We currently initialize the vm_page array in three passes: one to zero
the array, one to initialize the "order" field of each page (necessary
when inserting them into the vm_phys buddy allocator one-by-one), and
one to initialize the remaining non-zero fields and individually insert
each page into the allocator.
Merge the three passes into one following a suggestion from alc:
initialize vm_page fields in a single pass, and use vm_phys_free_contig()
to efficiently insert physical memory segments into the buddy allocator.
This reduces the initialization time to a third or a quarter of what it
was before on most systems that I tested.
Reviewed by: alc, kib
MFC after: 3 weeks
Differential Revision: https://reviews.freebsd.org/D12248
While these locks are guarnteed to not share their respective cache lines,
their current placement leaves unnecessary holes in lines which preceeded them.
For instance the annotation of vm_page_queue_free_mtx allows 2 neighbour
cachelines (previously separate by the lock) to be collapsed into 1.
The annotation is only effective on architectures which have it implemented in
their linker script (currently only amd64). Thus locks are not converted to
their not-padaligned variants as to not affect the rest.
MFC after: 1 week
In swp_pager_meta_build(), if the requested operation results in
freeing the last swap pointer in the swblk, free the trie node. Other
swap pager code does not expect to find completely empty swblk.
Reviewed by: alc, markj
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
swapblk for our index while we dropped the object lock.
Noted by: jeff
Reviewed by: alc, markj
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
The function return value is not used. Its argument is always
swap_total/PAGE_SIZE, so make it not take any arguments.
Submitted by: ota@j.email.ne.jp
PR: 221356
MFC after: 1 week
Before r207410, the hold count of a page in a page queue was protected
by the queue lock, and, before laundering a page, the page daemon
removed managed writeable mappings of the page before releasing the
queue lock. This ensured that other threads could not concurrently
create transient writeable mappings using pmap_extract_and_hold() on a
user map, as is done for example by vmapbuf(). With that revision,
however, a race can allow the creation of such a mapping, meaning that
the page might be modified as it is being laundered, potentially
resulting in it being marked clean when its contents do not match
those given to the pager. Close the race by using the page lock to
synchronize the hold count check in vm_pageout_cluster() with the
removal of writeable managed mappings.
Reported by: alc
Reviewed by: alc, kib
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D12084
new use of the vm_object's lock to synchronize updates to a radix trie
mapping per-vm object page indices to on-disk swap blocks.
Fix a typo in a nearby comment.
Reviewed by: kib, markj
X-MFC with: r322913
Differential Revision: https://reviews.freebsd.org/D12134
vm_object page indices to on-disk swap space (r322913) has changed the
synchronization requirements for a couple swap pager functions. Whereas
before a read lock on the vm object sufficed because of the global mutex
on the hash table, a write lock on the vm object may now be required. In
particular, calls to vm_pager_page_unswapped() now require a write lock on
the vm_object. Consequently, vm_fault()'s fast path cannot call
vm_pager_page_unswapped(). The swap space will have to be released at a
later point.
Reviewed by: kib, markj
X-MFC with: r322913
Differential Revision: https://reviews.freebsd.org/D12134
blocks assigned to the object pages.
- The global swhash_mtx is removed, trie is synchronized by the
corresponding object lock.
- The swp_pager_meta_free_all() function used during object
termination is optimized by only looking at the trie instead of
having to search whole hash for the swap blocks owned by the object.
- On swap_pager_swapoff(), instead of iterating over the swhash,
global object list have to be inspected. There, we have to ensure
that we do see valid trie content if we see that the object type is
swap.
Sizing of the swblk zone is same as for swblock zone, each swblk maps
SWAP_META_PAGES pages.
Proposed by: alc
Reviewed by: alc, markj (previous version)
Tested by: alc, pho (previous version)
Sponsored by: The FreeBSD Foundation
MFC after: 1 month
Differential revision: https://reviews.freebsd.org/D11435
Setting this flag allows us to skip pages removal from VM object queue
during object termination and to leave that for cdev_pg_dtor function.
Move pages removal code to separate function vm_object_terminate_pages()
as comments does not survive indentation.
This will be required for Intel SGX support where we will have to remove
pages from VM object manually.
Reviewed by: kib, alc
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D11688
This is a variant of vm_page_alloc() which accepts an additional parameter:
the page in the object with largest index that is smaller than the requested
index. vm_page_alloc() finds this page using a lookup in the object's radix
tree, but in some cases its identity is already known, allowing the lookup
to be elided.
Modify kmem_back() and vm_page_grab_pages() to use vm_page_alloc_after().
vm_page_alloc() is converted into a trivial wrapper of
vm_page_alloc_after().
Suggested by: alc
Reviewed by: alc, kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D11984
We can remove some unnecessary object radix tree lookups by using the
object memq to iterate over pages in the specified range. This does not,
however, eliminate the lookup needed in vm_page_free_toq() to remove each
tree entry.
Reviewed by: alc, kib (previous revision)
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D11945
vm_page_grab() on consecutive page indices. Besides simplifying the code
in the caller, vm_page_grab_pages() allows for batching optimizations.
For example, the current implementation replaces calls to vm_page_lookup()
on consecutive page indices by cheaper calls to vm_page_next().
Reviewed by: kib, markj
Tested by: pho (an earlier version)
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D11926
Suppose that a file on NFS has partially filled last page, and this
page is dirty. NFS VOP_PAGEOUT() method only marks the the page clean
up to the block of the last written byte, leaving other blocks dirty.
Also any page which erronously exists in the vnode vm_object past EOF
is also left marked as dirty.
With the introduction of the buf-cache coherent pager, each pass of
syncer over the object with such page results in creation of B_DELWRI
buffer due to VOP_WRITE() call. This buffer is noted on next syncer
pass, which results e.g. a visible manifestation of shutdown never
finishing vnode sync. Note that before buf-cache coherency commit, a
dirty page might left never synced to server if a partial writes
occur.
Fix this by clearing dirty bits after EOF. Only blocks of the partial
page which are completely after EOF are marked clean, to avoid
possible user data loss.
Reported by: mav
Reviewed by: alc, markj
Tested by: mav, pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D11697
Differential Revision discusses the benefits of this change.)
Add a function, vm_reserv_to_superpage(), that returns the superpage
containing the specified base page.
Reviewed by: kib, markj
Tested by: pho
MFC after: 10 days
Differential Revision: https://reviews.freebsd.org/D11556
add support for explicitly requesting that pmap_enter() create a 2MB page
mapping. (Essentially, this feature allows the machine-independent layer to
create superpage mappings preemptively, and not wait for automatic promotion
to occur.)
Export pmap_ps_enabled() to the machine-independent layer.
Add a flag to pmap_pv_insert_pde() that specifies whether it should fail or
reclaim a PV entry when one is not available.
Refactor pmap_enter_pde() into two functions, one by the same name, that is
a general-purpose function for creating PDE PG_PS mappings, and another,
pmap_enter_2mpage(), that is used to prefault 2MB read- and/or execute-only
mappings for execve(2), mmap(2), and shmat(2).
Submitted by: Yufeng Zhou <yz70@rice.edu> (an earlier version)
Reviewed by: kib, markj
Tested by: pho
MFC after: 10 days
Differential Revision: https://reviews.freebsd.org/D11556
superpage all belong to the same object. To date, that check has not been
needed, but upcoming changes require it. (See the Differential Revision.)
Reviewed by: kib, markj
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D11556
vm_radix trie.
Existing vm_radix_init() function is renamed to vm_radix_zinit().
Inlines moved out of the _ headers.
Reviewed by: alc, markj (previous version)
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D11661
Commit message for r321173 incorrectly stated that the change disables
automatic stack growth from the AIO daemons contexts, with explanation
that this is currently prevents applying wrong resource limits. Fix
this by actually disabling the growth.
Noted by: alc
Reviewed by: alc, jhb
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
check blocking grow from other processes accesses.
Debugger may access stack grow area with ptrace(2). In this case,
real state of the process is to not have the stack grown, which
provides more accurate inspection. Technical reason to avoid the grow
is to avoid applying wrong process (debugger) stack limit.
This change also has a consequence of making aio workers accesses past
the bottom of stacks into EFAULT, arguably the situation is a
programmers mistake.
Reported by: jhb
Discussed with: alc, jhb
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Reported by: antoine
Tested by: Stefan Ehmann <shoesoft@gmx.net>,
Jan Kokemueller <jan.kokemueller@gmail.com>
PR: 220493
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
gap entry in the vm map being smaller than the sysctl-derived stack guard
size. Otherwise, the value of max_grow can suffer from overflow, and the
roundup(grow_amount, sgrowsiz) will not be properly capped, resulting in
an assertion failure.
In collaboration with: kib
MFC after: 3 days
recycles the current vm space. Otherwise, an mlockall(MCL_FUTURE) could
still be in effect on the process after an execve(2), which violates the
specification for mlockall(2).
It's pointless for vm_map_stack() to check the MEMLOCK limit. It will
never be asked to wire the stack. Moreover, it doesn't even implement
wiring of the stack.
Reviewed by: kib, markj
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D11421
a hint.
Right now, for non-fixed mmap(2) calls, addr is de-facto interpreted
as the absolute minimal address of the range where the mapping is
created. The VA allocator only allocates in the range [addr,
VM_MAXUSER_ADDRESS]. This is too restrictive, the mmap(2) call might
unduly fail if there is no free addresses above addr but a lot of
usable space below it.
Lift this implementation limitation by allocating VA in two passes.
First, try to allocate above addr, as before. If that fails, do the
second pass with less restrictive constraints for the start of
allocation by specifying minimal allocation address at the max bss
end, if this limit is less than addr.
One important case where this change makes a difference is the
allocation of the stacks for new threads in libthr. Under some
configuration conditions, libthr tries to hint kernel to reuse the
main thread stack grow area for the new stacks. This cannot work by
design now after grow area is converted to stack, and there is no
unallocated VA above the main stack. Interpreting requested stack
base address as the hint provides compatibility with old libthr and
with (mis-)configured current libthr.
Reviewed by: alc
Tested by: dim (previous version)
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
the requested protection.
The syscall returns success without changing the protection of the
guard. This is consistent with the current mprotect(2) behaviour on
the unmapped ranges. More important, the calls performed by libc and
libthr to allow execution of stacks, if requested by the loaded ELF
objects, do the expected change instead of failing on the grow space
guard.
Reviewed by: alc, markj
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
If mmap(2) is called with the MAP_STACK flag and the size which is
less or equal to the initial stack mapping size plus guard,
calculation of the mapping layout created zero-sized guard. Attempt
to create such entry failed in vm_map_insert(), causing the whole
mmap(2) call to fail.
Fix it by adjusting the initial mapping size to have space for
non-empty guard. Reject MAP_STACK requests which are shorter or equal
to the configured guard pages size.
Reported and tested by: Manfred Antar <null@pozo.com>
Reviewed by: alc, markj
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Decouple the pageout cluster size from the size of the hash table entry
used by the swap pager for mapping (object, pindex) to a block on the
swap device(s), and keep the size of a hash table entry at its current
size.
Eliminate a pointless macro.
Reviewed by: kib, markj (an earlier version)
MFC after: 4 weeks
Differential Revision: https://reviews.freebsd.org/D11305
Guard, requested by the MAP_GUARD mmap(2) flag, prevents the reuse of
the allocated address space, but does not allow instantiation of the
pages in the range. It is useful for more explicit support for usual
two-stage reserve then commit allocators, since it prevents accidental
instantiation of the mapping, e.g. by mprotect(2).
Use guards to reimplement stack grow code. Explicitely track stack
grow area with the guard, including the stack guard page. On stack
grow, trivial shift of the guard map entry and stack map entry limits
makes the stack expansion. Move the code to detect stack grow and
call vm_map_growstack(), from vm_fault() into vm_map_lookup().
As result, it is impossible to get random mapping to occur in the
stack grow area, or to overlap the stack guard page.
Enable stack guard page by default.
Reviewed by: alc, markj
Man page update reviewed by: alc, bjk, emaste, markj, pho
Tested by: pho, Qualys
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D11306 (man pages)
The issue is catched by "vm_map_wire: alien wire" KASSERT at the end
of the vm_map_wire(). We currently check for MAP_ENTRY_WIRE_SKIPPED
flag before ensuring that the wiring_thread is curthread. For HOLESOK
wiring, this means that we might see WIRE_SKIPPED entry from different
wiring.
The fix it by only checking WIRE_SKIPPED if the entry is put
IN_TRANSITION by us. Also fixed a typo in the comment explaining the
situation.
Reported and tested by: pho
Reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
instantiated.
Calling pmap_copy() on non-faulted anonymous memory entries is useless.
Noted and reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
VM_MAP_WIRE_SYSTEM mode when wiring the newly grown stack.
System maps do not create auto-grown stack. Any stack we handled,
even for P_SYSTEM, must be for user address space. P_SYSTEM processes
with mapped user space is either init(8) or an aio worker attached to
other user process with aio buffer pointing into stack area. In either
case, VM_MAP_WIRE_USER mode should be used.
Noted and reviewed by: alc
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
dirty. Assert that they are fully dirty rather than redundantly calling
vm_page_dirty() on them.
Reviewed by: kib, markj
MFC after: 1 week
X-MFC after: r319932
- Add asserts that the pages to write are dirty. The last page, if
partially written, is only required to be dirty, while completely
written pages should have all dirty bit set.
- Use uintmax_t to print vm_page pindexes.
- Use NULL instead of casted zero.
- Remove if () test which duplicated the loop ending condition.
- Miscellaneous style fixes.
Reviewed by: alc, markj (previous version)
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
internal zones only. This allows to create new zones at early stages
of boot, without need to mark them as internal to UMA, which isn't
always true.
Reviewed by: alc
r31386 changed how the size of the VM page array was calculated to be
less wasteful. For most systems, the amount of memory is divided by
the overhead required by each page (a page of data plus a struct vm_page)
to determine the maximum number of available pages. However, if the
remainder for the first non-available page was at least a page of data
(so that the only memory missing was a struct vm_page), this last page
was left in phys_avail[] but was not allocated an entry in the VM page
array. Handle this case by explicitly excluding the page from
phys_avail[].
Reviewed by: alc
Sponsored by: DARPA / AFRL
Differential Revision: https://reviews.freebsd.org/D11000
beginning of a swap area for a disk label. However, neither r118390 nor
r118544, which increased the reservation from one to two blocks, correctly
accounted for these blocks when updating the variable "swap_pager_avail".
This change corrects that error.
Reviewed by: kib
MFC after: 5 days
pager used a different scheme for striping the allocation of swap space
across multiple devices. And, although blist_fill() was intended to support
fill operations with large counts, the old striping scheme never performed a
fill larger than the stripe size. Consequently, the misplacement of a
sanity check in blst_meta_fill() went undetected. Now, moving forward in
time to r118390, a new scheme for striping was introduced that maintained a
blist allocator per device, but as noted in r318995, swapoff_one() was not
fully and correctly converted to the new scheme. This change completes what
was started in r318995 by fixing the underlying bug in blst_meta_fill() that
stops swapoff_one() from simply performing a single blist_fill() operation.
Reviewed by: kib
MFC after: 5 days
Differential Revision: https://reviews.freebsd.org/D11043
