and die.

Signed-off-by: Kaige Li <likaige@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: SGI-IP30: Reorder the macros in war.h
Joshua Kinard [Sun, 17 May 2020 23:24:39 +0000 (19:24 -0400)]
MIPS: SGI-IP30: Reorder the macros in war.h

Fix the ordering of the macros in arch/mips/mach-ip30/war.h to match
those in arch/mips/mach-ip27/war.h.

Signed-off-by: Joshua Kinard <kumba@gentoo.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: emulate CPUCFG instruction on older Loongson64 cores
WANG Xuerui [Sat, 23 May 2020 13:37:01 +0000 (21:37 +0800)]
MIPS: emulate CPUCFG instruction on older Loongson64 cores

CPUCFG is the instruction for querying processor characteristics on
newer Loongson processors, much like CPUID of x86. Since the instruction
is supposedly designed to provide a unified way to do feature detection
(without having to, for example, parse /proc/cpuinfo which is too
heavyweight), it is important to provide compatibility for older cores
without native support. Fortunately, most of the fields can be
synthesized without changes to semantics. Performance is not really big
a concern, because feature detection logic is not expected to be
invoked very often in typical userland applications.

The instruction can't be emulated on LOONGSON_2EF cores, according to
FlyGoat's experiments. Because the LWC2 opcode is assigned to other
valid instructions on 2E and 2F, no RI exception is raised for us to
intercept. So compatibility is only extended back furthest to
Loongson-3A1000. Loongson-2K is covered too, as it is basically a remix
of various blocks from the 3A/3B models from a kernel perspective.

This is lightly based on Loongson's work on their Linux 3.10 fork, for
being the authority on the right feature flags to fill in, where things
aren't otherwise discoverable.

Signed-off-by: WANG Xuerui <git@xen0n.name>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: Huacai Chen <chenhc@lemote.com>
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Tidy up CP0.Config6 bits definition
Huacai Chen [Sat, 23 May 2020 07:51:45 +0000 (15:51 +0800)]
MIPS: Tidy up CP0.Config6 bits definition

CP0.Config6 is a Vendor-defined register whose bits definitions are
different from one to another. Recently, Xuerui's Loongson-3 patch and
Serge's P5600 patch make the definitions inconsistency and unclear.

To make life easy, this patch tidy the definition up:
1, Add a _MTI_ infix for proAptiv/P5600 feature bits;
2, Add a _LOONGSON_ infix for Loongson-3 feature bits;
3, Add bit6/bit7 definition for Loongson-3 which will be used later.

All existing users of these macros are updated.

Cc: WANG Xuerui <git@xen0n.name>
Cc: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: DTS: Only build subdir of current platform
Tiezhu Yang [Thu, 21 May 2020 02:15:06 +0000 (10:15 +0800)]
MIPS: DTS: Only build subdir of current platform

Add config check in Makefile to only build the subdir of current platform.

E.g. without this patch:

  AR      arch/mips/built-in.a
  AR      arch/mips/boot/dts/brcm/built-in.a
  AR      arch/mips/boot/dts/cavium-octeon/built-in.a
  AR      arch/mips/boot/dts/img/built-in.a
  AR      arch/mips/boot/dts/ingenic/built-in.a
  AR      arch/mips/boot/dts/lantiq/built-in.a
  DTC     arch/mips/boot/dts/loongson/loongson3_4core_rs780e.dtb
  DTB     arch/mips/boot/dts/loongson/loongson3_4core_rs780e.dtb.S
  AS      arch/mips/boot/dts/loongson/loongson3_4core_rs780e.dtb.o
  DTC     arch/mips/boot/dts/loongson/loongson3_8core_rs780e.dtb
  DTB     arch/mips/boot/dts/loongson/loongson3_8core_rs780e.dtb.S
  AS      arch/mips/boot/dts/loongson/loongson3_8core_rs780e.dtb.o
  AR      arch/mips/boot/dts/loongson/built-in.a
  AR      arch/mips/boot/dts/mscc/built-in.a
  AR      arch/mips/boot/dts/mti/built-in.a
  AR      arch/mips/boot/dts/netlogic/built-in.a
  AR      arch/mips/boot/dts/ni/built-in.a
  AR      arch/mips/boot/dts/pic32/built-in.a
  AR      arch/mips/boot/dts/qca/built-in.a
  AR      arch/mips/boot/dts/ralink/built-in.a
  AR      arch/mips/boot/dts/xilfpga/built-in.a
  AR      arch/mips/boot/dts/built-in.a

With this patch:

  AR      arch/mips/built-in.a
  DTC     arch/mips/boot/dts/loongson/loongson3_4core_rs780e.dtb
  DTB     arch/mips/boot/dts/loongson/loongson3_4core_rs780e.dtb.S
  AS      arch/mips/boot/dts/loongson/loongson3_4core_rs780e.dtb.o
  DTC     arch/mips/boot/dts/loongson/loongson3_8core_rs780e.dtb
  DTB     arch/mips/boot/dts/loongson/loongson3_8core_rs780e.dtb.S
  AS      arch/mips/boot/dts/loongson/loongson3_8core_rs780e.dtb.o
  AR      arch/mips/boot/dts/loongson/built-in.a
  AR      arch/mips/boot/dts/built-in.a

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Fix exception handler memcpy()
Ben Hutchings [Sat, 23 May 2020 15:50:34 +0000 (23:50 +0800)]
MIPS: Fix exception handler memcpy()

The exception handler subroutines are declared as a single char, but
when copied to the required addresses the copy length is 0x80.

When range checks are enabled for memcpy() this results in a build
failure, with error messages such as:

In file included from arch/mips/mti-malta/malta-init.c:15:
In function 'memcpy',
    inlined from 'mips_nmi_setup' at arch/mips/mti-malta/malta-init.c:98:2:
include/linux/string.h:376:4: error: call to '__read_overflow2' declared with attribute error: detected read beyond size of object passed as 2nd parameter
  376 |    __read_overflow2();
      |    ^~~~~~~~~~~~~~~~~~

Change the declarations to use type char[].

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Signed-off-by: YunQiang Su <syq@debian.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: cevt-r4k: Update the r4k-clockevent frequency in sync with CPU
Serge Semin [Thu, 21 May 2020 14:07:24 +0000 (17:07 +0300)]
mips: cevt-r4k: Update the r4k-clockevent frequency in sync with CPU

Due to being embedded into the CPU cores MIPS count/compare timer
frequency is changed together with the CPU clocks alteration.
In case if frequency really changes the kernel clockevent framework
must be notified, otherwise the kernel timers won't work correctly.
Fix this by calling clockevents_update_freq() for each r4k clockevent
handlers registered per available CPUs.

Traditionally MIPS r4k-clock are clocked with CPU frequency divided by 2.
But this isn't true for some of the platforms. Due to this we have to save
the basic CPU frequency, so then use it to scale the initial timer
frequency (mips_hpt_frequency) and pass the updated value further to the
clockevent framework.

Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: csrc-r4k: Mark R4K timer as unstable if CPU freq changes
Serge Semin [Thu, 21 May 2020 14:07:23 +0000 (17:07 +0300)]
mips: csrc-r4k: Mark R4K timer as unstable if CPU freq changes

Commit 07d69579e7fe ("MIPS: Don't register r4k sched clock when CPUFREQ
enabled") disabled the r4k-clock usage for scheduler ticks counting due
to the scheduler being non-tolerant for unstable clocks sources. For the
same reason the clock should be used in the system clocksource framework
with care. As soon as CPU frequency changes the clocksource framework
should be notified about this by marking the R4K timer being unstable
(which it really is, since the ticks rate has been changed synchronously
with the CPU frequency).

Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Add udelay lpj numbers adjustment
Serge Semin [Thu, 21 May 2020 14:07:22 +0000 (17:07 +0300)]
mips: Add udelay lpj numbers adjustment

Loops-per-jiffies is a special number which represents a number of
noop-loop cycles per CPU-scheduler quantum - jiffies. As you
understand aside from CPU-specific implementation it depends on
the CPU frequency. So when a platform has the CPU frequency fixed,
we have no problem and the current udelay interface will work
just fine. But as soon as CPU-freq driver is enabled and the cores
frequency changes, we'll end up with distorted udelay's. In order
to fix this we have to accordinly adjust the per-CPU udelay_val
(the same as the global loops_per_jiffy) number. This can be done
in the CPU-freq transition event handler. We subscribe to that event
in the MIPS arch time-inititalization method.

Co-developed-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Add CPS_NS16550_WIDTH config
Serge Semin [Thu, 21 May 2020 14:07:19 +0000 (17:07 +0300)]
mips: Add CPS_NS16550_WIDTH config

On some platforms IO-memory might require to use a proper load/store
instructions (like Baikal-T1 IO-memory). To fix the cps-vec UART debug
printout let's add the CONFIG_CPS_NS16550_WIDTH config to determine which
instructions lb/sb, lh/sh or lw/sw are required for MMIO operations.

Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Add CONFIG/CONFIG6/Cause reg fields macro
Serge Semin [Thu, 21 May 2020 14:07:18 +0000 (17:07 +0300)]
mips: Add CONFIG/CONFIG6/Cause reg fields macro

There are bit fields which persist in the MIPS CONFIG and CONFIG6
registers, but haven't been described in the generic mipsregs.h
header so far. In particular, the generic CONFIG bitfields are
BE - endian mode, BM - burst mode, SB - SimpleBE, OCP interface mode
indicator, UDI - user-defined "CorExtend" instructions, DSP - data
scratch pad RAM present, ISP - instruction scratch pad RAM present,
etc. The core-specific CONFIG6 bitfields are JRCD - jump register
cache prediction disable, R6 - MIPSr6 extensions enable, IFUPerfCtl -
IFU performance control, SPCD - sleep state performance counter, DLSB -
disable load/store bonding. A new exception code reported in the
ExcCode field of the Cause register: 30 - Parity/ECC error exception
happened on either fetch, load or cache refill. Lets add them to the
mipsregs.h header to be used in future platform code, which have them
utilized.

Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Add CP0 Write Merge config support
Serge Semin [Thu, 21 May 2020 14:07:17 +0000 (17:07 +0300)]
mips: Add CP0 Write Merge config support

CP0 config register may indicate whether write-through merging
is allowed. Currently there are two types of the merging available:
SysAD Valid and Full modes. Whether each of them are supported by
the core is implementation dependent. Moreover whether the ability
to change the mode also depends on the chip family instance. Taking
into account all of this we created a dedicated mm_config() method
to detect and enable merging if it's supported. It is called for
MIPS-type processors at CPU-probe stage and attempts to detect whether
the write merging is available. If it's known to be supported and
switchable, then switch on the full mode. Otherwise just perform the
CP0.Config.MM field analysis.

In addition there are platforms like InterAptiv/ProAptiv, which do have
the MM bit field set by default, but having write-through cacheing
unsupported makes write-merging also unsupported. In this case we just
ignore the MM field value.

Co-developed-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Fix cpu_has_mips64r1/2 activation for MIPS32 CPUs
Serge Semin [Thu, 21 May 2020 14:07:16 +0000 (17:07 +0300)]
mips: Fix cpu_has_mips64r1/2 activation for MIPS32 CPUs

Commit 1aeba347b3a9 ("MIPS: Hardcode cpu_has_mips* where target ISA
allows") updated the cpu_has_mips* macro to be replaced with a constant
expression where it's possible. By mistake it wasn't done correctly
for cpu_has_mips64r1/cpu_has_mips64r2 macro. They are defined to
be replaced with conditional expression __isa_range_or_flag(), which
means either ISA revision being within the range or the corresponding
CPU options flag was set at the probe stage or both being true at the
same time. But the ISA level value doesn't indicate whether the ISA is
MIPS32 or MIPS64. Due to this if we select MIPS32r1 - MIPS32r5
architectures the __isa_range() macro will activate the
cpu_has_mips64rX flags, which is incorrect. In order to fix the
problem we make sure the 64bits CPU support is enabled by means of
checking the flag cpu_has_64bits aside with proper ISA range and specific
Revision flag being set.

Fixes: 1aeba347b3a9 ("MIPS: Hardcode cpu_has_mips* where target ISA allows")
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Add MIPS Warrior P5600 support
Serge Semin [Thu, 21 May 2020 14:07:15 +0000 (17:07 +0300)]
mips: Add MIPS Warrior P5600 support

This is a MIPS32 Release 5 based IP core with XPA, EVA, dual/quad issue
exec pipes, MMU with two-levels TLB, UCA, MSA, MDU core level features
and system level features like up to six P5600 calculation cores, CM2
with L2 cache, IOCU/IOMMU (though might be unused depending on the
system-specific IP core configuration), GIC, CPC, virtualisation module,
eJTAG and PDtrace.

As being MIPS32 Release 5 based core it provides all the features
available by the CPU_MIPS32_R5 config, while adding a few more like
UCA attribute support, availability of CPU-freq (by means of L2/CM
clock ratio setting), EI/VI GIC modes detection at runtime.

In addition to this if P5600 architecture is enabled modern GNU GCC
provides a specific tuning for P5600 processors with respect to the
classic MIPS32 Release 5. First of all branch-likely avoidance is
activated only when the code is compiled with the speed optimization
(avoidance is always enabled for the pure MIPS32 Release 5
architecture). Secondly the madd/msub avoidance is enabled since
madd/msub utilization isn't profitable due to overhead of getting the
result out of the HI/LO registers. Multiply-accumulate instructions are
activated and utilized together with the necessary code reorder when
multiply-add/multiply-subtract statements are met. Finally load/store
bonding is activated by default. All of these optimizations may make
the code relatively faster than if just MIP32 release 5 architecture
was requested.

Co-developed-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Add MIPS Release 5 support
Serge Semin [Thu, 21 May 2020 14:07:14 +0000 (17:07 +0300)]
mips: Add MIPS Release 5 support

There are five MIPS32/64 architecture releases currently available:
from 1 to 6 except fourth one, which was intentionally skipped.
Three of them can be called as major: 1st, 2nd and 6th, that not only
have some system level alterations, but also introduced significant
core/ISA level updates. The rest of the MIPS architecture releases are
minor.

Even though they don't have as much ISA/system/core level changes
as the major ones with respect to the previous releases, they still
provide a set of updates (I'd say they were intended to be the
intermediate releases before a major one) that might be useful for the
kernel and user-level code, when activated by the kernel or compiler.
In particular the following features were introduced or ended up being
available at/after MIPS32/64 Release 5 architecture:
+ the last release of the misaligned memory access instructions,
+ virtualisation - VZ ASE - is optional component of the arch,
+ SIMD - MSA ASE - is optional component of the arch,
+ DSP ASE is optional component of the arch,
+ CP0.Status.FR=1 for CP1.FIR.F64=1 (pure 64-bit FPU general registers)
  must be available if FPU is implemented,
+ CP1.FIR.Has2008 support is required so CP1.FCSR.{ABS2008,NAN2008} bits
  are available.
+ UFR/UNFR aliases to access CP0.Status.FR from user-space by means of
  ctc1/cfc1 instructions (enabled by CP0.Config5.UFR),
+ CP0.COnfig5.LLB=1 and eretnc instruction are implemented to without
  accidentally clearing LL-bit when returning from an interrupt,
  exception, or error trap,
+ XPA feature together with extended versions of CPx registers is
  introduced, which needs to have mfhc0/mthc0 instructions available.

So due to these changes GNU GCC provides an extended instructions set
support for MIPS32/64 Release 5 by default like eretnc/mfhc0/mthc0. Even
though the architecture alteration isn't that big, it still worth to be
taken into account by the kernel software. Finally we can't deny that
some optimization/limitations might be found in future and implemented
on some level in kernel or compiler. In this case having even
intermediate MIPS architecture releases support would be more than
useful.

So the most of the changes provided by this commit can be split into
either compile- or runtime configs related. The compile-time related
changes are caused by adding the new CONFIG_CPU_MIPS32_R5/CONFIG_CPU_MIPSR5
configs and concern the code activating MIPSR2 or MIPSR6 already
implemented features (like eretnc/LLbit, mthc0/mfhc0). In addition
CPU_HAS_MSA can be now freely enabled for MIPS32/64 release 5 based
platforms as this is done for CPU_MIPS32_R6 CPUs. The runtime changes
concerns the features which are handled with respect to the MIPS ISA
revision detected at run-time by means of CP0.Config.{AT,AR} bits. Alas
these fields can be used to detect either r1 or r2 or r6 releases.
But since we know which CPUs in fact support the R5 arch, we can manually
set MIPS_CPU_ISA_M32R5/MIPS_CPU_ISA_M64R5 bit of c->isa_level and then
use cpu_has_mips32r5/cpu_has_mips64r5 where it's appropriate.

Since XPA/EVA provide too complex alterationss and to have them used with
MIPS32 Release 2 charged kernels (for compatibility with current platform
configs) they are left to be setup as a separate kernel configs.

Co-developed-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: MAAR: Use more precise address mask
Serge Semin [Thu, 21 May 2020 00:34:37 +0000 (03:34 +0300)]
mips: MAAR: Use more precise address mask

Indeed according to the MIPS32 Privileged Resource Architecgture the MAAR
pair register address field either takes [12:31] bits for non-XPA systems
and [12:55] otherwise. In any case the current address mask is just
wrong for 64-bit and 32-bits XPA chips. So lets extend it to 59-bits
of physical address value. This shall cover the 64-bits architecture and
systems with XPA enabled, and won't cause any problem for non-XPA 32-bit
systems, since address values exceeding the architecture specific MAAR
mask will be just truncated with setting zeros in the unsupported upper
bits.

Co-developed-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: SGI-IP27: Remove not used definition TICK_SIZE in ip27-timer.c
Tiezhu Yang [Wed, 20 May 2020 07:08:02 +0000 (15:08 +0800)]
MIPS: SGI-IP27: Remove not used definition TICK_SIZE in ip27-timer.c

After commit f5ff0a280201 ("[MIPS] Use generic NTP code for all MIPS
platforms"), TICK_SIZE is not used in ip27-timer.c for many years,
remove it.

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: Add missing include
Paul Cercueil [Tue, 19 May 2020 21:22:30 +0000 (23:22 +0200)]
MIPS: ingenic: Add missing include

Add missing include which adds the prototype to plat_time_init().

Fixes: f932449c11da ("MIPS: ingenic: Drop obsolete code, merge the rest in setup.c")
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: SGI-IP27: Remove not used includes and comment in ip27-timer.c
Tiezhu Yang [Wed, 20 May 2020 05:12:37 +0000 (13:12 +0800)]
MIPS: SGI-IP27: Remove not used includes and comment in ip27-timer.c

After commit 0ce5ebd24d25 ("mfd: ioc3: Add driver for SGI IOC3 chip"),
the related includes and comment about ioc3 are not used any more in
ip27-timer.c, remove them.

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: MAAR: Add XPA mode support
Serge Semin [Wed, 6 May 2020 17:42:30 +0000 (20:42 +0300)]
mips: MAAR: Add XPA mode support

When XPA mode is enabled the normally 32-bits MAAR pair registers
are extended to be of 64-bits width as in pure 64-bits MIPS
architecture. In this case the MAAR registers can enable the
speculative loads/stores for addresses of up to 39-bits width.
But in this case the process of the MAAR initialization changes a bit.
The upper 32-bits of the registers are supposed to be accessed by mean
of the dedicated instructions mfhc0/mthc0 and there is a CP0.MAAR.VH
bit which should be set together with CP0.MAAR.VL as indication
of the boundary validity. All of these peculiarities were taken into
account in this commit so the speculative loads/stores would work
when XPA mode is enabled.

Co-developed-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: linux-pm@vger.kernel.org
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Remove useless parameter of bootcmdline_init()
Zhi Li [Mon, 18 May 2020 07:08:08 +0000 (15:08 +0800)]
MIPS: Remove useless parameter of bootcmdline_init()

The parameter "cmdline_p" is useless in bootcmdline_init(),remove it.

Signed-off-by: Zhi Li <lizhi01@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: SGI-IP27: Remove duplicated include in ip27-timer.c
Tiezhu Yang [Tue, 19 May 2020 12:28:11 +0000 (20:28 +0800)]
MIPS: SGI-IP27: Remove duplicated include in ip27-timer.c

After commit 9d0aaf98dc24 ("MIPS: SGI-IP27: Move all shared IP27
declarations to ip27-common.h"), ip27-common.h is included more
than once in ip27-timer.c, remove it.

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: SGI-IP30: Remove R5432_CP0_INTERRUPT_WAR from war.h
Joshua Kinard [Sun, 17 May 2020 23:24:38 +0000 (19:24 -0400)]
MIPS: SGI-IP30: Remove R5432_CP0_INTERRUPT_WAR from war.h

Remove an old macro that no longer exists anywhere else in the tree
that snuck in when IP30 support was added

Signed-off-by: Joshua Kinard <kumba@gentoo.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson: Enable devicetree based probing for 8250 ports in defconfig
Tiezhu Yang [Mon, 18 May 2020 02:11:01 +0000 (10:11 +0800)]
MIPS: Loongson: Enable devicetree based probing for 8250 ports in defconfig

After commit 87fcfa7b7fe6 ("MIPS: Loongson64: Add generic dts"),
there already exists the node and property of Loongson CPU UART0
in loongson3-package.dtsi:

cpu_uart0: serial@1fe001e0 {
        compatible = "ns16550a";
        reg = <0 0x1fe001e0 0x8>;
        clock-frequency = <33000000>;
        interrupt-parent = <&liointc>;
        interrupts = <10 IRQ_TYPE_LEVEL_HIGH>;
        no-loopback-test;
};

In order to support for serial console on the Loongson platform,
add CONFIG_SERIAL_OF_PLATFORM=y to loongson3_defconfig.

With this patch, we can see the following boot messages:

[    1.877745] printk: console [ttyS0] disabled
[    1.881979] 1fe001e0.serial: ttyS0 at MMIO 0x1fe001e0 (irq = 16, base_baud = 2062500) is a 16550A
[    1.890838] printk: console [ttyS0] enabled

And also, we can login normally from the serial console.

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Reviewed-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips/mm: Add page soft dirty tracking
Guoyun Sun [Wed, 29 Apr 2020 08:44:32 +0000 (16:44 +0800)]
mips/mm: Add page soft dirty tracking

User space checkpoint and restart tool (CRIU) needs the page's change
to be soft tracked. This allows to do a pre checkpoint and then dump
only touched pages.

Signed-off-by: Guoyun Sun <sunguoyun@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: define more Loongson CP0.Config6 and CP0.Diag feature bits
WANG Xuerui [Sun, 3 May 2020 10:33:04 +0000 (18:33 +0800)]
MIPS: define more Loongson CP0.Config6 and CP0.Diag feature bits

These are exposed to userland alternatively via the new CPUCFG
instruction on Loongson-3A R4 and above. Add definitions for readback
on older cores.

Signed-off-by: WANG Xuerui <git@xen0n.name>
Cc: Huacai Chen <chenhc@lemote.com>
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: define offsets and known revisions for some CPUCFG features
WANG Xuerui [Sun, 3 May 2020 10:33:03 +0000 (18:33 +0800)]
MIPS: Loongson64: define offsets and known revisions for some CPUCFG features

Add the constants for easier and maintainable composition of CPUCFG
values.

Signed-off-by: WANG Xuerui <git@xen0n.name>
Cc: Huacai Chen <chenhc@lemote.com>
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: fix typos in loongson_regs.h
WANG Xuerui [Sun, 3 May 2020 10:33:02 +0000 (18:33 +0800)]
MIPS: Loongson64: fix typos in loongson_regs.h

Fix some symbol names to align with Loongson's User Manual wording. Also
correct the comment in csr_readq() suggesting the wrong instruction in
use.

Fixes: 6a6f9b7dafd50efc ("MIPS: Loongson: Add CFUCFG&CSR support")
Signed-off-by: WANG Xuerui <git@xen0n.name>
Cc: Huacai Chen <chenhc@lemote.com>
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Remove not used 8250-platform.c
Tiezhu Yang [Sat, 16 May 2020 02:15:49 +0000 (10:15 +0800)]
MIPS: Remove not used 8250-platform.c

When CONFIG_HAVE_STD_PC_SERIAL_PORT is set, there exists build errors
of 8250-platform.c due to linux/module.h is not included.

CONFIG_HAVE_STD_PC_SERIAL_PORT is not used in arch/mips for many years,
8250-platform.c is also not built and used, so it is not necessary to
fix the build errors, just remove the not used file 8250-platform.c and
the related code in Kconfig and Makefile.

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson: Build ATI Radeon GPU driver as module
Tiezhu Yang [Sat, 16 May 2020 02:15:48 +0000 (10:15 +0800)]
MIPS: Loongson: Build ATI Radeon GPU driver as module

When ATI Radeon GPU driver has been compiled directly into the kernel
instead of as a module, we should make sure the firmware for the model
(check available ones in /lib/firmware/radeon) is built-in to the kernel
as well, otherwise there exists the following fatal error during GPU init,
change CONFIG_DRM_RADEON=y to CONFIG_DRM_RADEON=m to fix it.

[    1.900997] [drm] Loading RS780 Microcode
[    1.905077] radeon 0000:01:05.0: Direct firmware load for radeon/RS780_pfp.bin failed with error -2
[    1.914140] r600_cp: Failed to load firmware "radeon/RS780_pfp.bin"
[    1.920405] [drm:r600_init] *ERROR* Failed to load firmware!
[    1.926069] radeon 0000:01:05.0: Fatal error during GPU init
[    1.931729] [drm] radeon: finishing device.

Fixes: 024e6a8b5bb1 ("MIPS: Loongson: Add a Loongson-3 default config file")
Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ath79: ar9331: rename uart to serial node
Oleksij Rempel [Sun, 10 May 2020 06:03:24 +0000 (08:03 +0200)]
MIPS: ath79: ar9331: rename uart to serial node

schema violation was detected by the dtbs_check

Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ath79: ar9331_dpt_module: update led nodes
Oleksij Rempel [Sun, 10 May 2020 06:03:23 +0000 (08:03 +0200)]
MIPS: ath79: ar9331_dpt_module: update led nodes

Fit led nodes to the latest naming schema.

Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: SGI-IP27: Use the _AC() macro in spaces.h
Joshua Kinard [Fri, 15 May 2020 03:24:22 +0000 (23:24 -0400)]
MIPS: SGI-IP27: Use the _AC() macro in spaces.h

The attached patch wraps several of the macros in IP27's spaces.h
header file with the _AC() macro.  This matches most of the other
spaces.h files in the MIPS tree.

Signed-off-by: Joshua Kinard <kumba@gentoo.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ralink: Don't include objects twice
Thomas Bogendoerfer [Thu, 14 May 2020 14:31:22 +0000 (16:31 +0200)]
MIPS: ralink: Don't include objects twice

With the change of platform file inclusion object were included via
platform-y and core-y. Remove the core-y part to fix it.

Fixes: 26bff9eb49201aeb ("MIPS: Only include the platformfile needed")
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Fix builds for VR41xx platforms
Thomas Bogendoerfer [Thu, 14 May 2020 11:07:23 +0000 (13:07 +0200)]
MIPS: Fix builds for VR41xx platforms

Changing inclusion of Platform files, broke VR41xx platforms. Add Makefile
to vr41xx directory and traverse subdirs from it.

Fixes: 26bff9eb49201aeb ("MIPS: Only include the platformfile needed")
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoKVM: MIPS/Emulate: Remove unneeded semicolon
Zou Wei [Thu, 30 Apr 2020 03:19:36 +0000 (11:19 +0800)]
KVM: MIPS/Emulate: Remove unneeded semicolon

Fixes coccicheck warnings:

arch/mips/kvm/emulate.c:1793:3-4: Unneeded semicolon
arch/mips/kvm/emulate.c:1968:3-4: Unneeded semicolon

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zou Wei <zou_wei@huawei.com>
Reviewed-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoKVM: MIPS/VZ: Remove unneeded semicolon
Zou Wei [Thu, 30 Apr 2020 03:14:50 +0000 (11:14 +0800)]
KVM: MIPS/VZ: Remove unneeded semicolon

Fixes coccicheck warnings:

arch/mips/kvm/vz.c:1186:4-5: Unneeded semicolon
arch/mips/kvm/vz.c:1195:3-4: Unneeded semicolon
arch/mips/kvm/vz.c:1949:3-4: Unneeded semicolon
arch/mips/kvm/vz.c:1121:2-3: Unneeded semicolon
arch/mips/kvm/vz.c:2188:3-4: Unneeded semicolon

Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Zou Wei <zou_wei@huawei.com>
Reviewed-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoKVM: MIPS/TLB: Remove Unneeded semicolon in tlb.c
Jason Yan [Tue, 28 Apr 2020 06:32:45 +0000 (14:32 +0800)]
KVM: MIPS/TLB: Remove Unneeded semicolon in tlb.c

Fix the following coccicheck warning:

arch/mips/kvm/tlb.c:472:2-3: Unneeded semicolon
arch/mips/kvm/tlb.c:489:2-3: Unneeded semicolon

Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoKVM: MIPS: use true,false for bool variable
Jason Yan [Wed, 29 Apr 2020 14:09:35 +0000 (22:09 +0800)]
KVM: MIPS: use true,false for bool variable

Fix the following coccicheck warning:

arch/mips/kvm/mips.c:82:1-28: WARNING: Assignment of 0/1 to bool
variable
arch/mips/kvm/mips.c:88:1-28: WARNING: Assignment of 0/1 to bool
variable

Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: dts: mscc: Updated changed name for miim pinctrl function
Lars Povlsen [Wed, 13 May 2020 13:23:47 +0000 (15:23 +0200)]
MIPS: dts: mscc: Updated changed name for miim pinctrl function

This is an add-on patch to the main SoC Sparx5 series
(Message-ID: <20200513125532.24585-1-lars.povlsen@microchip.com>).

This changes the miim pinctrl function name from "miim1" to "miim" due
to refactoring in the driver, obsoleting the instance number.

The change in the driver was to better fit new platforms, as the
instance number is redundant information. Specifically, support for
the Microchip Sparx5 SoC is being submitted, where this change became
necessary.

Reviewed-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Lars Povlsen <lars.povlsen@microchip.com>
Acked-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Fix "make clean" error due to recent changes
Huacai Chen [Wed, 13 May 2020 08:38:41 +0000 (16:38 +0800)]
MIPS: Fix "make clean" error due to recent changes

Commit 26bff9eb49201aeb ("MIPS: Only include the platformfile needed")
moves platform-(CONFIG_XYZ) from arch/mips/xyz/Platform to arch/mips/
Kbuild.platforms. This change causes an error when "make clean":

  ./scripts/Makefile.clean:15: arch/mips/vr41xx/Makefile: No such file or directory
  make[3]: *** No rule to make target `arch/mips/vr41xx/Makefile'.  Stop.
  make[2]: *** [arch/mips/vr41xx] Error 2
  make[1]: *** [_clean_arch/mips] Error 2
  make: *** [sub-make] Error 2

Clean-files are defined in arch/mips/Kbuild:

  obj- := $(platform-)

Due to the movement of platform-(CONFIG_XYZ), "make clean" will enter
arch/mips/vr41xx/ whether CONFIG_MACH_VR41XX is defined or not. Because
there is no Makefile in arch/mips/vr41xx/, "make clean" fails. I don't
know what is the best way to fix it, but it seems like we can avoid this
error by changing the obj- definition:

  obj- := $(platform-y)

Fixes: 26bff9eb49201aeb ("MIPS: Only include the platformfile needed")
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Fix typos in arch/mips/Kbuild.platforms
Huacai Chen [Wed, 13 May 2020 08:38:40 +0000 (16:38 +0800)]
MIPS: Fix typos in arch/mips/Kbuild.platforms

Commit 26bff9eb49201aeb ("MIPS: Only include the platform file needed")
misspelled "txx9" to "tx99", so fix it.

Fixes: 26bff9eb49201aeb ("MIPS: Only include the platform file needed")
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Only include the platform file needed
Thomas Bogendoerfer [Wed, 6 May 2020 16:24:02 +0000 (18:24 +0200)]
MIPS: Only include the platform file needed

Instead of including all Platform files, we simply include the
needed one and avoid clashes with makefile variables.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: VDSO: Allow ld.lld to link the VDSO
Nathan Chancellor [Tue, 28 Apr 2020 22:14:19 +0000 (15:14 -0700)]
MIPS: VDSO: Allow ld.lld to link the VDSO

Currently, when linking with ld.lld, this warning pops up:

    arch/mips/vdso/Makefile:70: MIPS VDSO requires binutils >= 2.25

CONFIG_LD_VERSION is set with scripts/ld-version.sh, which is specific
to GNU ld. It returns 0 for ld.lld so CONFIG_MIPS_LD_CAN_LINK_VDSO does
not set.

ld.lld has a completely different versioning scheme (as it follows
LLVM's versioning) and it does not have the issue mentioned in the
comment block so it should be allowed to link the VDSO.

With this patch, the VDSO successfully links and shows P_MIPS_PC32 in
vgettimeofday.o.

$ llvm-objdump -Dr arch/mips/vdso/vgettimeofday.o | grep R_MIPS_PC32
00000024:  R_MIPS_PC32 _start
000000b0:  R_MIPS_PC32 _start
000002bc:  R_MIPS_PC32 _start
0000036c:  R_MIPS_PC32 _start
00000468:  R_MIPS_PC32 _start

Reported-by: Dmitry Golovin <dima@golovin.in>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/785
Link: https://github.com/llvm/llvm-project/commit/e364e2e9ce50c12eb2bf093560e1a1a8544d455a
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: VDSO: Use $(LD) instead of $(CC) to link VDSO
Nathan Chancellor [Tue, 28 Apr 2020 22:14:18 +0000 (15:14 -0700)]
MIPS: VDSO: Use $(LD) instead of $(CC) to link VDSO

Currently, the VDSO is being linked through $(CC). This does not match
how the rest of the kernel links objects, which is through the $(LD)
variable.

When clang is built in a default configuration, it first attempts to use
the target triple's default linker then the system's default linker,
unless told otherwise through -fuse-ld=... We do not use -fuse-ld=
because it can be brittle and we have support for invoking $(LD)
directly. See commit fe00e50b2db8c ("ARM: 8858/1: vdso: use $(LD)
instead of $(CC) to link VDSO") and commit 691efbedc60d2 ("arm64: vdso:
use $(LD) instead of $(CC) to link VDSO") for examples of doing this in
the VDSO.

Do the same thing here. Replace the custom linking logic with $(cmd_ld)
and ldflags-y so that $(LD) is respected. We need to explicitly add two
flags to the linker that were implicitly passed by the compiler:
-G 0 (which comes from ccflags-vdso) and --eh-frame-hdr.

Before this patch (generated by adding '-v' to VDSO_LDFLAGS):

<gcc_prefix>/libexec/gcc/mips64-linux/9.3.0/collect2 \
-plugin <gcc_prefix>/libexec/gcc/mips64-linux/9.3.0/liblto_plugin.so \
-plugin-opt=<gcc_prefix>/libexec/gcc/mips64-linux/9.3.0/lto-wrapper \
-plugin-opt=-fresolution=/tmp/ccGEi5Ka.res \
--eh-frame-hdr \
-G 0 \
-EB \
-mips64r2 \
-shared \
-melf64btsmip \
-o arch/mips/vdso/vdso.so.dbg.raw \
-L<gcc_prefix>/lib/gcc/mips64-linux/9.3.0/64 \
-L<gcc_prefix>/lib/gcc/mips64-linux/9.3.0 \
-L<gcc_prefix>/lib/gcc/mips64-linux/9.3.0/../../../../mips64-linux/lib \
-Bsymbolic \
--no-undefined \
-soname=linux-vdso.so.1 \
-EB \
--hash-style=sysv \
--build-id \
-T arch/mips/vdso/vdso.lds \
arch/mips/vdso/elf.o \
arch/mips/vdso/vgettimeofday.o \
arch/mips/vdso/sigreturn.o

After this patch:

<gcc_prefix>/bin/mips64-linux-ld \
-m elf64btsmip \
-Bsymbolic \
--no-undefined \
-soname=linux-vdso.so.1 \
-EB \
-nostdlib \
-shared \
-G 0 \
--eh-frame-hdr \
--hash-style=sysv \
--build-id \
-T  arch/mips/vdso/vdso.lds \
arch/mips/vdso/elf.o \
arch/mips/vdso/vgettimeofday.o
arch/mips/vdso/sigreturn.o \
-o arch/mips/vdso/vdso.so.dbg.raw

Note that we leave behind -mips64r2. Turns out that ld ignores it (see
get_emulation in ld/ldmain.c). This is true of current trunk and 2.23,
which is the minimum supported version for the kernel:

https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=ld/ldmain.c;hb=aa4209e7b679afd74a3860ce25659e71cc4847d5#l593
https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=ld/ldmain.c;hb=a55e30b51bc6227d8d41f707654d0a5620978dcf#l641

Before this patch, LD=ld.lld did nothing:

$ llvm-readelf -p.comment arch/mips/vdso/vdso.so.dbg | sed 's/(.*//'
String dump of section '.comment':
[     0] ClangBuiltLinux clang version 11.0.0

After this patch, it does:

$ llvm-readelf -p.comment arch/mips/vdso/vdso.so.dbg | sed 's/(.*//'
String dump of section '.comment':
[     0] Linker: LLD 11.0.0
[    62] ClangBuiltLinux clang version 11.0.0

Link: https://github.com/ClangBuiltLinux/linux/issues/785
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Unconditionally specify '-EB' or '-EL'
Nathan Chancellor [Tue, 28 Apr 2020 22:14:17 +0000 (15:14 -0700)]
MIPS: Unconditionally specify '-EB' or '-EL'

This was all done to work around a GCC bug that has been fixed after
4.2. The kernel requires GCC 4.6 or newer so remove all of these hacks
and just use the traditional flags.

 $ mips64-linux-gcc --version | head -n1
 mips64-linux-gcc (GCC) 4.6.3

 $ mips64-linux-gcc -EB -dM -E -C -x c /dev/null | grep MIPSE
 #define MIPSEB 1
 #define __MIPSEB__ 1
 #define _MIPSEB 1
 #define __MIPSEB 1

 $ mips64-linux-gcc -EL -dM -E -C -x c /dev/null | grep MIPSE
 #define __MIPSEL__ 1
 #define MIPSEL 1
 #define _MIPSEL 1
 #define __MIPSEL 1

This is necessary when converting the MIPS VDSO to use $(LD) instead of
$(CC) to link because the OUTPUT_FORMAT is defaulted to little endian
and only flips to big endian when '-EB' is set on the command line.
There is no issue currently because the compiler explicitly passes
'-EB' or '-EL' to the linker regardless of whether or not it was
provided by the user. Passing '-v' to VDSO_LDFLAGS shows:

<gcc_prefix>/libexec/gcc/mips64-linux/9.3.0/collect2 ... -EB ...

even though '-EB' is nowhere to be found in KBUILD_CFLAGS. The VDSO
Makefile already supports getting '-EB' or '-EL' from KBUILD_CFLAGS
through a filter directive but '-EB' or '-EL' is not always present.

If we do not do this, we will see the following error when compiling
for big endian:

$ make -j$(nproc) ARCH=mips CROSS_COMPILE=mips64-linux- \
  64r2el_defconfig arch/mips/vdso/
...
mips64-linux-ld: arch/mips/vdso/elf.o: compiled for a big endian system
and target is little endian
mips64-linux-ld: arch/mips/vdso/elf.o: endianness incompatible with that
of the selected emulation
mips64-linux-ld: failed to merge target specific data of file
arch/mips/vdso/elf.o
...

Remove this legacy hack and just use '-EB' and '-EL' unconditionally.

Reported-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: VDSO: Move disabling the VDSO logic to Kconfig
Nathan Chancellor [Tue, 28 Apr 2020 22:14:16 +0000 (15:14 -0700)]
MIPS: VDSO: Move disabling the VDSO logic to Kconfig

After commit 9553d16fa671 ("init/kconfig: Add LD_VERSION Kconfig"), we
have access to GNU ld's version at configuration time. As a result, we
can make it clearer under what configuration circumstances the MIPS VDSO
needs to be disabled.

This is a prerequisite for getting rid of the MIPS VDSO binutils
warning and linking the VDSO when LD is ld.lld. Wrapping the call to
ld-ifversion with CONFIG_LD_IS_LLD does not work because the config
values are wiped away during 'make clean'.

Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agokbuild: add CONFIG_LD_IS_LLD
Sami Tolvanen [Tue, 28 Apr 2020 22:14:15 +0000 (15:14 -0700)]
kbuild: add CONFIG_LD_IS_LLD

Similarly to the CC_IS_CLANG config, add LD_IS_LLD to avoid GNU ld
specific logic such as ld-version or ld-ifversion and gain the
ability to select potential features that depend on the linker at
configuration time such as LTO.

Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Acked-by: Masahiro Yamada <masahiroy@kernel.org>
[nc: Reword commit message]
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Reviewed-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: unexport __flush_icache_user_range
Christoph Hellwig [Sun, 10 May 2020 07:54:42 +0000 (09:54 +0200)]
MIPS: unexport __flush_icache_user_range

__flush_icache_user_range is not used in modular code, so unexport it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Remove NEC MARKEINS/EMMA
Thomas Bogendoerfer [Mon, 20 Apr 2020 12:37:24 +0000 (14:37 +0200)]
MIPS: Remove NEC MARKEINS/EMMA

No (active) developer owns this hardware, so let's remove Linux support.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Remove PMC MSP71xx platform
Thomas Bogendoerfer [Mon, 20 Apr 2020 12:30:35 +0000 (14:30 +0200)]
MIPS: Remove PMC MSP71xx platform

No (active) developer owns this hardware, so let's remove Linux support.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Remove support for LASAT
Thomas Bogendoerfer [Mon, 20 Apr 2020 12:22:29 +0000 (14:22 +0200)]
MIPS: Remove support for LASAT

All LASAT has probably gone bad, so let's remove Linux support.

Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Replace zero-length array with flexible-array
Gustavo A. R. Silva [Thu, 7 May 2020 19:00:52 +0000 (14:00 -0500)]
MIPS: Replace zero-length array with flexible-array

The current codebase makes use of the zero-length array language
extension to the C90 standard, but the preferred mechanism to declare
variable-length types such as these ones is a flexible array member[1][2],
introduced in C99:

struct foo {
        int stuff;
        struct boo array[];
};

By making use of the mechanism above, we will get a compiler warning
in case the flexible array does not occur last in the structure, which
will help us prevent some kind of undefined behavior bugs from being
inadvertently introduced[3] to the codebase from now on.

Also, notice that, dynamic memory allocations won't be affected by
this change:

"Flexible array members have incomplete type, and so the sizeof operator
may not be applied. As a quirk of the original implementation of
zero-length arrays, sizeof evaluates to zero."[1]

sizeof(flexible-array-member) triggers a warning because flexible array
members have incomplete type[1]. There are some instances of code in
which the sizeof operator is being incorrectly/erroneously applied to
zero-length arrays and the result is zero. Such instances may be hiding
some bugs. So, this work (flexible-array member conversions) will also
help to get completely rid of those sorts of issues.

This issue was found with the help of Coccinelle.

[1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
[2] https://github.com/KSPP/linux/issues/21
[3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour")

Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson: Add DMA support for LS7A
Tiezhu Yang [Fri, 8 May 2020 08:36:05 +0000 (16:36 +0800)]
MIPS: Loongson: Add DMA support for LS7A

In the current market, the most used bridge chip on the Loongson platform
are RS780E and LS7A, the RS780E bridge chip is already supported by the
mainline kernel.

If use the default implementation of __phys_to_dma() and __dma_to_phys()
in dma-direct.h when CONFIG_ARCH_HAS_PHYS_TO_DMA is not set, it works
well used with LS7A on the Loongson single-way and multi-way platform,
and also works well used with RS780E on the Loongson single-way platform,
but the DMA address will be wrong on the non-node0 used with RS780E on
the Loongson multi-way platform.

Just as the description in the code comment, the devices get node id from
40 bit of HyperTransport bus, so we extract 2 bit node id (bit 44~45) from
48 bit address space of Loongson CPU and embed it into HyperTransport bus
(bit 37-38), this operation can be done only at the software level used
with RS780E on the Loongson multi-way platform, because it has no hardware
function to translate address of node id, this is a hardware compatibility
problem.

Device
    |
    | DMA address
    |
Host Bridge
    |
    | HT bus address (40 bit)
    |
   CPU
    |
    | physical address (48 bit)
    |
   RAM

The LS7A has dma_node_id_offset field in the DMA route config register,
the hardware can use the dma_node_id_offset to translate address of
node id automatically, so we can get correct address when just use the
dma_pfn_offset field in struct device.

For the above reasons, in order to maintain downward compatibility
to support the RS780E bridge chip, it is better to use the platform
dependent implementation of __phys_to_dma() and __dma_to_phys().

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: inst.h: Stop including asm.h to avoid various build failures
Huacai Chen [Fri, 8 May 2020 09:30:12 +0000 (17:30 +0800)]
MIPS: inst.h: Stop including asm.h to avoid various build failures

Commit d339cd02b888eb8 ("MIPS: Move unaligned load/store helpers to
inst.h") causes a lot of build failures because macros in asm.h conflict
with various subsystems. Some of these conflictions has been fixed (such
as LONG, PANIC and PRINT) by adjusting asm.h, but some of them is nearly
impossible to fix (such as PTR and END). The only reason of including
asm.h in inst.h is that we need the PTR macro which is used by unaligned
load/store helpers. So in this patch we define a new PTR_STR macro and
use it to replace STR(PTR), then we can stop including asm.h to avoid
various build failures.

Fixes: d339cd02b888eb8 ("MIPS: Move unaligned load/store helpers to inst.h")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: asm: Rename some macros to avoid build errors
Huacai Chen [Thu, 7 May 2020 09:49:18 +0000 (17:49 +0800)]
MIPS: asm: Rename some macros to avoid build errors

Use ASM_ prefix to rename some macros (PANIC and PRINT), in order to
avoid build errors (all users are updated as well):

1, PANIC conflicts with drivers/scsi/smartpqi/smartpqi_init.c
2, PRINT conflicts with net/netfilter/nf_conntrack_h323_asn1.c and net/
   mac80211/debugfs_sta.c

Fixes: d339cd02b888eb8 ("MIPS: Move unaligned load/store helpers to inst.h")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: cm: Add L2 ECC/parity errors reporting
Serge Semin [Wed, 6 May 2020 17:42:23 +0000 (20:42 +0300)]
mips: cm: Add L2 ECC/parity errors reporting

According to the MIPS32 InterAptiv software manual error codes 24 - 26
of CM2 indicate L2 ECC/parity error with switching to a corresponding
errors info fields. This patch provides these errors parsing code,
which handles the read/write uncorrectable and correctable ECC/parity
errors, and prints instruction causing the fault, RAM array type, cache
way/dword and syndrome associated with the faulty data.

Co-developed-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: linux-pm@vger.kernel.org
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: cm: Fix an invalid error code of INTVN_*_ERR
Serge Semin [Wed, 6 May 2020 17:42:22 +0000 (20:42 +0300)]
mips: cm: Fix an invalid error code of INTVN_*_ERR

Commit 3885c2b463f6 ("MIPS: CM: Add support for reporting CM cache
errors") adds cm2_causes[] array with map of error type ID and
pointers to the short description string. There is a mistake in
the table, since according to MIPS32 manual CM2_ERROR_TYPE = {17,18}
correspond to INTVN_WR_ERR and INTVN_RD_ERR, while the table
claims they have {0x17,0x18} codes. This is obviously hex-dec
copy-paste bug. Moreover codes {0x18 - 0x1a} indicate L2 ECC errors.

Fixes: 3885c2b463f6 ("MIPS: CM: Add support for reporting CM cache errors")
Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Cc: Alexey Malahov <Alexey.Malahov@baikalelectronics.ru>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Rob Herring <robh+dt@kernel.org>
Cc: linux-pm@vger.kernel.org
Cc: devicetree@vger.kernel.org
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: Replace <linux/clk-provider.h> by <linux/of_clk.h>
Geert Uytterhoeven [Wed, 6 May 2020 09:04:52 +0000 (11:04 +0200)]
MIPS: ingenic: Replace <linux/clk-provider.h> by <linux/of_clk.h>

The JZ4740 setup code is not a clock provider, and just needs to call
of_clk_init().

Hence it can include <linux/of_clk.h> instead of <linux/clk-provider.h>.

Fixes: f932449c11dabb4b ("MIPS: ingenic: Drop obsolete code, merge the rest in setup.c")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Use fallthrough for arch/mips
Liangliang Huang [Mon, 4 May 2020 08:51:29 +0000 (16:51 +0800)]
MIPS: Use fallthrough for arch/mips

Convert the various /* fallthrough */ comments to the pseudo-keyword
fallthrough;

Done via script:
https://lore.kernel.org/lkml/b56602fcf79f849e733e7b521bb0e17895d390fa.1582230379.git.joe@perches.com/

Signed-off-by: Liangliang Huang <huangll@lemote.com>
Reviewed-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Truncate link address into 32bit for 32bit kernel
Jiaxun Yang [Wed, 6 May 2020 05:52:45 +0000 (13:52 +0800)]
MIPS: Truncate link address into 32bit for 32bit kernel

LLD failed to link vmlinux with 64bit load address for 32bit ELF
while bfd will strip 64bit address into 32bit silently.
To fix LLD build, we should truncate load address provided by platform
into 32bit for 32bit kernel.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/786
Link: https://sourceware.org/bugzilla/show_bug.cgi?id=25784
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Tested-by: Nathan Chancellor <natechancellor@gmail.com>
Cc: Maciej W. Rozycki <macro@linux-mips.org>
Tested-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Remove dead code in pci.h
Jason Yan [Wed, 6 May 2020 13:42:50 +0000 (21:42 +0800)]
MIPS: Remove dead code in pci.h

This code has been marked dead for more than 10 years. Seems no need to
keep it now.

Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: CFE: Remove dead code in cfe_getfwinfo()
Jason Yan [Wed, 6 May 2020 13:43:07 +0000 (21:43 +0800)]
MIPS: CFE: Remove dead code in cfe_getfwinfo()

This code has been marked dead since the beginning of the git history.
Seems no need to keep it now.

Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Octeon: Remove dead code in __cvmx_helper_npi_probe()
Jason Yan [Wed, 6 May 2020 13:43:15 +0000 (21:43 +0800)]
MIPS: Octeon: Remove dead code in __cvmx_helper_npi_probe()

This code has been marked dead for more than 10 years. Seems no need to
keep it now.

Signed-off-by: Jason Yan <yanaijie@huawei.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: tools: Move "returns" after "loongson3-llsc-check"
Tiezhu Yang [Sun, 3 May 2020 00:05:02 +0000 (08:05 +0800)]
MIPS: tools: Move "returns" after "loongson3-llsc-check"

Just move "returns" after "loongson3-llsc-check", no function changes.

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: perf: Remove unnecessary "fallthrough" pseudo keywords
Huacai Chen [Sat, 2 May 2020 10:46:24 +0000 (18:46 +0800)]
MIPS: perf: Remove unnecessary "fallthrough" pseudo keywords

The last branch of switch-case doesn't need a "fallthrough" pseudo
keyword, and it will cause errors when building a kernel with -Werror:

   arch/mips/kernel/perf_event_mipsxx.c: In function 'reset_counters':
   include/linux/compiler_attributes.h:200:41: error: attribute 'fallthrough' not preceding a case label or default label [-Werror]
     200 | # define fallthrough                    __attribute__((__fallthrough__))
         |                                         ^~~~~~~~~~~~~
>> arch/mips/kernel/perf_event_mipsxx.c:932:3: note: in expansion of macro 'fallthrough'
     932 |   fallthrough;
         |   ^~~~~~~~~~~
   arch/mips/kernel/perf_event_mipsxx.c: In function 'loongson3_reset_counters':
   include/linux/compiler_attributes.h:200:41: error: attribute 'fallthrough' not preceding a case label or default label [-Werror]
     200 | # define fallthrough                    __attribute__((__fallthrough__))
         |                                         ^~~~~~~~~~~~~
   arch/mips/kernel/perf_event_mipsxx.c:903:3: note: in expansion of macro 'fallthrough'
     903 |   fallthrough;
         |   ^~~~~~~~~~~
   cc1: all warnings being treated as errors

Fix it by removing unnecessary "fallthrough" pseudo keywords.

Fixes: e9dfbaaeef1c9fe ("MIPS: perf: Add hardware perf events support for new Loongson-3")
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agomips: Drop CONFIG_MTD_M25P80 in various defconfig files
Bin Meng [Sat, 2 May 2020 04:25:49 +0000 (21:25 -0700)]
mips: Drop CONFIG_MTD_M25P80 in various defconfig files

Drop CONFIG_MTD_M25P80 that was removed in
commit b35b9a10362d ("mtd: spi-nor: Move m25p80 code in spi-nor.c")

Signed-off-by: Bin Meng <bin.meng@windriver.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: Correct TLB type for Loongson-3 Classic
Jiaxun Yang [Thu, 30 Apr 2020 16:48:02 +0000 (00:48 +0800)]
MIPS: Loongson64: Correct TLB type for Loongson-3 Classic

Huacai just informed me that some early Loongson-3A2000 had wrong
TLB type in Config0 register. That means we have to correct it via
PRID.

It looks like I shoudn't drop MIPS_CPU_FTLB flag in PRID case for
Loongson-3 Classic.

Fixes: da1bd29742b1 ("MIPS: Loongson64: Probe CPU features via CPUCFG")
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Reported-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: Probe CPU features via CPUCFG
Jiaxun Yang [Thu, 30 Apr 2020 03:18:35 +0000 (11:18 +0800)]
MIPS: Loongson64: Probe CPU features via CPUCFG

CPUCFG is a Loongson self-defined instruction used to mark CPU
features for Loongson processors started from Loongson-3A4000.

Slightly adjust cpu_probe_loongson function as well. Remove features
that already probed via decode_configs in processor's PRID case
and add a comment about TLBINV.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: perf: Add hardware perf events support for new Loongson-3
Huacai Chen [Thu, 30 Apr 2020 09:45:16 +0000 (17:45 +0800)]
MIPS: perf: Add hardware perf events support for new Loongson-3

New Loongson-3 means Loongson-3A R2 (Loongson-3A2000) and newer CPUs.
Loongson-3 processors have three types of PMU types (so there are three
event maps): Loongson-3A1000/Loonngson-3B1000/Loongson-3B1500 is Type-1,
Loongson-3A2000/Loongson-3A3000 is Type-2, Loongson-3A4000+ is Type-3.

Signed-off-by: Huacai Chen <chenhc@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: tools: Show result for loongson3-llsc-check
Tiezhu Yang [Thu, 30 Apr 2020 12:36:24 +0000 (20:36 +0800)]
MIPS: tools: Show result for loongson3-llsc-check

It is better to show the result before loongson3-llsc-check exit,
otherwise we can see nothing if the return status is EXIT_SUCCESS,
it seems confusing.

E.g. without this patch:

[loongson@localhost tools]$ ./loongson3-llsc-check ../../../vmlinux
[loongson@localhost tools]$

With this patch:

[loongson@localhost tools]$ ./loongson3-llsc-check ../../../vmlinux
loongson3-llsc-check returns success
[loongson@localhost tools]$

Signed-off-by: Tiezhu Yang <yangtiezhu@loongson.cn>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: Drop obsolete code, merge the rest in setup.c
Paul Cercueil [Mon, 13 Apr 2020 15:26:33 +0000 (17:26 +0200)]
MIPS: ingenic: Drop obsolete code, merge the rest in setup.c

Drop a bootload of 10-years-old dirty code, that is not used anymore, as
it has been replaced with clean code over the ages.

Merge the very few bits left inside setup.c, so that everything is clean
and tidy now.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: GCW0: Update defconfig
Paul Cercueil [Mon, 13 Apr 2020 15:26:32 +0000 (17:26 +0200)]
MIPS: ingenic: GCW0: Update defconfig

Enable support for the new hardware that was added in the devicetree.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: CI20: enable OST, PWM drivers in defconfig
Paul Cercueil [Mon, 13 Apr 2020 15:26:31 +0000 (17:26 +0200)]
MIPS: ingenic: CI20: enable OST, PWM drivers in defconfig

The OST driver provides a clocksource and sched_clock that are much more
accurate than the default ones.

The PWM driver allows to use the PWM pins on the external header of the
board.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: DTS: Update GCW0 support
Paul Cercueil [Mon, 13 Apr 2020 15:26:30 +0000 (17:26 +0200)]
MIPS: ingenic: DTS: Update GCW0 support

Add support for the face buttons, the ACT8600 PMUC, the LCD panel
with backlight, the rumble, internal/external SD readers, and other
things.

Note that the otg-phy node was dropped in the process as it was neither
useful nor used, and was inside a non-compliant board "bus".

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: DTS: Update JZ4770 support
Paul Cercueil [Mon, 13 Apr 2020 15:26:29 +0000 (17:26 +0200)]
MIPS: ingenic: DTS: Update JZ4770 support

Add support for the RTC, AIC, CODEC, MMC 0/1/2, ADC, GPU, LCD,
USB OTG, USB PHY controllers.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: DTS: Add nodes for the watchdog/PWM/OST
Paul Cercueil [Mon, 13 Apr 2020 15:26:28 +0000 (17:26 +0200)]
MIPS: ingenic: DTS: Add nodes for the watchdog/PWM/OST

Add the TCU nodes to the JZ4780, JZ4770 and JZ4740 devicetree files.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: DTS: Respect cell count of common properties
Paul Cercueil [Mon, 13 Apr 2020 15:26:27 +0000 (17:26 +0200)]
MIPS: ingenic: DTS: Respect cell count of common properties

If N fields of X cells should be provided, then that's what the
devicetree should represent, instead of having one single field of
(N*X) cells.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: ingenic: DTS: Fix invalid value in #dma-cells
Paul Cercueil [Mon, 13 Apr 2020 15:26:26 +0000 (17:26 +0200)]
MIPS: ingenic: DTS: Fix invalid value in #dma-cells

The driver requires two cells and not just one.

Since these nodes are both disabled as no hardware currently use them,
this fix does not really requires a Fixes: tag.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
4 years agoMIPS: Loongson64: Switch the order of RS780E and LS7A
Liangliang Huang [Wed, 29 Apr 2020 09:04:17 +0000 (17:04 +0800)]
MIPS: Loongson64: Switch the order of RS780E and LS7A

Sort the members of enum in alphabetical order is better to avoid
duplicate mistakes (because the list may be grow very large), so
fix it by exchanging the order.

Signed-off-by: Liangliang Huang <huangll@lemote.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>