[PATCH v3 0/4] powerpc/64: copy_tofrom_user exception handling improvements

[PATCH v3 0/4] powerpc/64: copy_tofrom_user exception handling improvements

[Paul Mackerras]

This is a repost of a series that I posted back in 2016 but which was
never applied.  It aims to make the exception handling code in
__copy_tofrom_user_base clearer and easier to verify, and strengthens
the selftests for the user copy code to test all the paths and to test
the exception handling.  Finally it then fixes a deficiency in that
when copying to userspace we don't always copy quite as many bytes as
we could.

I have rebased this series on top of the powerpc next branch as of
today.

Paul.

 arch/powerpc/lib/copyuser_64.S                     | 585 +++++++++------------
 arch/powerpc/lib/copyuser_power7.S                 |  21 +-
 arch/powerpc/lib/memcpy_64.S                       |   9 +-
 arch/powerpc/lib/memcpy_power7.S                   |  22 +-
 .../testing/selftests/powerpc/copyloops/.gitignore |  17 +-
 tools/testing/selftests/powerpc/copyloops/Makefile |  44 +-
 .../selftests/powerpc/copyloops/asm/ppc_asm.h      |  44 +-
 .../powerpc/copyloops/copy_tofrom_user_reference.S |  24 +
 .../selftests/powerpc/copyloops/exc_validate.c     | 124 +++++
 tools/testing/selftests/powerpc/copyloops/stubs.S  |  19 +
 10 files changed, 516 insertions(+), 393 deletions(-)

[PATCH v3 1/4] powerpc/64: Make exception table clearer in __copy_tofrom_user_base

[Paul Mackerras]

This aims to make the generation of exception table entries for the
loads and stores in __copy_tofrom_user_base clearer and easier to
verify.  Instead of having a series of local labels on the loads and
stores, with a series of corresponding labels later for the exception
handlers, we now use macros to generate exception table entries at the
point of each load and store that could potentially trap.  We do this
with the macros lex (load exception) and stex (store exception).
These macros are used right before the load or store to which they
apply.

Some complexity is introduced by the fact that we have some more work
to do after hitting an exception, because we need to calculate and
return the number of bytes not copied.  The code uses r3 as the
current pointer into the destination buffer, that is, the address of
the first byte of the destination that has not been modified.
However, at various points in the copy loops, r3 can be 4, 8, 16 or 24
bytes behind that point.

To express this offset in an understandable way, we define a symbol
r3_offset which is updated at various points so that it equal to the
difference between the address of the first unmodified byte of the
destination and the value in r3.  (In fact it only needs to be
accurate at the point of each lex or stex macro invocation.)

The rules for updating r3_offset are as follows:

* It starts out at 0
* An addi r3,r3,N instruction decreases r3_offset by N
* A store instruction (stb, sth, stw, std) to N(r3)
  increases r3_offset by the width of the store (1, 2, 4, 8)
* A store with update instruction (stbu, sthu, stwu, stdu) to N(r3)
  sets r3_offset to the width of the store.

There is some trickiness to the way that the lex and stex macros and
the associated exception handlers work.  I would have liked to use
the current value of r3_offset in the name of the symbol used as
the exception handler, as in ".Lld_exc_$(r3_offset)" and then
have symbols .Lld_exc_0, .Lld_exc_8, .Lld_exc_16 etc. corresponding
to the offsets that needed to be added to r3.  However, I couldn't
see a way to do that with gas.

Instead, the exception handler address is .Lld_exc - r3_offset or
.Lst_exc - r3_offset, that is, the distance ahead of .Lld_exc/.Lst_exc
that we start executing is equal to the amount that we need to add to
r3.  This works because r3_offset is always a small multiple of 4,
and our instructions are 4 bytes long.  This means that before
.Lld_exc and .Lst_exc, we have a sequence of instructions that
increments r3 by 4, 8, 16 or 24 depending on where we start.  The
sequence increments r3 by 4 per instruction (on average).

We also replace the exception table for the 4k copy loop by a
macro per load or store.  These loads and stores all use exactly
the same exception handler, which simply resets the argument registers
r3, r4 and r5 to there original values and re-does the whole copy
using the slower loop.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
---
 arch/powerpc/lib/copyuser_64.S | 551 +++++++++++++++++------------------------
 1 file changed, 225 insertions(+), 326 deletions(-)

diff --git a/arch/powerpc/lib/copyuser_64.S b/arch/powerpc/lib/copyuser_64.S
index 2d6f128..8c5033c 100644
--- a/arch/powerpc/lib/copyuser_64.S
+++ b/arch/powerpc/lib/copyuser_64.S
@@ -20,6 +20,28 @@
 #define sHd sld		/* Shift towards high-numbered address. */
 #endif
 
+/*
+ * These macros are used to generate exception table entries.
+ * The exception handlers below use the original arguments
+ * (stored on the stack) and the point where we're up to in
+ * the destination buffer, i.e. the address of the first
+ * unmodified byte.  Generally r3 points into the destination
+ * buffer, but the first unmodified byte is at a variable
+ * offset from r3.  In the code below, the symbol r3_offset
+ * is set to indicate the current offset at each point in
+ * the code.  This offset is then used as a negative offset
+ * from the exception handler code, and those instructions
+ * before the exception handlers are addi instructions that
+ * adjust r3 to point to the correct place.
+ */
+	.macro	lex		/* exception handler for load */
+100:	EX_TABLE(100b, .Lld_exc - r3_offset)
+	.endm
+
+	.macro	stex		/* exception handler for store */
+100:	EX_TABLE(100b, .Lst_exc - r3_offset)
+	.endm
+
 	.align	7
 _GLOBAL_TOC(__copy_tofrom_user)
 #ifdef CONFIG_PPC_BOOK3S_64
@@ -30,7 +52,7 @@ FTR_SECTION_ELSE
 ALT_FTR_SECTION_END_IFCLR(CPU_FTR_VMX_COPY)
 #endif
 _GLOBAL(__copy_tofrom_user_base)
-	/* first check for a whole page copy on a page boundary */
+	/* first check for a 4kB copy on a 4kB boundary */
 	cmpldi	cr1,r5,16
 	cmpdi	cr6,r5,4096
 	or	r0,r3,r4
@@ -59,6 +81,7 @@ ALT_FTR_SECTION_END(CPU_FTR_UNALIGNED_LD_STD | CPU_FTR_CP_USE_DCBTZ, \
 		    CPU_FTR_UNALIGNED_LD_STD)
 .Ldst_aligned:
 	addi	r3,r3,-16
+r3_offset = 16
 BEGIN_FTR_SECTION
 	andi.	r0,r4,7
 	bne	.Lsrc_unaligned
@@ -66,57 +89,69 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 	blt	cr1,.Ldo_tail		/* if < 16 bytes to copy */
 	srdi	r0,r5,5
 	cmpdi	cr1,r0,0
-20:	ld	r7,0(r4)
-220:	ld	r6,8(r4)
+lex;	ld	r7,0(r4)
+lex;	ld	r6,8(r4)
 	addi	r4,r4,16
 	mtctr	r0
 	andi.	r0,r5,0x10
 	beq	22f
 	addi	r3,r3,16
+r3_offset = 0
 	addi	r4,r4,-16
 	mr	r9,r7
 	mr	r8,r6
 	beq	cr1,72f
-21:	ld	r7,16(r4)
-221:	ld	r6,24(r4)
+21:
+lex;	ld	r7,16(r4)
+lex;	ld	r6,24(r4)
 	addi	r4,r4,32
-70:	std	r9,0(r3)
-270:	std	r8,8(r3)
-22:	ld	r9,0(r4)
-222:	ld	r8,8(r4)
-71:	std	r7,16(r3)
-271:	std	r6,24(r3)
+stex;	std	r9,0(r3)
+r3_offset = 8
+stex;	std	r8,8(r3)
+r3_offset = 16
+22:
+lex;	ld	r9,0(r4)
+lex;	ld	r8,8(r4)
+stex;	std	r7,16(r3)
+r3_offset = 24
+stex;	std	r6,24(r3)
 	addi	r3,r3,32
+r3_offset = 0
 	bdnz	21b
-72:	std	r9,0(r3)
-272:	std	r8,8(r3)
+72:
+stex;	std	r9,0(r3)
+r3_offset = 8
+stex;	std	r8,8(r3)
+r3_offset = 16
 	andi.	r5,r5,0xf
 	beq+	3f
 	addi	r4,r4,16
 .Ldo_tail:
 	addi	r3,r3,16
+r3_offset = 0
 	bf	cr7*4+0,246f
-244:	ld	r9,0(r4)
+lex;	ld	r9,0(r4)
 	addi	r4,r4,8
-245:	std	r9,0(r3)
+stex;	std	r9,0(r3)
 	addi	r3,r3,8
 246:	bf	cr7*4+1,1f
-23:	lwz	r9,0(r4)
+lex;	lwz	r9,0(r4)
 	addi	r4,r4,4
-73:	stw	r9,0(r3)
+stex;	stw	r9,0(r3)
 	addi	r3,r3,4
 1:	bf	cr7*4+2,2f
-44:	lhz	r9,0(r4)
+lex;	lhz	r9,0(r4)
 	addi	r4,r4,2
-74:	sth	r9,0(r3)
+stex;	sth	r9,0(r3)
 	addi	r3,r3,2
 2:	bf	cr7*4+3,3f
-45:	lbz	r9,0(r4)
-75:	stb	r9,0(r3)
+lex;	lbz	r9,0(r4)
+stex;	stb	r9,0(r3)
 3:	li	r3,0
 	blr
 
 .Lsrc_unaligned:
+r3_offset = 16
 	srdi	r6,r5,3
 	addi	r5,r5,-16
 	subf	r4,r0,r4
@@ -129,58 +164,69 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 	add	r5,r5,r0
 	bt	cr7*4+0,28f
 
-24:	ld	r9,0(r4)	/* 3+2n loads, 2+2n stores */
-25:	ld	r0,8(r4)
+lex;	ld	r9,0(r4)	/* 3+2n loads, 2+2n stores */
+lex;	ld	r0,8(r4)
 	sLd	r6,r9,r10
-26:	ldu	r9,16(r4)
+lex;	ldu	r9,16(r4)
 	sHd	r7,r0,r11
 	sLd	r8,r0,r10
 	or	r7,r7,r6
 	blt	cr6,79f
-27:	ld	r0,8(r4)
+lex;	ld	r0,8(r4)
 	b	2f
 
-28:	ld	r0,0(r4)	/* 4+2n loads, 3+2n stores */
-29:	ldu	r9,8(r4)
+28:
+lex;	ld	r0,0(r4)	/* 4+2n loads, 3+2n stores */
+lex;	ldu	r9,8(r4)
 	sLd	r8,r0,r10
 	addi	r3,r3,-8
+r3_offset = 24
 	blt	cr6,5f
-30:	ld	r0,8(r4)
+lex;	ld	r0,8(r4)
 	sHd	r12,r9,r11
 	sLd	r6,r9,r10
-31:	ldu	r9,16(r4)
+lex;	ldu	r9,16(r4)
 	or	r12,r8,r12
 	sHd	r7,r0,r11
 	sLd	r8,r0,r10
 	addi	r3,r3,16
+r3_offset = 8
 	beq	cr6,78f
 
 1:	or	r7,r7,r6
-32:	ld	r0,8(r4)
-76:	std	r12,8(r3)
+lex;	ld	r0,8(r4)
+stex;	std	r12,8(r3)
+r3_offset = 16
 2:	sHd	r12,r9,r11
 	sLd	r6,r9,r10
-33:	ldu	r9,16(r4)
+lex;	ldu	r9,16(r4)
 	or	r12,r8,r12
-77:	stdu	r7,16(r3)
+stex;	stdu	r7,16(r3)
+r3_offset = 8
 	sHd	r7,r0,r11
 	sLd	r8,r0,r10
 	bdnz	1b
 
-78:	std	r12,8(r3)
+78:
+stex;	std	r12,8(r3)
+r3_offset = 16
 	or	r7,r7,r6
-79:	std	r7,16(r3)
+79:
+stex;	std	r7,16(r3)
+r3_offset = 24
 5:	sHd	r12,r9,r11
 	or	r12,r8,r12
-80:	std	r12,24(r3)
+stex;	std	r12,24(r3)
+r3_offset = 32
 	bne	6f
 	li	r3,0
 	blr
 6:	cmpwi	cr1,r5,8
 	addi	r3,r3,32
+r3_offset = 0
 	sLd	r9,r9,r10
 	ble	cr1,7f
-34:	ld	r0,8(r4)
+lex;	ld	r0,8(r4)
 	sHd	r7,r0,r11
 	or	r9,r7,r9
 7:
@@ -188,7 +234,7 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 #ifdef __BIG_ENDIAN__
 	rotldi	r9,r9,32
 #endif
-94:	stw	r9,0(r3)
+stex;	stw	r9,0(r3)
 #ifdef __LITTLE_ENDIAN__
 	rotrdi	r9,r9,32
 #endif
@@ -197,7 +243,7 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 #ifdef __BIG_ENDIAN__
 	rotldi	r9,r9,16
 #endif
-95:	sth	r9,0(r3)
+stex;	sth	r9,0(r3)
 #ifdef __LITTLE_ENDIAN__
 	rotrdi	r9,r9,16
 #endif
@@ -206,7 +252,7 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 #ifdef __BIG_ENDIAN__
 	rotldi	r9,r9,8
 #endif
-96:	stb	r9,0(r3)
+stex;	stb	r9,0(r3)
 #ifdef __LITTLE_ENDIAN__
 	rotrdi	r9,r9,8
 #endif
@@ -214,47 +260,55 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 	blr
 
 .Ldst_unaligned:
+r3_offset = 0
 	PPC_MTOCRF(0x01,r6)		/* put #bytes to 8B bdry into cr7 */
 	subf	r5,r6,r5
 	li	r7,0
 	cmpldi	cr1,r5,16
 	bf	cr7*4+3,1f
-35:	lbz	r0,0(r4)
-81:	stb	r0,0(r3)
+100:	EX_TABLE(100b, .Lld_exc_r7)
+	lbz	r0,0(r4)
+100:	EX_TABLE(100b, .Lst_exc_r7)
+	stb	r0,0(r3)
 	addi	r7,r7,1
 1:	bf	cr7*4+2,2f
-36:	lhzx	r0,r7,r4
-82:	sthx	r0,r7,r3
+100:	EX_TABLE(100b, .Lld_exc_r7)
+	lhzx	r0,r7,r4
+100:	EX_TABLE(100b, .Lst_exc_r7)
+	sthx	r0,r7,r3
 	addi	r7,r7,2
 2:	bf	cr7*4+1,3f
-37:	lwzx	r0,r7,r4
-83:	stwx	r0,r7,r3
+100:	EX_TABLE(100b, .Lld_exc_r7)
+	lwzx	r0,r7,r4
+100:	EX_TABLE(100b, .Lst_exc_r7)
+	stwx	r0,r7,r3
 3:	PPC_MTOCRF(0x01,r5)
 	add	r4,r6,r4
 	add	r3,r6,r3
 	b	.Ldst_aligned
 
 .Lshort_copy:
+r3_offset = 0
 	bf	cr7*4+0,1f
-38:	lwz	r0,0(r4)
-39:	lwz	r9,4(r4)
+lex;	lwz	r0,0(r4)
+lex;	lwz	r9,4(r4)
 	addi	r4,r4,8
-84:	stw	r0,0(r3)
-85:	stw	r9,4(r3)
+stex;	stw	r0,0(r3)
+stex;	stw	r9,4(r3)
 	addi	r3,r3,8
 1:	bf	cr7*4+1,2f
-40:	lwz	r0,0(r4)
+lex;	lwz	r0,0(r4)
 	addi	r4,r4,4
-86:	stw	r0,0(r3)
+stex;	stw	r0,0(r3)
 	addi	r3,r3,4
 2:	bf	cr7*4+2,3f
-41:	lhz	r0,0(r4)
+lex;	lhz	r0,0(r4)
 	addi	r4,r4,2
-87:	sth	r0,0(r3)
+stex;	sth	r0,0(r3)
 	addi	r3,r3,2
 3:	bf	cr7*4+3,4f
-42:	lbz	r0,0(r4)
-88:	stb	r0,0(r3)
+lex;	lbz	r0,0(r4)
+stex;	stb	r0,0(r3)
 4:	li	r3,0
 	blr
 
@@ -262,48 +316,34 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
  * exception handlers follow
  * we have to return the number of bytes not copied
  * for an exception on a load, we set the rest of the destination to 0
+ * Note that the number of bytes of instructions for adjusting r3 needs
+ * to equal the amount of the adjustment, due to the trick of using
+ * .Lld_exc - r3_offset as the handler address.
  */
 
-136:
-137:
+.Lld_exc_r7:
 	add	r3,r3,r7
-	b	1f
-130:
-131:
+	b	.Lld_exc
+
+	/* adjust by 24 */
 	addi	r3,r3,8
-120:
-320:
-122:
-322:
-124:
-125:
-126:
-127:
-128:
-129:
-133:
+	nop
+	/* adjust by 16 */
 	addi	r3,r3,8
-132:
+	nop
+	/* adjust by 8 */
 	addi	r3,r3,8
-121:
-321:
-344:
-134:
-135:
-138:
-139:
-140:
-141:
-142:
-123:
-144:
-145:
+	nop
 
 /*
- * here we have had a fault on a load and r3 points to the first
- * unmodified byte of the destination
+ * Here we have had a fault on a load and r3 points to the first
+ * unmodified byte of the destination.  We use the original arguments
+ * and r3 to work out how much wasn't copied.  Since we load some
+ * distance ahead of the stores, we continue copying byte-by-byte until
+ * we hit the load fault again in order to copy as much as possible.
  */
-1:	ld	r6,-24(r1)
+.Lld_exc:
+	ld	r6,-24(r1)
 	ld	r4,-16(r1)
 	ld	r5,-8(r1)
 	subf	r6,r6,r3
@@ -314,9 +354,11 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
  * first see if we can copy any more bytes before hitting another exception
  */
 	mtctr	r5
+r3_offset = 0
+100:	EX_TABLE(100b, .Ldone)
 43:	lbz	r0,0(r4)
 	addi	r4,r4,1
-89:	stb	r0,0(r3)
+stex;	stb	r0,0(r3)
 	addi	r3,r3,1
 	bdnz	43b
 	li	r3,0		/* huh? all copied successfully this time? */
@@ -325,116 +367,46 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 /*
  * here we have trapped again, amount remaining is in ctr.
  */
-143:	mfctr	r3
+.Ldone:
+	mfctr	r3
 	blr
 
 /*
  * exception handlers for stores: we just need to work
  * out how many bytes weren't copied
+ * Note that the number of bytes of instructions for adjusting r3 needs
+ * to equal the amount of the adjustment, due to the trick of using
+ * .Lst_exc - r3_offset as the handler address.
  */
-182:
-183:
+.Lst_exc_r7:
 	add	r3,r3,r7
-	b	1f
-371:
-180:
+	b	.Lst_exc
+
+	/* adjust by 24 */
 	addi	r3,r3,8
-171:
-177:
-179:
+	nop
+	/* adjust by 16 */
 	addi	r3,r3,8
-370:
-372:
-176:
-178:
+	nop
+	/* adjust by 8 */
 	addi	r3,r3,4
-185:
+	/* adjust by 4 */
 	addi	r3,r3,4
-170:
-172:
-345:
-173:
-174:
-175:
-181:
-184:
-186:
-187:
-188:
-189:	
-194:
-195:
-196:
-1:
+.Lst_exc:
 	ld	r6,-24(r1)
 	ld	r5,-8(r1)
 	add	r6,r6,r5
-	subf	r3,r3,r6	/* #bytes not copied */
+	subf	r3,r3,r6	/* #bytes not copied in r3 */
 	blr
 
-	EX_TABLE(20b,120b)
-	EX_TABLE(220b,320b)
-	EX_TABLE(21b,121b)
-	EX_TABLE(221b,321b)
-	EX_TABLE(70b,170b)
-	EX_TABLE(270b,370b)
-	EX_TABLE(22b,122b)
-	EX_TABLE(222b,322b)
-	EX_TABLE(71b,171b)
-	EX_TABLE(271b,371b)
-	EX_TABLE(72b,172b)
-	EX_TABLE(272b,372b)
-	EX_TABLE(244b,344b)
-	EX_TABLE(245b,345b)
-	EX_TABLE(23b,123b)
-	EX_TABLE(73b,173b)
-	EX_TABLE(44b,144b)
-	EX_TABLE(74b,174b)
-	EX_TABLE(45b,145b)
-	EX_TABLE(75b,175b)
-	EX_TABLE(24b,124b)
-	EX_TABLE(25b,125b)
-	EX_TABLE(26b,126b)
-	EX_TABLE(27b,127b)
-	EX_TABLE(28b,128b)
-	EX_TABLE(29b,129b)
-	EX_TABLE(30b,130b)
-	EX_TABLE(31b,131b)
-	EX_TABLE(32b,132b)
-	EX_TABLE(76b,176b)
-	EX_TABLE(33b,133b)
-	EX_TABLE(77b,177b)
-	EX_TABLE(78b,178b)
-	EX_TABLE(79b,179b)
-	EX_TABLE(80b,180b)
-	EX_TABLE(34b,134b)
-	EX_TABLE(94b,194b)
-	EX_TABLE(95b,195b)
-	EX_TABLE(96b,196b)
-	EX_TABLE(35b,135b)
-	EX_TABLE(81b,181b)
-	EX_TABLE(36b,136b)
-	EX_TABLE(82b,182b)
-	EX_TABLE(37b,137b)
-	EX_TABLE(83b,183b)
-	EX_TABLE(38b,138b)
-	EX_TABLE(39b,139b)
-	EX_TABLE(84b,184b)
-	EX_TABLE(85b,185b)
-	EX_TABLE(40b,140b)
-	EX_TABLE(86b,186b)
-	EX_TABLE(41b,141b)
-	EX_TABLE(87b,187b)
-	EX_TABLE(42b,142b)
-	EX_TABLE(88b,188b)
-	EX_TABLE(43b,143b)
-	EX_TABLE(89b,189b)
-
 /*
  * Routine to copy a whole page of data, optimized for POWER4.
  * On POWER4 it is more than 50% faster than the simple loop
  * above (following the .Ldst_aligned label).
  */
+	.macro	exc
+100:	EX_TABLE(100b, .Labort)
+	.endm
 .Lcopy_page_4K:
 	std	r31,-32(1)
 	std	r30,-40(1)
@@ -453,86 +425,86 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 	li	r0,5
 0:	addi	r5,r5,-24
 	mtctr	r0
-20:	ld	r22,640(4)
-21:	ld	r21,512(4)
-22:	ld	r20,384(4)
-23:	ld	r11,256(4)
-24:	ld	r9,128(4)
-25:	ld	r7,0(4)
-26:	ld	r25,648(4)
-27:	ld	r24,520(4)
-28:	ld	r23,392(4)
-29:	ld	r10,264(4)
-30:	ld	r8,136(4)
-31:	ldu	r6,8(4)
+exc;	ld	r22,640(4)
+exc;	ld	r21,512(4)
+exc;	ld	r20,384(4)
+exc;	ld	r11,256(4)
+exc;	ld	r9,128(4)
+exc;	ld	r7,0(4)
+exc;	ld	r25,648(4)
+exc;	ld	r24,520(4)
+exc;	ld	r23,392(4)
+exc;	ld	r10,264(4)
+exc;	ld	r8,136(4)
+exc;	ldu	r6,8(4)
 	cmpwi	r5,24
 1:
-32:	std	r22,648(3)
-33:	std	r21,520(3)
-34:	std	r20,392(3)
-35:	std	r11,264(3)
-36:	std	r9,136(3)
-37:	std	r7,8(3)
-38:	ld	r28,648(4)
-39:	ld	r27,520(4)
-40:	ld	r26,392(4)
-41:	ld	r31,264(4)
-42:	ld	r30,136(4)
-43:	ld	r29,8(4)
-44:	std	r25,656(3)
-45:	std	r24,528(3)
-46:	std	r23,400(3)
-47:	std	r10,272(3)
-48:	std	r8,144(3)
-49:	std	r6,16(3)
-50:	ld	r22,656(4)
-51:	ld	r21,528(4)
-52:	ld	r20,400(4)
-53:	ld	r11,272(4)
-54:	ld	r9,144(4)
-55:	ld	r7,16(4)
-56:	std	r28,664(3)
-57:	std	r27,536(3)
-58:	std	r26,408(3)
-59:	std	r31,280(3)
-60:	std	r30,152(3)
-61:	stdu	r29,24(3)
-62:	ld	r25,664(4)
-63:	ld	r24,536(4)
-64:	ld	r23,408(4)
-65:	ld	r10,280(4)
-66:	ld	r8,152(4)
-67:	ldu	r6,24(4)
+exc;	std	r22,648(3)
+exc;	std	r21,520(3)
+exc;	std	r20,392(3)
+exc;	std	r11,264(3)
+exc;	std	r9,136(3)
+exc;	std	r7,8(3)
+exc;	ld	r28,648(4)
+exc;	ld	r27,520(4)
+exc;	ld	r26,392(4)
+exc;	ld	r31,264(4)
+exc;	ld	r30,136(4)
+exc;	ld	r29,8(4)
+exc;	std	r25,656(3)
+exc;	std	r24,528(3)
+exc;	std	r23,400(3)
+exc;	std	r10,272(3)
+exc;	std	r8,144(3)
+exc;	std	r6,16(3)
+exc;	ld	r22,656(4)
+exc;	ld	r21,528(4)
+exc;	ld	r20,400(4)
+exc;	ld	r11,272(4)
+exc;	ld	r9,144(4)
+exc;	ld	r7,16(4)
+exc;	std	r28,664(3)
+exc;	std	r27,536(3)
+exc;	std	r26,408(3)
+exc;	std	r31,280(3)
+exc;	std	r30,152(3)
+exc;	stdu	r29,24(3)
+exc;	ld	r25,664(4)
+exc;	ld	r24,536(4)
+exc;	ld	r23,408(4)
+exc;	ld	r10,280(4)
+exc;	ld	r8,152(4)
+exc;	ldu	r6,24(4)
 	bdnz	1b
-68:	std	r22,648(3)
-69:	std	r21,520(3)
-70:	std	r20,392(3)
-71:	std	r11,264(3)
-72:	std	r9,136(3)
-73:	std	r7,8(3)
-74:	addi	r4,r4,640
-75:	addi	r3,r3,648
+exc;	std	r22,648(3)
+exc;	std	r21,520(3)
+exc;	std	r20,392(3)
+exc;	std	r11,264(3)
+exc;	std	r9,136(3)
+exc;	std	r7,8(3)
+	addi	r4,r4,640
+	addi	r3,r3,648
 	bge	0b
 	mtctr	r5
-76:	ld	r7,0(4)
-77:	ld	r8,8(4)
-78:	ldu	r9,16(4)
+exc;	ld	r7,0(4)
+exc;	ld	r8,8(4)
+exc;	ldu	r9,16(4)
 3:
-79:	ld	r10,8(4)
-80:	std	r7,8(3)
-81:	ld	r7,16(4)
-82:	std	r8,16(3)
-83:	ld	r8,24(4)
-84:	std	r9,24(3)
-85:	ldu	r9,32(4)
-86:	stdu	r10,32(3)
+exc;	ld	r10,8(4)
+exc;	std	r7,8(3)
+exc;	ld	r7,16(4)
+exc;	std	r8,16(3)
+exc;	ld	r8,24(4)
+exc;	std	r9,24(3)
+exc;	ldu	r9,32(4)
+exc;	stdu	r10,32(3)
 	bdnz	3b
 4:
-87:	ld	r10,8(4)
-88:	std	r7,8(3)
-89:	std	r8,16(3)
-90:	std	r9,24(3)
-91:	std	r10,32(3)
+exc;	ld	r10,8(4)
+exc;	std	r7,8(3)
+exc;	std	r8,16(3)
+exc;	std	r9,24(3)
+exc;	std	r10,32(3)
 9:	ld	r20,-120(1)
 	ld	r21,-112(1)
 	ld	r22,-104(1)
@@ -552,7 +524,8 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
  * on an exception, reset to the beginning and jump back into the
  * standard __copy_tofrom_user
  */
-100:	ld	r20,-120(1)
+.Labort:
+	ld	r20,-120(1)
 	ld	r21,-112(1)
 	ld	r22,-104(1)
 	ld	r23,-96(1)
@@ -568,78 +541,4 @@ END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
 	ld	r4,-16(r1)
 	li	r5,4096
 	b	.Ldst_aligned
-
-	EX_TABLE(20b,100b)
-	EX_TABLE(21b,100b)
-	EX_TABLE(22b,100b)
-	EX_TABLE(23b,100b)
-	EX_TABLE(24b,100b)
-	EX_TABLE(25b,100b)
-	EX_TABLE(26b,100b)
-	EX_TABLE(27b,100b)
-	EX_TABLE(28b,100b)
-	EX_TABLE(29b,100b)
-	EX_TABLE(30b,100b)
-	EX_TABLE(31b,100b)
-	EX_TABLE(32b,100b)
-	EX_TABLE(33b,100b)
-	EX_TABLE(34b,100b)
-	EX_TABLE(35b,100b)
-	EX_TABLE(36b,100b)
-	EX_TABLE(37b,100b)
-	EX_TABLE(38b,100b)
-	EX_TABLE(39b,100b)
-	EX_TABLE(40b,100b)
-	EX_TABLE(41b,100b)
-	EX_TABLE(42b,100b)
-	EX_TABLE(43b,100b)
-	EX_TABLE(44b,100b)
-	EX_TABLE(45b,100b)
-	EX_TABLE(46b,100b)
-	EX_TABLE(47b,100b)
-	EX_TABLE(48b,100b)
-	EX_TABLE(49b,100b)
-	EX_TABLE(50b,100b)
-	EX_TABLE(51b,100b)
-	EX_TABLE(52b,100b)
-	EX_TABLE(53b,100b)
-	EX_TABLE(54b,100b)
-	EX_TABLE(55b,100b)
-	EX_TABLE(56b,100b)
-	EX_TABLE(57b,100b)
-	EX_TABLE(58b,100b)
-	EX_TABLE(59b,100b)
-	EX_TABLE(60b,100b)
-	EX_TABLE(61b,100b)
-	EX_TABLE(62b,100b)
-	EX_TABLE(63b,100b)
-	EX_TABLE(64b,100b)
-	EX_TABLE(65b,100b)
-	EX_TABLE(66b,100b)
-	EX_TABLE(67b,100b)
-	EX_TABLE(68b,100b)
-	EX_TABLE(69b,100b)
-	EX_TABLE(70b,100b)
-	EX_TABLE(71b,100b)
-	EX_TABLE(72b,100b)
-	EX_TABLE(73b,100b)
-	EX_TABLE(74b,100b)
-	EX_TABLE(75b,100b)
-	EX_TABLE(76b,100b)
-	EX_TABLE(77b,100b)
-	EX_TABLE(78b,100b)
-	EX_TABLE(79b,100b)
-	EX_TABLE(80b,100b)
-	EX_TABLE(81b,100b)
-	EX_TABLE(82b,100b)
-	EX_TABLE(83b,100b)
-	EX_TABLE(84b,100b)
-	EX_TABLE(85b,100b)
-	EX_TABLE(86b,100b)
-	EX_TABLE(87b,100b)
-	EX_TABLE(88b,100b)
-	EX_TABLE(89b,100b)
-	EX_TABLE(90b,100b)
-	EX_TABLE(91b,100b)
-
 EXPORT_SYMBOL(__copy_tofrom_user)
-- 
2.7.4

Re: [v3, 1/4] powerpc/64: Make exception table clearer in __copy_tofrom_user_base

[Michael Ellerman]

On Fri, 2018-08-03 at 10:13:03 UTC, Paul Mackerras wrote:
> This aims to make the generation of exception table entries for the
> loads and stores in __copy_tofrom_user_base clearer and easier to
> verify.  Instead of having a series of local labels on the loads and
> stores, with a series of corresponding labels later for the exception
> handlers, we now use macros to generate exception table entries at the
> point of each load and store that could potentially trap.  We do this
> with the macros lex (load exception) and stex (store exception).
> These macros are used right before the load or store to which they
> apply.
> 
> Some complexity is introduced by the fact that we have some more work
> to do after hitting an exception, because we need to calculate and
> return the number of bytes not copied.  The code uses r3 as the
> current pointer into the destination buffer, that is, the address of
> the first byte of the destination that has not been modified.
> However, at various points in the copy loops, r3 can be 4, 8, 16 or 24
> bytes behind that point.
> 
> To express this offset in an understandable way, we define a symbol
> r3_offset which is updated at various points so that it equal to the
> difference between the address of the first unmodified byte of the
> destination and the value in r3.  (In fact it only needs to be
> accurate at the point of each lex or stex macro invocation.)
> 
> The rules for updating r3_offset are as follows:
> 
> * It starts out at 0
> * An addi r3,r3,N instruction decreases r3_offset by N
> * A store instruction (stb, sth, stw, std) to N(r3)
>   increases r3_offset by the width of the store (1, 2, 4, 8)
> * A store with update instruction (stbu, sthu, stwu, stdu) to N(r3)
>   sets r3_offset to the width of the store.
> 
> There is some trickiness to the way that the lex and stex macros and
> the associated exception handlers work.  I would have liked to use
> the current value of r3_offset in the name of the symbol used as
> the exception handler, as in ".Lld_exc_$(r3_offset)" and then
> have symbols .Lld_exc_0, .Lld_exc_8, .Lld_exc_16 etc. corresponding
> to the offsets that needed to be added to r3.  However, I couldn't
> see a way to do that with gas.
> 
> Instead, the exception handler address is .Lld_exc - r3_offset or
> .Lst_exc - r3_offset, that is, the distance ahead of .Lld_exc/.Lst_exc
> that we start executing is equal to the amount that we need to add to
> r3.  This works because r3_offset is always a small multiple of 4,
> and our instructions are 4 bytes long.  This means that before
> .Lld_exc and .Lst_exc, we have a sequence of instructions that
> increments r3 by 4, 8, 16 or 24 depending on where we start.  The
> sequence increments r3 by 4 per instruction (on average).
> 
> We also replace the exception table for the 4k copy loop by a
> macro per load or store.  These loads and stores all use exactly
> the same exception handler, which simply resets the argument registers
> r3, r4 and r5 to there original values and re-does the whole copy
> using the slower loop.
> 
> Signed-off-by: Paul Mackerras <paulus@ozlabs.org>

Series applied to powerpc next, thanks.

https://git.kernel.org/powerpc/c/a7c81ce398e2ad304f61d6167155f3

cheers

[PATCH v3 2/4] selftests/powerpc/64: Test all paths through copy routines

[Paul Mackerras]

The hand-coded assembler 64-bit copy routines include feature sections
that select one code path or another depending on which CPU we are
executing on.  The self-tests for these copy routines end up testing
just one path.  This adds a mechanism for selecting any desired code
path at compile time, and makes 2 or 3 versions of each test, each
using a different code path, so as to cover all the possible paths.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
---
 arch/powerpc/lib/copyuser_64.S                     |  7 +++++
 arch/powerpc/lib/copyuser_power7.S                 | 21 ++++++-------
 arch/powerpc/lib/memcpy_64.S                       |  9 ++++--
 arch/powerpc/lib/memcpy_power7.S                   | 22 +++++++-------
 .../testing/selftests/powerpc/copyloops/.gitignore | 14 ++++++---
 tools/testing/selftests/powerpc/copyloops/Makefile | 34 ++++++++++++++++++----
 .../selftests/powerpc/copyloops/asm/ppc_asm.h      | 21 +++++++------
 7 files changed, 84 insertions(+), 44 deletions(-)

diff --git a/arch/powerpc/lib/copyuser_64.S b/arch/powerpc/lib/copyuser_64.S
index 8c5033c..2197a35 100644
--- a/arch/powerpc/lib/copyuser_64.S
+++ b/arch/powerpc/lib/copyuser_64.S
@@ -12,6 +12,11 @@
 #include <asm/asm-compat.h>
 #include <asm/feature-fixups.h>
 
+#ifndef SELFTEST_CASE
+/* 0 == most CPUs, 1 == POWER6, 2 == Cell */
+#define SELFTEST_CASE	0
+#endif
+
 #ifdef __BIG_ENDIAN__
 #define sLd sld		/* Shift towards low-numbered address. */
 #define sHd srd		/* Shift towards high-numbered address. */
@@ -73,6 +78,7 @@ _GLOBAL(__copy_tofrom_user_base)
  * At the time of writing the only CPU that has this combination of bits
  * set is Power6.
  */
+test_feature = (SELFTEST_CASE == 1)
 BEGIN_FTR_SECTION
 	nop
 FTR_SECTION_ELSE
@@ -82,6 +88,7 @@ ALT_FTR_SECTION_END(CPU_FTR_UNALIGNED_LD_STD | CPU_FTR_CP_USE_DCBTZ, \
 .Ldst_aligned:
 	addi	r3,r3,-16
 r3_offset = 16
+test_feature = (SELFTEST_CASE == 0)
 BEGIN_FTR_SECTION
 	andi.	r0,r4,7
 	bne	.Lsrc_unaligned
diff --git a/arch/powerpc/lib/copyuser_power7.S b/arch/powerpc/lib/copyuser_power7.S
index 215e476..1a1fe18 100644
--- a/arch/powerpc/lib/copyuser_power7.S
+++ b/arch/powerpc/lib/copyuser_power7.S
@@ -19,6 +19,11 @@
  */
 #include <asm/ppc_asm.h>
 
+#ifndef SELFTEST_CASE
+/* 0 == don't use VMX, 1 == use VMX */
+#define SELFTEST_CASE	0
+#endif
+
 #ifdef __BIG_ENDIAN__
 #define LVS(VRT,RA,RB)		lvsl	VRT,RA,RB
 #define VPERM(VRT,VRA,VRB,VRC)	vperm	VRT,VRA,VRB,VRC
@@ -80,7 +85,6 @@
 
 
 _GLOBAL(__copy_tofrom_user_power7)
-#ifdef CONFIG_ALTIVEC
 	cmpldi	r5,16
 	cmpldi	cr1,r5,3328
 
@@ -89,15 +93,12 @@ _GLOBAL(__copy_tofrom_user_power7)
 	std	r5,-STACKFRAMESIZE+STK_REG(R29)(r1)
 
 	blt	.Lshort_copy
-	bge	cr1,.Lvmx_copy
-#else
-	cmpldi	r5,16
 
-	std	r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
-	std	r4,-STACKFRAMESIZE+STK_REG(R30)(r1)
-	std	r5,-STACKFRAMESIZE+STK_REG(R29)(r1)
-
-	blt	.Lshort_copy
+#ifdef CONFIG_ALTIVEC
+test_feature = SELFTEST_CASE
+BEGIN_FTR_SECTION
+	bgt	cr1,.Lvmx_copy
+END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
 #endif
 
 .Lnonvmx_copy:
@@ -278,8 +279,8 @@ err1;	stb	r0,0(r3)
 	addi	r1,r1,STACKFRAMESIZE
 	b	.Lnonvmx_copy
 
-#ifdef CONFIG_ALTIVEC
 .Lvmx_copy:
+#ifdef CONFIG_ALTIVEC
 	mflr	r0
 	std	r0,16(r1)
 	stdu	r1,-STACKFRAMESIZE(r1)
diff --git a/arch/powerpc/lib/memcpy_64.S b/arch/powerpc/lib/memcpy_64.S
index 94650d6..273ea67 100644
--- a/arch/powerpc/lib/memcpy_64.S
+++ b/arch/powerpc/lib/memcpy_64.S
@@ -12,6 +12,11 @@
 #include <asm/asm-compat.h>
 #include <asm/feature-fixups.h>
 
+#ifndef SELFTEST_CASE
+/* For big-endian, 0 == most CPUs, 1 == POWER6, 2 == Cell */
+#define SELFTEST_CASE	0
+#endif
+
 	.align	7
 _GLOBAL_TOC(memcpy)
 BEGIN_FTR_SECTION
@@ -22,10 +27,8 @@ BEGIN_FTR_SECTION
 #endif
 FTR_SECTION_ELSE
 #ifdef CONFIG_PPC_BOOK3S_64
-#ifndef SELFTEST
 	b	memcpy_power7
 #endif
-#endif
 ALT_FTR_SECTION_END_IFCLR(CPU_FTR_VMX_COPY)
 #ifdef __LITTLE_ENDIAN__
 	/* dumb little-endian memcpy that will get replaced at runtime */
@@ -49,6 +52,7 @@ ALT_FTR_SECTION_END_IFCLR(CPU_FTR_VMX_COPY)
    cleared.
    At the time of writing the only CPU that has this combination of bits
    set is Power6. */
+test_feature = (SELFTEST_CASE == 1)
 BEGIN_FTR_SECTION
 	nop
 FTR_SECTION_ELSE
@@ -57,6 +61,7 @@ ALT_FTR_SECTION_END(CPU_FTR_UNALIGNED_LD_STD | CPU_FTR_CP_USE_DCBTZ, \
                     CPU_FTR_UNALIGNED_LD_STD)
 .Ldst_aligned:
 	addi	r3,r3,-16
+test_feature = (SELFTEST_CASE == 0)
 BEGIN_FTR_SECTION
 	andi.	r0,r4,7
 	bne	.Lsrc_unaligned
diff --git a/arch/powerpc/lib/memcpy_power7.S b/arch/powerpc/lib/memcpy_power7.S
index 070cdf6..89bfefc 100644
--- a/arch/powerpc/lib/memcpy_power7.S
+++ b/arch/powerpc/lib/memcpy_power7.S
@@ -19,7 +19,10 @@
  */
 #include <asm/ppc_asm.h>
 
-_GLOBAL(memcpy_power7)
+#ifndef SELFTEST_CASE
+/* 0 == don't use VMX, 1 == use VMX */
+#define SELFTEST_CASE	0
+#endif
 
 #ifdef __BIG_ENDIAN__
 #define LVS(VRT,RA,RB)		lvsl	VRT,RA,RB
@@ -29,20 +32,17 @@ _GLOBAL(memcpy_power7)
 #define VPERM(VRT,VRA,VRB,VRC)	vperm	VRT,VRB,VRA,VRC
 #endif
 
-#ifdef CONFIG_ALTIVEC
+_GLOBAL(memcpy_power7)
 	cmpldi	r5,16
 	cmpldi	cr1,r5,4096
-
 	std	r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
-
 	blt	.Lshort_copy
-	bgt	cr1,.Lvmx_copy
-#else
-	cmpldi	r5,16
-
-	std	r3,-STACKFRAMESIZE+STK_REG(R31)(r1)
 
-	blt	.Lshort_copy
+#ifdef CONFIG_ALTIVEC
+test_feature = SELFTEST_CASE
+BEGIN_FTR_SECTION
+	bgt	cr1, .Lvmx_copy
+END_FTR_SECTION_IFSET(CPU_FTR_ALTIVEC)
 #endif
 
 .Lnonvmx_copy:
@@ -223,8 +223,8 @@ _GLOBAL(memcpy_power7)
 	addi	r1,r1,STACKFRAMESIZE
 	b	.Lnonvmx_copy
 
-#ifdef CONFIG_ALTIVEC
 .Lvmx_copy:
+#ifdef CONFIG_ALTIVEC
 	mflr	r0
 	std	r4,-STACKFRAMESIZE+STK_REG(R30)(r1)
 	std	r5,-STACKFRAMESIZE+STK_REG(R29)(r1)
diff --git a/tools/testing/selftests/powerpc/copyloops/.gitignore b/tools/testing/selftests/powerpc/copyloops/.gitignore
index 25a192f..bc6c4ba 100644
--- a/tools/testing/selftests/powerpc/copyloops/.gitignore
+++ b/tools/testing/selftests/powerpc/copyloops/.gitignore
@@ -1,4 +1,10 @@
-copyuser_64
-copyuser_power7
-memcpy_64
-memcpy_power7
+copyuser_64_t0
+copyuser_64_t1
+copyuser_64_t2
+copyuser_power7_t0
+copyuser_power7_t1
+memcpy_64_t0
+memcpy_64_t1
+memcpy_64_t2
+memcpy_power7_t0
+memcpy_power7_t1
diff --git a/tools/testing/selftests/powerpc/copyloops/Makefile b/tools/testing/selftests/powerpc/copyloops/Makefile
index eedce33..9a5b0d2 100644
--- a/tools/testing/selftests/powerpc/copyloops/Makefile
+++ b/tools/testing/selftests/powerpc/copyloops/Makefile
@@ -8,14 +8,36 @@ CFLAGS += -maltivec
 # Use our CFLAGS for the implicit .S rule & set the asm machine type
 ASFLAGS = $(CFLAGS) -Wa,-mpower4
 
-TEST_GEN_PROGS := copyuser_64 copyuser_power7 memcpy_64 memcpy_power7
+TEST_GEN_PROGS := copyuser_64_t0 copyuser_64_t1 copyuser_64_t2 \
+		copyuser_p7_t0 copyuser_p7_t1 \
+		memcpy_64_t0 memcpy_64_t1 memcpy_64_t2 \
+		memcpy_p7_t0 memcpy_p7_t1
+
 EXTRA_SOURCES := validate.c ../harness.c
 
 include ../../lib.mk
 
-$(OUTPUT)/copyuser_64:     CPPFLAGS += -D COPY_LOOP=test___copy_tofrom_user_base
-$(OUTPUT)/copyuser_power7: CPPFLAGS += -D COPY_LOOP=test___copy_tofrom_user_power7
-$(OUTPUT)/memcpy_64:       CPPFLAGS += -D COPY_LOOP=test_memcpy
-$(OUTPUT)/memcpy_power7:   CPPFLAGS += -D COPY_LOOP=test_memcpy_power7
+$(OUTPUT)/copyuser_64_t%:	copyuser_64.S $(EXTRA_SOURCES)
+	$(CC) $(CPPFLAGS) $(CFLAGS) \
+		-D COPY_LOOP=test___copy_tofrom_user_base \
+		-D SELFTEST_CASE=$(subst copyuser_64_t,,$(notdir $@)) \
+		-o $@ $^
+
+$(OUTPUT)/copyuser_p7_t%:	copyuser_power7.S $(EXTRA_SOURCES)
+	$(CC) $(CPPFLAGS) $(CFLAGS) \
+		-D COPY_LOOP=test___copy_tofrom_user_power7 \
+		-D SELFTEST_CASE=$(subst copyuser_p7_t,,$(notdir $@)) \
+		-o $@ $^
+
+# Strictly speaking, we only need the memcpy_64 test cases for big-endian
+$(OUTPUT)/memcpy_64_t%:	memcpy_64.S $(EXTRA_SOURCES)
+	$(CC) $(CPPFLAGS) $(CFLAGS) \
+		-D COPY_LOOP=test_memcpy \
+		-D SELFTEST_CASE=$(subst memcpy_64_t,,$(notdir $@)) \
+		-o $@ $^
 
-$(TEST_GEN_PROGS): $(EXTRA_SOURCES)
+$(OUTPUT)/memcpy_p7_t%:	memcpy_power7.S $(EXTRA_SOURCES)
+	$(CC) $(CPPFLAGS) $(CFLAGS) \
+		-D COPY_LOOP=test_memcpy_power7 \
+		-D SELFTEST_CASE=$(subst memcpy_p7_t,,$(notdir $@)) \
+		-o $@ $^
diff --git a/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h b/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
index dfce161..91bc403 100644
--- a/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
+++ b/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
@@ -43,17 +43,16 @@ FUNC_START(enter_vmx_ops)
 FUNC_START(exit_vmx_ops)
 	blr
 
-FUNC_START(memcpy_power7)
-	blr
-
-FUNC_START(__copy_tofrom_user_power7)
-	blr
-
 FUNC_START(__copy_tofrom_user_base)
 	blr
 
-#define BEGIN_FTR_SECTION
-#define FTR_SECTION_ELSE
-#define ALT_FTR_SECTION_END_IFCLR(x)
-#define ALT_FTR_SECTION_END(x, y)
-#define END_FTR_SECTION_IFCLR(x)
+#define BEGIN_FTR_SECTION		.if test_feature
+#define FTR_SECTION_ELSE		.else
+#define ALT_FTR_SECTION_END_IFCLR(x)	.endif
+#define ALT_FTR_SECTION_END_IFSET(x)	.endif
+#define ALT_FTR_SECTION_END(x, y)	.endif
+#define END_FTR_SECTION_IFCLR(x)	.endif
+#define END_FTR_SECTION_IFSET(x)	.endif
+
+/* Default to taking the first of any alternative feature sections */
+test_feature = 1
-- 
2.7.4

[PATCH v3 3/4] selftests/powerpc/64: Test exception cases in copy_tofrom_user

[Paul Mackerras]

From: Michael Ellerman <mpe@ellerman.id.au>

This adds a set of test cases to test the behaviour of
copy_tofrom_user when exceptions are encountered accessing the
source or destination.  Currently, copy_tofrom_user does not always
copy as many bytes as possible when an exception occurs on a store
to the destination, and that is reflected in failures in these tests.

Based on a test program from Anton Blanchard.

[paulus@ozlabs.org - test all three paths, wrote commit description,
 made EX_TABLE create an exception table.]

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
---
 .../testing/selftests/powerpc/copyloops/.gitignore |   3 +
 tools/testing/selftests/powerpc/copyloops/Makefile |  12 +-
 .../selftests/powerpc/copyloops/asm/ppc_asm.h      |  27 ++---
 .../powerpc/copyloops/copy_tofrom_user_reference.S |  24 ++++
 .../selftests/powerpc/copyloops/exc_validate.c     | 124 +++++++++++++++++++++
 tools/testing/selftests/powerpc/copyloops/stubs.S  |  19 ++++
 6 files changed, 188 insertions(+), 21 deletions(-)
 create mode 100644 tools/testing/selftests/powerpc/copyloops/copy_tofrom_user_reference.S
 create mode 100644 tools/testing/selftests/powerpc/copyloops/exc_validate.c
 create mode 100644 tools/testing/selftests/powerpc/copyloops/stubs.S

diff --git a/tools/testing/selftests/powerpc/copyloops/.gitignore b/tools/testing/selftests/powerpc/copyloops/.gitignore
index bc6c4ba..ce12cd0 100644
--- a/tools/testing/selftests/powerpc/copyloops/.gitignore
+++ b/tools/testing/selftests/powerpc/copyloops/.gitignore
@@ -8,3 +8,6 @@ memcpy_64_t1
 memcpy_64_t2
 memcpy_power7_t0
 memcpy_power7_t1
+copyuser_64_exc_t0
+copyuser_64_exc_t1
+copyuser_64_exc_t2
diff --git a/tools/testing/selftests/powerpc/copyloops/Makefile b/tools/testing/selftests/powerpc/copyloops/Makefile
index 9a5b0d2..9da0010 100644
--- a/tools/testing/selftests/powerpc/copyloops/Makefile
+++ b/tools/testing/selftests/powerpc/copyloops/Makefile
@@ -11,9 +11,10 @@ ASFLAGS = $(CFLAGS) -Wa,-mpower4
 TEST_GEN_PROGS := copyuser_64_t0 copyuser_64_t1 copyuser_64_t2 \
 		copyuser_p7_t0 copyuser_p7_t1 \
 		memcpy_64_t0 memcpy_64_t1 memcpy_64_t2 \
-		memcpy_p7_t0 memcpy_p7_t1
+		memcpy_p7_t0 memcpy_p7_t1 \
+		copyuser_64_exc_t0 copyuser_64_exc_t1 copyuser_64_exc_t2
 
-EXTRA_SOURCES := validate.c ../harness.c
+EXTRA_SOURCES := validate.c ../harness.c stubs.S
 
 include ../../lib.mk
 
@@ -41,3 +42,10 @@ $(OUTPUT)/memcpy_p7_t%:	memcpy_power7.S $(EXTRA_SOURCES)
 		-D COPY_LOOP=test_memcpy_power7 \
 		-D SELFTEST_CASE=$(subst memcpy_p7_t,,$(notdir $@)) \
 		-o $@ $^
+
+$(OUTPUT)/copyuser_64_exc_t%: copyuser_64.S exc_validate.c ../harness.c \
+		copy_tofrom_user_reference.S stubs.S
+	$(CC) $(CPPFLAGS) $(CFLAGS) \
+		-D COPY_LOOP=test___copy_tofrom_user_base \
+		-D SELFTEST_CASE=$(subst copyuser_64_exc_t,,$(notdir $@)) \
+		-o $@ $^
diff --git a/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h b/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
index 91bc403..0605df8 100644
--- a/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
+++ b/tools/testing/selftests/powerpc/copyloops/asm/ppc_asm.h
@@ -1,4 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __SELFTESTS_POWERPC_PPC_ASM_H
+#define __SELFTESTS_POWERPC_PPC_ASM_H
 #include <ppc-asm.h>
 
 #define CONFIG_ALTIVEC
@@ -26,25 +28,10 @@
 
 #define PPC_MTOCRF(A, B)	mtocrf A, B
 
-#define EX_TABLE(x, y)
-
-FUNC_START(enter_vmx_usercopy)
-	li	r3,1
-	blr
-
-FUNC_START(exit_vmx_usercopy)
-	li	r3,0
-	blr
-
-FUNC_START(enter_vmx_ops)
-	li	r3,1
-	blr
-
-FUNC_START(exit_vmx_ops)
-	blr
-
-FUNC_START(__copy_tofrom_user_base)
-	blr
+#define EX_TABLE(x, y)			\
+	.section __ex_table,"a";	\
+	.8byte	x, y;			\
+	.previous
 
 #define BEGIN_FTR_SECTION		.if test_feature
 #define FTR_SECTION_ELSE		.else
@@ -56,3 +43,5 @@ FUNC_START(__copy_tofrom_user_base)
 
 /* Default to taking the first of any alternative feature sections */
 test_feature = 1
+
+#endif /* __SELFTESTS_POWERPC_PPC_ASM_H */
diff --git a/tools/testing/selftests/powerpc/copyloops/copy_tofrom_user_reference.S b/tools/testing/selftests/powerpc/copyloops/copy_tofrom_user_reference.S
new file mode 100644
index 0000000..3363b86
--- /dev/null
+++ b/tools/testing/selftests/powerpc/copyloops/copy_tofrom_user_reference.S
@@ -0,0 +1,24 @@
+#include <asm/ppc_asm.h>
+
+_GLOBAL(copy_tofrom_user_reference)
+	cmpdi	r5,0
+	beq	4f
+
+	mtctr	r5
+
+1:	lbz	r6,0(r4)
+2:	stb	r6,0(r3)
+	addi	r3,r3,1
+	addi	r4,r4,1
+	bdnz	1b
+
+3:	mfctr	r3
+	blr
+
+4:	mr	r3,r5
+	blr
+
+.section __ex_table,"a"
+	.llong	1b,3b
+	.llong	2b,3b
+.text
diff --git a/tools/testing/selftests/powerpc/copyloops/exc_validate.c b/tools/testing/selftests/powerpc/copyloops/exc_validate.c
new file mode 100644
index 0000000..c896ea9
--- /dev/null
+++ b/tools/testing/selftests/powerpc/copyloops/exc_validate.c
@@ -0,0 +1,124 @@
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+#include <signal.h>
+#include <unistd.h>
+#include <sys/mman.h>
+
+#include "utils.h"
+
+extern char __start___ex_table[];
+extern char __stop___ex_table[];
+
+#if defined(__powerpc64__)
+#define UCONTEXT_NIA(UC)	(UC)->uc_mcontext.gp_regs[PT_NIP]
+#elif defined(__powerpc__)
+#define UCONTEXT_NIA(UC)	(UC)->uc_mcontext.uc_regs->gregs[PT_NIP]
+#else
+#error implement UCONTEXT_NIA
+#endif
+
+static void segv_handler(int signr, siginfo_t *info, void *ptr)
+{
+	ucontext_t *uc = (ucontext_t *)ptr;
+	unsigned long addr = (unsigned long)info->si_addr;
+	unsigned long *ip = &UCONTEXT_NIA(uc);
+	unsigned long *ex_p = (unsigned long *)__start___ex_table;
+
+	while (ex_p < (unsigned long *)__stop___ex_table) {
+		unsigned long insn, fixup;
+
+		insn = *ex_p++;
+		fixup = *ex_p++;
+
+		if (insn == *ip) {
+			*ip = fixup;
+			return;
+		}
+	}
+
+	printf("No exception table match for NIA %lx ADDR %lx\n", *ip, addr);
+	abort();
+}
+
+static void setup_segv_handler(void)
+{
+	struct sigaction action;
+
+	memset(&action, 0, sizeof(action));
+	action.sa_sigaction = segv_handler;
+	action.sa_flags = SA_SIGINFO;
+	sigaction(SIGSEGV, &action, NULL);
+}
+
+unsigned long COPY_LOOP(void *to, const void *from, unsigned long size);
+unsigned long test_copy_tofrom_user_reference(void *to, const void *from, unsigned long size);
+
+static int total_passed;
+static int total_failed;
+
+static void do_one_test(char *dstp, char *srcp, unsigned long len)
+{
+	unsigned long got, expected;
+
+	got = COPY_LOOP(dstp, srcp, len);
+	expected = test_copy_tofrom_user_reference(dstp, srcp, len);
+
+	if (got != expected) {
+		total_failed++;
+		printf("FAIL from=%p to=%p len=%ld returned %ld, expected %ld\n",
+		       srcp, dstp, len, got, expected);
+		//abort();
+	} else
+		total_passed++;
+}
+
+//#define MAX_LEN 512
+#define MAX_LEN 16
+
+int test_copy_exception(void)
+{
+	int page_size;
+	static char *p, *q;
+	unsigned long src, dst, len;
+
+	page_size = getpagesize();
+	p = mmap(NULL, page_size * 2, PROT_READ|PROT_WRITE,
+		MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
+
+	if (p == MAP_FAILED) {
+		perror("mmap");
+		exit(1);
+	}
+
+	memset(p, 0, page_size);
+
+	setup_segv_handler();
+
+	if (mprotect(p + page_size, page_size, PROT_NONE)) {
+		perror("mprotect");
+		exit(1);
+	}
+
+	q = p + page_size - MAX_LEN;
+
+	for (src = 0; src < MAX_LEN; src++) {
+		for (dst = 0; dst < MAX_LEN; dst++) {
+			for (len = 0; len < MAX_LEN+1; len++) {
+				// printf("from=%p to=%p len=%ld\n", q+dst, q+src, len);
+				do_one_test(q+dst, q+src, len);
+			}
+		}
+	}
+
+	printf("Totals:\n");
+	printf("  Pass: %d\n", total_passed);
+	printf("  Fail: %d\n", total_failed);
+
+	return 0;
+}
+
+int main(void)
+{
+	return test_harness(test_copy_exception, str(COPY_LOOP));
+}
diff --git a/tools/testing/selftests/powerpc/copyloops/stubs.S b/tools/testing/selftests/powerpc/copyloops/stubs.S
new file mode 100644
index 0000000..ec8bcf2
--- /dev/null
+++ b/tools/testing/selftests/powerpc/copyloops/stubs.S
@@ -0,0 +1,19 @@
+#include <asm/ppc_asm.h>
+
+FUNC_START(enter_vmx_usercopy)
+	li	r3,1
+	blr
+
+FUNC_START(exit_vmx_usercopy)
+	li	r3,0
+	blr
+
+FUNC_START(enter_vmx_ops)
+	li	r3,1
+	blr
+
+FUNC_START(exit_vmx_ops)
+	blr
+
+FUNC_START(__copy_tofrom_user_base)
+	blr
-- 
2.7.4

[PATCH v3 4/4] powerpc/64: Copy as much as possible in __copy_tofrom_user

[Paul Mackerras]

In __copy_tofrom_user, if we encounter an exception on a store, we
stop copying and return the number of bytes not copied.  However,
if the store is wider than one byte and is to an unaligned address,
it is possible that the store operand overlaps a page boundary
and the exception occurred on the latter part of the store operand,
meaning that it would be possible to copy a few more bytes.  Since
copy_to_user is generally expected to copy as much as possible,
it would be better to copy those extra few bytes.  This adds code
to do that.  Since this edge case is not performance-critical,
the code has been written to be compact rather than as fast as
possible.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>
---
 arch/powerpc/lib/copyuser_64.S | 29 +++++++++++++++++++++++------
 1 file changed, 23 insertions(+), 6 deletions(-)

diff --git a/arch/powerpc/lib/copyuser_64.S b/arch/powerpc/lib/copyuser_64.S
index 2197a35..96c514b 100644
--- a/arch/powerpc/lib/copyuser_64.S
+++ b/arch/powerpc/lib/copyuser_64.S
@@ -379,8 +379,8 @@ stex;	stb	r0,0(r3)
 	blr
 
 /*
- * exception handlers for stores: we just need to work
- * out how many bytes weren't copied
+ * exception handlers for stores: we need to work out how many bytes
+ * weren't copied, and we may need to copy some more.
  * Note that the number of bytes of instructions for adjusting r3 needs
  * to equal the amount of the adjustment, due to the trick of using
  * .Lst_exc - r3_offset as the handler address.
@@ -400,10 +400,27 @@ stex;	stb	r0,0(r3)
 	/* adjust by 4 */
 	addi	r3,r3,4
 .Lst_exc:
-	ld	r6,-24(r1)
-	ld	r5,-8(r1)
-	add	r6,r6,r5
-	subf	r3,r3,r6	/* #bytes not copied in r3 */
+	ld	r6,-24(r1)	/* original destination pointer */
+	ld	r4,-16(r1)	/* original source pointer */
+	ld	r5,-8(r1)	/* original number of bytes */
+	add	r7,r6,r5
+	/*
+	 * If the destination pointer isn't 8-byte aligned,
+	 * we may have got the exception as a result of a
+	 * store that overlapped a page boundary, so we may be
+	 * able to copy a few more bytes.
+	 */
+17:	andi.	r0,r3,7
+	beq	19f
+	subf	r8,r6,r3	/* #bytes copied */
+100:	EX_TABLE(100b,19f)
+	lbzx	r0,r8,r4
+100:	EX_TABLE(100b,19f)
+	stb	r0,0(r3)
+	addi	r3,r3,1
+	cmpld	r3,r7
+	blt	17b
+19:	subf	r3,r3,r7	/* #bytes not copied in r3 */
 	blr
 
 /*
-- 
2.7.4