[PATCH v2 00/10] virtio-mem: paravirtualized memory
From: David Hildenbrand
Date: Wed Mar 11 2020 - 13:15:14 EST
This series is based on latest linux-next. The patches are located at:
https://github.com/davidhildenbrand/linux.git virtio-mem-v2
I now have acks for all !virtio-mem changes. I'll be happy to get review
feedback, testing reports, etc. for the virtio-mem changes. If there are
no further comments, I guess this is good to go as a v1 soon.
The basic idea of virtio-mem is to provide a flexible,
cross-architecture memory hot(un)plug solution that avoids many limitations
imposed by existing technologies, architectures, and interfaces. More
details can be found below and in linked material.
It's currently only enabled for x86-64, however, should theoretically work
on any architecture that supports virtio and implements memory hot(un)plug
under Linux - like s390x, powerpc64, and arm64. On x86-64, it is currently
possible to add/remove memory to the system in >= 4MB granularity.
Memory hotplug works very reliably. For memory unplug, there are no
guarantees how much memory can actually get unplugged, it depends on the
setup (especially: fragmentation of physical memory).
I am currently getting the QEMU side into shape (which will be posted as
RFC soon, see below for a link to the current state). Experimental Kata
support is in the works [4]. Also, a cloud-hypervisor implementation is
under discussion [5].
--------------------------------------------------------------------------
1. virtio-mem
--------------------------------------------------------------------------
The basic idea behind virtio-mem was presented at KVM Forum 2018. The
slides can be found at [1]. The previous RFC can be found at [2]. The
first RFC can be found at [3]. However, the concept evolved over time. The
KVM Forum slides roughly match the current design.
Patch #2 ("virtio-mem: Paravirtualized memory hotplug") contains quite some
information, especially in "include/uapi/linux/virtio_mem.h":
Each virtio-mem device manages a dedicated region in physical address
space. Each device can belong to a single NUMA node, multiple devices
for a single NUMA node are possible. A virtio-mem device is like a
"resizable DIMM" consisting of small memory blocks that can be plugged
or unplugged. The device driver is responsible for (un)plugging memory
blocks on demand.
Virtio-mem devices can only operate on their assigned memory region in
order to (un)plug memory. A device cannot (un)plug memory belonging to
other devices.
The "region_size" corresponds to the maximum amount of memory that can
be provided by a device. The "size" corresponds to the amount of memory
that is currently plugged. "requested_size" corresponds to a request
from the device to the device driver to (un)plug blocks. The
device driver should try to (un)plug blocks in order to reach the
"requested_size". It is impossible to plug more memory than requested.
The "usable_region_size" represents the memory region that can actually
be used to (un)plug memory. It is always at least as big as the
"requested_size" and will grow dynamically. It will only shrink when
explicitly triggered (VIRTIO_MEM_REQ_UNPLUG).
There are no guarantees what will happen if unplugged memory is
read/written. Such memory should, in general, not be touched. E.g.,
even writing might succeed, but the values will simply be discarded at
random points in time.
It can happen that the device cannot process a request, because it is
busy. The device driver has to retry later.
Usually, during system resets all memory will get unplugged, so the
device driver can start with a clean state. However, in specific
scenarios (if the device is busy) it can happen that the device still
has memory plugged. The device driver can request to unplug all memory
(VIRTIO_MEM_REQ_UNPLUG) - which might take a while to succeed if the
device is busy.
--------------------------------------------------------------------------
2. Linux Implementation
--------------------------------------------------------------------------
Memory blocks (e.g., 128MB) are added/removed on demand. Within these
memory blocks, subblocks (e.g., 4MB) are plugged/unplugged. The sizes
depend on the target architecture, MAX_ORDER, pageblock_order, and
the block size of a virtio-mem device.
add_memory()/try_remove_memory() is used to add/remove memory blocks.
virtio-mem will not online memory blocks itself. This has to be done by
user space, or configured into the kernel
(CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE). virtio-mem will only unplug memory
that was online to the ZONE_NORMAL. Memory is suggested to be onlined to
the ZONE_NORMAL for now.
The memory hotplug notifier is used to properly synchronize against
onlining/offlining of memory blocks and to track the states of memory
blocks (including the zone memory blocks are onlined to).
The set_online_page() callback is used to keep unplugged subblocks
of a memory block fake-offline when onlining the memory block.
generic_online_page() is used to fake-online plugged subblocks. This
handling is similar to the Hyper-V balloon driver.
PG_offline is used to mark unplugged subblocks as offline, so e.g.,
dumping tools (makedumpfile) will skip these pages. This is similar to
other balloon drivers like virtio-balloon and Hyper-V.
Memory offlining code is extended to allow drivers to drop their reference
to PG_offline pages when MEM_GOING_OFFLINE, so these pages can be skipped
when offlining memory blocks. This allows to offline memory blocks that
have partially unplugged (allocated e.g., via alloc_contig_range())
subblocks - or are completely unplugged.
alloc_contig_range()/free_contig_range() [now exposed] is used to
unplug/plug subblocks of memory blocks the are already exposed to Linux.
offline_and_remove_memory() [new] is used to offline a fully unplugged
memory block and remove it from Linux.
--------------------------------------------------------------------------
3. Changes v1 -> v2
--------------------------------------------------------------------------
- "virtio-mem: Paravirtualized memory hotplug"
-- Use "__u64" and friends in uapi header
-- Split out ACPI PXM handling
- "virtio-mem: Allow to specify an ACPI PXM as nid"
-- Squash of the ACPI PXM handling and previous "ACPI: NUMA: export
pxm_to_node"
- "virtio-mem: Paravirtualized memory hotunplug part 2"
-- Squashed previous "mm: Export alloc_contig_range() /
free_contig_range()"
- "virtio-mem: Allow to offline partially unplugged memory blocks"
-- WARN and dump_page() in case somebody has a reference to an unplugged
page
- "virtio-mem: Better retry handling"
-- Use retry interval of 5s -> 5m
- Tweaked some patch descriptions
--------------------------------------------------------------------------
4. Future work
--------------------------------------------------------------------------
One of the next TODO things besides the QEMU part is writing a virtio-mem
spec - however, that might still take some time.
virtio-mem extensions (via new feature flags):
- Indicate the guest status (e.g., initialized, working, all memory is
busy when unplugging, too many memory blocks are offline when plugging,
etc.)
- Guest-triggered shrinking of the usable region (e.g., whenever the
highest memory block is removed).
- Exchange of plugged<->unplugged block for defragmentation.
Memory hotplug:
- Reduce the amount of memory resources if that tunes out to be an
issue. Or try to speed up relevant code paths to deal with many
resources.
- Allocate vmemmap from added memory.
Memory hotunplug:
- Performance improvements:
-- Sense (lockless) if it make sense to try alloc_contig_range() at all
before directly trying to isolate and taking locks.
-- Try to unplug bigger chunks within a memory block first.
- Make unplug more likely to succeed:
-- There are various idea to limit fragmentation on memory block
granularity. (e.g., ZONE_PREFER_MOVABLE and smart balancing)
-- Allocate vmemmap from added memory.
- OOM handling, e.g., via an OOM handler/shrinker.
- Defragmentation
- Support for < MAX_ORDER - 1 blocks (esp. pageblock_order)
--------------------------------------------------------------------------
5. Example Usage
--------------------------------------------------------------------------
A QEMU implementation (without protection of unplugged memory, but with
resizable memory regions and optimized migration) is available at (kept
updated):
https://github.com/davidhildenbrand/qemu.git virtio-mem
Start QEMU with two virtio-mem devices (one per NUMA node):
$ qemu-system-x86_64 -m 4G,maxmem=204G \
-smp sockets=2,cores=2 \
-numa node,nodeid=0,cpus=0-1 -numa node,nodeid=1,cpus=2-3 \
[...]
-object memory-backend-ram,id=mem0,size=100G,managed-size=on \
-device virtio-mem-pci,id=vm0,memdev=mem0,node=0,requested-size=0M \
-object memory-backend-ram,id=mem1,size=100G,managed-size=on \
-device virtio-mem-pci,id=vm1,memdev=mem1,node=1,requested-size=1G
Query the configuration:
QEMU 4.2.50 monitor - type 'help' for more information
(qemu) info memory-devices
Memory device [virtio-mem]: "vm0"
memaddr: 0x140000000
node: 0
requested-size: 0
size: 0
max-size: 107374182400
block-size: 2097152
memdev: /objects/mem0
Memory device [virtio-mem]: "vm1"
memaddr: 0x1a40000000
node: 1
requested-size: 1073741824
size: 1073741824
max-size: 107374182400
block-size: 2097152
memdev: /objects/mem1
Add some memory to node 0:
QEMU 4.2.50 monitor - type 'help' for more information
(qemu) qom-set vm0 requested-size 1G
Remove some memory from node 1:
QEMU 4.2.50 monitor - type 'help' for more information
(qemu) qom-set vm1 requested-size 64M
Query the configuration again:
QEMU 4.2.50 monitor - type 'help' for more information
(qemu) info memory-devices
Memory device [virtio-mem]: "vm0"
memaddr: 0x140000000
node: 0
requested-size: 1073741824
size: 1073741824
max-size: 107374182400
block-size: 2097152
memdev: /objects/mem0
Memory device [virtio-mem]: "vm1"
memaddr: 0x1a40000000
node: 1
requested-size: 67108864
size: 67108864
max-size: 107374182400
block-size: 2097152
memdev: /objects/mem1
--------------------------------------------------------------------------
6. Q/A
--------------------------------------------------------------------------
Q: Why add/remove parts ("subblocks") of memory blocks/sections?
A: Flexibility (section size depends on the architecture) - e.g., some
architectures have a section size of 2GB. Also, the memory block size
is variable (e.g., on x86-64). I want to avoid any such restrictions.
Some use cases want to add/remove memory in smaller granularity to a
VM (e.g., the Hyper-V balloon also implements this) - especially smaller
VMs like used for kata containers. Also, on memory unplug, it is more
reliable to free-up and unplug multiple small chunks instead
of one big chunk. E.g., if one page of a DIMM is either unmovable or
pinned, the DIMM can't get unplugged. This approach is basically a
compromise between DIMM-based memory hot(un)plug and balloon
inflation/deflation, which works mostly on page granularity.
Q: Why care about memory blocks?
A: They are the way to tell user space about new memory. This way,
memory can get onlined/offlined by user space. Also, e.g., kdump
relies on udev events to reload kexec when memory blocks are
onlined/offlined. Memory blocks are the "real" memory hot(un)plug
granularity. Everything that's smaller has to be emulated "on top".
Q: Won't memory unplug of subblocks fragment memory?
A: Yes and no. Unplugging e.g., >=4MB subblocks on x86-64 will not really
fragment memory like unplugging random pages like a balloon driver does.
Buddy merging will not be limited. However, any allocation that requires
bigger consecutive memory chunks (e.g., gigantic pages) might observe
the fragmentation. Possible solutions: Allocate gigantic huge pages
before unplugging memory, don't unplug memory, combine virtio-mem with
DIMM based memory or bigger initial memory. Remember, a virtio-mem
device will only unplug on the memory range it manages, not on other
DIMMs. Unplug of single memory blocks will result in similar
fragmentation in respect to gigantic huge pages.
Q: How reliable is memory unplug?
A: There are no guarantees on how much memory can get unplugged
again. However, it is more likely to find 4MB chunks to unplug than
e.g., 128MB chunks. If memory is terribly fragmented, there is nothing
we can do - for now. I consider memory hotplug the first primary use
of virtio-mem. Memory unplug might usually work, but we want to improve
the performance and the amount of memory we can actually unplug later.
Q: Why not unplug from the ZONE_MOVABLE?
A: Unplugged memory chunks are unmovable. Unmovable data must not end up
on the ZONE_MOVABLE - similar to gigantic pages - they will never be
allocated from ZONE_MOVABLE. virtio-mem added memory can be onlined
to the ZONE_MOVABLE, but subblocks will not get unplugged from it.
Q: How big should the initial (!virtio-mem) memory of a VM be?
A: virtio-mem memory will not go to the DMA zones. So to avoid running out
of DMA memory, I suggest something like 2-3GB on x86-64. But many
VMs can most probably deal with less DMA memory - depends on the use
case.
[1] https://events.linuxfoundation.org/wp-content/uploads/2017/12/virtio-mem-Paravirtualized-Memory-David-Hildenbrand-Red-Hat-1.pdf
[2] https://lkml.kernel.org/r/20190919142228.5483-1-david@xxxxxxxxxx
[3] https://lkml.kernel.org/r/547865a9-d6c2-7140-47e2-5af01e7d761d@xxxxxxxxxx
[4] https://github.com/kata-containers/documentation/pull/592
[5] https://github.com/cloud-hypervisor/cloud-hypervisor/pull/837
Cc: Sebastien Boeuf <sebastien.boeuf@xxxxxxxxx>
Cc: Samuel Ortiz <samuel.ortiz@xxxxxxxxx>
Cc: Robert Bradford <robert.bradford@xxxxxxxxx>
Cc: Luiz Capitulino <lcapitulino@xxxxxxxxxx>
Cc: Pankaj Gupta <pankaj.gupta.linux@xxxxxxxxx>
Cc: teawater <teawaterz@xxxxxxxxxxxxxxxxx>
Cc: Igor Mammedov <imammedo@xxxxxxxxxx>
Cc: Dr. David Alan Gilbert <dgilbert@xxxxxxxxxx>
David Hildenbrand (10):
virtio-mem: Paravirtualized memory hotplug
virtio-mem: Allow to specify an ACPI PXM as nid
virtio-mem: Paravirtualized memory hotunplug part 1
virtio-mem: Paravirtualized memory hotunplug part 2
mm: Allow to offline unmovable PageOffline() pages via
MEM_GOING_OFFLINE
virtio-mem: Allow to offline partially unplugged memory blocks
mm/memory_hotplug: Introduce offline_and_remove_memory()
virtio-mem: Offline and remove completely unplugged memory blocks
virtio-mem: Better retry handling
MAINTAINERS: Add myself as virtio-mem maintainer
MAINTAINERS | 7 +
drivers/acpi/numa/srat.c | 1 +
drivers/virtio/Kconfig | 18 +
drivers/virtio/Makefile | 1 +
drivers/virtio/virtio_mem.c | 1910 +++++++++++++++++++++++++++++++
include/linux/memory_hotplug.h | 1 +
include/linux/page-flags.h | 10 +
include/uapi/linux/virtio_ids.h | 1 +
include/uapi/linux/virtio_mem.h | 208 ++++
mm/memory_hotplug.c | 81 +-
mm/page_alloc.c | 26 +
mm/page_isolation.c | 9 +
12 files changed, 2263 insertions(+), 10 deletions(-)
create mode 100644 drivers/virtio/virtio_mem.c
create mode 100644 include/uapi/linux/virtio_mem.h
--
2.24.1