5 By: Peter Zijlstra <a.p.zijlstra@chello.nl>
12 High memory (highmem) is used when the size of physical memory approaches or
13 exceeds the maximum size of virtual memory. At that point it becomes
14 impossible for the kernel to keep all of the available physical memory mapped
15 at all times. This means the kernel needs to start using temporary mappings of
16 the pieces of physical memory that it wants to access.
18 The part of (physical) memory not covered by a permanent mapping is what we
19 refer to as 'highmem'. There are various architecture dependent constraints on
20 where exactly that border lies.
22 In the i386 arch, for example, we choose to map the kernel into every process's
23 VM space so that we don't have to pay the full TLB invalidation costs for
24 kernel entry/exit. This means the available virtual memory space (4GiB on
25 i386) has to be divided between user and kernel space.
27 The traditional split for architectures using this approach is 3:1, 3GiB for
28 userspace and the top 1GiB for kernel space::
38 This means that the kernel can at most map 1GiB of physical memory at any one
39 time, but because we need virtual address space for other things - including
40 temporary maps to access the rest of the physical memory - the actual direct
41 map will typically be less (usually around ~896MiB).
43 Other architectures that have mm context tagged TLBs can have separate kernel
44 and user maps. Some hardware (like some ARMs), however, have limited virtual
45 space when they use mm context tags.
48 Temporary Virtual Mappings
49 ==========================
51 The kernel contains several ways of creating temporary mappings. The following
52 list shows them in order of preference of use.
54 * kmap_local_page(). This function is used to require short term mappings.
55 It can be invoked from any context (including interrupts) but the mappings
56 can only be used in the context which acquired them.
58 This function should always be used, whereas kmap_atomic() and kmap() have
61 These mappings are thread-local and CPU-local, meaning that the mapping
62 can only be accessed from within this thread and the thread is bound to the
63 CPU while the mapping is active. Although preemption is never disabled by
64 this function, the CPU can not be unplugged from the system via
65 CPU-hotplug until the mapping is disposed.
67 It's valid to take pagefaults in a local kmap region, unless the context
68 in which the local mapping is acquired does not allow it for other reasons.
70 As said, pagefaults and preemption are never disabled. There is no need to
71 disable preemption because, when context switches to a different task, the
72 maps of the outgoing task are saved and those of the incoming one are
75 kmap_local_page() always returns a valid virtual address and it is assumed
76 that kunmap_local() will never fail.
78 On CONFIG_HIGHMEM=n kernels and for low memory pages this returns the
79 virtual address of the direct mapping. Only real highmem pages are
80 temporarily mapped. Therefore, users may call a plain page_address()
81 for pages which are known to not come from ZONE_HIGHMEM. However, it is
82 always safe to use kmap_local_page() / kunmap_local().
84 While it is significantly faster than kmap(), for the highmem case it
85 comes with restrictions about the pointers validity. Contrary to kmap()
86 mappings, the local mappings are only valid in the context of the caller
87 and cannot be handed to other contexts. This implies that users must
88 be absolutely sure to keep the use of the return address local to the
89 thread which mapped it.
91 Most code can be designed to use thread local mappings. User should
92 therefore try to design their code to avoid the use of kmap() by mapping
93 pages in the same thread the address will be used and prefer
96 Nesting kmap_local_page() and kmap_atomic() mappings is allowed to a certain
97 extent (up to KMAP_TYPE_NR) but their invocations have to be strictly ordered
98 because the map implementation is stack based. See kmap_local_page() kdocs
99 (included in the "Functions" section) for details on how to manage nested
102 * kmap_atomic(). This function has been deprecated; use kmap_local_page().
104 NOTE: Conversions to kmap_local_page() must take care to follow the mapping
105 restrictions imposed on kmap_local_page(). Furthermore, the code between
106 calls to kmap_atomic() and kunmap_atomic() may implicitly depend on the side
107 effects of atomic mappings, i.e. disabling page faults or preemption, or both.
108 In that case, explicit calls to pagefault_disable() or preempt_disable() or
109 both must be made in conjunction with the use of kmap_local_page().
111 [Legacy documentation]
113 This permits a very short duration mapping of a single page. Since the
114 mapping is restricted to the CPU that issued it, it performs well, but
115 the issuing task is therefore required to stay on that CPU until it has
116 finished, lest some other task displace its mappings.
118 kmap_atomic() may also be used by interrupt contexts, since it does not
119 sleep and the callers too may not sleep until after kunmap_atomic() is
122 Each call of kmap_atomic() in the kernel creates a non-preemptible section
123 and disable pagefaults. This could be a source of unwanted latency. Therefore
124 users should prefer kmap_local_page() instead of kmap_atomic().
126 It is assumed that k[un]map_atomic() won't fail.
128 * kmap(). This function has been deprecated; use kmap_local_page().
130 NOTE: Conversions to kmap_local_page() must take care to follow the mapping
131 restrictions imposed on kmap_local_page(). In particular, it is necessary to
132 make sure that the kernel virtual memory pointer is only valid in the thread
135 [Legacy documentation]
137 This should be used to make short duration mapping of a single page with no
138 restrictions on preemption or migration. It comes with an overhead as mapping
139 space is restricted and protected by a global lock for synchronization. When
140 mapping is no longer needed, the address that the page was mapped to must be
141 released with kunmap().
143 Mapping changes must be propagated across all the CPUs. kmap() also
144 requires global TLB invalidation when the kmap's pool wraps and it might
145 block when the mapping space is fully utilized until a slot becomes
146 available. Therefore, kmap() is only callable from preemptible context.
148 All the above work is necessary if a mapping must last for a relatively
149 long time but the bulk of high-memory mappings in the kernel are
150 short-lived and only used in one place. This means that the cost of
151 kmap() is mostly wasted in such cases. kmap() was not intended for long
152 term mappings but it has morphed in that direction and its use is
153 strongly discouraged in newer code and the set of the preceding functions
156 On 64-bit systems, calls to kmap_local_page(), kmap_atomic() and kmap() have
157 no real work to do because a 64-bit address space is more than sufficient to
158 address all the physical memory whose pages are permanently mapped.
160 * vmap(). This can be used to make a long duration mapping of multiple
161 physical pages into a contiguous virtual space. It needs global
162 synchronization to unmap.
165 Cost of Temporary Mappings
166 ==========================
168 The cost of creating temporary mappings can be quite high. The arch has to
169 manipulate the kernel's page tables, the data TLB and/or the MMU's registers.
171 If CONFIG_HIGHMEM is not set, then the kernel will try and create a mapping
172 simply with a bit of arithmetic that will convert the page struct address into
173 a pointer to the page contents rather than juggling mappings about. In such a
174 case, the unmap operation may be a null operation.
176 If CONFIG_MMU is not set, then there can be no temporary mappings and no
177 highmem. In such a case, the arithmetic approach will also be used.
183 The i386 arch, under some circumstances, will permit you to stick up to 64GiB
184 of RAM into your 32-bit machine. This has a number of consequences:
186 * Linux needs a page-frame structure for each page in the system and the
187 pageframes need to live in the permanent mapping, which means:
189 * you can have 896M/sizeof(struct page) page-frames at most; with struct
190 page being 32-bytes that would end up being something in the order of 112G
191 worth of pages; the kernel, however, needs to store more than just
192 page-frames in that memory...
194 * PAE makes your page tables larger - which slows the system down as more
195 data has to be accessed to traverse in TLB fills and the like. One
196 advantage is that PAE has more PTE bits and can provide advanced features
199 The general recommendation is that you don't use more than 8GiB on a 32-bit
200 machine - although more might work for you and your workload, you're pretty
201 much on your own - don't expect kernel developers to really care much if things
208 .. kernel-doc:: include/linux/highmem.h
209 .. kernel-doc:: mm/highmem.c
210 .. kernel-doc:: include/linux/highmem-internal.h