static DEFINE_PER_CPU(struct lguest *, last_guest);
/*S:010
- * We are getting close to the Switcher.
+ * We approach 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
* 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)
+static void copy_in_guest_info(struct lg_cpu *cpu, struct lguest_pages *pages)
{
+ struct lguest *lg = cpu->lg;
/* 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
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);
+ map_switcher_in_guest(cpu, 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). */
}
/* 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)
+static void run_guest_once(struct lg_cpu *cpu, struct lguest_pages *pages)
{
/* This is a dummy value we need for GCC's sake. */
unsigned int clobber;
+ struct lguest *lg = cpu->lg;
/* Copy the guest-specific information into this CPU's "struct
* lguest_pages". */
- copy_in_guest_info(lg, pages);
+ copy_in_guest_info(cpu, 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;
+ cpu->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 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... */
+ * exactly match the stack layout created by 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. */
}
/*:*/
+/*M:002 There are hooks in the scheduler which we can register to tell when we
+ * get kicked off the CPU (preempt_notifier_register()). This would allow us
+ * to lazily disable SYSENTER which would regain some performance, and should
+ * also simplify copy_in_guest_info(). Note that we'd still need to restore
+ * things when we exit to Launcher userspace, but that's fairly easy.
+ *
+ * The hooks were designed for KVM, but we can also put them to good use. :*/
+
/*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)
+void lguest_arch_run_guest(struct lg_cpu *cpu)
{
- /* 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. */
+ struct lguest *lg = cpu->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. */
+ /* 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)
+ /* Now we actually run the Guest. It will return when something
+ * interesting happens, and we can examine its registers to see what it
+ * was doing. */
+ run_guest_once(cpu, lguest_pages(raw_smp_processor_id()));
+
+ /* Note that 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 (cpu->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)
+ else if (cpu->regs->trapnum == 7)
math_state_restore();
/* Restore SYSENTER if it's supposed to be on. */
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.
+/*H:130 Now we've examined the hypercall code; our Guest can make requests.
+ * Our Guest is usually so well behaved; it never tries to do things it isn't
+ * allowed to, and uses hypercalls instead. 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
+ * When the Guest uses one of these instructions, we get a trap (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)
+static int emulate_insn(struct lg_cpu *cpu)
{
+ struct lguest *lg = cpu->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);
+ unsigned long physaddr = guest_pa(lg, cpu->regs->eip);
/* This must be the Guest kernel trying to do something, not userspace!
* The bottom two bits of the CS segment register are the privilege
* level. */
- if ((lg->regs->cs & 3) != GUEST_PL)
+ if ((cpu->regs->cs & 3) != GUEST_PL)
return 0;
/* Decoding x86 instructions is icky. */
if (in) {
/* Lower bit tells is whether it's a 16 or 32 bit access */
if (insn & 0x1)
- lg->regs->eax = 0xFFFFFFFF;
+ cpu->regs->eax = 0xFFFFFFFF;
else
- lg->regs->eax |= (0xFFFF << shift);
+ cpu->regs->eax |= (0xFFFF << shift);
}
/* Finally, we've "done" the instruction, so move past it. */
- lg->regs->eip += insnlen;
+ cpu->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)
+void lguest_arch_handle_trap(struct lg_cpu *cpu)
{
- 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))
+ struct lguest *lg = cpu->lg;
+ switch (cpu->regs->trapnum) {
+ case 13: /* We've intercepted a General Protection Fault. */
+ /* 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 (cpu->regs->errcode == 0) {
+ if (emulate_insn(cpu))
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))
+ 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, cpu->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 */
+ /* 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 LHCALL_LGUEST_INIT hypercall, so
+ * lg->lguest_data could 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 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 LGUEST_TRAP_ENTRY:
/* Our 'struct hcall_args' maps directly over our regs: we set
* up the pointer now to indicate a hypercall is pending. */
- lg->hcall = (struct hcall_args *)lg->regs;
+ cpu->hcall = (struct hcall_args *)cpu->regs;
return;
}
/* We didn't handle the trap, so it needs to go to the Guest. */
- if (!deliver_trap(lg, lg->regs->trapnum))
+ if (!deliver_trap(cpu, cpu->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);
+ cpu->regs->trapnum, cpu->regs->eip,
+ cpu->regs->trapnum == 14 ? lg->arch.last_pagefault
+ : cpu->regs->errcode);
}
/* Now we can look at each of the routines this calls, in increasing order of
/* We don't need the complexity of CPUs coming and going while we're
* doing this. */
- lock_cpu_hotplug();
+ get_online_cpus();
if (cpu_has_pge) { /* We have a broader idea of "global". */
/* Remember that this was originally set (for cleanup). */
cpu_had_pge = 1;
/* Turn off the feature in the global feature set. */
clear_bit(X86_FEATURE_PGE, boot_cpu_data.x86_capability);
}
- unlock_cpu_hotplug();
+ put_online_cpus();
};
/*:*/
void __exit lguest_arch_host_fini(void)
{
/* If we had PGE before we started, turn it back on now. */
- lock_cpu_hotplug();
+ get_online_cpus();
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();
+ put_online_cpus();
}
/*H:122 The i386-specific hypercalls simply farm out to the right functions. */
-int lguest_arch_do_hcall(struct lguest *lg, struct hcall_args *args)
+int lguest_arch_do_hcall(struct lg_cpu *cpu, struct hcall_args *args)
{
+ struct lguest *lg = cpu->lg;
+
switch (args->arg0) {
case LHCALL_LOAD_GDT:
load_guest_gdt(lg, args->arg1, args->arg2);
}
/*H:126 i386-specific hypercall initialization: */
-int lguest_arch_init_hypercalls(struct lguest *lg)
+int lguest_arch_init_hypercalls(struct lg_cpu *cpu)
{
u32 tsc_speed;
+ struct lguest *lg = cpu->lg;
/* The pointer to the Guest's "struct lguest_data" is the only
* argument. We check that address now. */
- if (!lguest_address_ok(lg, lg->hcall->arg1, sizeof(*lg->lguest_data)))
+ if (!lguest_address_ok(lg, cpu->hcall->arg1, sizeof(*lg->lguest_data)))
return -EFAULT;
/* Having checked it, we simply set lg->lguest_data to point straight
* copy_to_user/from_user from now on, instead of lgread/write. I put
* this in to show that I'm not immune to writing stupid
* optimizations. */
- lg->lguest_data = lg->mem_base + lg->hcall->arg1;
+ lg->lguest_data = lg->mem_base + cpu->hcall->arg1;
/* We insist that the Time Stamp Counter exist and doesn't change with
* cpu frequency. Some devious chip manufacturers decided that TSC
return 0;
}
-/* Now we've examined the hypercall code; our Guest can make requests. There
- * is one other way we can do things for the Guest, as we see in
- * emulate_insn(). :*/
/*L:030 lguest_arch_setup_regs()
*
* Most of the Guest's registers are left alone: we used get_zeroed_page() to
* allocate the structure, so they will be 0. */
-void lguest_arch_setup_regs(struct lguest *lg, unsigned long start)
+void lguest_arch_setup_regs(struct lg_cpu *cpu, unsigned long start)
{
- struct lguest_regs *regs = lg->regs;
+ struct lguest_regs *regs = cpu->regs;
/* There are four "segment" registers which the Guest needs to boot:
* The "code segment" register (cs) refers to the kernel code segment
* is supposed to always be "1". Bit 9 (0x200) controls whether
* interrupts are enabled. We always leave interrupts enabled while
* running the Guest. */
- regs->eflags = 0x202;
+ regs->eflags = X86_EFLAGS_IF | 0x2;
/* The "Extended Instruction Pointer" register says where the Guest is
* running. */
/* %esi points to our boot information, at physical address 0, so don't
* touch it. */
+
/* There are a couple of GDT entries the Guest expects when first
* booting. */
-
- setup_guest_gdt(lg);
+ setup_guest_gdt(cpu->lg);
}
projects / linux-2.6 / blobdiff