svn commit: r366904 - in head/sys: amd64/ia32 i386/i386

[Konstantin Belousov]

Author: kib
Date: Wed Oct 21 00:15:12 2020
New Revision: 366904
URL: https://svnweb.freebsd.org/changeset/base/366904

Log:
  Improve FPU Tag Word reconstruction on i386 to indicate register states.
  
  Improve the code reconstructing en_tw in struct fpreg32 from FXSAVE
  results so that all register states are indicated correctly.  The
  previous code unconditionally mapped non-empty register state to
  'normalized value' constant.  The new code explicitly distinguishes
  the 'zero value' and 'special value' constants as well.  This improves
  consistency between real FSAVE and translation from FXSAVE, and
  ensures that tests using PT_GETFPREGS can rely on a single correct
  value independently of the underlying implementation.
  
  PR:	250454
  Sponsored by:	The FreeBSD Foundation
  Obtained from:	Moritz Systems
  Submitted by:	Michał Górny <mgorny at moritz.systems>
  Discussed with:	emaste
  MFC after:	1 week
  Differential revision:	https://reviews.freebsd.org/D26856

Modified:
  head/sys/amd64/ia32/ia32_reg.c
  head/sys/i386/i386/npx.c

Modified: head/sys/amd64/ia32/ia32_reg.c
==============================================================================
--- head/sys/amd64/ia32/ia32_reg.c	Tue Oct 20 20:59:13 2020	(r366903)
+++ head/sys/amd64/ia32/ia32_reg.c	Wed Oct 21 00:15:12 2020	(r366904)
@@ -145,7 +145,11 @@ fill_fpregs32(struct thread *td, struct fpreg32 *regs)
 	struct save87 *sv_87;
 	struct env87 *penv_87;
 	struct envxmm *penv_xmm;
-	int i;
+	struct fpacc87 *fx_reg;
+	int i, st;
+	uint64_t mantissa;
+	uint16_t tw, exp;
+	uint8_t ab_tw;
 
 	bzero(regs, sizeof(*regs));
 	sv_87 = (struct save87 *)regs;
@@ -171,13 +175,39 @@ fill_fpregs32(struct thread *td, struct fpreg32 *regs)
 	/* Entry into the kernel always sets TF_HASSEGS */
 	penv_87->en_fos = td->td_frame->tf_ds;
 
-	/* FPU registers and tags */
-	penv_87->en_tw = 0xffff;
-	for (i = 0; i < 8; ++i) {
-		sv_87->sv_ac[i] = sv_fpu->sv_fp[i].fp_acc;
-		if ((penv_xmm->en_tw & (1 << i)) != 0)
-			penv_87->en_tw &= ~(3 << i * 2);
+	/*
+	 * FPU registers and tags.
+	 * For ST(i), i = fpu_reg - top; we start with fpu_reg=7.
+	 */
+	st = 7 - ((penv_xmm->en_sw >> 11) & 7);
+	ab_tw = penv_xmm->en_tw;
+	tw = 0;
+	for (i = 0x80; i != 0; i >>= 1) {
+		sv_87->sv_ac[st] = sv_fpu->sv_fp[st].fp_acc;
+		tw <<= 2;
+		if ((ab_tw & i) != 0) {
+			/* Non-empty - we need to check ST(i) */
+			fx_reg = &sv_fpu->sv_fp[st].fp_acc;
+			/* The first 64 bits contain the mantissa. */
+			mantissa = *((uint64_t *)fx_reg->fp_bytes);
+			/*
+			 * The final 16 bits contain the sign bit and the exponent.
+			 * Mask the sign bit since it is of no consequence to these
+			 * tests.
+			 */
+			exp = *((uint16_t *)&fx_reg->fp_bytes[8]) & 0x7fff;
+			if (exp == 0) {
+				if (mantissa == 0)
+					tw |= 1; /* Zero */
+				else
+					tw |= 2; /* Denormal */
+			} else if (exp == 0x7fff)
+				tw |= 2; /* Infinity or NaN */
+		} else
+			tw |= 3; /* Empty */
+		st = (st - 1) & 7;
 	}
+	penv_87->en_tw = tw;
 
 	return (0);
 }

Modified: head/sys/i386/i386/npx.c
==============================================================================
--- head/sys/i386/i386/npx.c	Tue Oct 20 20:59:13 2020	(r366903)
+++ head/sys/i386/i386/npx.c	Wed Oct 21 00:15:12 2020	(r366904)
@@ -1154,7 +1154,11 @@ npx_fill_fpregs_xmm1(struct savexmm *sv_xmm, struct sa
 {
 	struct env87 *penv_87;
 	struct envxmm *penv_xmm;
-	int i;
+	struct fpacc87 *fx_reg;
+	int i, st;
+	uint64_t mantissa;
+	uint16_t tw, exp;
+	uint8_t ab_tw;
 
 	penv_87 = &sv_87->sv_env;
 	penv_xmm = &sv_xmm->sv_env;
@@ -1168,14 +1172,39 @@ npx_fill_fpregs_xmm1(struct savexmm *sv_xmm, struct sa
 	penv_87->en_foo = penv_xmm->en_foo;
 	penv_87->en_fos = penv_xmm->en_fos;
 
-	/* FPU registers and tags */
-	penv_87->en_tw = 0xffff;
-	for (i = 0; i < 8; ++i) {
-		sv_87->sv_ac[i] = sv_xmm->sv_fp[i].fp_acc;
-		if ((penv_xmm->en_tw & (1 << i)) != 0)
-			/* zero and special are set as valid */
-			penv_87->en_tw &= ~(3 << i * 2);
+	/*
+	 * FPU registers and tags.
+	 * For ST(i), i = fpu_reg - top; we start with fpu_reg=7.
+	 */
+	st = 7 - ((penv_xmm->en_sw >> 11) & 7);
+	ab_tw = penv_xmm->en_tw;
+	tw = 0;
+	for (i = 0x80; i != 0; i >>= 1) {
+		sv_87->sv_ac[st] = sv_xmm->sv_fp[st].fp_acc;
+		tw <<= 2;
+		if (ab_tw & i) {
+			/* Non-empty - we need to check ST(i) */
+			fx_reg = &sv_xmm->sv_fp[st].fp_acc;
+			/* The first 64 bits contain the mantissa. */
+			mantissa = *((uint64_t *)fx_reg->fp_bytes);
+			/*
+			 * The final 16 bits contain the sign bit and the exponent.
+			 * Mask the sign bit since it is of no consequence to these
+			 * tests.
+			 */
+			exp = *((uint16_t *)&fx_reg->fp_bytes[8]) & 0x7fff;
+			if (exp == 0) {
+				if (mantissa == 0)
+					tw |= 1; /* Zero */
+				else
+					tw |= 2; /* Denormal */
+			} else if (exp == 0x7fff)
+				tw |= 2; /* Infinity or NaN */
+		} else
+			tw |= 3; /* Empty */
+		st = (st - 1) & 7;
 	}
+	penv_87->en_tw = tw;
 }
 
 void