Stuff for 4.2 (including a revert)

Hi all-

Patch 1 is a revert, Cc'd to stable. I haven't tried to write a real test for the bug, but the code definitely looks buggy. I intent to fix it for real as part of converting more of the syscall entry code to C, but this is a better fix for now.

Patch 2 fixes selftests fallout from the sigcontext revert.

Patch 3 is a new test that I thought I added but apparently didn't. It's probably 4.3 material, but it's related to patch 1, so I'm sending them together.

Andy Lutomirski (3):
Revert "sched/x86_64: Don't save flags on context switch" selftests/x86: Disable sigreturn_64 selftests/x86: Add syscall_nt selftest

 arch/x86/include/asm/switch_to.h         | 12 +++----
 tools/testing/selftests/x86/Makefile     |  4 +--
 tools/testing/selftests/x86/syscall_nt.c | 54 ++++++++++++++++++++++++++++++++
 3 files changed, 60 insertions(+), 10 deletions(-)
 create mode 100644 tools/testing/selftests/x86/syscall_nt.c

2.4.3

This reverts commit 2c7577a7583747c9b71f26dced7f696b739da745.

It was a nice speedup. It's also not quite correct: SYSENTER enables interrupts too early.

We can re-add this optimization once the SYSENTER code is beaten into shape, which should happen in 4.3 or 4.4.

---
 arch/x86/include/asm/switch_to.h | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)

index 751bf4b7bf11..d7f3b3b78ac3 100644
@@ -79,12 +79,12 @@ do {									\
 #else /* CONFIG_X86_32 */
 
 /* frame pointer must be last for get_wchan */
-#define SAVE_CONTEXT    "pushq %%rbp ; movq %%rsi,%%rbp\n\t"
-#define RESTORE_CONTEXT "movq %%rbp,%%rsi ; popq %%rbp\t"
+#define SAVE_CONTEXT    "pushf ; pushq %%rbp ; movq %%rsi,%%rbp\n\t"
+#define RESTORE_CONTEXT "movq %%rbp,%%rsi ; popq %%rbp ; popf\t"
 
 #define __EXTRA_CLOBBER  \
 	, "rcx", "rbx", "rdx", "r8", "r9", "r10", "r11", \
-	  "r12", "r13", "r14", "r15", "flags"
+	  "r12", "r13", "r14", "r15"
 
 #ifdef CONFIG_CC_STACKPROTECTOR
 #define __switch_canary							  \

@@ -100,11 +100,7 @@ do {									\
 #define __switch_canary_iparam
 #endif	/* CC_STACKPROTECTOR */
 
-/*
- * There is no need to save or restore flags, because flags are always
- * clean in kernel mode, with the possible exception of IOPL.  Kernel IOPL
- * has no effect.
- */
+/* Save restore flags to clear handle leaking NT */
 #define switch_to(prev, next, last) \
 	asm volatile(SAVE_CONTEXT					  \
 	     "movq %%rsp,%P[threadrsp](%[prev])\n\t" /* save RSP */	  \

2.4.3

I've had this sitting around for a while. Add it to the selftests tree. Far Cry running under Wine depends on this behavior.

---
 tools/testing/selftests/x86/Makefile     |  2 +-
 tools/testing/selftests/x86/syscall_nt.c | 54 ++++++++++++++++++++++++++++++++
 2 files changed, 55 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/selftests/x86/syscall_nt.c

index 1cda40be06ca..2221f2ba449d 100644
@@ -4,7 +4,7 @@ include ../lib.mk
 
 .PHONY: all all_32 all_64 warn_32bit_failure clean
 
-TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs
+TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs syscall_nt
 TARGETS_C_32BIT_ONLY := entry_from_vm86 sigreturn
 
 TARGETS_C_32BIT_ALL := $(TARGETS_C_BOTHBITS) $(TARGETS_C_32BIT_ONLY)

2.4.3

sigreturn_64 was broken by ed596cde9425 ("Revert x86 sigcontext cleanups"). Turn it off until we have a better fix.

---
 tools/testing/selftests/x86/Makefile | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

index caa60d56d7d1..1cda40be06ca 100644
@@ -4,8 +4,8 @@ include ../lib.mk
 
 .PHONY: all all_32 all_64 warn_32bit_failure clean
 
-TARGETS_C_BOTHBITS := sigreturn single_step_syscall sysret_ss_attrs
-TARGETS_C_32BIT_ONLY := entry_from_vm86
+TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs
+TARGETS_C_32BIT_ONLY := entry_from_vm86 sigreturn
 
 TARGETS_C_32BIT_ALL := $(TARGETS_C_BOTHBITS) $(TARGETS_C_32BIT_ONLY)
 BINARIES_32 := $(TARGETS_C_32BIT_ALL:%=%_32)

2.4.3