diff options
author | Stas Sergeev <stsp@aknet.ru> | 2006-12-07 02:14:01 +0100 |
---|---|---|
committer | Andi Kleen <andi@basil.nowhere.org> | 2006-12-07 02:14:01 +0100 |
commit | be44d2aabce2d62f72d5751d1871b6212bf7a1c7 (patch) | |
tree | 3f190dd5b5747ee83b50c4596b4801ce6c6b551c /arch/i386 | |
parent | bb81a09e55eaf7e5f798468ab971469b6f66a259 (diff) |
[PATCH] i386: espfix cleanup
Clean up the espfix code:
- Introduced PER_CPU() macro to be used from asm
- Introduced GET_DESC_BASE() macro to be used from asm
- Rewrote the fixup code in asm, as calling a C code with the altered %ss
appeared to be unsafe
- No longer altering the stack from a .fixup section
- 16bit per-cpu stack is no longer used, instead the stack segment base
is patched the way so that the high word of the kernel and user %esp
are the same.
- Added the limit-patching for the espfix segment. (Chuck Ebbert)
[jeremy@goop.org: use the x86 scaling addressing mode rather than shifting]
Signed-off-by: Stas Sergeev <stsp@aknet.ru>
Signed-off-by: Andi Kleen <ak@suse.de>
Acked-by: Zachary Amsden <zach@vmware.com>
Acked-by: Chuck Ebbert <76306.1226@compuserve.com>
Acked-by: Jan Beulich <jbeulich@novell.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Diffstat (limited to 'arch/i386')
-rw-r--r-- | arch/i386/kernel/asm-offsets.c | 5 | ||||
-rw-r--r-- | arch/i386/kernel/cpu/common.c | 11 | ||||
-rw-r--r-- | arch/i386/kernel/entry.S | 73 | ||||
-rw-r--r-- | arch/i386/kernel/head.S | 2 | ||||
-rw-r--r-- | arch/i386/kernel/traps.c | 57 |
5 files changed, 55 insertions, 93 deletions
diff --git a/arch/i386/kernel/asm-offsets.c b/arch/i386/kernel/asm-offsets.c index c80271f8f08..e94d910a28b 100644 --- a/arch/i386/kernel/asm-offsets.c +++ b/arch/i386/kernel/asm-offsets.c @@ -58,6 +58,11 @@ void foo(void) OFFSET(TI_sysenter_return, thread_info, sysenter_return); BLANK(); + OFFSET(GDS_size, Xgt_desc_struct, size); + OFFSET(GDS_address, Xgt_desc_struct, address); + OFFSET(GDS_pad, Xgt_desc_struct, pad); + BLANK(); + OFFSET(EXEC_DOMAIN_handler, exec_domain, handler); OFFSET(RT_SIGFRAME_sigcontext, rt_sigframe, uc.uc_mcontext); BLANK(); diff --git a/arch/i386/kernel/cpu/common.c b/arch/i386/kernel/cpu/common.c index d9f3e3c31f0..5532fc4e1bf 100644 --- a/arch/i386/kernel/cpu/common.c +++ b/arch/i386/kernel/cpu/common.c @@ -24,9 +24,6 @@ DEFINE_PER_CPU(struct Xgt_desc_struct, cpu_gdt_descr); EXPORT_PER_CPU_SYMBOL(cpu_gdt_descr); -DEFINE_PER_CPU(unsigned char, cpu_16bit_stack[CPU_16BIT_STACK_SIZE]); -EXPORT_PER_CPU_SYMBOL(cpu_16bit_stack); - static int cachesize_override __cpuinitdata = -1; static int disable_x86_fxsr __cpuinitdata; static int disable_x86_serial_nr __cpuinitdata = 1; @@ -603,7 +600,6 @@ void __cpuinit cpu_init(void) struct tss_struct * t = &per_cpu(init_tss, cpu); struct thread_struct *thread = ¤t->thread; struct desc_struct *gdt; - __u32 stk16_off = (__u32)&per_cpu(cpu_16bit_stack, cpu); struct Xgt_desc_struct *cpu_gdt_descr = &per_cpu(cpu_gdt_descr, cpu); if (cpu_test_and_set(cpu, cpu_initialized)) { @@ -651,13 +647,6 @@ old_gdt: * and set up the GDT descriptor: */ memcpy(gdt, cpu_gdt_table, GDT_SIZE); - - /* Set up GDT entry for 16bit stack */ - *(__u64 *)(&gdt[GDT_ENTRY_ESPFIX_SS]) |= - ((((__u64)stk16_off) << 16) & 0x000000ffffff0000ULL) | - ((((__u64)stk16_off) << 32) & 0xff00000000000000ULL) | - (CPU_16BIT_STACK_SIZE - 1); - cpu_gdt_descr->size = GDT_SIZE - 1; cpu_gdt_descr->address = (unsigned long)gdt; diff --git a/arch/i386/kernel/entry.S b/arch/i386/kernel/entry.S index 5a63d6fdb70..c38d801ba0b 100644 --- a/arch/i386/kernel/entry.S +++ b/arch/i386/kernel/entry.S @@ -48,6 +48,7 @@ #include <asm/smp.h> #include <asm/page.h> #include <asm/desc.h> +#include <asm/percpu.h> #include <asm/dwarf2.h> #include "irq_vectors.h" @@ -418,23 +419,18 @@ ldt_ss: * This is an "official" bug of all the x86-compatible * CPUs, which we can try to work around to make * dosemu and wine happy. */ - subl $8, %esp # reserve space for switch16 pointer - CFI_ADJUST_CFA_OFFSET 8 + movl OLDESP(%esp), %eax + movl %esp, %edx + call patch_espfix_desc + pushl $__ESPFIX_SS + CFI_ADJUST_CFA_OFFSET 4 + pushl %eax + CFI_ADJUST_CFA_OFFSET 4 DISABLE_INTERRUPTS TRACE_IRQS_OFF - movl %esp, %eax - /* Set up the 16bit stack frame with switch32 pointer on top, - * and a switch16 pointer on top of the current frame. */ - call setup_x86_bogus_stack - CFI_ADJUST_CFA_OFFSET -8 # frame has moved - TRACE_IRQS_IRET - RESTORE_REGS - lss 20+4(%esp), %esp # switch to 16bit stack -1: INTERRUPT_RETURN -.section __ex_table,"a" - .align 4 - .long 1b,iret_exc -.previous + lss (%esp), %esp + CFI_ADJUST_CFA_OFFSET -8 + jmp restore_nocheck CFI_ENDPROC # perform work that needs to be done immediately before resumption @@ -524,30 +520,30 @@ syscall_badsys: CFI_ENDPROC #define FIXUP_ESPFIX_STACK \ - movl %esp, %eax; \ - /* switch to 32bit stack using the pointer on top of 16bit stack */ \ - lss %ss:CPU_16BIT_STACK_SIZE-8, %esp; \ - /* copy data from 16bit stack to 32bit stack */ \ - call fixup_x86_bogus_stack; \ - /* put ESP to the proper location */ \ - movl %eax, %esp; -#define UNWIND_ESPFIX_STACK \ + /* since we are on a wrong stack, we cant make it a C code :( */ \ + GET_THREAD_INFO(%ebp); \ + movl TI_cpu(%ebp), %ebx; \ + PER_CPU(cpu_gdt_descr, %ebx); \ + movl GDS_address(%ebx), %ebx; \ + GET_DESC_BASE(GDT_ENTRY_ESPFIX_SS, %ebx, %eax, %ax, %al, %ah); \ + addl %esp, %eax; \ + pushl $__KERNEL_DS; \ + CFI_ADJUST_CFA_OFFSET 4; \ pushl %eax; \ CFI_ADJUST_CFA_OFFSET 4; \ + lss (%esp), %esp; \ + CFI_ADJUST_CFA_OFFSET -8; +#define UNWIND_ESPFIX_STACK \ movl %ss, %eax; \ - /* see if on 16bit stack */ \ + /* see if on espfix stack */ \ cmpw $__ESPFIX_SS, %ax; \ - je 28f; \ -27: popl %eax; \ - CFI_ADJUST_CFA_OFFSET -4; \ -.section .fixup,"ax"; \ -28: movl $__KERNEL_DS, %eax; \ + jne 27f; \ + movl $__KERNEL_DS, %eax; \ movl %eax, %ds; \ movl %eax, %es; \ - /* switch to 32bit stack */ \ + /* switch to normal stack */ \ FIXUP_ESPFIX_STACK; \ - jmp 27b; \ -.previous +27:; /* * Build the entry stubs and pointer table with @@ -614,7 +610,6 @@ error_code: pushl %eax CFI_ADJUST_CFA_OFFSET 4 CFI_REL_OFFSET eax, 0 - xorl %eax, %eax pushl %ebp CFI_ADJUST_CFA_OFFSET 4 CFI_REL_OFFSET ebp, 0 @@ -627,7 +622,6 @@ error_code: pushl %edx CFI_ADJUST_CFA_OFFSET 4 CFI_REL_OFFSET edx, 0 - decl %eax # eax = -1 pushl %ecx CFI_ADJUST_CFA_OFFSET 4 CFI_REL_OFFSET ecx, 0 @@ -644,7 +638,7 @@ error_code: /*CFI_REGISTER es, ecx*/ movl ES(%esp), %edi # get the function address movl ORIG_EAX(%esp), %edx # get the error code - movl %eax, ORIG_EAX(%esp) + movl $-1, ORIG_EAX(%esp) movl %ecx, ES(%esp) /*CFI_REL_OFFSET es, ES*/ movl $(__USER_DS), %ecx @@ -754,7 +748,7 @@ KPROBE_ENTRY(nmi) cmpw $__ESPFIX_SS, %ax popl %eax CFI_ADJUST_CFA_OFFSET -4 - je nmi_16bit_stack + je nmi_espfix_stack cmpl $sysenter_entry,(%esp) je nmi_stack_fixup pushl %eax @@ -797,7 +791,7 @@ nmi_debug_stack_check: FIX_STACK(24,nmi_stack_correct, 1) jmp nmi_stack_correct -nmi_16bit_stack: +nmi_espfix_stack: /* We have a RING0_INT_FRAME here. * * create the pointer to lss back @@ -806,7 +800,6 @@ nmi_16bit_stack: CFI_ADJUST_CFA_OFFSET 4 pushl %esp CFI_ADJUST_CFA_OFFSET 4 - movzwl %sp, %esp addw $4, (%esp) /* copy the iret frame of 12 bytes */ .rept 3 @@ -817,11 +810,11 @@ nmi_16bit_stack: CFI_ADJUST_CFA_OFFSET 4 SAVE_ALL FIXUP_ESPFIX_STACK # %eax == %esp - CFI_ADJUST_CFA_OFFSET -20 # the frame has now moved xorl %edx,%edx # zero error code call do_nmi RESTORE_REGS - lss 12+4(%esp), %esp # back to 16bit stack + lss 12+4(%esp), %esp # back to espfix stack + CFI_ADJUST_CFA_OFFSET -24 1: INTERRUPT_RETURN CFI_ENDPROC .section __ex_table,"a" diff --git a/arch/i386/kernel/head.S b/arch/i386/kernel/head.S index ca31f18d277..b1f1df11fcc 100644 --- a/arch/i386/kernel/head.S +++ b/arch/i386/kernel/head.S @@ -584,7 +584,7 @@ ENTRY(cpu_gdt_table) .quad 0x00009a000000ffff /* 0xc0 APM CS 16 code (16 bit) */ .quad 0x004092000000ffff /* 0xc8 APM DS data */ - .quad 0x0000920000000000 /* 0xd0 - ESPFIX 16-bit SS */ + .quad 0x00c0920000000000 /* 0xd0 - ESPFIX SS */ .quad 0x0000000000000000 /* 0xd8 - unused */ .quad 0x0000000000000000 /* 0xe0 - unused */ .quad 0x0000000000000000 /* 0xe8 - unused */ diff --git a/arch/i386/kernel/traps.c b/arch/i386/kernel/traps.c index 56655ea8d98..f9bb1f89d68 100644 --- a/arch/i386/kernel/traps.c +++ b/arch/i386/kernel/traps.c @@ -1088,49 +1088,24 @@ fastcall void do_spurious_interrupt_bug(struct pt_regs * regs, #endif } -fastcall void setup_x86_bogus_stack(unsigned char * stk) +fastcall unsigned long patch_espfix_desc(unsigned long uesp, + unsigned long kesp) { - unsigned long *switch16_ptr, *switch32_ptr; - struct pt_regs *regs; - unsigned long stack_top, stack_bot; - unsigned short iret_frame16_off; int cpu = smp_processor_id(); - /* reserve the space on 32bit stack for the magic switch16 pointer */ - memmove(stk, stk + 8, sizeof(struct pt_regs)); - switch16_ptr = (unsigned long *)(stk + sizeof(struct pt_regs)); - regs = (struct pt_regs *)stk; - /* now the switch32 on 16bit stack */ - stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu); - stack_top = stack_bot + CPU_16BIT_STACK_SIZE; - switch32_ptr = (unsigned long *)(stack_top - 8); - iret_frame16_off = CPU_16BIT_STACK_SIZE - 8 - 20; - /* copy iret frame on 16bit stack */ - memcpy((void *)(stack_bot + iret_frame16_off), ®s->eip, 20); - /* fill in the switch pointers */ - switch16_ptr[0] = (regs->esp & 0xffff0000) | iret_frame16_off; - switch16_ptr[1] = __ESPFIX_SS; - switch32_ptr[0] = (unsigned long)stk + sizeof(struct pt_regs) + - 8 - CPU_16BIT_STACK_SIZE; - switch32_ptr[1] = __KERNEL_DS; -} - -fastcall unsigned char * fixup_x86_bogus_stack(unsigned short sp) -{ - unsigned long *switch32_ptr; - unsigned char *stack16, *stack32; - unsigned long stack_top, stack_bot; - int len; - int cpu = smp_processor_id(); - stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu); - stack_top = stack_bot + CPU_16BIT_STACK_SIZE; - switch32_ptr = (unsigned long *)(stack_top - 8); - /* copy the data from 16bit stack to 32bit stack */ - len = CPU_16BIT_STACK_SIZE - 8 - sp; - stack16 = (unsigned char *)(stack_bot + sp); - stack32 = (unsigned char *) - (switch32_ptr[0] + CPU_16BIT_STACK_SIZE - 8 - len); - memcpy(stack32, stack16, len); - return stack32; + struct Xgt_desc_struct *cpu_gdt_descr = &per_cpu(cpu_gdt_descr, cpu); + struct desc_struct *gdt = (struct desc_struct *)cpu_gdt_descr->address; + unsigned long base = (kesp - uesp) & -THREAD_SIZE; + unsigned long new_kesp = kesp - base; + unsigned long lim_pages = (new_kesp | (THREAD_SIZE - 1)) >> PAGE_SHIFT; + __u64 desc = *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS]; + /* Set up base for espfix segment */ + desc &= 0x00f0ff0000000000ULL; + desc |= ((((__u64)base) << 16) & 0x000000ffffff0000ULL) | + ((((__u64)base) << 32) & 0xff00000000000000ULL) | + ((((__u64)lim_pages) << 32) & 0x000f000000000000ULL) | + (lim_pages & 0xffff); + *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS] = desc; + return new_kesp; } /* |