13 files changed, 117 insertions, 61 deletions

diff --git a/arch/sparc/include/asm/compat.h b/arch/sparc/include/asm/compat.h
index cef99fbc0a21..830502fe62b4 100644
--- a/arch/sparc/include/asm/compat.h
+++ b/arch/sparc/include/asm/compat.h
@@ -232,9 +232,10 @@ static inline void __user *arch_compat_alloc_user_space(long len)
         struct pt_regs *regs = current_thread_info()->kregs;
         unsigned long usp = regs->u_regs[UREG_I6];
 
-        if (!(test_thread_flag(TIF_32BIT)))
+        if (test_thread_64bit_stack(usp))
                 usp += STACK_BIAS;
-        else
+
+        if (test_thread_flag(TIF_32BIT))
                 usp &= 0xffffffffUL;
 
         usp -= len;
diff --git a/arch/sparc/include/asm/thread_info_64.h b/arch/sparc/include/asm/thread_info_64.h
index 4e2276631081..a3fe4dcc0aa6 100644
--- a/arch/sparc/include/asm/thread_info_64.h
+++ b/arch/sparc/include/asm/thread_info_64.h
@@ -259,6 +259,11 @@ static inline bool test_and_clear_restore_sigmask(void)
 
 #define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG)
 
+#define thread32_stack_is_64bit(__SP) (((__SP) & 0x1) != 0)
+#define test_thread_64bit_stack(__SP) \
+        ((test_thread_flag(TIF_32BIT) && !thread32_stack_is_64bit(__SP)) ? \
+         false : true)
+
 #endif  /* !__ASSEMBLY__ */
 
 #endif /* __KERNEL__ */
diff --git a/arch/sparc/include/asm/ttable.h b/arch/sparc/include/asm/ttable.h
index 48f2807d3265..71b5a67522ab 100644
--- a/arch/sparc/include/asm/ttable.h
+++ b/arch/sparc/include/asm/ttable.h
@@ -372,7 +372,9 @@ etrap_spill_fixup_64bit:				\
 
 /* Normal 32bit spill */
 #define SPILL_2_GENERIC(ASI)                            \
-        srl     %sp, 0, %sp;                            \
+        and     %sp, 1, %g3;                            \
+        brnz,pn %g3, (. - (128 + 4));                   \
+         srl    %sp, 0, %sp;                            \
         stwa    %l0, [%sp + %g0] ASI;                   \
         mov     0x04, %g3;                              \
         stwa    %l1, [%sp + %g3] ASI;                   \
@@ -398,14 +400,16 @@ etrap_spill_fixup_64bit:				\
         stwa    %i6, [%g1 + %g0] ASI;                   \
         stwa    %i7, [%g1 + %g3] ASI;                   \
         saved;                                          \
-        retry; nop; nop;                                \
+        retry;                                          \
         b,a,pt  %xcc, spill_fixup_dax;                  \
         b,a,pt  %xcc, spill_fixup_mna;                  \
         b,a,pt  %xcc, spill_fixup;
 
 #define SPILL_2_GENERIC_ETRAP           \
 etrap_user_spill_32bit:                 \
-        srl     %sp, 0, %sp;            \
+        and     %sp, 1, %g3;            \
+        brnz,pn %g3, etrap_user_spill_64bit;    \
+         srl    %sp, 0, %sp;            \
         stwa    %l0, [%sp + 0x00] %asi; \
         stwa    %l1, [%sp + 0x04] %asi; \
         stwa    %l2, [%sp + 0x08] %asi; \
@@ -427,7 +431,7 @@ etrap_user_spill_32bit:			\
         ba,pt   %xcc, etrap_save;       \
          wrpr   %g1, %cwp;              \
         nop; nop; nop; nop;             \
-        nop; nop; nop; nop;             \
+        nop; nop;                       \
         ba,a,pt %xcc, etrap_spill_fixup_32bit; \
         ba,a,pt %xcc, etrap_spill_fixup_32bit; \
         ba,a,pt %xcc, etrap_spill_fixup_32bit;
@@ -592,7 +596,9 @@ user_rtt_fill_64bit:					\
 
 /* Normal 32bit fill */
 #define FILL_2_GENERIC(ASI)                             \
-        srl     %sp, 0, %sp;                            \
+        and     %sp, 1, %g3;                            \
+        brnz,pn %g3, (. - (128 + 4));                   \
+         srl    %sp, 0, %sp;                            \
         lduwa   [%sp + %g0] ASI, %l0;                   \
         mov     0x04, %g2;                              \
         mov     0x08, %g3;                              \
@@ -616,14 +622,16 @@ user_rtt_fill_64bit:					\
         lduwa   [%g1 + %g3] ASI, %i6;                   \
         lduwa   [%g1 + %g5] ASI, %i7;                   \
         restored;                                       \
-        retry; nop; nop; nop; nop;                      \
+        retry; nop; nop;                                \
         b,a,pt  %xcc, fill_fixup_dax;                   \
         b,a,pt  %xcc, fill_fixup_mna;                   \
         b,a,pt  %xcc, fill_fixup;
 
 #define FILL_2_GENERIC_RTRAP                            \
 user_rtt_fill_32bit:                                    \
-        srl     %sp, 0, %sp;                            \
+        and     %sp, 1, %g3;                            \
+        brnz,pn %g3, user_rtt_fill_64bit;               \
+         srl    %sp, 0, %sp;                            \
         lduwa   [%sp + 0x00] %asi, %l0;                 \
         lduwa   [%sp + 0x04] %asi, %l1;                 \
         lduwa   [%sp + 0x08] %asi, %l2;                 \
@@ -643,7 +651,7 @@ user_rtt_fill_32bit:					\
         ba,pt   %xcc, user_rtt_pre_restore;             \
          restored;                                      \
         nop; nop; nop; nop; nop;                        \
-        nop; nop; nop; nop; nop;                        \
+        nop; nop; nop;                                  \
         ba,a,pt %xcc, user_rtt_fill_fixup;              \
         ba,a,pt %xcc, user_rtt_fill_fixup;              \
         ba,a,pt %xcc, user_rtt_fill_fixup;
diff --git a/arch/sparc/include/uapi/asm/unistd.h b/arch/sparc/include/uapi/asm/unistd.h
index 8974ef7ae920..bed86a820d09 100644
--- a/arch/sparc/include/uapi/asm/unistd.h
+++ b/arch/sparc/include/uapi/asm/unistd.h
@@ -405,8 +405,12 @@
 #define __NR_setns              337
 #define __NR_process_vm_readv   338
 #define __NR_process_vm_writev  339
+#define __NR_kern_features      340
 
-#define NR_syscalls             340
+#define NR_syscalls             341
+
+/* Bitmask values returned from kern_features system call.  */
+#define KERN_FEATURE_MIXED_MODE_STACK   0x00000001
 
 #ifdef __32bit_syscall_numbers__
 /* Sparc 32-bit only has the "setresuid32", "getresuid32" variants,
diff --git a/arch/sparc/kernel/perf_event.c b/arch/sparc/kernel/perf_event.c
index 885a8af74064..b5c38faa4ead 100644
--- a/arch/sparc/kernel/perf_event.c
+++ b/arch/sparc/kernel/perf_event.c
@@ -1762,15 +1762,25 @@ static void perf_callchain_user_32(struct perf_callchain_entry *entry,
 
         ufp = regs->u_regs[UREG_I6] & 0xffffffffUL;
         do {
-                struct sparc_stackf32 *usf, sf;
                 unsigned long pc;
 
-                usf = (struct sparc_stackf32 *) ufp;
-                if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
-                        break;
+                if (thread32_stack_is_64bit(ufp)) {
+                        struct sparc_stackf *usf, sf;
 
-                pc = sf.callers_pc;
-                ufp = (unsigned long)sf.fp;
+                        ufp += STACK_BIAS;
+                        usf = (struct sparc_stackf *) ufp;
+                        if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
+                                break;
+                        pc = sf.callers_pc & 0xffffffff;
+                        ufp = ((unsigned long) sf.fp) & 0xffffffff;
+                } else {
+                        struct sparc_stackf32 *usf, sf;
+                        usf = (struct sparc_stackf32 *) ufp;
+                        if (__copy_from_user_inatomic(&sf, usf, sizeof(sf)))
+                                break;
+                        pc = sf.callers_pc;
+                        ufp = (unsigned long)sf.fp;
+                }
                 perf_callchain_store(entry, pc);
         } while (entry->nr < PERF_MAX_STACK_DEPTH);
 }
diff --git a/arch/sparc/kernel/process_64.c b/arch/sparc/kernel/process_64.c
index d778248ef3f8..c6e0c2910043 100644
--- a/arch/sparc/kernel/process_64.c
+++ b/arch/sparc/kernel/process_64.c
@@ -452,13 +452,16 @@ void flush_thread(void)
 /* It's a bit more tricky when 64-bit tasks are involved... */
 static unsigned long clone_stackframe(unsigned long csp, unsigned long psp)
 {
+        bool stack_64bit = test_thread_64bit_stack(psp);
         unsigned long fp, distance, rval;
 
-        if (!(test_thread_flag(TIF_32BIT))) {
+        if (stack_64bit) {
                 csp += STACK_BIAS;
                 psp += STACK_BIAS;
                 __get_user(fp, &(((struct reg_window __user *)psp)->ins[6]));
                 fp += STACK_BIAS;
+                if (test_thread_flag(TIF_32BIT))
+                        fp &= 0xffffffff;
         } else
                 __get_user(fp, &(((struct reg_window32 __user *)psp)->ins[6]));
 
@@ -472,7 +475,7 @@ static unsigned long clone_stackframe(unsigned long csp, unsigned long psp)
         rval = (csp - distance);
         if (copy_in_user((void __user *) rval, (void __user *) psp, distance))
                 rval = 0;
-        else if (test_thread_flag(TIF_32BIT)) {
+        else if (!stack_64bit) {
                 if (put_user(((u32)csp),
                              &(((struct reg_window32 __user *)rval)->ins[6])))
                         rval = 0;
@@ -507,18 +510,18 @@ void synchronize_user_stack(void)
 
         flush_user_windows();
         if ((window = get_thread_wsaved()) != 0) {
-                int winsize = sizeof(struct reg_window);
-                int bias = 0;
-
-                if (test_thread_flag(TIF_32BIT))
-                        winsize = sizeof(struct reg_window32);
-                else
-                        bias = STACK_BIAS;
-
                 window -= 1;
                 do {
-                        unsigned long sp = (t->rwbuf_stkptrs[window] + bias);
                         struct reg_window *rwin = &t->reg_window[window];
+                        int winsize = sizeof(struct reg_window);
+                        unsigned long sp;
+
+                        sp = t->rwbuf_stkptrs[window];
+
+                        if (test_thread_64bit_stack(sp))
+                                sp += STACK_BIAS;
+                        else
+                                winsize = sizeof(struct reg_window32);
 
                         if (!copy_to_user((char __user *)sp, rwin, winsize)) {
                                 shift_window_buffer(window, get_thread_wsaved() - 1, t);
@@ -544,13 +547,6 @@ void fault_in_user_windows(void)
 {
         struct thread_info *t = current_thread_info();
         unsigned long window;
-        int winsize = sizeof(struct reg_window);
-        int bias = 0;
-
-        if (test_thread_flag(TIF_32BIT))
-                winsize = sizeof(struct reg_window32);
-        else
-                bias = STACK_BIAS;
 
         flush_user_windows();
         window = get_thread_wsaved();
@@ -558,8 +554,16 @@ void fault_in_user_windows(void)
         if (likely(window != 0)) {
                 window -= 1;
                 do {
-                        unsigned long sp = (t->rwbuf_stkptrs[window] + bias);
                         struct reg_window *rwin = &t->reg_window[window];
+                        int winsize = sizeof(struct reg_window);
+                        unsigned long sp;
+
+                        sp = t->rwbuf_stkptrs[window];
+
+                        if (test_thread_64bit_stack(sp))
+                                sp += STACK_BIAS;
+                        else
+                                winsize = sizeof(struct reg_window32);
 
                         if (unlikely(sp & 0x7UL))
                                 stack_unaligned(sp);
diff --git a/arch/sparc/kernel/ptrace_64.c b/arch/sparc/kernel/ptrace_64.c
index 484dabac7045..7ff45e4ba681 100644
--- a/arch/sparc/kernel/ptrace_64.c
+++ b/arch/sparc/kernel/ptrace_64.c
@@ -151,7 +151,7 @@ static int regwindow64_get(struct task_struct *target,
 {
         unsigned long rw_addr = regs->u_regs[UREG_I6];
 
-        if (test_tsk_thread_flag(current, TIF_32BIT)) {
+        if (!test_thread_64bit_stack(rw_addr)) {
                 struct reg_window32 win32;
                 int i;
 
@@ -176,7 +176,7 @@ static int regwindow64_set(struct task_struct *target,
 {
         unsigned long rw_addr = regs->u_regs[UREG_I6];
 
-        if (test_tsk_thread_flag(current, TIF_32BIT)) {
+        if (!test_thread_64bit_stack(rw_addr)) {
                 struct reg_window32 win32;
                 int i;
 
diff --git a/arch/sparc/kernel/sys_sparc_64.c b/arch/sparc/kernel/sys_sparc_64.c
index 11c6c9603e71..878ef3d5fec5 100644
--- a/arch/sparc/kernel/sys_sparc_64.c
+++ b/arch/sparc/kernel/sys_sparc_64.c
@@ -751,3 +751,8 @@ int kernel_execve(const char *filename,
                       : "cc");
         return __res;
 }
+
+asmlinkage long sys_kern_features(void)
+{
+        return KERN_FEATURE_MIXED_MODE_STACK;
+}
diff --git a/arch/sparc/kernel/systbls_64.S b/arch/sparc/kernel/systbls_64.S
index 3a58e0d66f51..45ce6be088e4 100644
--- a/arch/sparc/kernel/systbls_64.S
+++ b/arch/sparc/kernel/systbls_64.S
@@ -86,6 +86,7 @@ sys_call_table32:
         .word compat_sys_pwritev, compat_sys_rt_tgsigqueueinfo, sys_perf_event_open, compat_sys_recvmmsg, sys_fanotify_init
 /*330*/ .word sys32_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, compat_sys_open_by_handle_at, compat_sys_clock_adjtime
         .word sys_syncfs, compat_sys_sendmmsg, sys_setns, compat_sys_process_vm_readv, compat_sys_process_vm_writev
+/*340*/ .word sys_kern_features
 
 #endif /* CONFIG_COMPAT */
 
@@ -163,3 +164,4 @@ sys_call_table:
         .word sys_pwritev, sys_rt_tgsigqueueinfo, sys_perf_event_open, sys_recvmmsg, sys_fanotify_init
 /*330*/ .word sys_fanotify_mark, sys_prlimit64, sys_name_to_handle_at, sys_open_by_handle_at, sys_clock_adjtime
         .word sys_syncfs, sys_sendmmsg, sys_setns, sys_process_vm_readv, sys_process_vm_writev
+/*340*/ .word sys_kern_features
diff --git a/arch/sparc/kernel/unaligned_64.c b/arch/sparc/kernel/unaligned_64.c
index f81d038f7340..8201c25e7669 100644
--- a/arch/sparc/kernel/unaligned_64.c
+++ b/arch/sparc/kernel/unaligned_64.c
@@ -113,21 +113,24 @@ static inline long sign_extend_imm13(long imm)
 
 static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 {
-        unsigned long value;
+        unsigned long value, fp;
         
         if (reg < 16)
                 return (!reg ? 0 : regs->u_regs[reg]);
+
+        fp = regs->u_regs[UREG_FP];
+
         if (regs->tstate & TSTATE_PRIV) {
                 struct reg_window *win;
-                win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                win = (struct reg_window *)(fp + STACK_BIAS);
                 value = win->locals[reg - 16];
-        } else if (test_thread_flag(TIF_32BIT)) {
+        } else if (!test_thread_64bit_stack(fp)) {
                 struct reg_window32 __user *win32;
-                win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+                win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
                 get_user(value, &win32->locals[reg - 16]);
         } else {
                 struct reg_window __user *win;
-                win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                win = (struct reg_window __user *)(fp + STACK_BIAS);
                 get_user(value, &win->locals[reg - 16]);
         }
         return value;
@@ -135,19 +138,24 @@ static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 
 static unsigned long *fetch_reg_addr(unsigned int reg, struct pt_regs *regs)
 {
+        unsigned long fp;
+
         if (reg < 16)
                 return &regs->u_regs[reg];
+
+        fp = regs->u_regs[UREG_FP];
+
         if (regs->tstate & TSTATE_PRIV) {
                 struct reg_window *win;
-                win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                win = (struct reg_window *)(fp + STACK_BIAS);
                 return &win->locals[reg - 16];
-        } else if (test_thread_flag(TIF_32BIT)) {
+        } else if (!test_thread_64bit_stack(fp)) {
                 struct reg_window32 *win32;
-                win32 = (struct reg_window32 *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+                win32 = (struct reg_window32 *)((unsigned long)((u32)fp));
                 return (unsigned long *)&win32->locals[reg - 16];
         } else {
                 struct reg_window *win;
-                win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                win = (struct reg_window *)(fp + STACK_BIAS);
                 return &win->locals[reg - 16];
         }
 }
@@ -392,13 +400,15 @@ int handle_popc(u32 insn, struct pt_regs *regs)
                 if (rd)
                         regs->u_regs[rd] = ret;
         } else {
-                if (test_thread_flag(TIF_32BIT)) {
+                unsigned long fp = regs->u_regs[UREG_FP];
+
+                if (!test_thread_64bit_stack(fp)) {
                         struct reg_window32 __user *win32;
-                        win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+                        win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
                         put_user(ret, &win32->locals[rd - 16]);
                 } else {
                         struct reg_window __user *win;
-                        win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                        win = (struct reg_window __user *)(fp + STACK_BIAS);
                         put_user(ret, &win->locals[rd - 16]);
                 }
         }
@@ -554,7 +564,7 @@ void handle_ld_nf(u32 insn, struct pt_regs *regs)
                 reg[0] = 0;
                 if ((insn & 0x780000) == 0x180000)
                         reg[1] = 0;
-        } else if (test_thread_flag(TIF_32BIT)) {
+        } else if (!test_thread_64bit_stack(regs->u_regs[UREG_FP])) {
                 put_user(0, (int __user *) reg);
                 if ((insn & 0x780000) == 0x180000)
                         put_user(0, ((int __user *) reg) + 1);
diff --git a/arch/sparc/kernel/visemul.c b/arch/sparc/kernel/visemul.c
index 08e074b7eb6a..c096c624ac4d 100644
--- a/arch/sparc/kernel/visemul.c
+++ b/arch/sparc/kernel/visemul.c
@@ -149,21 +149,24 @@ static inline void maybe_flush_windows(unsigned int rs1, unsigned int rs2,
 
 static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 {
-        unsigned long value;
+        unsigned long value, fp;
         
         if (reg < 16)
                 return (!reg ? 0 : regs->u_regs[reg]);
+
+        fp = regs->u_regs[UREG_FP];
+
         if (regs->tstate & TSTATE_PRIV) {
                 struct reg_window *win;
-                win = (struct reg_window *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                win = (struct reg_window *)(fp + STACK_BIAS);
                 value = win->locals[reg - 16];
-        } else if (test_thread_flag(TIF_32BIT)) {
+        } else if (!test_thread_64bit_stack(fp)) {
                 struct reg_window32 __user *win32;
-                win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+                win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
                 get_user(value, &win32->locals[reg - 16]);
         } else {
                 struct reg_window __user *win;
-                win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                win = (struct reg_window __user *)(fp + STACK_BIAS);
                 get_user(value, &win->locals[reg - 16]);
         }
         return value;
@@ -172,16 +175,18 @@ static unsigned long fetch_reg(unsigned int reg, struct pt_regs *regs)
 static inline unsigned long __user *__fetch_reg_addr_user(unsigned int reg,
                                                           struct pt_regs *regs)
 {
+        unsigned long fp = regs->u_regs[UREG_FP];
+
         BUG_ON(reg < 16);
         BUG_ON(regs->tstate & TSTATE_PRIV);
 
-        if (test_thread_flag(TIF_32BIT)) {
+        if (!test_thread_64bit_stack(fp)) {
                 struct reg_window32 __user *win32;
-                win32 = (struct reg_window32 __user *)((unsigned long)((u32)regs->u_regs[UREG_FP]));
+                win32 = (struct reg_window32 __user *)((unsigned long)((u32)fp));
                 return (unsigned long __user *)&win32->locals[reg - 16];
         } else {
                 struct reg_window __user *win;
-                win = (struct reg_window __user *)(regs->u_regs[UREG_FP] + STACK_BIAS);
+                win = (struct reg_window __user *)(fp + STACK_BIAS);
                 return &win->locals[reg - 16];
         }
 }
@@ -204,7 +209,7 @@ static void store_reg(struct pt_regs *regs, unsigned long val, unsigned long rd)
         } else {
                 unsigned long __user *rd_user = __fetch_reg_addr_user(rd, regs);
 
-                if (test_thread_flag(TIF_32BIT))
+                if (!test_thread_64bit_stack(regs->u_regs[UREG_FP]))
                         __put_user((u32)val, (u32 __user *)rd_user);
                 else
                         __put_user(val, rd_user);
diff --git a/arch/sparc/kernel/winfixup.S b/arch/sparc/kernel/winfixup.S
index a6b0863c27df..1e67ce958369 100644
--- a/arch/sparc/kernel/winfixup.S
+++ b/arch/sparc/kernel/winfixup.S
@@ -43,6 +43,8 @@ spill_fixup_mna:
 spill_fixup_dax:
         TRAP_LOAD_THREAD_REG(%g6, %g1)
         ldx     [%g6 + TI_FLAGS], %g1
+        andcc   %sp, 0x1, %g0
+        movne   %icc, 0, %g1
         andcc   %g1, _TIF_32BIT, %g0
         ldub    [%g6 + TI_WSAVED], %g1
         sll     %g1, 3, %g3
diff --git a/arch/sparc/math-emu/math_64.c b/arch/sparc/math-emu/math_64.c
index 1704068da928..034aadbff036 100644
--- a/arch/sparc/math-emu/math_64.c
+++ b/arch/sparc/math-emu/math_64.c
@@ -320,7 +320,7 @@ int do_mathemu(struct pt_regs *regs, struct fpustate *f, bool illegal_insn_trap)
                                         XR = 0;
                                 else if (freg < 16)
                                         XR = regs->u_regs[freg];
-                                else if (test_thread_flag(TIF_32BIT)) {
+                                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]));