[PATCH 14/30] cr: x86_64 support

From: Alexey Dobriyan
Date: Thu Apr 09 2009 - 22:40:00 EST


Now x86 matrix of migration is:

task/kernel kernel
------------------------------
i386/i386 => i386
i386/i386 => x86_64
i386/x86_64 => i386
i386/x86_64 => x86_64
x86_64/x86_64 => x86_64

FIXME: support more that busyloop

Signed-off-by: Alexey Dobriyan <adobriyan@xxxxxxxxx>
---

arch/x86/ia32/ia32entry.S | 2
arch/x86/include/asm/unistd_64.h | 4
include/linux/cr.h | 41 ++++
kernel/cr/Makefile | 1
kernel/cr/cr-x86_64.c | 365 +++++++++++++++++++++++++++++++++++++++
kernel/cr/cr.h | 2
6 files changed, 414 insertions(+), 1 deletion(-)

--- a/arch/x86/ia32/ia32entry.S
+++ b/arch/x86/ia32/ia32entry.S
@@ -830,4 +830,6 @@ ia32_sys_call_table:
.quad sys_inotify_init1
.quad compat_sys_preadv
.quad compat_sys_pwritev
+ .quad sys_checkpoint /* 335 */
+ .quad sys_restart
ia32_syscall_end:
--- a/arch/x86/include/asm/unistd_64.h
+++ b/arch/x86/include/asm/unistd_64.h
@@ -657,6 +657,10 @@ __SYSCALL(__NR_inotify_init1, sys_inotify_init1)
__SYSCALL(__NR_preadv, sys_preadv)
#define __NR_pwritev 296
__SYSCALL(__NR_pwritev, sys_pwritev)
+#define __NR_checkpoint 297
+__SYSCALL(__NR_checkpoint, sys_checkpoint)
+#define __NR_restart 298
+__SYSCALL(__NR_restart, sys_restart)


#ifndef __NO_STUBS
--- a/include/linux/cr.h
+++ b/include/linux/cr.h
@@ -18,6 +18,7 @@ struct cr_image_header {
/* Mutable part. */
/* Arch of the kernel which dumped the image. */
#define CR_ARCH_X86_32 1
+#define CR_ARCH_X86_64 2
__le32 cr_arch;
/*
* Distributions are expected to leave image version alone and
@@ -101,6 +102,46 @@ struct cr_image_arch_x86_32 {
/* __u8 cr_xstate[cr_len_xstate]; */
} __packed;

+struct cr_image_arch_x86_64 {
+ __u64 cr_r15;
+ __u64 cr_r14;
+ __u64 cr_r13;
+ __u64 cr_r12;
+ __u64 cr_rbp;
+ __u64 cr_rbx;
+ __u64 cr_r11;
+ __u64 cr_r10;
+ __u64 cr_r9;
+ __u64 cr_r8;
+ __u64 cr_rax;
+ __u64 cr_rcx;
+ __u64 cr_rdx;
+ __u64 cr_rsi;
+ __u64 cr_rdi;
+ __u64 cr_orig_rax;
+ __u64 cr_rip;
+ __u64 cr_rflags;
+ __u64 cr_rsp;
+
+ __u64 cr_fs;
+ __u64 cr_gs;
+ __u16 cr_cs;
+ __u16 cr_ds;
+ __u16 cr_es;
+ __u16 cr_fsindex;
+ __u16 cr_gsindex;
+ __u16 cr_ss;
+
+ __u64 cr_dr0;
+ __u64 cr_dr1;
+ __u64 cr_dr2;
+ __u64 cr_dr3;
+ __u64 cr_dr6;
+ __u64 cr_dr7;
+
+ __u64 cr_tls_array[3];
+} __packed;
+
struct cr_image_mm_struct {
struct cr_object_header cr_hdr;

--- a/kernel/cr/Makefile
+++ b/kernel/cr/Makefile
@@ -5,3 +5,4 @@ cr-y += cr-file.o
cr-y += cr-mm.o
cr-y += cr-task.o
cr-$(CONFIG_X86_32) += cr-x86_32.o
+cr-$(CONFIG_X86_64) += cr-x86_64.o
new file mode 100644
--- /dev/null
+++ b/kernel/cr/cr-x86_64.c
@@ -0,0 +1,365 @@
+/* Copyright (C) 2000-2009 Parallels Holdings, Ltd. */
+#include <linux/sched.h>
+#include <asm/i387.h>
+
+#include <linux/cr.h>
+#include "cr.h"
+
+__u32 cr_image_header_arch(void)
+{
+ return CR_ARCH_X86_64;
+}
+
+int cr_arch_check_image_header(struct cr_image_header *i)
+{
+ if (i->cr_arch == cpu_to_le32(CR_ARCH_X86_64))
+ return 0;
+#ifdef CONFIG_COMPAT
+ if (i->cr_arch == cpu_to_le32(CR_ARCH_X86_32))
+ return 0;
+#endif
+ return -EINVAL;
+}
+
+__u32 cr_task_struct_arch(struct task_struct *tsk)
+{
+#ifdef CONFIG_COMPAT
+ if (test_tsk_thread_flag(tsk, TIF_IA32))
+ return CR_ARCH_X86_32;
+#endif
+ return CR_ARCH_X86_64;
+}
+
+int cr_arch_check_image_task_struct(struct cr_image_task_struct *i)
+{
+ if (i->cr_tsk_arch == CR_ARCH_X86_64)
+ return 0;
+#ifdef CONFIG_COMPAT
+ if (i->cr_tsk_arch == CR_ARCH_X86_32)
+ return 0;
+#endif
+ return -EINVAL;
+}
+
+unsigned int cr_arch_len_task_struct(struct task_struct *tsk)
+{
+ unsigned int len;
+
+ len = sizeof(struct cr_image_arch_x86_64);
+#ifdef CONFIG_COMPAT
+ if (test_tsk_thread_flag(tsk, TIF_IA32))
+ len = sizeof(struct cr_image_arch_x86_32);
+#endif
+ return len;
+}
+
+int cr_arch_check_task_struct(struct task_struct *tsk)
+{
+ if (tsk->thread.xstate) {
+ WARN_ON(1);
+ return -EINVAL;
+ }
+ if (tsk->thread.io_bitmap_ptr) {
+ WARN_ON(1);
+ return -EINVAL;
+ }
+#ifdef CONFIG_X86_DS
+ if (tsk->thread.ds_ctx) {
+ WARN_ON(1);
+ return -EINVAL;
+ }
+#endif
+ return 0;
+}
+
+static __u16 encode_segment(u16 reg)
+{
+ if (reg == 0)
+ return CR_SEG_NULL;
+ BUG_ON((reg & 3) != 3);
+ if (reg & 4)
+ return CR_SEG_LDT | (reg >> 3);
+
+ if (reg == __USER_CS)
+ return CR_SEG_USER64_CS;
+ if (reg == __USER_DS)
+ return CR_SEG_USER64_DS;
+#ifdef CONFIG_COMPAT
+ if (reg == __USER32_CS)
+ return CR_SEG_USER32_CS;
+ if (reg == __USER32_DS)
+ return CR_SEG_USER32_DS;
+#endif
+
+ if (GDT_ENTRY_TLS_MIN <= (reg >> 3) && (reg >> 3) <= GDT_ENTRY_TLS_MAX)
+ return CR_SEG_TLS | ((reg >> 3) - GDT_ENTRY_TLS_MIN);
+ BUG();
+}
+
+static u16 decode_segment(__u16 reg)
+{
+ if (reg == CR_SEG_NULL)
+ return 0;
+
+ if (reg == CR_SEG_USER64_CS)
+ return __USER_CS;
+ if (reg == CR_SEG_USER64_DS)
+ return __USER_DS;
+#ifdef CONFIG_COMPAT
+ if (reg == CR_SEG_USER32_CS)
+ return __USER32_CS;
+ if (reg == CR_SEG_USER32_DS)
+ return __USER32_DS;
+#endif
+
+ BUILD_BUG_ON(GDT_ENTRY_TLS_MAX - GDT_ENTRY_TLS_MIN + 1 != 3);
+ if ((reg & CR_SEG_TLS) == CR_SEG_TLS) {
+ reg &= ~CR_SEG_TLS;
+ if (reg <= GDT_ENTRY_TLS_MAX - GDT_ENTRY_TLS_MIN)
+ return ((GDT_ENTRY_TLS_MIN + reg) << 3) | 3;
+ }
+ if ((reg & CR_SEG_LDT) == CR_SEG_LDT) {
+ reg &= ~CR_SEG_LDT;
+ return (reg << 3) | 7;
+ }
+ BUG();
+}
+
+#ifdef CONFIG_COMPAT
+static int cr_dump_task_struct_x86_32(struct cr_context *ctx, struct task_struct *tsk)
+{
+ struct cr_image_arch_x86_32 *i;
+ struct pt_regs *regs = task_pt_regs(tsk);
+ int rv;
+
+ i = kzalloc(sizeof(*i), GFP_KERNEL);
+ if (!i)
+ return -ENOMEM;
+
+ i->cr_ebx = regs->bx;
+ i->cr_ecx = regs->cx;
+ i->cr_edx = regs->dx;
+ i->cr_esi = regs->si;
+ i->cr_edi = regs->di;
+ i->cr_ebp = regs->bp;
+ i->cr_eax = regs->ax;
+ i->cr_orig_eax = regs->orig_ax;
+ i->cr_eip = regs->ip;
+ i->cr_eflags = regs->flags;
+ i->cr_esp = regs->sp;
+
+ i->cr_cs = encode_segment(regs->cs);
+ i->cr_ds = encode_segment(tsk->thread.ds);
+ i->cr_es = encode_segment(tsk->thread.es);
+ i->cr_fs = encode_segment(tsk->thread.fsindex);
+ i->cr_gs = encode_segment(tsk->thread.gsindex);
+ i->cr_ss = encode_segment(regs->ss);
+
+ i->cr_dr0 = tsk->thread.debugreg0;
+ i->cr_dr1 = tsk->thread.debugreg1;
+ i->cr_dr2 = tsk->thread.debugreg2;
+ i->cr_dr3 = tsk->thread.debugreg3;
+ i->cr_dr6 = tsk->thread.debugreg6;
+ i->cr_dr7 = tsk->thread.debugreg7;
+
+ BUILD_BUG_ON(sizeof(tsk->thread.tls_array[0]) != 8);
+ BUILD_BUG_ON(sizeof(tsk->thread.tls_array) != 3 * 8);
+ memcpy(i->cr_tls_array, tsk->thread.tls_array, sizeof(i->cr_tls_array));
+
+ rv = cr_write(ctx, i, sizeof(*i));
+ kfree(i);
+ return rv;
+}
+#endif
+
+static int cr_dump_task_struct_x86_64(struct cr_context *ctx, struct task_struct *tsk)
+{
+ struct cr_image_arch_x86_64 *i;
+ struct pt_regs *regs = task_pt_regs(tsk);
+ int rv;
+
+ i = kzalloc(sizeof(*i), GFP_KERNEL);
+ if (!i)
+ return -ENOMEM;
+
+ i->cr_r15 = regs->r15;
+ i->cr_r14 = regs->r14;
+ i->cr_r13 = regs->r13;
+ i->cr_r12 = regs->r12;
+ i->cr_rbp = regs->bp;
+ i->cr_rbx = regs->bx;
+ i->cr_r11 = regs->r11;
+ i->cr_r10 = regs->r10;
+ i->cr_r9 = regs->r9;
+ i->cr_r8 = regs->r8;
+ i->cr_rax = regs->ax;
+ i->cr_rcx = regs->cx;
+ i->cr_rdx = regs->dx;
+ i->cr_rsi = regs->si;
+ i->cr_rdi = regs->di;
+ i->cr_orig_rax = regs->orig_ax;
+ i->cr_rip = regs->ip;
+ i->cr_rflags = regs->flags;
+ i->cr_rsp = regs->sp;
+
+ i->cr_fs = tsk->thread.fs;
+ i->cr_gs = tsk->thread.gs;
+ i->cr_cs = encode_segment(regs->cs);
+ i->cr_ds = encode_segment(tsk->thread.ds);
+ i->cr_es = encode_segment(tsk->thread.es);
+ i->cr_fsindex = encode_segment(tsk->thread.fsindex);
+ i->cr_gsindex = encode_segment(tsk->thread.gsindex);
+ i->cr_ss = encode_segment(regs->ss);
+
+ i->cr_dr0 = tsk->thread.debugreg0;
+ i->cr_dr1 = tsk->thread.debugreg1;
+ i->cr_dr2 = tsk->thread.debugreg2;
+ i->cr_dr3 = tsk->thread.debugreg3;
+ i->cr_dr6 = tsk->thread.debugreg6;
+ i->cr_dr7 = tsk->thread.debugreg7;
+
+ rv = cr_write(ctx, i, sizeof(*i));
+ kfree(i);
+ return rv;
+}
+
+int cr_arch_dump_task_struct(struct cr_context *ctx, struct task_struct *tsk)
+{
+#ifdef CONFIG_COMPAT
+ if (test_tsk_thread_flag(tsk, TIF_IA32))
+ return cr_dump_task_struct_x86_32(ctx, tsk);
+#endif
+ return cr_dump_task_struct_x86_64(ctx, tsk);
+}
+
+#ifdef CONFIG_COMPAT
+static int cr_restore_task_struct_x86_32(struct task_struct *tsk, struct cr_image_arch_x86_32 *i)
+{
+ struct pt_regs *regs = task_pt_regs(tsk);
+
+ tsk->thread.sp = (unsigned long)regs;
+ tsk->thread.sp0 = (unsigned long)(regs + 1);
+
+ regs->bx = i->cr_ebx;
+ regs->cx = i->cr_ecx;
+ regs->dx = i->cr_edx;
+ regs->si = i->cr_esi;
+ regs->di = i->cr_edi;
+ regs->bp = i->cr_ebp;
+ regs->ax = i->cr_eax;
+ regs->orig_ax = i->cr_orig_eax;
+ regs->ip = i->cr_eip;
+ regs->flags = i->cr_eflags;
+ regs->sp = i->cr_esp;
+ tsk->thread.usersp = regs->sp;
+
+ regs->cs = decode_segment(i->cr_cs);
+ tsk->thread.ds = decode_segment(i->cr_ds);
+ tsk->thread.es = decode_segment(i->cr_es);
+ tsk->thread.fs = 0;
+ tsk->thread.fsindex = decode_segment(i->cr_fs);
+ tsk->thread.gs = 0;
+ tsk->thread.gsindex = decode_segment(i->cr_gs);
+ regs->ss = decode_segment(i->cr_ss);
+
+ tsk->thread.debugreg0 = i->cr_dr0;
+ tsk->thread.debugreg1 = i->cr_dr1;
+ tsk->thread.debugreg2 = i->cr_dr2;
+ tsk->thread.debugreg3 = i->cr_dr3;
+ tsk->thread.debugreg6 = i->cr_dr6;
+ tsk->thread.debugreg7 = i->cr_dr7;
+
+ memcpy(tsk->thread.tls_array, i->cr_tls_array, 3 * 8);
+
+ set_tsk_thread_flag(tsk, TIF_FORK);
+ set_tsk_thread_flag(tsk, TIF_IA32);
+ return 0;
+}
+#endif
+
+static int cr_restore_task_struct_x86_64(struct task_struct *tsk, struct cr_image_arch_x86_64 *i)
+{
+ struct pt_regs *regs = task_pt_regs(tsk);
+
+ tsk->thread.sp = (unsigned long)regs;
+ tsk->thread.sp0 = (unsigned long)(regs + 1);
+
+ regs->r15 = i->cr_r15;
+ regs->r14 = i->cr_r14;
+ regs->r13 = i->cr_r13;
+ regs->r12 = i->cr_r12;
+ regs->bp = i->cr_rbp;
+ regs->bx = i->cr_rbx;
+ regs->r11 = i->cr_r11;
+ regs->r10 = i->cr_r10;
+ regs->r9 = i->cr_r9;
+ regs->r8 = i->cr_r8;
+ regs->ax = i->cr_rax;
+ regs->cx = i->cr_rcx;
+ regs->dx = i->cr_rdx;
+ regs->si = i->cr_rsi;
+ regs->di = i->cr_rdi;
+ regs->orig_ax = i->cr_orig_rax;
+ regs->ip = i->cr_rip;
+ regs->flags = i->cr_rflags;
+ regs->sp = i->cr_rsp;
+ tsk->thread.usersp = regs->sp;
+
+ tsk->thread.fs = i->cr_fs;
+ tsk->thread.gs = i->cr_gs;
+ regs->cs = decode_segment(i->cr_cs);
+ tsk->thread.ds = decode_segment(i->cr_ds);
+ tsk->thread.es = decode_segment(i->cr_es);
+ tsk->thread.fsindex = decode_segment(i->cr_fsindex);
+ tsk->thread.gsindex = decode_segment(i->cr_gsindex);
+ regs->ss = decode_segment(i->cr_ss);
+
+ tsk->thread.debugreg0 = i->cr_dr0;
+ tsk->thread.debugreg1 = i->cr_dr1;
+ tsk->thread.debugreg2 = i->cr_dr2;
+ tsk->thread.debugreg3 = i->cr_dr3;
+ tsk->thread.debugreg6 = i->cr_dr6;
+ tsk->thread.debugreg7 = i->cr_dr7;
+
+ set_tsk_thread_flag(tsk, TIF_FORK);
+ return 0;
+}
+
+int cr_arch_restore_task_struct(struct task_struct *tsk, struct cr_image_task_struct *i)
+{
+ if (i->cr_tsk_arch == CR_ARCH_X86_64) {
+ struct cr_image_arch_x86_64 *arch_i = (struct cr_image_arch_x86_64 *)(i + 1);
+
+ return cr_restore_task_struct_x86_64(tsk, arch_i);
+ }
+#ifdef CONFIG_COMPAT
+ if (i->cr_tsk_arch == CR_ARCH_X86_32) {
+ struct cr_image_arch_x86_32 *arch_i = (struct cr_image_arch_x86_32 *)(i + 1);
+
+ return cr_restore_task_struct_x86_32(tsk, arch_i);
+ }
+#endif
+ BUG();
+}
+
+int cr_arch_check_mm_struct(struct mm_struct *mm)
+{
+ mutex_lock(&mm->context.lock);
+ if (mm->context.ldt || mm->context.size != 0) {
+ mutex_unlock(&mm->context.lock);
+ WARN_ON(1);
+ return -EINVAL;
+ }
+ mutex_unlock(&mm->context.lock);
+ return 0;
+}
+
+unsigned int cr_arch_len_mm_struct(struct mm_struct *mm)
+{
+ return 0;
+}
+
+int cr_arch_dump_mm_struct(struct cr_context *ctx, struct mm_struct *mm)
+{
+ return 0;
+}
--- a/kernel/cr/cr.h
+++ b/kernel/cr/cr.h
@@ -78,7 +78,7 @@ int cr_restore_file(struct cr_context *ctx, loff_t pos);
int cr_restore_mm_struct(struct cr_context *ctx, loff_t pos);
int cr_restore_task_struct(struct cr_context *ctx, loff_t pos);

-#if defined(CONFIG_X86_32)
+#if defined(CONFIG_X86_32) || defined(CONFIG_X86_64)
__u32 cr_image_header_arch(void);
int cr_arch_check_image_header(struct cr_image_header *i);

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/