Save and restore segment registers on amd64 when entering and leaving

the kernel on amd64. Fill and read segment registers for mcontext and
signals. Handle traps caused by restoration of the
invalidated selectors.

Implement user-mode creation and manipulation of the process-specific
LDT descriptors for amd64, see sysarch(2).

Implement support for TSS i/o port access permission bitmap for amd64.

Context-switch LDT and TSS. Do not save and restore segment registers on
the context switch, that is handled by kernel enter/leave trampolines
now. Remove segment restore code from the signal trampolines for
freebsd/amd64, freebsd/ia32 and linux/i386 for the same reason.

Implement amd64-specific compat shims for sysarch.

Linuxolator (temporary ?) switched to use gsbase for thread_area pointer.

TODO:
Currently, gdb is not adapted to show segment registers from struct reg.
Also, no machine-depended ptrace command is added to set segment
registers for debugged process.

In collaboration with:	pho
Discussed with:	peter
Reviewed by:	jhb
Linuxolator tested by:	dchagin

1 files changed, 59 insertions, 94 deletions

diff --git a/sys/amd64/amd64/cpu_switch.S b/sys/amd64/amd64/cpu_switch.S
index 0c59703..6fc8290 100644
--- a/sys/amd64/amd64/cpu_switch.S
+++ b/sys/amd64/amd64/cpu_switch.S
@@ -75,8 +75,6 @@ ENTRY(cpu_throw)
 1:
 	movq	TD_PCB(%rdi),%r8		/* Old pcb */
 	movl	PCPU(CPUID), %eax
-	movq	PCB_FSBASE(%r8),%r9
-	movq	PCB_GSBASE(%r8),%r10
 	/* release bit from old pm_active */
 	movq	TD_PROC(%rdi), %rdx		/* oldtd->td_proc */
 	movq	P_VMSPACE(%rdx), %rdx		/* proc->p_vmspace */
@@ -110,28 +108,6 @@ ENTRY(cpu_switch)
 	movq	%rbx,PCB_RBX(%r8)
 	movq	%rax,PCB_RIP(%r8)
 
-	/*
-	 * Reread fs and gs bases. Explicit fs segment register load
-	 * by the usermode code may change actual fs base without
-	 * updating pcb_{fs,gs}base.
-	 *
-	 * %rdx still contains the mtx, save %rdx around rdmsr.
-	 */
-	movq	%rdx,%r11
-	movl	$MSR_FSBASE,%ecx
-	rdmsr
-	shlq	$32,%rdx
-	leaq	(%rax,%rdx),%r9
-	movl	$MSR_KGSBASE,%ecx
-	rdmsr
-	shlq	$32,%rdx
-	leaq	(%rax,%rdx),%r10
-	movq	%r11,%rdx
-
-	testl	$PCB_32BIT,PCB_FLAGS(%r8)
-	jnz	store_seg
-done_store_seg:
-
 	testl	$PCB_DBREGS,PCB_FLAGS(%r8)
 	jnz	store_dr			/* static predict not taken */
 done_store_dr:
@@ -192,36 +168,47 @@ sw1:
 	testl	$TDP_KTHREAD,TD_PFLAGS(%rsi)
 	jnz	do_kthread
 
-	testl	$PCB_32BIT,PCB_FLAGS(%r8)
-	jnz	load_seg
-done_load_seg:
+	/*
+	 * Load ldt register
+	 */
+	movq	TD_PROC(%rsi),%rcx
+	cmpq	$0, P_MD+MD_LDT(%rcx)
+	jne	do_ldt
+	xorl	%eax,%eax
+ld_ldt:	lldt	%ax
 
-	cmpq	PCB_FSBASE(%r8),%r9
-	jz	1f
-	/* Restore userland %fs */
-restore_fsbase:
-	movl	$MSR_FSBASE,%ecx
+	/* Restore fs base in GDT */
 	movl	PCB_FSBASE(%r8),%eax
-	movl	PCB_FSBASE+4(%r8),%edx
-	wrmsr
-1:
-	cmpq	PCB_GSBASE(%r8),%r10
-	jz	2f
-	/* Restore userland %gs */
-	movl	$MSR_KGSBASE,%ecx
+	movq	PCPU(FS32P),%rdx
+	movw	%ax,2(%rdx)
+	shrl	$16,%eax
+	movb	%al,4(%rdx)
+	shrl	$8,%eax
+	movb	%al,7(%rdx)
+
+	/* Restore gs base in GDT */
 	movl	PCB_GSBASE(%r8),%eax
-	movl	PCB_GSBASE+4(%r8),%edx
-	wrmsr
-2:
+	movq	PCPU(GS32P),%rdx
+	movw	%ax,2(%rdx)
+	shrl	$16,%eax
+	movb	%al,4(%rdx)
+	shrl	$8,%eax
+	movb	%al,7(%rdx)
 
-do_tss:
+do_kthread:
+	/* Do we need to reload tss ? */
+	movq	PCPU(TSSP),%rax
+	movq	PCB_TSSP(%r8),%rdx
+	testq	%rdx,%rdx
+	cmovzq	PCPU(COMMONTSSP),%rdx
+	cmpq	%rax,%rdx
+	jne	do_tss
+done_tss:
+	movq	%r8,PCPU(RSP0)
+	movq	%r8,PCPU(CURPCB)
 	/* Update the TSS_RSP0 pointer for the next interrupt */
-	movq	PCPU(TSSP), %rax
-	movq	%r8, PCPU(RSP0)
-	movq	%r8, PCPU(CURPCB)
-	addq	$COMMON_TSS_RSP0, %rax
-	movq	%rsi, PCPU(CURTHREAD)		/* into next thread */
-	movq	%r8, (%rax)
+	movq	%r8,COMMON_TSS_RSP0(%rdx)
+	movq	%rsi,PCPU(CURTHREAD)		/* into next thread */
 
 	/* Test if debug registers should be restored. */
 	testl	$PCB_DBREGS,PCB_FLAGS(%r8)
@@ -250,45 +237,6 @@ done_load_dr:
 	 * We use jumps rather than call in order to avoid the stack.
 	 */
 
-do_kthread:
-	/*
-	 * Copy old fs/gsbase to new kthread pcb for future switches
-	 * This maintains curpcb->pcb_[fg]sbase as caches of the MSR
-	 */
-	movq	%r9,PCB_FSBASE(%r8)
-	movq	%r10,PCB_GSBASE(%r8)
-	jmp	do_tss
-
-store_seg:
-	mov	%gs,PCB_GS(%r8)
-	testl	$PCB_GS32BIT,PCB_FLAGS(%r8)
-	jnz	2f
-1:	mov	%ds,PCB_DS(%r8)
-	mov	%es,PCB_ES(%r8)
-	mov	%fs,PCB_FS(%r8)
-	jmp	done_store_seg
-2:	movq	PCPU(GS32P),%rax
-	movq	(%rax),%rax
-	movq	%rax,PCB_GS32SD(%r8)
-	jmp	1b
-
-load_seg:
-	testl	$PCB_GS32BIT,PCB_FLAGS(%r8)
-	jnz	2f
-1:	movl	$MSR_GSBASE,%ecx
-	rdmsr
-	mov	PCB_GS(%r8),%gs
-	wrmsr
-	mov	PCB_DS(%r8),%ds
-	mov	PCB_ES(%r8),%es
-	mov	PCB_FS(%r8),%fs
-	jmp	restore_fsbase
-	/* Restore userland %gs while preserving kernel gsbase */
-2:	movq	PCPU(GS32P),%rax
-	movq	PCB_GS32SD(%r8),%rcx
-	movq	%rcx,(%rax)
-	jmp	1b
-
 store_dr:
 	movq	%dr7,%rax			/* yes, do the save */
 	movq	%dr0,%r15
@@ -325,6 +273,29 @@ load_dr:
 	movq	%r11,%dr6
 	movq	%rax,%dr7
 	jmp	done_load_dr
+
+do_tss:	movq	%rdx,PCPU(TSSP)
+	movq	%rdx,%rcx
+	movq	PCPU(TSS),%rax
+	movw	%rcx,2(%rax)
+	shrq	$16,%rcx
+	movb	%cl,4(%rax)
+	shrq	$8,%rcx
+	movb	%cl,7(%rax)
+	shrq	$8,%rcx
+	movl	%ecx,8(%rax)
+	movb	$0x89,5(%rax)	/* unset busy */
+	movl	$TSSSEL,%eax
+	ltr	%ax
+	jmp	done_tss
+
+do_ldt:	movq	PCPU(LDT),%rax
+	movq	P_MD+MD_LDT_SD(%rcx),%rdx
+	movq	%rdx,(%rax)
+	movq	P_MD+MD_LDT_SD+8(%rcx),%rdx
+	movq	%rdx,8(%rax)
+	movl	$LDTSEL,%eax
+	jmp	ld_ldt
 END(cpu_switch)
 
 /*
@@ -398,12 +369,6 @@ ENTRY(savectx2)
 	movq	(%rsp),%rax
 	movq	%rax,PCB_RIP(%r8)
 
-	mov	%ds,PCB_DS(%r8)
-	mov	%es,PCB_ES(%r8)
-	mov	%ss,XPCB_SS(%r8)
-	mov	%fs,PCB_FS(%r8)
-	mov	%gs,PCB_GS(%r8)
-
 	movq	%rbx,PCB_RBX(%r8)
 	movq	%rsp,PCB_RSP(%r8)
 	movq	%rbp,PCB_RBP(%r8)