Implement an approximation of Linux MADV_DONTNEED semantics.

Linux MADV_DONTNEED is not advisory: it has side effects for anonymous
memory, and some system software depends on that.  In particular,
MADV_DONTNEED causes anonymous pages to be discarded.  If the mapping is
a private mapping of a named object then subsequent faults are to
repopulate the range from that object, otherwise pages will be
zero-filled.  For mappings of non-anonymous objects, Linux MADV_DONTNEED
can be implemented in the same way as our MADV_DONTNEED.

This implementation differs from Linux semantics in its handling of
private mappings, inherited through fork(), of non-anonymous objects.
After applying MADV_DONTNEED, subsequent faults will repopulate the
mapping from the parent object rather than the root of the shadow chain.

PR:		230160
Reviewed by:	alc, kib
Sponsored by:	The FreeBSD Foundation
MFC after:	2 weeks
Differential Revision:	https://reviews.freebsd.org/D25330
This commit is contained in:
Mark Johnston 2020-06-25 20:30:30 +00:00
parent 701267ad19
commit f4134e3d87

View File

@ -38,9 +38,11 @@ __FBSDID("$FreeBSD$");
#include <sys/file.h>
#include <sys/imgact.h>
#include <sys/ktr.h>
#include <sys/lock.h>
#include <sys/mman.h>
#include <sys/proc.h>
#include <sys/resourcevar.h>
#include <sys/rwlock.h>
#include <sys/syscallsubr.h>
#include <sys/sysent.h>
#include <sys/sysproto.h>
@ -48,6 +50,7 @@ __FBSDID("$FreeBSD$");
#include <vm/pmap.h>
#include <vm/vm_extern.h>
#include <vm/vm_map.h>
#include <vm/vm_object.h>
#include <compat/linux/linux_emul.h>
#include <compat/linux/linux_mmap.h>
@ -242,6 +245,98 @@ linux_mprotect_common(struct thread *td, uintptr_t addr, size_t len, int prot)
return (kern_mprotect(td, addr, len, prot));
}
/*
* Implement Linux madvise(MADV_DONTNEED), which has unusual semantics: for
* anonymous memory, pages in the range are immediately discarded.
*/
static int
linux_madvise_dontneed(struct thread *td, vm_offset_t start, vm_offset_t end)
{
vm_map_t map;
vm_map_entry_t entry;
vm_object_t backing_object, object;
vm_offset_t estart, eend;
vm_pindex_t pstart, pend;
int error;
map = &td->td_proc->p_vmspace->vm_map;
if (!vm_map_range_valid(map, start, end))
return (EINVAL);
start = trunc_page(start);
end = round_page(end);
error = 0;
vm_map_lock_read(map);
if (!vm_map_lookup_entry(map, start, &entry))
entry = vm_map_entry_succ(entry);
for (; entry->start < end; entry = vm_map_entry_succ(entry)) {
if ((entry->eflags & MAP_ENTRY_IS_SUB_MAP) != 0)
continue;
if (entry->wired_count != 0) {
error = EINVAL;
break;
}
object = entry->object.vm_object;
if (object == NULL)
continue;
pstart = OFF_TO_IDX(entry->offset);
if (start > entry->start) {
pstart += atop(start - entry->start);
estart = start;
} else {
estart = entry->start;
}
pend = OFF_TO_IDX(entry->offset) +
atop(entry->end - entry->start);
if (entry->end > end) {
pend -= atop(entry->end - end);
eend = end;
} else {
eend = entry->end;
}
if ((object->flags & (OBJ_ANON | OBJ_ONEMAPPING)) ==
(OBJ_ANON | OBJ_ONEMAPPING)) {
/*
* Singly-mapped anonymous memory is discarded. This
* does not match Linux's semantics when the object
* belongs to a shadow chain of length > 1, since
* subsequent faults may retrieve pages from an
* intermediate anonymous object. However, handling
* this case correctly introduces a fair bit of
* complexity.
*/
VM_OBJECT_WLOCK(object);
if ((object->flags & OBJ_ONEMAPPING) != 0) {
vm_object_collapse(object);
vm_object_page_remove(object, pstart, pend, 0);
backing_object = object->backing_object;
if (backing_object != NULL &&
(backing_object->flags & OBJ_ANON) != 0)
linux_msg(td,
"possibly incorrect MADV_DONTNEED");
VM_OBJECT_WUNLOCK(object);
continue;
}
VM_OBJECT_WUNLOCK(object);
}
/*
* Handle shared mappings. Remove them outright instead of
* calling pmap_advise(), for consistency with Linux.
*/
pmap_remove(map->pmap, estart, eend);
vm_object_madvise(object, pstart, pend, MADV_DONTNEED);
}
vm_map_unlock_read(map);
return (error);
}
int
linux_madvise_common(struct thread *td, uintptr_t addr, size_t len, int behav)
{
@ -256,7 +351,7 @@ linux_madvise_common(struct thread *td, uintptr_t addr, size_t len, int behav)
case LINUX_MADV_WILLNEED:
return (kern_madvise(td, addr, len, MADV_WILLNEED));
case LINUX_MADV_DONTNEED:
return (kern_madvise(td, addr, len, MADV_DONTNEED));
return (linux_madvise_dontneed(td, addr, addr + len));
case LINUX_MADV_FREE:
return (kern_madvise(td, addr, len, MADV_FREE));
case LINUX_MADV_REMOVE: