static int
phys_pager_getpages(vm_object_t object, vm_page_t *m, int count, int reqpage)
{
int i, s;
s = splvm();
/*
* Fill as many pages as vm_fault has allocated for us.
*/
for (i = 0; i < count; i++) {
if ((m[i]->flags & PG_ZERO) == 0)
pmap_zero_page(m[i]);
vm_page_flag_set(m[i], PG_ZERO);
/* Switch off pv_entries */
vm_page_lock_queues();
vm_page_unmanage(m[i]);
vm_page_unlock_queues();
m[i]->valid = VM_PAGE_BITS_ALL;
m[i]->dirty = 0;
/* The requested page must remain busy, the others not. */
if (reqpage != i) {
vm_page_flag_clear(m[i], PG_BUSY);
m[i]->busy = 0;
}
}
splx(s);
return (VM_PAGER_OK);
}
vm_page_t
shmem_read_mapping_page(vm_object_t object, vm_pindex_t pindex)
{
vm_page_t m;
int rv;
VM_OBJECT_LOCK_ASSERT_OWNED(object);
m = vm_page_grab(object, pindex, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
if (m->valid != VM_PAGE_BITS_ALL) {
if (vm_pager_has_page(object, pindex)) {
rv = vm_pager_get_page(object, &m, 1);
m = vm_page_lookup(object, pindex);
if (m == NULL)
return ERR_PTR(-ENOMEM);
if (rv != VM_PAGER_OK) {
vm_page_free(m);
return ERR_PTR(-ENOMEM);
}
} else {
pmap_zero_page(VM_PAGE_TO_PHYS(m));
m->valid = VM_PAGE_BITS_ALL;
m->dirty = 0;
}
}
vm_page_wire(m);
vm_page_wakeup(m);
return (m);
}
/*
* Fill as many pages as vm_fault has allocated for us.
*/
static int
phys_pager_getpages(vm_object_t object, vm_page_t *m, int count, int reqpage)
{
int i;
VM_OBJECT_ASSERT_WLOCKED(object);
for (i = 0; i < count; i++) {
if (m[i]->valid == 0) {
if ((m[i]->flags & PG_ZERO) == 0)
pmap_zero_page(m[i]);
m[i]->valid = VM_PAGE_BITS_ALL;
}
KASSERT(m[i]->valid == VM_PAGE_BITS_ALL,
("phys_pager_getpages: partially valid page %p", m[i]));
KASSERT(m[i]->dirty == 0,
("phys_pager_getpages: dirty page %p", m[i]));
/* The requested page must remain busy, the others not. */
if (i == reqpage) {
vm_page_lock(m[i]);
vm_page_flash(m[i]);
vm_page_unlock(m[i]);
} else
vm_page_xunbusy(m[i]);
}
return (VM_PAGER_OK);
}
/*
* Fill as many pages as vm_fault has allocated for us.
*/
static int
phys_pager_getpages(vm_object_t object, vm_page_t *m, int count, int reqpage)
{
int i;
VM_OBJECT_LOCK_ASSERT(object, MA_OWNED);
for (i = 0; i < count; i++) {
if (m[i]->valid == 0) {
if ((m[i]->flags & PG_ZERO) == 0)
pmap_zero_page(m[i]);
m[i]->valid = VM_PAGE_BITS_ALL;
}
KASSERT(m[i]->valid == VM_PAGE_BITS_ALL,
("phys_pager_getpages: partially valid page %p", m[i]));
}
vm_page_lock_queues();
for (i = 0; i < count; i++) {
/* Switch off pv_entries */
vm_page_unmanage(m[i]);
m[i]->dirty = 0;
/* The requested page must remain busy, the others not. */
if (reqpage != i) {
vm_page_flag_clear(m[i], PG_BUSY);
m[i]->busy = 0;
}
}
vm_page_unlock_queues();
return (VM_PAGER_OK);
}
/*
* Allocates a region from the kernel address map and physically
* contiguous pages within the specified address range to the kernel
* object. Creates a wired mapping from this region to these pages, and
* returns the region's starting virtual address. If M_ZERO is specified
* through the given flags, then the pages are zeroed before they are
* mapped.
*/
vm_offset_t
kmem_alloc_contig_domain(int domain, vm_size_t size, int flags, vm_paddr_t low,
vm_paddr_t high, u_long alignment, vm_paddr_t boundary,
vm_memattr_t memattr)
{
vmem_t *vmem;
vm_object_t object = kernel_object;
vm_offset_t addr, offset, tmp;
vm_page_t end_m, m;
u_long npages;
int pflags, tries;
size = round_page(size);
vmem = vm_dom[domain].vmd_kernel_arena;
if (vmem_alloc(vmem, size, flags | M_BESTFIT, &addr))
return (0);
offset = addr - VM_MIN_KERNEL_ADDRESS;
pflags = malloc2vm_flags(flags) | VM_ALLOC_NOBUSY | VM_ALLOC_WIRED;
pflags &= ~(VM_ALLOC_NOWAIT | VM_ALLOC_WAITOK | VM_ALLOC_WAITFAIL);
pflags |= VM_ALLOC_NOWAIT;
npages = atop(size);
VM_OBJECT_WLOCK(object);
tries = 0;
retry:
m = vm_page_alloc_contig_domain(object, atop(offset), domain, pflags,
npages, low, high, alignment, boundary, memattr);
if (m == NULL) {
VM_OBJECT_WUNLOCK(object);
if (tries < ((flags & M_NOWAIT) != 0 ? 1 : 3)) {
if (!vm_page_reclaim_contig_domain(domain, pflags,
npages, low, high, alignment, boundary) &&
(flags & M_WAITOK) != 0)
vm_wait_domain(domain);
VM_OBJECT_WLOCK(object);
tries++;
goto retry;
}
vmem_free(vmem, addr, size);
return (0);
}
KASSERT(vm_phys_domain(m) == domain,
("kmem_alloc_contig_domain: Domain mismatch %d != %d",
vm_phys_domain(m), domain));
end_m = m + npages;
tmp = addr;
for (; m < end_m; m++) {
if ((flags & M_ZERO) && (m->flags & PG_ZERO) == 0)
pmap_zero_page(m);
m->valid = VM_PAGE_BITS_ALL;
pmap_enter(kernel_pmap, tmp, m, VM_PROT_ALL,
VM_PROT_ALL | PMAP_ENTER_WIRED, 0);
tmp += PAGE_SIZE;
}
VM_OBJECT_WUNLOCK(object);
return (addr);
}
/*
* kmem_back:
*
* Allocate physical pages for the specified virtual address range.
*/
int
kmem_back_domain(int domain, vm_object_t object, vm_offset_t addr,
vm_size_t size, int flags)
{
vm_offset_t offset, i;
vm_page_t m, mpred;
int pflags;
KASSERT(object == kernel_object,
("kmem_back_domain: only supports kernel object."));
offset = addr - VM_MIN_KERNEL_ADDRESS;
pflags = malloc2vm_flags(flags) | VM_ALLOC_NOBUSY | VM_ALLOC_WIRED;
pflags &= ~(VM_ALLOC_NOWAIT | VM_ALLOC_WAITOK | VM_ALLOC_WAITFAIL);
if (flags & M_WAITOK)
pflags |= VM_ALLOC_WAITFAIL;
i = 0;
VM_OBJECT_WLOCK(object);
retry:
mpred = vm_radix_lookup_le(&object->rtree, atop(offset + i));
for (; i < size; i += PAGE_SIZE, mpred = m) {
m = vm_page_alloc_domain_after(object, atop(offset + i),
domain, pflags, mpred);
/*
* Ran out of space, free everything up and return. Don't need
* to lock page queues here as we know that the pages we got
* aren't on any queues.
*/
if (m == NULL) {
if ((flags & M_NOWAIT) == 0)
goto retry;
VM_OBJECT_WUNLOCK(object);
kmem_unback(object, addr, i);
return (KERN_NO_SPACE);
}
KASSERT(vm_phys_domain(m) == domain,
("kmem_back_domain: Domain mismatch %d != %d",
vm_phys_domain(m), domain));
if (flags & M_ZERO && (m->flags & PG_ZERO) == 0)
pmap_zero_page(m);
KASSERT((m->oflags & VPO_UNMANAGED) != 0,
("kmem_malloc: page %p is managed", m));
m->valid = VM_PAGE_BITS_ALL;
pmap_enter(kernel_pmap, addr + i, m, VM_PROT_ALL,
VM_PROT_ALL | PMAP_ENTER_WIRED, 0);
}
VM_OBJECT_WUNLOCK(object);
return (KERN_SUCCESS);
}
/*
* Allocates a region from the kernel address map and physical pages
* within the specified address range to the kernel object. Creates a
* wired mapping from this region to these pages, and returns the
* region's starting virtual address. The allocated pages are not
* necessarily physically contiguous. If M_ZERO is specified through the
* given flags, then the pages are zeroed before they are mapped.
*/
vm_offset_t
kmem_alloc_attr_domain(int domain, vm_size_t size, int flags, vm_paddr_t low,
vm_paddr_t high, vm_memattr_t memattr)
{
vmem_t *vmem;
vm_object_t object = kernel_object;
vm_offset_t addr, i, offset;
vm_page_t m;
int pflags, tries;
size = round_page(size);
vmem = vm_dom[domain].vmd_kernel_arena;
if (vmem_alloc(vmem, size, M_BESTFIT | flags, &addr))
return (0);
offset = addr - VM_MIN_KERNEL_ADDRESS;
pflags = malloc2vm_flags(flags) | VM_ALLOC_NOBUSY | VM_ALLOC_WIRED;
pflags &= ~(VM_ALLOC_NOWAIT | VM_ALLOC_WAITOK | VM_ALLOC_WAITFAIL);
pflags |= VM_ALLOC_NOWAIT;
VM_OBJECT_WLOCK(object);
for (i = 0; i < size; i += PAGE_SIZE) {
tries = 0;
retry:
m = vm_page_alloc_contig_domain(object, atop(offset + i),
domain, pflags, 1, low, high, PAGE_SIZE, 0, memattr);
if (m == NULL) {
VM_OBJECT_WUNLOCK(object);
if (tries < ((flags & M_NOWAIT) != 0 ? 1 : 3)) {
if (!vm_page_reclaim_contig_domain(domain,
pflags, 1, low, high, PAGE_SIZE, 0) &&
(flags & M_WAITOK) != 0)
vm_wait_domain(domain);
VM_OBJECT_WLOCK(object);
tries++;
goto retry;
}
kmem_unback(object, addr, i);
vmem_free(vmem, addr, size);
return (0);
}
KASSERT(vm_phys_domain(m) == domain,
("kmem_alloc_attr_domain: Domain mismatch %d != %d",
vm_phys_domain(m), domain));
if ((flags & M_ZERO) && (m->flags & PG_ZERO) == 0)
pmap_zero_page(m);
m->valid = VM_PAGE_BITS_ALL;
pmap_enter(kernel_pmap, addr + i, m, VM_PROT_ALL,
VM_PROT_ALL | PMAP_ENTER_WIRED, 0);
}
VM_OBJECT_WUNLOCK(object);
return (addr);
}
void
kmem_init_zero_region(void)
{
vm_offset_t addr, i;
vm_page_t m;
/*
* Map a single physical page of zeros to a larger virtual range.
* This requires less looping in places that want large amounts of
* zeros, while not using much more physical resources.
*/
addr = kva_alloc(ZERO_REGION_SIZE);
m = vm_page_alloc(NULL, 0, VM_ALLOC_NORMAL |
VM_ALLOC_NOOBJ | VM_ALLOC_WIRED | VM_ALLOC_ZERO);
if ((m->flags & PG_ZERO) == 0)
pmap_zero_page(m);
for (i = 0; i < ZERO_REGION_SIZE; i += PAGE_SIZE)
pmap_qenter(addr + i, &m, 1);
pmap_protect(kernel_pmap, addr, addr + ZERO_REGION_SIZE, VM_PROT_READ);
zero_region = (const void *)addr;
}
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));
}
int
vm_fault_hold(vm_map_t map, vm_offset_t vaddr, vm_prot_t fault_type,
int fault_flags, vm_page_t *m_hold)
{
vm_prot_t prot;
long ahead, behind;
int alloc_req, era, faultcount, nera, reqpage, result;
boolean_t growstack, is_first_object_locked, wired;
int map_generation;
vm_object_t next_object;
vm_page_t marray[VM_FAULT_READ_MAX];
int hardfault;
struct faultstate fs;
struct vnode *vp;
int locked, error;
hardfault = 0;
growstack = TRUE;
PCPU_INC(cnt.v_vm_faults);
fs.vp = NULL;
faultcount = reqpage = 0;
RetryFault:;
/*
* Find the backing store object and offset into it to begin the
* search.
*/
fs.map = map;
result = vm_map_lookup(&fs.map, vaddr, fault_type, &fs.entry,
&fs.first_object, &fs.first_pindex, &prot, &wired);
if (result != KERN_SUCCESS) {
if (growstack && result == KERN_INVALID_ADDRESS &&
map != kernel_map) {
result = vm_map_growstack(curproc, vaddr);
if (result != KERN_SUCCESS)
return (KERN_FAILURE);
growstack = FALSE;
goto RetryFault;
}
return (result);
}
map_generation = fs.map->timestamp;
if (fs.entry->eflags & MAP_ENTRY_NOFAULT) {
panic("vm_fault: fault on nofault entry, addr: %lx",
(u_long)vaddr);
}
/*
* Make a reference to this object to prevent its disposal while we
* are messing with it. Once we have the reference, the map is free
* to be diddled. Since objects reference their shadows (and copies),
* they will stay around as well.
*
* Bump the paging-in-progress count to prevent size changes (e.g.
* truncation operations) during I/O. This must be done after
* obtaining the vnode lock in order to avoid possible deadlocks.
*/
VM_OBJECT_WLOCK(fs.first_object);
vm_object_reference_locked(fs.first_object);
vm_object_pip_add(fs.first_object, 1);
fs.lookup_still_valid = TRUE;
if (wired)
fault_type = prot | (fault_type & VM_PROT_COPY);
fs.first_m = NULL;
/*
* Search for the page at object/offset.
*/
fs.object = fs.first_object;
fs.pindex = fs.first_pindex;
while (TRUE) {
/*
* If the object is dead, we stop here
*/
if (fs.object->flags & OBJ_DEAD) {
unlock_and_deallocate(&fs);
return (KERN_PROTECTION_FAILURE);
}
/*
* See if page is resident
*/
fs.m = vm_page_lookup(fs.object, fs.pindex);
if (fs.m != NULL) {
/*
* check for page-based copy on write.
* We check fs.object == fs.first_object so
* as to ensure the legacy COW mechanism is
* used when the page in question is part of
* a shadow object. Otherwise, vm_page_cowfault()
* removes the page from the backing object,
* which is not what we want.
*/
vm_page_lock(fs.m);
if ((fs.m->cow) &&
(fault_type & VM_PROT_WRITE) &&
(fs.object == fs.first_object)) {
vm_page_cowfault(fs.m);
unlock_and_deallocate(&fs);
goto RetryFault;
}
/*
* Wait/Retry if the page is busy. We have to do this
* if the page is busy via either VPO_BUSY or
* vm_page_t->busy because the vm_pager may be using
* vm_page_t->busy for pageouts ( and even pageins if
* it is the vnode pager ), and we could end up trying
* to pagein and pageout the same page simultaneously.
*
* We can theoretically allow the busy case on a read
* fault if the page is marked valid, but since such
* pages are typically already pmap'd, putting that
* special case in might be more effort then it is
* worth. We cannot under any circumstances mess
* around with a vm_page_t->busy page except, perhaps,
* to pmap it.
*/
if ((fs.m->oflags & VPO_BUSY) || fs.m->busy) {
/*
* Reference the page before unlocking and
* sleeping so that the page daemon is less
* likely to reclaim it.
*/
vm_page_aflag_set(fs.m, PGA_REFERENCED);
vm_page_unlock(fs.m);
if (fs.object != fs.first_object) {
if (!VM_OBJECT_TRYWLOCK(
fs.first_object)) {
VM_OBJECT_WUNLOCK(fs.object);
VM_OBJECT_WLOCK(fs.first_object);
VM_OBJECT_WLOCK(fs.object);
}
vm_page_lock(fs.first_m);
vm_page_free(fs.first_m);
vm_page_unlock(fs.first_m);
vm_object_pip_wakeup(fs.first_object);
VM_OBJECT_WUNLOCK(fs.first_object);
fs.first_m = NULL;
}
unlock_map(&fs);
if (fs.m == vm_page_lookup(fs.object,
fs.pindex)) {
vm_page_sleep_if_busy(fs.m, TRUE,
"vmpfw");
}
vm_object_pip_wakeup(fs.object);
VM_OBJECT_WUNLOCK(fs.object);
PCPU_INC(cnt.v_intrans);
vm_object_deallocate(fs.first_object);
goto RetryFault;
}
vm_page_remque(fs.m);
vm_page_unlock(fs.m);
/*
* Mark page busy for other processes, and the
* pagedaemon. If it still isn't completely valid
* (readable), jump to readrest, else break-out ( we
* found the page ).
*/
vm_page_busy(fs.m);
if (fs.m->valid != VM_PAGE_BITS_ALL)
goto readrest;
break;
}
/*
* Page is not resident, If this is the search termination
* or the pager might contain the page, allocate a new page.
*/
if (TRYPAGER || fs.object == fs.first_object) {
if (fs.pindex >= fs.object->size) {
unlock_and_deallocate(&fs);
return (KERN_PROTECTION_FAILURE);
}
/*
* Allocate a new page for this object/offset pair.
*
* Unlocked read of the p_flag is harmless. At
* worst, the P_KILLED might be not observed
* there, and allocation can fail, causing
* restart and new reading of the p_flag.
*/
fs.m = NULL;
if (!vm_page_count_severe() || P_KILLED(curproc)) {
#if VM_NRESERVLEVEL > 0
if ((fs.object->flags & OBJ_COLORED) == 0) {
fs.object->flags |= OBJ_COLORED;
fs.object->pg_color = atop(vaddr) -
fs.pindex;
}
#endif
alloc_req = P_KILLED(curproc) ?
VM_ALLOC_SYSTEM : VM_ALLOC_NORMAL;
if (fs.object->type != OBJT_VNODE &&
fs.object->backing_object == NULL)
alloc_req |= VM_ALLOC_ZERO;
fs.m = vm_page_alloc(fs.object, fs.pindex,
alloc_req);
}
if (fs.m == NULL) {
unlock_and_deallocate(&fs);
VM_WAITPFAULT;
goto RetryFault;
} else if (fs.m->valid == VM_PAGE_BITS_ALL)
break;
}
readrest:
/*
* We have found a valid page or we have allocated a new page.
* The page thus may not be valid or may not be entirely
* valid.
*
* Attempt to fault-in the page if there is a chance that the
* pager has it, and potentially fault in additional pages
* at the same time.
*/
if (TRYPAGER) {
int rv;
u_char behavior = vm_map_entry_behavior(fs.entry);
if (behavior == MAP_ENTRY_BEHAV_RANDOM ||
P_KILLED(curproc)) {
behind = 0;
ahead = 0;
} else if (behavior == MAP_ENTRY_BEHAV_SEQUENTIAL) {
behind = 0;
ahead = atop(fs.entry->end - vaddr) - 1;
if (ahead > VM_FAULT_READ_AHEAD_MAX)
ahead = VM_FAULT_READ_AHEAD_MAX;
if (fs.pindex == fs.entry->next_read)
vm_fault_cache_behind(&fs,
VM_FAULT_READ_MAX);
} else {
/*
* If this is a sequential page fault, then
* arithmetically increase the number of pages
* in the read-ahead window. Otherwise, reset
* the read-ahead window to its smallest size.
*/
behind = atop(vaddr - fs.entry->start);
if (behind > VM_FAULT_READ_BEHIND)
behind = VM_FAULT_READ_BEHIND;
ahead = atop(fs.entry->end - vaddr) - 1;
era = fs.entry->read_ahead;
if (fs.pindex == fs.entry->next_read) {
nera = era + behind;
if (nera > VM_FAULT_READ_AHEAD_MAX)
nera = VM_FAULT_READ_AHEAD_MAX;
behind = 0;
if (ahead > nera)
ahead = nera;
if (era == VM_FAULT_READ_AHEAD_MAX)
vm_fault_cache_behind(&fs,
VM_FAULT_CACHE_BEHIND);
} else if (ahead > VM_FAULT_READ_AHEAD_MIN)
ahead = VM_FAULT_READ_AHEAD_MIN;
if (era != ahead)
fs.entry->read_ahead = ahead;
}
/*
* Call the pager to retrieve the data, if any, after
* releasing the lock on the map. We hold a ref on
* fs.object and the pages are VPO_BUSY'd.
*/
unlock_map(&fs);
if (fs.object->type == OBJT_VNODE) {
vp = fs.object->handle;
if (vp == fs.vp)
goto vnode_locked;
else if (fs.vp != NULL) {
vput(fs.vp);
fs.vp = NULL;
}
locked = VOP_ISLOCKED(vp);
if (locked != LK_EXCLUSIVE)
locked = LK_SHARED;
/* Do not sleep for vnode lock while fs.m is busy */
error = vget(vp, locked | LK_CANRECURSE |
LK_NOWAIT, curthread);
if (error != 0) {
vhold(vp);
release_page(&fs);
unlock_and_deallocate(&fs);
error = vget(vp, locked | LK_RETRY |
LK_CANRECURSE, curthread);
vdrop(vp);
fs.vp = vp;
KASSERT(error == 0,
("vm_fault: vget failed"));
goto RetryFault;
}
fs.vp = vp;
}
vnode_locked:
KASSERT(fs.vp == NULL || !fs.map->system_map,
("vm_fault: vnode-backed object mapped by system map"));
/*
* now we find out if any other pages should be paged
* in at this time this routine checks to see if the
* pages surrounding this fault reside in the same
* object as the page for this fault. If they do,
* then they are faulted in also into the object. The
* array "marray" returned contains an array of
* vm_page_t structs where one of them is the
* vm_page_t passed to the routine. The reqpage
* return value is the index into the marray for the
* vm_page_t passed to the routine.
*
* fs.m plus the additional pages are VPO_BUSY'd.
*/
faultcount = vm_fault_additional_pages(
fs.m, behind, ahead, marray, &reqpage);
rv = faultcount ?
vm_pager_get_pages(fs.object, marray, faultcount,
reqpage) : VM_PAGER_FAIL;
if (rv == VM_PAGER_OK) {
/*
* Found the page. Leave it busy while we play
* with it.
*/
/*
* Relookup in case pager changed page. Pager
* is responsible for disposition of old page
* if moved.
*/
fs.m = vm_page_lookup(fs.object, fs.pindex);
if (!fs.m) {
unlock_and_deallocate(&fs);
goto RetryFault;
}
hardfault++;
break; /* break to PAGE HAS BEEN FOUND */
}
/*
* Remove the bogus page (which does not exist at this
* object/offset); before doing so, we must get back
* our object lock to preserve our invariant.
*
* Also wake up any other process that may want to bring
* in this page.
*
* If this is the top-level object, we must leave the
* busy page to prevent another process from rushing
* past us, and inserting the page in that object at
* the same time that we are.
*/
if (rv == VM_PAGER_ERROR)
printf("vm_fault: pager read error, pid %d (%s)\n",
curproc->p_pid, curproc->p_comm);
/*
* Data outside the range of the pager or an I/O error
*/
/*
* XXX - the check for kernel_map is a kludge to work
* around having the machine panic on a kernel space
* fault w/ I/O error.
*/
if (((fs.map != kernel_map) && (rv == VM_PAGER_ERROR)) ||
(rv == VM_PAGER_BAD)) {
vm_page_lock(fs.m);
vm_page_free(fs.m);
vm_page_unlock(fs.m);
fs.m = NULL;
unlock_and_deallocate(&fs);
return ((rv == VM_PAGER_ERROR) ? KERN_FAILURE : KERN_PROTECTION_FAILURE);
}
if (fs.object != fs.first_object) {
vm_page_lock(fs.m);
vm_page_free(fs.m);
vm_page_unlock(fs.m);
fs.m = NULL;
/*
* XXX - we cannot just fall out at this
* point, m has been freed and is invalid!
*/
}
}
/*
* We get here if the object has default pager (or unwiring)
* or the pager doesn't have the page.
*/
if (fs.object == fs.first_object)
fs.first_m = fs.m;
/*
* Move on to the next object. Lock the next object before
* unlocking the current one.
*/
fs.pindex += OFF_TO_IDX(fs.object->backing_object_offset);
next_object = fs.object->backing_object;
if (next_object == NULL) {
/*
* If there's no object left, fill the page in the top
* object with zeros.
*/
if (fs.object != fs.first_object) {
vm_object_pip_wakeup(fs.object);
VM_OBJECT_WUNLOCK(fs.object);
fs.object = fs.first_object;
fs.pindex = fs.first_pindex;
fs.m = fs.first_m;
VM_OBJECT_WLOCK(fs.object);
}
fs.first_m = NULL;
/*
* Zero the page if necessary and mark it valid.
*/
if ((fs.m->flags & PG_ZERO) == 0) {
pmap_zero_page(fs.m);
} else {
PCPU_INC(cnt.v_ozfod);
}
PCPU_INC(cnt.v_zfod);
fs.m->valid = VM_PAGE_BITS_ALL;
break; /* break to PAGE HAS BEEN FOUND */
} else {
KASSERT(fs.object != next_object,
("object loop %p", next_object));
VM_OBJECT_WLOCK(next_object);
vm_object_pip_add(next_object, 1);
if (fs.object != fs.first_object)
vm_object_pip_wakeup(fs.object);
VM_OBJECT_WUNLOCK(fs.object);
fs.object = next_object;
}
}
KASSERT((fs.m->oflags & VPO_BUSY) != 0,
("vm_fault: not busy after main loop"));
/*
* PAGE HAS BEEN FOUND. [Loop invariant still holds -- the object lock
* is held.]
*/
/*
* If the page is being written, but isn't already owned by the
* top-level object, we have to copy it into a new page owned by the
* top-level object.
*/
if (fs.object != fs.first_object) {
/*
* We only really need to copy if we want to write it.
*/
if ((fault_type & (VM_PROT_COPY | VM_PROT_WRITE)) != 0) {
/*
* This allows pages to be virtually copied from a
* backing_object into the first_object, where the
* backing object has no other refs to it, and cannot
* gain any more refs. Instead of a bcopy, we just
* move the page from the backing object to the
* first object. Note that we must mark the page
* dirty in the first object so that it will go out
* to swap when needed.
*/
is_first_object_locked = FALSE;
if (
/*
* Only one shadow object
*/
(fs.object->shadow_count == 1) &&
/*
* No COW refs, except us
*/
(fs.object->ref_count == 1) &&
/*
* No one else can look this object up
*/
(fs.object->handle == NULL) &&
/*
* No other ways to look the object up
*/
((fs.object->type == OBJT_DEFAULT) ||
(fs.object->type == OBJT_SWAP)) &&
(is_first_object_locked = VM_OBJECT_TRYWLOCK(fs.first_object)) &&
/*
* We don't chase down the shadow chain
*/
fs.object == fs.first_object->backing_object) {
/*
* get rid of the unnecessary page
*/
vm_page_lock(fs.first_m);
vm_page_free(fs.first_m);
vm_page_unlock(fs.first_m);
/*
* grab the page and put it into the
* process'es object. The page is
* automatically made dirty.
*/
vm_page_lock(fs.m);
vm_page_rename(fs.m, fs.first_object, fs.first_pindex);
vm_page_unlock(fs.m);
vm_page_busy(fs.m);
fs.first_m = fs.m;
fs.m = NULL;
PCPU_INC(cnt.v_cow_optim);
} else {
/*
* Oh, well, lets copy it.
*/
pmap_copy_page(fs.m, fs.first_m);
fs.first_m->valid = VM_PAGE_BITS_ALL;
if (wired && (fault_flags &
VM_FAULT_CHANGE_WIRING) == 0) {
vm_page_lock(fs.first_m);
vm_page_wire(fs.first_m);
vm_page_unlock(fs.first_m);
vm_page_lock(fs.m);
vm_page_unwire(fs.m, FALSE);
vm_page_unlock(fs.m);
}
/*
* We no longer need the old page or object.
*/
release_page(&fs);
}
/*
* fs.object != fs.first_object due to above
* conditional
*/
vm_object_pip_wakeup(fs.object);
VM_OBJECT_WUNLOCK(fs.object);
/*
* Only use the new page below...
*/
fs.object = fs.first_object;
fs.pindex = fs.first_pindex;
fs.m = fs.first_m;
if (!is_first_object_locked)
VM_OBJECT_WLOCK(fs.object);
PCPU_INC(cnt.v_cow_faults);
curthread->td_cow++;
} else {
prot &= ~VM_PROT_WRITE;
}
}
/*
* We must verify that the maps have not changed since our last
* lookup.
*/
if (!fs.lookup_still_valid) {
vm_object_t retry_object;
vm_pindex_t retry_pindex;
vm_prot_t retry_prot;
if (!vm_map_trylock_read(fs.map)) {
release_page(&fs);
unlock_and_deallocate(&fs);
goto RetryFault;
}
fs.lookup_still_valid = TRUE;
if (fs.map->timestamp != map_generation) {
result = vm_map_lookup_locked(&fs.map, vaddr, fault_type,
&fs.entry, &retry_object, &retry_pindex, &retry_prot, &wired);
/*
* If we don't need the page any longer, put it on the inactive
* list (the easiest thing to do here). If no one needs it,
* pageout will grab it eventually.
*/
if (result != KERN_SUCCESS) {
release_page(&fs);
unlock_and_deallocate(&fs);
/*
* If retry of map lookup would have blocked then
* retry fault from start.
*/
if (result == KERN_FAILURE)
goto RetryFault;
return (result);
}
if ((retry_object != fs.first_object) ||
(retry_pindex != fs.first_pindex)) {
release_page(&fs);
unlock_and_deallocate(&fs);
goto RetryFault;
}
/*
* Check whether the protection has changed or the object has
* been copied while we left the map unlocked. Changing from
* read to write permission is OK - we leave the page
* write-protected, and catch the write fault. Changing from
* write to read permission means that we can't mark the page
* write-enabled after all.
*/
prot &= retry_prot;
}
}
/*
* If the page was filled by a pager, update the map entry's
* last read offset. Since the pager does not return the
* actual set of pages that it read, this update is based on
* the requested set. Typically, the requested and actual
* sets are the same.
*
* XXX The following assignment modifies the map
* without holding a write lock on it.
*/
if (hardfault)
fs.entry->next_read = fs.pindex + faultcount - reqpage;
if ((prot & VM_PROT_WRITE) != 0 ||
(fault_flags & VM_FAULT_DIRTY) != 0) {
vm_object_set_writeable_dirty(fs.object);
/*
* If this is a NOSYNC mmap we do not want to set VPO_NOSYNC
* if the page is already dirty to prevent data written with
* the expectation of being synced from not being synced.
* Likewise if this entry does not request NOSYNC then make
* sure the page isn't marked NOSYNC. Applications sharing
* data should use the same flags to avoid ping ponging.
*/
if (fs.entry->eflags & MAP_ENTRY_NOSYNC) {
if (fs.m->dirty == 0)
fs.m->oflags |= VPO_NOSYNC;
} else {
fs.m->oflags &= ~VPO_NOSYNC;
}
/*
* If the fault is a write, we know that this page is being
* written NOW so dirty it explicitly to save on
* pmap_is_modified() calls later.
*
* Also tell the backing pager, if any, that it should remove
* any swap backing since the page is now dirty.
*/
if (((fault_type & VM_PROT_WRITE) != 0 &&
(fault_flags & VM_FAULT_CHANGE_WIRING) == 0) ||
(fault_flags & VM_FAULT_DIRTY) != 0) {
vm_page_dirty(fs.m);
vm_pager_page_unswapped(fs.m);
}
}
/*
* Page had better still be busy
*/
KASSERT(fs.m->oflags & VPO_BUSY,
("vm_fault: page %p not busy!", fs.m));
/*
* Page must be completely valid or it is not fit to
* map into user space. vm_pager_get_pages() ensures this.
*/
KASSERT(fs.m->valid == VM_PAGE_BITS_ALL,
("vm_fault: page %p partially invalid", fs.m));
VM_OBJECT_WUNLOCK(fs.object);
/*
* Put this page into the physical map. We had to do the unlock above
* because pmap_enter() may sleep. We don't put the page
* back on the active queue until later so that the pageout daemon
* won't find it (yet).
*/
pmap_enter(fs.map->pmap, vaddr, fault_type, fs.m, prot, wired);
if ((fault_flags & VM_FAULT_CHANGE_WIRING) == 0 && wired == 0)
vm_fault_prefault(fs.map->pmap, vaddr, fs.entry);
VM_OBJECT_WLOCK(fs.object);
vm_page_lock(fs.m);
/*
* If the page is not wired down, then put it where the pageout daemon
* can find it.
*/
if (fault_flags & VM_FAULT_CHANGE_WIRING) {
if (wired)
vm_page_wire(fs.m);
else
vm_page_unwire(fs.m, 1);
} else
vm_page_activate(fs.m);
if (m_hold != NULL) {
*m_hold = fs.m;
vm_page_hold(fs.m);
}
vm_page_unlock(fs.m);
vm_page_wakeup(fs.m);
/*
* Unlock everything, and return
*/
unlock_and_deallocate(&fs);
if (hardfault) {
PCPU_INC(cnt.v_io_faults);
curthread->td_ru.ru_majflt++;
} else
curthread->td_ru.ru_minflt++;
return (KERN_SUCCESS);
}
/*
* This is now called from local media FS's to operate against their
* own vnodes if they fail to implement VOP_GETPAGES.
*/
int
vnode_pager_generic_getpages(struct vnode *vp, vm_page_t *m, int count,
int *a_rbehind, int *a_rahead, vop_getpages_iodone_t iodone, void *arg)
{
vm_object_t object;
struct bufobj *bo;
struct buf *bp;
off_t foff;
#ifdef INVARIANTS
off_t blkno0;
#endif
int bsize, pagesperblock, *freecnt;
int error, before, after, rbehind, rahead, poff, i;
int bytecount, secmask;
KASSERT(vp->v_type != VCHR && vp->v_type != VBLK,
("%s does not support devices", __func__));
if (vp->v_iflag & VI_DOOMED)
return (VM_PAGER_BAD);
object = vp->v_object;
foff = IDX_TO_OFF(m[0]->pindex);
bsize = vp->v_mount->mnt_stat.f_iosize;
pagesperblock = bsize / PAGE_SIZE;
KASSERT(foff < object->un_pager.vnp.vnp_size,
("%s: page %p offset beyond vp %p size", __func__, m[0], vp));
KASSERT(count <= sizeof(bp->b_pages),
("%s: requested %d pages", __func__, count));
/*
* The last page has valid blocks. Invalid part can only
* exist at the end of file, and the page is made fully valid
* by zeroing in vm_pager_get_pages().
*/
if (m[count - 1]->valid != 0 && --count == 0) {
if (iodone != NULL)
iodone(arg, m, 1, 0);
return (VM_PAGER_OK);
}
/*
* Synchronous and asynchronous paging operations use different
* free pbuf counters. This is done to avoid asynchronous requests
* to consume all pbufs.
* Allocate the pbuf at the very beginning of the function, so that
* if we are low on certain kind of pbufs don't even proceed to BMAP,
* but sleep.
*/
freecnt = iodone != NULL ?
&vnode_async_pbuf_freecnt : &vnode_pbuf_freecnt;
bp = getpbuf(freecnt);
/*
* Get the underlying device blocks for the file with VOP_BMAP().
* If the file system doesn't support VOP_BMAP, use old way of
* getting pages via VOP_READ.
*/
error = VOP_BMAP(vp, foff / bsize, &bo, &bp->b_blkno, &after, &before);
if (error == EOPNOTSUPP) {
relpbuf(bp, freecnt);
VM_OBJECT_WLOCK(object);
for (i = 0; i < count; i++) {
PCPU_INC(cnt.v_vnodein);
PCPU_INC(cnt.v_vnodepgsin);
error = vnode_pager_input_old(object, m[i]);
if (error)
break;
}
VM_OBJECT_WUNLOCK(object);
return (error);
} else if (error != 0) {
relpbuf(bp, freecnt);
return (VM_PAGER_ERROR);
}
/*
* If the file system supports BMAP, but blocksize is smaller
* than a page size, then use special small filesystem code.
*/
if (pagesperblock == 0) {
relpbuf(bp, freecnt);
for (i = 0; i < count; i++) {
PCPU_INC(cnt.v_vnodein);
PCPU_INC(cnt.v_vnodepgsin);
error = vnode_pager_input_smlfs(object, m[i]);
if (error)
break;
}
return (error);
}
/*
* A sparse file can be encountered only for a single page request,
* which may not be preceded by call to vm_pager_haspage().
*/
if (bp->b_blkno == -1) {
KASSERT(count == 1,
("%s: array[%d] request to a sparse file %p", __func__,
count, vp));
relpbuf(bp, freecnt);
pmap_zero_page(m[0]);
KASSERT(m[0]->dirty == 0, ("%s: page %p is dirty",
__func__, m[0]));
VM_OBJECT_WLOCK(object);
m[0]->valid = VM_PAGE_BITS_ALL;
VM_OBJECT_WUNLOCK(object);
return (VM_PAGER_OK);
}
#ifdef INVARIANTS
blkno0 = bp->b_blkno;
#endif
bp->b_blkno += (foff % bsize) / DEV_BSIZE;
/* Recalculate blocks available after/before to pages. */
poff = (foff % bsize) / PAGE_SIZE;
before *= pagesperblock;
before += poff;
after *= pagesperblock;
after += pagesperblock - (poff + 1);
if (m[0]->pindex + after >= object->size)
after = object->size - 1 - m[0]->pindex;
KASSERT(count <= after + 1, ("%s: %d pages asked, can do only %d",
__func__, count, after + 1));
after -= count - 1;
/* Trim requested rbehind/rahead to possible values. */
rbehind = a_rbehind ? *a_rbehind : 0;
rahead = a_rahead ? *a_rahead : 0;
rbehind = min(rbehind, before);
rbehind = min(rbehind, m[0]->pindex);
rahead = min(rahead, after);
rahead = min(rahead, object->size - m[count - 1]->pindex);
/*
* Check that total amount of pages fit into buf. Trim rbehind and
* rahead evenly if not.
*/
if (rbehind + rahead + count > nitems(bp->b_pages)) {
int trim, sum;
trim = rbehind + rahead + count - nitems(bp->b_pages) + 1;
sum = rbehind + rahead;
if (rbehind == before) {
/* Roundup rbehind trim to block size. */
rbehind -= roundup(trim * rbehind / sum, pagesperblock);
if (rbehind < 0)
rbehind = 0;
} else
rbehind -= trim * rbehind / sum;
rahead -= trim * rahead / sum;
}
KASSERT(rbehind + rahead + count <= nitems(bp->b_pages),
("%s: behind %d ahead %d count %d", __func__,
rbehind, rahead, count));
/*
* Fill in the bp->b_pages[] array with requested and optional
* read behind or read ahead pages. Read behind pages are looked
* up in a backward direction, down to a first cached page. Same
* for read ahead pages, but there is no need to shift the array
* in case of encountering a cached page.
*/
i = bp->b_npages = 0;
if (rbehind) {
vm_pindex_t startpindex, tpindex;
vm_page_t p;
VM_OBJECT_WLOCK(object);
startpindex = m[0]->pindex - rbehind;
if ((p = TAILQ_PREV(m[0], pglist, listq)) != NULL &&
p->pindex >= startpindex)
startpindex = p->pindex + 1;
/* tpindex is unsigned; beware of numeric underflow. */
for (tpindex = m[0]->pindex - 1;
tpindex >= startpindex && tpindex < m[0]->pindex;
tpindex--, i++) {
p = vm_page_alloc(object, tpindex, VM_ALLOC_NORMAL);
if (p == NULL) {
/* Shift the array. */
for (int j = 0; j < i; j++)
bp->b_pages[j] = bp->b_pages[j +
tpindex + 1 - startpindex];
break;
}
bp->b_pages[tpindex - startpindex] = p;
}
bp->b_pgbefore = i;
bp->b_npages += i;
bp->b_blkno -= IDX_TO_OFF(i) / DEV_BSIZE;
} else
bp->b_pgbefore = 0;
/* Requested pages. */
for (int j = 0; j < count; j++, i++)
bp->b_pages[i] = m[j];
bp->b_npages += count;
if (rahead) {
vm_pindex_t endpindex, tpindex;
vm_page_t p;
if (!VM_OBJECT_WOWNED(object))
VM_OBJECT_WLOCK(object);
endpindex = m[count - 1]->pindex + rahead + 1;
if ((p = TAILQ_NEXT(m[count - 1], listq)) != NULL &&
p->pindex < endpindex)
endpindex = p->pindex;
if (endpindex > object->size)
endpindex = object->size;
for (tpindex = m[count - 1]->pindex + 1;
tpindex < endpindex; i++, tpindex++) {
p = vm_page_alloc(object, tpindex, VM_ALLOC_NORMAL);
if (p == NULL)
break;
bp->b_pages[i] = p;
}
bp->b_pgafter = i - bp->b_npages;
bp->b_npages = i;
} else
bp->b_pgafter = 0;
if (VM_OBJECT_WOWNED(object))
VM_OBJECT_WUNLOCK(object);
/* Report back actual behind/ahead read. */
if (a_rbehind)
*a_rbehind = bp->b_pgbefore;
if (a_rahead)
*a_rahead = bp->b_pgafter;
#ifdef INVARIANTS
KASSERT(bp->b_npages <= nitems(bp->b_pages),
("%s: buf %p overflowed", __func__, bp));
for (int j = 1; j < bp->b_npages; j++)
KASSERT(bp->b_pages[j]->pindex - 1 ==
bp->b_pages[j - 1]->pindex,
("%s: pages array not consecutive, bp %p", __func__, bp));
#endif
/*
* Recalculate first offset and bytecount with regards to read behind.
* Truncate bytecount to vnode real size and round up physical size
* for real devices.
*/
foff = IDX_TO_OFF(bp->b_pages[0]->pindex);
bytecount = bp->b_npages << PAGE_SHIFT;
if ((foff + bytecount) > object->un_pager.vnp.vnp_size)
bytecount = object->un_pager.vnp.vnp_size - foff;
secmask = bo->bo_bsize - 1;
KASSERT(secmask < PAGE_SIZE && secmask > 0,
("%s: sector size %d too large", __func__, secmask + 1));
bytecount = (bytecount + secmask) & ~secmask;
/*
* And map the pages to be read into the kva, if the filesystem
* requires mapped buffers.
*/
if ((vp->v_mount->mnt_kern_flag & MNTK_UNMAPPED_BUFS) != 0 &&
unmapped_buf_allowed) {
bp->b_data = unmapped_buf;
bp->b_offset = 0;
} else {
bp->b_data = bp->b_kvabase;
pmap_qenter((vm_offset_t)bp->b_data, bp->b_pages, bp->b_npages);
}
/* Build a minimal buffer header. */
bp->b_iocmd = BIO_READ;
KASSERT(bp->b_rcred == NOCRED, ("leaking read ucred"));
KASSERT(bp->b_wcred == NOCRED, ("leaking write ucred"));
bp->b_rcred = crhold(curthread->td_ucred);
bp->b_wcred = crhold(curthread->td_ucred);
pbgetbo(bo, bp);
bp->b_vp = vp;
bp->b_bcount = bp->b_bufsize = bp->b_runningbufspace = bytecount;
bp->b_iooffset = dbtob(bp->b_blkno);
KASSERT(IDX_TO_OFF(m[0]->pindex - bp->b_pages[0]->pindex) ==
(blkno0 - bp->b_blkno) * DEV_BSIZE +
IDX_TO_OFF(m[0]->pindex) % bsize,
("wrong offsets bsize %d m[0] %ju b_pages[0] %ju "
"blkno0 %ju b_blkno %ju", bsize,
(uintmax_t)m[0]->pindex, (uintmax_t)bp->b_pages[0]->pindex,
(uintmax_t)blkno0, (uintmax_t)bp->b_blkno));
atomic_add_long(&runningbufspace, bp->b_runningbufspace);
PCPU_INC(cnt.v_vnodein);
PCPU_ADD(cnt.v_vnodepgsin, bp->b_npages);
if (iodone != NULL) { /* async */
bp->b_pgiodone = iodone;
bp->b_caller1 = arg;
bp->b_iodone = vnode_pager_generic_getpages_done_async;
bp->b_flags |= B_ASYNC;
BUF_KERNPROC(bp);
bstrategy(bp);
return (VM_PAGER_OK);
} else {
bp->b_iodone = bdone;
bstrategy(bp);
bwait(bp, PVM, "vnread");
error = vnode_pager_generic_getpages_done(bp);
for (i = 0; i < bp->b_npages; i++)
bp->b_pages[i] = NULL;
bp->b_vp = NULL;
pbrelbo(bp);
relpbuf(bp, &vnode_pbuf_freecnt);
return (error != 0 ? VM_PAGER_ERROR : VM_PAGER_OK);
}
}
/*
* vm_contig_pg_kmap:
*
* Map previously allocated (vm_contig_pg_alloc) range of pages from
* vm_page_array[] into the KVA. Once mapped, the pages are part of
* the Kernel, and are to free'ed with kmem_free(&kernel_map, addr, size).
*
* No requirements.
*/
vm_offset_t
vm_contig_pg_kmap(int start, u_long size, vm_map_t map, int flags)
{
vm_offset_t addr, tmp_addr;
vm_page_t pga = vm_page_array;
int i, count;
size = round_page(size);
if (size == 0)
panic("vm_contig_pg_kmap: size must not be 0");
crit_enter();
lwkt_gettoken(&vm_token);
/*
* We've found a contiguous chunk that meets our requirements.
* Allocate KVM, and assign phys pages and return a kernel VM
* pointer.
*/
count = vm_map_entry_reserve(MAP_RESERVE_COUNT);
vm_map_lock(map);
if (vm_map_findspace(map, vm_map_min(map), size, PAGE_SIZE, 0, &addr) !=
KERN_SUCCESS) {
/*
* XXX We almost never run out of kernel virtual
* space, so we don't make the allocated memory
* above available.
*/
vm_map_unlock(map);
vm_map_entry_release(count);
lwkt_reltoken(&vm_token);
crit_exit();
return (0);
}
/*
* kernel_object maps 1:1 to kernel_map.
*/
vm_object_hold(&kernel_object);
vm_object_reference(&kernel_object);
vm_map_insert(map, &count,
&kernel_object, addr,
addr, addr + size,
VM_MAPTYPE_NORMAL,
VM_PROT_ALL, VM_PROT_ALL,
0);
vm_map_unlock(map);
vm_map_entry_release(count);
tmp_addr = addr;
for (i = start; i < (start + size / PAGE_SIZE); i++) {
vm_page_t m = &pga[i];
vm_page_insert(m, &kernel_object, OFF_TO_IDX(tmp_addr));
if ((flags & M_ZERO) && !(m->flags & PG_ZERO))
pmap_zero_page(VM_PAGE_TO_PHYS(m));
m->flags = 0;
tmp_addr += PAGE_SIZE;
}
vm_map_wire(map, addr, addr + size, 0);
vm_object_drop(&kernel_object);
lwkt_reltoken(&vm_token);
crit_exit();
return (addr);
}
/*
* This is now called from local media FS's to operate against their
* own vnodes if they fail to implement VOP_GETPAGES.
*/
int
vnode_pager_generic_getpages(struct vnode *vp, vm_page_t *m, int bytecount,
int reqpage, vop_getpages_iodone_t iodone, void *arg)
{
vm_object_t object;
struct bufobj *bo;
struct buf *bp;
daddr_t firstaddr, reqblock;
off_t foff, pib;
int pbefore, pafter, i, size, bsize, first, last, *freecnt;
int count, error, before, after, secmask;
KASSERT(vp->v_type != VCHR && vp->v_type != VBLK,
("vnode_pager_generic_getpages does not support devices"));
if (vp->v_iflag & VI_DOOMED)
return (VM_PAGER_BAD);
object = vp->v_object;
count = bytecount / PAGE_SIZE;
bsize = vp->v_mount->mnt_stat.f_iosize;
/*
* Synchronous and asynchronous paging operations use different
* free pbuf counters. This is done to avoid asynchronous requests
* to consume all pbufs.
* Allocate the pbuf at the very beginning of the function, so that
* if we are low on certain kind of pbufs don't even proceed to BMAP,
* but sleep.
*/
freecnt = iodone != NULL ?
&vnode_async_pbuf_freecnt : &vnode_pbuf_freecnt;
bp = getpbuf(freecnt);
/*
* Get the underlying device blocks for the file with VOP_BMAP().
* If the file system doesn't support VOP_BMAP, use old way of
* getting pages via VOP_READ.
*/
error = VOP_BMAP(vp, IDX_TO_OFF(m[reqpage]->pindex) / bsize, &bo,
&reqblock, &after, &before);
if (error == EOPNOTSUPP) {
relpbuf(bp, freecnt);
VM_OBJECT_WLOCK(object);
for (i = 0; i < count; i++)
if (i != reqpage) {
vm_page_lock(m[i]);
vm_page_free(m[i]);
vm_page_unlock(m[i]);
}
PCPU_INC(cnt.v_vnodein);
PCPU_INC(cnt.v_vnodepgsin);
error = vnode_pager_input_old(object, m[reqpage]);
VM_OBJECT_WUNLOCK(object);
return (error);
} else if (error != 0) {
relpbuf(bp, freecnt);
vm_pager_free_nonreq(object, m, reqpage, count, FALSE);
return (VM_PAGER_ERROR);
/*
* If the blocksize is smaller than a page size, then use
* special small filesystem code.
*/
} else if ((PAGE_SIZE / bsize) > 1) {
relpbuf(bp, freecnt);
vm_pager_free_nonreq(object, m, reqpage, count, FALSE);
PCPU_INC(cnt.v_vnodein);
PCPU_INC(cnt.v_vnodepgsin);
return (vnode_pager_input_smlfs(object, m[reqpage]));
}
/*
* Since the caller has busied the requested page, that page's valid
* field will not be changed by other threads.
*/
vm_page_assert_xbusied(m[reqpage]);
/*
* If we have a completely valid page available to us, we can
* clean up and return. Otherwise we have to re-read the
* media.
*/
if (m[reqpage]->valid == VM_PAGE_BITS_ALL) {
relpbuf(bp, freecnt);
vm_pager_free_nonreq(object, m, reqpage, count, FALSE);
return (VM_PAGER_OK);
} else if (reqblock == -1) {
relpbuf(bp, freecnt);
pmap_zero_page(m[reqpage]);
KASSERT(m[reqpage]->dirty == 0,
("vnode_pager_generic_getpages: page %p is dirty", m));
VM_OBJECT_WLOCK(object);
m[reqpage]->valid = VM_PAGE_BITS_ALL;
vm_pager_free_nonreq(object, m, reqpage, count, TRUE);
VM_OBJECT_WUNLOCK(object);
return (VM_PAGER_OK);
} else if (m[reqpage]->valid != 0) {
VM_OBJECT_WLOCK(object);
m[reqpage]->valid = 0;
VM_OBJECT_WUNLOCK(object);
}
pib = IDX_TO_OFF(m[reqpage]->pindex) % bsize;
pbefore = ((daddr_t)before * bsize + pib) / PAGE_SIZE;
pafter = ((daddr_t)(after + 1) * bsize - pib) / PAGE_SIZE - 1;
first = reqpage < pbefore ? 0 : reqpage - pbefore;
last = reqpage + pafter >= count ? count - 1 : reqpage + pafter;
if (first > 0 || last + 1 < count) {
VM_OBJECT_WLOCK(object);
for (i = 0; i < first; i++) {
vm_page_lock(m[i]);
vm_page_free(m[i]);
vm_page_unlock(m[i]);
}
for (i = last + 1; i < count; i++) {
vm_page_lock(m[i]);
vm_page_free(m[i]);
vm_page_unlock(m[i]);
}
VM_OBJECT_WUNLOCK(object);
}
/*
* here on direct device I/O
*/
firstaddr = reqblock;
firstaddr += pib / DEV_BSIZE;
firstaddr -= IDX_TO_OFF(reqpage - first) / DEV_BSIZE;
/*
* The first and last page have been calculated now, move
* input pages to be zero based, and adjust the count.
*/
m += first;
reqpage -= first;
count = last - first + 1;
/*
* calculate the file virtual address for the transfer
*/
foff = IDX_TO_OFF(m[0]->pindex);
/*
* calculate the size of the transfer
*/
size = count * PAGE_SIZE;
KASSERT(count > 0, ("zero count"));
if ((foff + size) > object->un_pager.vnp.vnp_size)
size = object->un_pager.vnp.vnp_size - foff;
KASSERT(size > 0, ("zero size"));
/*
* round up physical size for real devices.
*/
secmask = bo->bo_bsize - 1;
KASSERT(secmask < PAGE_SIZE && secmask > 0,
("vnode_pager_generic_getpages: sector size %d too large",
secmask + 1));
size = (size + secmask) & ~secmask;
/*
* and map the pages to be read into the kva, if the filesystem
* requires mapped buffers.
*/
if ((vp->v_mount->mnt_kern_flag & MNTK_UNMAPPED_BUFS) != 0 &&
unmapped_buf_allowed) {
bp->b_data = unmapped_buf;
bp->b_offset = 0;
} else {
bp->b_data = bp->b_kvabase;
pmap_qenter((vm_offset_t)bp->b_data, m, count);
}
/* build a minimal buffer header */
bp->b_iocmd = BIO_READ;
KASSERT(bp->b_rcred == NOCRED, ("leaking read ucred"));
KASSERT(bp->b_wcred == NOCRED, ("leaking write ucred"));
bp->b_rcred = crhold(curthread->td_ucred);
bp->b_wcred = crhold(curthread->td_ucred);
bp->b_blkno = firstaddr;
pbgetbo(bo, bp);
bp->b_vp = vp;
bp->b_bcount = size;
bp->b_bufsize = size;
bp->b_runningbufspace = bp->b_bufsize;
for (i = 0; i < count; i++)
bp->b_pages[i] = m[i];
bp->b_npages = count;
bp->b_pager.pg_reqpage = reqpage;
atomic_add_long(&runningbufspace, bp->b_runningbufspace);
PCPU_INC(cnt.v_vnodein);
PCPU_ADD(cnt.v_vnodepgsin, count);
/* do the input */
bp->b_iooffset = dbtob(bp->b_blkno);
if (iodone != NULL) { /* async */
bp->b_pager.pg_iodone = iodone;
bp->b_caller1 = arg;
bp->b_iodone = vnode_pager_generic_getpages_done_async;
bp->b_flags |= B_ASYNC;
BUF_KERNPROC(bp);
bstrategy(bp);
/* Good bye! */
} else {
bp->b_iodone = bdone;
bstrategy(bp);
bwait(bp, PVM, "vnread");
error = vnode_pager_generic_getpages_done(bp);
for (i = 0; i < bp->b_npages; i++)
bp->b_pages[i] = NULL;
bp->b_vp = NULL;
pbrelbo(bp);
relpbuf(bp, &vnode_pbuf_freecnt);
}
return (error != 0 ? VM_PAGER_ERROR : VM_PAGER_OK);
}
