Commit 7f3591cf authored by Linus Torvalds's avatar Linus Torvalds
Browse files

Merge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-lguest

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-lguest: (31 commits)
  lguest: add support for indirect ring entries
  lguest: suppress notifications in example Launcher
  lguest: try to batch interrupts on network receive
  lguest: avoid sending interrupts to Guest when no activity occurs.
  lguest: implement deferred interrupts in example Launcher
  lguest: remove obsolete LHREQ_BREAK call
  lguest: have example Launcher service all devices in separate threads
  lguest: use eventfds for device notification
  eventfd: export eventfd_signal and eventfd_fget for lguest
  lguest: allow any process to send interrupts
  lguest: PAE fixes
  lguest: PAE support
  lguest: Add support for kvm_hypercall4()
  lguest: replace hypercall name LHCALL_SET_PMD with LHCALL_SET_PGD
  lguest: use native_set_* macros, which properly handle 64-bit entries when PAE is activated
  lguest: map switcher with executable page table entries
  lguest: fix writev returning short on console output
  lguest: clean up length-used value in example launcher
  lguest: Segment selectors are 16-bit long. Fix lg_cpu.ss1 definition.
  lguest: beyond ARRAY_SIZE of cpu->arch.gdt
  ...
parents 16ffc3ee d1f0132e
# This creates the demonstration utility "lguest" which runs a Linux guest.
CFLAGS:=-Wall -Wmissing-declarations -Wmissing-prototypes -O3 -I../../include -I../../arch/x86/include -U_FORTIFY_SOURCE
LDLIBS:=-lz
CFLAGS:=-m32 -Wall -Wmissing-declarations -Wmissing-prototypes -O3 -I../../include -I../../arch/x86/include -U_FORTIFY_SOURCE
all: lguest
......
This diff is collapsed.
......@@ -37,7 +37,6 @@ Running Lguest:
"Paravirtualized guest support" = Y
"Lguest guest support" = Y
"High Memory Support" = off/4GB
"PAE (Physical Address Extension) Support" = N
"Alignment value to which kernel should be aligned" = 0x100000
(CONFIG_PARAVIRT=y, CONFIG_LGUEST_GUEST=y, CONFIG_HIGHMEM64G=n and
CONFIG_PHYSICAL_ALIGN=0x100000)
......
......@@ -17,8 +17,13 @@
/* Pages for switcher itself, then two pages per cpu */
#define TOTAL_SWITCHER_PAGES (SHARED_SWITCHER_PAGES + 2 * nr_cpu_ids)
/* We map at -4M for ease of mapping into the guest (one PTE page). */
/* We map at -4M (-2M when PAE is activated) for ease of mapping
* into the guest (one PTE page). */
#ifdef CONFIG_X86_PAE
#define SWITCHER_ADDR 0xFFE00000
#else
#define SWITCHER_ADDR 0xFFC00000
#endif
/* Found in switcher.S */
extern unsigned long default_idt_entries[];
......
......@@ -12,11 +12,13 @@
#define LHCALL_TS 8
#define LHCALL_SET_CLOCKEVENT 9
#define LHCALL_HALT 10
#define LHCALL_SET_PMD 13
#define LHCALL_SET_PTE 14
#define LHCALL_SET_PMD 15
#define LHCALL_SET_PGD 15
#define LHCALL_LOAD_TLS 16
#define LHCALL_NOTIFY 17
#define LHCALL_LOAD_GDT_ENTRY 18
#define LHCALL_SEND_INTERRUPTS 19
#define LGUEST_TRAP_ENTRY 0x1F
......@@ -32,10 +34,10 @@
* operations? There are two ways: the direct way is to make a "hypercall",
* to make requests of the Host Itself.
*
* We use the KVM hypercall mechanism. Eighteen hypercalls are
* We use the KVM hypercall mechanism. Seventeen hypercalls are
* available: the hypercall number is put in the %eax register, and the
* arguments (when required) are placed in %ebx, %ecx and %edx. If a return
* value makes sense, it's returned in %eax.
* arguments (when required) are placed in %ebx, %ecx, %edx and %esi.
* If a return value makes sense, it's returned in %eax.
*
* Grossly invalid calls result in Sudden Death at the hands of the vengeful
* Host, rather than returning failure. This reflects Winston Churchill's
......@@ -47,8 +49,9 @@
#define LHCALL_RING_SIZE 64
struct hcall_args {
/* These map directly onto eax, ebx, ecx, edx in struct lguest_regs */
unsigned long arg0, arg1, arg2, arg3;
/* These map directly onto eax, ebx, ecx, edx and esi
* in struct lguest_regs */
unsigned long arg0, arg1, arg2, arg3, arg4;
};
#endif /* !__ASSEMBLY__ */
......
......@@ -126,6 +126,7 @@ void foo(void)
#if defined(CONFIG_LGUEST) || defined(CONFIG_LGUEST_GUEST) || defined(CONFIG_LGUEST_MODULE)
BLANK();
OFFSET(LGUEST_DATA_irq_enabled, lguest_data, irq_enabled);
OFFSET(LGUEST_DATA_irq_pending, lguest_data, irq_pending);
OFFSET(LGUEST_DATA_pgdir, lguest_data, pgdir);
BLANK();
......
......@@ -2,7 +2,6 @@ config LGUEST_GUEST
bool "Lguest guest support"
select PARAVIRT
depends on X86_32
depends on !X86_PAE
select VIRTIO
select VIRTIO_RING
select VIRTIO_CONSOLE
......
......@@ -87,7 +87,7 @@ struct lguest_data lguest_data = {
/*G:037 async_hcall() is pretty simple: I'm quite proud of it really. We have a
* ring buffer of stored hypercalls which the Host will run though next time we
* do a normal hypercall. Each entry in the ring has 4 slots for the hypercall
* do a normal hypercall. Each entry in the ring has 5 slots for the hypercall
* arguments, and a "hcall_status" word which is 0 if the call is ready to go,
* and 255 once the Host has finished with it.
*
......@@ -96,7 +96,8 @@ struct lguest_data lguest_data = {
* effect of causing the Host to run all the stored calls in the ring buffer
* which empties it for next time! */
static void async_hcall(unsigned long call, unsigned long arg1,
unsigned long arg2, unsigned long arg3)
unsigned long arg2, unsigned long arg3,
unsigned long arg4)
{
/* Note: This code assumes we're uniprocessor. */
static unsigned int next_call;
......@@ -108,12 +109,13 @@ static void async_hcall(unsigned long call, unsigned long arg1,
local_irq_save(flags);
if (lguest_data.hcall_status[next_call] != 0xFF) {
/* Table full, so do normal hcall which will flush table. */
kvm_hypercall3(call, arg1, arg2, arg3);
kvm_hypercall4(call, arg1, arg2, arg3, arg4);
} else {
lguest_data.hcalls[next_call].arg0 = call;
lguest_data.hcalls[next_call].arg1 = arg1;
lguest_data.hcalls[next_call].arg2 = arg2;
lguest_data.hcalls[next_call].arg3 = arg3;
lguest_data.hcalls[next_call].arg4 = arg4;
/* Arguments must all be written before we mark it to go */
wmb();
lguest_data.hcall_status[next_call] = 0;
......@@ -141,7 +143,7 @@ static void lazy_hcall1(unsigned long call,
if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE)
kvm_hypercall1(call, arg1);
else
async_hcall(call, arg1, 0, 0);
async_hcall(call, arg1, 0, 0, 0);
}
static void lazy_hcall2(unsigned long call,
......@@ -151,7 +153,7 @@ static void lazy_hcall2(unsigned long call,
if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE)
kvm_hypercall2(call, arg1, arg2);
else
async_hcall(call, arg1, arg2, 0);
async_hcall(call, arg1, arg2, 0, 0);
}
static void lazy_hcall3(unsigned long call,
......@@ -162,9 +164,23 @@ static void lazy_hcall3(unsigned long call,
if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE)
kvm_hypercall3(call, arg1, arg2, arg3);
else
async_hcall(call, arg1, arg2, arg3);
async_hcall(call, arg1, arg2, arg3, 0);
}
#ifdef CONFIG_X86_PAE
static void lazy_hcall4(unsigned long call,
unsigned long arg1,
unsigned long arg2,
unsigned long arg3,
unsigned long arg4)
{
if (paravirt_get_lazy_mode() == PARAVIRT_LAZY_NONE)
kvm_hypercall4(call, arg1, arg2, arg3, arg4);
else
async_hcall(call, arg1, arg2, arg3, arg4);
}
#endif
/* When lazy mode is turned off reset the per-cpu lazy mode variable and then
* issue the do-nothing hypercall to flush any stored calls. */
static void lguest_leave_lazy_mmu_mode(void)
......@@ -179,7 +195,7 @@ static void lguest_end_context_switch(struct task_struct *next)
paravirt_end_context_switch(next);
}
/*G:033
/*G:032
* After that diversion we return to our first native-instruction
* replacements: four functions for interrupt control.
*
......@@ -199,30 +215,28 @@ static unsigned long save_fl(void)
{
return lguest_data.irq_enabled;
}
PV_CALLEE_SAVE_REGS_THUNK(save_fl);
/* restore_flags() just sets the flags back to the value given. */
static void restore_fl(unsigned long flags)
{
lguest_data.irq_enabled = flags;
}
PV_CALLEE_SAVE_REGS_THUNK(restore_fl);
/* Interrupts go off... */
static void irq_disable(void)
{
lguest_data.irq_enabled = 0;
}
/* Let's pause a moment. Remember how I said these are called so often?
* Jeremy Fitzhardinge optimized them so hard early in 2009 that he had to
* break some rules. In particular, these functions are assumed to save their
* own registers if they need to: normal C functions assume they can trash the
* eax register. To use normal C functions, we use
* PV_CALLEE_SAVE_REGS_THUNK(), which pushes %eax onto the stack, calls the
* C function, then restores it. */
PV_CALLEE_SAVE_REGS_THUNK(save_fl);
PV_CALLEE_SAVE_REGS_THUNK(irq_disable);
/*:*/
/* Interrupts go on... */
static void irq_enable(void)
{
lguest_data.irq_enabled = X86_EFLAGS_IF;
}
PV_CALLEE_SAVE_REGS_THUNK(irq_enable);
/* These are in i386_head.S */
extern void lg_irq_enable(void);
extern void lg_restore_fl(unsigned long flags);
/*:*/
/*M:003 Note that we don't check for outstanding interrupts when we re-enable
* them (or when we unmask an interrupt). This seems to work for the moment,
* since interrupts are rare and we'll just get the interrupt on the next timer
......@@ -368,8 +382,8 @@ static void lguest_cpuid(unsigned int *ax, unsigned int *bx,
case 1: /* Basic feature request. */
/* We only allow kernel to see SSE3, CMPXCHG16B and SSSE3 */
*cx &= 0x00002201;
/* SSE, SSE2, FXSR, MMX, CMOV, CMPXCHG8B, TSC, FPU. */
*dx &= 0x07808111;
/* SSE, SSE2, FXSR, MMX, CMOV, CMPXCHG8B, TSC, FPU, PAE. */
*dx &= 0x07808151;
/* The Host can do a nice optimization if it knows that the
* kernel mappings (addresses above 0xC0000000 or whatever
* PAGE_OFFSET is set to) haven't changed. But Linux calls
......@@ -388,6 +402,11 @@ static void lguest_cpuid(unsigned int *ax, unsigned int *bx,
if (*ax > 0x80000008)
*ax = 0x80000008;
break;
case 0x80000001:
/* Here we should fix nx cap depending on host. */
/* For this version of PAE, we just clear NX bit. */
*dx &= ~(1 << 20);
break;
}
}
......@@ -521,25 +540,52 @@ static void lguest_write_cr4(unsigned long val)
static void lguest_pte_update(struct mm_struct *mm, unsigned long addr,
pte_t *ptep)
{
#ifdef CONFIG_X86_PAE
lazy_hcall4(LHCALL_SET_PTE, __pa(mm->pgd), addr,
ptep->pte_low, ptep->pte_high);
#else
lazy_hcall3(LHCALL_SET_PTE, __pa(mm->pgd), addr, ptep->pte_low);
#endif
}
static void lguest_set_pte_at(struct mm_struct *mm, unsigned long addr,
pte_t *ptep, pte_t pteval)
{
*ptep = pteval;
native_set_pte(ptep, pteval);
lguest_pte_update(mm, addr, ptep);
}
/* The Guest calls this to set a top-level entry. Again, we set the entry then
* tell the Host which top-level page we changed, and the index of the entry we
* changed. */
/* The Guest calls lguest_set_pud to set a top-level entry and lguest_set_pmd
* to set a middle-level entry when PAE is activated.
* Again, we set the entry then tell the Host which page we changed,
* and the index of the entry we changed. */
#ifdef CONFIG_X86_PAE
static void lguest_set_pud(pud_t *pudp, pud_t pudval)
{
native_set_pud(pudp, pudval);
/* 32 bytes aligned pdpt address and the index. */
lazy_hcall2(LHCALL_SET_PGD, __pa(pudp) & 0xFFFFFFE0,
(__pa(pudp) & 0x1F) / sizeof(pud_t));
}
static void lguest_set_pmd(pmd_t *pmdp, pmd_t pmdval)
{
*pmdp = pmdval;
native_set_pmd(pmdp, pmdval);
lazy_hcall2(LHCALL_SET_PMD, __pa(pmdp) & PAGE_MASK,
(__pa(pmdp) & (PAGE_SIZE - 1)) / 4);
(__pa(pmdp) & (PAGE_SIZE - 1)) / sizeof(pmd_t));
}
#else
/* The Guest calls lguest_set_pmd to set a top-level entry when PAE is not
* activated. */
static void lguest_set_pmd(pmd_t *pmdp, pmd_t pmdval)
{
native_set_pmd(pmdp, pmdval);
lazy_hcall2(LHCALL_SET_PGD, __pa(pmdp) & PAGE_MASK,
(__pa(pmdp) & (PAGE_SIZE - 1)) / sizeof(pmd_t));
}
#endif
/* There are a couple of legacy places where the kernel sets a PTE, but we
* don't know the top level any more. This is useless for us, since we don't
......@@ -552,11 +598,31 @@ static void lguest_set_pmd(pmd_t *pmdp, pmd_t pmdval)
* which brings boot back to 0.25 seconds. */
static void lguest_set_pte(pte_t *ptep, pte_t pteval)
{
*ptep = pteval;
native_set_pte(ptep, pteval);
if (cr3_changed)
lazy_hcall1(LHCALL_FLUSH_TLB, 1);
}
#ifdef CONFIG_X86_PAE
static void lguest_set_pte_atomic(pte_t *ptep, pte_t pte)
{
native_set_pte_atomic(ptep, pte);
if (cr3_changed)
lazy_hcall1(LHCALL_FLUSH_TLB, 1);
}
void lguest_pte_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
{
native_pte_clear(mm, addr, ptep);
lguest_pte_update(mm, addr, ptep);
}
void lguest_pmd_clear(pmd_t *pmdp)
{
lguest_set_pmd(pmdp, __pmd(0));
}
#endif
/* Unfortunately for Lguest, the pv_mmu_ops for page tables were based on
* native page table operations. On native hardware you can set a new page
* table entry whenever you want, but if you want to remove one you have to do
......@@ -628,13 +694,12 @@ static void __init lguest_init_IRQ(void)
{
unsigned int i;
for (i = 0; i < LGUEST_IRQS; i++) {
int vector = FIRST_EXTERNAL_VECTOR + i;
for (i = FIRST_EXTERNAL_VECTOR; i < NR_VECTORS; i++) {
/* Some systems map "vectors" to interrupts weirdly. Lguest has
* a straightforward 1 to 1 mapping, so force that here. */
__get_cpu_var(vector_irq)[vector] = i;
if (vector != SYSCALL_VECTOR)
set_intr_gate(vector, interrupt[i]);
__get_cpu_var(vector_irq)[i] = i - FIRST_EXTERNAL_VECTOR;
if (i != SYSCALL_VECTOR)
set_intr_gate(i, interrupt[i - FIRST_EXTERNAL_VECTOR]);
}
/* This call is required to set up for 4k stacks, where we have
* separate stacks for hard and soft interrupts. */
......@@ -973,10 +1038,10 @@ static void lguest_restart(char *reason)
*
* Our current solution is to allow the paravirt back end to optionally patch
* over the indirect calls to replace them with something more efficient. We
* patch the four most commonly called functions: disable interrupts, enable
* interrupts, restore interrupts and save interrupts. We usually have 6 or 10
* bytes to patch into: the Guest versions of these operations are small enough
* that we can fit comfortably.
* patch two of the simplest of the most commonly called functions: disable
* interrupts and save interrupts. We usually have 6 or 10 bytes to patch
* into: the Guest versions of these operations are small enough that we can
* fit comfortably.
*
* First we need assembly templates of each of the patchable Guest operations,
* and these are in i386_head.S. */
......@@ -987,8 +1052,6 @@ static const struct lguest_insns
const char *start, *end;
} lguest_insns[] = {
[PARAVIRT_PATCH(pv_irq_ops.irq_disable)] = { lgstart_cli, lgend_cli },
[PARAVIRT_PATCH(pv_irq_ops.irq_enable)] = { lgstart_sti, lgend_sti },
[PARAVIRT_PATCH(pv_irq_ops.restore_fl)] = { lgstart_popf, lgend_popf },
[PARAVIRT_PATCH(pv_irq_ops.save_fl)] = { lgstart_pushf, lgend_pushf },
};
......@@ -1026,6 +1089,7 @@ __init void lguest_init(void)
pv_info.name = "lguest";
pv_info.paravirt_enabled = 1;
pv_info.kernel_rpl = 1;
pv_info.shared_kernel_pmd = 1;
/* We set up all the lguest overrides for sensitive operations. These
* are detailed with the operations themselves. */
......@@ -1033,9 +1097,9 @@ __init void lguest_init(void)
/* interrupt-related operations */
pv_irq_ops.init_IRQ = lguest_init_IRQ;
pv_irq_ops.save_fl = PV_CALLEE_SAVE(save_fl);
pv_irq_ops.restore_fl = PV_CALLEE_SAVE(restore_fl);
pv_irq_ops.restore_fl = __PV_IS_CALLEE_SAVE(lg_restore_fl);
pv_irq_ops.irq_disable = PV_CALLEE_SAVE(irq_disable);
pv_irq_ops.irq_enable = PV_CALLEE_SAVE(irq_enable);
pv_irq_ops.irq_enable = __PV_IS_CALLEE_SAVE(lg_irq_enable);
pv_irq_ops.safe_halt = lguest_safe_halt;
/* init-time operations */
......@@ -1071,6 +1135,12 @@ __init void lguest_init(void)
pv_mmu_ops.set_pte = lguest_set_pte;
pv_mmu_ops.set_pte_at = lguest_set_pte_at;
pv_mmu_ops.set_pmd = lguest_set_pmd;
#ifdef CONFIG_X86_PAE
pv_mmu_ops.set_pte_atomic = lguest_set_pte_atomic;
pv_mmu_ops.pte_clear = lguest_pte_clear;
pv_mmu_ops.pmd_clear = lguest_pmd_clear;
pv_mmu_ops.set_pud = lguest_set_pud;
#endif
pv_mmu_ops.read_cr2 = lguest_read_cr2;
pv_mmu_ops.read_cr3 = lguest_read_cr3;
pv_mmu_ops.lazy_mode.enter = paravirt_enter_lazy_mmu;
......
......@@ -46,10 +46,64 @@ ENTRY(lguest_entry)
.globl lgstart_##name; .globl lgend_##name
LGUEST_PATCH(cli, movl $0, lguest_data+LGUEST_DATA_irq_enabled)
LGUEST_PATCH(sti, movl $X86_EFLAGS_IF, lguest_data+LGUEST_DATA_irq_enabled)
LGUEST_PATCH(popf, movl %eax, lguest_data+LGUEST_DATA_irq_enabled)
LGUEST_PATCH(pushf, movl lguest_data+LGUEST_DATA_irq_enabled, %eax)
/*:*/
/*G:033 But using those wrappers is inefficient (we'll see why that doesn't
* matter for save_fl and irq_disable later). If we write our routines
* carefully in assembler, we can avoid clobbering any registers and avoid
* jumping through the wrapper functions.
*
* I skipped over our first piece of assembler, but this one is worth studying
* in a bit more detail so I'll describe in easy stages. First, the routine
* to enable interrupts: */
ENTRY(lg_irq_enable)
/* The reverse of irq_disable, this sets lguest_data.irq_enabled to
* X86_EFLAGS_IF (ie. "Interrupts enabled"). */
movl $X86_EFLAGS_IF, lguest_data+LGUEST_DATA_irq_enabled
/* But now we need to check if the Host wants to know: there might have
* been interrupts waiting to be delivered, in which case it will have
* set lguest_data.irq_pending to X86_EFLAGS_IF. If it's not zero, we
* jump to send_interrupts, otherwise we're done. */
testl $0, lguest_data+LGUEST_DATA_irq_pending
jnz send_interrupts
/* One cool thing about x86 is that you can do many things without using
* a register. In this case, the normal path hasn't needed to save or
* restore any registers at all! */
ret
send_interrupts:
/* OK, now we need a register: eax is used for the hypercall number,
* which is LHCALL_SEND_INTERRUPTS.
*
* We used not to bother with this pending detection at all, which was
* much simpler. Sooner or later the Host would realize it had to
* send us an interrupt. But that turns out to make performance 7
* times worse on a simple tcp benchmark. So now we do this the hard
* way. */
pushl %eax
movl $LHCALL_SEND_INTERRUPTS, %eax
/* This is a vmcall instruction (same thing that KVM uses). Older
* assembler versions might not know the "vmcall" instruction, so we
* create one manually here. */
.byte 0x0f,0x01,0xc1 /* KVM_HYPERCALL */
popl %eax
ret
/* Finally, the "popf" or "restore flags" routine. The %eax register holds the
* flags (in practice, either X86_EFLAGS_IF or 0): if it's X86_EFLAGS_IF we're
* enabling interrupts again, if it's 0 we're leaving them off. */
ENTRY(lg_restore_fl)
/* This is just "lguest_data.irq_enabled = flags;" */
movl %eax, lguest_data+LGUEST_DATA_irq_enabled
/* Now, if the %eax value has enabled interrupts and
* lguest_data.irq_pending is set, we want to tell the Host so it can
* deliver any outstanding interrupts. Fortunately, both values will
* be X86_EFLAGS_IF (ie. 512) in that case, and the "testl"
* instruction will AND them together for us. If both are set, we
* jump to send_interrupts. */
testl lguest_data+LGUEST_DATA_irq_pending, %eax
jnz send_interrupts
/* Again, the normal path has used no extra registers. Clever, huh? */
ret
/* These demark the EIP range where host should never deliver interrupts. */
.global lguest_noirq_start
......
config LGUEST
tristate "Linux hypervisor example code"
depends on X86_32 && EXPERIMENTAL && !X86_PAE && FUTEX
depends on X86_32 && EXPERIMENTAL && EVENTFD
select HVC_DRIVER
---help---
This is a very simple module which allows you to run
......
......@@ -95,7 +95,7 @@ static __init int map_switcher(void)
* array of struct pages. It increments that pointer, but we don't
* care. */
pagep = switcher_page;
err = map_vm_area(switcher_vma, PAGE_KERNEL, &pagep);
err = map_vm_area(switcher_vma, PAGE_KERNEL_EXEC, &pagep);
if (err) {
printk("lguest: map_vm_area failed: %i\n", err);
goto free_vma;
......@@ -188,6 +188,9 @@ int run_guest(struct lg_cpu *cpu, unsigned long __user *user)
{
/* We stop running once the Guest is dead. */
while (!cpu->lg->dead) {
unsigned int irq;
bool more;
/* First we run any hypercalls the Guest wants done. */
if (cpu->hcall)
do_hypercalls(cpu);
......@@ -195,23 +198,23 @@ int run_guest(struct lg_cpu *cpu, unsigned long __user *user)
/* It's possible the Guest did a NOTIFY hypercall to the
* Launcher, in which case we return from the read() now. */
if (cpu->pending_notify) {
if (put_user(cpu->pending_notify, user))
return -EFAULT;
return sizeof(cpu->pending_notify);
if (!send_notify_to_eventfd(cpu)) {
if (put_user(cpu->pending_notify, user))
return -EFAULT;
return sizeof(cpu->pending_notify);
}
}
/* Check for signals */
if (signal_pending(current))
return -ERESTARTSYS;
/* If Waker set break_out, return to Launcher. */
if (cpu->break_out)
return -EAGAIN;
/* Check if there are any interrupts which can be delivered now:
* if so, this sets up the hander to be executed when we next
* run the Guest. */
maybe_do_interrupt(cpu);
irq = interrupt_pending(cpu, &more);
if (irq < LGUEST_IRQS)
try_deliver_interrupt(cpu, irq, more);
/* All long-lived kernel loops need to check with this horrible
* thing called the freezer. If the Host is trying to suspend,
......@@ -224,10 +227,15 @@ int run_guest(struct lg_cpu *cpu, unsigned long __user *user)
break;
/* If the Guest asked to be stopped, we sleep. The Guest's
* clock timer or LHREQ_BREAK from the Waker will wake us. */
* clock timer will wake us. */
if (cpu->halted) {
set_current_state(TASK_INTERRUPTIBLE);
schedule();
/* Just before we sleep, make sure no interrupt snuck in
* which we should be doing. */
if (interrupt_pending(cpu, &more) < LGUEST_IRQS)
set_current_state(TASK_RUNNING);
else
schedule();
continue;
}
......
......@@ -37,6 +37,10 @@ static void do_hcall(struct lg_cpu *cpu, struct hcall_args *args)
/* This call does nothing, except by breaking out of the Guest
* it makes us process all the asynchronous hypercalls. */
break;
case LHCALL_SEND_INTERRUPTS:
/* This call does nothing too, but by breaking out of the Guest
* it makes us process any pending interrupts. */
break;
case LHCALL_LGUEST_INIT:
/* You can't get here unless you're already initialized. Don't
* do that. */
......@@ -73,11 +77,21 @@ static void do_hcall(struct lg_cpu *cpu, struct hcall_args *args)
guest_set_stack(cpu, args->arg1, args->arg2, args->arg3);
break;
case LHCALL_SET_PTE:
#ifdef CONFIG_X86_PAE
guest_set_pte(cpu, args->arg1, args->arg2,
__pte(args->arg3 | (u64)args->arg4 << 32));
#else
guest_set_pte(cpu, args->arg1, args->arg2, __pte(args->arg3));
#endif
break;
case LHCALL_SET_PGD:
guest_set_pgd(cpu->lg, args->arg1, args->arg2);
break;
#ifdef CONFIG_X86_PAE
case LHCALL_SET_PMD:
guest_set_pmd(cpu->lg, args->arg1, args->arg2);
break;
#endif
case LHCALL_SET_CLOCKEVENT:
guest_set_clockevent(cpu, args->arg1);
break;
......
......@@ -128,30 +128,39 @@ static void set_guest_interrupt(struct lg_cpu *cpu, u32 lo, u32 hi,
/*H:205
* Virtual Interrupts.
*
* maybe_do_interrupt() gets called before every entry to the Guest, to see if
* we should divert the Guest to running an interrupt handler. */
void maybe_do_interrupt(struct lg_cpu *cpu)
* interrupt_pending() returns the first pending interrupt which isn't blocked
* by the Guest. It is called before every entry to the Guest, and just before
* we go to sleep when the Guest has halted itself. */
unsigned int interrupt_pending(struct lg_cpu *cpu, bool *more)
{
unsigned int irq;
DECLARE_BITMAP(blk, LGUEST_IRQS);
struct desc_struct *idt;
/* If the Guest hasn't even initialized yet, we can do nothing. */
if (!cpu->lg->lguest_data)
return;
return LGUEST_IRQS;
/* Take our "irqs_pending" array and remove any interrupts the Guest
* wants blocked: the result ends up in "blk". */
if (copy_from_user(&blk, cpu->lg->lguest_data->blocked_interrupts,
sizeof(blk)))
return;
return LGUEST_IRQS;
bitmap_andnot(blk, cpu->irqs_pending, blk