aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorPaul Mundt <lethal@linux-sh.org>2007-11-26 20:38:36 +0900
committerPaul Mundt <lethal@linux-sh.org>2008-01-28 13:18:57 +0900
commit74d99a5e262229ee865f6f68528d10b82471ead6 (patch)
tree1e2382c8779e1ee06226f9cc9acbf9a63eb5fc44
parenta8f67f4b4d4b74cd14d3540ade8657ebee543340 (diff)
sh: SH-2A FPU support.
Signed-off-by: Kieran Bingham <kbingham@mpc-data.co.uk> Signed-off-by: Paul Mundt <lethal@linux-sh.org>
-rw-r--r--arch/sh/Kconfig2
-rw-r--r--arch/sh/Makefile4
-rw-r--r--arch/sh/kernel/cpu/sh2/entry.S17
-rw-r--r--arch/sh/kernel/cpu/sh2a/Makefile2
-rw-r--r--arch/sh/kernel/cpu/sh2a/fpu.c633
-rw-r--r--arch/sh/kernel/cpu/sh3/ex.S2
-rw-r--r--arch/sh/kernel/cpu/sh4/fpu.c23
-rw-r--r--arch/sh/kernel/traps_32.c13
-rw-r--r--include/asm-sh/fpu.h2
-rw-r--r--include/asm-sh/sigcontext.h3
-rw-r--r--include/asm-sh/system.h2
11 files changed, 673 insertions, 30 deletions
diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index c18a5512ac8..2dc3b177193 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -161,6 +161,7 @@ config CPU_SUBTYPE_SH7619
config CPU_SUBTYPE_SH7203
bool "Support SH7203 processor"
select CPU_SH2A
+ select CPU_HAS_FPU
config CPU_SUBTYPE_SH7206
bool "Support SH7206 processor"
@@ -169,6 +170,7 @@ config CPU_SUBTYPE_SH7206
config CPU_SUBTYPE_SH7263
bool "Support SH7263 processor"
select CPU_SH2A
+ select CPU_HAS_FPU
# SH-3 Processor Support
diff --git a/arch/sh/Makefile b/arch/sh/Makefile
index f7cbc13fd2a..292d8618248 100644
--- a/arch/sh/Makefile
+++ b/arch/sh/Makefile
@@ -20,10 +20,6 @@ isa-$(CONFIG_CPU_SH4AL_DSP) := sh4al
isa-$(CONFIG_CPU_SH5) := shmedia
isa-$(CONFIG_SH_DSP) := $(isa-y)-dsp
-ifndef CONFIG_MMU
-isa-y := $(isa-y)-nommu
-endif
-
ifndef CONFIG_SH_DSP
ifndef CONFIG_SH_FPU
isa-y := $(isa-y)-nofpu
diff --git a/arch/sh/kernel/cpu/sh2/entry.S b/arch/sh/kernel/cpu/sh2/entry.S
index 4ff2334b4a3..7a26569e795 100644
--- a/arch/sh/kernel/cpu/sh2/entry.S
+++ b/arch/sh/kernel/cpu/sh2/entry.S
@@ -149,6 +149,14 @@ ENTRY(exception_handler)
mov #32,r8
cmp/hs r8,r9
bt trap_entry ! 64 > vec >= 32 is trap
+
+#if defined(CONFIG_SH_FPU)
+ mov #13,r8
+ cmp/eq r8,r9
+ bt 10f ! fpu
+ nop
+#endif
+
mov.l 4f,r8
mov r9,r4
shll2 r9
@@ -158,6 +166,10 @@ ENTRY(exception_handler)
cmp/eq r9,r8
bf 3f
mov.l 8f,r8 ! unhandled exception
+#if defined(CONFIG_SH_FPU)
+10:
+ mov.l 9f, r8 ! unhandled exception
+#endif
3:
mov.l 5f,r10
jmp @r8
@@ -177,7 +189,10 @@ interrupt_entry:
6: .long ret_from_irq
7: .long do_IRQ
8: .long do_exception_error
-
+#ifdef CONFIG_SH_FPU
+9: .long fpu_error_trap_handler
+#endif
+
trap_entry:
mov #0x30,r8
cmp/ge r8,r9 ! vector 0x20-0x2f is systemcall
diff --git a/arch/sh/kernel/cpu/sh2a/Makefile b/arch/sh/kernel/cpu/sh2a/Makefile
index 50e4d0ffdd6..b279cdc3a23 100644
--- a/arch/sh/kernel/cpu/sh2a/Makefile
+++ b/arch/sh/kernel/cpu/sh2a/Makefile
@@ -6,6 +6,8 @@ obj-y := common.o probe.o opcode_helper.o
common-y += $(addprefix ../sh2/, ex.o entry.o)
+obj-$(CONFIG_SH_FPU) += fpu.o
+
obj-$(CONFIG_CPU_SUBTYPE_SH7206) += setup-sh7206.o clock-sh7206.o
obj-$(CONFIG_CPU_SUBTYPE_SH7203) += setup-sh7203.o clock-sh7203.o
obj-$(CONFIG_CPU_SUBTYPE_SH7263) += setup-sh7203.o clock-sh7203.o
diff --git a/arch/sh/kernel/cpu/sh2a/fpu.c b/arch/sh/kernel/cpu/sh2a/fpu.c
new file mode 100644
index 00000000000..ff99562456f
--- /dev/null
+++ b/arch/sh/kernel/cpu/sh2a/fpu.c
@@ -0,0 +1,633 @@
+/*
+ * Save/restore floating point context for signal handlers.
+ *
+ * Copyright (C) 1999, 2000 Kaz Kojima & Niibe Yutaka
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * FIXME! These routines can be optimized in big endian case.
+ */
+#include <linux/sched.h>
+#include <linux/signal.h>
+#include <asm/processor.h>
+#include <asm/io.h>
+
+/* The PR (precision) bit in the FP Status Register must be clear when
+ * an frchg instruction is executed, otherwise the instruction is undefined.
+ * Executing frchg with PR set causes a trap on some SH4 implementations.
+ */
+
+#define FPSCR_RCHG 0x00000000
+
+
+/*
+ * Save FPU registers onto task structure.
+ * Assume called with FPU enabled (SR.FD=0).
+ */
+void
+save_fpu(struct task_struct *tsk, struct pt_regs *regs)
+{
+ unsigned long dummy;
+
+ clear_tsk_thread_flag(tsk, TIF_USEDFPU);
+ enable_fpu();
+ asm volatile("sts.l fpul, @-%0\n\t"
+ "sts.l fpscr, @-%0\n\t"
+ "fmov.s fr15, @-%0\n\t"
+ "fmov.s fr14, @-%0\n\t"
+ "fmov.s fr13, @-%0\n\t"
+ "fmov.s fr12, @-%0\n\t"
+ "fmov.s fr11, @-%0\n\t"
+ "fmov.s fr10, @-%0\n\t"
+ "fmov.s fr9, @-%0\n\t"
+ "fmov.s fr8, @-%0\n\t"
+ "fmov.s fr7, @-%0\n\t"
+ "fmov.s fr6, @-%0\n\t"
+ "fmov.s fr5, @-%0\n\t"
+ "fmov.s fr4, @-%0\n\t"
+ "fmov.s fr3, @-%0\n\t"
+ "fmov.s fr2, @-%0\n\t"
+ "fmov.s fr1, @-%0\n\t"
+ "fmov.s fr0, @-%0\n\t"
+ "lds %3, fpscr\n\t"
+ : "=r" (dummy)
+ : "0" ((char *)(&tsk->thread.fpu.hard.status)),
+ "r" (FPSCR_RCHG),
+ "r" (FPSCR_INIT)
+ : "memory");
+
+ disable_fpu();
+ release_fpu(regs);
+}
+
+static void
+restore_fpu(struct task_struct *tsk)
+{
+ unsigned long dummy;
+
+ enable_fpu();
+ asm volatile("fmov.s @%0+, fr0\n\t"
+ "fmov.s @%0+, fr1\n\t"
+ "fmov.s @%0+, fr2\n\t"
+ "fmov.s @%0+, fr3\n\t"
+ "fmov.s @%0+, fr4\n\t"
+ "fmov.s @%0+, fr5\n\t"
+ "fmov.s @%0+, fr6\n\t"
+ "fmov.s @%0+, fr7\n\t"
+ "fmov.s @%0+, fr8\n\t"
+ "fmov.s @%0+, fr9\n\t"
+ "fmov.s @%0+, fr10\n\t"
+ "fmov.s @%0+, fr11\n\t"
+ "fmov.s @%0+, fr12\n\t"
+ "fmov.s @%0+, fr13\n\t"
+ "fmov.s @%0+, fr14\n\t"
+ "fmov.s @%0+, fr15\n\t"
+ "lds.l @%0+, fpscr\n\t"
+ "lds.l @%0+, fpul\n\t"
+ : "=r" (dummy)
+ : "0" (&tsk->thread.fpu), "r" (FPSCR_RCHG)
+ : "memory");
+ disable_fpu();
+}
+
+/*
+ * Load the FPU with signalling NANS. This bit pattern we're using
+ * has the property that no matter wether considered as single or as
+ * double precission represents signaling NANS.
+ */
+
+static void
+fpu_init(void)
+{
+ enable_fpu();
+ asm volatile("lds %0, fpul\n\t"
+ "fsts fpul, fr0\n\t"
+ "fsts fpul, fr1\n\t"
+ "fsts fpul, fr2\n\t"
+ "fsts fpul, fr3\n\t"
+ "fsts fpul, fr4\n\t"
+ "fsts fpul, fr5\n\t"
+ "fsts fpul, fr6\n\t"
+ "fsts fpul, fr7\n\t"
+ "fsts fpul, fr8\n\t"
+ "fsts fpul, fr9\n\t"
+ "fsts fpul, fr10\n\t"
+ "fsts fpul, fr11\n\t"
+ "fsts fpul, fr12\n\t"
+ "fsts fpul, fr13\n\t"
+ "fsts fpul, fr14\n\t"
+ "fsts fpul, fr15\n\t"
+ "lds %2, fpscr\n\t"
+ : /* no output */
+ : "r" (0), "r" (FPSCR_RCHG), "r" (FPSCR_INIT));
+ disable_fpu();
+}
+
+/*
+ * Emulate arithmetic ops on denormalized number for some FPU insns.
+ */
+
+/* denormalized float * float */
+static int denormal_mulf(int hx, int hy)
+{
+ unsigned int ix, iy;
+ unsigned long long m, n;
+ int exp, w;
+
+ ix = hx & 0x7fffffff;
+ iy = hy & 0x7fffffff;
+ if (iy < 0x00800000 || ix == 0)
+ return ((hx ^ hy) & 0x80000000);
+
+ exp = (iy & 0x7f800000) >> 23;
+ ix &= 0x007fffff;
+ iy = (iy & 0x007fffff) | 0x00800000;
+ m = (unsigned long long)ix * iy;
+ n = m;
+ w = -1;
+ while (n) { n >>= 1; w++; }
+
+ /* FIXME: use guard bits */
+ exp += w - 126 - 46;
+ if (exp > 0)
+ ix = ((int) (m >> (w - 23)) & 0x007fffff) | (exp << 23);
+ else if (exp + 22 >= 0)
+ ix = (int) (m >> (w - 22 - exp)) & 0x007fffff;
+ else
+ ix = 0;
+
+ ix |= (hx ^ hy) & 0x80000000;
+ return ix;
+}
+
+/* denormalized double * double */
+static void mult64(unsigned long long x, unsigned long long y,
+ unsigned long long *highp, unsigned long long *lowp)
+{
+ unsigned long long sub0, sub1, sub2, sub3;
+ unsigned long long high, low;
+
+ sub0 = (x >> 32) * (unsigned long) (y >> 32);
+ sub1 = (x & 0xffffffffLL) * (unsigned long) (y >> 32);
+ sub2 = (x >> 32) * (unsigned long) (y & 0xffffffffLL);
+ sub3 = (x & 0xffffffffLL) * (unsigned long) (y & 0xffffffffLL);
+ low = sub3;
+ high = 0LL;
+ sub3 += (sub1 << 32);
+ if (low > sub3)
+ high++;
+ low = sub3;
+ sub3 += (sub2 << 32);
+ if (low > sub3)
+ high++;
+ low = sub3;
+ high += (sub1 >> 32) + (sub2 >> 32);
+ high += sub0;
+ *lowp = low;
+ *highp = high;
+}
+
+static inline long long rshift64(unsigned long long mh,
+ unsigned long long ml, int n)
+{
+ if (n >= 64)
+ return mh >> (n - 64);
+ return (mh << (64 - n)) | (ml >> n);
+}
+
+static long long denormal_muld(long long hx, long long hy)
+{
+ unsigned long long ix, iy;
+ unsigned long long mh, ml, nh, nl;
+ int exp, w;
+
+ ix = hx & 0x7fffffffffffffffLL;
+ iy = hy & 0x7fffffffffffffffLL;
+ if (iy < 0x0010000000000000LL || ix == 0)
+ return ((hx ^ hy) & 0x8000000000000000LL);
+
+ exp = (iy & 0x7ff0000000000000LL) >> 52;
+ ix &= 0x000fffffffffffffLL;
+ iy = (iy & 0x000fffffffffffffLL) | 0x0010000000000000LL;
+ mult64(ix, iy, &mh, &ml);
+ nh = mh;
+ nl = ml;
+ w = -1;
+ if (nh) {
+ while (nh) { nh >>= 1; w++;}
+ w += 64;
+ } else
+ while (nl) { nl >>= 1; w++;}
+
+ /* FIXME: use guard bits */
+ exp += w - 1022 - 52 * 2;
+ if (exp > 0)
+ ix = (rshift64(mh, ml, w - 52) & 0x000fffffffffffffLL)
+ | ((long long)exp << 52);
+ else if (exp + 51 >= 0)
+ ix = rshift64(mh, ml, w - 51 - exp) & 0x000fffffffffffffLL;
+ else
+ ix = 0;
+
+ ix |= (hx ^ hy) & 0x8000000000000000LL;
+ return ix;
+}
+
+/* ix - iy where iy: denormal and ix, iy >= 0 */
+static int denormal_subf1(unsigned int ix, unsigned int iy)
+{
+ int frac;
+ int exp;
+
+ if (ix < 0x00800000)
+ return ix - iy;
+
+ exp = (ix & 0x7f800000) >> 23;
+ if (exp - 1 > 31)
+ return ix;
+ iy >>= exp - 1;
+ if (iy == 0)
+ return ix;
+
+ frac = (ix & 0x007fffff) | 0x00800000;
+ frac -= iy;
+ while (frac < 0x00800000) {
+ if (--exp == 0)
+ return frac;
+ frac <<= 1;
+ }
+
+ return (exp << 23) | (frac & 0x007fffff);
+}
+
+/* ix + iy where iy: denormal and ix, iy >= 0 */
+static int denormal_addf1(unsigned int ix, unsigned int iy)
+{
+ int frac;
+ int exp;
+
+ if (ix < 0x00800000)
+ return ix + iy;
+
+ exp = (ix & 0x7f800000) >> 23;
+ if (exp - 1 > 31)
+ return ix;
+ iy >>= exp - 1;
+ if (iy == 0)
+ return ix;
+
+ frac = (ix & 0x007fffff) | 0x00800000;
+ frac += iy;
+ if (frac >= 0x01000000) {
+ frac >>= 1;
+ ++exp;
+ }
+
+ return (exp << 23) | (frac & 0x007fffff);
+}
+
+static int denormal_addf(int hx, int hy)
+{
+ unsigned int ix, iy;
+ int sign;
+
+ if ((hx ^ hy) & 0x80000000) {
+ sign = hx & 0x80000000;
+ ix = hx & 0x7fffffff;
+ iy = hy & 0x7fffffff;
+ if (iy < 0x00800000) {
+ ix = denormal_subf1(ix, iy);
+ if (ix < 0) {
+ ix = -ix;
+ sign ^= 0x80000000;
+ }
+ } else {
+ ix = denormal_subf1(iy, ix);
+ sign ^= 0x80000000;
+ }
+ } else {
+ sign = hx & 0x80000000;
+ ix = hx & 0x7fffffff;
+ iy = hy & 0x7fffffff;
+ if (iy < 0x00800000)
+ ix = denormal_addf1(ix, iy);
+ else
+ ix = denormal_addf1(iy, ix);
+ }
+
+ return sign | ix;
+}
+
+/* ix - iy where iy: denormal and ix, iy >= 0 */
+static long long denormal_subd1(unsigned long long ix, unsigned long long iy)
+{
+ long long frac;
+ int exp;
+
+ if (ix < 0x0010000000000000LL)
+ return ix - iy;
+
+ exp = (ix & 0x7ff0000000000000LL) >> 52;
+ if (exp - 1 > 63)
+ return ix;
+ iy >>= exp - 1;
+ if (iy == 0)
+ return ix;
+
+ frac = (ix & 0x000fffffffffffffLL) | 0x0010000000000000LL;
+ frac -= iy;
+ while (frac < 0x0010000000000000LL) {
+ if (--exp == 0)
+ return frac;
+ frac <<= 1;
+ }
+
+ return ((long long)exp << 52) | (frac & 0x000fffffffffffffLL);
+}
+
+/* ix + iy where iy: denormal and ix, iy >= 0 */
+static long long denormal_addd1(unsigned long long ix, unsigned long long iy)
+{
+ long long frac;
+ long long exp;
+
+ if (ix < 0x0010000000000000LL)
+ return ix + iy;
+
+ exp = (ix & 0x7ff0000000000000LL) >> 52;
+ if (exp - 1 > 63)
+ return ix;
+ iy >>= exp - 1;
+ if (iy == 0)
+ return ix;
+
+ frac = (ix & 0x000fffffffffffffLL) | 0x0010000000000000LL;
+ frac += iy;
+ if (frac >= 0x0020000000000000LL) {
+ frac >>= 1;
+ ++exp;
+ }
+
+ return (exp << 52) | (frac & 0x000fffffffffffffLL);
+}
+
+static long long denormal_addd(long long hx, long long hy)
+{
+ unsigned long long ix, iy;
+ long long sign;
+
+ if ((hx ^ hy) & 0x8000000000000000LL) {
+ sign = hx & 0x8000000000000000LL;
+ ix = hx & 0x7fffffffffffffffLL;
+ iy = hy & 0x7fffffffffffffffLL;
+ if (iy < 0x0010000000000000LL) {
+ ix = denormal_subd1(ix, iy);
+ if (ix < 0) {
+ ix = -ix;
+ sign ^= 0x8000000000000000LL;
+ }
+ } else {
+ ix = denormal_subd1(iy, ix);
+ sign ^= 0x8000000000000000LL;
+ }
+ } else {
+ sign = hx & 0x8000000000000000LL;
+ ix = hx & 0x7fffffffffffffffLL;
+ iy = hy & 0x7fffffffffffffffLL;
+ if (iy < 0x0010000000000000LL)
+ ix = denormal_addd1(ix, iy);
+ else
+ ix = denormal_addd1(iy, ix);
+ }
+
+ return sign | ix;
+}
+
+/**
+ * denormal_to_double - Given denormalized float number,
+ * store double float
+ *
+ * @fpu: Pointer to sh_fpu_hard structure
+ * @n: Index to FP register
+ */
+static void
+denormal_to_double (struct sh_fpu_hard_struct *fpu, int n)
+{
+ unsigned long du, dl;
+ unsigned long x = fpu->fpul;
+ int exp = 1023 - 126;
+
+ if (x != 0 && (x & 0x7f800000) == 0) {
+ du = (x & 0x80000000);
+ while ((x & 0x00800000) == 0) {
+ x <<= 1;
+ exp--;
+ }
+ x &= 0x007fffff;
+ du |= (exp << 20) | (x >> 3);
+ dl = x << 29;
+
+ fpu->fp_regs[n] = du;
+ fpu->fp_regs[n+1] = dl;
+ }
+}
+
+/**
+ * ieee_fpe_handler - Handle denormalized number exception
+ *
+ * @regs: Pointer to register structure
+ *
+ * Returns 1 when it's handled (should not cause exception).
+ */
+static int
+ieee_fpe_handler (struct pt_regs *regs)
+{
+ unsigned short insn = *(unsigned short *) regs->pc;
+ unsigned short finsn;
+ unsigned long nextpc;
+ int nib[4] = {
+ (insn >> 12) & 0xf,
+ (insn >> 8) & 0xf,
+ (insn >> 4) & 0xf,
+ insn & 0xf};
+
+ if (nib[0] == 0xb ||
+ (nib[0] == 0x4 && nib[2] == 0x0 && nib[3] == 0xb)) /* bsr & jsr */
+ regs->pr = regs->pc + 4;
+ if (nib[0] == 0xa || nib[0] == 0xb) { /* bra & bsr */
+ nextpc = regs->pc + 4 + ((short) ((insn & 0xfff) << 4) >> 3);
+ finsn = *(unsigned short *) (regs->pc + 2);
+ } else if (nib[0] == 0x8 && nib[1] == 0xd) { /* bt/s */
+ if (regs->sr & 1)
+ nextpc = regs->pc + 4 + ((char) (insn & 0xff) << 1);
+ else
+ nextpc = regs->pc + 4;
+ finsn = *(unsigned short *) (regs->pc + 2);
+ } else if (nib[0] == 0x8 && nib[1] == 0xf) { /* bf/s */
+ if (regs->sr & 1)
+ nextpc = regs->pc + 4;
+ else
+ nextpc = regs->pc + 4 + ((char) (insn & 0xff) << 1);
+ finsn = *(unsigned short *) (regs->pc + 2);
+ } else if (nib[0] == 0x4 && nib[3] == 0xb &&
+ (nib[2] == 0x0 || nib[2] == 0x2)) { /* jmp & jsr */
+ nextpc = regs->regs[nib[1]];
+ finsn = *(unsigned short *) (regs->pc + 2);
+ } else if (nib[0] == 0x0 && nib[3] == 0x3 &&
+ (nib[2] == 0x0 || nib[2] == 0x2)) { /* braf & bsrf */
+ nextpc = regs->pc + 4 + regs->regs[nib[1]];
+ finsn = *(unsigned short *) (regs->pc + 2);
+ } else if (insn == 0x000b) { /* rts */
+ nextpc = regs->pr;
+ finsn = *(unsigned short *) (regs->pc + 2);
+ } else {
+ nextpc = regs->pc + 2;
+ finsn = insn;
+ }
+
+#define FPSCR_FPU_ERROR (1 << 17)
+
+ if ((finsn & 0xf1ff) == 0xf0ad) { /* fcnvsd */
+ struct task_struct *tsk = current;
+
+ if ((tsk->thread.fpu.hard.fpscr & FPSCR_FPU_ERROR)) {
+ /* FPU error */
+ denormal_to_double (&tsk->thread.fpu.hard,
+ (finsn >> 8) & 0xf);
+ } else
+ return 0;
+
+ regs->pc = nextpc;
+ return 1;
+ } else if ((finsn & 0xf00f) == 0xf002) { /* fmul */
+ struct task_struct *tsk = current;
+ int fpscr;
+ int n, m, prec;
+ unsigned int hx, hy;
+
+ n = (finsn >> 8) & 0xf;
+ m = (finsn >> 4) & 0xf;
+ hx = tsk->thread.fpu.hard.fp_regs[n];
+ hy = tsk->thread.fpu.hard.fp_regs[m];
+ fpscr = tsk->thread.fpu.hard.fpscr;
+ prec = fpscr & (1 << 19);
+
+ if ((fpscr & FPSCR_FPU_ERROR)
+ && (prec && ((hx & 0x7fffffff) < 0x00100000
+ || (hy & 0x7fffffff) < 0x00100000))) {
+ long long llx, lly;
+
+ /* FPU error because of denormal */
+ llx = ((long long) hx << 32)
+ | tsk->thread.fpu.hard.fp_regs[n+1];
+ lly = ((long long) hy << 32)
+ | tsk->thread.fpu.hard.fp_regs[m+1];
+ if ((hx & 0x7fffffff) >= 0x00100000)
+ llx = denormal_muld(lly, llx);
+ else
+ llx = denormal_muld(llx, lly);
+ tsk->thread.fpu.hard.fp_regs[n] = llx >> 32;
+ tsk->thread.fpu.hard.fp_regs[n+1] = llx & 0xffffffff;
+ } else if ((fpscr & FPSCR_FPU_ERROR)
+ && (!prec && ((hx & 0x7fffffff) < 0x00800000
+ || (hy & 0x7fffffff) < 0x00800000))) {
+ /* FPU error because of denormal */
+ if ((hx & 0x7fffffff) >= 0x00800000)
+ hx = denormal_mulf(hy, hx);
+ else
+ hx = denormal_mulf(hx, hy);
+ tsk->thread.fpu.hard.fp_regs[n] = hx;
+ } else
+ return 0;
+
+ regs->pc = nextpc;
+ return 1;
+ } else if ((finsn & 0xf00e) == 0xf000) { /* fadd, fsub */
+ struct task_struct *tsk = current;
+ int fpscr;
+ int n, m, prec;
+ unsigned int hx, hy;
+
+ n = (finsn >> 8) & 0xf;
+ m = (finsn >> 4) & 0xf;
+ hx = tsk->thread.fpu.hard.fp_regs[n];
+ hy = tsk->thread.fpu.hard.fp_regs[m];
+ fpscr = tsk->thread.fpu.hard.fpscr;
+ prec = fpscr & (1 << 19);
+
+ if ((fpscr & FPSCR_FPU_ERROR)
+ && (prec && ((hx & 0x7fffffff) < 0x00100000
+ || (hy & 0x7fffffff) < 0x00100000))) {
+ long long llx, lly;
+
+ /* FPU error because of denormal */
+ llx = ((long long) hx << 32)
+ | tsk->thread.fpu.hard.fp_regs[n+1];
+ lly = ((long long) hy << 32)
+ | tsk->thread.fpu.hard.fp_regs[m+1];
+ if ((finsn & 0xf00f) == 0xf000)
+ llx = denormal_addd(llx, lly);
+ else
+ llx = denormal_addd(llx, lly ^ (1LL << 63));
+ tsk->thread.fpu.hard.fp_regs[n] = llx >> 32;
+ tsk->thread.fpu.hard.fp_regs[n+1] = llx & 0xffffffff;
+ } else if ((fpscr & FPSCR_FPU_ERROR)
+ && (!prec && ((hx & 0x7fffffff) < 0x00800000
+ || (hy & 0x7fffffff) < 0x00800000))) {
+ /* FPU error because of denormal */
+ if ((finsn & 0xf00f) == 0xf000)
+ hx = denormal_addf(hx, hy);
+ else
+ hx = denormal_addf(hx, hy ^ 0x80000000);
+ tsk->thread.fpu.hard.fp_regs[n] = hx;
+ } else
+ return 0;
+
+ regs->pc = nextpc;
+ return 1;
+ }
+
+ return 0;
+}
+
+BUILD_TRAP_HANDLER(fpu_error)
+{
+ struct task_struct *tsk = current;
+ TRAP_HANDLER_DECL;
+
+ save_fpu(tsk, regs);
+ if (ieee_fpe_handler(regs)) {
+ tsk->thread.fpu.hard.fpscr &=
+ ~(FPSCR_CAUSE_MASK | FPSCR_FLAG_MASK);
+ grab_fpu(regs);
+ restore_fpu(tsk);
+ set_tsk_thread_flag(tsk, TIF_USEDFPU);
+ return;
+ }
+
+ force_sig(SIGFPE, tsk);
+}
+
+BUILD_TRAP_HANDLER(fpu_state_restore)
+{
+ struct task_struct *tsk = current;
+ TRAP_HANDLER_DECL;
+
+ grab_fpu(regs);
+ if (!user_mode(regs)) {
+ printk(KERN_ERR "BUG: FPU is used in kernel mode.\n");
+ return;
+ }
+
+ if (used_math()) {
+ /* Using the FPU again. */
+ restore_fpu(tsk);
+ } else {
+ /* First time FPU user. */
+ fpu_init();
+ set_used_math();
+ }
+ set_tsk_thread_flag(tsk, TIF_USEDFPU);
+}
diff --git a/arch/sh/kernel/cpu/sh3/ex.S b/arch/sh/kernel/cpu/sh3/ex.S
index b6abf38d3a8..11b6d9c6eda 100644
--- a/arch/sh/kernel/cpu/sh3/ex.S
+++ b/arch/sh/kernel/cpu/sh3/ex.S
@@ -36,7 +36,7 @@ ENTRY(exception_handling_table)
.long exception_error ! address error store /* 100 */
#endif
#if defined(CONFIG_SH_FPU)
- .long do_fpu_error /* 120 */
+ .long fpu_error_trap_handler /* 120 */
#else
.long exception_error /* 120 */
#endif
diff --git a/arch/sh/kernel/cpu/sh4/fpu.c b/arch/sh/kernel/cpu/sh4/fpu.c
index c5a4fc77fa0..e624180b446 100644
--- a/arch/sh/kernel/cpu/sh4/fpu.c
+++ b/arch/sh/kernel/cpu/sh4/fpu.c
@@ -82,8 +82,8 @@ save_fpu(struct task_struct *tsk, struct pt_regs *regs)
"r" (FPSCR_INIT)
: "memory");
- disable_fpu();
- release_fpu(regs);
+ disable_fpu();
+ release_fpu(regs);
}
static void
@@ -91,7 +91,7 @@ restore_fpu(struct task_struct *tsk)
{
unsigned long dummy;
- enable_fpu();
+ enable_fpu();
asm volatile("lds %2, fpscr\n\t"
"fmov.s @%0+, fr0\n\t"
"fmov.s @%0+, fr1\n\t"
@@ -138,7 +138,7 @@ restore_fpu(struct task_struct *tsk)
/*
* Load the FPU with signalling NANS. This bit pattern we're using
* has the property that no matter wether considered as single or as
- * double precision represents signaling NANS.
+ * double precision represents signaling NANS.
*/
static void
@@ -184,7 +184,7 @@ fpu_init(void)
"lds %2, fpscr\n\t"
: /* no output */
: "r" (0), "r" (FPSCR_RCHG), "r" (FPSCR_INIT));
- disable_fpu();
+ disable_fpu();
}
/**
@@ -238,7 +238,6 @@ ieee_fpe_handler (struct pt_regs *regs)
if (nib[0] == 0xb ||
(nib[0] == 0x4 && nib[2] == 0x0 && nib[3] == 0xb)) /* bsr & jsr */
regs->pr = regs->pc + 4;
-
if (nib[0] == 0xa || nib[0] == 0xb) { /* bra & bsr */
nextpc = regs->pc + 4 + ((short) ((insn & 0xfff) << 4) >> 3);
finsn = *(unsigned short *) (regs->pc + 2);
@@ -293,12 +292,10 @@ ieee_fpe_handler (struct pt_regs *regs)
return 0;
}
-asmlinkage void
-do_fpu_error(unsigned long r4, unsigned long r5, unsigned long r6,
- unsigned long r7, struct pt_regs __regs)
+BUILD_TRAP_HANDLER(fpu_error)
{
- struct pt_regs *regs = RELOC_HIDE(&__regs, 0);
struct task_struct *tsk = current;
+ TRAP_HANDLER_DECL;
if (ieee_fpe_handler(regs))
return;
@@ -308,12 +305,10 @@ do_fpu_error(unsigned long r4, unsigned long r5, unsigned long r6,
force_sig(SIGFPE, tsk);
}
-asmlinkage void
-do_fpu_state_restore(unsigned long r4, unsigned long r5, unsigned long r6,
- unsigned long r7, struct pt_regs __regs)
+BUILD_TRAP_HANDLER(fpu_state_restore)
{
- struct pt_regs *regs = RELOC_HIDE(&__regs, 0);
struct task_struct *tsk = current;
+ TRAP_HANDLER_DECL;
grab_fpu(regs);
if (!user_mode(regs)) {
diff --git a/arch/sh/kernel/traps_32.c b/arch/sh/kernel/traps_32.c
index 0d05fb3c48e..2e58f7a6b74 100644
--- a/arch/sh/kernel/traps_32.c
+++ b/arch/sh/kernel/traps_32.c
@@ -662,11 +662,6 @@ asmlinkage void do_divide_error(unsigned long r4, unsigned long r5,
}
#endif
-/* arch/sh/kernel/cpu/sh4/fpu.c */
-extern int do_fpu_inst(unsigned short, struct pt_regs *);
-extern asmlinkage void do_fpu_state_restore(unsigned long r4, unsigned long r5,
- unsigned long r6, unsigned long r7, struct pt_regs __regs);
-
asmlinkage void do_reserved_inst(unsigned long r4, unsigned long r5,
unsigned long r6, unsigned long r7,
struct pt_regs __regs)
@@ -853,11 +848,11 @@ void __init trap_init(void)
set_exception_table_evt(0x820, do_illegal_slot_inst);
#elif defined(CONFIG_SH_FPU)
#ifdef CONFIG_CPU_SUBTYPE_SHX3
- set_exception_table_evt(0xd80, do_fpu_state_restore);
- set_exception_table_evt(0xda0, do_fpu_state_restore);
+ set_exception_table_evt(0xd80, fpu_state_restore_trap_handler);
+ set_exception_table_evt(0xda0, fpu_state_restore_trap_handler);
#else
- set_exception_table_evt(0x800, do_fpu_state_restore);
- set_exception_table_evt(0x820, do_fpu_state_restore);
+ set_exception_table_evt(0x800, fpu_state_restore_trap_handler);
+ set_exception_table_evt(0x820, fpu_state_restore_trap_handler);
#endif
#endif
diff --git a/include/asm-sh/fpu.h b/include/asm-sh/fpu.h
index 33db698a6b4..f8429880a27 100644
--- a/include/asm-sh/fpu.h
+++ b/include/asm-sh/fpu.h
@@ -26,6 +26,8 @@ extern void save_fpu(struct task_struct *__tsk, struct pt_regs *regs);
#define save_fpu(tsk, regs) do { } while (0)
#endif
+extern int do_fpu_inst(unsigned short, struct pt_regs *);
+
#define unlazy_fpu(tsk, regs) do { \
if (test_tsk_thread_flag(tsk, TIF_USEDFPU)) { \
save_fpu(tsk, regs); \
diff --git a/include/asm-sh/sigcontext.h b/include/asm-sh/sigcontext.h
index 8583143fa28..8ce1435bc0b 100644
--- a/include/asm-sh/sigcontext.h
+++ b/include/asm-sh/sigcontext.h
@@ -25,7 +25,8 @@ struct sigcontext {
unsigned long sc_mach;
unsigned long sc_macl;
-#if defined(__SH4__) || defined(CONFIG_CPU_SH4)
+#if defined(__SH4__) || defined(CONFIG_CPU_SH4) || \
+ defined(__SH2A__) || defined(CONFIG_CPU_SH2A)
/* FPU registers */
unsigned long sc_fpregs[16];
unsigned long sc_xfpregs[16];
diff --git a/include/asm-sh/system.h b/include/asm-sh/system.h
index 8b01fc4a56a..ad3d2a63613 100644
--- a/include/asm-sh/system.h
+++ b/include/asm-sh/system.h
@@ -205,6 +205,8 @@ asmlinkage void name##_trap_handler(unsigned int vec, struct pt_regs *regs)
BUILD_TRAP_HANDLER(address_error);
BUILD_TRAP_HANDLER(debug);
BUILD_TRAP_HANDLER(bug);
+BUILD_TRAP_HANDLER(fpu_error);
+BUILD_TRAP_HANDLER(fpu_state_restore);
#define arch_align_stack(x) (x)