[PATCH] i386: Convert PDA into the percpu section

Currently x86 (similar to x84-64) has a special per-cpu structure
called "i386_pda" which can be easily and efficiently referenced via
the %fs register.  An ELF section is more flexible than a structure,
allowing any piece of code to use this area.  Indeed, such a section
already exists: the per-cpu area.

So this patch:
(1) Removes the PDA and uses per-cpu variables for each current member.
(2) Replaces the __KERNEL_PDA segment with __KERNEL_PERCPU.
(3) Creates a per-cpu mirror of __per_cpu_offset called this_cpu_off, which
    can be used to calculate addresses for this CPU's variables.
(4) Simplifies startup, because %fs doesn't need to be loaded with a
    special segment at early boot; it can be deferred until the first
    percpu area is allocated (or never for UP).

The result is less code and one less x86-specific concept.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>

10 files changed, 38 insertions, 74 deletions

diff --git a/arch/i386/kernel/asm-offsets.c b/arch/i386/kernel/asm-offsets.c
index d558adfc293..b05e85fd1c1 100644
--- a/arch/i386/kernel/asm-offsets.c
+++ b/arch/i386/kernel/asm-offsets.c
@@ -15,7 +15,6 @@
 #include <asm/processor.h>
 #include <asm/thread_info.h>
 #include <asm/elf.h>
-#include <asm/pda.h>
 
 #define DEFINE(sym, val) \
         asm volatile("\n->" #sym " %0 " #val : : "i" (val))
@@ -101,10 +100,6 @@ void foo(void)
 
 	OFFSET(crypto_tfm_ctx_offset, crypto_tfm, __crt_ctx);
 
-	BLANK();
- 	OFFSET(PDA_cpu, i386_pda, cpu_number);
-	OFFSET(PDA_pcurrent, i386_pda, pcurrent);
-
 #ifdef CONFIG_PARAVIRT
 	BLANK();
 	OFFSET(PARAVIRT_enabled, paravirt_ops, paravirt_enabled);
diff --git a/arch/i386/kernel/cpu/common.c b/arch/i386/kernel/cpu/common.c
index 7a4c036d93c..27e00565f5e 100644
--- a/arch/i386/kernel/cpu/common.c
+++ b/arch/i386/kernel/cpu/common.c
@@ -18,7 +18,6 @@
 #include <asm/apic.h>
 #include <mach_apic.h>
 #endif
-#include <asm/pda.h>
 
 #include "cpu.h"
 
@@ -47,13 +46,10 @@ DEFINE_PER_CPU(struct gdt_page, gdt_page) = { .gdt = {
 	[GDT_ENTRY_APMBIOS_BASE+2] = { 0x0000ffff, 0x00409200 }, /* data */
 
 	[GDT_ENTRY_ESPFIX_SS] = { 0x00000000, 0x00c09200 },
-	[GDT_ENTRY_PDA] = { 0x00000000, 0x00c09200 }, /* set in setup_pda */
+	[GDT_ENTRY_PERCPU] = { 0x00000000, 0x00000000 },
 } };
 EXPORT_PER_CPU_SYMBOL_GPL(gdt_page);
 
-DEFINE_PER_CPU(struct i386_pda, _cpu_pda);
-EXPORT_PER_CPU_SYMBOL(_cpu_pda);
-
 static int cachesize_override __cpuinitdata = -1;
 static int disable_x86_fxsr __cpuinitdata;
 static int disable_x86_serial_nr __cpuinitdata = 1;
@@ -634,21 +630,14 @@ void __init early_cpu_init(void)
 #endif
 }
 
-/* Make sure %gs is initialized properly in idle threads */
+/* Make sure %fs is initialized properly in idle threads */
 struct pt_regs * __devinit idle_regs(struct pt_regs *regs)
 {
 	memset(regs, 0, sizeof(struct pt_regs));
-	regs->xfs = __KERNEL_PDA;
+	regs->xfs = __KERNEL_PERCPU;
 	return regs;
 }
 
-/* Initial PDA used by boot CPU */
-struct i386_pda boot_pda = {
-	._pda = &boot_pda,
-	.cpu_number = 0,
-	.pcurrent = &init_task,
-};
-
 /*
  * cpu_init() initializes state that is per-CPU. Some data is already
  * initialized (naturally) in the bootstrap process, such as the GDT
diff --git a/arch/i386/kernel/entry.S b/arch/i386/kernel/entry.S
index 3e4aa1fd33e..7f92ceb428a 100644
--- a/arch/i386/kernel/entry.S
+++ b/arch/i386/kernel/entry.S
@@ -132,7 +132,7 @@ VM_MASK		= 0x00020000
 	movl $(__USER_DS), %edx; \
 	movl %edx, %ds; \
 	movl %edx, %es; \
-	movl $(__KERNEL_PDA), %edx; \
+	movl $(__KERNEL_PERCPU), %edx; \
 	movl %edx, %fs
 
 #define RESTORE_INT_REGS \
@@ -556,7 +556,6 @@ END(syscall_badsys)
 
 #define FIXUP_ESPFIX_STACK \
 	/* since we are on a wrong stack, we cant make it a C code :( */ \
-	movl %fs:PDA_cpu, %ebx; \
 	PER_CPU(gdt_page, %ebx); \
 	GET_DESC_BASE(GDT_ENTRY_ESPFIX_SS, %ebx, %eax, %ax, %al, %ah); \
 	addl %esp, %eax; \
@@ -681,7 +680,7 @@ error_code:
 	pushl %fs
 	CFI_ADJUST_CFA_OFFSET 4
 	/*CFI_REL_OFFSET fs, 0*/
-	movl $(__KERNEL_PDA), %ecx
+	movl $(__KERNEL_PERCPU), %ecx
 	movl %ecx, %fs
 	UNWIND_ESPFIX_STACK
 	popl %ecx
diff --git a/arch/i386/kernel/head.S b/arch/i386/kernel/head.S
index bb36c24311b..12277d8938d 100644
--- a/arch/i386/kernel/head.S
+++ b/arch/i386/kernel/head.S
@@ -317,12 +317,12 @@ is386:	movl $2,%ecx		# set MP
 	movl %eax,%cr0
 
 	call check_x87
-	call setup_pda
 	lgdt early_gdt_descr
 	lidt idt_descr
 	ljmp $(__KERNEL_CS),$1f
 1:	movl $(__KERNEL_DS),%eax	# reload all the segment registers
 	movl %eax,%ss			# after changing gdt.
+	movl %eax,%fs			# gets reset once there's real percpu
 
 	movl $(__USER_DS),%eax		# DS/ES contains default USER segment
 	movl %eax,%ds
@@ -332,16 +332,17 @@ is386:	movl $2,%ecx		# set MP
 	movl %eax,%gs
 	lldt %ax
 
-	movl $(__KERNEL_PDA),%eax
-	mov  %eax,%fs
-
 	cld			# gcc2 wants the direction flag cleared at all times
 	pushl $0		# fake return address for unwinder
 #ifdef CONFIG_SMP
 	movb ready, %cl
 	movb $1, ready
 	cmpb $0,%cl		# the first CPU calls start_kernel
-	jne initialize_secondary # all other CPUs call initialize_secondary
+	je   1f
+	movl $(__KERNEL_PERCPU), %eax
+	movl %eax,%fs		# set this cpu's percpu
+	jmp initialize_secondary # all other CPUs call initialize_secondary
+1:
 #endif /* CONFIG_SMP */
 	jmp start_kernel
 
@@ -365,23 +366,6 @@ check_x87:
 	ret
 
 /*
- * Point the GDT at this CPU's PDA.  On boot this will be
- * cpu_gdt_table and boot_pda; for secondary CPUs, these will be
- * that CPU's GDT and PDA.
- */
-ENTRY(setup_pda)
-	/* get the PDA pointer */
-	movl start_pda, %eax
-
-	/* slot the PDA address into the GDT */
-	mov early_gdt_descr+2, %ecx
-	mov %ax, (__KERNEL_PDA+0+2)(%ecx)		/* base & 0x0000ffff */
-	shr $16, %eax
-	mov %al, (__KERNEL_PDA+4+0)(%ecx)		/* base & 0x00ff0000 */
-	mov %ah, (__KERNEL_PDA+4+3)(%ecx)		/* base & 0xff000000 */
-	ret
-
-/*
  *  setup_idt
  *
  *  sets up a idt with 256 entries pointing to
@@ -553,9 +537,6 @@ ENTRY(empty_zero_page)
  * This starts the data section.
  */
 .data
-ENTRY(start_pda)
-	.long boot_pda
-
 ENTRY(stack_start)
 	.long init_thread_union+THREAD_SIZE
 	.long __BOOT_DS
diff --git a/arch/i386/kernel/i386_ksyms.c b/arch/i386/kernel/i386_ksyms.c
index 4afe26e8626..e3d4b73bfdb 100644
--- a/arch/i386/kernel/i386_ksyms.c
+++ b/arch/i386/kernel/i386_ksyms.c
@@ -28,5 +28,3 @@ EXPORT_SYMBOL(__read_lock_failed);
 #endif
 
 EXPORT_SYMBOL(csum_partial);
-
-EXPORT_SYMBOL(_proxy_pda);
diff --git a/arch/i386/kernel/irq.c b/arch/i386/kernel/irq.c
index 8db8d514c9c..d2daf672f4a 100644
--- a/arch/i386/kernel/irq.c
+++ b/arch/i386/kernel/irq.c
@@ -24,6 +24,9 @@
 DEFINE_PER_CPU(irq_cpustat_t, irq_stat) ____cacheline_internodealigned_in_smp;
 EXPORT_PER_CPU_SYMBOL(irq_stat);
 
+DEFINE_PER_CPU(struct pt_regs *, irq_regs);
+EXPORT_PER_CPU_SYMBOL(irq_regs);
+
 /*
  * 'what should we do if we get a hw irq event on an illegal vector'.
  * each architecture has to answer this themselves.
diff --git a/arch/i386/kernel/process.c b/arch/i386/kernel/process.c
index 5fb9524c6f4..61999479b7a 100644
--- a/arch/i386/kernel/process.c
+++ b/arch/i386/kernel/process.c
@@ -39,6 +39,7 @@
 #include <linux/random.h>
 #include <linux/personality.h>
 #include <linux/tick.h>
+#include <linux/percpu.h>
 
 #include <asm/uaccess.h>
 #include <asm/pgtable.h>
@@ -57,7 +58,6 @@
 
 #include <asm/tlbflush.h>
 #include <asm/cpu.h>
-#include <asm/pda.h>
 
 asmlinkage void ret_from_fork(void) __asm__("ret_from_fork");
 
@@ -66,6 +66,12 @@ static int hlt_counter;
 unsigned long boot_option_idle_override = 0;
 EXPORT_SYMBOL(boot_option_idle_override);
 
+DEFINE_PER_CPU(struct task_struct *, current_task) = &init_task;
+EXPORT_PER_CPU_SYMBOL(current_task);
+
+DEFINE_PER_CPU(int, cpu_number);
+EXPORT_PER_CPU_SYMBOL(cpu_number);
+
 /*
  * Return saved PC of a blocked thread.
  */
@@ -342,7 +348,7 @@ int kernel_thread(int (*fn)(void *), void * arg, unsigned long flags)
 
 	regs.xds = __USER_DS;
 	regs.xes = __USER_DS;
-	regs.xfs = __KERNEL_PDA;
+	regs.xfs = __KERNEL_PERCPU;
 	regs.orig_eax = -1;
 	regs.eip = (unsigned long) kernel_thread_helper;
 	regs.xcs = __KERNEL_CS | get_kernel_rpl();
@@ -711,7 +717,7 @@ struct task_struct fastcall * __switch_to(struct task_struct *prev_p, struct tas
 	if (prev->gs | next->gs)
 		loadsegment(gs, next->gs);
 
-	write_pda(pcurrent, next_p);
+	x86_write_percpu(current_task, next_p);
 
 	return prev_p;
 }
diff --git a/arch/i386/kernel/smpboot.c b/arch/i386/kernel/smpboot.c
index 61e2842add3..f79b6233db7 100644
--- a/arch/i386/kernel/smpboot.c
+++ b/arch/i386/kernel/smpboot.c
@@ -53,7 +53,6 @@
 #include <asm/desc.h>
 #include <asm/arch_hooks.h>
 #include <asm/nmi.h>
-#include <asm/pda.h>
 
 #include <mach_apic.h>
 #include <mach_wakecpu.h>
@@ -99,6 +98,9 @@ EXPORT_SYMBOL(x86_cpu_to_apicid);
 
 u8 apicid_2_node[MAX_APICID];
 
+DEFINE_PER_CPU(unsigned long, this_cpu_off);
+EXPORT_PER_CPU_SYMBOL(this_cpu_off);
+
 /*
  * Trampoline 80x86 program as an array.
  */
@@ -456,7 +458,6 @@ extern struct {
 	void * esp;
 	unsigned short ss;
 } stack_start;
-extern struct i386_pda *start_pda;
 
 #ifdef CONFIG_NUMA
 
@@ -784,20 +785,17 @@ static inline struct task_struct * alloc_idle_task(int cpu)
 /* Initialize the CPU's GDT.  This is either the boot CPU doing itself
    (still using the master per-cpu area), or a CPU doing it for a
    secondary which will soon come up. */
-static __cpuinit void init_gdt(int cpu, struct task_struct *idle)
+static __cpuinit void init_gdt(int cpu)
 {
 	struct desc_struct *gdt = get_cpu_gdt_table(cpu);
-	struct i386_pda *pda = &per_cpu(_cpu_pda, cpu);
 
-	pack_descriptor((u32 *)&gdt[GDT_ENTRY_PDA].a,
-			(u32 *)&gdt[GDT_ENTRY_PDA].b,
-			(unsigned long)pda, sizeof(*pda) - 1,
-			0x80 | DESCTYPE_S | 0x2, 0); /* present read-write data segment */
+	pack_descriptor((u32 *)&gdt[GDT_ENTRY_PERCPU].a,
+			(u32 *)&gdt[GDT_ENTRY_PERCPU].b,
+			__per_cpu_offset[cpu], 0xFFFFF,
+			0x80 | DESCTYPE_S | 0x2, 0x8);
 
-	memset(pda, 0, sizeof(*pda));
-	pda->_pda = pda;
-	pda->cpu_number = cpu;
-	pda->pcurrent = idle;
+	per_cpu(this_cpu_off, cpu) = __per_cpu_offset[cpu];
+	per_cpu(cpu_number, cpu) = cpu;
 }
 
 /* Defined in head.S */
@@ -824,9 +822,9 @@ static int __cpuinit do_boot_cpu(int apicid, int cpu)
 	if (IS_ERR(idle))
 		panic("failed fork for CPU %d", cpu);
 
-	init_gdt(cpu, idle);
+	init_gdt(cpu);
+ 	per_cpu(current_task, cpu) = idle;
 	early_gdt_descr.address = (unsigned long)get_cpu_gdt_table(cpu);
-	start_pda = cpu_pda(cpu);
 
 	idle->thread.eip = (unsigned long) start_secondary;
 	/* start_eip had better be page-aligned! */
@@ -1188,14 +1186,14 @@ static inline void switch_to_new_gdt(void)
 	gdt_descr.address = (long)get_cpu_gdt_table(smp_processor_id());
 	gdt_descr.size = GDT_SIZE - 1;
 	load_gdt(&gdt_descr);
-	asm volatile ("mov %0, %%fs" : : "r" (__KERNEL_PDA) : "memory");
+	asm("mov %0, %%fs" : : "r" (__KERNEL_PERCPU) : "memory");
 }
 
 void __init native_smp_prepare_boot_cpu(void)
 {
 	unsigned int cpu = smp_processor_id();
 
-	init_gdt(cpu, current);
+	init_gdt(cpu);
 	switch_to_new_gdt();
 
 	cpu_set(cpu, cpu_online_map);
diff --git a/arch/i386/kernel/vmi.c b/arch/i386/kernel/vmi.c
index ccad7ee960a..12312988c62 100644
--- a/arch/i386/kernel/vmi.c
+++ b/arch/i386/kernel/vmi.c
@@ -504,8 +504,6 @@ static void vmi_pmd_clear(pmd_t *pmd)
 #endif
 
 #ifdef CONFIG_SMP
-extern void setup_pda(void);
-
 static void __devinit
 vmi_startup_ipi_hook(int phys_apicid, unsigned long start_eip,
 		     unsigned long start_esp)
@@ -530,13 +528,11 @@ vmi_startup_ipi_hook(int phys_apicid, unsigned long start_eip,
 
 	ap.ds = __USER_DS;
 	ap.es = __USER_DS;
-	ap.fs = __KERNEL_PDA;
+	ap.fs = __KERNEL_PERCPU;
 	ap.gs = 0;
 
 	ap.eflags = 0;
 
-	setup_pda();
-
 #ifdef CONFIG_X86_PAE
 	/* efer should match BSP efer. */
 	if (cpu_has_nx) {
diff --git a/arch/i386/kernel/vmlinux.lds.S b/arch/i386/kernel/vmlinux.lds.S
index 2ce4aa185fc..d125784ddf5 100644
--- a/arch/i386/kernel/vmlinux.lds.S
+++ b/arch/i386/kernel/vmlinux.lds.S
@@ -26,7 +26,6 @@ OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
 OUTPUT_ARCH(i386)
 ENTRY(phys_startup_32)
 jiffies = jiffies_64;
-_proxy_pda = 1;
 
 PHDRS {
 	text PT_LOAD FLAGS(5);	/* R_E */