Commit Graph

51 Commits

Author SHA1 Message Date
Konstantin Belousov
f2a496d667 MI VM: Make it possible to set size of superpage at boot instead of compile time.
In order to allow single kernel to use PAE pagetables on i386 if
hardware supports it, and fall back to classic two-level paging
structures if not, superpage code should be able to adopt to either 2M
or 4M superpages size.  There I make MI VM structures large enough to
track the biggest possible superpage, by allowing architecture to
define VM_NFREEORDER_MAX and VM_LEVEL_0_ORDER_MAX constants.
Corresponding VM_NFREEORDER and VM_LEVEL_0_ORDER symbols can be
defined as runtime values and must be less than the _MAX constants.
If architecture does not define _MAXs, it is assumed that _MAX ==
normal constant.

Reviewed by:	markj
Tested by:	pho (as part of the larger patch)
Sponsored by:	The FreeBSD Foundation
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D18853
2019-01-18 13:35:06 +00:00
Jeff Roberson
2ef6727edd Use the ticks since the last update to reduce hysteresis in the partpopq and
contention on the vm_reserv_domain lock.

This gives a roughly 8x speedup on will-it-scale fault1 on a 16 core machine.

Reviewed by:	alc, kib, markj
2018-07-07 01:54:45 +00:00
Jeff Roberson
2d3f4181de Fix two compliation problems on non-amd64 architectures. 2018-03-23 18:24:02 +00:00
Cy Schubert
72346b2232 Fix build on i386 without INVARIANTS following r331369.
--- vm_reserv.o ---
In file included from /opt/src/svn-current/sys/vm/vm_reserv.c:48:
In file included from /opt/src/svn-current/sys/sys/counter.h:37:
./machine/counter.h:174:3: error: implicit declaration of function
'critical_enter' is invalid in C99 [-Werror,-Wimplicit-function-declarat
ion]
                critical_enter();

Reviewed by:	jeff@
2018-03-23 03:22:30 +00:00
Jeff Roberson
5c930c894d Lock reservations with a dedicated lock in each reservation. Protect the
vmd_free_count with atomics.

This allows us to allocate and free from reservations without the free lock
except where a superpage is allocated from the physical layer, which is
roughly 1/512 of the operations on amd64.

Use the counter api to eliminate cache conention on counters.

Reviewed by:	markj
Tested by:	pho
Sponsored by:	Netflix, Dell/EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D14707
2018-03-22 19:21:11 +00:00
Jeff Roberson
30fbfdda6c Eliminate pageout wakeup races. Take another step towards lockless
vmd_free_count manipulation.  Reduce the scope of the free lock by
using a pageout lock to synchronize sleep and wakeup.  Only trigger
the pageout daemon on transitions between states.  Drive all wakeup
operations directly as side-effects from freeing memory rather than
requiring an additional function call.

Reviewed by:	markj, kib
Tested by:	pho
Sponsored by:	Netflix, Dell/EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D14612
2018-03-15 19:23:07 +00:00
Jeff Roberson
f4af595964 Don't assert that the domain free lock is held until we're certain that
there is a valid reservation.  This can trip erroneously when memory
falls within a domain but doesn't have the reservation initialized because
it does not meet size or alignment requirements.

Reported by:	pho, mjg
Sponsored by:	Netflix, Dell/EMC Isilon
2018-03-07 22:04:27 +00:00
Mark Johnston
5f70fb1425 Correct some comments after r328954.
Reviewed by:	kib
Differential Revision:	https://reviews.freebsd.org/D14486
2018-02-23 23:27:53 +00:00
Konstantin Belousov
ada27a3bb8 Cleanup unused page argument for vm_reserv_break().
Reviewed by:	markj
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D14364
2018-02-14 00:34:02 +00:00
Konstantin Belousov
c4be9169c0 Do not leak rv->psind in some specific situations.
Suppose that we have an object with a mapped superpage, and that all
pages in the superpages are held (by some driver).  Additionally,
suppose that the object is terminated, e.g. because the only process
mapping it is exiting.  Then the reservation is broken, but the pages
cannot be freed until later, when they are unheld.  In this situation,
the reservation code cannot clean psind, since no pages are freed, and
the page is freed and then reused with invalid psind.

Clean psind on vm_reserv_break() to avoid the situation.

Reported and tested by:	Slava Shwartsman
Reviewed by:	markj
Sponsored by:	Mellanox Technologies
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D14335
2018-02-13 15:36:28 +00:00
Jeff Roberson
e2068d0bcd Use per-domain locks for vm page queue free. Move paging control from
global to per-domain state.  Protect reservations with the free lock
from the domain that they belong to.  Refactor to make vm domains more
of a first class object.

Reviewed by:    markj, kib, gallatin
Tested by:      pho
Sponsored by:   Netflix, Dell/EMC Isilon
Differential Revision:  https://reviews.freebsd.org/D14000
2018-02-06 22:10:07 +00:00
Jeff Roberson
7a469c8ef3 Implement NUMA policy for kmem_*(9). This maintains compatibility with
reservations by giving each memory domain its own KVA space in vmem that
is naturally aligned on superpage boundaries.

Reviewed by:	alc, markj, kib  (some objections)
Sponsored by:	Netflix, Dell/EMC Isilon
Tested by;	pho
Differential Revision:	https://reviews.freebsd.org/D13289
2018-01-12 23:13:55 +00:00
Jeff Roberson
ef435ae7de Move domain iterators into the page layer where domain selection should take
place.  This makes the majority of the phys layer explicitly domain specific.

Reviewed by:	markj, kib (some objections)
Discussed with:	alc
Tested by:	pho
Sponsored by:	Netflix & Dell EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D13014
2017-11-28 23:18:35 +00:00
Pedro F. Giffuni
fe267a5590 sys: general adoption of SPDX licensing ID tags.
Mainly focus on files that use BSD 2-Clause license, however the tool I
was using misidentified many licenses so this was mostly a manual - error
prone - task.

The Software Package Data Exchange (SPDX) group provides a specification
to make it easier for automated tools to detect and summarize well known
opensource licenses. We are gradually adopting the specification, noting
that the tags are considered only advisory and do not, in any way,
superceed or replace the license texts.

No functional change intended.
2017-11-27 15:23:17 +00:00
Alan Cox
8b5e1472d2 Utilize pmap_enter(..., psind=1) in vm_fault_soft_fast() on amd64. (The
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
2017-07-23 16:28:13 +00:00
Gleb Smirnoff
9ed01c32e0 All these files need sys/vmmeter.h, but now they got it implicitly
included via sys/pcpu.h.
2017-04-17 17:07:00 +00:00
Alan Cox
920da7e4d2 Relax the object type restrictions on vm_page_alloc_contig(). Specifically,
add support for object types that were previously prohibited because they
could contain PG_CACHED pages.

Roughly halve the number of radix trie operations performed by
vm_page_alloc_contig() using the same approach that is employed by
vm_page_alloc().  Also, eliminate the radix trie lookup performed with the
free page queues lock held.

Tidy up the handling of radix trie insert failures in vm_page_alloc() and
vm_page_alloc_contig().

Reviewed by:	kib, markj
Tested by:	pho
Sponsored by:	Dell EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D8878
2016-12-28 18:32:13 +00:00
Alan Cox
3453bca864 Eliminate every mention of PG_CACHED pages from the comments in the machine-
independent layer of the virtual memory system.  Update some of the nearby
comments to eliminate redundancy and improve clarity.

In vm/vm_reserv.c, do not use hyphens after adverbs ending in -ly per
The Chicago Manual of Style.

Update the comment in vm/vm_page.h defining the four types of page queues to
reflect the elimination of PG_CACHED pages and the introduction of the
laundry queue.

Reviewed by:	kib, markj
Sponsored by:	Dell EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D8752
2016-12-12 17:47:09 +00:00
Alan Cox
7667839a7e Remove most of the code for implementing PG_CACHED pages. (This change does
not remove user-space visible fields from vm_cnt or all of the references to
cached pages from comments.  Those changes will come later.)

Reviewed by:	kib, markj
Tested by:	pho
Sponsored by:	Dell EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D8497
2016-11-15 18:22:50 +00:00
Alan Cox
c869e67208 Introduce a new mechanism for relocating virtual pages to a new physical
address and use this mechanism when:

1. kmem_alloc_{attr,contig}() can't find suitable free pages in the physical
   memory allocator's free page lists.  This replaces the long-standing
   approach of scanning the inactive and inactive queues, converting clean
   pages into PG_CACHED pages and laundering dirty pages.  In contrast, the
   new mechanism does not use PG_CACHED pages nor does it trigger a large
   number of I/O operations.

2. on 32-bit MIPS processors, uma_small_alloc() and the pmap can't find
   free pages in the physical memory allocator's free page lists that are
   covered by the direct map.  Tested by: adrian

3. ttm_bo_global_init() and ttm_vm_page_alloc_dma32() can't find suitable
   free pages in the physical memory allocator's free page lists.

In the coming months, I expect that this new mechanism will be applied in
other places.  For example, balloon drivers should use relocation to
minimize fragmentation of the guest physical address space.

Make vm_phys_alloc_contig() a little smarter (and more efficient in some
cases).  Specifically, use vm_phys_segs[] earlier to avoid scanning free
page lists that can't possibly contain suitable pages.

Reviewed by:	kib, markj
Glanced at:	jhb
Discussed with:	jeff
Sponsored by:	EMC / Isilon Storage Division
Differential Revision:	https://reviews.freebsd.org/D4444
2015-12-19 18:42:50 +00:00
Alan Cox
67b7e4345e Correct an error in vm_reserv_reclaim_contig(). In the highly unusual
case that the reservation contained "low", the starting position in the
popmap for the free page search was incorrectly calculated.  The most
likely (and visible) symptom of this error was the assertion failure,
"vm_reserv_reclaim_contig: pa is too low".
2015-11-26 19:12:18 +00:00
Alan Cox
e0a63baae4 Introduce a sysctl for reporting the number of fully populated reservations. 2015-08-06 21:27:50 +00:00
Alan Cox
966272ca33 Retire VM_FREEPOOL_CACHE as the next step in eliminating PG_CACHE pages.
Differential Revision:	https://reviews.freebsd.org/D2712
Reviewed by:	kib
Sponsored by:	EMC / Isilon Storage Division
2015-06-08 04:59:32 +00:00
Alan Cox
6a93e36b5f Correct an off-by-one error in vm_reserv_reclaim_contig() that results in
an infinite loop.

Submitted by:	Svatopluk Kraus
MFC after:	1 week
2015-04-11 22:57:13 +00:00
Ian Lepore
ed9dd64b8c Revert r279932; this is going to be fixed in the sbuf code instead.
PR:		195668
2015-03-14 13:00:37 +00:00
Ian Lepore
f3b9fcf251 Nullterminate strings returned via sysctl.
PR:		195668
2015-03-12 18:06:30 +00:00
Konstantin Belousov
79d7993d98 Fix function name in the panic message.
Sponsored by:	The FreeBSD Foundation
MFC after:	1 week
2015-03-08 02:13:46 +00:00
Alan Cox
09e5f3c4b8 By the time that vm_reserv_init() runs, vm_phys_segs[] is initialized. Use
it instead of phys_avail[].

Discussed with:	Svatopluk Kraus
2014-11-22 17:46:30 +00:00
Alan Cox
64f096eeb2 Fix a boundary case error in vm_reserv_alloc_contig(): If a reservation
isn't being allocated for the last of the requested pages, because a
reservation won't fit in the gap between allocated pages, then the
reservation structure shouldn't be initialized.

While I'm here, improve the nearby comments.

Reported by:	jeff, pho
MFC after:	1 week
Sponsored by:	EMC / Isilon Storage Division
2014-09-10 05:52:30 +00:00
Alan Cox
3180f7573a Correct a bug in the management of the population map on big-endian
machines.  Specifically, there was a mismatch between how the routine
allocation and deallocation operations accessed the population map
and how the aggressively optimized reservation-breaking operation
accessed it.  So, problems only occurred when reservations were broken.
This change makes the routine operations access the population map in
the same way as the reservation breaking operation.

This bug was introduced in r259999.

PR:		187080
Tested by:	jmg (on an "armeb" machine)
Sponsored by:	EMC / Isilon Storage Division
2014-06-11 16:11:12 +00:00
Alan Cox
dd05fa1945 Add a page size field to struct vm_page. Increase the page size field when
a partially populated reservation becomes fully populated, and decrease this
field when a fully populated reservation becomes partially populated.

Use this field to simplify the implementation of pmap_enter_object() on
amd64, arm, and i386.

On all architectures where we support superpages, the cost of creating a
superpage mapping is roughly the same as creating a base page mapping.  For
example, both kinds of mappings entail the creation of a single PTE and PV
entry.  With this in mind, use the page size field to make the
implementation of vm_map_pmap_enter(..., MAP_PREFAULT_PARTIAL) a little
smarter.  Previously, if MAP_PREFAULT_PARTIAL was specified to
vm_map_pmap_enter(), that function would only map base pages.  Now, it will
create up to 96 base page or superpage mappings.

Reviewed by:	kib
Sponsored by:	EMC / Isilon Storage Division
2014-06-07 17:12:26 +00:00
Alan Cox
a08c151546 Add "popmap" assertions: The page being freed isn't already free, and the
page being allocated isn't already allocated.

Sponsored by:	EMC / Isilon Storage Division
2013-12-29 04:54:52 +00:00
Alan Cox
ec17932242 MFp4 alc_popmap
Change the way that reservations keep track of which pages are in use.
  Instead of using the page's PG_CACHED and PG_FREE flags, maintain a bit
  vector within the reservation.  This approach has a couple benefits.
  First, it makes breaking reservations much cheaper because there are
  fewer cache misses to identify the unused pages.  Second, it is a pre-
  requisite for supporting two or more reservation sizes.
2013-12-28 04:28:35 +00:00
Konstantin Belousov
9eab548476 PG_SLAB no longer serves a useful purpose, since m->object is no
longer abused to store pointer to slab. Remove it.

Reviewed by:    alc
Sponsored by:   The FreeBSD Foundation
Approved by:	re (hrs)
2013-09-17 07:35:26 +00:00
Alan Cox
404eb1b3fd Refactor vm_page_alloc()'s interactions with vm_reserv_alloc_page() and
vm_page_insert() so that (1) vm_radix_lookup_le() is never called while the
free page queues lock is held and (2) vm_radix_lookup_le() is called at most
once.  This change reduces the average time that the free page queues lock
is held by vm_page_alloc() as well as vm_page_alloc()'s average overall
running time.

Sponsored by:	EMC / Isilon Storage Division
2013-05-12 16:50:18 +00:00
Attilio Rao
774d251d99 Sync back vmcontention branch into HEAD:
Replace the per-object resident and cached pages splay tree with a
path-compressed multi-digit radix trie.
Along with this, switch also the x86-specific handling of idle page
tables to using the radix trie.

This change is supposed to do the following:
- Allowing the acquisition of read locking for lookup operations of the
  resident/cached pages collections as the per-vm_page_t splay iterators
  are now removed.
- Increase the scalability of the operations on the page collections.

The radix trie does rely on the consumers locking to ensure atomicity of
its operations.  In order to avoid deadlocks the bisection nodes are
pre-allocated in the UMA zone.  This can be done safely because the
algorithm needs at maximum one new node per insert which means the
maximum number of the desired nodes is the number of available physical
frames themselves.  However, not all the times a new bisection node is
really needed.

The radix trie implements path-compression because UFS indirect blocks
can lead to several objects with a very sparse trie, increasing the number
of levels to usually scan.  It also helps in the nodes pre-fetching by
introducing the single node per-insert property.

This code is not generalized (yet) because of the possible loss of
performance by having much of the sizes in play configurable.
However, efforts to make this code more general and then reusable in
further different consumers might be really done.

The only KPI change is the removal of the function vm_page_splay() which
is now reaped.
The only KBI change, instead, is the removal of the left/right iterators
from struct vm_page, which are now reaped.

Further technical notes broken into mealpieces can be retrieved from the
svn branch:
http://svn.freebsd.org/base/user/attilio/vmcontention/

Sponsored by:	EMC / Isilon storage division
In collaboration with:	alc, jeff
Tested by:	flo, pho, jhb, davide
Tested by:	ian (arm)
Tested by:	andreast (powerpc)
2013-03-18 00:25:02 +00:00
Attilio Rao
89f6b8632c Switch the vm_object mutex to be a rwlock. This will enable in the
future further optimizations where the vm_object lock will be held
in read mode most of the time the page cache resident pool of pages
are accessed for reading purposes.

The change is mostly mechanical but few notes are reported:
* The KPI changes as follow:
  - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK()
  - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK()
  - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK()
  - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED()
    (in order to avoid visibility of implementation details)
  - The read-mode operations are added:
    VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(),
    VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED()
* The vm/vm_pager.h namespace pollution avoidance (forcing requiring
  sys/mutex.h in consumers directly to cater its inlining functions
  using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h
  consumers now must include also sys/rwlock.h.
* zfs requires a quite convoluted fix to include FreeBSD rwlocks into
  the compat layer because the name clash between FreeBSD and solaris
  versions must be avoided.
  At this purpose zfs redefines the vm_object locking functions
  directly, isolating the FreeBSD components in specific compat stubs.

The KPI results heavilly broken by this commit.  Thirdy part ports must
be updated accordingly (I can think off-hand of VirtualBox, for example).

Sponsored by:	EMC / Isilon storage division
Reviewed by:	jeff
Reviewed by:	pjd (ZFS specific review)
Discussed with:	alc
Tested by:	pho
2013-03-09 02:32:23 +00:00
Alan Cox
476f7f2423 Correct an off-by-one error in vm_reserv_alloc_contig() that resulted in
the last reservation of a multi-reservation allocation not being
initialized.
2012-07-15 21:46:19 +00:00
Alan Cox
908e3da10e If a page belonging a reservation is cached, then mark the reservation so
that it will be freed to the cache pool rather than the default pool.
Otherwise, the cached pages within the reservation may be recycled sooner
than necessary.

Reported by:	Andrey Zonov
2012-04-08 17:00:46 +00:00
Alan Cox
c68c35372e Introduce vm_reserv_alloc_contig() and teach vm_page_alloc_contig() how to
use superpage reservations.  So, for the first time, kernel virtual memory
that is allocated by contigmalloc(), kmem_alloc_attr(), and
kmem_alloc_contig() can be promoted to superpages.  In fact, even a series
of small contigmalloc() allocations may collectively result in a promoted
superpage.

Eliminate some duplication of code in vm_reserv_alloc_page().

Change the type of vm_reserv_reclaim_contig()'s first parameter in order
that it be consistent with other vm_*_contig() functions.

Tested by:	marius (sparc64)
2011-12-05 18:29:25 +00:00
Alan Cox
5c1f2cc4c2 Eliminate vm_phys_bootstrap_alloc(). It was a failed attempt at
eliminating duplicated code in the various pmap implementations.

Micro-optimize vm_phys_free_pages().

Introduce vm_phys_free_contig().  It is fast routine for freeing an
arbitrary number of physically contiguous pages.  In particular, it
doesn't require the number of pages to be a power of two.

Use "u_long" instead of "unsigned long".

Bruce Evans (bde@) has convinced me that the "boundary" parameters
to kmem_alloc_contig(), vm_phys_alloc_contig(), and
vm_reserv_reclaim_contig() should be of type "vm_paddr_t" and not
"u_long".  Make this change.
2011-10-30 05:06:14 +00:00
Matthew D Fleming
00f0e671ff Explicitly wire the user buffer rather than doing it implicitly in
sbuf_new_for_sysctl(9).  This allows using an sbuf with a SYSCTL_OUT
drain for extremely large amounts of data where the caller knows that
appropriate references are held, and sleeping is not an issue.

Inspired by:	rwatson
2011-01-27 00:34:12 +00:00
Max Laier
85f2a0c91e Off by one page in vm_reserv_reclaim_contig(): Also reclaim reservations
with only a single free page if that satisfies the requested size.

MFC after:	3 days
Reviewed by:	alc
2010-11-19 04:30:33 +00:00
Alan Cox
2cf36c8f67 Enable reservation-based physical memory allocation. Even without the
creation of large page mappings in the pmap, it can provide modest
performance benefits.  In particular, for a "buildworld" on a 2x 1GHz
Ultrasparc IIIi it reduced the wall clock time by 2.2% and the system
time by 12.6%.

Tested by:	marius@
2010-11-10 17:57:34 +00:00
Alan Cox
d689bc0082 Correct some format strings used by sysctls.
MFC after:	1 week
2010-10-30 18:00:53 +00:00
Matthew D Fleming
4e6571599b Re-add r212370 now that the LOR in powerpc64 has been resolved:
Add a drain function for struct sysctl_req, and use it for a variety
of handlers, some of which had to do awkward things to get a large
enough SBUF_FIXEDLEN buffer.

Note that some sysctl handlers were explicitly outputting a trailing
NUL byte.  This behaviour was preserved, though it should not be
necessary.

Reviewed by:    phk (original patch)
2010-09-16 16:13:12 +00:00
Matthew D Fleming
404a593e28 Revert r212370, as it causes a LOR on powerpc. powerpc does a few
unexpected things in copyout(9) and so wiring the user buffer is not
sufficient to perform a copyout(9) while holding a random mutex.

Requested by: nwhitehorn
2010-09-13 18:48:23 +00:00
Matthew D Fleming
dd67e2103c Add a drain function for struct sysctl_req, and use it for a variety of
handlers, some of which had to do awkward things to get a large enough
FIXEDLEN buffer.

Note that some sysctl handlers were explicitly outputting a trailing NUL
byte.  This behaviour was preserved, though it should not be necessary.

Reviewed by:	phk
2010-09-09 18:33:46 +00:00
Alan Cox
ab5378cf11 Previously, when vm_page_free_toq() was performed on a page belonging to
a reservation, unless all of the reservation's pages were free, the
reservation was moved to the head of the partially-populated reservations
queue, where it would be the next reservation to be broken in case the
free page queues were emptied.  Now, instead, I am moving it to the tail.
Very likely this reservation is in the process of being freed in its
entirety, so placing it at the tail of the queue makes it more likely that
the underlying physical memory will be returned to the free page queues as
one contiguous chunk.  If a reservation must be broken, it will, instead,
be the longest unchanged reservation, which is arguably the reservation
that is least likely to ever achieve promotion or be freed in its entirety.

MFC after:	6 weeks
2009-04-11 09:09:00 +00:00
Alan Cox
44aab2c3de Introduce vm_reserv_reclaim_contig(). This function is used by
contigmalloc(9) as a last resort to steal pages from an inactive,
partially-used superpage reservation.

Rename vm_reserv_reclaim() to vm_reserv_reclaim_inactive() and
refactor it so that a separate subroutine is responsible for breaking
the selected reservation.  This subroutine is also used by
vm_reserv_reclaim_contig().
2008-04-06 18:09:28 +00:00