/*
* Define the code needed before returning to user mode, for trap and
* syscall.
*/
void
userret(struct thread *td, struct trapframe *frame)
{
struct proc *p = td->td_proc;
CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid,
td->td_name);
KASSERT((p->p_flag & P_WEXIT) == 0,
("Exiting process returns to usermode"));
#if 0
#ifdef DIAGNOSTIC
/* Check that we called signotify() enough. */
PROC_LOCK(p);
thread_lock(td);
if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 ||
(td->td_flags & TDF_ASTPENDING) == 0))
printf("failed to set signal flags properly for ast()\n");
thread_unlock(td);
PROC_UNLOCK(p);
#endif
#endif
#ifdef KTRACE
KTRUSERRET(td);
#endif
/*
* If this thread tickled GEOM, we need to wait for the giggling to
* stop before we return to userland
*/
if (td->td_pflags & TDP_GEOM)
g_waitidle();
/*
* Charge system time if profiling.
*/
if (p->p_flag & P_PROFIL)
addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio);
/*
* Let the scheduler adjust our priority etc.
*/
sched_userret(td);
KASSERT(td->td_locks == 0,
("userret: Returning with %d locks held.", td->td_locks));
#ifdef VIMAGE
/* Unfortunately td_vnet_lpush needs VNET_DEBUG. */
VNET_ASSERT(curvnet == NULL,
("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s",
__func__, td, p->p_pid, td->td_name, curvnet,
(td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A"));
#endif
#ifdef XEN
PT_UPDATES_FLUSH();
#endif
}
static void
cpu_initialize_context(unsigned int cpu)
{
/* vcpu_guest_context_t is too large to allocate on the stack.
* Hence we allocate statically and protect it with a lock */
vm_page_t m[NPGPTD + 2];
static vcpu_guest_context_t ctxt;
vm_offset_t boot_stack;
vm_offset_t newPTD;
vm_paddr_t ma[NPGPTD];
int i;
/*
* Page 0,[0-3] PTD
* Page 1, [4] boot stack
* Page [5] PDPT
*
*/
for (i = 0; i < NPGPTD + 2; i++) {
m[i] = vm_page_alloc(NULL, 0,
VM_ALLOC_NORMAL | VM_ALLOC_NOOBJ | VM_ALLOC_WIRED |
VM_ALLOC_ZERO);
pmap_zero_page(m[i]);
}
boot_stack = kmem_alloc_nofault(kernel_map, PAGE_SIZE);
newPTD = kmem_alloc_nofault(kernel_map, NPGPTD * PAGE_SIZE);
ma[0] = VM_PAGE_TO_MACH(m[0])|PG_V;
#ifdef PAE
pmap_kenter(boot_stack, VM_PAGE_TO_PHYS(m[NPGPTD + 1]));
for (i = 0; i < NPGPTD; i++) {
((vm_paddr_t *)boot_stack)[i] =
ma[i] = VM_PAGE_TO_MACH(m[i])|PG_V;
}
#endif
/*
* Copy cpu0 IdlePTD to new IdlePTD - copying only
* kernel mappings
*/
pmap_qenter(newPTD, m, 4);
memcpy((uint8_t *)newPTD + KPTDI*sizeof(vm_paddr_t),
(uint8_t *)PTOV(IdlePTD) + KPTDI*sizeof(vm_paddr_t),
nkpt*sizeof(vm_paddr_t));
pmap_qremove(newPTD, 4);
kmem_free(kernel_map, newPTD, 4 * PAGE_SIZE);
/*
* map actual idle stack to boot_stack
*/
pmap_kenter(boot_stack, VM_PAGE_TO_PHYS(m[NPGPTD]));
xen_pgdpt_pin(VM_PAGE_TO_MACH(m[NPGPTD + 1]));
rw_wlock(&pvh_global_lock);
for (i = 0; i < 4; i++) {
int pdir = (PTDPTDI + i) / NPDEPG;
int curoffset = (PTDPTDI + i) % NPDEPG;
xen_queue_pt_update((vm_paddr_t)
((ma[pdir] & ~PG_V) + (curoffset*sizeof(vm_paddr_t))),
ma[i]);
}
PT_UPDATES_FLUSH();
rw_wunlock(&pvh_global_lock);
memset(&ctxt, 0, sizeof(ctxt));
ctxt.flags = VGCF_IN_KERNEL;
ctxt.user_regs.ds = GSEL(GDATA_SEL, SEL_KPL);
ctxt.user_regs.es = GSEL(GDATA_SEL, SEL_KPL);
ctxt.user_regs.fs = GSEL(GPRIV_SEL, SEL_KPL);
ctxt.user_regs.gs = GSEL(GDATA_SEL, SEL_KPL);
ctxt.user_regs.cs = GSEL(GCODE_SEL, SEL_KPL);
ctxt.user_regs.ss = GSEL(GDATA_SEL, SEL_KPL);
ctxt.user_regs.eip = (unsigned long)init_secondary;
ctxt.user_regs.eflags = PSL_KERNEL | 0x1000; /* IOPL_RING1 */
memset(&ctxt.fpu_ctxt, 0, sizeof(ctxt.fpu_ctxt));
smp_trap_init(ctxt.trap_ctxt);
ctxt.ldt_ents = 0;
ctxt.gdt_frames[0] = (uint32_t)((uint64_t)vtomach(bootAPgdt) >> PAGE_SHIFT);
ctxt.gdt_ents = 512;
#ifdef __i386__
ctxt.user_regs.esp = boot_stack + PAGE_SIZE;
ctxt.kernel_ss = GSEL(GDATA_SEL, SEL_KPL);
ctxt.kernel_sp = boot_stack + PAGE_SIZE;
ctxt.event_callback_cs = GSEL(GCODE_SEL, SEL_KPL);
ctxt.event_callback_eip = (unsigned long)Xhypervisor_callback;
ctxt.failsafe_callback_cs = GSEL(GCODE_SEL, SEL_KPL);
ctxt.failsafe_callback_eip = (unsigned long)failsafe_callback;
ctxt.ctrlreg[3] = VM_PAGE_TO_MACH(m[NPGPTD + 1]);
#else /* __x86_64__ */
ctxt.user_regs.esp = idle->thread.rsp0 - sizeof(struct pt_regs);
ctxt.kernel_ss = GSEL(GDATA_SEL, SEL_KPL);
ctxt.kernel_sp = idle->thread.rsp0;
ctxt.event_callback_eip = (unsigned long)hypervisor_callback;
ctxt.failsafe_callback_eip = (unsigned long)failsafe_callback;
ctxt.syscall_callback_eip = (unsigned long)system_call;
ctxt.ctrlreg[3] = xen_pfn_to_cr3(virt_to_mfn(init_level4_pgt));
ctxt.gs_base_kernel = (unsigned long)(cpu_pda(cpu));
#endif
printf("gdtpfn=%lx pdptpfn=%lx\n",
ctxt.gdt_frames[0],
ctxt.ctrlreg[3] >> PAGE_SHIFT);
PANIC_IF(HYPERVISOR_vcpu_op(VCPUOP_initialise, cpu, &ctxt));
DELAY(3000);
PANIC_IF(HYPERVISOR_vcpu_op(VCPUOP_up, cpu, NULL));
}
/*
* Define the code needed before returning to user mode, for trap and
* syscall.
*/
void
userret(struct thread *td, struct trapframe *frame)
{
struct proc *p = td->td_proc;
CTR3(KTR_SYSC, "userret: thread %p (pid %d, %s)", td, p->p_pid,
td->td_name);
KASSERT((p->p_flag & P_WEXIT) == 0,
("Exiting process returns to usermode"));
#if 0
#ifdef DIAGNOSTIC
/* Check that we called signotify() enough. */
PROC_LOCK(p);
thread_lock(td);
if (SIGPENDING(td) && ((td->td_flags & TDF_NEEDSIGCHK) == 0 ||
(td->td_flags & TDF_ASTPENDING) == 0))
printf("failed to set signal flags properly for ast()\n");
thread_unlock(td);
PROC_UNLOCK(p);
#endif
#endif
#ifdef KTRACE
KTRUSERRET(td);
#endif
/*
* If this thread tickled GEOM, we need to wait for the giggling to
* stop before we return to userland
*/
if (td->td_pflags & TDP_GEOM)
g_waitidle();
/*
* Charge system time if profiling.
*/
if (p->p_flag & P_PROFIL)
addupc_task(td, TRAPF_PC(frame), td->td_pticks * psratio);
/*
* Let the scheduler adjust our priority etc.
*/
sched_userret(td);
#ifdef XEN
PT_UPDATES_FLUSH();
#endif
/*
* Check for misbehavior.
*
* In case there is a callchain tracing ongoing because of
* hwpmc(4), skip the scheduler pinning check.
* hwpmc(4) subsystem, infact, will collect callchain informations
* at ast() checkpoint, which is past userret().
*/
WITNESS_WARN(WARN_PANIC, NULL, "userret: returning");
KASSERT(td->td_critnest == 0,
("userret: Returning in a critical section"));
KASSERT(td->td_locks == 0,
("userret: Returning with %d locks held", td->td_locks));
KASSERT(td->td_rw_rlocks == 0,
("userret: Returning with %d rwlocks held in read mode",
td->td_rw_rlocks));
KASSERT((td->td_pflags & TDP_NOFAULTING) == 0,
("userret: Returning with pagefaults disabled"));
KASSERT(td->td_no_sleeping == 0,
("userret: Returning with sleep disabled"));
KASSERT(td->td_pinned == 0 || (td->td_pflags & TDP_CALLCHAIN) != 0,
("userret: Returning with with pinned thread"));
KASSERT(td->td_vp_reserv == 0,
("userret: Returning while holding vnode reservation"));
KASSERT((td->td_flags & TDF_SBDRY) == 0,
("userret: Returning with stop signals deferred"));
#ifdef VIMAGE
/* Unfortunately td_vnet_lpush needs VNET_DEBUG. */
VNET_ASSERT(curvnet == NULL,
("%s: Returning on td %p (pid %d, %s) with vnet %p set in %s",
__func__, td, p->p_pid, td->td_name, curvnet,
(td->td_vnet_lpush != NULL) ? td->td_vnet_lpush : "N/A"));
#endif
#ifdef RACCT
PROC_LOCK(p);
while (p->p_throttled == 1)
msleep(p->p_racct, &p->p_mtx, 0, "racct", 0);
PROC_UNLOCK(p);
#endif
}
/* Full PV mode suspension. */
static void
xctrl_suspend()
{
int i, j, k, fpp, suspend_cancelled;
unsigned long max_pfn, start_info_mfn;
EVENTHANDLER_INVOKE(power_suspend);
#ifdef SMP
struct thread *td;
cpuset_t map;
u_int cpuid;
/*
* Bind us to CPU 0 and stop any other VCPUs.
*/
td = curthread;
thread_lock(td);
sched_bind(td, 0);
thread_unlock(td);
cpuid = PCPU_GET(cpuid);
KASSERT(cpuid == 0, ("xen_suspend: not running on cpu 0"));
map = all_cpus;
CPU_CLR(cpuid, &map);
CPU_NAND(&map, &stopped_cpus);
if (!CPU_EMPTY(&map))
stop_cpus(map);
#endif
/*
* Be sure to hold Giant across DEVICE_SUSPEND/RESUME since non-MPSAFE
* drivers need this.
*/
mtx_lock(&Giant);
if (DEVICE_SUSPEND(root_bus) != 0) {
mtx_unlock(&Giant);
printf("%s: device_suspend failed\n", __func__);
#ifdef SMP
if (!CPU_EMPTY(&map))
restart_cpus(map);
#endif
return;
}
mtx_unlock(&Giant);
local_irq_disable();
xencons_suspend();
gnttab_suspend();
intr_suspend();
max_pfn = HYPERVISOR_shared_info->arch.max_pfn;
void *shared_info = HYPERVISOR_shared_info;
HYPERVISOR_shared_info = NULL;
pmap_kremove((vm_offset_t) shared_info);
PT_UPDATES_FLUSH();
xen_start_info->store_mfn = MFNTOPFN(xen_start_info->store_mfn);
xen_start_info->console.domU.mfn = MFNTOPFN(xen_start_info->console.domU.mfn);
/*
* We'll stop somewhere inside this hypercall. When it returns,
* we'll start resuming after the restore.
*/
start_info_mfn = VTOMFN(xen_start_info);
pmap_suspend();
suspend_cancelled = HYPERVISOR_suspend(start_info_mfn);
pmap_resume();
pmap_kenter_ma((vm_offset_t) shared_info, xen_start_info->shared_info);
HYPERVISOR_shared_info = shared_info;
HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list =
VTOMFN(xen_pfn_to_mfn_frame_list_list);
fpp = PAGE_SIZE/sizeof(unsigned long);
for (i = 0, j = 0, k = -1; i < max_pfn; i += fpp, j++) {
if ((j % fpp) == 0) {
k++;
xen_pfn_to_mfn_frame_list_list[k] =
VTOMFN(xen_pfn_to_mfn_frame_list[k]);
j = 0;
}
xen_pfn_to_mfn_frame_list[k][j] =
VTOMFN(&xen_phys_machine[i]);
}
HYPERVISOR_shared_info->arch.max_pfn = max_pfn;
gnttab_resume();
intr_resume(suspend_cancelled != 0);
local_irq_enable();
xencons_resume();
#ifdef CONFIG_SMP
for_each_cpu(i)
vcpu_prepare(i);
#endif
/*
* Only resume xenbus /after/ we've prepared our VCPUs; otherwise
* the VCPU hotplug callback can race with our vcpu_prepare
*/
mtx_lock(&Giant);
DEVICE_RESUME(root_bus);
mtx_unlock(&Giant);
#ifdef SMP
thread_lock(curthread);
sched_unbind(curthread);
thread_unlock(curthread);
if (!CPU_EMPTY(&map))
restart_cpus(map);
#endif
EVENTHANDLER_INVOKE(power_resume);
}