Move i386 part of core.c to x86/core.c.

Separate i386 architecture specific from core.c and move it to
x86/core.c and add x86/lguest.h header file to match.

Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

2 files changed, 478 insertions, 1 deletions

diff --git a/drivers/lguest/x86/core.c b/drivers/lguest/x86/core.c
new file mode 100644
index 00000000000..e2f46b16ce3
--- /dev/null
+++ b/drivers/lguest/x86/core.c
@@ -0,0 +1,476 @@
+/*
+ * Copyright (C) 2006, Rusty Russell <rusty@rustcorp.com.au> IBM Corporation.
+ * Copyright (C) 2007, Jes Sorensen <jes@sgi.com> SGI.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
+ * NON INFRINGEMENT.  See the GNU General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+#include <linux/kernel.h>
+#include <linux/start_kernel.h>
+#include <linux/string.h>
+#include <linux/console.h>
+#include <linux/screen_info.h>
+#include <linux/irq.h>
+#include <linux/interrupt.h>
+#include <linux/clocksource.h>
+#include <linux/clockchips.h>
+#include <linux/cpu.h>
+#include <linux/lguest.h>
+#include <linux/lguest_launcher.h>
+#include <linux/lguest_bus.h>
+#include <asm/paravirt.h>
+#include <asm/param.h>
+#include <asm/page.h>
+#include <asm/pgtable.h>
+#include <asm/desc.h>
+#include <asm/setup.h>
+#include <asm/lguest.h>
+#include <asm/uaccess.h>
+#include <asm/i387.h>
+#include "../lg.h"
+
+static int cpu_had_pge;
+
+static struct {
+	unsigned long offset;
+	unsigned short segment;
+} lguest_entry;
+
+/* Offset from where switcher.S was compiled to where we've copied it */
+static unsigned long switcher_offset(void)
+{
+	return SWITCHER_ADDR - (unsigned long)start_switcher_text;
+}
+
+/* This cpu's struct lguest_pages. */
+static struct lguest_pages *lguest_pages(unsigned int cpu)
+{
+	return &(((struct lguest_pages *)
+		  (SWITCHER_ADDR + SHARED_SWITCHER_PAGES*PAGE_SIZE))[cpu]);
+}
+
+static DEFINE_PER_CPU(struct lguest *, last_guest);
+
+/*S:010
+ * We are getting close to the Switcher.
+ *
+ * Remember that each CPU has two pages which are visible to the Guest when it
+ * runs on that CPU.  This has to contain the state for that Guest: we copy the
+ * state in just before we run the Guest.
+ *
+ * Each Guest has "changed" flags which indicate what has changed in the Guest
+ * since it last ran.  We saw this set in interrupts_and_traps.c and
+ * segments.c.
+ */
+static void copy_in_guest_info(struct lguest *lg, struct lguest_pages *pages)
+{
+	/* Copying all this data can be quite expensive.  We usually run the
+	 * same Guest we ran last time (and that Guest hasn't run anywhere else
+	 * meanwhile).  If that's not the case, we pretend everything in the
+	 * Guest has changed. */
+	if (__get_cpu_var(last_guest) != lg || lg->last_pages != pages) {
+		__get_cpu_var(last_guest) = lg;
+		lg->last_pages = pages;
+		lg->changed = CHANGED_ALL;
+	}
+
+	/* These copies are pretty cheap, so we do them unconditionally: */
+	/* Save the current Host top-level page directory. */
+	pages->state.host_cr3 = __pa(current->mm->pgd);
+	/* Set up the Guest's page tables to see this CPU's pages (and no
+	 * other CPU's pages). */
+	map_switcher_in_guest(lg, pages);
+	/* Set up the two "TSS" members which tell the CPU what stack to use
+	 * for traps which do directly into the Guest (ie. traps at privilege
+	 * level 1). */
+	pages->state.guest_tss.esp1 = lg->esp1;
+	pages->state.guest_tss.ss1 = lg->ss1;
+
+	/* Copy direct-to-Guest trap entries. */
+	if (lg->changed & CHANGED_IDT)
+		copy_traps(lg, pages->state.guest_idt, default_idt_entries);
+
+	/* Copy all GDT entries which the Guest can change. */
+	if (lg->changed & CHANGED_GDT)
+		copy_gdt(lg, pages->state.guest_gdt);
+	/* If only the TLS entries have changed, copy them. */
+	else if (lg->changed & CHANGED_GDT_TLS)
+		copy_gdt_tls(lg, pages->state.guest_gdt);
+
+	/* Mark the Guest as unchanged for next time. */
+	lg->changed = 0;
+}
+
+/* Finally: the code to actually call into the Switcher to run the Guest. */
+static void run_guest_once(struct lguest *lg, struct lguest_pages *pages)
+{
+	/* This is a dummy value we need for GCC's sake. */
+	unsigned int clobber;
+
+	/* Copy the guest-specific information into this CPU's "struct
+	 * lguest_pages". */
+	copy_in_guest_info(lg, pages);
+
+	/* Set the trap number to 256 (impossible value).  If we fault while
+	 * switching to the Guest (bad segment registers or bug), this will
+	 * cause us to abort the Guest. */
+	lg->regs->trapnum = 256;
+
+	/* Now: we push the "eflags" register on the stack, then do an "lcall".
+	 * This is how we change from using the kernel code segment to using
+	 * the dedicated lguest code segment, as well as jumping into the
+	 * Switcher.
+	 *
+	 * The lcall also pushes the old code segment (KERNEL_CS) onto the
+	 * stack, then the address of this call.  This stack layout happens to
+	 * exactly match the stack of an interrupt... */
+	asm volatile("pushf; lcall *lguest_entry"
+		     /* This is how we tell GCC that %eax ("a") and %ebx ("b")
+		      * are changed by this routine.  The "=" means output. */
+		     : "=a"(clobber), "=b"(clobber)
+		     /* %eax contains the pages pointer.  ("0" refers to the
+		      * 0-th argument above, ie "a").  %ebx contains the
+		      * physical address of the Guest's top-level page
+		      * directory. */
+		     : "0"(pages), "1"(__pa(lg->pgdirs[lg->pgdidx].pgdir))
+		     /* We tell gcc that all these registers could change,
+		      * which means we don't have to save and restore them in
+		      * the Switcher. */
+		     : "memory", "%edx", "%ecx", "%edi", "%esi");
+}
+/*:*/
+
+/*H:040 This is the i386-specific code to setup and run the Guest.  Interrupts
+ * are disabled: we own the CPU. */
+void lguest_arch_run_guest(struct lguest *lg)
+{
+	/* Remember the awfully-named TS bit?  If the Guest has asked
+	 * to set it we set it now, so we can trap and pass that trap
+	 * to the Guest if it uses the FPU. */
+	if (lg->ts)
+		lguest_set_ts();
+
+	/* SYSENTER is an optimized way of doing system calls.  We
+	 * can't allow it because it always jumps to privilege level 0.
+	 * A normal Guest won't try it because we don't advertise it in
+	 * CPUID, but a malicious Guest (or malicious Guest userspace
+	 * program) could, so we tell the CPU to disable it before
+	 * running the Guest. */
+	if (boot_cpu_has(X86_FEATURE_SEP))
+		wrmsr(MSR_IA32_SYSENTER_CS, 0, 0);
+
+	/* Now we actually run the Guest.  It will pop back out when
+	 * something interesting happens, and we can examine its
+	 * registers to see what it was doing. */
+	run_guest_once(lg, lguest_pages(raw_smp_processor_id()));
+
+	/* The "regs" pointer contains two extra entries which are not
+	 * really registers: a trap number which says what interrupt or
+	 * trap made the switcher code come back, and an error code
+	 * which some traps set.  */
+
+	/* If the Guest page faulted, then the cr2 register will tell
+	 * us the bad virtual address.  We have to grab this now,
+	 * because once we re-enable interrupts an interrupt could
+	 * fault and thus overwrite cr2, or we could even move off to a
+	 * different CPU. */
+	if (lg->regs->trapnum == 14)
+		lg->arch.last_pagefault = read_cr2();
+	/* Similarly, if we took a trap because the Guest used the FPU,
+	 * we have to restore the FPU it expects to see. */
+	else if (lg->regs->trapnum == 7)
+		math_state_restore();
+
+	/* Restore SYSENTER if it's supposed to be on. */
+	if (boot_cpu_has(X86_FEATURE_SEP))
+		wrmsr(MSR_IA32_SYSENTER_CS, __KERNEL_CS, 0);
+}
+
+/*H:130 Our Guest is usually so well behaved; it never tries to do things it
+ * isn't allowed to.  Unfortunately, Linux's paravirtual infrastructure isn't
+ * quite complete, because it doesn't contain replacements for the Intel I/O
+ * instructions.  As a result, the Guest sometimes fumbles across one during
+ * the boot process as it probes for various things which are usually attached
+ * to a PC.
+ *
+ * When the Guest uses one of these instructions, we get trap #13 (General
+ * Protection Fault) and come here.  We see if it's one of those troublesome
+ * instructions and skip over it.  We return true if we did. */
+static int emulate_insn(struct lguest *lg)
+{
+	u8 insn;
+	unsigned int insnlen = 0, in = 0, shift = 0;
+	/* The eip contains the *virtual* address of the Guest's instruction:
+	 * guest_pa just subtracts the Guest's page_offset. */
+	unsigned long physaddr = guest_pa(lg, lg->regs->eip);
+
+	/* The guest_pa() function only works for Guest kernel addresses, but
+	 * that's all we're trying to do anyway. */
+	if (lg->regs->eip < lg->page_offset)
+		return 0;
+
+	/* Decoding x86 instructions is icky. */
+	lgread(lg, &insn, physaddr, 1);
+
+	/* 0x66 is an "operand prefix".  It means it's using the upper 16 bits
+	   of the eax register. */
+	if (insn == 0x66) {
+		shift = 16;
+		/* The instruction is 1 byte so far, read the next byte. */
+		insnlen = 1;
+		lgread(lg, &insn, physaddr + insnlen, 1);
+	}
+
+	/* We can ignore the lower bit for the moment and decode the 4 opcodes
+	 * we need to emulate. */
+	switch (insn & 0xFE) {
+	case 0xE4: /* in     <next byte>,%al */
+		insnlen += 2;
+		in = 1;
+		break;
+	case 0xEC: /* in     (%dx),%al */
+		insnlen += 1;
+		in = 1;
+		break;
+	case 0xE6: /* out    %al,<next byte> */
+		insnlen += 2;
+		break;
+	case 0xEE: /* out    %al,(%dx) */
+		insnlen += 1;
+		break;
+	default:
+		/* OK, we don't know what this is, can't emulate. */
+		return 0;
+	}
+
+	/* If it was an "IN" instruction, they expect the result to be read
+	 * into %eax, so we change %eax.  We always return all-ones, which
+	 * traditionally means "there's nothing there". */
+	if (in) {
+		/* Lower bit tells is whether it's a 16 or 32 bit access */
+		if (insn & 0x1)
+			lg->regs->eax = 0xFFFFFFFF;
+		else
+			lg->regs->eax |= (0xFFFF << shift);
+	}
+	/* Finally, we've "done" the instruction, so move past it. */
+	lg->regs->eip += insnlen;
+	/* Success! */
+	return 1;
+}
+
+/*H:050 Once we've re-enabled interrupts, we look at why the Guest exited. */
+void lguest_arch_handle_trap(struct lguest *lg)
+{
+	switch (lg->regs->trapnum) {
+	case 13: /* We've intercepted a GPF. */
+		 /* Check if this was one of those annoying IN or OUT
+		  * instructions which we need to emulate.  If so, we
+		  * just go back into the Guest after we've done it. */
+		if (lg->regs->errcode == 0) {
+			if (emulate_insn(lg))
+				return;
+		}
+		break;
+	case 14: /* We've intercepted a page fault. */
+		 /* The Guest accessed a virtual address that wasn't
+		  * mapped.  This happens a lot: we don't actually set
+		  * up most of the page tables for the Guest at all when
+		  * we start: as it runs it asks for more and more, and
+		  * we set them up as required. In this case, we don't
+		  * even tell the Guest that the fault happened.
+		  *
+		  * The errcode tells whether this was a read or a
+		  * write, and whether kernel or userspace code. */
+		if (demand_page(lg, lg->arch.last_pagefault, lg->regs->errcode))
+			return;
+
+		 /* OK, it's really not there (or not OK): the Guest
+		  * needs to know.  We write out the cr2 value so it
+		  * knows where the fault occurred.
+		  *
+		  * Note that if the Guest were really messed up, this
+		  * could happen before it's done the INITIALIZE
+		  * hypercall, so lg->lguest_data will be NULL */
+		if (lg->lguest_data &&
+		    put_user(lg->arch.last_pagefault, &lg->lguest_data->cr2))
+			kill_guest(lg, "Writing cr2");
+		break;
+	case 7: /* We've intercepted a Device Not Available fault. */
+		/* If the Guest doesn't want to know, we already
+		 * restored the Floating Point Unit, so we just
+		 * continue without telling it. */
+		if (!lg->ts)
+			return;
+		break;
+	case 32 ... 255:
+		/* These values mean a real interrupt occurred, in
+		 * which case the Host handler has already been run.
+		 * We just do a friendly check if another process
+		 * should now be run, then fall through to loop
+		 * around: */
+		cond_resched();
+	case LGUEST_TRAP_ENTRY: /* Handled before re-entering Guest */
+		return;
+	}
+
+	/* We didn't handle the trap, so it needs to go to the Guest. */
+	if (!deliver_trap(lg, lg->regs->trapnum))
+		/* If the Guest doesn't have a handler (either it hasn't
+		 * registered any yet, or it's one of the faults we don't let
+		 * it handle), it dies with a cryptic error message. */
+		kill_guest(lg, "unhandled trap %li at %#lx (%#lx)",
+			   lg->regs->trapnum, lg->regs->eip,
+			   lg->regs->trapnum == 14 ? lg->arch.last_pagefault
+			   : lg->regs->errcode);
+}
+
+/* Now we can look at each of the routines this calls, in increasing order of
+ * complexity: do_hypercalls(), emulate_insn(), maybe_do_interrupt(),
+ * deliver_trap() and demand_page().  After all those, we'll be ready to
+ * examine the Switcher, and our philosophical understanding of the Host/Guest
+ * duality will be complete. :*/
+static void adjust_pge(void *on)
+{
+	if (on)
+		write_cr4(read_cr4() | X86_CR4_PGE);
+	else
+		write_cr4(read_cr4() & ~X86_CR4_PGE);
+}
+
+/*H:020 Now the Switcher is mapped and every thing else is ready, we need to do
+ * some more i386-specific initialization. */
+void __init lguest_arch_host_init(void)
+{
+	int i;
+
+	/* Most of the i386/switcher.S doesn't care that it's been moved; on
+	 * Intel, jumps are relative, and it doesn't access any references to
+	 * external code or data.
+	 *
+	 * The only exception is the interrupt handlers in switcher.S: their
+	 * addresses are placed in a table (default_idt_entries), so we need to
+	 * update the table with the new addresses.  switcher_offset() is a
+	 * convenience function which returns the distance between the builtin
+	 * switcher code and the high-mapped copy we just made. */
+	for (i = 0; i < IDT_ENTRIES; i++)
+		default_idt_entries[i] += switcher_offset();
+
+	/*
+	 * Set up the Switcher's per-cpu areas.
+	 *
+	 * Each CPU gets two pages of its own within the high-mapped region
+	 * (aka. "struct lguest_pages").  Much of this can be initialized now,
+	 * but some depends on what Guest we are running (which is set up in
+	 * copy_in_guest_info()).
+	 */
+	for_each_possible_cpu(i) {
+		/* lguest_pages() returns this CPU's two pages. */
+		struct lguest_pages *pages = lguest_pages(i);
+		/* This is a convenience pointer to make the code fit one
+		 * statement to a line. */
+		struct lguest_ro_state *state = &pages->state;
+
+		/* The Global Descriptor Table: the Host has a different one
+		 * for each CPU.  We keep a descriptor for the GDT which says
+		 * where it is and how big it is (the size is actually the last
+		 * byte, not the size, hence the "-1"). */
+		state->host_gdt_desc.size = GDT_SIZE-1;
+		state->host_gdt_desc.address = (long)get_cpu_gdt_table(i);
+
+		/* All CPUs on the Host use the same Interrupt Descriptor
+		 * Table, so we just use store_idt(), which gets this CPU's IDT
+		 * descriptor. */
+		store_idt(&state->host_idt_desc);
+
+		/* The descriptors for the Guest's GDT and IDT can be filled
+		 * out now, too.  We copy the GDT & IDT into ->guest_gdt and
+		 * ->guest_idt before actually running the Guest. */
+		state->guest_idt_desc.size = sizeof(state->guest_idt)-1;
+		state->guest_idt_desc.address = (long)&state->guest_idt;
+		state->guest_gdt_desc.size = sizeof(state->guest_gdt)-1;
+		state->guest_gdt_desc.address = (long)&state->guest_gdt;
+
+		/* We know where we want the stack to be when the Guest enters
+		 * the switcher: in pages->regs.  The stack grows upwards, so
+		 * we start it at the end of that structure. */
+		state->guest_tss.esp0 = (long)(&pages->regs + 1);
+		/* And this is the GDT entry to use for the stack: we keep a
+		 * couple of special LGUEST entries. */
+		state->guest_tss.ss0 = LGUEST_DS;
+
+		/* x86 can have a finegrained bitmap which indicates what I/O
+		 * ports the process can use.  We set it to the end of our
+		 * structure, meaning "none". */
+		state->guest_tss.io_bitmap_base = sizeof(state->guest_tss);
+
+		/* Some GDT entries are the same across all Guests, so we can
+		 * set them up now. */
+		setup_default_gdt_entries(state);
+		/* Most IDT entries are the same for all Guests, too.*/
+		setup_default_idt_entries(state, default_idt_entries);
+
+		/* The Host needs to be able to use the LGUEST segments on this
+		 * CPU, too, so put them in the Host GDT. */
+		get_cpu_gdt_table(i)[GDT_ENTRY_LGUEST_CS] = FULL_EXEC_SEGMENT;
+		get_cpu_gdt_table(i)[GDT_ENTRY_LGUEST_DS] = FULL_SEGMENT;
+	}
+
+	/* In the Switcher, we want the %cs segment register to use the
+	 * LGUEST_CS GDT entry: we've put that in the Host and Guest GDTs, so
+	 * it will be undisturbed when we switch.  To change %cs and jump we
+	 * need this structure to feed to Intel's "lcall" instruction. */
+	lguest_entry.offset = (long)switch_to_guest + switcher_offset();
+	lguest_entry.segment = LGUEST_CS;
+
+	/* Finally, we need to turn off "Page Global Enable".  PGE is an
+	 * optimization where page table entries are specially marked to show
+	 * they never change.  The Host kernel marks all the kernel pages this
+	 * way because it's always present, even when userspace is running.
+	 *
+	 * Lguest breaks this: unbeknownst to the rest of the Host kernel, we
+	 * switch to the Guest kernel.  If you don't disable this on all CPUs,
+	 * you'll get really weird bugs that you'll chase for two days.
+	 *
+	 * I used to turn PGE off every time we switched to the Guest and back
+	 * on when we return, but that slowed the Switcher down noticibly. */
+
+	/* We don't need the complexity of CPUs coming and going while we're
+	 * doing this. */
+	lock_cpu_hotplug();
+	if (cpu_has_pge) { /* We have a broader idea of "global". */
+		/* Remember that this was originally set (for cleanup). */
+		cpu_had_pge = 1;
+		/* adjust_pge is a helper function which sets or unsets the PGE
+		 * bit on its CPU, depending on the argument (0 == unset). */
+		on_each_cpu(adjust_pge, (void *)0, 0, 1);
+		/* Turn off the feature in the global feature set. */
+		clear_bit(X86_FEATURE_PGE, boot_cpu_data.x86_capability);
+	}
+	unlock_cpu_hotplug();
+};
+/*:*/
+
+void __exit lguest_arch_host_fini(void)
+{
+	/* If we had PGE before we started, turn it back on now. */
+	lock_cpu_hotplug();
+	if (cpu_had_pge) {
+		set_bit(X86_FEATURE_PGE, boot_cpu_data.x86_capability);
+		/* adjust_pge's argument "1" means set PGE. */
+		on_each_cpu(adjust_pge, (void *)1, 0, 1);
+	}
+	unlock_cpu_hotplug();
+}
diff --git a/drivers/lguest/x86/switcher_32.S b/drivers/lguest/x86/switcher_32.S
index a3d23f79cba..e66cec5ac24 100644
--- a/drivers/lguest/x86/switcher_32.S
+++ b/drivers/lguest/x86/switcher_32.S
@@ -48,7 +48,8 @@
 #include <linux/linkage.h>
 #include <asm/asm-offsets.h>
 #include <asm/page.h>
-#include "../lg.h"
+#include <asm/segment.h>
+#include <asm/lguest.h>
 
 // We mark the start of the code to copy
 // It's placed in .text tho it's never run here