| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Loongson-3's COP2 is Multi-Media coprocessor, it is disabled in kernel
mode by default. However, gslq/gssq (16-bytes load/store instructions)
overrides the instruction format of lwc2/swc2. If we wan't to use gslq/
gssq for optimization in kernel, we should enable COP2 usage in kernel.
Please pay attention that in this patch we only enable COP2 in kernel,
which means it will lose ST0_CU2 when a process go to user space (try
to use COP2 in user space will trigger an exception and then grab COP2,
which is similar to FPU). And as a result, we need to modify the context
switching code because the new scheduled process doesn't contain ST0_CU2
in its THERAD_STATUS probably.
For zboot, we disable gslq/gssq be generated by toolchain.
Signed-off-by: Huacai Chen <[email protected]>
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
Use a new config option to enable MIPS 34K ITLB workaround and remove
define from different war.h files.
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
Remove another unused MIPS platform.
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
Newer Loongson cores (Loongson-3A R2 and newer) use the
implementation-dependent ExcCode 16 to signal Loongson-specific
exceptions. The extended cause is put in the non-standard CP0.Diag1
register which is CP0 Register 22 Select 1, called GSCause in Loongson
manuals. Inside is an exception code bitfield called GSExcCode, only
codes 0 to 6 inclusive are documented (so far, in the Loongson 3A3000
User Manual, Volume 2).
During experiments, it was found that some undocumented unprivileged
instructions can trigger the also-undocumented GSExcCode 8 on Loongson
3A4000. Processor state is not corrupted, but we cannot continue without
further knowledge, and Loongson is not providing that information as of
this writing. So we send SIGILL on seeing this exception code to thwart
easy local DoS attacks.
Other exception codes are made fatal, partly because of insufficient
knowledge, also partly because they are not as easily reproduced. None
of them are encountered in the wild with upstream kernels and userspace
so far.
Some older cores (Loongson-3A1000 and Loongson-3B1500) have ExcCode 16
too, but the semantic is equivalent to GSExcCode 0. Because the
respective manuals did not mention the CP0.Diag1 register or its read
behavior, these cores are not covered in this patch, as MFC0 from
non-existent CP0 registers is UNDEFINED according to the MIPS
architecture spec.
Reviewed-by: Huacai Chen <[email protected]>
Signed-off-by: WANG Xuerui <[email protected]>
Cc: Huacai Chen <[email protected]>
Cc: Jiaxun Yang <[email protected]>
Cc: Tiezhu Yang <[email protected]>
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
This 32-bit CP0 register is named GSCause in Loongson manuals. It carries
Loongson extended exception information. We name it Diag1 because we
fear the "GSCause" name might get changed in the future.
Reviewed-by: Huacai Chen <[email protected]>
Signed-off-by: WANG Xuerui <[email protected]>
Cc: Huacai Chen <[email protected]>
Cc: Jiaxun Yang <[email protected]>
Cc: Tiezhu Yang <[email protected]>
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
Other vendor-defined registers use the vendor name as a prefix, not an
infix, so unify the naming style of CP0.Config6 bits.
Suggested-by: Maciej W. Rozycki" <[email protected]>
Signed-off-by: Huacai Chen <[email protected]>
Reviewed-by: Maciej W. Rozycki <[email protected]>
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
Loongson-3 has CONFIG6 and DIAG registers which need to be emulated.
CONFIG6 is mostly used to enable/disable FTLB and SFB, while DIAG is
mostly used to flush BTB, ITLB, DTLB, VTLB and FTLB.
Acked-by: Thomas Bogendoerfer <[email protected]>
Reviewed-by: Aleksandar Markovic <[email protected]>
Signed-off-by: Huacai Chen <[email protected]>
Co-developed-by: Jiaxun Yang <[email protected]>
Message-Id: <[email protected]>
Signed-off-by: Paolo Bonzini <[email protected]>
|
|
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 <[email protected]>
Cc: Serge Semin <[email protected]>
Signed-off-by: Huacai Chen <[email protected]>
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
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 <[email protected]>
Cc: Alexey Malahov <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
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 <[email protected]>
Signed-off-by: Alexey Malahov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
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 <[email protected]>
Signed-off-by: Alexey Malahov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
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 <[email protected]>
Signed-off-by: Alexey Malahov <[email protected]>
Signed-off-by: Serge Semin <[email protected]>
Cc: Thomas Bogendoerfer <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Arnd Bergmann <[email protected]>
Cc: Rob Herring <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
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 <[email protected]>
Cc: Huacai Chen <[email protected]>
Cc: Jiaxun Yang <[email protected]>
Signed-off-by: Thomas Bogendoerfer <[email protected]>
|
|
MAC2008 means the processor implemented IEEE754 style Fused MADD
instruction. It was introduced in Release3 but removed in Release5.
The toolchain support of MAC2008 have never landed except for Loongson
processors.
This patch aimed to disabled the MAC2008 if it's optional. For
MAC2008 only processors, we corrected math-emu behavior to align
with actual hardware behavior.
Signed-off-by: Jiaxun Yang <[email protected]>
[[email protected]: Fixup MIPSr2-r5 check in cpu_set_fpu_2008.]
Signed-off-by: Paul Burton <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
|
|
JZ4760/JZ4770/JZ4775/X1000/X1500 has an abandoned huge page tlb,
this mode is not compatible with the MIPS standard, it will cause
tlbmiss and into an infinite loop (line 21 in the tlb-funcs.S)
when starting the init process. write 0xa9000000 to cp0 register 5
sel 4 to disable this function to prevent getting stuck. Confirmed
by Ingenic, this operation will not adversely affect processors
without HPTLB function.
Signed-off-by: Zhou Yanjie <[email protected]>
Acked-by: Paul Cercueil <[email protected]>
Signed-off-by: Paul Burton <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
|
|
In order to further reduce power consumption, the XBurst core
by default attempts to avoid branch target buffer lookups by
detecting & special casing loops. This feature will cause
BogoMIPS and lpj calculate in error. Set cp0 config7 bit 4 to
disable this feature.
Signed-off-by: Zhou Yanjie <[email protected]>
Signed-off-by: Paul Burton <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
|
|
Introduce support for using MemoryMapIDs (MMIDs) as an alternative to
Address Space IDs (ASIDs). The major difference between the two is that
MMIDs are global - ie. an MMID uniquely identifies an address space
across all coherent CPUs. In contrast ASIDs are non-global per-CPU IDs,
wherein each address space is allocated a separate ASID for each CPU
upon which it is used. This global namespace allows a new GINVT
instruction be used to globally invalidate TLB entries associated with a
particular MMID across all coherent CPUs in the system, removing the
need for IPIs to invalidate entries with separate ASIDs on each CPU.
The allocation scheme used here is largely borrowed from arm64 (see
arch/arm64/mm/context.c). In essence we maintain a bitmap to track
available MMIDs, and MMIDs in active use at the time of a rollover to a
new MMID version are preserved in the new version. The allocation scheme
requires efficient 64 bit atomics in order to perform reasonably, so
this support depends upon CONFIG_GENERIC_ATOMIC64=n (ie. currently it
will only be included in MIPS64 kernels).
The first, and currently only, available CPU with support for MMIDs is
the MIPS I6500. This CPU supports 16 bit MMIDs, and so for now we cap
our MMIDs to 16 bits wide in order to prevent the bitmap growing to
absurd sizes if any future CPU does implement 32 bit MMIDs as the
architecture manuals suggest is recommended.
When MMIDs are in use we also make use of GINVT instruction which is
available due to the global nature of MMIDs. By executing a sequence of
GINVT & SYNC 0x14 instructions we can avoid the overhead of an IPI to
each remote CPU in many cases. One complication is that GINVT will
invalidate wired entries (in all cases apart from type 0, which targets
the entire TLB). In order to avoid GINVT invalidating any wired TLB
entries we set up, we make sure to create those entries using a reserved
MMID (0) that we never associate with any address space.
Also of note is that KVM will require further work in order to support
MMIDs & GINVT, since KVM is involved in allocating IDs for guests & in
configuring the MMU. That work is not part of this patch, so for now
when MMIDs are in use KVM is disabled.
Signed-off-by: Paul Burton <[email protected]>
Cc: [email protected]
|
|
Add a family of ginvt_* functions making it easy to emit a GINVT
instruction to globally invalidate TLB entries. We make use of the
_ASM_MACRO infrastructure to support emitting the instructions even if
the assembler isn't new enough to support them natively.
An associated STYPE_GINV definition & sync_ginv() function are added to
emit a sync instruction of type 0x14, which operates as a completion
barrier for these new GINVT (and GINVI) instructions.
Signed-off-by: Paul Burton <[email protected]>
Cc: [email protected]
|
|
We currently have 2 commonly used methods for switching ISA within
assembly code, then restoring the original ISA.
1) Using a pair of .set push & .set pop directives. For example:
.set push
.set mips32r2
<some_insn>
.set pop
2) Using .set mips0 to restore the ISA originally specified on the
command line. For example:
.set mips32r2
<some_insn>
.set mips0
Unfortunately method 2 does not work with nanoMIPS toolchains, where the
assembler rejects the .set mips0 directive like so:
Error: cannot change ISA from nanoMIPS to mips0
In preparation for supporting nanoMIPS builds, switch all instances of
method 2 in generic non-platform-specific code to use push & pop as in
method 1 instead. The .set push & .set pop is arguably cleaner anyway,
and if nothing else it's good to consistently use one method.
Signed-off-by: Paul Burton <[email protected]>
Patchwork: https://patchwork.linux-mips.org/patch/21037/
Cc: [email protected]
|
|
Currently we hardcode a list of files for which we specify that the
toolchain has DSP ASE support when building for MIPSr2 only. This has a
number of problems:
1) It doesn't actually ensure that the toolchain supports the DSP ASE
at all.
2) It's fragile if we try to use DSP ASE macros in other files.
3) It makes no provision for MIPSr6 & later systems which also support
the DSP ASE & end up using the .word directive implementation of
the DSP macros.
Fix this by detecting assembler support for the DSP ASE globally, not
just for a small set of files, and not just for MIPSr2. This now exposes
use of toolchain DSP support to kernel builds targeting MIPSr1 and
older, so we add .set MIPS_ISA_LEVEL directives prior to all .set dsp
directives in order to prevent the assembler from complaining that the
DSP ASE is only supported with MIPSr2 & higher.
Signed-off-by: Paul Burton <[email protected]>
Patchwork: https://patchwork.linux-mips.org/patch/20901/
Cc: [email protected]
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux
Pull MIPS updates from Paul Burton:
"Here are the main MIPS changes for 4.19.
An overview of the general architecture changes:
- Massive DMA ops refactoring from Christoph Hellwig (huzzah for
deleting crufty code!).
- We introduce NT_MIPS_DSP & NT_MIPS_FP_MODE ELF notes &
corresponding regsets to expose DSP ASE & floating point mode state
respectively, both for live debugging & core dumps.
- We better optimize our code by hard-coding cpu_has_* macros at
compile time where their values are known due to the ISA revision
that the kernel build is targeting.
- The EJTAG exception handler now better handles SMP systems, where
it was previously possible for CPUs to clobber a register value
saved by another CPU.
- Our implementation of memset() gained a couple of fixes for MIPSr6
systems to return correct values in some cases where stores fault.
- We now implement ioremap_wc() using the uncached-accelerated cache
coherency attribute where supported, which is detected during boot,
and fall back to plain uncached access where necessary. The
MIPS-specific (and unused in tree) ioremap_uncached_accelerated() &
ioremap_cacheable_cow() are removed.
- The prctl(PR_SET_FP_MODE, ...) syscall is better supported for SMP
systems by reworking the way we ensure remote CPUs that may be
running threads within the affected process switch mode.
- Systems using the MIPS Coherence Manager will now set the
MIPS_IC_SNOOPS_REMOTE flag to avoid some unnecessary cache
maintenance overhead when flushing the icache.
- A few fixes were made for building with clang/LLVM, which now
sucessfully builds kernels for many of our platforms.
- Miscellaneous cleanups all over.
And some platform-specific changes:
- ar7 gained stubs for a few clock API functions to fix build
failures for some drivers.
- ath79 gained support for a few new SoCs, a few fixes & better
gpio-keys support.
- Ci20 now exposes its SPI bus using the spi-gpio driver.
- The generic platform can now auto-detect a suitable value for
PHYS_OFFSET based upon the memory map described by the device tree,
allowing us to avoid wasting memory on page book-keeping for
systems where RAM starts at a non-zero physical address.
- Ingenic systems using the jz4740 platform code now link their
vmlinuz higher to allow for kernels of a realistic size.
- Loongson32 now builds the kernel targeting MIPSr1 rather than
MIPSr2 to avoid CPU errata.
- Loongson64 gains a couple of fixes, a workaround for a write
buffering issue & support for the Loongson 3A R3.1 CPU.
- Malta now uses the piix4-poweroff driver to handle powering down.
- Microsemi Ocelot gained support for its SPI bus & NOR flash, its
second MDIO bus and can now be supported by a FIT/.itb image.
- Octeon saw a bunch of header cleanups which remove a lot of
duplicate or unused code"
* tag 'mips_4.19' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux: (123 commits)
MIPS: Remove remnants of UASM_ISA
MIPS: netlogic: xlr: Remove erroneous check in nlm_fmn_send()
MIPS: VDSO: Force link endianness
MIPS: Always specify -EB or -EL when using clang
MIPS: Use dins to simplify __write_64bit_c0_split()
MIPS: Use read-write output operand in __write_64bit_c0_split()
MIPS: Avoid using array as parameter to write_c0_kpgd()
MIPS: vdso: Allow clang's --target flag in VDSO cflags
MIPS: genvdso: Remove GOT checks
MIPS: Remove obsolete MIPS checks for DST node "chosen@0"
MIPS: generic: Remove input symbols from defconfig
MIPS: Delete unused code in linux32.c
MIPS: Remove unused sys_32_mmap2
MIPS: Remove nabi_no_regargs
mips: dts: mscc: enable spi and NOR flash support on ocelot PCB123
mips: dts: mscc: Add spi on Ocelot
MIPS: Loongson: Merge load addresses
MIPS: Loongson: Set Loongson32 to MIPS32R1
MIPS: mscc: ocelot: add interrupt controller properties to GPIO controller
MIPS: generic: Select MIPS_AUTO_PFN_OFFSET
...
|
|
The code in __write_64bit_c0_split() is used by MIPS32 kernels running
on MIPS64 CPUs to write a 64-bit value to a 64-bit coprocessor 0
register using a single 64-bit dmtc0 instruction. It does this by
combining the 2x 32-bit registers used to hold the 64-bit value into a
single register, which in the existing code involves three steps:
1) Zero extend register A which holds bits 31:0 of our data, since it
may have previously held a sign-extended value.
2) Shift register B which holds bits 63:32 of our data in bits 31:0
left by 32 bits, such that the bits forming our data are in the
position they'll be in the final 64-bit value & bits 31:0 of the
register are zero.
3) Or the two registers together to form the 64-bit value in one
64-bit register.
From MIPS r2 onwards we have a dins instruction which can effectively
perform all 3 of those steps using a single instruction.
Add a path for MIPS r2 & beyond which uses dins to take bits 31:0 from
register B & insert them into bits 63:32 of register A, giving us our
full 64-bit value in register A with one instruction.
Since we know that MIPS r2 & above support the sel field for the dmtc0
instruction, we don't bother special casing sel==0. Omiting the sel
field would assemble to exactly the same instruction as when we
explicitly specify that it equals zero.
Signed-off-by: Paul Burton <[email protected]>
|
|
Commit c22c80431055 ("MIPS: Fix input modify in
__write_64bit_c0_split()") modified __write_64bit_c0_split() constraints
such that we have both an input & an output which we hope to assign to
the same registers, and modify the output rather than incorrectly
clobbering an input.
The way in which we use both an output & an input parameter with the
input constrained to share the output registers is a little convoluted &
also problematic for clang, which complains if the input & output values
have different widths. For example:
In file included from kernel/fork.c:98:
./arch/mips/include/asm/mmu_context.h:149:19: error: unsupported
inline asm: input with type 'unsigned long' matching output with
type 'unsigned long long'
write_c0_entryhi(cpu_asid(cpu, next));
~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~
./arch/mips/include/asm/mmu_context.h:93:2: note: expanded from macro
'cpu_asid'
(cpu_context((cpu), (mm)) & cpu_asid_mask(&cpu_data[cpu]))
^
./arch/mips/include/asm/mipsregs.h:1617:65: note: expanded from macro
'write_c0_entryhi'
#define write_c0_entryhi(val) __write_ulong_c0_register($10, 0, val)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~
./arch/mips/include/asm/mipsregs.h:1430:39: note: expanded from macro
'__write_ulong_c0_register'
__write_64bit_c0_register(reg, sel, val); \
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~
./arch/mips/include/asm/mipsregs.h:1400:41: note: expanded from macro
'__write_64bit_c0_register'
__write_64bit_c0_split(register, sel, value); \
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~
./arch/mips/include/asm/mipsregs.h:1498:13: note: expanded from macro
'__write_64bit_c0_split'
: "r,0" (val)); \
^~~
We can both fix this build failure & simplify the code somewhat by
assigning the __tmp variable with the input value in C prior to our
inline assembly, and then using a single read-write output operand (ie.
a constraint beginning with +) to provide this value to our assembly.
Signed-off-by: Paul Burton <[email protected]>
|
|
This reverts commit 2a027b47dba6 ("MIPS: BCM47XX: Enable 74K Core
ExternalSync for PCIe erratum").
Enabling ExternalSync caused a regression for BCM4718A1 (used e.g. in
Netgear E3000 and ASUS RT-N16): it simply hangs during PCIe
initialization. It's likely that BCM4717A1 is also affected.
I didn't notice that earlier as the only BCM47XX devices with PCIe I
own are:
1) BCM4706 with 2 x 14e4:4331
2) BCM4706 with 14e4:4360 and 14e4:4331
it appears that BCM4706 is unaffected.
While BCM5300X-ES300-RDS.pdf seems to document that erratum and its
workarounds (according to quotes provided by Tokunori) it seems not even
Broadcom follows them.
According to the provided info Broadcom should define CONF7_ES in their
SDK's mipsinc.h and implement workaround in the si_mips_init(). Checking
both didn't reveal such code. It *could* mean Broadcom also had some
problems with the given workaround.
Signed-off-by: Rafał Miłecki <[email protected]>
Signed-off-by: Paul Burton <[email protected]>
Reported-by: Michael Marley <[email protected]>
Patchwork: https://patchwork.linux-mips.org/patch/20032/
URL: https://bugs.openwrt.org/index.php?do=details&task_id=1688
Cc: Tokunori Ikegami <[email protected]>
Cc: Hauke Mehrtens <[email protected]>
Cc: Chris Packham <[email protected]>
Cc: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
|
|
Defines CP0_CONFIG3, CP0_CONFIG6, CP0_PAGEGRAIN and use them in
kernel-entry-init.h for Loongson64.
Signed-off-by: Huacai Chen <[email protected]>
Signed-off-by: Paul Burton <[email protected]>
Patchwork: https://patchwork.linux-mips.org/patch/19264/
Cc: Ralf Baechle <[email protected]>
Cc: James Hogan <[email protected]>
Cc: [email protected]
Cc: Fuxin Zhang <[email protected]>
Cc: Zhangjin Wu <[email protected]>
Cc: Huacai Chen <[email protected]>
|
|
The erratum and workaround are described by BCM5300X-ES300-RDS.pdf as
below.
R10: PCIe Transactions Periodically Fail
Description: The BCM5300X PCIe does not maintain transaction ordering.
This may cause PCIe transaction failure.
Fix Comment: Add a dummy PCIe configuration read after a PCIe
configuration write to ensure PCIe configuration access
ordering. Set ES bit of CP0 configu7 register to enable
sync function so that the sync instruction is functional.
Resolution: hndpci.c: extpci_write_config()
hndmips.c: si_mips_init()
mipsinc.h CONF7_ES
This is fixed by the CFE MIPS bcmsi chipset driver also for BCM47XX.
Also the dummy PCIe configuration read is already implemented in the
Linux BCMA driver.
Enable ExternalSync in Config7 when CONFIG_BCMA_DRIVER_PCI_HOSTMODE=y
too so that the sync instruction is externalised.
Signed-off-by: Tokunori Ikegami <[email protected]>
Reviewed-by: Paul Burton <[email protected]>
Acked-by: Hauke Mehrtens <[email protected]>
Cc: Chris Packham <[email protected]>
Cc: Rafał Miłecki <[email protected]>
Cc: [email protected]
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/19461/
Signed-off-by: James Hogan <[email protected]>
|
|
Processors implementing the MIPS MT ASE may have performance counters
implemented per core or per TC. Processors implemented by MIPS
Technologies signify presence per TC through a bit in the implementation
specific Config7 register. Currently the code which probes for their
presence blindly reads a magic number corresponding to this bit, despite
it potentially having a different meaning in the CPU implementation.
Since CPU features are generally detected by cpu-probe.c, perform the
detection here instead. Introduce cpu_set_mt_per_tc_perf which checks
the bit in config7 and call it from MIPS CPUs known to implement this
bit and the MT ASE, specifically, the 34K, 1004K and interAptiv.
Once the presence of the per-tc counter is indicated in cpu_data, tests
for it can be updated to use this flag.
Suggested-by: James Hogan <[email protected]>
Signed-off-by: Matt Redfearn <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Florian Fainelli <[email protected]>
Cc: Matt Redfearn <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: Maciej W. Rozycki <[email protected]>
Cc: [email protected]>
Patchwork: https://patchwork.linux-mips.org/patch/19136/
Signed-off-by: James Hogan <[email protected]>
|
|
Indicate that CRC32 and CRC32C instuctions are supported by the CPU
through elf_hwcap flags.
This will be used by a follow-up commit that introduces crc32(c) crypto
acceleration modules and is required by GENERIC_CPU_AUTOPROBE feature.
Signed-off-by: Marcin Nowakowski <[email protected]>
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/18600/
|
|
Now that we are using assembler macros to implement XPA instructions on
toolchains which don't support them, pass Cop0 register names to the
__{readx,writex}_32bit_c0_register macros in $n format rather than
register numbers. Also pass a register select which may be useful in
future (for example for MemoryMapID field of WatchHi registers on
I6500).
This is to make them consistent with the normal Cop0 register access
macros which they were originally based on.
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17777/
|
|
Tweak __writex_32bit_c0_register() to allow the compiler to use $0 (the
zero register) as an input to the mthc0 instruction.
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17774/
|
|
Utilise XPA instructions MFHC0 & MTHC0 in inline assembly instead of
directly encoding them with the _ASM_INSN* macros, and transparently
implement these instructions as assembler macros if the toolchain
doesn't support them natively, using the recently introduced assembler
macro helpers.
The old direct encodings were restricted to using the register $at, so
this allows the extra register moves to go away (saving a grand total of
24 bytes).
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17775/
|
|
Now that we are using assembler macros to implement VZ instructions on
toolchains which don't support them, pass VZ guest Cop0 register names
to the __{read,write}_{32bit,ulong,64bit}_gc0_register macros in $n
format rather than register numbers. This is to make them consistent
with the normal root Cop0 register access macros which they were
originally based on.
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17773/
|
|
Update VZ guest register & guest TLB access helpers to use the new
assembly macros for parsing register names and creating custom assembly
macro instructions, which has a number of advantages:
- Better code can be generated on toolchains which don't support VZ,
more closely matching those which do, since there is no need to
bounce values via the $at register. Some differences still remain due
to the inability to safely fill branch delay slots and R6 compact
branch forbidden slots with explicitly encoded instructions,
resulting in some extra NOPs added by the assembler.
- Some code duplication between toolchains which do and don't support
VZ instructions is removed, since the helpers are only implemented
once. When the toolchain doesn't implement the instruction an
assembly macro implements it instead.
- Instruction encodings are kept together in the source.
On a generic kernel with KVM VZ support enabled this change saves about
2.5KiB of kernel code when TOOLCHAIN_SUPPORTS_VIRT=n, bringing it down
to about 0.5KiB more than when TOOLCHAIN_SUPPORTS_VIRT=y on r6, and just
68 bytes more on r2.
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17772/
|
|
Implement a parse_r assembler macro in asm/mipsregs.h to parse a
register in $n form, and a few C macros for defining assembler macro
instructions. These can be used to more transparently support older
binutils versions which don't support for example the msa, virt, xpa, or
crc instructions.
In particular they overcome the difficulty of turning a register name in
$n form into an instruction encoding suitable for giving to .word /
.hword, which is particularly problematic when needed from inline
assembly where the compiler is responsible for register allocation.
Traditionally this had required the use of $at and an extra MOV
instruction, but for CRC instructions with multiple GP register operands
that approach becomes more difficult.
Three assembler macro creation helpers are added:
- _ASM_MACRO_0(OP, ENC)
This is to define an assembler macro for an instruction which has no
operands, for example the VZ TLBGR instruction.
- _ASM_MACRO_2R(OP, R1, R2, ENC)
This is to define an assembler macro for an instruction which has 2
register operands, for example the CFCMSA instruction.
- _ASM_MACRO_3R(OP, R1, R2, R3, ENC)
This is to define an assembler macro for an instruction which has 3
register operands, for example the crc32 instructions.
- _ASM_MACRO_2R_1S(OP, R1, R2, SEL3, ENC)
This is to define an assembler macro for a Cop0 move instruction,
with 2 register operands and an optional register select operand
which defaults to 0, for example the VZ MFGC0 instruction.
Suggested-by: Ralf Baechle <[email protected]>
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Marcin Nowakowski <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17770/
|
|
Make read_c0_prid() use the new constant accessor macros so that it can
potentially be optimised or removed by the compiler. This is
particularly important under virtualisation, where even with hardware
assisted virtualisation (VZ), access to the PRid register may need to be
emulated by the hypervisor.
In particular this helps eliminate the read of the PRid register in the
rather frequently called add_interrupt_randomness() (which calls into
arch/mips/include/asm/timex.h) when the prid is unused but the read
can't be removed due to the inline asm being marked __volatile__.
Reported-by: Yann LeDu <[email protected]>
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Maciej W. Rozycki <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17923/
|
|
Some Cop0 registers are constant and have no side effects when read.
There is no need for the inline asm to read these to be marked
__volatile__, and doing so prevents them from being removed by the
compiler.
Add a few new accessor macros to handle these registers more efficiently
(especially for the sake of running in a guest where redundant access to
the register may trap to the hypervisor):
__read_const_32bit_c0_register()
__read_const_64bit_c0_register()
__read_const_ulong_c0_register()
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Maciej W. Rozycki <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17922/
|
|
Optimize `__read_64bit_c0_split' and reduce the instruction count by 1,
observing that a DSLL/DSRA pair by 32, is equivalent to SLL by 0, which
architecturally truncates the value requested to 32 bits on 64-bit MIPS
hardware regardless of whether the input operand is or is not a properly
sign-extended 32-bit value.
Signed-off-by: Maciej W. Rozycki <[email protected]>
Reviewed-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17399/
Signed-off-by: James Hogan <[email protected]>
|
|
The inline asm in __write_64bit_c0_split() modifies the 64-bit input
operand by shifting the high register left by 32, and constructing the
full 64-bit value in the low register (even on a 32-bit kernel), so if
that value is used again it could cause breakage as GCC would assume the
registers haven't changed when they have.
To quote the GCC extended asm documentation:
> Warning: Do not modify the contents of input-only operands (except for
> inputs tied to outputs). The compiler assumes that on exit from the
> asm statement these operands contain the same values as they had
> before executing the statement.
Avoid modifying the input by using a temporary variable as an output
which is modified instead of the input and not otherwise used. The asm
is always __volatile__ so GCC shouldn't optimise it out. The low
register of the temporary output is written before the high register of
the input is read, so we have two constraint alternatives, one where
both use the same registers (for when the input value isn't subsequently
used), and one with an early clobber on the output in case the low
output uses the same register as the high input. This allows the
resulting assembly to remain mostly unchanged.
A diff of a MIPS32r6 kernel reveals only three differences, two in
relation to write_c0_r10k_diag() in cpu_probe() (register allocation
rearranged slightly but otherwise identical), and one in relation to
write_c0_cvmmemctl2() in kvm_vz_local_flush_guesttlb_all(), but the
octeon CPU is only supported on 64-bit kernels where
__write_64bit_c0_split() isn't used so that shouldn't matter in
practice. So there currently doesn't appear to be anything broken by
this bug.
Signed-off-by: James Hogan <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17315/
Signed-off-by: Ralf Baechle <[email protected]>
|
|
MIPSr6 introduces a GlobalNumber register, which is required when VPs
are implemented (ie. when multi-threading is supported) but otherwise
optional. The register contains sufficient information to uniquely
identify a VP within a system using its cluster number, core number & VP
ID.
In preparation for using this register & its fields, introduce an
accessor macro for it & define its various bits with the typical style
preprocessor macros.
Signed-off-by: Paul Burton <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/17007/
Signed-off-by: Ralf Baechle <[email protected]>
|
|
Identify the presence of the MIPS16e2 ASE as per the architecture
specification[1], by checking for CP0 Config5.CA2 bit being 1[2].
References:
[1] "MIPS32 Architecture for Programmers: MIPS16e2 Application-Specific
Extension Technical Reference Manual", Imagination Technologies
Ltd., Document Number: MD01172, Revision 01.00, April 26, 2016,
Section 1.2 "Software Detection of the ASE", p. 5
[2] "MIPS32 interAptiv Multiprocessing System Software User's Manual",
Imagination Technologies Ltd., Document Number: MD00904, Revision
02.01, June 15, 2016, Section 2.2.1.6 "Device Configuration 5 --
Config5 (CP0 Register 16, Select 5)", pp. 71-72
Signed-off-by: Maciej W. Rozycki <[email protected]>
Reviewed-by: James Hogan <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/16094/
Signed-off-by: Ralf Baechle <[email protected]>
|
|
Octeon III implements a read-only guest CP0_PRid register, so add cases
to the KVM register access API for Octeon to ensure the correct value is
read and writes are ignored.
Signed-off-by: James Hogan <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: "Radim Krčmář" <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: David Daney <[email protected]>
Cc: Andreas Herrmann <[email protected]>
Cc: [email protected]
Cc: [email protected]
|
|
Add accessors for some VZ related Cavium Octeon III specific COP0
registers, along with field definitions. These will mostly be used by
KVM to set up interrupt routing and partition the TLB between root and
guest.
Signed-off-by: James Hogan <[email protected]>
Acked-by: Ralf Baechle <[email protected]>
Cc: David Daney <[email protected]>
Cc: Andreas Herrmann <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: "Radim Krčmář" <[email protected]>
Cc: [email protected]
Cc: [email protected]
|
|
Add some missing guest accessors and register field definitions for KVM
for MIPS VZ to make use of.
Guest CP0_LLAddr register accessors and definitions for the LLB field
allow KVM to clear the guest LLB to cancel in-progress LL/SC atomics on
restore, and to emulate accesses by the guest to the CP0_LLAddr
register.
Bitwise modifiers and definitions for the guest CP0_Wired and
CP0_Config1 registers allow KVM to modify fields within the CP0_Wired
and CP0_Config1 registers.
Finally a definition for the CP0_Config5.SBRI bit allows KVM to
initialise and allow modification of the guest version of the SBRI bit.
Signed-off-by: James Hogan <[email protected]>
Acked-by: Ralf Baechle <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: "Radim Krčmář" <[email protected]>
Cc: [email protected]
Cc: [email protected]
|
|
The MAAR V bit has been renamed VL since another bit called VH is added
at the top of the register when it is extended to 64-bits on a 32-bit
processor with XPA. Rename the V definition, fix the various users, and
add definitions for the VH bit. Also add a definition for the MAARI
Index field.
Signed-off-by: James Hogan <[email protected]>
Acked-by: Ralf Baechle <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: "Radim Krčmář" <[email protected]>
Cc: [email protected]
Cc: [email protected]
|
|
Unify definitions for MIPS performance counter register fields in
mipsregs.h rather than duplicating them in perf_events and oprofile.
This will allow future patches to use them to expose performance
counters to KVM guests.
Signed-off-by: James Hogan <[email protected]>
Cc: Ralf Baechle <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Arnaldo Carvalho de Melo <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Robert Richter <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/15212/
Signed-off-by: James Hogan <[email protected]>
|
|
Since MIPSr6 the Wired register is split into 2 fields, with the upper
16 bits of the register indicating a limit on the value that the wired
entry count in the bottom 16 bits of the register can take. This means
that simply reading the wired register doesn't get us a valid TLB entry
index any longer, and we instead need to retrieve only the lower 16 bits
of the register. Introduce a new num_wired_entries() function which does
this on MIPSr6 or higher and simply returns the value of the wired
register on older architecture revisions, and make use of it when
reading the number of wired entries.
Since commit e710d6668309 ("MIPS: tlb-r4k: If there are wired entries,
don't use TLBINVF") we have been using a non-zero number of wired
entries to determine whether we should avoid use of the tlbinvf
instruction (which would invalidate wired entries) and instead loop over
TLB entries in local_flush_tlb_all(). This loop begins with the number
of wired entries, or before this patch some large bogus TLB index on
MIPSr6 systems. Thus since the aforementioned commit some MIPSr6 systems
with FTLBs have been prone to leaving stale address translations in the
FTLB & crashing in various weird & wonderful ways when we later observe
the wrong memory.
Signed-off-by: Paul Burton <[email protected]>
Cc: Matt Redfearn <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/14557/
Signed-off-by: Ralf Baechle <[email protected]>
|
|
The FTLBP field in Config7 for the I6400 is intended as chicken bits for
debugging rather than as a field that software actually makes use of.
For best performance, FTLBP should be left at its default value of 0
with all TLB writes hitting the FTLB by default.
Additionally, since set_ftlb_enable is called from decode_configs before
decode_config4 which determines the size of the TLBs, this was
previously always setting FTLBP=3 for a 3:1 FTLB:VTLB write ratio which
makes abysmal use of the available FTLB resources.
This effectively reverts b0c4e1b79d8a ("MIPS: Set up FTLB probability
for I6400").
Signed-off-by: Paul Burton <[email protected]>
Fixes: b0c4e1b79d8a ("MIPS: Set up FTLB probability for I6400")
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/14021/
Signed-off-by: Ralf Baechle <[email protected]>
|
|
The Config.VI bit specifies that the instruction cache is virtually
tagged, which is checked in c-r4k.c's probe_pcache(). Add a proper
definition for it in mipsregs.h and make use of it.
Signed-off-by: James Hogan <[email protected]>
Acked-by: Ralf Baechle <[email protected]>
Cc: [email protected]
Signed-off-by: Paolo Bonzini <[email protected]>
|
|
No preprocessor definitions are used in the handling of the registers
accessible with the RDHWR instruction, nor the corresponding bits in the
CP0 HWREna register.
Add definitions for both the register numbers (MIPS_HWR_*) and HWREna
bits (MIPS_HWRENA_*) in asm/mipsregs.h and make use of them in the
initialisation of HWREna and emulation of the RDHWR instruction.
Signed-off-by: James Hogan <[email protected]>
Acked-by: Ralf Baechle <[email protected]>
Cc: David Daney <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Cc: Radim Krčmář <[email protected]>
Cc: [email protected]
Cc: [email protected]
Signed-off-by: Paolo Bonzini <[email protected]>
|
|
Add field definitions for some of the 64-bit specific Hardware page
Table Walker (HTW) register fields in PWSize and PWCtl, in preparation
for fixing the 64-bit HTW configuration.
Also print these fields out along with the others in print_htw_config().
Signed-off-by: James Hogan <[email protected]>
Cc: Paul Burton <[email protected]>
Cc: [email protected]
Patchwork: https://patchwork.linux-mips.org/patch/13363/
Signed-off-by: Ralf Baechle <[email protected]>
|