[PATCH] paravirt: Patch inline replacements for paravirt intercepts

It turns out that the most called ops, by several orders of magnitude, are the interrupt manipulation ops. These are obvious candidates for patching, so mark them up and create infrastructure for it. The method used is that the ops structure has a patch function, which is called for each place which needs to be patched: this returns a number of instructions (the rest are NOP-padded). Usually we can spare a register (%eax) for the binary patched code to use, but in a couple of critical places in entry.S we can't: we make the clobbers explicit at the call site, and manually clobber the allowed registers in debug mode as an extra check. And: Don't abuse CONFIG_DEBUG_KERNEL, add CONFIG_DEBUG_PARAVIRT. And: AK: Fix warnings in x86-64 alternative.c build And: AK: Fix compilation with defconfig And: ^From: Andrew Morton <akpm@osdl.org> Some binutlises still like to emit references to __stop_parainstructions and __start_parainstructions. And: AK: Fix warnings about unused variables when PARAVIRT is disabled. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com> Signed-off-by: Chris Wright <chrisw@sous-sol.org> Signed-off-by: Zachary Amsden <zach@vmware.com> Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org>

[PATCH] paravirt: Patch inline replacements for paravirt intercepts
It turns out that the most called ops, by several orders of magnitude, are the interrupt manipulation ops. These are obvious candidates for patching, so mark them up and create infrastructure for it. The method used is that the ops structure has a patch function, which is called for each place which needs to be patched: this returns a number of instructions (the rest are NOP-padded). Usually we can spare a register (%eax) for the binary patched code to use, but in a couple of critical places in entry.S we can't: we make the clobbers explicit at the call site, and manually clobber the allowed registers in debug mode as an extra check. And: Don't abuse CONFIG_DEBUG_KERNEL, add CONFIG_DEBUG_PARAVIRT. And: AK: Fix warnings in x86-64 alternative.c build And: AK: Fix compilation with defconfig And: ^From: Andrew Morton <akpm@osdl.org> Some binutlises still like to emit references to __stop_parainstructions and __start_parainstructions. And: AK: Fix warnings about unused variables when PARAVIRT is disabled. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com> Signed-off-by: Chris Wright <chrisw@sous-sol.org> Signed-off-by: Zachary Amsden <zach@vmware.com> Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
Rusty Russell · Andi Kleen
1 parent d3561b7fa0
Showing 14 changed files with 459 additions and 188 deletions Side-by-side Diff
arch/i386/Kconfig.debug
arch/i386/kernel/alternative.c
arch/i386/kernel/entry.S
arch/i386/kernel/module.c
arch/i386/kernel/paravirt.c
arch/i386/kernel/vmlinux.lds.S
include/asm-i386/alternative.h
include/asm-i386/desc.h
include/asm-i386/irqflags.h
include/asm-i386/paravirt.h
include/asm-i386/processor.h
include/asm-i386/spinlock.h
include/asm-x86_64/alternative.h
scripts/mod/modpost.c
@@ -85,5 +85,15 @@
           option saves about 4k and might cause you much additional grey
           hair.
  
+config DEBUG_PARAVIRT
+	bool "Enable some paravirtualization debugging"
+	default y
+	depends on PARAVIRT && DEBUG_KERNEL
+	help
+	  Currently deliberately clobbers regs which are allowed to be
+	  clobbered in inlined paravirt hooks, even in native mode.
+	  If turning this off solves a problem, then DISABLE_INTERRUPTS() or
+	  ENABLE_INTERRUPTS() is lying about what registers can be clobbered.
+
 endmenu
@@ -124,6 +124,20 @@
  
 #endif /* CONFIG_X86_64 */
  
+static void nop_out(void *insns, unsigned int len)
+{
+	unsigned char **noptable = find_nop_table();
+
+	while (len > 0) {
+		unsigned int noplen = len;
+		if (noplen > ASM_NOP_MAX)
+			noplen = ASM_NOP_MAX;
+		memcpy(insns, noptable[noplen], noplen);
+		insns += noplen;
+		len -= noplen;
+	}
+}
+
 extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
 extern struct alt_instr __smp_alt_instructions[], __smp_alt_instructions_end[];
 extern u8 *__smp_locks[], *__smp_locks_end[];
  
@@ -138,10 +152,9 @@
  
 void apply_alternatives(struct alt_instr *start, struct alt_instr *end)
 {
-	unsigned char **noptable = find_nop_table();
 	struct alt_instr *a;
 	u8 *instr;
-	int diff, i, k;
+	int diff;
  
 	DPRINTK("%s: alt table %p -> %p\n", __FUNCTION__, start, end);
 	for (a = start; a < end; a++) {
@@ -159,13 +172,7 @@
 #endif
 		memcpy(instr, a->replacement, a->replacementlen);
 		diff = a->instrlen - a->replacementlen;
-		/* Pad the rest with nops */
-		for (i = a->replacementlen; diff > 0; diff -= k, i += k) {
-			k = diff;
-			if (k > ASM_NOP_MAX)
-				k = ASM_NOP_MAX;
-			memcpy(a->instr + i, noptable[k], k);
-		}
+		nop_out(instr + a->replacementlen, diff);
 	}
 }
  
@@ -209,7 +216,6 @@
  
 static void alternatives_smp_unlock(u8 **start, u8 **end, u8 *text, u8 *text_end)
 {
-	unsigned char **noptable = find_nop_table();
 	u8 **ptr;
  
 	for (ptr = start; ptr < end; ptr++) {
@@ -217,7 +223,7 @@
 			continue;
 		if (*ptr > text_end)
 			continue;
-		**ptr = noptable[1][0];
+		nop_out(*ptr, 1);
 	};
 }
  
@@ -343,6 +349,40 @@
  
 #endif
  
+#ifdef CONFIG_PARAVIRT
+void apply_paravirt(struct paravirt_patch *start, struct paravirt_patch *end)
+{
+	struct paravirt_patch *p;
+
+	for (p = start; p < end; p++) {
+		unsigned int used;
+
+		used = paravirt_ops.patch(p->instrtype, p->clobbers, p->instr,
+					  p->len);
+#ifdef CONFIG_DEBUG_PARAVIRT
+		{
+		int i;
+		/* Deliberately clobber regs using "not %reg" to find bugs. */
+		for (i = 0; i < 3; i++) {
+			if (p->len - used >= 2 && (p->clobbers & (1 << i))) {
+				memcpy(p->instr + used, "\xf7\xd0", 2);
+				p->instr[used+1] |= i;
+				used += 2;
+			}
+		}
+		}
+#endif
+		/* Pad the rest with nops */
+		nop_out(p->instr + used, p->len - used);
+	}
+
+	/* Sync to be conservative, in case we patched following instructions */
+	sync_core();
+}
+extern struct paravirt_patch __start_parainstructions[],
+	__stop_parainstructions[];
+#endif	/* CONFIG_PARAVIRT */
+
 void __init alternative_instructions(void)
 {
 	unsigned long flags;
@@ -390,6 +430,7 @@
 		alternatives_smp_switch(0);
 	}
 #endif
+ 	apply_paravirt(__start_parainstructions, __stop_parainstructions);
 	local_irq_restore(flags);
 }
@@ -53,6 +53,19 @@
 #include <asm/dwarf2.h>
 #include "irq_vectors.h"
  
+/*
+ * We use macros for low-level operations which need to be overridden
+ * for paravirtualization.  The following will never clobber any registers:
+ *   INTERRUPT_RETURN (aka. "iret")
+ *   GET_CR0_INTO_EAX (aka. "movl %cr0, %eax")
+ *   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).
+ * Allowing a register to be clobbered can shrink the paravirt replacement
+ * enough to patch inline, increasing performance.
+ */
+
 #define nr_syscalls ((syscall_table_size)/4)
  
 CF_MASK		= 0x00000001
  
@@ -63,9 +76,9 @@
 VM_MASK		= 0x00020000
  
 #ifdef CONFIG_PREEMPT
-#define preempt_stop		DISABLE_INTERRUPTS; TRACE_IRQS_OFF
+#define preempt_stop(clobbers)	DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF
 #else
-#define preempt_stop
+#define preempt_stop(clobbers)
 #define resume_kernel		restore_nocheck
 #endif
  
@@ -226,7 +239,7 @@
 	ALIGN
 	RING0_PTREGS_FRAME
 ret_from_exception:
-	preempt_stop
+	preempt_stop(CLBR_ANY)
 ret_from_intr:
 	GET_THREAD_INFO(%ebp)
 check_userspace:
@@ -237,7 +250,7 @@
 	jb resume_kernel		# not returning to v8086 or userspace
  
 ENTRY(resume_userspace)
- 	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
+ 	DISABLE_INTERRUPTS(CLBR_ANY)	# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	movl TI_flags(%ebp), %ecx
@@ -248,7 +261,7 @@
  
 #ifdef CONFIG_PREEMPT
 ENTRY(resume_kernel)
-	DISABLE_INTERRUPTS
+	DISABLE_INTERRUPTS(CLBR_ANY)
 	cmpl $0,TI_preempt_count(%ebp)	# non-zero preempt_count ?
 	jnz restore_nocheck
 need_resched:
@@ -277,7 +290,7 @@
 	 * No need to follow this irqs on/off section: the syscall
 	 * disabled irqs and here we enable it straight after entry:
 	 */
-	ENABLE_INTERRUPTS
+	ENABLE_INTERRUPTS(CLBR_NONE)
 	pushl $(__USER_DS)
 	CFI_ADJUST_CFA_OFFSET 4
 	/*CFI_REL_OFFSET ss, 0*/
@@ -322,7 +335,7 @@
 	jae syscall_badsys
 	call *sys_call_table(,%eax,4)
 	movl %eax,PT_EAX(%esp)
-	DISABLE_INTERRUPTS
+	DISABLE_INTERRUPTS(CLBR_ECX|CLBR_EDX)
 	TRACE_IRQS_OFF
 	movl TI_flags(%ebp), %ecx
 	testw $_TIF_ALLWORK_MASK, %cx
@@ -364,7 +377,7 @@
 	call *sys_call_table(,%eax,4)
 	movl %eax,PT_EAX(%esp)		# store the return value
 syscall_exit:
-	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
+	DISABLE_INTERRUPTS(CLBR_ANY)	# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	TRACE_IRQS_OFF
@@ -393,7 +406,7 @@
 .section .fixup,"ax"
 iret_exc:
 	TRACE_IRQS_ON
-	ENABLE_INTERRUPTS
+	ENABLE_INTERRUPTS(CLBR_NONE)
 	pushl $0			# no error code
 	pushl $do_iret_error
 	jmp error_code
@@ -436,7 +449,7 @@
 	CFI_ADJUST_CFA_OFFSET 4
 	pushl %eax
 	CFI_ADJUST_CFA_OFFSET 4
-	DISABLE_INTERRUPTS
+	DISABLE_INTERRUPTS(CLBR_EAX)
 	TRACE_IRQS_OFF
 	lss (%esp), %esp
 	CFI_ADJUST_CFA_OFFSET -8
@@ -451,7 +464,7 @@
 	jz work_notifysig
 work_resched:
 	call schedule
-	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
+	DISABLE_INTERRUPTS(CLBR_ANY)	# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	TRACE_IRQS_OFF
@@ -509,7 +522,7 @@
 	testb $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP), %cl
 	jz work_pending
 	TRACE_IRQS_ON
-	ENABLE_INTERRUPTS		# could let do_syscall_trace() call
+	ENABLE_INTERRUPTS(CLBR_ANY)	# could let do_syscall_trace() call
 					# schedule() instead
 	movl %esp, %eax
 	movl $1, %edx
@@ -693,7 +706,7 @@
 	GET_CR0_INTO_EAX
 	testl $0x4, %eax		# EM (math emulation bit)
 	jne device_not_available_emulate
-	preempt_stop
+	preempt_stop(CLBR_ANY)
 	call math_state_restore
 	jmp ret_from_exception
 device_not_available_emulate:
@@ -108,7 +108,8 @@
 		    const Elf_Shdr *sechdrs,
 		    struct module *me)
 {
-	const Elf_Shdr *s, *text = NULL, *alt = NULL, *locks = NULL;
+	const Elf_Shdr *s, *text = NULL, *alt = NULL, *locks = NULL,
+		*para = NULL;
 	char *secstrings = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset;
  
 	for (s = sechdrs; s < sechdrs + hdr->e_shnum; s++) { 
@@ -118,6 +119,8 @@
 			alt = s;
 		if (!strcmp(".smp_locks", secstrings + s->sh_name))
 			locks= s;
+		if (!strcmp(".parainstructions", secstrings + s->sh_name))
+			para = s;
 	}
  
 	if (alt) {
@@ -132,6 +135,12 @@
 					    lseg, lseg + locks->sh_size,
 					    tseg, tseg + text->sh_size);
 	}
+
+	if (para) {
+		void *pseg = (void *)para->sh_addr;
+		apply_paravirt(pseg, pseg + para->sh_size);
+	}
+
 	return 0;
 }
  
@@ -45,6 +45,49 @@
 	return paravirt_ops.memory_setup();
 }
  
+/* Simple instruction patching code. */
+#define DEF_NATIVE(name, code)					\
+	extern const char start_##name[], end_##name[];		\
+	asm("start_" #name ": " code "; end_" #name ":")
+DEF_NATIVE(cli, "cli");
+DEF_NATIVE(sti, "sti");
+DEF_NATIVE(popf, "push %eax; popf");
+DEF_NATIVE(pushf, "pushf; pop %eax");
+DEF_NATIVE(pushf_cli, "pushf; pop %eax; cli");
+DEF_NATIVE(iret, "iret");
+DEF_NATIVE(sti_sysexit, "sti; sysexit");
+
+static const struct native_insns
+{
+	const char *start, *end;
+} native_insns[] = {
+	[PARAVIRT_IRQ_DISABLE] = { start_cli, end_cli },
+	[PARAVIRT_IRQ_ENABLE] = { start_sti, end_sti },
+	[PARAVIRT_RESTORE_FLAGS] = { start_popf, end_popf },
+	[PARAVIRT_SAVE_FLAGS] = { start_pushf, end_pushf },
+	[PARAVIRT_SAVE_FLAGS_IRQ_DISABLE] = { start_pushf_cli, end_pushf_cli },
+	[PARAVIRT_INTERRUPT_RETURN] = { start_iret, end_iret },
+	[PARAVIRT_STI_SYSEXIT] = { start_sti_sysexit, end_sti_sysexit },
+};
+
+static unsigned native_patch(u8 type, u16 clobbers, void *insns, unsigned len)
+{
+	unsigned int insn_len;
+
+	/* Don't touch it if we don't have a replacement */
+	if (type >= ARRAY_SIZE(native_insns) || !native_insns[type].start)
+		return len;
+
+	insn_len = native_insns[type].end - native_insns[type].start;
+
+	/* Similarly if we can't fit replacement. */
+	if (len < insn_len)
+		return len;
+
+	memcpy(insns, native_insns[type].start, insn_len);
+	return insn_len;
+}
+
 static fastcall unsigned long native_get_debugreg(int regno)
 {
 	unsigned long val = 0; 	/* Damn you, gcc! */
@@ -349,6 +392,7 @@
 	.paravirt_enabled = 0,
 	.kernel_rpl = 0,
  
+ 	.patch = native_patch,
 	.banner = default_banner,
 	.arch_setup = native_nop,
 	.memory_setup = machine_specific_memory_setup,
@@ -165,6 +165,12 @@
   .altinstr_replacement : AT(ADDR(.altinstr_replacement) - LOAD_OFFSET) {
 	*(.altinstr_replacement)
   }
+  . = ALIGN(4);
+  __start_parainstructions = .;
+  .parainstructions : AT(ADDR(.parainstructions) - LOAD_OFFSET) {
+	*(.parainstructions)
+  }
+  __stop_parainstructions = .;
   /* .exit.text is discard at runtime, not link time, to deal with references
      from .altinstructions and .eh_frame */
   .exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) { *(.exit.text) }
@@ -4,7 +4,7 @@
 #ifdef __KERNEL__
  
 #include <asm/types.h>
-
+#include <linux/stddef.h>
 #include <linux/types.h>
  
 struct alt_instr {
@@ -116,6 +116,17 @@
  
 #else /* ! CONFIG_SMP */
 #define LOCK_PREFIX ""
+#endif
+
+struct paravirt_patch;
+#ifdef CONFIG_PARAVIRT
+void apply_paravirt(struct paravirt_patch *start, struct paravirt_patch *end);
+#else
+static inline void
+apply_paravirt(struct paravirt_patch *start, struct paravirt_patch *end)
+{}
+#define __start_parainstructions NULL
+#define __stop_parainstructions NULL
 #endif
  
 #endif /* _I386_ALTERNATIVE_H */
@@ -81,6 +81,10 @@
 #undef C
 }
  
+#define write_ldt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
+#define write_gdt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
+#define write_idt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
+
 static inline void write_dt_entry(void *dt, int entry, __u32 entry_a, __u32 entry_b)
 {
 	__u32 *lp = (__u32 *)((char *)dt + entry*8);
@@ -88,26 +92,6 @@
 	*(lp+1) = entry_b;
 }
  
-#define write_ldt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
-#define write_gdt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
-#define write_idt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
-
-static inline void _set_gate(int gate, unsigned int type, void *addr, unsigned short seg)
-{
-	__u32 a, b;
-	pack_gate(&a, &b, (unsigned long)addr, seg, type, 0);
-	write_idt_entry(idt_table, gate, a, b);
-}
-
-static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, const void *addr)
-{
-	__u32 a, b;
-	pack_descriptor(&a, &b, (unsigned long)addr,
-			offsetof(struct tss_struct, __cacheline_filler) - 1,
-			DESCTYPE_TSS, 0);
-	write_gdt_entry(get_cpu_gdt_table(cpu), entry, a, b);
-}
-
 #define set_ldt native_set_ldt
 #endif /* CONFIG_PARAVIRT */
  
@@ -127,6 +111,23 @@
 		__asm__ __volatile__("lldt %w0"::"q" (GDT_ENTRY_LDT*8));
 	}
 }
+
+static inline void _set_gate(int gate, unsigned int type, void *addr, unsigned short seg)
+{
+	__u32 a, b;
+	pack_gate(&a, &b, (unsigned long)addr, seg, type, 0);
+	write_idt_entry(idt_table, gate, a, b);
+}
+
+static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, const void *addr)
+{
+	__u32 a, b;
+	pack_descriptor(&a, &b, (unsigned long)addr,
+			offsetof(struct tss_struct, __cacheline_filler) - 1,
+			DESCTYPE_TSS, 0);
+	write_gdt_entry(get_cpu_gdt_table(cpu), entry, a, b);
+}
+
  
 #define set_tss_desc(cpu,addr) __set_tss_desc(cpu, GDT_ENTRY_TSS, addr)
  
@@ -79,8 +79,8 @@
 }
  
 #else
-#define DISABLE_INTERRUPTS		cli
-#define ENABLE_INTERRUPTS		sti
+#define DISABLE_INTERRUPTS(clobbers)	cli
+#define ENABLE_INTERRUPTS(clobbers)	sti
 #define ENABLE_INTERRUPTS_SYSEXIT	sti; sysexit
 #define INTERRUPT_RETURN		iret
 #define GET_CR0_INTO_EAX		movl %cr0, %eax
@@ -3,8 +3,26 @@
 /* Various instructions on x86 need to be replaced for
  * para-virtualization: those hooks are defined here. */
 #include <linux/linkage.h>
+#include <linux/stringify.h>
  
 #ifdef CONFIG_PARAVIRT
+/* These are the most performance critical ops, so we want to be able to patch
+ * callers */
+#define PARAVIRT_IRQ_DISABLE 0
+#define PARAVIRT_IRQ_ENABLE 1
+#define PARAVIRT_RESTORE_FLAGS 2
+#define PARAVIRT_SAVE_FLAGS 3
+#define PARAVIRT_SAVE_FLAGS_IRQ_DISABLE 4
+#define PARAVIRT_INTERRUPT_RETURN 5
+#define PARAVIRT_STI_SYSEXIT 6
+
+/* Bitmask of what can be clobbered: usually at least eax. */
+#define CLBR_NONE 0x0
+#define CLBR_EAX 0x1
+#define CLBR_ECX 0x2
+#define CLBR_EDX 0x4
+#define CLBR_ANY 0x7
+
 #ifndef __ASSEMBLY__
 struct thread_struct;
 struct Xgt_desc_struct;
@@ -15,6 +33,15 @@
  	int paravirt_enabled;
 	const char *name;
  
+	/*
+	 * Patch may replace one of the defined code sequences with arbitrary
+	 * code, subject to the same register constraints.  This generally
+	 * means the code is not free to clobber any registers other than EAX.
+	 * The patch function should return the number of bytes of code
+	 * generated, as we nop pad the rest in generic code.
+	 */
+	unsigned (*patch)(u8 type, u16 clobber, void *firstinsn, unsigned len);
+
 	void (*arch_setup)(void);
 	char *(*memory_setup)(void);
 	void (*init_IRQ)(void);
@@ -147,35 +174,6 @@
 #define read_cr4_safe(x) paravirt_ops.read_cr4_safe()
 #define write_cr4(x) paravirt_ops.write_cr4(x)
  
-static inline unsigned long __raw_local_save_flags(void)
-{
-	return paravirt_ops.save_fl();
-}
-
-static inline void raw_local_irq_restore(unsigned long flags)
-{
-	return paravirt_ops.restore_fl(flags);
-}
-
-static inline void raw_local_irq_disable(void)
-{
-	paravirt_ops.irq_disable();
-}
-
-static inline void raw_local_irq_enable(void)
-{
-	paravirt_ops.irq_enable();
-}
-
-static inline unsigned long __raw_local_irq_save(void)
-{
-	unsigned long flags = paravirt_ops.save_fl();
-
-	paravirt_ops.irq_disable();
-
-	return flags;
-}
-
 static inline void raw_safe_halt(void)
 {
 	paravirt_ops.safe_halt();
  
@@ -267,15 +265,134 @@
 #endif
 }
  
-#define CLI_STRING	"pushl %eax; pushl %ecx; pushl %edx; call *paravirt_ops+PARAVIRT_irq_disable; popl %edx; popl %ecx; popl %eax"
-#define STI_STRING	"pushl %eax; pushl %ecx; pushl %edx; call *paravirt_ops+PARAVIRT_irq_enable; popl %edx; popl %ecx; popl %eax"
+/* These all sit in the .parainstructions section to tell us what to patch. */
+struct paravirt_patch {
+	u8 *instr; 		/* original instructions */
+	u8 instrtype;		/* type of this instruction */
+	u8 len;			/* length of original instruction */
+	u16 clobbers;		/* what registers you may clobber */
+};
+
+#define paravirt_alt(insn_string, typenum, clobber)	\
+	"771:\n\t" insn_string "\n" "772:\n"		\
+	".pushsection .parainstructions,\"a\"\n"	\
+	"  .long 771b\n"				\
+	"  .byte " __stringify(typenum) "\n"		\
+	"  .byte 772b-771b\n"				\
+	"  .short " __stringify(clobber) "\n"		\
+	".popsection"
+
+static inline unsigned long __raw_local_save_flags(void)
+{
+	unsigned long f;
+
+	__asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
+					   "call *%1;"
+					   "popl %%edx; popl %%ecx",
+					  PARAVIRT_SAVE_FLAGS, CLBR_NONE)
+			     : "=a"(f): "m"(paravirt_ops.save_fl)
+			     : "memory", "cc");
+	return f;
+}
+
+static inline void raw_local_irq_restore(unsigned long f)
+{
+	__asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
+					   "call *%1;"
+					   "popl %%edx; popl %%ecx",
+					  PARAVIRT_RESTORE_FLAGS, CLBR_EAX)
+			     : "=a"(f) : "m" (paravirt_ops.restore_fl), "0"(f)
+			     : "memory", "cc");
+}
+
+static inline void raw_local_irq_disable(void)
+{
+	__asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
+					   "call *%0;"
+					   "popl %%edx; popl %%ecx",
+					  PARAVIRT_IRQ_DISABLE, CLBR_EAX)
+			     : : "m" (paravirt_ops.irq_disable)
+			     : "memory", "eax", "cc");
+}
+
+static inline void raw_local_irq_enable(void)
+{
+	__asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
+					   "call *%0;"
+					   "popl %%edx; popl %%ecx",
+					  PARAVIRT_IRQ_ENABLE, CLBR_EAX)
+			     : : "m" (paravirt_ops.irq_enable)
+			     : "memory", "eax", "cc");
+}
+
+static inline unsigned long __raw_local_irq_save(void)
+{
+	unsigned long f;
+
+	__asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
+					   "call *%1; pushl %%eax;"
+					   "call *%2; popl %%eax;"
+					   "popl %%edx; popl %%ecx",
+					  PARAVIRT_SAVE_FLAGS_IRQ_DISABLE,
+					  CLBR_NONE)
+			     : "=a"(f)
+			     : "m" (paravirt_ops.save_fl),
+			       "m" (paravirt_ops.irq_disable)
+			     : "memory", "cc");
+	return f;
+}
+
+#define CLI_STRING paravirt_alt("pushl %%ecx; pushl %%edx;"		\
+		     "call *paravirt_ops+%c[irq_disable];"		\
+		     "popl %%edx; popl %%ecx",				\
+		     PARAVIRT_IRQ_DISABLE, CLBR_EAX)
+
+#define STI_STRING paravirt_alt("pushl %%ecx; pushl %%edx;"		\
+		     "call *paravirt_ops+%c[irq_enable];"		\
+		     "popl %%edx; popl %%ecx",				\
+		     PARAVIRT_IRQ_ENABLE, CLBR_EAX)
+#define CLI_STI_CLOBBERS , "%eax"
+#define CLI_STI_INPUT_ARGS \
+	,								\
+	[irq_disable] "i" (offsetof(struct paravirt_ops, irq_disable)),	\
+	[irq_enable] "i" (offsetof(struct paravirt_ops, irq_enable))
+
 #else  /* __ASSEMBLY__ */
  
-#define INTERRUPT_RETURN	jmp *%cs:paravirt_ops+PARAVIRT_iret
-#define DISABLE_INTERRUPTS	pushl %eax; pushl %ecx; pushl %edx; call *paravirt_ops+PARAVIRT_irq_disable; popl %edx; popl %ecx; popl %eax
-#define ENABLE_INTERRUPTS	pushl %eax; pushl %ecx; pushl %edx; call *%cs:paravirt_ops+PARAVIRT_irq_enable; popl %edx; popl %ecx; popl %eax
-#define ENABLE_INTERRUPTS_SYSEXIT	jmp *%cs:paravirt_ops+PARAVIRT_irq_enable_sysexit
-#define GET_CR0_INTO_EAX	call *paravirt_ops+PARAVIRT_read_cr0
+#define PARA_PATCH(ptype, clobbers, ops)	\
+771:;						\
+	ops;					\
+772:;						\
+	.pushsection .parainstructions,"a";	\
+	 .long 771b;				\
+	 .byte ptype;				\
+	 .byte 772b-771b;			\
+	 .short clobbers;			\
+	.popsection
+
+#define INTERRUPT_RETURN				\
+	PARA_PATCH(PARAVIRT_INTERRUPT_RETURN, CLBR_ANY,	\
+	jmp *%cs:paravirt_ops+PARAVIRT_iret)
+
+#define DISABLE_INTERRUPTS(clobbers)			\
+	PARA_PATCH(PARAVIRT_IRQ_DISABLE, clobbers,	\
+	pushl %ecx; pushl %edx;				\
+	call *paravirt_ops+PARAVIRT_irq_disable;	\
+	popl %edx; popl %ecx)				\
+
+#define ENABLE_INTERRUPTS(clobbers)			\
+	PARA_PATCH(PARAVIRT_IRQ_ENABLE, clobbers,	\
+	pushl %ecx; pushl %edx;				\
+	call *%cs:paravirt_ops+PARAVIRT_irq_enable;	\
+	popl %edx; popl %ecx)
+
+#define ENABLE_INTERRUPTS_SYSEXIT			\
+	PARA_PATCH(PARAVIRT_STI_SYSEXIT, CLBR_ANY,	\
+	jmp *%cs:paravirt_ops+PARAVIRT_irq_enable_sysexit)
+
+#define GET_CR0_INTO_EAX			\
+	call *paravirt_ops+PARAVIRT_read_cr0
+
 #endif /* __ASSEMBLY__ */
 #endif /* CONFIG_PARAVIRT */
 #endif	/* __ASM_PARAVIRT_H */
@@ -156,59 +156,6 @@
 		: "0" (*eax), "2" (*ecx));
 }
  
-/*
- * Generic CPUID function
- * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx
- * resulting in stale register contents being returned.
- */
-static inline void cpuid(unsigned int op, unsigned int *eax, unsigned int *ebx, unsigned int *ecx, unsigned int *edx)
-{
-	*eax = op;
-	*ecx = 0;
-	__cpuid(eax, ebx, ecx, edx);
-}
-
-/* Some CPUID calls want 'count' to be placed in ecx */
-static inline void cpuid_count(int op, int count, int *eax, int *ebx, int *ecx,
-			       int *edx)
-{
-	*eax = op;
-	*ecx = count;
-	__cpuid(eax, ebx, ecx, edx);
-}
-
-/*
- * CPUID functions returning a single datum
- */
-static inline unsigned int cpuid_eax(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return eax;
-}
-static inline unsigned int cpuid_ebx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return ebx;
-}
-static inline unsigned int cpuid_ecx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return ecx;
-}
-static inline unsigned int cpuid_edx(unsigned int op)
-{
-	unsigned int eax, ebx, ecx, edx;
-
-	cpuid(op, &eax, &ebx, &ecx, &edx);
-	return edx;
-}
-
 #define load_cr3(pgdir) write_cr3(__pa(pgdir))
  
 /*
@@ -491,22 +438,6 @@
 	.io_bitmap	= { [ 0 ... IO_BITMAP_LONGS] = ~0 },		\
 }
  
-#ifdef CONFIG_PARAVIRT
-#include <asm/paravirt.h>
-#else
-#define paravirt_enabled() 0
-#define __cpuid native_cpuid
-
-static inline void load_esp0(struct tss_struct *tss, struct thread_struct *thread)
-{
-	tss->esp0 = thread->esp0;
-	/* This can only happen when SEP is enabled, no need to test "SEP"arately */
-	if (unlikely(tss->ss1 != thread->sysenter_cs)) {
-		tss->ss1 = thread->sysenter_cs;
-		wrmsr(MSR_IA32_SYSENTER_CS, thread->sysenter_cs, 0);
-	}
-}
-
 #define start_thread(regs, new_eip, new_esp) do {		\
 	__asm__("movl %0,%%fs": :"r" (0));			\
 	regs->xgs = 0;						\
@@ -519,36 +450,6 @@
 	regs->esp = new_esp;					\
 } while (0)
  
-/*
- * These special macros can be used to get or set a debugging register
- */
-#define get_debugreg(var, register)				\
-		__asm__("movl %%db" #register ", %0"		\
-			:"=r" (var))
-#define set_debugreg(value, register)			\
-		__asm__("movl %0,%%db" #register		\
-			: /* no output */			\
-			:"r" (value))
-
-#define set_iopl_mask native_set_iopl_mask
-#endif /* CONFIG_PARAVIRT */
-
-/*
- * Set IOPL bits in EFLAGS from given mask
- */
-static fastcall inline void native_set_iopl_mask(unsigned mask)
-{
-	unsigned int reg;
-	__asm__ __volatile__ ("pushfl;"
-			      "popl %0;"
-			      "andl %1, %0;"
-			      "orl %2, %0;"
-			      "pushl %0;"
-			      "popfl"
-				: "=&r" (reg)
-				: "i" (~X86_EFLAGS_IOPL), "r" (mask));
-}
-
 /* Forward declaration, a strange C thing */
 struct task_struct;
 struct mm_struct;
@@ -639,6 +540,105 @@
 }
  
 #define cpu_relax()	rep_nop()
+
+#ifdef CONFIG_PARAVIRT
+#include <asm/paravirt.h>
+#else
+#define paravirt_enabled() 0
+#define __cpuid native_cpuid
+
+static inline void load_esp0(struct tss_struct *tss, struct thread_struct *thread)
+{
+	tss->esp0 = thread->esp0;
+	/* This can only happen when SEP is enabled, no need to test "SEP"arately */
+	if (unlikely(tss->ss1 != thread->sysenter_cs)) {
+		tss->ss1 = thread->sysenter_cs;
+		wrmsr(MSR_IA32_SYSENTER_CS, thread->sysenter_cs, 0);
+	}
+}
+
+/*
+ * These special macros can be used to get or set a debugging register
+ */
+#define get_debugreg(var, register)				\
+		__asm__("movl %%db" #register ", %0"		\
+			:"=r" (var))
+#define set_debugreg(value, register)			\
+		__asm__("movl %0,%%db" #register		\
+			: /* no output */			\
+			:"r" (value))
+
+#define set_iopl_mask native_set_iopl_mask
+#endif /* CONFIG_PARAVIRT */
+
+/*
+ * Set IOPL bits in EFLAGS from given mask
+ */
+static fastcall inline void native_set_iopl_mask(unsigned mask)
+{
+	unsigned int reg;
+	__asm__ __volatile__ ("pushfl;"
+			      "popl %0;"
+			      "andl %1, %0;"
+			      "orl %2, %0;"
+			      "pushl %0;"
+			      "popfl"
+				: "=&r" (reg)
+				: "i" (~X86_EFLAGS_IOPL), "r" (mask));
+}
+
+/*
+ * Generic CPUID function
+ * clear %ecx since some cpus (Cyrix MII) do not set or clear %ecx
+ * resulting in stale register contents being returned.
+ */
+static inline void cpuid(unsigned int op, unsigned int *eax, unsigned int *ebx, unsigned int *ecx, unsigned int *edx)
+{
+	*eax = op;
+	*ecx = 0;
+	__cpuid(eax, ebx, ecx, edx);
+}
+
+/* Some CPUID calls want 'count' to be placed in ecx */
+static inline void cpuid_count(int op, int count, int *eax, int *ebx, int *ecx,
+			       int *edx)
+{
+	*eax = op;
+	*ecx = count;
+	__cpuid(eax, ebx, ecx, edx);
+}
+
+/*
+ * CPUID functions returning a single datum
+ */
+static inline unsigned int cpuid_eax(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return eax;
+}
+static inline unsigned int cpuid_ebx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return ebx;
+}
+static inline unsigned int cpuid_ecx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return ecx;
+}
+static inline unsigned int cpuid_edx(unsigned int op)
+{
+	unsigned int eax, ebx, ecx, edx;
+
+	cpuid(op, &eax, &ebx, &ecx, &edx);
+	return edx;
+}
  
 /* generic versions from gas */
 #define GENERIC_NOP1	".byte 0x90\n"
@@ -12,6 +12,8 @@
 #else
 #define CLI_STRING	"cli"
 #define STI_STRING	"sti"
+#define CLI_STI_CLOBBERS
+#define CLI_STI_INPUT_ARGS
 #endif /* CONFIG_PARAVIRT */
  
 /*
  
  
  
  
@@ -57,25 +59,28 @@
 {
 	asm volatile(
 		"\n1:\t"
-		LOCK_PREFIX " ; decb %0\n\t"
+		LOCK_PREFIX " ; decb %[slock]\n\t"
 		"jns 5f\n"
 		"2:\t"
-		"testl $0x200, %1\n\t"
+		"testl $0x200, %[flags]\n\t"
 		"jz 4f\n\t"
 		STI_STRING "\n"
 		"3:\t"
 		"rep;nop\n\t"
-		"cmpb $0, %0\n\t"
+		"cmpb $0, %[slock]\n\t"
 		"jle 3b\n\t"
 		CLI_STRING "\n\t"
 		"jmp 1b\n"
 		"4:\t"
 		"rep;nop\n\t"
-		"cmpb $0, %0\n\t"
+		"cmpb $0, %[slock]\n\t"
 		"jg 1b\n\t"
 		"jmp 4b\n"
 		"5:\n\t"
-		: "+m" (lock->slock) : "r" (flags) : "memory");
+		: [slock] "+m" (lock->slock)
+		: [flags] "r" (flags)
+	 	  CLI_STI_INPUT_ARGS
+		: "memory" CLI_STI_CLOBBERS);
 }
 #endif
  
@@ -4,6 +4,7 @@
 #ifdef __KERNEL__
  
 #include <linux/types.h>
+#include <linux/stddef.h>
 #include <asm/cpufeature.h>
  
 struct alt_instr {
@@ -131,6 +132,17 @@
  
 #else /* ! CONFIG_SMP */
 #define LOCK_PREFIX ""
+#endif
+
+struct paravirt_patch;
+#ifdef CONFIG_PARAVIRT
+void apply_paravirt(struct paravirt_patch *start, struct paravirt_patch *end);
+#else
+static inline void
+apply_paravirt(struct paravirt_patch *start, struct paravirt_patch *end)
+{}
+#define __start_parainstructions NULL
+#define __stop_parainstructions NULL
 #endif
  
 #endif /* _X86_64_ALTERNATIVE_H */
@@ -911,6 +911,7 @@
 		".toc1",  /* used by ppc64 */
 		".stab",
 		".rodata",
+		".parainstructions",
 		".text.lock",
 		"__bug_table", /* used by powerpc for BUG() */
 		".pci_fixup_header",
@@ -931,6 +932,7 @@
 		".altinstructions",
 		".eh_frame",
 		".debug",
+		".parainstructions",
 		NULL
 	};
 	/* part of section name */
...	...	@@ -85,5 +85,15 @@
85	85	option saves about 4k and might cause you much additional grey
86	86	hair.
87	87
	88	+config DEBUG_PARAVIRT
	89	+ bool "Enable some paravirtualization debugging"
	90	+ default y
	91	+ depends on PARAVIRT && DEBUG_KERNEL
	92	+ help
	93	+ Currently deliberately clobbers regs which are allowed to be
	94	+ clobbered in inlined paravirt hooks, even in native mode.
	95	+ If turning this off solves a problem, then DISABLE_INTERRUPTS() or
	96	+ ENABLE_INTERRUPTS() is lying about what registers can be clobbered.
	97	+
88	98	endmenu
...	...	@@ -124,6 +124,20 @@
124	124
125	125	#endif /* CONFIG_X86_64 */
126	126
	127	+static void nop_out(void *insns, unsigned int len)
	128	+{
	129	+ unsigned char **noptable = find_nop_table();
	130	+
	131	+ while (len > 0) {
	132	+ unsigned int noplen = len;
	133	+ if (noplen > ASM_NOP_MAX)
	134	+ noplen = ASM_NOP_MAX;
	135	+ memcpy(insns, noptable[noplen], noplen);
	136	+ insns += noplen;
	137	+ len -= noplen;
	138	+ }
	139	+}
	140	+
127	141	extern struct alt_instr __alt_instructions[], __alt_instructions_end[];
128	142	extern struct alt_instr __smp_alt_instructions[], __smp_alt_instructions_end[];
129	143	extern u8 __smp_locks[], __smp_locks_end[];
130	144
...	...	@@ -138,10 +152,9 @@
138	152
139	153	void apply_alternatives(struct alt_instr start, struct alt_instr end)
140	154	{
141		- unsigned char **noptable = find_nop_table();
142	155	struct alt_instr *a;
143	156	u8 *instr;
144		- int diff, i, k;
	157	+ int diff;
145	158
146	159	DPRINTK("%s: alt table %p -> %p\n", __FUNCTION__, start, end);
147	160	for (a = start; a < end; a++) {
...	...	@@ -159,13 +172,7 @@
159	172	#endif
160	173	memcpy(instr, a->replacement, a->replacementlen);
161	174	diff = a->instrlen - a->replacementlen;
162		- /* Pad the rest with nops */
163		- for (i = a->replacementlen; diff > 0; diff -= k, i += k) {
164		- k = diff;
165		- if (k > ASM_NOP_MAX)
166		- k = ASM_NOP_MAX;
167		- memcpy(a->instr + i, noptable[k], k);
168		- }
	175	+ nop_out(instr + a->replacementlen, diff);
169	176	}
170	177	}
171	178
...	...	@@ -209,7 +216,6 @@
209	216
210	217	static void alternatives_smp_unlock(u8 start, u8 end, u8 text, u8 text_end)
211	218	{
212		- unsigned char **noptable = find_nop_table();
213	219	u8 **ptr;
214	220
215	221	for (ptr = start; ptr < end; ptr++) {
...	...	@@ -217,7 +223,7 @@
217	223	continue;
218	224	if (*ptr > text_end)
219	225	continue;
220		- **ptr = noptable[1][0];
	226	+ nop_out(*ptr, 1);
221	227	};
222	228	}
223	229
...	...	@@ -343,6 +349,40 @@
343	349
344	350	#endif
345	351
	352	+#ifdef CONFIG_PARAVIRT
	353	+void apply_paravirt(struct paravirt_patch start, struct paravirt_patch end)
	354	+{
	355	+ struct paravirt_patch *p;
	356	+
	357	+ for (p = start; p < end; p++) {
	358	+ unsigned int used;
	359	+
	360	+ used = paravirt_ops.patch(p->instrtype, p->clobbers, p->instr,
	361	+ p->len);
	362	+#ifdef CONFIG_DEBUG_PARAVIRT
	363	+ {
	364	+ int i;
	365	+ /* Deliberately clobber regs using "not %reg" to find bugs. */
	366	+ for (i = 0; i < 3; i++) {
	367	+ if (p->len - used >= 2 && (p->clobbers & (1 << i))) {
	368	+ memcpy(p->instr + used, "\xf7\xd0", 2);
	369	+ p->instr[used+1] \|= i;
	370	+ used += 2;
	371	+ }
	372	+ }
	373	+ }
	374	+#endif
	375	+ /* Pad the rest with nops */
	376	+ nop_out(p->instr + used, p->len - used);
	377	+ }
	378	+
	379	+ /* Sync to be conservative, in case we patched following instructions */
	380	+ sync_core();
	381	+}
	382	+extern struct paravirt_patch __start_parainstructions[],
	383	+ __stop_parainstructions[];
	384	+#endif /* CONFIG_PARAVIRT */
	385	+
346	386	void __init alternative_instructions(void)
347	387	{
348	388	unsigned long flags;
...	...	@@ -390,6 +430,7 @@
390	430	alternatives_smp_switch(0);
391	431	}
392	432	#endif
	433	+ apply_paravirt(__start_parainstructions, __stop_parainstructions);
393	434	local_irq_restore(flags);
394	435	}
...	...	@@ -53,6 +53,19 @@
53	53	#include <asm/dwarf2.h>
54	54	#include "irq_vectors.h"
55	55
	56	+/*
	57	+ * We use macros for low-level operations which need to be overridden
	58	+ * for paravirtualization. The following will never clobber any registers:
	59	+ * INTERRUPT_RETURN (aka. "iret")
	60	+ * GET_CR0_INTO_EAX (aka. "movl %cr0, %eax")
	61	+ * ENABLE_INTERRUPTS_SYSEXIT (aka "sti; sysexit").
	62	+ *
	63	+ * For DISABLE_INTERRUPTS/ENABLE_INTERRUPTS (aka "cli"/"sti"), you must
	64	+ * specify what registers can be overwritten (CLBR_NONE, CLBR_EAX/EDX/ECX/ANY).
	65	+ * Allowing a register to be clobbered can shrink the paravirt replacement
	66	+ * enough to patch inline, increasing performance.
	67	+ */
	68	+
56	69	#define nr_syscalls ((syscall_table_size)/4)
57	70
58	71	CF_MASK = 0x00000001
59	72
...	...	@@ -63,9 +76,9 @@
63	76	VM_MASK = 0x00020000
64	77
65	78	#ifdef CONFIG_PREEMPT
66		-#define preempt_stop DISABLE_INTERRUPTS; TRACE_IRQS_OFF
	79	+#define preempt_stop(clobbers) DISABLE_INTERRUPTS(clobbers); TRACE_IRQS_OFF
67	80	#else
68		-#define preempt_stop
	81	+#define preempt_stop(clobbers)
69	82	#define resume_kernel restore_nocheck
70	83	#endif
71	84
...	...	@@ -226,7 +239,7 @@
226	239	ALIGN
227	240	RING0_PTREGS_FRAME
228	241	ret_from_exception:
229		- preempt_stop
	242	+ preempt_stop(CLBR_ANY)
230	243	ret_from_intr:
231	244	GET_THREAD_INFO(%ebp)
232	245	check_userspace:
...	...	@@ -237,7 +250,7 @@
237	250	jb resume_kernel # not returning to v8086 or userspace
238	251
239	252	ENTRY(resume_userspace)
240		- DISABLE_INTERRUPTS # make sure we don't miss an interrupt
	253	+ DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
241	254	# setting need_resched or sigpending
242	255	# between sampling and the iret
243	256	movl TI_flags(%ebp), %ecx
...	...	@@ -248,7 +261,7 @@
248	261
249	262	#ifdef CONFIG_PREEMPT
250	263	ENTRY(resume_kernel)
251		- DISABLE_INTERRUPTS
	264	+ DISABLE_INTERRUPTS(CLBR_ANY)
252	265	cmpl $0,TI_preempt_count(%ebp) # non-zero preempt_count ?
253	266	jnz restore_nocheck
254	267	need_resched:
...	...	@@ -277,7 +290,7 @@
277	290	* No need to follow this irqs on/off section: the syscall
278	291	* disabled irqs and here we enable it straight after entry:
279	292	*/
280		- ENABLE_INTERRUPTS
	293	+ ENABLE_INTERRUPTS(CLBR_NONE)
281	294	pushl $(__USER_DS)
282	295	CFI_ADJUST_CFA_OFFSET 4
283	296	/CFI_REL_OFFSET ss, 0/
...	...	@@ -322,7 +335,7 @@
322	335	jae syscall_badsys
323	336	call *sys_call_table(,%eax,4)
324	337	movl %eax,PT_EAX(%esp)
325		- DISABLE_INTERRUPTS
	338	+ DISABLE_INTERRUPTS(CLBR_ECX\|CLBR_EDX)
326	339	TRACE_IRQS_OFF
327	340	movl TI_flags(%ebp), %ecx
328	341	testw $_TIF_ALLWORK_MASK, %cx
...	...	@@ -364,7 +377,7 @@
364	377	call *sys_call_table(,%eax,4)
365	378	movl %eax,PT_EAX(%esp) # store the return value
366	379	syscall_exit:
367		- DISABLE_INTERRUPTS # make sure we don't miss an interrupt
	380	+ DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
368	381	# setting need_resched or sigpending
369	382	# between sampling and the iret
370	383	TRACE_IRQS_OFF
...	...	@@ -393,7 +406,7 @@
393	406	.section .fixup,"ax"
394	407	iret_exc:
395	408	TRACE_IRQS_ON
396		- ENABLE_INTERRUPTS
	409	+ ENABLE_INTERRUPTS(CLBR_NONE)
397	410	pushl $0 # no error code
398	411	pushl $do_iret_error
399	412	jmp error_code
...	...	@@ -436,7 +449,7 @@
436	449	CFI_ADJUST_CFA_OFFSET 4
437	450	pushl %eax
438	451	CFI_ADJUST_CFA_OFFSET 4
439		- DISABLE_INTERRUPTS
	452	+ DISABLE_INTERRUPTS(CLBR_EAX)
440	453	TRACE_IRQS_OFF
441	454	lss (%esp), %esp
442	455	CFI_ADJUST_CFA_OFFSET -8
...	...	@@ -451,7 +464,7 @@
451	464	jz work_notifysig
452	465	work_resched:
453	466	call schedule
454		- DISABLE_INTERRUPTS # make sure we don't miss an interrupt
	467	+ DISABLE_INTERRUPTS(CLBR_ANY) # make sure we don't miss an interrupt
455	468	# setting need_resched or sigpending
456	469	# between sampling and the iret
457	470	TRACE_IRQS_OFF
...	...	@@ -509,7 +522,7 @@
509	522	testb $(_TIF_SYSCALL_TRACE\|_TIF_SYSCALL_AUDIT\|_TIF_SINGLESTEP), %cl
510	523	jz work_pending
511	524	TRACE_IRQS_ON
512		- ENABLE_INTERRUPTS # could let do_syscall_trace() call
	525	+ ENABLE_INTERRUPTS(CLBR_ANY) # could let do_syscall_trace() call
513	526	# schedule() instead
514	527	movl %esp, %eax
515	528	movl $1, %edx
...	...	@@ -693,7 +706,7 @@
693	706	GET_CR0_INTO_EAX
694	707	testl $0x4, %eax # EM (math emulation bit)
695	708	jne device_not_available_emulate
696		- preempt_stop
	709	+ preempt_stop(CLBR_ANY)
697	710	call math_state_restore
698	711	jmp ret_from_exception
699	712	device_not_available_emulate:
...	...	@@ -108,7 +108,8 @@
108	108	const Elf_Shdr *sechdrs,
109	109	struct module *me)
110	110	{
111		- const Elf_Shdr s, text = NULL, alt = NULL, locks = NULL;
	111	+ const Elf_Shdr s, text = NULL, alt = NULL, locks = NULL,
	112	+ *para = NULL;
112	113	char secstrings = (void )hdr + sechdrs[hdr->e_shstrndx].sh_offset;
113	114
114	115	for (s = sechdrs; s < sechdrs + hdr->e_shnum; s++) {
...	...	@@ -118,6 +119,8 @@
118	119	alt = s;
119	120	if (!strcmp(".smp_locks", secstrings + s->sh_name))
120	121	locks= s;
	122	+ if (!strcmp(".parainstructions", secstrings + s->sh_name))
	123	+ para = s;
121	124	}
122	125
123	126	if (alt) {
...	...	@@ -132,6 +135,12 @@
132	135	lseg, lseg + locks->sh_size,
133	136	tseg, tseg + text->sh_size);
134	137	}
	138	+
	139	+ if (para) {
	140	+ void pseg = (void )para->sh_addr;
	141	+ apply_paravirt(pseg, pseg + para->sh_size);
	142	+ }
	143	+
135	144	return 0;
136	145	}
137	146
...	...	@@ -45,6 +45,49 @@
45	45	return paravirt_ops.memory_setup();
46	46	}
47	47
	48	+/* Simple instruction patching code. */
	49	+#define DEF_NATIVE(name, code) \
	50	+ extern const char start_##name[], end_##name[]; \
	51	+ asm("start_" #name ": " code "; end_" #name ":")
	52	+DEF_NATIVE(cli, "cli");
	53	+DEF_NATIVE(sti, "sti");
	54	+DEF_NATIVE(popf, "push %eax; popf");
	55	+DEF_NATIVE(pushf, "pushf; pop %eax");
	56	+DEF_NATIVE(pushf_cli, "pushf; pop %eax; cli");
	57	+DEF_NATIVE(iret, "iret");
	58	+DEF_NATIVE(sti_sysexit, "sti; sysexit");
	59	+
	60	+static const struct native_insns
	61	+{
	62	+ const char start, end;
	63	+} native_insns[] = {
	64	+ [PARAVIRT_IRQ_DISABLE] = { start_cli, end_cli },
	65	+ [PARAVIRT_IRQ_ENABLE] = { start_sti, end_sti },
	66	+ [PARAVIRT_RESTORE_FLAGS] = { start_popf, end_popf },
	67	+ [PARAVIRT_SAVE_FLAGS] = { start_pushf, end_pushf },
	68	+ [PARAVIRT_SAVE_FLAGS_IRQ_DISABLE] = { start_pushf_cli, end_pushf_cli },
	69	+ [PARAVIRT_INTERRUPT_RETURN] = { start_iret, end_iret },
	70	+ [PARAVIRT_STI_SYSEXIT] = { start_sti_sysexit, end_sti_sysexit },
	71	+};
	72	+
	73	+static unsigned native_patch(u8 type, u16 clobbers, void *insns, unsigned len)
	74	+{
	75	+ unsigned int insn_len;
	76	+
	77	+ /* Don't touch it if we don't have a replacement */
	78	+ if (type >= ARRAY_SIZE(native_insns) \|\| !native_insns[type].start)
	79	+ return len;
	80	+
	81	+ insn_len = native_insns[type].end - native_insns[type].start;
	82	+
	83	+ /* Similarly if we can't fit replacement. */
	84	+ if (len < insn_len)
	85	+ return len;
	86	+
	87	+ memcpy(insns, native_insns[type].start, insn_len);
	88	+ return insn_len;
	89	+}
	90	+
48	91	static fastcall unsigned long native_get_debugreg(int regno)
49	92	{
50	93	unsigned long val = 0; /* Damn you, gcc! */
...	...	@@ -349,6 +392,7 @@
349	392	.paravirt_enabled = 0,
350	393	.kernel_rpl = 0,
351	394
	395	+ .patch = native_patch,
352	396	.banner = default_banner,
353	397	.arch_setup = native_nop,
354	398	.memory_setup = machine_specific_memory_setup,
...	...	@@ -165,6 +165,12 @@
165	165	.altinstr_replacement : AT(ADDR(.altinstr_replacement) - LOAD_OFFSET) {
166	166	*(.altinstr_replacement)
167	167	}
	168	+ . = ALIGN(4);
	169	+ __start_parainstructions = .;
	170	+ .parainstructions : AT(ADDR(.parainstructions) - LOAD_OFFSET) {
	171	+ *(.parainstructions)
	172	+ }
	173	+ __stop_parainstructions = .;
168	174	/* .exit.text is discard at runtime, not link time, to deal with references
169	175	from .altinstructions and .eh_frame */
170	176	.exit.text : AT(ADDR(.exit.text) - LOAD_OFFSET) { *(.exit.text) }
...	...	@@ -4,7 +4,7 @@
4	4	#ifdef __KERNEL__
5	5
6	6	#include <asm/types.h>
7		-
	7	+#include <linux/stddef.h>
8	8	#include <linux/types.h>
9	9
10	10	struct alt_instr {
...	...	@@ -116,6 +116,17 @@
116	116
117	117	#else /* ! CONFIG_SMP */
118	118	#define LOCK_PREFIX ""
	119	+#endif
	120	+
	121	+struct paravirt_patch;
	122	+#ifdef CONFIG_PARAVIRT
	123	+void apply_paravirt(struct paravirt_patch start, struct paravirt_patch end);
	124	+#else
	125	+static inline void
	126	+apply_paravirt(struct paravirt_patch start, struct paravirt_patch end)
	127	+{}
	128	+#define __start_parainstructions NULL
	129	+#define __stop_parainstructions NULL
119	130	#endif
120	131
121	132	#endif /* _I386_ALTERNATIVE_H */
...	...	@@ -81,6 +81,10 @@
81	81	#undef C
82	82	}
83	83
	84	+#define write_ldt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
	85	+#define write_gdt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
	86	+#define write_idt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
	87	+
84	88	static inline void write_dt_entry(void *dt, int entry, __u32 entry_a, __u32 entry_b)
85	89	{
86	90	__u32 lp = (__u32 )((char )dt + entry8);
...	...	@@ -88,26 +92,6 @@
88	92	*(lp+1) = entry_b;
89	93	}
90	94
91		-#define write_ldt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
92		-#define write_gdt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
93		-#define write_idt_entry(dt, entry, a, b) write_dt_entry(dt, entry, a, b)
94		-
95		-static inline void _set_gate(int gate, unsigned int type, void *addr, unsigned short seg)
96		-{
97		- __u32 a, b;
98		- pack_gate(&a, &b, (unsigned long)addr, seg, type, 0);
99		- write_idt_entry(idt_table, gate, a, b);
100		-}
101		-
102		-static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, const void *addr)
103		-{
104		- __u32 a, b;
105		- pack_descriptor(&a, &b, (unsigned long)addr,
106		- offsetof(struct tss_struct, __cacheline_filler) - 1,
107		- DESCTYPE_TSS, 0);
108		- write_gdt_entry(get_cpu_gdt_table(cpu), entry, a, b);
109		-}
110		-
111	95	#define set_ldt native_set_ldt
112	96	#endif /* CONFIG_PARAVIRT */
113	97
...	...	@@ -127,6 +111,23 @@
127	111	__asm__ __volatile__("lldt %w0"::"q" (GDT_ENTRY_LDT*8));
128	112	}
129	113	}
	114	+
	115	+static inline void _set_gate(int gate, unsigned int type, void *addr, unsigned short seg)
	116	+{
	117	+ __u32 a, b;
	118	+ pack_gate(&a, &b, (unsigned long)addr, seg, type, 0);
	119	+ write_idt_entry(idt_table, gate, a, b);
	120	+}
	121	+
	122	+static inline void __set_tss_desc(unsigned int cpu, unsigned int entry, const void *addr)
	123	+{
	124	+ __u32 a, b;
	125	+ pack_descriptor(&a, &b, (unsigned long)addr,
	126	+ offsetof(struct tss_struct, __cacheline_filler) - 1,
	127	+ DESCTYPE_TSS, 0);
	128	+ write_gdt_entry(get_cpu_gdt_table(cpu), entry, a, b);
	129	+}
	130	+
130	131
131	132	#define set_tss_desc(cpu,addr) __set_tss_desc(cpu, GDT_ENTRY_TSS, addr)
132	133
...	...	@@ -79,8 +79,8 @@
79	79	}
80	80
81	81	#else
82		-#define DISABLE_INTERRUPTS cli
83		-#define ENABLE_INTERRUPTS sti
	82	+#define DISABLE_INTERRUPTS(clobbers) cli
	83	+#define ENABLE_INTERRUPTS(clobbers) sti
84	84	#define ENABLE_INTERRUPTS_SYSEXIT sti; sysexit
85	85	#define INTERRUPT_RETURN iret
86	86	#define GET_CR0_INTO_EAX movl %cr0, %eax
...	...	@@ -3,8 +3,26 @@
3	3	/* Various instructions on x86 need to be replaced for
4	4	* para-virtualization: those hooks are defined here. */
5	5	#include <linux/linkage.h>
	6	+#include <linux/stringify.h>
6	7
7	8	#ifdef CONFIG_PARAVIRT
	9	+/* These are the most performance critical ops, so we want to be able to patch
	10	+ * callers */
	11	+#define PARAVIRT_IRQ_DISABLE 0
	12	+#define PARAVIRT_IRQ_ENABLE 1
	13	+#define PARAVIRT_RESTORE_FLAGS 2
	14	+#define PARAVIRT_SAVE_FLAGS 3
	15	+#define PARAVIRT_SAVE_FLAGS_IRQ_DISABLE 4
	16	+#define PARAVIRT_INTERRUPT_RETURN 5
	17	+#define PARAVIRT_STI_SYSEXIT 6
	18	+
	19	+/* Bitmask of what can be clobbered: usually at least eax. */
	20	+#define CLBR_NONE 0x0
	21	+#define CLBR_EAX 0x1
	22	+#define CLBR_ECX 0x2
	23	+#define CLBR_EDX 0x4
	24	+#define CLBR_ANY 0x7
	25	+
8	26	#ifndef __ASSEMBLY__
9	27	struct thread_struct;
10	28	struct Xgt_desc_struct;
...	...	@@ -15,6 +33,15 @@
15	33	int paravirt_enabled;
16	34	const char *name;
17	35
	36	+ /*
	37	+ * Patch may replace one of the defined code sequences with arbitrary
	38	+ * code, subject to the same register constraints. This generally
	39	+ * means the code is not free to clobber any registers other than EAX.
	40	+ * The patch function should return the number of bytes of code
	41	+ * generated, as we nop pad the rest in generic code.
	42	+ */
	43	+ unsigned (patch)(u8 type, u16 clobber, void firstinsn, unsigned len);
	44	+
18	45	void (*arch_setup)(void);
19	46	char (memory_setup)(void);
20	47	void (*init_IRQ)(void);
...	...	@@ -147,35 +174,6 @@
147	174	#define read_cr4_safe(x) paravirt_ops.read_cr4_safe()
148	175	#define write_cr4(x) paravirt_ops.write_cr4(x)
149	176
150		-static inline unsigned long __raw_local_save_flags(void)
151		-{
152		- return paravirt_ops.save_fl();
153		-}
154		-
155		-static inline void raw_local_irq_restore(unsigned long flags)
156		-{
157		- return paravirt_ops.restore_fl(flags);
158		-}
159		-
160		-static inline void raw_local_irq_disable(void)
161		-{
162		- paravirt_ops.irq_disable();
163		-}
164		-
165		-static inline void raw_local_irq_enable(void)
166		-{
167		- paravirt_ops.irq_enable();
168		-}
169		-
170		-static inline unsigned long __raw_local_irq_save(void)
171		-{
172		- unsigned long flags = paravirt_ops.save_fl();
173		-
174		- paravirt_ops.irq_disable();
175		-
176		- return flags;
177		-}
178		-
179	177	static inline void raw_safe_halt(void)
180	178	{
181	179	paravirt_ops.safe_halt();
182	180
...	...	@@ -267,15 +265,134 @@
267	265	#endif
268	266	}
269	267
270		-#define CLI_STRING "pushl %eax; pushl %ecx; pushl %edx; call *paravirt_ops+PARAVIRT_irq_disable; popl %edx; popl %ecx; popl %eax"
271		-#define STI_STRING "pushl %eax; pushl %ecx; pushl %edx; call *paravirt_ops+PARAVIRT_irq_enable; popl %edx; popl %ecx; popl %eax"
	268	+/* These all sit in the .parainstructions section to tell us what to patch. */
	269	+struct paravirt_patch {
	270	+ u8 instr; / original instructions */
	271	+ u8 instrtype; /* type of this instruction */
	272	+ u8 len; /* length of original instruction */
	273	+ u16 clobbers; /* what registers you may clobber */
	274	+};
	275	+
	276	+#define paravirt_alt(insn_string, typenum, clobber) \
	277	+ "771:\n\t" insn_string "\n" "772:\n" \
	278	+ ".pushsection .parainstructions,\"a\"\n" \
	279	+ " .long 771b\n" \
	280	+ " .byte " __stringify(typenum) "\n" \
	281	+ " .byte 772b-771b\n" \
	282	+ " .short " __stringify(clobber) "\n" \
	283	+ ".popsection"
	284	+
	285	+static inline unsigned long __raw_local_save_flags(void)
	286	+{
	287	+ unsigned long f;
	288	+
	289	+ __asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
	290	+ "call *%1;"
	291	+ "popl %%edx; popl %%ecx",
	292	+ PARAVIRT_SAVE_FLAGS, CLBR_NONE)
	293	+ : "=a"(f): "m"(paravirt_ops.save_fl)
	294	+ : "memory", "cc");
	295	+ return f;
	296	+}
	297	+
	298	+static inline void raw_local_irq_restore(unsigned long f)
	299	+{
	300	+ __asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
	301	+ "call *%1;"
	302	+ "popl %%edx; popl %%ecx",
	303	+ PARAVIRT_RESTORE_FLAGS, CLBR_EAX)
	304	+ : "=a"(f) : "m" (paravirt_ops.restore_fl), "0"(f)
	305	+ : "memory", "cc");
	306	+}
	307	+
	308	+static inline void raw_local_irq_disable(void)
	309	+{
	310	+ __asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
	311	+ "call *%0;"
	312	+ "popl %%edx; popl %%ecx",
	313	+ PARAVIRT_IRQ_DISABLE, CLBR_EAX)
	314	+ : : "m" (paravirt_ops.irq_disable)
	315	+ : "memory", "eax", "cc");
	316	+}
	317	+
	318	+static inline void raw_local_irq_enable(void)
	319	+{
	320	+ __asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
	321	+ "call *%0;"
	322	+ "popl %%edx; popl %%ecx",
	323	+ PARAVIRT_IRQ_ENABLE, CLBR_EAX)
	324	+ : : "m" (paravirt_ops.irq_enable)
	325	+ : "memory", "eax", "cc");
	326	+}
	327	+
	328	+static inline unsigned long __raw_local_irq_save(void)
	329	+{
	330	+ unsigned long f;
	331	+
	332	+ __asm__ __volatile__(paravirt_alt( "pushl %%ecx; pushl %%edx;"
	333	+ "call *%1; pushl %%eax;"
	334	+ "call *%2; popl %%eax;"
	335	+ "popl %%edx; popl %%ecx",
	336	+ PARAVIRT_SAVE_FLAGS_IRQ_DISABLE,
	337	+ CLBR_NONE)
	338	+ : "=a"(f)
	339	+ : "m" (paravirt_ops.save_fl),
	340	+ "m" (paravirt_ops.irq_disable)
	341	+ : "memory", "cc");
	342	+ return f;
	343	+}
	344	+
	345	+#define CLI_STRING paravirt_alt("pushl %%ecx; pushl %%edx;" \
	346	+ "call *paravirt_ops+%c[irq_disable];" \
	347	+ "popl %%edx; popl %%ecx", \
	348	+ PARAVIRT_IRQ_DISABLE, CLBR_EAX)
	349	+
	350	+#define STI_STRING paravirt_alt("pushl %%ecx; pushl %%edx;" \
	351	+ "call *paravirt_ops+%c[irq_enable];" \
	352	+ "popl %%edx; popl %%ecx", \
	353	+ PARAVIRT_IRQ_ENABLE, CLBR_EAX)
	354	+#define CLI_STI_CLOBBERS , "%eax"
	355	+#define CLI_STI_INPUT_ARGS \
	356	+ , \
	357	+ [irq_disable] "i" (offsetof(struct paravirt_ops, irq_disable)), \
	358	+ [irq_enable] "i" (offsetof(struct paravirt_ops, irq_enable))
	359	+
272	360	#else /* __ASSEMBLY__ */
273	361
274		-#define INTERRUPT_RETURN jmp *%cs:paravirt_ops+PARAVIRT_iret
275		-#define DISABLE_INTERRUPTS pushl %eax; pushl %ecx; pushl %edx; call *paravirt_ops+PARAVIRT_irq_disable; popl %edx; popl %ecx; popl %eax
276		-#define ENABLE_INTERRUPTS pushl %eax; pushl %ecx; pushl %edx; call *%cs:paravirt_ops+PARAVIRT_irq_enable; popl %edx; popl %ecx; popl %eax
277		-#define ENABLE_INTERRUPTS_SYSEXIT jmp *%cs:paravirt_ops+PARAVIRT_irq_enable_sysexit
278		-#define GET_CR0_INTO_EAX call *paravirt_ops+PARAVIRT_read_cr0
	362	+#define PARA_PATCH(ptype, clobbers, ops) \
	363	+771:; \
	364	+ ops; \
	365	+772:; \
	366	+ .pushsection .parainstructions,"a"; \
	367	+ .long 771b; \
	368	+ .byte ptype; \
	369	+ .byte 772b-771b; \
	370	+ .short clobbers; \
	371	+ .popsection
	372	+
	373	+#define INTERRUPT_RETURN \
	374	+ PARA_PATCH(PARAVIRT_INTERRUPT_RETURN, CLBR_ANY, \
	375	+ jmp *%cs:paravirt_ops+PARAVIRT_iret)
	376	+
	377	+#define DISABLE_INTERRUPTS(clobbers) \
	378	+ PARA_PATCH(PARAVIRT_IRQ_DISABLE, clobbers, \
	379	+ pushl %ecx; pushl %edx; \
	380	+ call *paravirt_ops+PARAVIRT_irq_disable; \
	381	+ popl %edx; popl %ecx) \
	382	+
	383	+#define ENABLE_INTERRUPTS(clobbers) \
	384	+ PARA_PATCH(PARAVIRT_IRQ_ENABLE, clobbers, \
	385	+ pushl %ecx; pushl %edx; \
	386	+ call *%cs:paravirt_ops+PARAVIRT_irq_enable; \
	387	+ popl %edx; popl %ecx)
	388	+
	389	+#define ENABLE_INTERRUPTS_SYSEXIT \
	390	+ PARA_PATCH(PARAVIRT_STI_SYSEXIT, CLBR_ANY, \
	391	+ jmp *%cs:paravirt_ops+PARAVIRT_irq_enable_sysexit)
	392	+
	393	+#define GET_CR0_INTO_EAX \
	394	+ call *paravirt_ops+PARAVIRT_read_cr0
	395	+
279	396	#endif /* __ASSEMBLY__ */
280	397	#endif /* CONFIG_PARAVIRT */
281	398	#endif /* __ASM_PARAVIRT_H */