freebsd-dev/sys/sparc64/include/vmparam.h

252 lines
8.7 KiB
C
Raw Normal View History

/*-
* Copyright (c) 1990 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 1994 John S. Dyson
* All rights reserved.
*
* This code is derived from software contributed to Berkeley by
* William Jolitz.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* from: @(#)vmparam.h 5.9 (Berkeley) 5/12/91
* from: FreeBSD: src/sys/i386/include/vmparam.h,v 1.33 2000/03/30
* $FreeBSD$
*/
#ifndef _MACHINE_VMPARAM_H_
#define _MACHINE_VMPARAM_H_
/*
* Virtual memory related constants, all in bytes
*/
#ifndef MAXTSIZ
#define MAXTSIZ (1*1024*1024*1024) /* max text size */
#endif
#ifndef DFLDSIZ
#define DFLDSIZ (128*1024*1024) /* initial data size limit */
#endif
#ifndef MAXDSIZ
#define MAXDSIZ (1*1024*1024*1024) /* max data size */
#endif
#ifndef DFLSSIZ
#define DFLSSIZ (128*1024*1024) /* initial stack size limit */
#endif
#ifndef MAXSSIZ
#define MAXSSIZ (1*1024*1024*1024) /* max stack size */
#endif
#ifndef SGROWSIZ
#define SGROWSIZ (128*1024) /* amount to grow stack */
#endif
/*
* The physical address space is sparsely populated.
*/
#define VM_PHYSSEG_SPARSE
/*
* The number of PHYSSEG entries must be one greater than the number
* of phys_avail entries because the phys_avail entry that spans the
* largest physical address that is accessible by ISA DMA is split
2011-06-21 20:50:55 +00:00
* into two PHYSSEG entries.
*/
#define VM_PHYSSEG_MAX 64
/*
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
* Create three free page pools: VM_FREEPOOL_DEFAULT is the default pool
* from which physical pages are allocated and VM_FREEPOOL_DIRECT is
* the pool from which physical pages for small UMA objects are
* allocated.
*/
Change the management of cached pages (PQ_CACHE) in two fundamental ways: (1) Cached pages are no longer kept in the object's resident page splay tree and memq. Instead, they are kept in a separate per-object splay tree of cached pages. However, access to this new per-object splay tree is synchronized by the _free_ page queues lock, not to be confused with the heavily contended page queues lock. Consequently, a cached page can be reclaimed by vm_page_alloc(9) without acquiring the object's lock or the page queues lock. This solves a problem independently reported by tegge@ and Isilon. Specifically, they observed the page daemon consuming a great deal of CPU time because of pages bouncing back and forth between the cache queue (PQ_CACHE) and the inactive queue (PQ_INACTIVE). The source of this problem turned out to be a deadlock avoidance strategy employed when selecting a cached page to reclaim in vm_page_select_cache(). However, the root cause was really that reclaiming a cached page required the acquisition of an object lock while the page queues lock was already held. Thus, this change addresses the problem at its root, by eliminating the need to acquire the object's lock. Moreover, keeping cached pages in the object's primary splay tree and memq was, in effect, optimizing for the uncommon case. Cached pages are reclaimed far, far more often than they are reactivated. Instead, this change makes reclamation cheaper, especially in terms of synchronization overhead, and reactivation more expensive, because reactivated pages will have to be reentered into the object's primary splay tree and memq. (2) Cached pages are now stored alongside free pages in the physical memory allocator's buddy queues, increasing the likelihood that large allocations of contiguous physical memory (i.e., superpages) will succeed. Finally, as a result of this change long-standing restrictions on when and where a cached page can be reclaimed and returned by vm_page_alloc(9) are eliminated. Specifically, calls to vm_page_alloc(9) specifying VM_ALLOC_INTERRUPT can now reclaim and return a formerly cached page. Consequently, a call to malloc(9) specifying M_NOWAIT is less likely to fail. Discussed with: many over the course of the summer, including jeff@, Justin Husted @ Isilon, peter@, tegge@ Tested by: an earlier version by kris@ Approved by: re (kensmith)
2007-09-25 06:25:06 +00:00
#define VM_NFREEPOOL 3
#define VM_FREEPOOL_CACHE 2
#define VM_FREEPOOL_DEFAULT 0
#define VM_FREEPOOL_DIRECT 1
/*
* Create two free page lists: VM_FREELIST_DEFAULT is for physical
* pages that are above the largest physical address that is
* accessible by ISA DMA and VM_FREELIST_ISADMA is for physical pages
* that are below that address.
*/
#define VM_NFREELIST 2
#define VM_FREELIST_DEFAULT 0
#define VM_FREELIST_ISADMA 1
/*
* An allocation size of 16MB is supported in order to optimize the
* use of the direct map by UMA. Specifically, a cache line contains
* at most four TTEs, collectively mapping 16MB of physical memory.
* By reducing the number of distinct 16MB "pages" that are used by UMA,
* the physical memory allocator reduces the likelihood of both 4MB
* page TLB misses and cache misses caused by 4MB page TLB misses.
*/
#define VM_NFREEORDER 12
Very rough first cut at NUMA support for the physical page allocator. For now it uses a very dumb first-touch allocation policy. This will change in the future. - Each architecture indicates the maximum number of supported memory domains via a new VM_NDOMAIN parameter in <machine/vmparam.h>. - Each cpu now has a PCPU_GET(domain) member to indicate the memory domain a CPU belongs to. Domain values are dense and numbered from 0. - When a platform supports multiple domains, the default freelist (VM_FREELIST_DEFAULT) is split up into N freelists, one for each domain. The MD code is required to populate an array of mem_affinity structures. Each entry in the array defines a range of memory (start and end) and a domain for the range. Multiple entries may be present for a single domain. The list is terminated by an entry where all fields are zero. This array of structures is used to split up phys_avail[] regions that fall in VM_FREELIST_DEFAULT into per-domain freelists. - Each memory domain has a separate lookup-array of freelists that is used when fulfulling a physical memory allocation. Right now the per-domain freelists are listed in a round-robin order for each domain. In the future a table such as the ACPI SLIT table may be used to order the per-domain lookup lists based on the penalty for each memory domain relative to a specific domain. The lookup lists may be examined via a new vm.phys.lookup_lists sysctl. - The first-touch policy is implemented by using PCPU_GET(domain) to pick a lookup list when allocating memory. Reviewed by: alc
2010-07-27 20:33:50 +00:00
/*
* Only one memory domain.
*/
#ifndef VM_NDOMAIN
#define VM_NDOMAIN 1
#endif
/*
* Enable superpage reservations: 1 level.
*/
#ifndef VM_NRESERVLEVEL
#define VM_NRESERVLEVEL 1
#endif
/*
* Level 0 reservations consist of 512 pages.
*/
#ifndef VM_LEVEL_0_ORDER
#define VM_LEVEL_0_ORDER 9
#endif
/**
* Address space layout.
*
* UltraSPARC I and II implement a 44 bit virtual address space. The address
* space is split into 2 regions at each end of the 64 bit address space, with
* an out of range "hole" in the middle. UltraSPARC III implements the full
* 64 bit virtual address space, but we don't really have any use for it and
* 43 bits of user address space is considered to be "enough", so we ignore it.
*
* Upper region: 0xffffffffffffffff
2011-06-21 20:50:55 +00:00
* 0xfffff80000000000
*
* Hole: 0xfffff7ffffffffff
2011-06-21 20:50:55 +00:00
* 0x0000080000000000
*
* Lower region: 0x000007ffffffffff
2011-06-21 20:50:55 +00:00
* 0x0000000000000000
*
* In general we ignore the upper region, and use the lower region as mappable
* space.
*
* We define some interesting address constants:
*
* VM_MIN_ADDRESS and VM_MAX_ADDRESS define the start and end of the entire
* 64 bit address space, mostly just for convenience.
*
* VM_MIN_DIRECT_ADDRESS and VM_MAX_DIRECT_ADDRESS define the start and end
* of the direct mapped region. This maps virtual addresses to physical
* addresses directly using 4mb tlb entries, with the physical address encoded
* in the lower 43 bits of virtual address. These mappings are convenient
* because they do not require page tables, and because they never change they
* do not require tlb flushes. However, since these mappings are cacheable,
* we must ensure that all pages accessed this way are either not double
* mapped, or that all other mappings have virtual color equal to physical
* color, in order to avoid creating illegal aliases in the data cache.
*
* VM_MIN_KERNEL_ADDRESS and VM_MAX_KERNEL_ADDRESS define the start and end of
* mappable kernel virtual address space. VM_MIN_KERNEL_ADDRESS is basically
* arbitrary, a convenient address is chosen which allows both the kernel text
* and data and the prom's address space to be mapped with 1 4mb tsb page.
* VM_MAX_KERNEL_ADDRESS is variable, computed at startup time based on the
* amount of physical memory available. Each 4mb tsb page provides 1g of
* virtual address space, with the only practical limit being available
* phsyical memory.
*
* VM_MIN_PROM_ADDRESS and VM_MAX_PROM_ADDRESS define the start and end of the
* prom address space. On startup the prom's mappings are duplicated in the
* kernel tsb, to allow prom memory to be accessed normally by the kernel.
*
* VM_MIN_USER_ADDRESS and VM_MAX_USER_ADDRESS define the start and end of the
* user address space. There are some hardware errata about using addresses
* at the boundary of the va hole, so we allow just under 43 bits of user
* address space. Note that the kernel and user address spaces overlap, but
* this doesn't matter because they use different tlb contexts, and because
* the kernel address space is not mapped into each process' address space.
*/
#define VM_MIN_ADDRESS (0x0000000000000000UL)
#define VM_MAX_ADDRESS (0xffffffffffffffffUL)
#define VM_MIN_DIRECT_ADDRESS (0xfffff80000000000UL)
#define VM_MAX_DIRECT_ADDRESS (VM_MAX_ADDRESS)
#define VM_MIN_KERNEL_ADDRESS (0x00000000c0000000UL)
#define VM_MAX_KERNEL_ADDRESS (vm_max_kernel_address)
#define VM_MIN_PROM_ADDRESS (0x00000000f0000000UL)
#define VM_MAX_PROM_ADDRESS (0x00000000ffffffffUL)
#define VM_MIN_USER_ADDRESS (0x0000000000000000UL)
#define VM_MAX_USER_ADDRESS (0x000007fe00000000UL)
#define VM_MINUSER_ADDRESS (VM_MIN_USER_ADDRESS)
#define VM_MAXUSER_ADDRESS (VM_MAX_USER_ADDRESS)
#define KERNBASE (VM_MIN_KERNEL_ADDRESS)
#define PROMBASE (VM_MIN_PROM_ADDRESS)
#define USRSTACK (VM_MAX_USER_ADDRESS)
/*
* Virtual size (bytes) for various kernel submaps.
*/
#ifndef VM_KMEM_SIZE
#define VM_KMEM_SIZE (16*1024*1024)
#endif
/*
* How many physical pages per KVA page allocated.
* min(max(max(VM_KMEM_SIZE, Physical memory/VM_KMEM_SIZE_SCALE),
* VM_KMEM_SIZE_MIN), VM_KMEM_SIZE_MAX)
* is the total KVA space allocated for kmem_map.
*/
#ifndef VM_KMEM_SIZE_SCALE
#define VM_KMEM_SIZE_SCALE (tsb_kernel_ldd_phys == 0 ? 3 : 2)
#endif
/*
* Ceiling on amount of kmem_map kva space.
*/
#ifndef VM_KMEM_SIZE_MAX
#define VM_KMEM_SIZE_MAX ((VM_MAX_KERNEL_ADDRESS - \
VM_MIN_KERNEL_ADDRESS + 1) * 3 / 5)
#endif
/*
* Initial pagein size of beginning of executable file.
*/
#ifndef VM_INITIAL_PAGEIN
#define VM_INITIAL_PAGEIN 16
#endif
#define UMA_MD_SMALL_ALLOC
extern u_int tsb_kernel_ldd_phys;
extern vm_offset_t vm_max_kernel_address;
/*
* Older sparc64 machines have a virtually indexed L1 data cache of 16KB.
* Consequently, mapping the same physical page multiple times may have
* caching disabled.
*/
#define ZERO_REGION_SIZE PAGE_SIZE
#endif /* !_MACHINE_VMPARAM_H_ */