Propagate sNaN payload in quieting in the legacy-NaN mode as well. If
clearing the quiet bit would produce infinity, then set the next lower
trailing significand field bit, matching the SB-1 and BMIPS5000 hardware
implementations. Some other MIPS FPU hardware implementations do
produce the default qNaN bit pattern instead.
This reverts some changes made for semantics preservation with commit
dc3ddf42 [MIPS: math-emu: Update sNaN quieting handlers], consequently
bringing back most of the semantics from before commit fdffbafb [Lots of
FPU bug fixes from Kjeld Borch Egevang.], except from the qNaN produced
in the infinity case. Previously the default qNaN bit pattern was
produced in that case.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Matthew Fortune <Matthew.Fortune@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11483/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Remove irrelevant content from the description of the emulation frame in
`mips_dsemul', referring to bare-metal configurations. Update the text,
reflecting the change made with commit ba3049ed40 ("MIPS: Switch FPU
emulator trap to BREAK instruction."), where we switched from using an
address error exception on an unaligned access to the use of a BREAK 514
instruction causing a breakpoint exception instead.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/12176/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Emulate the microMIPS ADDIUPC instruction directly in `mips_dsemul'. If
executed in the emulation frame, this instruction produces an incorrect
result, because the value of the PC there is not the same as where the
instruction originated.
Reshape code so as to handle all microMIPS cases together.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/12175/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Complement commit 102cedc32a ("MIPS: microMIPS: Floating point
support.") which introduced microMIPS FPU emulation, but did not adjust
the encoding of the BREAK instruction used to terminate the branch delay
slot emulation frame. Consequently the execution of any such frame is
indeterminate and, depending on CPU configuration, will result in random
code execution or an offending program being terminated with SIGILL.
This is because the regular MIPS BREAK instruction is encoded with the 0
major and the 0xd minor opcode, however in the microMIPS instruction set
this major/minor opcode pair denotes an encoding reserved for the DSP
ASE. Instead the microMIPS BREAK instruction is encoded with the 0
major and the 0x7 minor opcode.
Use the correct BREAK encoding for microMIPS FPU emulation then.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/12174/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Correct formatting breakage introduced with commit 102cedc32a ("MIPS:
microMIPS: Floating point support."), so that further changes to this
code can be consistent.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/12173/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Fix an issue introduced with commit 9ab4471c9f ("MIPS: math-emu:
Correct delay-slot exception propagation") where the emulation of a NOP
instruction signals the need to terminate the emulation loop. This in
turn, if the PC has not changed from the entry to the loop, will cause
the kernel to terminate the program with SIGILL.
Consider this program:
static double div(double d)
{
do
d /= 2.0;
while (d > .5);
return d;
}
int main(int argc, char **argv)
{
return div(argc);
}
which gets compiled to the following binary code:
00400490 <main>:
400490: 44840000 mtc1 a0,$f0
400494: 3c020040 lui v0,0x40
400498: d44207f8 ldc1 $f2,2040(v0)
40049c: 46800021 cvt.d.w $f0,$f0
4004a0: 46220002 mul.d $f0,$f0,$f2
4004a4: 4620103c c.lt.d $f2,$f0
4004a8: 4501fffd bc1t 4004a0 <main+0x10>
4004ac: 00000000 nop
4004b0: 4620000d trunc.w.d $f0,$f0
4004b4: 03e00008 jr ra
4004b8: 44020000 mfc1 v0,$f0
4004bc: 00000000 nop
Where the FPU emulator is used, depending on the number of command-line
arguments this code will either run to completion or terminate with
SIGILL.
If no arguments are specified, then BC1T will not be taken, NOP will not
be emulated and code will complete successfully.
If one argument is specified, then BC1T will be taken once and NOP will
be emulated. At this point the entry PC value will be 0x400498 and the
new PC value, set by `mips_dsemul' will be 0x4004a0, the target of BC1T.
The emulation loop will terminate, but SIGILL will not be issued,
because the PC has changed. The FPU emulator will be entered again and
on the second execution BC1T will not be taken, NOP will not be emulated
and code will complete successfully.
If two or more arguments are specified, then the first execution of BC1T
will proceed as above. Upon reentering the FPU emulator the emulation
loop will continue to BC1T, at which point the branch will be taken and
NOP emulated again. At this point however the entry PC value will be
0x4004a0, the same as the target of BC1T. This will make the emulator
conclude that execution has not advanced and therefore an unsupported
FPU instruction has been encountered, and SIGILL will be sent to the
process.
Fix the problem by extending the internal API of `mips_dsemul', making
it return -1 if no delay slot emulation frame has been made, the
instruction has been handled and execution of the emulation loop needs
to continue as if nothing happened. Remove code from `mips_dsemul' to
reproduce steps made by the emulation loop at the conclusion of each
iteration, as those will be reached normally now. Adjust call sites
accordingly. Document the API.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Aurelien Jarno <aurelien@aurel32.net>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/12172/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Implement IEEE Std 754-2008 NaN encoding wired to the state of the
FCSR.NAN2008 bit. Make the interpretation of the quiet bit in NaN data
as follows:
* in the legacy mode originally defined by the MIPS architecture the
value of 1 denotes an sNaN whereas the value of 0 denotes a qNaN,
* in the 2008 mode introduced with revision 5 of the MIPS architecture
the value of 0 denotes an sNaN whereas the value of 1 denotes a qNaN,
following the definition of the preferred NaN encoding introduced with
IEEE Std 754-2008.
In the 2008 mode, following the requirement of the said standard, quiet
an sNaN where needed by setting the quiet bit to 1 and leaving all the
NaN payload bits unchanged.
Update format conversion operations according to the rules set by IEEE
Std 754-2008 and the MIPS architecture. Specifically:
* propagate NaN payload bits through conversions between floating-point
formats such that as much information as possible is preserved and
specifically a conversion from a narrower format to a wider format and
then back to the original format does not change a qNaN payload in any
way,
* conversions from a floating-point to an integer format where the
source is a NaN, infinity or a value that would convert to an integer
outside the range of the result format produce, under the default
exception handling, the respective values defined by the MIPS
architecture.
In full FPU emulation set the FIR.HAS2008 bit to 1, however do not make
any further FCSR bits writable.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Matthew Fortune <Matthew.Fortune@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11477/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Implement IEEE Std 754-2008 non-arithmetic ABS.fmt and NEG.fmt emulation
wired to the state of the FCSR.ABS2008 bit. In the non-arithmetic mode
the sign bit is altered according to the operation requested regardless
of the datum encoded in the input operand, no other bits are changed,
the resulting bit pattern is written to the output operand and no
exception is ever signalled.
Signed-off-by: Maciej W. Rozycki <macro@imgtec.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Matthew Fortune <Matthew.Fortune@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/11476/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
We have many extern declarations of mips_debugfs_dir through arch/mips/
in various C files. Unify them by declaring mips_debugfs_dir in a
header, including it in each affected C file & removing the duplicate
declarations.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: Steven J. Hill <Steven.Hill@imgtec.com>
Cc: Alexander Sverdlin <alexander.sverdlin@nokia.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
Cc: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-kernel@vger.kernel.org
Cc: Joe Perches <joe@perches.com>
Cc: Jaedon Shin <jaedon.shin@gmail.com>
Cc: James Hogan <james.hogan@imgtec.com>
Cc: David Daney <david.daney@cavium.com>
Cc: Zubair Lutfullah Kakakhel <Zubair.Kakakhel@imgtec.com>
Cc: Markos Chandras <markos.chandras@imgtec.com>
Cc: James Cowgill <James.Cowgill@imgtec.com>
Patchwork: https://patchwork.linux-mips.org/patch/11181/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
handle_dsemul does not exist and it's not being used in the code at all
so remove its declaration. The deliberate DS emulation exception is
handled by the do_dsemulret C code.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10950/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R6 introduced the following instruction:
Scalar Floating-Point Maximum and
Scalar Floating-Point argument with Maximum Absolute Value
MAX.fmt writes the maximum value of the inputs fs and ft to the
destination fd.
MAXA.fmt takes input arguments fs and ft and writes the argument with
the maximum absolute value to the destination fd.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10961/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R6 introduced the following instruction:
Scalar Floating-Point Minimum and
Scalar Floating-Point argument with Minimum Absolute Value
MIN.fmt writes the minimum value of the inputs fs and ft to the
destination fd.
MINA.fmt takes input arguments fs and ft and writes the argument with
the minimum absolute value to the destination fd.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10960/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R6 introduced the following instruction:
Stores in fd a bit mask reflecting the floating-point class of the
floating point scalar value fs.
CLASS.fmt: FPR[fd] = class(FPR[fs])
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10959/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R6 introduced the following instruction:
Floating-Point Round to Integral
Scalar floating-point round to integral floating point value.
RINT.fmt: FPR[fd] = round_int(FPR[fs])
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10958/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R6 introduced the following instruction:
SELNEZ.fmt: FPR[fd] FPR[ft].bit0 ? FPR[fs] : 0
Add support for emulating the single and double precision
formats of the said instruction.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10955/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R6 introduced the following instruction:
SELEQZ.fmt: FPR[fd] FPR[ft].bit0 ? 0 : FPR[fs]
Add support for emulating the single and double precision formats
of the said instruction.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10954/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Add support for emulating the new CMP.condn.fmt R6 instructions and
return SIGILL for the old C.cond.fmt if R2 emulation is not enabled
since it's not supported by R6.
The functionality of the new CMP.condn.fmt is the following one:
If the comparison specified by the condn field of the instruction
is true for the operand values, the result is true; otherwise, the
result is false. If no exception is taken, the result is written into
FPR fd; true is all 1s and false is all 0s repeated the operand width
of fmt. All other bits beyond the operand width fmt are UNPREDICTABLE.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10953/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit c8a34581ec ("MIPS: Emulate the BC1{EQ,NE}Z FPU instructions")
added support for emulating the new R6 BC1{EQ,NE}Z branches but it missed
the case where the instruction that caused the exception was not on a DS.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Fixes: c8a34581ec ("MIPS: Emulate the BC1{EQ,NE}Z FPU instructions")
Cc: <stable@vger.kernel.org> # 4.0+
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10738/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The mfhc/mthc instructions are supported on MIPS R6 so emulate
them if needed.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: <stable@vger.kernel.org> # 4.0+
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10737/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The double format (d_fmt) case uses an opening bracket which then
closes at the end of the word format (w_fmt). This can be rather confusing
so add the closing bracket at the end of the d_fmt case and use another one
for the w_fmt one.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10733/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit 5f9f41c474 ("MIPS: kernel: Prepare
the JR instruction for emulation on MIPS R6") added support for
emulating the JR instruction on MIPS R6 cores but that introduced a bug
which could be triggered when hitting a JALR opcode because the code used
the wrong field in the 'r_format' struct to determine the instruction
opcode. This lead to crashes because an emulated JALR instruction was
treated as a JR one when the R6 emulator was turned off.
Fixes: 5f9f41c474 ("MIPS: kernel: Prepare the JR instruction for emulation on MIPS R6")
Cc: <stable@vger.kernel.org> # 4.0+
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: stable@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/10583/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commits f1b44067c1 ("MIPS: Emulate the
new MIPS R6 B{L,G}T{Z,}{AL,}C instructions") and commit
a8ff66f52d ("MIPS: Emulate the new MIPS
R6 B{L,G}E{Z,}{AL,}C instructions") added support for emulating various
branch compact instructions. However, it missed the case for those which
use the old BLEZL and BGTZL opcodes leading to random crashes when the R6
emulator is disabled. We fix this by ensuring that the 'rt' field is not
zero which is always true for these branch compact instructions.
Fixes: f1b44067c1 ("MIPS: Emulate the new MIPS R6 B{L,G}T{Z,}{AL,}C instructions")
Fixes: a8ff66f52d ("MIPS: Emulate the new MIPS R6 B{L,G}E{Z,}{AL,}C instructions")
Cc: <stable@vger.kernel.org> # 4.0+
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: Markos Chandras <markos.chandras@imgtec.com>
Patchwork: https://patchwork.linux-mips.org/patch/10582/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Fix "BUG: using smp_processor_id() in preemptible" reported in accesses
to thread's FPU defaults: the value to initialise FSCR to at program
startup, the FCSR r/w mask and the contents of FIR in full FPU
emulation, removing a regression introduced with 9b26616c [MIPS: Respect
the ISA level in FCSR handling] and f6843626 [MIPS: math-emu: Set FIR
feature flags for full emulation].
Use `boot_cpu_data' to obtain the data from, following the approach that
`cpu_has_*' macros take and avoiding the call to `smp_processor_id' made
in the reference to `current_cpu_data'. The contents of FSCR have to be
consistent across processors in an SMP system, the settings there must
not change as a thread is migrated across processors. And the contents
of FIR are guaranteed to be consistent in FPU emulation, by definition.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Tested-by: Ezequiel Garcia <ezequiel.garcia@imgtec.com>
Tested-by: Paul Martin <paul.martin@codethink.co.uk>
Cc: Markos Chandras <Markos.Chandras@imgtec.com>
Cc: James Hogan <james.hogan@imgtec.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/10030/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Define the central place the default FCSR value is set from, initialised
in `cpu_probe'. Determine the FCSR mask applied to values written to
the register with CTC1 in the full emulation mode and via ptrace(2),
according to the ISA level of processor hardware or the writability of
bits 31:18 if actual FPU hardware is used.
Software may rely on FCSR bits whose functions our emulator does not
implement, so it should not allow them to be set or software may get
confused. For ptrace(2) it's just sanity.
[ralf@linux-mips.org: Fixed double inclusion of <asm/current.h>.]
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9711/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The ABS.fmt and NEG.fmt instructions have been specified as arithmetic
in the MIPS architecture, which in particular implies handling NaN data
in the usual way with qNaN bit patterns propagated unchanged and sNaN
bit patterns signalling the usual IEEE 754 Invalid Operation exception
and quieted by default.
A series of changes applied over time to our implementation:
c5033d78 [MIPS] ieee754[sd]p_neg workaround
cea2be44 MIPS: Fix abs.[sd] and neg.[sd] emulation for NaN operands
has led to the current situation where the sign bit is updated according
to the operation requested even for NaN inputs. This is according to
these commits a workaround so that broken binaries produced by GCC
disregarding the properties of these instructions have a chance to work.
For sNaN inputs this remains within IEEE Std 754 as the standard leaves
the choice of output qNaN bit patterns produced under the default
Invalid Operation exception handling for individual sNaN input bit
patterns to implementer's discretion, even though it still recommends as
much NaN input information to be preserved in NaN outputs.
For qNaN inputs however it violates the standard as it requires a qNaN
input bit patterns to propagate unchanged to output.
This is also unlike real MIPS FPU hardware behaves where sNaN and/or
qNaN processing has been fully implemented with no Unimplemented
Operation exception signalled. Such hardware propagates any input qNaN
bit pattern unchanged. It also quiets any input sNaN bit pattern in an
implementer-specific manner, for example the MIPS 74Kf processor returns
the default qNaN pattern with the sign bit always clear and the Broadcom
SB-1 and BMIPS5000 processors propagate the input sNaN bit pattern with
the sign bit unchanged and the quiet bit first cleared in the trailing
significand field and then the next lower bit set if clearing the quiet
bit left the field with no other bit set.
Especially the latter observation indicates the limited usefulness of
the workaround as it will cover many hardware configurations, but not
all of them, only making it harder to discover such broken binaries that
need to be recompiled with GCC told to avoid the use of ABS.fmt and
NEG.fmt instructions where non-arithmetic semantics is required by the
algorithm used.
Revert the damage done by the series of changes then, and take the
opportunity to simplify implementation by calling `ieee754dp_sub' and
`ieee754dp_add' as required and also the rounding mode set towards -Inf
temporarily so that the sign of 0 is correctly handled.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9710/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Define IEEE 754-2008 feature control bits: FIR.HAS2008, FCSR.ABS2008 and
FCSR.NAN2008, and update the `_ieee754_csr' structure accordingly.
For completeness define FIR.UFRP too.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9709/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Implement the FCCR, FEXR and FENR "shadow" FPU registers for the
architecture levels that include them, for the CFC1 and CTC1
instructions in the full emulation mode.
For completeness add macros for the CP1 UFR and UNFR registers too, no
actual implementation though.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9708/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Implement FIR feature flags in the FPU emulator according to features
supported and architecture level requirements. The W, L and F64 bits
have only been added at level #2 even though the features they refer to
were also included with the MIPS64r1 ISA and the W fixed-point format
also with the MIPS32r1 ISA.
This is only relevant for the full emulation mode and the emulated CFC1
instruction as well as ptrace(2) accesses.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9707/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Complement 593d33fe [MIPS: math-emu: Move various objects into an ar
library.] and also move sp_tlong.o, sp_flong.o, dp_tlong.o, and
dp_flong.o into an `ar' library. These objects implement long
fixed-point format support that can be omitted from MIPS I, MIPS II and
MIPS32r1 configurations.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9702/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Restore EPC at the branch whose delay slot is emulated if the delay-slot
instruction signals. This is so that code in `fpu_emulator_cop1Handler'
does not see EPC having advanced and mistakenly successfully resume
userland execution from the location at the branch target in that case.
Restoring EPC guarantees an immediate exit from the emulation loop and
if EPC hasn't advanced at all since entering the loop, also issuing the
signal reported by the delay-slot instruction.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9701/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Correct ISA requirements for floating-point instructions:
* the CU3 exception signifies a real COP3 instruction in MIPS I & II,
* the BC1FL and BC1TL instructions are not supported in MIPS I,
* the SQRT.fmt instructions are indeed supported in MIPS II,
* the LDC1 and SDC1 instructions are indeed supported in MIPS32r1,
* the CEIL.W.fmt, FLOOR.W.fmt, ROUND.W.fmt and TRUNC.W.fmt instructions
are indeed supported in MIPS32,
* the CVT.L.fmt and CVT.fmt.L instructions are indeed supported in
MIPS32r2 and MIPS32r6,
* the CEIL.L.fmt, FLOOR.L.fmt, ROUND.L.fmt and TRUNC.L.fmt instructions
are indeed supported in MIPS32r2 and MIPS32r6,
* the RSQRT.fmt and RECIP.fmt instructions are indeed supported in
MIPS64r1,
Also simplify conditionals for MIPS III and MIPS IV FPU instructions and
the handling of the MOVCI minor opcode.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9700/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Correct a cache coherency regression introduced with be1664c4 [Another
round of fixes for the fp emulator.] for the emulation frame used in
delay-slot emulation.
Two instructions are copied into the frame and as from the commit
referred a cache synchronisation call is made for the second instruction
aka `badinst' of the two only. The `flush_cache_sigtramp' interface is
reused that guarantees that synchronisation will be made for 8 bytes or
2 instructions starting from the address requested, although if cache
lines are wider then a larger area may be synchronised.
Change the call to point to the first of the two instructions aka `emul'
instead, removing unpredictable behaviour resulting from cache
incoherency.
This bug only ever manifested itself on systems implementing 4-byte
cache lines, typically MIPS I systems, causing all kinds of weirdness.
This is because the sequence of two instructions starting from `emul' is
8-byte aligned and for 8-byte or wider cache lines the line synchronised
will span both, so the vast majority of systems have escaped unharmed.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9698/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The `ieee754sp_isnan' and `ieee754dp_isnan' NaN classifiers are now no
longer externally referred, remove their header prototypes and make them
local to the two only respective places still making use of them.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9693/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Rewrite qNaN handling in `ieee754sp_fdp' using the `ieee754_class_nan'
helper recently added, removing the external call to `ieee754sp_isnan'
and reducing the size of code by 16 instructions or 64 bytes.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9692/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Remove a redundant call to `ieee754_setandtestcx' in `ieee754sp_cmp' and
`ieee754dp_cmp'. The IEEE 754 exception requested will have already
been set by a call to `ieee754_setcx' immediately above, because `sig'
has to be non-zero to reach here, and the comparison result returned
will be 0 regardless of the result from the call. Simplify the return
expression remaining. All this reducing the size of code by 16 and 12
instructions or 64 and 48 bytes respectively.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9690/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
We have the input operands already classified in `ieee754sp_cmp' and
`ieee754dp_cmp' comparison operations, so use the class obtained to tell
NaNs and numbers apart rather than classifying inputs again for this
purpose, reducing the size of code by 24 and 40 instructions or 96 and
160 bytes respectively.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9689/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Revert the changes made by commit fdffbafb [Lots of FPU bug fixes from
Kjeld Borch Egevang.] to `ieee754sp_nanxcpt' and `ieee754dp_nanxcpt'
sNaN quieting handlers and their callers so that sNaN processing is done
within the handlers againg. Pass the sNaN causing an IEEE 754 invalid
operation exception down to the relevant handler. Pass the sNaN in `fs'
where two sNaNs are supplied to a binary operation.
Set the Invalid Operation FCSR exception bits in the quieting handlers
rather than at their call sites throughout. Make the handlers exclusive
for sNaN processing.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9688/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Don't call the `ieee754sp_nanxcpt' and `ieee754dp_nanxcpt' sNaN quieting
handlers for a qNaN supplied to floating-point format conversions or
SQRT.S/SQRT.D instructions, or for a qNaN produced out of a negative
operand supplied to SQRT.S/SQRT.D instructions. Return the qNaN right
away in these cases.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9687/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit fdffbafb [Lots of FPU bug fixes from Kjeld Borch Egevang.]
replaced the two single `ieee754sp_nanxcpt' and `ieee754dp_nanxcpt'
places, where sNaN quieting used to happen for single and double
floating-point operations respectively, with individual qNaN
instantiations across all the call sites instead. It also made most of
these two functions dead code as where called on a qNaN they return
right away.
To revert the damage and make sNaN quieting uniform again first rewrite
`ieee754sp_nanxcpt' and `ieee754dp_nanxcpt' to do the same quieting all
the call sites do, that is return the default qNaN encoding for all
input sNaN values; never propagate any sNaN payload bits from its
trailing significand field.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9685/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Move CFC1/CTC1 emulation code to separate functions to avoid excessive
indentation in forthcoming changes. Adjust formatting in a minor way
and remove extraneous round brackets.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9682/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit 56a64733 [MIPS: math-emu: Switch to using the MIPS rounding
modes.] removed the distinction between hardware and emulator rounding
mode encodings, the hardware encoding is now used in emulation as well.
Complement the change and remove the `modeindex' macro previously used
for indexing into encoding translation tables, it now does nothing and
only obfuscates code by reinserting the value extracted from FCSR.
Adjust comments accordingly.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9680/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Add the remaining missing comments for IEEE 754 special value array
indices. Reindent macro definitions for consistency.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9671/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
IEEE754_SPCVAL_NMIN denotes the index into the special value array where
the closest to zero negative normal number expressible is stored.
Similarly IEEE754_SPCVAL_NMIND denotes such index for the closest to
zero negative subnormal number expressible. Make comments match that.
Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/9670/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Delay slot emulation in the FPU emulator is the only kernel user of an
executable stack, it is also very slow. Add a counter so we can see
how many of these emulations are done.
Signed-off-by: David Daney <david.daney@cavium.com>
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/8634/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Remove variable self-assignments.
This silences a bunch of -Wself-assign warnings reported by clang.
The changed code can be compiled without warnings by both gcc and clang.
Signed-off-by: Toma Tabacu <toma.tabacu@imgtec.com>
Signed-off-by: Daniel Sanders <daniel.sanders@imgtec.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/9314/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
MIPS R2 FPU instructions are also present in MIPS R6 so amend the
preprocessor definitions to take MIPS R6 into consideration.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
MIPS R6 removed quite a few R2 instructions. However, there
is plenty of <R6 userland code so we add an in-kernel emulator
so we can still be able to execute all R2 userland out there.
The emulator comes with a handy debugfs under /mips/ directory
(r2-emul-stats) to provide some basic statistics of the
instructions that are being emulated.
Below are some statistics from booting a minimal buildroot image:
Instruction Total BDslot
------------------------------
movs 236969 0
hilo 56686 0
muls 55279 0
divs 10941 0
dsps 0 0
bops 1 0
traps 0 0
fpus 0 0
loads 214981 17
stores 103364 0
llsc 56898 0
dsemul 150418 0
jr 370158
bltzl 43
bgezl 1594
bltzll 0
bgezll 0
bltzal 39
bgezal 39
beql 14503
bnel 138741
blezl 0
bgtzl 3988
Signed-off-by: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
MIPS R6 added the following four instructions which share the
BGTZ and BGTZL opcode:
BLTZALC: Compact branch-and-link if GPR rt is < to zero
BGTZALC: Compact branch-and-link if GPR rt is > to zero
BLTZL : Compact branch if GPR rt is < to zero
BGTZL : Compact branch if GPR rt is > to zero
BLTC : Compact branch if GPR rs is less than GPR rt
BLTUC : Similar to BLTC but unsigned
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
MIPS R6 added the following four instructions which share the
BLEZ and BLEZL opcodes:
BLEZALC: Compact branch-and-link if GPR rt is <= to zero
BGEZALC: Compact branch-and-link if GPR rt is >= to zero
BLEZC : Compact branch if GPR rt is <= to zero
BGEZC : Compact branch if GPR rt is >= to zero
BGEC : Compact branch if GPR rs is less than or equal to GPR rt
BGEUC : Similar to BGEC but unsigned.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
MIPS R6 introduced the following two branch instructions for COP1:
BC1EQZ: Branch if Cop1 (FPR) Register Bit 0 is Equal to Zero
BC1NEZ: Branch if Cop1 (FPR) Register Bit 0 is Not Equal to Zero
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
MIPS R6 removed the BLTZL, BGEZL, BLTZAL, BGEZAL, BEQL, BNEL, BLEZL,
BGTZL branch likely instructions so we must not try to emulate them on
MIPS R6 if the R2-to-R6 emulator is not present.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
The MIPS R6 JR instruction is an alias to the JALR one, so it may
need emulation for non-R6 userlands.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Hybrid FPRs is a scheme where scalar FP registers are 64b wide, but
accesses to odd indexed single registers use bits 63:32 of the
preceeding even indexed 64b register. In this mode all FP code
except that built for the plain FP64 ABI can execute correctly. Most
notably a combination of FP64A & FP32 code can execute correctly,
allowing for existing FP32 binaries to be linked with new FP64A binaries
that can make use of 64 bit FP & MSA.
Hybrid FPRs are implemented by setting both the FR & FRE bits, trapping
& emulating single precision FP instructions (via Reserved Instruction
exceptions) whilst allowing others to execute natively. It therefore has
a penalty in terms of execution speed, and should only be used when no
fully native mode can be. As more binaries are recompiled to use either
the FPXX or FP64(A) ABIs, the need for hybrid FPRs should diminish.
However in the short to mid term it allows for a gradual transition
towards that world, rather than a complete ABI break which is not
feasible for some users & not desirable for many.
A task will be executed using the hybrid FPR scheme when its
TIF_HYBRID_FPREGS flag is set & TIF_32BIT_FPREGS is clear. A further
patch will set the flags as necessary, this patch simply adds the
infrastructure necessary for the hybrid FPR mode to work.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: linux-fsdevel@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/7683/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Based on the spatch
@@
expression e;
@@
- return (e);
+ return e;
with heavy hand editing because some of the changes are either whitespace
or identation only or result in excessivly long lines.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Starting with version 2.24.51.20140728 MIPS binutils complain loudly
about mixing soft-float and hard-float object files, leading to this
build failure since GCC is invoked with "-msoft-float" on MIPS:
{standard input}: Warning: .gnu_attribute 4,3 requires `softfloat'
LD arch/mips/alchemy/common/built-in.o
mipsel-softfloat-linux-gnu-ld: Warning: arch/mips/alchemy/common/built-in.o
uses -msoft-float (set by arch/mips/alchemy/common/prom.o),
arch/mips/alchemy/common/sleeper.o uses -mhard-float
To fix this, we detect if GAS is new enough to support "-msoft-float" command
option, and if it does, we can let GCC pass it to GAS; but then we also need
to sprinkle the files which make use of floating point registers with the
necessary ".set hardfloat" directives.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com>
Cc: Linux-MIPS <linux-mips@linux-mips.org>
Cc: Matthew Fortune <Matthew.Fortune@imgtec.com>
Cc: Markos Chandras <Markos.Chandras@imgtec.com>
Cc: Maciej W. Rozycki <macro@linux-mips.org>
Patchwork: https://patchwork.linux-mips.org/patch/8355/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit 08a07904e1 ("MIPS: math-emu: Remove most ifdefery") removed
the #ifdef ISA conditions and switched to runtime detection. However,
according to the instruction set manual, the cop1x_op instructions are
available in >=MIPS32r2 as well. This fixes a problem on MIPS32r2
with the ntpd package which failed to execute with a SIGILL exit code due
to the fact that a madd.d instruction was not being emulated.
Signed-off-by: Markos Chandras <markos.chandras@imgtec.com>
Fixes: 08a07904e1 ("MIPS: math-emu: Remove most ifdefery")
Cc: <stable@vger.kernel.org> # v3.16+
Cc: linux-mips@linux-mips.org
Reviewed-by: Paul Burton <paul.burton@imgtec.com>
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Cc: Markos Chandras <markos.chandras@imgtec.com>
Patchwork: https://patchwork.linux-mips.org/patch/8173/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Commit bbd426f542 "MIPS: Simplify FP context access" modified the
SIFROMREG & SIFROMHREG macros such that they return unsigned rather
than signed 32b integers. I had believed that to be fine, but
inadvertently missed the MFC1 & MFHC1 cases which write to a struct
pt_regs regs element. On MIPS32 this is fine, but on 64 bit those
saved regs' fields are 64 bit wide. Using unsigned values caused the
32 bit value from the FP register to be zero rather than sign extended
as the architecture specifies, causing incorrect emulation of the
MFC1 & MFHc1 instructions. Fix by reintroducing the casts to signed
integers, and therefore the sign extension.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
Cc: stable@vger.kernel.org # v3.15+
Cc: linux-mips@linux-mips.org
Patchwork: https://patchwork.linux-mips.org/patch/7848/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
The fix in the preceeding commit did do exactly the same thing in two
places showing some code cleanup was due.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Older GCC doesn't get named initializations of anonymous structs right,
that is members are not initializable in the containing structure through
name however old style initializations are working fine.
The issue exists with gcc up to 4.5.x.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Some versions of the assembler will not assemble CFC1 for OCTEON, so
override the ISA for these.
Add r4k_fpu.o to handle low level FPU initialization.
Modify octeon_switch.S to save the FPU registers. And include
r4k_switch.S to pick up more FPU support.
Get rid of "#define cpu_has_fpu 0"
Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: Andreas Herrmann <andreas.herrmann@caviumnetworks.com>
Cc: linux-mips@linux-mips.org
Cc: James Hogan <james.hogan@imgtec.com>
Cc: kvm@vger.kernel.org
Patchwork: https://patchwork.linux-mips.org/patch/7006/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Sometimes it's useful to let the user, while doing performance research,
know what in the IEEE754 exceptions has caused many times of FP emulation
when running a specific application. This patch adds 5 more files to
/sys/kernel/debug/mips/fpuemustats/, whose filenames begin with "ieee754".
These stats are in addition to the existing cp1ops, cp1xops, errors, loads
and stores, which may not be useful in understanding the reasons of ieee754
exceptions.
[ralf@linux-mips.org: Fixed reject due to other changes to the kernel
FP assist software.]
Signed-off-by: Deng-Cheng Zhu <dengcheng.zhu@imgtec.com>
Cc: linux-mips@linux-mips.org
Cc: Steven.Hill@imgtec.com
Cc: james.hogan@imgtec.com
Patchwork: http://patchwork.linux-mips.org/patch/7044/
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Move microMIPS32_to_MIPS32() to a separate file which only gets built
for mipsMIPS configurations; for other configurations the optimizer
eleminates calls to microMIPS32_to_MIPS32().
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Previously math-emu was using the IEEE-754 constants internally. These
were differing by having the constants for rounding to +/- infinity
switched, so a conversion was necessary. This would be entirely
avoidable if the MIPS constants were used throughout, so get rid of
the bloat.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
mm_isBranchInstr() did reside in the math emu code even though it logically
is separate and also is used outside the math emu code. In addition GCC 4.9.0
leaves the following unnnecessarily bloated function body for a non-microMIPS
configuration:
<mm_isBranchInstr>:
105c: afa50004 sw a1,4(sp)
1060: afa60008 sw a2,8(sp)
1064: afa7000c sw a3,12(sp)
1068: 03e00008 jr ra
106c: 00001021 move v0,zero
which stores arguments that are never going to be used on the stack frame.
Move mm_isBranchInstr() from cp1emu.c to branch.c, then split mm_isBranchInstr()
into a __mm_isBranchInstr() core and a mm_isBranchInstr() wrapper inline function
which only invokes __mm_isBranchInstr() on microMIPS configurations.
This shaves off 112 bytes off the kernel and improves code flow a bit.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
o Only define variables in the outermost block
o One empty line at most
o Format comments as per CodingStyle
o Update FSF address in licensing term comment
o Spell FPU and MIPS in all capitals.
o Remove ####-type of lines in comments.
o Try to make things a bit most consistent between sp_*.c / dp_*.c files.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
There are two version of get_rounding(), one for single precision, one
for double precision. Add a ieee754sp_ rsp. ieee754dp_ prefix for
clarity.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
ieee754d.o contains only debug code and dp_sqrt.o and sp_sqrt.o contain
code which for MIPS I/II/III systems we don't want to link. Again the
savings can be considerable for some systems:
$ mips-linux-size --totals ieee754d.o dp_sqrt.o sp_sqrt.o
text data bss dec hex filename
1624 0 0 1624 658 ieee754d.o
2016 0 0 2016 7e0 dp_sqrt.o
736 0 0 736 2e0 sp_sqrt.o
4376 0 0 4376 1118 (TOTALS)
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Both are unused since lmo commit fdffbafbb38723618626c70ffdc6ff9175cdffa2
[Lots of FPU bug fixes from Kjeld Borch Egevang.]
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
LONG_LONG_MAX is a symbol defined in <limits.h> which may not be available
so better rely on something provided by a kernel header. While at it,
turn these function-like macros into inline functions.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Maciej W. Rozycki