x86/paravirt: split sysret and sysexit

Don't conflate sysret and sysexit; they're different instructions with
different semantics, and may be in use at the same time (at least
within the same kernel, depending on whether its an Intel or AMD
system).

sysexit - just return to userspace, does no register restoration of
    any kind; must explicitly atomically enable interrupts.

sysret - reloads flags from r11, so no need to explicitly enable
    interrupts on 64-bit, responsible for restoring usermode %gs
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citirx.com>
Cc: xen-devel <xen-devel@lists.xensource.com>
Cc: Stephen Tweedie <sct@redhat.com>
Cc: Eduardo Habkost <ehabkost@redhat.com>
Cc: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

arch/x86/kernel/asm-offsets_32.c

@@ -111,7 +111,7 @@ void foo(void)
 	OFFSET(PV_IRQ_irq_disable, pv_irq_ops, irq_disable);
 	OFFSET(PV_IRQ_irq_enable, pv_irq_ops, irq_enable);
 	OFFSET(PV_CPU_iret, pv_cpu_ops, iret);
-	OFFSET(PV_CPU_irq_enable_syscall_ret, pv_cpu_ops, irq_enable_syscall_ret);
+	OFFSET(PV_CPU_irq_enable_sysexit, pv_cpu_ops, irq_enable_sysexit);
 	OFFSET(PV_CPU_read_cr0, pv_cpu_ops, read_cr0);
 #endif
 

arch/x86/kernel/asm-offsets_64.c

@@ -62,7 +62,7 @@ int main(void)
 	OFFSET(PV_IRQ_irq_disable, pv_irq_ops, irq_disable);
 	OFFSET(PV_IRQ_irq_enable, pv_irq_ops, irq_enable);
 	OFFSET(PV_CPU_iret, pv_cpu_ops, iret);
-	OFFSET(PV_CPU_irq_enable_syscall_ret, pv_cpu_ops, irq_enable_syscall_ret);
+	OFFSET(PV_CPU_usersp_sysret, pv_cpu_ops, usersp_sysret);
 	OFFSET(PV_CPU_swapgs, pv_cpu_ops, swapgs);
 	OFFSET(PV_MMU_read_cr2, pv_mmu_ops, read_cr2);
 #endif

arch/x86/kernel/entry_32.S

@@ -58,7 +58,7 @@
  * for paravirtualization.  The following will never clobber any registers:
  *   INTERRUPT_RETURN (aka. "iret")
  *   GET_CR0_INTO_EAX (aka. "movl %cr0, %eax")
- *   ENABLE_INTERRUPTS_SYSCALL_RET (aka "sti; sysexit").
+ *   ENABLE_INTERRUPTS_SYSEXIT (aka "sti; sysexit").
  *
  * For DISABLE_INTERRUPTS/ENABLE_INTERRUPTS (aka "cli"/"sti"), you must
  * specify what registers can be overwritten (CLBR_NONE, CLBR_EAX/EDX/ECX/ANY).
@@ -349,7 +349,7 @@ sysenter_past_esp:
 	xorl %ebp,%ebp
 	TRACE_IRQS_ON
 1:	mov  PT_FS(%esp), %fs
-	ENABLE_INTERRUPTS_SYSCALL_RET
+	ENABLE_INTERRUPTS_SYSEXIT
 	CFI_ENDPROC
 .pushsection .fixup,"ax"
 2:	movl $0,PT_FS(%esp)
@@ -874,10 +874,10 @@ ENTRY(native_iret)
 .previous
 END(native_iret)
 
-ENTRY(native_irq_enable_syscall_ret)
+ENTRY(native_irq_enable_sysexit)
 	sti
 	sysexit
-END(native_irq_enable_syscall_ret)
+END(native_irq_enable_sysexit)
 #endif
 
 KPROBE_ENTRY(int3)

arch/x86/kernel/entry_64.S

@@ -59,7 +59,7 @@
 #endif	
 
 #ifdef CONFIG_PARAVIRT
-ENTRY(native_irq_enable_syscall_ret)
+ENTRY(native_usersp_sysret)
 	movq	%gs:pda_oldrsp,%rsp
 	swapgs
 	sysretq
@@ -275,7 +275,7 @@ sysret_check:
 	CFI_REGISTER	rip,rcx
 	RESTORE_ARGS 0,-ARG_SKIP,1
 	/*CFI_REGISTER	rflags,r11*/
-	ENABLE_INTERRUPTS_SYSCALL_RET
+	USERSP_SYSRET
 
 	CFI_RESTORE_STATE
 	/* Handle reschedules */

arch/x86/kernel/paravirt.c

@@ -140,7 +140,8 @@ unsigned paravirt_patch_default(u8 type, u16 clobbers, void *insnbuf,
 		/* If the operation is a nop, then nop the callsite */
 		ret = paravirt_patch_nop();
 	else if (type == PARAVIRT_PATCH(pv_cpu_ops.iret) ||
-		 type == PARAVIRT_PATCH(pv_cpu_ops.irq_enable_syscall_ret))
+		 type == PARAVIRT_PATCH(pv_cpu_ops.irq_enable_sysexit) ||
+		 type == PARAVIRT_PATCH(pv_cpu_ops.usersp_sysret))
 		/* If operation requires a jmp, then jmp */
 		ret = paravirt_patch_jmp(insnbuf, opfunc, addr, len);
 	else
@@ -191,7 +192,8 @@ static void native_flush_tlb_single(unsigned long addr)
 
 /* These are in entry.S */
 extern void native_iret(void);
-extern void native_irq_enable_syscall_ret(void);
+extern void native_irq_enable_sysexit(void);
+extern void native_usersp_sysret(void);
 
 static int __init print_banner(void)
 {
@@ -327,7 +329,11 @@ struct pv_cpu_ops pv_cpu_ops = {
 	.write_idt_entry = native_write_idt_entry,
 	.load_sp0 = native_load_sp0,
 
-	.irq_enable_syscall_ret = native_irq_enable_syscall_ret,
+#ifdef CONFIG_X86_32
+	.irq_enable_sysexit = native_irq_enable_sysexit,
+#else
+	.usersp_sysret = native_usersp_sysret,
+#endif
 	.iret = native_iret,
 	.swapgs = native_swapgs,
 

arch/x86/kernel/paravirt_patch_32.c

@@ -5,7 +5,7 @@ DEF_NATIVE(pv_irq_ops, irq_enable, "sti");
 DEF_NATIVE(pv_irq_ops, restore_fl, "push %eax; popf");
 DEF_NATIVE(pv_irq_ops, save_fl, "pushf; pop %eax");
 DEF_NATIVE(pv_cpu_ops, iret, "iret");
-DEF_NATIVE(pv_cpu_ops, irq_enable_syscall_ret, "sti; sysexit");
+DEF_NATIVE(pv_cpu_ops, irq_enable_sysexit, "sti; sysexit");
 DEF_NATIVE(pv_mmu_ops, read_cr2, "mov %cr2, %eax");
 DEF_NATIVE(pv_mmu_ops, write_cr3, "mov %eax, %cr3");
 DEF_NATIVE(pv_mmu_ops, read_cr3, "mov %cr3, %eax");
@@ -29,7 +29,7 @@ unsigned native_patch(u8 type, u16 clobbers, void *ibuf,
 		PATCH_SITE(pv_irq_ops, restore_fl);
 		PATCH_SITE(pv_irq_ops, save_fl);
 		PATCH_SITE(pv_cpu_ops, iret);
-		PATCH_SITE(pv_cpu_ops, irq_enable_syscall_ret);
+		PATCH_SITE(pv_cpu_ops, irq_enable_sysexit);
 		PATCH_SITE(pv_mmu_ops, read_cr2);
 		PATCH_SITE(pv_mmu_ops, read_cr3);
 		PATCH_SITE(pv_mmu_ops, write_cr3);

arch/x86/kernel/paravirt_patch_64.c

@@ -15,7 +15,7 @@ DEF_NATIVE(pv_cpu_ops, clts, "clts");
 DEF_NATIVE(pv_cpu_ops, wbinvd, "wbinvd");
 
 /* the three commands give us more control to how to return from a syscall */
-DEF_NATIVE(pv_cpu_ops, irq_enable_syscall_ret, "movq %gs:" __stringify(pda_oldrsp) ", %rsp; swapgs; sysretq;");
+DEF_NATIVE(pv_cpu_ops, usersp_sysret, "movq %gs:" __stringify(pda_oldrsp) ", %rsp; swapgs; sysretq;");
 DEF_NATIVE(pv_cpu_ops, swapgs, "swapgs");
 
 unsigned native_patch(u8 type, u16 clobbers, void *ibuf,
@@ -35,7 +35,7 @@ unsigned native_patch(u8 type, u16 clobbers, void *ibuf,
 		PATCH_SITE(pv_irq_ops, irq_enable);
 		PATCH_SITE(pv_irq_ops, irq_disable);
 		PATCH_SITE(pv_cpu_ops, iret);
-		PATCH_SITE(pv_cpu_ops, irq_enable_syscall_ret);
+		PATCH_SITE(pv_cpu_ops, usersp_sysret);
 		PATCH_SITE(pv_cpu_ops, swapgs);
 		PATCH_SITE(pv_mmu_ops, read_cr2);
 		PATCH_SITE(pv_mmu_ops, read_cr3);

arch/x86/kernel/vmi_32.c

@@ -151,7 +151,7 @@ static unsigned vmi_patch(u8 type, u16 clobbers, void *insns,
 					      insns, ip);
 		case PARAVIRT_PATCH(pv_cpu_ops.iret):
 			return patch_internal(VMI_CALL_IRET, len, insns, ip);
-		case PARAVIRT_PATCH(pv_cpu_ops.irq_enable_syscall_ret):
+		case PARAVIRT_PATCH(pv_cpu_ops.irq_enable_sysexit):
 			return patch_internal(VMI_CALL_SYSEXIT, len, insns, ip);
 		default:
 			break;
@@ -896,7 +896,7 @@ static inline int __init activate_vmi(void)
 	 * the backend.  They are performance critical anyway, so requiring
 	 * a patch is not a big problem.
 	 */
-	pv_cpu_ops.irq_enable_syscall_ret = (void *)0xfeedbab0;
+	pv_cpu_ops.irq_enable_sysexit = (void *)0xfeedbab0;
 	pv_cpu_ops.iret = (void *)0xbadbab0;
 
 #ifdef CONFIG_SMP

arch/x86/xen/enlighten.c

@@ -1089,7 +1089,7 @@ static const struct pv_cpu_ops xen_cpu_ops __initdata = {
 	.read_pmc = native_read_pmc,
 
 	.iret = xen_iret,
-	.irq_enable_syscall_ret = xen_sysexit,
+	.irq_enable_sysexit = xen_sysexit,
 
 	.load_tr_desc = paravirt_nop,
 	.set_ldt = xen_set_ldt,

include/asm-x86/irqflags.h

@@ -112,13 +112,13 @@ static inline unsigned long __raw_local_irq_save(void)
 
 #ifdef CONFIG_X86_64
 #define INTERRUPT_RETURN	iretq
-#define ENABLE_INTERRUPTS_SYSCALL_RET			\
+#define USERSP_SYSRET					\
 			movq	%gs:pda_oldrsp, %rsp;	\
 			swapgs;				\
 			sysretq;
 #else
 #define INTERRUPT_RETURN		iret
-#define ENABLE_INTERRUPTS_SYSCALL_RET	sti; sysexit
+#define ENABLE_INTERRUPTS_SYSEXIT	sti; sysexit
 #define GET_CR0_INTO_EAX		movl %cr0, %eax
 #endif
 

include/asm-x86/paravirt.h

@@ -141,8 +141,9 @@ struct pv_cpu_ops {
 	u64 (*read_pmc)(int counter);
 	unsigned long long (*read_tscp)(unsigned int *aux);
 
-	/* These two are jmp to, not actually called. */
-	void (*irq_enable_syscall_ret)(void);
+	/* These three are jmp to, not actually called. */
+	void (*irq_enable_sysexit)(void);
+	void (*usersp_sysret)(void);
 	void (*iret)(void);
 
 	void (*swapgs)(void);
@@ -1480,10 +1481,10 @@ static inline unsigned long __raw_local_irq_save(void)
 		  call PARA_INDIRECT(pv_irq_ops+PV_IRQ_irq_enable);	\
 		  PV_RESTORE_REGS;)
 
-#define ENABLE_INTERRUPTS_SYSCALL_RET					\
-	PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_irq_enable_syscall_ret),\
+#define ENABLE_INTERRUPTS_SYSEXIT					\
+	PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_irq_enable_sysexit),	\
 		  CLBR_NONE,						\
-		  jmp PARA_INDIRECT(pv_cpu_ops+PV_CPU_irq_enable_syscall_ret))
+		  jmp PARA_INDIRECT(pv_cpu_ops+PV_CPU_irq_enable_sysexit))
 
 
 #ifdef CONFIG_X86_32
@@ -1504,6 +1505,10 @@ static inline unsigned long __raw_local_irq_save(void)
 	movq %rax, %rcx;				\
 	xorq %rax, %rax;
 
+#define USERSP_SYSRET							\
+	PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_usersp_sysret),		\
+		  CLBR_NONE,						\
+		  jmp PARA_INDIRECT(pv_cpu_ops+PV_CPU_usersp_sysret))
 #endif
 
 #endif /* __ASSEMBLY__ */