linux-6.8/mm/Kconfig

1 # SPDX-License-Identifier: GPL-2.0-only
3 menu "Memory Management options"
7 # add proper SWAP support to them, in which case this can be remove.
16 	bool "Support for paging of anonymous memory (swap)"
20 	  This option allows you to choose whether you want to have support
22 	  used to provide more virtual memory than the actual RAM present
32 	  pages that are in the process of being swapped out and attempts to
33 	  compress them into a dynamically allocated RAM-based memory pool.
56 	  the zswap entry is invalidated at once, as opposed to leaving it
59 	  This avoids having two copies of the same page in memory
61 	  The cost is that if the page was never dirtied and needs to be
62 	  swapped out again, it will be re-compressed.
65 	bool "Shrink the zswap pool on memory pressure"
71 	  written back to the backing swap device) on memory pressure.
76 	  and consume memory indefinitely.
87 	  a particular compression algorithm please refer to the benchmarks
191 	  It is designed to store up to two compressed pages per physical
193 	  deterministic reclaim properties that make it preferable to a higher
201 	  It is designed to store up to three compressed pages per physical
210 	  zsmalloc is a slab-based memory allocator designed to store
219 	  This option enables code in the zsmalloc to collect various
221 	  information to userspace via debugfs.
225 	int "Maximum number of physical pages per-zspage"
249 	bool "Configure for minimal memory footprint"
253 	   Configures the slab allocator in a way to achieve minimal memory
262 	bool "Allow slab caches to be merged"
265 	  For reduced kernel memory fragmentation, slab caches can be
267 	  This carries a risk of kernel heap overflows being able to
269 	  cache layout), which makes such heap attacks easier to exploit
271 	  can usually only damage objects in the same cache. To disable
287 	  Many kernel heap attacks try to target slab cache metadata and
289 	  sacrifices to harden the kernel slab allocator against common
297 	  The statistics are useful to debug slab allocation behavior in
298 	  order find ways to optimize the allocator. This should never be
301 	  supports the determination of the most active slabs to figure
302 	  out which slabs are relevant to a particular load.
303 	  Try running: slabinfo -DA
311 	  that is local to a processor at the price of more indeterminism
323 	  on code address, which makes the attackers more difficult to spray
324 	  vulnerable memory objects on the heap for the purpose of exploiting
325 	  memory vulnerabilities.
327 	  Currently the number of copies is set to 16, a reasonably large value
328 	  that effectively diverges the memory objects allocated for different
330 	  limited degree of memory and CPU overhead that relates to hardware and
340 	  utilization of a direct-mapped memory-side-cache. See section
341 	  5.2.27 Heterogeneous Memory Attribute Table (HMAT) in the ACPI
343 	  the presence of a memory-side-cache. There are also incidental
345 	  allocations to compliment SLAB_FREELIST_RANDOM, but the
353 	  after runtime detection of a direct-mapped memory-side-cache.
365 	  This option changes the bootup default to heap randomization
367 	  /proc/sys/kernel/randomize_va_space to 2.
369 	  On non-ancient distros (post-2000 ones) N is usually a safe choice.
372 	bool "Allow mmapped anonymous memory to be uninitialized"
376 	  Normally, and according to the Linux spec, anonymous memory obtained
377 	  from mmap() has its contents cleared before it is passed to
378 	  userspace.  Enabling this config option allows you to request that
384 	  ELF-FDPIC binfmt's brk and stack allocator.
388 	  userspace.  Since that isn't generally a problem on no-MMU systems,
389 	  it is normally safe to say Y here.
391 	  See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
398 	prompt "Memory model"
403 	  This option allows you to change some of the ways that
404 	  Linux manages its memory internally. Most users will
409 	bool "Flat Memory"
412 	  This option is best suited for non-NUMA systems with
418 	  spaces and for features like NUMA and memory hotplug,
419 	  choose "Sparse Memory".
421 	  If unsure, choose this option (Flat Memory) over any other.
424 	bool "Sparse Memory"
428 	  memory hot-plug systems.  This is normal.
431 	  holes is their physical address space and allows memory
432 	  hot-plug and hot-remove.
434 	  If unsure, choose "Flat Memory" over this option.
472 	bool "Sparse Memory virtual memmap"
476 	  SPARSEMEM_VMEMMAP uses a virtually mapped memmap to optimise
481 # to enable the feature of HugeTLB/dev_dax vmemmap optimization.
496 # Don't discard allocated memory used to track "memory" and "reserved" memblocks
497 # after early boot, so it can still be used to test for validity of memory.
498 # Also, memblocks are updated with memory hot(un)plug.
502 # Keep arch NUMA mapping infrastructure post-init.
510 # IORESOURCE_EXCLUSIVE cannot be mapped to user space, for example, via
517 # Only be set on architectures that have completely implemented memory hotplug
531 	bool "Memory hotplug"
541 	bool "Online the newly added memory blocks by default"
544 	  This option sets the default policy setting for memory hotplug
545 	  onlining policy (/sys/devices/system/memory/auto_online_blocks) which
546 	  determines what happens to newly added memory regions. Policy setting
548 	  See Documentation/admin-guide/mm/memory-hotplug.rst for more information.
550 	  Say Y here if you want all hot-plugged memory blocks to appear in
552 	  Say N here if you want the default policy to keep all hot-plugged
553 	  memory blocks in 'offline' state.
556 	bool "Allow for memory hot remove"
571 # Heavily threaded applications may benefit from splitting the mm-wide
574 # Default to 4 for wider testing, though 8 might be more appropriate.
575 # ARM's adjust_pte (unused if VIPT) depends on mm-wide page_table_lock.
576 # PA-RISC 7xxx's spinlock_t would enlarge struct page from 32 to 44 bytes.
578 # a per-page lock leads to problems when multiple tables need to be locked
594 # support for memory balloon
599 # support for memory balloon compaction
601 	bool "Allow for balloon memory compaction/migration"
605 	  Memory fragmentation introduced by ballooning might reduce
606 	  significantly the number of 2MB contiguous memory blocks that can be
609 	  by the guest workload. Allowing the compaction & migration for memory
610 	  pages enlisted as being part of memory balloon devices avoids the
611 	  scenario aforementioned and helps improving memory defragmentation.
614 # support for memory compaction
616 	bool "Allow for memory compaction"
621 	  Compaction is the only memory management component to form
622 	  high order (larger physically contiguous) memory blocks
624 	  the lack of the feature can lead to unexpected OOM killer
625 	  invocations for high order memory requests. You shouldn't
627 	  it and then we would be really interested to hear about that at
628 	  linux-mm@kvack.org.
644 	  those pages to another entity, such as a hypervisor, so that the
645 	  memory can be freed within the host for other uses.
657 	  two situations. The first is on NUMA systems to put pages nearer
658 	  to the processors accessing. The second is when allocating huge
659 	  pages as migration can relocate pages to satisfy a huge page
674 	  Allows the pageblock_order value to be dynamic instead of just standard
679 	  clamped down to MAX_PAGE_ORDER.
685 	int "Maximum scale factor of PCP (Per-CPU pageset) batch allocate/free"
689 	  In page allocator, PCP (Per-CPU pageset) is refilled and drained in
690 	  batches.  The batch number is scaled automatically to improve page
692 	  latency.  This option sets the upper limit of scale factor to limit
704 	  memory available to the CPU. Enabled by default when HIGHMEM is
705 	  selected, but you may say n to override this.
720 	  saving memory until one or another app needs to modify the content.
724 	  root has set /sys/kernel/mm/ksm/run to 1 (if CONFIG_SYSFS is set).
727 	int "Low address space to protect from user allocation"
731 	  This is the portion of low virtual memory which should be protected
732 	  from userspace allocation.  Keeping a user from writing to low pages
738 	  Programs which use vm86 functionality or have some need to map
740 	  protection by setting the value to 0.
751 	bool "Enable recovery from hardware memory errors"
755 	  Enables code to recover from some memory failures on systems
756 	  with MCA recovery. This allows a system to continue running
757 	  even when some of its memory has uncorrected errors. This requires
758 	  special hardware support and typically ECC memory.
770 	  The NOMMU mmap() frequently needs to allocate large contiguous chunks
771 	  of memory on which to store mappings, but it can only ask the system
772 	  allocator for chunks in 2^N*PAGE_SIZE amounts - which is frequently
773 	  more than it requires.  To deal with this, mmap() is able to trim off
774 	  the excess and return it to the allocator.
776 	  If trimming is enabled, the excess is trimmed off and returned to the
781 	  long-term mappings means that the space is wasted.
786 	  no trimming is to occur.
791 	  See Documentation/admin-guide/mm/nommu-mmap.rst for more information.
805 	  Transparent Hugepages allows the kernel to use huge pages and
806 	  huge tlb transparently to the applications whenever possible.
807 	  This feature can improve computing performance to certain
808 	  applications by speeding up page faults during memory
812 	  If memory constrained on embedded, you may want to say N.
827 	  memory footprint of applications without a guaranteed
834 	  performance improvement benefit to the applications using
835 	  madvise(MADV_HUGEPAGE) but it won't risk to increase the
836 	  memory footprint of applications without a guaranteed
857 	bool "Read-only THP for filesystems (EXPERIMENTAL)"
861 	  Allow khugepaged to put read-only file-backed pages in THP.
890 	bool "Contiguous Memory Allocator"
895 	  This enables the Contiguous Memory Allocator which allows other
896 	  subsystems to allocate big physically-contiguous blocks of memory.
897 	  CMA reserves a region of memory and allows only movable pages to
898 	  be allocated from it. This way, the kernel can use the memory for
900 	  allocated pages are migrated away to serve the contiguous request.
923 	  This option exposes some sysfs attributes to get information
932 	  CMA allows to create CMA areas for particular purpose, mainly,
939 	bool "Track memory changes"
943 	  This option enables memory changes tracking by introducing a
944 	  soft-dirty bit on pte-s. This bit it set when someone writes
948 	  See Documentation/admin-guide/mm/soft-dirty.rst for more details.
954 	int "Default maximum user stack size for 32-bit processes (MB)"
959 	  This is the maximum stack size in Megabytes in the VM layout of 32-bit
966 	bool "Defer initialisation of struct pages to kthreads"
984 	  This adds PG_idle and PG_young flags to 'struct page'.  PTE Accessed
993 	  This feature allows to estimate the amount of user pages that have
995 	  be useful to tune memory cgroup limits and/or for job placement
998 	  See Documentation/admin-guide/mm/idle_page_tracking.rst for
1008 	  checking, an architecture-agnostic way to find the stack pointer
1029 	bool "Device memory (pmem, HMM, etc...) hotplug support"
1037 	  Device memory hotplug support allows for establishing pmem,
1038 	  or other device driver discovered memory regions, in the
1040 	  "device-physical" addresses which is needed for using a DAX
1046 # Helpers to mirror range of the CPU page tables of a process into device page
1058 	bool "Unaddressable device memory (GPU memory, ...)"
1063 	  Allows creation of struct pages to represent unaddressable device
1064 	  memory; i.e., memory that is only accessible from the device (or
1065 	  group of devices). You likely also want to select HMM_MIRROR.
1079 	  suitable for 64-bit architectures with CONFIG_FLATMEM or
1081 	  enough room for additional bits in page->flags.
1087 	  VM event counters are needed for event counts to be shown.
1093 	bool "Collect percpu memory statistics"
1097 	  be used to help understand percpu memory usage.
1100 	bool "Enable infrastructure for get_user_pages()-related unit tests"
1104 	  to make ioctl calls that can launch kernel-based unit tests for
1109 	  the non-_fast variants.
1111 	  There is also a sub-test that allows running dump_page() on any
1112 	  of up to eight pages (selected by command line args) within the
1113 	  range of user-space addresses. These pages are either pinned via
1119 comment "GUP_TEST needs to have DEBUG_FS enabled"
1126 	tristate "Enable a module to run time tests on dma_pool"
1130 	  various sizes and report how long it takes. This is intended to
1131 	  provide a consistent way to measure how changes to the
1139 # required to support multiple hugepage sizes. For example a4fe3ce76
1168 	  Enable the memfd_secret() system call with the ability to create
1169 	  memory areas visible only in the context of the owning process and
1170 	  not mapped to other processes and other kernel page tables.
1177 	  Allow naming anonymous virtual memory areas.
1179 	  This feature allows assigning names to virtual memory areas. Assigned
1181 	  and help identifying individual anonymous memory areas.
1182 	  Assigning a name to anonymous virtual memory area might prevent that
1183 	  area from being merged with adjacent virtual memory areas due to the
1200 	  Enable the userfaultfd() system call that allows to intercept and
1210 	  Allows to create marker PTEs for userfaultfd write protection
1211 	  purposes.  It is required to enable userfaultfd write protection on
1212 	  file-backed memory types like shmem and hugetlbfs.
1215 # multi-gen LRU {
1217 	bool "Multi-Gen LRU"
1219 	# make sure folio->flags has enough spare bits
1222 	  A high performance LRU implementation to overcommit memory. See
1223 	  Documentation/admin-guide/mm/multigen_lru.rst for details.
1229 	  This option enables the multi-gen LRU by default.
1235 	  Do not enable this option unless you plan to look at historical stats
1238 	  This option has a per-memcg and per-node memory overhead.
1252 	  Allow per-vma locking during page fault handling.
1254 	  This feature allows locking each virtual memory area separately when