summaryrefslogtreecommitdiff
path: root/arch/sparc/lib/NG2memcpy.S
diff options
context:
space:
mode:
authorRob Gardner <rob.gardner@oracle.com>2015-12-23 09:24:49 +0300
committerDavid S. Miller <davem@davemloft.net>2015-12-24 20:13:18 +0300
commita7c5724b5c17775ca8ea2fd9906d8a7e37337cce (patch)
tree802877bf40b98d95256551b97ac42492705291d3 /arch/sparc/lib/NG2memcpy.S
parent833526941f5945cf0b22a595bb8f3525b512f654 (diff)
downloadlinux-a7c5724b5c17775ca8ea2fd9906d8a7e37337cce.tar.xz
sparc64: fix FP corruption in user copy functions
Short story: Exception handlers used by some copy_to_user() and copy_from_user() functions do not diligently clean up floating point register usage, and this can result in a user process seeing invalid values in floating point registers. This sometimes makes the process fail. Long story: Several cpu-specific (NG4, NG2, U1, U3) memcpy functions use floating point registers and VIS alignaddr/faligndata to accelerate data copying when source and dest addresses don't align well. Linux uses a lazy scheme for saving floating point registers; It is not done upon entering the kernel since it's a very expensive operation. Rather, it is done only when needed. If the kernel ends up not using FP regs during the course of some trap or system call, then it can return to user space without saving or restoring them. The various memcpy functions begin their FP code with VISEntry (or a variation thereof), which saves the FP regs. They conclude their FP code with VISExit (or a variation) which essentially marks the FP regs "clean", ie, they contain no unsaved values. fprs.FPRS_FEF is turned off so that a lazy restore will be triggered when/if the user process accesses floating point regs again. The bug is that the user copy variants of memcpy, copy_from_user() and copy_to_user(), employ an exception handling mechanism to detect faults when accessing user space addresses, and when this handler is invoked, an immediate return from the function is forced, and VISExit is not executed, thus leaving the fprs register in an indeterminate state, but often with fprs.FPRS_FEF set and one or more dirty bits. This results in a return to user space with invalid values in the FP regs, and since fprs.FPRS_FEF is on, no lazy restore occurs. This bug affects copy_to_user() and copy_from_user() for NG4, NG2, U3, and U1. All are fixed by using a new exception handler for those loads and stores that are done during the time between VISEnter and VISExit. n.b. In NG4memcpy, the problematic code can be triggered by a copy size greater than 128 bytes and an unaligned source address. This bug is known to be the cause of random user process memory corruptions while perf is running with the callgraph option (ie, perf record -g). This occurs because perf uses copy_from_user() to read user stacks, and may fault when it follows a stack frame pointer off to an invalid page. Validation checks on the stack address just obscure the underlying problem. Signed-off-by: Rob Gardner <rob.gardner@oracle.com> Signed-off-by: Dave Aldridge <david.j.aldridge@oracle.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'arch/sparc/lib/NG2memcpy.S')
-rw-r--r--arch/sparc/lib/NG2memcpy.S118
1 files changed, 62 insertions, 56 deletions
diff --git a/arch/sparc/lib/NG2memcpy.S b/arch/sparc/lib/NG2memcpy.S
index 30eee6e8a81b..d5f585df2f3f 100644
--- a/arch/sparc/lib/NG2memcpy.S
+++ b/arch/sparc/lib/NG2memcpy.S
@@ -34,10 +34,16 @@
#ifndef EX_LD
#define EX_LD(x) x
#endif
+#ifndef EX_LD_FP
+#define EX_LD_FP(x) x
+#endif
#ifndef EX_ST
#define EX_ST(x) x
#endif
+#ifndef EX_ST_FP
+#define EX_ST_FP(x) x
+#endif
#ifndef EX_RETVAL
#define EX_RETVAL(x) x
@@ -134,40 +140,40 @@
fsrc2 %x6, %f12; \
fsrc2 %x7, %f14;
#define FREG_LOAD_1(base, x0) \
- EX_LD(LOAD(ldd, base + 0x00, %x0))
+ EX_LD_FP(LOAD(ldd, base + 0x00, %x0))
#define FREG_LOAD_2(base, x0, x1) \
- EX_LD(LOAD(ldd, base + 0x00, %x0)); \
- EX_LD(LOAD(ldd, base + 0x08, %x1));
+ EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+ EX_LD_FP(LOAD(ldd, base + 0x08, %x1));
#define FREG_LOAD_3(base, x0, x1, x2) \
- EX_LD(LOAD(ldd, base + 0x00, %x0)); \
- EX_LD(LOAD(ldd, base + 0x08, %x1)); \
- EX_LD(LOAD(ldd, base + 0x10, %x2));
+ EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+ EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+ EX_LD_FP(LOAD(ldd, base + 0x10, %x2));
#define FREG_LOAD_4(base, x0, x1, x2, x3) \
- EX_LD(LOAD(ldd, base + 0x00, %x0)); \
- EX_LD(LOAD(ldd, base + 0x08, %x1)); \
- EX_LD(LOAD(ldd, base + 0x10, %x2)); \
- EX_LD(LOAD(ldd, base + 0x18, %x3));
+ EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+ EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+ EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+ EX_LD_FP(LOAD(ldd, base + 0x18, %x3));
#define FREG_LOAD_5(base, x0, x1, x2, x3, x4) \
- EX_LD(LOAD(ldd, base + 0x00, %x0)); \
- EX_LD(LOAD(ldd, base + 0x08, %x1)); \
- EX_LD(LOAD(ldd, base + 0x10, %x2)); \
- EX_LD(LOAD(ldd, base + 0x18, %x3)); \
- EX_LD(LOAD(ldd, base + 0x20, %x4));
+ EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+ EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+ EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+ EX_LD_FP(LOAD(ldd, base + 0x18, %x3)); \
+ EX_LD_FP(LOAD(ldd, base + 0x20, %x4));
#define FREG_LOAD_6(base, x0, x1, x2, x3, x4, x5) \
- EX_LD(LOAD(ldd, base + 0x00, %x0)); \
- EX_LD(LOAD(ldd, base + 0x08, %x1)); \
- EX_LD(LOAD(ldd, base + 0x10, %x2)); \
- EX_LD(LOAD(ldd, base + 0x18, %x3)); \
- EX_LD(LOAD(ldd, base + 0x20, %x4)); \
- EX_LD(LOAD(ldd, base + 0x28, %x5));
+ EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+ EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+ EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+ EX_LD_FP(LOAD(ldd, base + 0x18, %x3)); \
+ EX_LD_FP(LOAD(ldd, base + 0x20, %x4)); \
+ EX_LD_FP(LOAD(ldd, base + 0x28, %x5));
#define FREG_LOAD_7(base, x0, x1, x2, x3, x4, x5, x6) \
- EX_LD(LOAD(ldd, base + 0x00, %x0)); \
- EX_LD(LOAD(ldd, base + 0x08, %x1)); \
- EX_LD(LOAD(ldd, base + 0x10, %x2)); \
- EX_LD(LOAD(ldd, base + 0x18, %x3)); \
- EX_LD(LOAD(ldd, base + 0x20, %x4)); \
- EX_LD(LOAD(ldd, base + 0x28, %x5)); \
- EX_LD(LOAD(ldd, base + 0x30, %x6));
+ EX_LD_FP(LOAD(ldd, base + 0x00, %x0)); \
+ EX_LD_FP(LOAD(ldd, base + 0x08, %x1)); \
+ EX_LD_FP(LOAD(ldd, base + 0x10, %x2)); \
+ EX_LD_FP(LOAD(ldd, base + 0x18, %x3)); \
+ EX_LD_FP(LOAD(ldd, base + 0x20, %x4)); \
+ EX_LD_FP(LOAD(ldd, base + 0x28, %x5)); \
+ EX_LD_FP(LOAD(ldd, base + 0x30, %x6));
.register %g2,#scratch
.register %g3,#scratch
@@ -275,11 +281,11 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
nop
/* fall through for 0 < low bits < 8 */
110: sub %o4, 64, %g2
- EX_LD(LOAD_BLK(%g2, %f0))
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+ EX_LD_FP(LOAD_BLK(%g2, %f0))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f2, f4, f6, f8, f10, f12, f14, f16)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_8(f16, f18, f20, f22, f24, f26, f28, f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -290,10 +296,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
120: sub %o4, 56, %g2
FREG_LOAD_7(%g2, f0, f2, f4, f6, f8, f10, f12)
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f2, f4, f6, f8, f10, f12, f16, f18)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_7(f18, f20, f22, f24, f26, f28, f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -304,10 +310,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
130: sub %o4, 48, %g2
FREG_LOAD_6(%g2, f0, f2, f4, f6, f8, f10)
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f2, f4, f6, f8, f10, f16, f18, f20)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_6(f20, f22, f24, f26, f28, f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -318,10 +324,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
140: sub %o4, 40, %g2
FREG_LOAD_5(%g2, f0, f2, f4, f6, f8)
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f2, f4, f6, f8, f16, f18, f20, f22)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_5(f22, f24, f26, f28, f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -332,10 +338,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
150: sub %o4, 32, %g2
FREG_LOAD_4(%g2, f0, f2, f4, f6)
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f2, f4, f6, f16, f18, f20, f22, f24)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_4(f24, f26, f28, f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -346,10 +352,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
160: sub %o4, 24, %g2
FREG_LOAD_3(%g2, f0, f2, f4)
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f2, f4, f16, f18, f20, f22, f24, f26)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_3(f26, f28, f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -360,10 +366,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
170: sub %o4, 16, %g2
FREG_LOAD_2(%g2, f0, f2)
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f2, f16, f18, f20, f22, f24, f26, f28)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_2(f28, f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -374,10 +380,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
180: sub %o4, 8, %g2
FREG_LOAD_1(%g2, f0)
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
- EX_LD(LOAD_BLK(%o4, %f16))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f16))
FREG_FROB(f0, f16, f18, f20, f22, f24, f26, f28, f30)
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
FREG_MOVE_1(f30)
subcc %g1, 64, %g1
add %o4, 64, %o4
@@ -387,10 +393,10 @@ FUNC_NAME: /* %o0=dst, %o1=src, %o2=len */
nop
190:
-1: EX_ST(STORE_INIT(%g0, %o4 + %g3))
+1: EX_ST_FP(STORE_INIT(%g0, %o4 + %g3))
subcc %g1, 64, %g1
- EX_LD(LOAD_BLK(%o4, %f0))
- EX_ST(STORE_BLK(%f0, %o4 + %g3))
+ EX_LD_FP(LOAD_BLK(%o4, %f0))
+ EX_ST_FP(STORE_BLK(%f0, %o4 + %g3))
add %o4, 64, %o4
bne,pt %xcc, 1b
LOAD(prefetch, %o4 + 64, #one_read)