Minki/linux
1
0
Fork 1
mirror of https://github.com/torvalds/linux.git synced 2024-11-06 20:21:57 +00:00
Code Issues Packages Projects Releases Wiki Activity
62d228b8c6
linux/arch/sparc/math-emu/math_64.c
David S. Miller 517ffce4e1 sparc64: Make montmul/montsqr/mpmul usable in 32-bit threads. 
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.

These values are 64-bit.  But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.

Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:

1) Load the top-32bits of a choosen integer register with a sentinel,
   say "-1".  This will be in the outer-most register window.

   The idea is that we're trying to see if the outer-most register
   window gets spilled, and thus the 64-bit values were truncated.

2) Load all the inputs for the montmul/montsqr/mpmul instruction,
   down to the inner-most register window.

3) Execute the opcode.

4) Traverse back up to the outer-most register window.

5) Check the sentinel, if it's still "-1" store the results.
   Otherwise retry the entire sequence.

This retry is extremely troublesome.  If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.

We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.

Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack.  Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.

So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.

Runtime detection of such a facility is tricky, and cumbersome at
best.  For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp).  Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.

With help from Andy Polyakov.

Signed-off-by: David S. Miller <davem@davemloft.net>
2012-10-26 15:18:37 -07:00

525 lines
16 KiB
C
Raw Blame History

/*
* arch/sparc64/math-emu/math.c
*
* Copyright (C) 1997,1999 Jakub Jelinek (jj@ultra.linux.cz)
* Copyright (C) 1999 David S. Miller (davem@redhat.com)
*
* Emulation routines originate from soft-fp package, which is part
* of glibc and has appropriate copyrights in it.
*/
#include <linux/types.h>
#include <linux/sched.h>
#include <linux/errno.h>
#include <linux/perf_event.h>
#include <asm/fpumacro.h>
#include <asm/ptrace.h>
#include <asm/uaccess.h>
#include <asm/cacheflush.h>
#include "sfp-util_64.h"
#include <math-emu/soft-fp.h>
#include <math-emu/single.h>
#include <math-emu/double.h>
#include <math-emu/quad.h>
/* QUAD - ftt == 3 */
#define FMOVQ 0x003
#define FNEGQ 0x007
#define FABSQ 0x00b
#define FSQRTQ 0x02b
#define FADDQ 0x043
#define FSUBQ 0x047
#define FMULQ 0x04b
#define FDIVQ 0x04f
#define FDMULQ 0x06e
#define FQTOX 0x083
#define FXTOQ 0x08c
#define FQTOS 0x0c7
#define FQTOD 0x0cb
#define FITOQ 0x0cc
#define FSTOQ 0x0cd
#define FDTOQ 0x0ce
#define FQTOI 0x0d3
/* SUBNORMAL - ftt == 2 */
#define FSQRTS 0x029
#define FSQRTD 0x02a
#define FADDS 0x041
#define FADDD 0x042
#define FSUBS 0x045
#define FSUBD 0x046
#define FMULS 0x049
#define FMULD 0x04a
#define FDIVS 0x04d
#define FDIVD 0x04e
#define FSMULD 0x069
#define FSTOX 0x081
#define FDTOX 0x082
#define FDTOS 0x0c6
#define FSTOD 0x0c9
#define FSTOI 0x0d1
#define FDTOI 0x0d2
#define FXTOS 0x084 /* Only Ultra-III generates this. */
#define FXTOD 0x088 /* Only Ultra-III generates this. */
#if 0 /* Optimized inline in sparc64/kernel/entry.S */
#define FITOS 0x0c4 /* Only Ultra-III generates this. */
#endif
#define FITOD 0x0c8 /* Only Ultra-III generates this. */
/* FPOP2 */
#define FCMPQ 0x053
#define FCMPEQ 0x057
#define FMOVQ0 0x003
#define FMOVQ1 0x043
#define FMOVQ2 0x083
#define FMOVQ3 0x0c3
#define FMOVQI 0x103
#define FMOVQX 0x183
#define FMOVQZ 0x027
#define FMOVQLE 0x047
#define FMOVQLZ 0x067
#define FMOVQNZ 0x0a7
#define FMOVQGZ 0x0c7
#define FMOVQGE 0x0e7
#define FSR_TEM_SHIFT 23UL
#define FSR_TEM_MASK (0x1fUL << FSR_TEM_SHIFT)
#define FSR_AEXC_SHIFT 5UL
#define FSR_AEXC_MASK (0x1fUL << FSR_AEXC_SHIFT)
#define FSR_CEXC_SHIFT 0UL
#define FSR_CEXC_MASK (0x1fUL << FSR_CEXC_SHIFT)
/* All routines returning an exception to raise should detect
* such exceptions _before_ rounding to be consistent with
* the behavior of the hardware in the implemented cases
* (and thus with the recommendations in the V9 architecture
* manual).
*
* We return 0 if a SIGFPE should be sent, 1 otherwise.
*/
static inline int record_exception(struct pt_regs *regs, int eflag)
{
u64 fsr = current_thread_info()->xfsr[0];
int would_trap;
/* Determine if this exception would have generated a trap. */
would_trap = (fsr & ((long)eflag << FSR_TEM_SHIFT)) != 0UL;
/* If trapping, we only want to signal one bit. */
if(would_trap != 0) {
eflag &= ((fsr & FSR_TEM_MASK) >> FSR_TEM_SHIFT);
if((eflag & (eflag - 1)) != 0) {
if(eflag & FP_EX_INVALID)
eflag = FP_EX_INVALID;
else if(eflag & FP_EX_OVERFLOW)
eflag = FP_EX_OVERFLOW;
else if(eflag & FP_EX_UNDERFLOW)
eflag = FP_EX_UNDERFLOW;
else if(eflag & FP_EX_DIVZERO)
eflag = FP_EX_DIVZERO;
else if(eflag & FP_EX_INEXACT)
eflag = FP_EX_INEXACT;
}
}
/* Set CEXC, here is the rule:
*
* In general all FPU ops will set one and only one
* bit in the CEXC field, this is always the case
* when the IEEE exception trap is enabled in TEM.
*/
fsr &= ~(FSR_CEXC_MASK);
fsr |= ((long)eflag << FSR_CEXC_SHIFT);
/* Set the AEXC field, rule is:
*
* If a trap would not be generated, the
* CEXC just generated is OR'd into the
* existing value of AEXC.
*/
if(would_trap == 0)
fsr |= ((long)eflag << FSR_AEXC_SHIFT);
/* If trapping, indicate fault trap type IEEE. */
if(would_trap != 0)
fsr |= (1UL << 14);
current_thread_info()->xfsr[0] = fsr;
/* If we will not trap, advance the program counter over
* the instruction being handled.
*/
if(would_trap == 0) {
regs->tpc = regs->tnpc;
regs->tnpc += 4;
}
return (would_trap ? 0 : 1);
}
typedef union {
u32 s;
u64 d;
u64 q[2];
} *argp;
int do_mathemu(struct pt_regs *regs, struct fpustate *f, bool illegal_insn_trap)
{
unsigned long pc = regs->tpc;
unsigned long tstate = regs->tstate;
u32 insn = 0;
int type = 0;
/* ftt tells which ftt it may happen in, r is rd, b is rs2 and a is rs1. The *u arg tells
whether the argument should be packed/unpacked (0 - do not unpack/pack, 1 - unpack/pack)
non-u args tells the size of the argument (0 - no argument, 1 - single, 2 - double, 3 - quad */
#define TYPE(ftt, r, ru, b, bu, a, au) type = (au << 2) | (a << 0) | (bu << 5) | (b << 3) | (ru << 8) | (r << 6) | (ftt << 9)
int freg;
static u64 zero[2] = { 0L, 0L };
int flags;
FP_DECL_EX;
FP_DECL_S(SA); FP_DECL_S(SB); FP_DECL_S(SR);
FP_DECL_D(DA); FP_DECL_D(DB); FP_DECL_D(DR);
FP_DECL_Q(QA); FP_DECL_Q(QB); FP_DECL_Q(QR);
int IR;
long XR, xfsr;
if (tstate & TSTATE_PRIV)
die_if_kernel("unfinished/unimplemented FPop from kernel", regs);
perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, 1, regs, 0);
if (test_thread_flag(TIF_32BIT))
pc = (u32)pc;
if (get_user(insn, (u32 __user *) pc) != -EFAULT) {
if ((insn & 0xc1f80000) == 0x81a00000) /* FPOP1 */ {
switch ((insn >> 5) & 0x1ff) {
/* QUAD - ftt == 3 */
case FMOVQ:
case FNEGQ:
case FABSQ: TYPE(3,3,0,3,0,0,0); break;
case FSQRTQ: TYPE(3,3,1,3,1,0,0); break;
case FADDQ:
case FSUBQ:
case FMULQ:
case FDIVQ: TYPE(3,3,1,3,1,3,1); break;
case FDMULQ: TYPE(3,3,1,2,1,2,1); break;
case FQTOX: TYPE(3,2,0,3,1,0,0); break;
case FXTOQ: TYPE(3,3,1,2,0,0,0); break;
case FQTOS: TYPE(3,1,1,3,1,0,0); break;
case FQTOD: TYPE(3,2,1,3,1,0,0); break;
case FITOQ: TYPE(3,3,1,1,0,0,0); break;
case FSTOQ: TYPE(3,3,1,1,1,0,0); break;
case FDTOQ: TYPE(3,3,1,2,1,0,0); break;
case FQTOI: TYPE(3,1,0,3,1,0,0); break;
/* We can get either unimplemented or unfinished
* for these cases. Pre-Niagara systems generate
* unfinished fpop for SUBNORMAL cases, and Niagara
* always gives unimplemented fpop for fsqrt{s,d}.
*/
case FSQRTS: {
unsigned long x = current_thread_info()->xfsr[0];
x = (x >> 14) & 0x7;
TYPE(x,1,1,1,1,0,0);
break;
}
case FSQRTD: {
unsigned long x = current_thread_info()->xfsr[0];
x = (x >> 14) & 0x7;
TYPE(x,2,1,2,1,0,0);
break;
}
/* SUBNORMAL - ftt == 2 */
case FADDD:
case FSUBD:
case FMULD:
case FDIVD: TYPE(2,2,1,2,1,2,1); break;
case FADDS:
case FSUBS:
case FMULS:
case FDIVS: TYPE(2,1,1,1,1,1,1); break;
case FSMULD: TYPE(2,2,1,1,1,1,1); break;
case FSTOX: TYPE(2,2,0,1,1,0,0); break;
case FDTOX: TYPE(2,2,0,2,1,0,0); break;
case FDTOS: TYPE(2,1,1,2,1,0,0); break;
case FSTOD: TYPE(2,2,1,1,1,0,0); break;
case FSTOI: TYPE(2,1,0,1,1,0,0); break;
case FDTOI: TYPE(2,1,0,2,1,0,0); break;
/* Only Ultra-III generates these */
case FXTOS: TYPE(2,1,1,2,0,0,0); break;
case FXTOD: TYPE(2,2,1,2,0,0,0); break;
#if 0 /* Optimized inline in sparc64/kernel/entry.S */
case FITOS: TYPE(2,1,1,1,0,0,0); break;
#endif
case FITOD: TYPE(2,2,1,1,0,0,0); break;
}
}
else if ((insn & 0xc1f80000) == 0x81a80000) /* FPOP2 */ {
IR = 2;
switch ((insn >> 5) & 0x1ff) {
case FCMPQ: TYPE(3,0,0,3,1,3,1); break;
case FCMPEQ: TYPE(3,0,0,3,1,3,1); break;
/* Now the conditional fmovq support */
case FMOVQ0:
case FMOVQ1:
case FMOVQ2:
case FMOVQ3:
/* fmovq %fccX, %fY, %fZ */
if (!((insn >> 11) & 3))
XR = current_thread_info()->xfsr[0] >> 10;
else
XR = current_thread_info()->xfsr[0] >> (30 + ((insn >> 10) & 0x6));
XR &= 3;
IR = 0;
switch ((insn >> 14) & 0x7) {
/* case 0: IR = 0; break; */ /* Never */
case 1: if (XR) IR = 1; break; /* Not Equal */
case 2: if (XR == 1 || XR == 2) IR = 1; break; /* Less or Greater */
case 3: if (XR & 1) IR = 1; break; /* Unordered or Less */
case 4: if (XR == 1) IR = 1; break; /* Less */
case 5: if (XR & 2) IR = 1; break; /* Unordered or Greater */
case 6: if (XR == 2) IR = 1; break; /* Greater */
case 7: if (XR == 3) IR = 1; break; /* Unordered */
}
if ((insn >> 14) & 8)
IR ^= 1;
break;
case FMOVQI:
case FMOVQX:
/* fmovq %[ix]cc, %fY, %fZ */
XR = regs->tstate >> 32;
if ((insn >> 5) & 0x80)
XR >>= 4;
XR &= 0xf;
IR = 0;
freg = ((XR >> 2) ^ XR) & 2;
switch ((insn >> 14) & 0x7) {
/* case 0: IR = 0; break; */ /* Never */
case 1: if (XR & 4) IR = 1; break; /* Equal */
case 2: if ((XR & 4) || freg) IR = 1; break; /* Less or Equal */
case 3: if (freg) IR = 1; break; /* Less */
case 4: if (XR & 5) IR = 1; break; /* Less or Equal Unsigned */
case 5: if (XR & 1) IR = 1; break; /* Carry Set */
case 6: if (XR & 8) IR = 1; break; /* Negative */
case 7: if (XR & 2) IR = 1; break; /* Overflow Set */
}
if ((insn >> 14) & 8)
IR ^= 1;
break;
case FMOVQZ:
case FMOVQLE:
case FMOVQLZ:
case FMOVQNZ:
case FMOVQGZ:
case FMOVQGE:
freg = (insn >> 14) & 0x1f;
if (!freg)
XR = 0;
else if (freg < 16)
XR = regs->u_regs[freg];
else if (!test_thread_64bit_stack(regs->u_regs[UREG_FP])) {
struct reg_window32 __user *win32;
flushw_user ();
win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
get_user(XR, &win32->locals[freg - 16]);
} else {
struct reg_window __user *win;
flushw_user ();
win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
get_user(XR, &win->locals[freg - 16]);
}
IR = 0;
switch ((insn >> 10) & 3) {
case 1: if (!XR) IR = 1; break; /* Register Zero */
case 2: if (XR <= 0) IR = 1; break; /* Register Less Than or Equal to Zero */
case 3: if (XR < 0) IR = 1; break; /* Register Less Than Zero */
}
if ((insn >> 10) & 4)
IR ^= 1;
break;
}
if (IR == 0) {
/* The fmov test was false. Do a nop instead */
current_thread_info()->xfsr[0] &= ~(FSR_CEXC_MASK);
regs->tpc = regs->tnpc;
regs->tnpc += 4;
return 1;
} else if (IR == 1) {
/* Change the instruction into plain fmovq */
insn = (insn & 0x3e00001f) | 0x81a00060;
TYPE(3,3,0,3,0,0,0);
}
}
}
if (type) {
argp rs1 = NULL, rs2 = NULL, rd = NULL;
/* Starting with UltraSPARC-T2, the cpu does not set the FP Trap
* Type field in the %fsr to unimplemented_FPop. Nor does it
* use the fp_exception_other trap. Instead it signals an
* illegal instruction and leaves the FP trap type field of
* the %fsr unchanged.
*/
if (!illegal_insn_trap) {
int ftt = (current_thread_info()->xfsr[0] >> 14) & 0x7;
if (ftt != (type >> 9))
goto err;
}
current_thread_info()->xfsr[0] &= ~0x1c000;
freg = ((insn >> 14) & 0x1f);
switch (type & 0x3) {
case 3: if (freg & 2) {
current_thread_info()->xfsr[0] |= (6 << 14) /* invalid_fp_register */;
goto err;
}
case 2: freg = ((freg & 1) << 5) | (freg & 0x1e);
case 1: rs1 = (argp)&f->regs[freg];
flags = (freg < 32) ? FPRS_DL : FPRS_DU;
if (!(current_thread_info()->fpsaved[0] & flags))
rs1 = (argp)&zero;
break;
}
switch (type & 0x7) {
case 7: FP_UNPACK_QP (QA, rs1); break;
case 6: FP_UNPACK_DP (DA, rs1); break;
case 5: FP_UNPACK_SP (SA, rs1); break;
}
freg = (insn & 0x1f);
switch ((type >> 3) & 0x3) {
case 3: if (freg & 2) {
current_thread_info()->xfsr[0] |= (6 << 14) /* invalid_fp_register */;
goto err;
}
case 2: freg = ((freg & 1) << 5) | (freg & 0x1e);
case 1: rs2 = (argp)&f->regs[freg];
flags = (freg < 32) ? FPRS_DL : FPRS_DU;
if (!(current_thread_info()->fpsaved[0] & flags))
rs2 = (argp)&zero;
break;
}
switch ((type >> 3) & 0x7) {
case 7: FP_UNPACK_QP (QB, rs2); break;
case 6: FP_UNPACK_DP (DB, rs2); break;
case 5: FP_UNPACK_SP (SB, rs2); break;
}
freg = ((insn >> 25) & 0x1f);
switch ((type >> 6) & 0x3) {
case 3: if (freg & 2) {
current_thread_info()->xfsr[0] |= (6 << 14) /* invalid_fp_register */;
goto err;
}
case 2: freg = ((freg & 1) << 5) | (freg & 0x1e);
case 1: rd = (argp)&f->regs[freg];
flags = (freg < 32) ? FPRS_DL : FPRS_DU;
if (!(current_thread_info()->fpsaved[0] & FPRS_FEF)) {
current_thread_info()->fpsaved[0] = FPRS_FEF;
current_thread_info()->gsr[0] = 0;
}
if (!(current_thread_info()->fpsaved[0] & flags)) {
if (freg < 32)
memset(f->regs, 0, 32*sizeof(u32));
else
memset(f->regs+32, 0, 32*sizeof(u32));
}
current_thread_info()->fpsaved[0] |= flags;
break;
}
switch ((insn >> 5) & 0x1ff) {
/* + */
case FADDS: FP_ADD_S (SR, SA, SB); break;
case FADDD: FP_ADD_D (DR, DA, DB); break;
case FADDQ: FP_ADD_Q (QR, QA, QB); break;
/* - */
case FSUBS: FP_SUB_S (SR, SA, SB); break;
case FSUBD: FP_SUB_D (DR, DA, DB); break;
case FSUBQ: FP_SUB_Q (QR, QA, QB); break;
/* * */
case FMULS: FP_MUL_S (SR, SA, SB); break;
case FSMULD: FP_CONV (D, S, 1, 1, DA, SA);
FP_CONV (D, S, 1, 1, DB, SB);
case FMULD: FP_MUL_D (DR, DA, DB); break;
case FDMULQ: FP_CONV (Q, D, 2, 1, QA, DA);
FP_CONV (Q, D, 2, 1, QB, DB);
case FMULQ: FP_MUL_Q (QR, QA, QB); break;
/* / */
case FDIVS: FP_DIV_S (SR, SA, SB); break;
case FDIVD: FP_DIV_D (DR, DA, DB); break;
case FDIVQ: FP_DIV_Q (QR, QA, QB); break;
/* sqrt */
case FSQRTS: FP_SQRT_S (SR, SB); break;
case FSQRTD: FP_SQRT_D (DR, DB); break;
case FSQRTQ: FP_SQRT_Q (QR, QB); break;
/* mov */
case FMOVQ: rd->q[0] = rs2->q[0]; rd->q[1] = rs2->q[1]; break;
case FABSQ: rd->q[0] = rs2->q[0] & 0x7fffffffffffffffUL; rd->q[1] = rs2->q[1]; break;
case FNEGQ: rd->q[0] = rs2->q[0] ^ 0x8000000000000000UL; rd->q[1] = rs2->q[1]; break;
/* float to int */
case FSTOI: FP_TO_INT_S (IR, SB, 32, 1); break;
case FDTOI: FP_TO_INT_D (IR, DB, 32, 1); break;
case FQTOI: FP_TO_INT_Q (IR, QB, 32, 1); break;
case FSTOX: FP_TO_INT_S (XR, SB, 64, 1); break;
case FDTOX: FP_TO_INT_D (XR, DB, 64, 1); break;
case FQTOX: FP_TO_INT_Q (XR, QB, 64, 1); break;
/* int to float */
case FITOQ: IR = rs2->s; FP_FROM_INT_Q (QR, IR, 32, int); break;
case FXTOQ: XR = rs2->d; FP_FROM_INT_Q (QR, XR, 64, long); break;
/* Only Ultra-III generates these */
case FXTOS: XR = rs2->d; FP_FROM_INT_S (SR, XR, 64, long); break;
case FXTOD: XR = rs2->d; FP_FROM_INT_D (DR, XR, 64, long); break;
#if 0 /* Optimized inline in sparc64/kernel/entry.S */
case FITOS: IR = rs2->s; FP_FROM_INT_S (SR, IR, 32, int); break;
#endif
case FITOD: IR = rs2->s; FP_FROM_INT_D (DR, IR, 32, int); break;
/* float to float */
case FSTOD: FP_CONV (D, S, 1, 1, DR, SB); break;
case FSTOQ: FP_CONV (Q, S, 2, 1, QR, SB); break;
case FDTOQ: FP_CONV (Q, D, 2, 1, QR, DB); break;
case FDTOS: FP_CONV (S, D, 1, 1, SR, DB); break;
case FQTOS: FP_CONV (S, Q, 1, 2, SR, QB); break;
case FQTOD: FP_CONV (D, Q, 1, 2, DR, QB); break;
/* comparison */
case FCMPQ:
case FCMPEQ:
FP_CMP_Q(XR, QB, QA, 3);
if (XR == 3 &&
(((insn >> 5) & 0x1ff) == FCMPEQ ||
FP_ISSIGNAN_Q(QA) ||
FP_ISSIGNAN_Q(QB)))
FP_SET_EXCEPTION (FP_EX_INVALID);
}
if (!FP_INHIBIT_RESULTS) {
switch ((type >> 6) & 0x7) {
case 0: xfsr = current_thread_info()->xfsr[0];
if (XR == -1) XR = 2;
switch (freg & 3) {
/* fcc0, 1, 2, 3 */
case 0: xfsr &= ~0xc00; xfsr |= (XR << 10); break;
case 1: xfsr &= ~0x300000000UL; xfsr |= (XR << 32); break;
case 2: xfsr &= ~0xc00000000UL; xfsr |= (XR << 34); break;
case 3: xfsr &= ~0x3000000000UL; xfsr |= (XR << 36); break;
}
current_thread_info()->xfsr[0] = xfsr;
break;
case 1: rd->s = IR; break;
case 2: rd->d = XR; break;
case 5: FP_PACK_SP (rd, SR); break;
case 6: FP_PACK_DP (rd, DR); break;
case 7: FP_PACK_QP (rd, QR); break;
}
}
if(_fex != 0)
return record_exception(regs, _fex);
/* Success and no exceptions detected. */
current_thread_info()->xfsr[0] &= ~(FSR_CEXC_MASK);
regs->tpc = regs->tnpc;
regs->tnpc += 4;
return 1;
}
err: return 0;
}
Reference in New Issue View Git Blame Copy Permalink