x86/paravirt: Make read_cr2() CALLEE_SAVE

The one paravirt read_cr2() implementation (Xen) is actually quite trivial
and doesn't need to clobber anything other than the return register.

Making read_cr2() CALLEE_SAVE avoids all the PUSH/POP nonsense and allows
more convenient use from assembly.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Juergen Gross <jgross@suse.com>
Cc: bp@alien8.de
Cc: rostedt@goodmis.org
Cc: luto@kernel.org
Cc: torvalds@linux-foundation.org
Cc: hpa@zytor.com
Cc: dave.hansen@linux.intel.com
Cc: zhe.he@windriver.com
Cc: joel@joelfernandes.org
Cc: devel@etsukata.com
Link: https://lkml.kernel.org/r/20190711114335.887392493@infradead.org

arch/x86/entry/calling.h

@@ -343,3 +343,9 @@ For 32-bit we have the following conventions - kernel is built with
 .Lafter_call_\@:
 #endif
 .endm
+
+#ifdef CONFIG_PARAVIRT_XXL
+#define GET_CR2_INTO(reg) GET_CR2_INTO_AX ; _ASM_MOV %_ASM_AX, reg
+#else
+#define GET_CR2_INTO(reg) _ASM_MOV %cr2, reg
+#endif

arch/x86/include/asm/paravirt.h

@@ -116,7 +116,7 @@ static inline void write_cr0(unsigned long x)
 
 static inline unsigned long read_cr2(void)
 {
-	return PVOP_CALL0(unsigned long, mmu.read_cr2);
+	return PVOP_CALLEE0(unsigned long, mmu.read_cr2);
 }
 
 static inline void write_cr2(unsigned long x)
@@ -909,13 +909,7 @@ extern void default_banner(void);
 		  ANNOTATE_RETPOLINE_SAFE;				\
 		  call PARA_INDIRECT(pv_ops+PV_CPU_swapgs);		\
 		 )
-#endif
-
-#define GET_CR2_INTO_RAX				\
-	ANNOTATE_RETPOLINE_SAFE;				\
-	call PARA_INDIRECT(pv_ops+PV_MMU_read_cr2);
 
-#ifdef CONFIG_PARAVIRT_XXL
 #define USERGS_SYSRET64							\
 	PARA_SITE(PARA_PATCH(PV_CPU_usergs_sysret64),			\
 		  ANNOTATE_RETPOLINE_SAFE;				\
@@ -929,9 +923,19 @@ extern void default_banner(void);
 		  call PARA_INDIRECT(pv_ops+PV_IRQ_save_fl);	    \
 		  PV_RESTORE_REGS(clobbers | CLBR_CALLEE_SAVE);)
 #endif
-#endif
+#endif /* CONFIG_PARAVIRT_XXL */
+#endif	/* CONFIG_X86_64 */
+
+#ifdef CONFIG_PARAVIRT_XXL
+
+#define GET_CR2_INTO_AX							\
+	PARA_SITE(PARA_PATCH(PV_MMU_read_cr2),				\
+		  ANNOTATE_RETPOLINE_SAFE;				\
+		  call PARA_INDIRECT(pv_ops+PV_MMU_read_cr2);		\
+		 )
+
+#endif /* CONFIG_PARAVIRT_XXL */
 
-#endif	/* CONFIG_X86_32 */
 
 #endif /* __ASSEMBLY__ */
 #else  /* CONFIG_PARAVIRT */

arch/x86/include/asm/paravirt_types.h

@@ -220,7 +220,7 @@ struct pv_mmu_ops {
 	void (*exit_mmap)(struct mm_struct *mm);
 
 #ifdef CONFIG_PARAVIRT_XXL
-	unsigned long (*read_cr2)(void);
+	struct paravirt_callee_save read_cr2;
 	void (*write_cr2)(unsigned long);
 
 	unsigned long (*read_cr3)(void);

arch/x86/kernel/asm-offsets.c

@@ -76,6 +76,7 @@ static void __used common(void)
 	BLANK();
 	OFFSET(XEN_vcpu_info_mask, vcpu_info, evtchn_upcall_mask);
 	OFFSET(XEN_vcpu_info_pending, vcpu_info, evtchn_upcall_pending);
+	OFFSET(XEN_vcpu_info_arch_cr2, vcpu_info, arch.cr2);
 #endif
 
 	BLANK();

arch/x86/kernel/head_64.S

@@ -29,9 +29,7 @@
 #ifdef CONFIG_PARAVIRT_XXL
 #include <asm/asm-offsets.h>
 #include <asm/paravirt.h>
-#define GET_CR2_INTO(reg) GET_CR2_INTO_RAX ; movq %rax, reg
 #else
-#define GET_CR2_INTO(reg) movq %cr2, reg
 #define INTERRUPT_RETURN iretq
 #endif
 
@@ -323,7 +321,7 @@ early_idt_handler_common:
 
 	cmpq $14,%rsi		/* Page fault? */
 	jnz 10f
-	GET_CR2_INTO(%rdi)	/* Can clobber any volatile register if pv */
+	GET_CR2_INTO(%rdi)	/* can clobber %rax if pv */
 	call early_make_pgtable
 	andl %eax,%eax
 	jz 20f			/* All good */

arch/x86/kernel/paravirt.c

@@ -370,7 +370,7 @@ struct paravirt_patch_template pv_ops = {
 	.mmu.exit_mmap		= paravirt_nop,
 
 #ifdef CONFIG_PARAVIRT_XXL
-	.mmu.read_cr2		= native_read_cr2,
+	.mmu.read_cr2		= __PV_IS_CALLEE_SAVE(native_read_cr2),
 	.mmu.write_cr2		= native_write_cr2,
 	.mmu.read_cr3		= __native_read_cr3,
 	.mmu.write_cr3		= native_write_cr3,

arch/x86/xen/enlighten_pv.c

@@ -998,7 +998,8 @@ void __init xen_setup_vcpu_info_placement(void)
 			__PV_IS_CALLEE_SAVE(xen_irq_disable_direct);
 		pv_ops.irq.irq_enable =
 			__PV_IS_CALLEE_SAVE(xen_irq_enable_direct);
-		pv_ops.mmu.read_cr2 = xen_read_cr2_direct;
+		pv_ops.mmu.read_cr2 =
+			__PV_IS_CALLEE_SAVE(xen_read_cr2_direct);
 	}
 }
 

arch/x86/xen/mmu_pv.c

@@ -1307,16 +1307,6 @@ static void xen_write_cr2(unsigned long cr2)
 	this_cpu_read(xen_vcpu)->arch.cr2 = cr2;
 }
 
-static unsigned long xen_read_cr2(void)
-{
-	return this_cpu_read(xen_vcpu)->arch.cr2;
-}
-
-unsigned long xen_read_cr2_direct(void)
-{
-	return this_cpu_read(xen_vcpu_info.arch.cr2);
-}
-
 static noinline void xen_flush_tlb(void)
 {
 	struct mmuext_op *op;
@@ -2397,7 +2387,7 @@ static void xen_leave_lazy_mmu(void)
 }
 
 static const struct pv_mmu_ops xen_mmu_ops __initconst = {
-	.read_cr2 = xen_read_cr2,
+	.read_cr2 = __PV_IS_CALLEE_SAVE(xen_read_cr2),
 	.write_cr2 = xen_write_cr2,
 
 	.read_cr3 = xen_read_cr3,

arch/x86/xen/xen-asm.S

@@ -10,6 +10,7 @@
 #include <asm/percpu.h>
 #include <asm/processor-flags.h>
 #include <asm/frame.h>
+#include <asm/asm.h>
 
 #include <linux/linkage.h>
 
@@ -135,3 +136,18 @@ ENTRY(check_events)
 	FRAME_END
 	ret
 ENDPROC(check_events)
+
+ENTRY(xen_read_cr2)
+	FRAME_BEGIN
+	_ASM_MOV PER_CPU_VAR(xen_vcpu), %_ASM_AX
+	_ASM_MOV XEN_vcpu_info_arch_cr2(%_ASM_AX), %_ASM_AX
+	FRAME_END
+	ret
+	ENDPROC(xen_read_cr2);
+
+ENTRY(xen_read_cr2_direct)
+	FRAME_BEGIN
+	_ASM_MOV PER_CPU_VAR(xen_vcpu_info) + XEN_vcpu_info_arch_cr2, %_ASM_AX
+	FRAME_END
+	ret
+	ENDPROC(xen_read_cr2_direct);

arch/x86/xen/xen-ops.h

@@ -134,6 +134,9 @@ __visible void xen_irq_disable_direct(void);
 __visible unsigned long xen_save_fl_direct(void);
 __visible void xen_restore_fl_direct(unsigned long);
 
+__visible unsigned long xen_read_cr2(void);
+__visible unsigned long xen_read_cr2_direct(void);
+
 /* These are not functions, and cannot be called normally */
 __visible void xen_iret(void);
 __visible void xen_sysret32(void);