Implement 'fast path' for the vm page fault handler. Or, it could be

called a scalable path.  When several preconditions hold, the vm
object lock for the object containing the faulted page is taken in
read mode, instead of write, which allows parallel faults processing
in the region.

Namely, the fast path is taken when the faulted page already exists
and does not need copy on write, is already fully valid, and not busy.
For technical reasons, fast path is avoided when the fault is the
first write on the vnode object, or when the fault is for wiring or
debugger read or write.

On the fast path, pmap_enter(9) is passed the PMAP_ENTER_NOSLEEP flag,
since object lock is kept.  Pmap might fail to create the entry, in
which case the fallback to slow path is performed.

Reviewed by:	alc
Tested by:	pho (previous version)
Hardware provided and hosted by:	The FreeBSD Foundation and
	 Sentex Data Communications
Sponsored by:	The FreeBSD Foundation
MFC after:	2 week
This commit is contained in:
kib 2014-08-15 07:30:14 +00:00
parent 45bdeab3db
commit ff74afd0bd

View File

@ -237,6 +237,7 @@ vm_fault_hold(vm_map_t map, vm_offset_t vaddr, vm_prot_t fault_type,
int hardfault;
struct faultstate fs;
struct vnode *vp;
vm_page_t m;
int locked, error;
hardfault = 0;
@ -290,6 +291,55 @@ RetryFault:;
goto RetryFault;
}
if (wired)
fault_type = prot | (fault_type & VM_PROT_COPY);
if (fs.vp == NULL /* avoid locked vnode leak */ &&
(fault_flags & (VM_FAULT_CHANGE_WIRING | VM_FAULT_DIRTY)) == 0 &&
/* avoid calling vm_object_set_writeable_dirty() */
((prot & VM_PROT_WRITE) == 0 ||
fs.first_object->type != OBJT_VNODE ||
(fs.first_object->flags & OBJ_MIGHTBEDIRTY) != 0)) {
VM_OBJECT_RLOCK(fs.first_object);
if ((prot & VM_PROT_WRITE) != 0 &&
fs.first_object->type == OBJT_VNODE &&
(fs.first_object->flags & OBJ_MIGHTBEDIRTY) == 0)
goto fast_failed;
m = vm_page_lookup(fs.first_object, fs.first_pindex);
if (m == NULL || vm_page_busied(m) ||
m->valid != VM_PAGE_BITS_ALL)
goto fast_failed;
result = pmap_enter(fs.map->pmap, vaddr, m, prot,
fault_type | PMAP_ENTER_NOSLEEP | (wired ? PMAP_ENTER_WIRED :
0), 0);
if (result != KERN_SUCCESS)
goto fast_failed;
if (m_hold != NULL) {
*m_hold = m;
vm_page_lock(m);
vm_page_hold(m);
vm_page_unlock(m);
}
if ((fault_type & VM_PROT_WRITE) != 0 &&
(m->oflags & VPO_UNMANAGED) == 0) {
vm_page_dirty(m);
vm_pager_page_unswapped(m);
}
VM_OBJECT_RUNLOCK(fs.first_object);
if (!wired)
vm_fault_prefault(&fs, vaddr, 0, 0);
vm_map_lookup_done(fs.map, fs.entry);
curthread->td_ru.ru_minflt++;
return (KERN_SUCCESS);
fast_failed:
if (!VM_OBJECT_TRYUPGRADE(fs.first_object)) {
VM_OBJECT_RUNLOCK(fs.first_object);
VM_OBJECT_WLOCK(fs.first_object);
}
} else {
VM_OBJECT_WLOCK(fs.first_object);
}
/*
* 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
@ -300,15 +350,11 @@ RetryFault:;
* 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;
/*