1.. _highmem: 2 3==================== 4High Memory Handling 5==================== 6 7By: Peter Zijlstra <a.p.zijlstra@chello.nl> 8 9.. contents:: :local: 10 11What Is High Memory? 12==================== 13 14High memory (highmem) is used when the size of physical memory approaches or 15exceeds the maximum size of virtual memory. At that point it becomes 16impossible for the kernel to keep all of the available physical memory mapped 17at all times. This means the kernel needs to start using temporary mappings of 18the pieces of physical memory that it wants to access. 19 20The part of (physical) memory not covered by a permanent mapping is what we 21refer to as 'highmem'. There are various architecture dependent constraints on 22where exactly that border lies. 23 24In the i386 arch, for example, we choose to map the kernel into every process's 25VM space so that we don't have to pay the full TLB invalidation costs for 26kernel entry/exit. This means the available virtual memory space (4GiB on 27i386) has to be divided between user and kernel space. 28 29The traditional split for architectures using this approach is 3:1, 3GiB for 30userspace and the top 1GiB for kernel space:: 31 32 +--------+ 0xffffffff 33 | Kernel | 34 +--------+ 0xc0000000 35 | | 36 | User | 37 | | 38 +--------+ 0x00000000 39 40This means that the kernel can at most map 1GiB of physical memory at any one 41time, but because we need virtual address space for other things - including 42temporary maps to access the rest of the physical memory - the actual direct 43map will typically be less (usually around ~896MiB). 44 45Other architectures that have mm context tagged TLBs can have separate kernel 46and user maps. Some hardware (like some ARMs), however, have limited virtual 47space when they use mm context tags. 48 49 50Temporary Virtual Mappings 51========================== 52 53The kernel contains several ways of creating temporary mappings. The following 54list shows them in order of preference of use. 55 56* kmap_local_page(). This function is used to require short term mappings. 57 It can be invoked from any context (including interrupts) but the mappings 58 can only be used in the context which acquired them. 59 60 This function should always be used, whereas kmap_atomic() and kmap() have 61 been deprecated. 62 63 These mappings are thread-local and CPU-local, meaning that the mapping 64 can only be accessed from within this thread and the thread is bound to the 65 CPU while the mapping is active. Although preemption is never disabled by 66 this function, the CPU can not be unplugged from the system via 67 CPU-hotplug until the mapping is disposed. 68 69 It's valid to take pagefaults in a local kmap region, unless the context 70 in which the local mapping is acquired does not allow it for other reasons. 71 72 As said, pagefaults and preemption are never disabled. There is no need to 73 disable preemption because, when context switches to a different task, the 74 maps of the outgoing task are saved and those of the incoming one are 75 restored. 76 77 kmap_local_page() always returns a valid virtual address and it is assumed 78 that kunmap_local() will never fail. 79 80 On CONFIG_HIGHMEM=n kernels and for low memory pages this returns the 81 virtual address of the direct mapping. Only real highmem pages are 82 temporarily mapped. Therefore, users may call a plain page_address() 83 for pages which are known to not come from ZONE_HIGHMEM. However, it is 84 always safe to use kmap_local_page() / kunmap_local(). 85 86 While it is significantly faster than kmap(), for the higmem case it 87 comes with restrictions about the pointers validity. Contrary to kmap() 88 mappings, the local mappings are only valid in the context of the caller 89 and cannot be handed to other contexts. This implies that users must 90 be absolutely sure to keep the use of the return address local to the 91 thread which mapped it. 92 93 Most code can be designed to use thread local mappings. User should 94 therefore try to design their code to avoid the use of kmap() by mapping 95 pages in the same thread the address will be used and prefer 96 kmap_local_page(). 97 98 Nesting kmap_local_page() and kmap_atomic() mappings is allowed to a certain 99 extent (up to KMAP_TYPE_NR) but their invocations have to be strictly ordered 100 because the map implementation is stack based. See kmap_local_page() kdocs 101 (included in the "Functions" section) for details on how to manage nested 102 mappings. 103 104* kmap_atomic(). This function has been deprecated; use kmap_local_page(). 105 106 NOTE: Conversions to kmap_local_page() must take care to follow the mapping 107 restrictions imposed on kmap_local_page(). Furthermore, the code between 108 calls to kmap_atomic() and kunmap_atomic() may implicitly depend on the side 109 effects of atomic mappings, i.e. disabling page faults or preemption, or both. 110 In that case, explicit calls to pagefault_disable() or preempt_disable() or 111 both must be made in conjunction with the use of kmap_local_page(). 112 113 [Legacy documentation] 114 115 This permits a very short duration mapping of a single page. Since the 116 mapping is restricted to the CPU that issued it, it performs well, but 117 the issuing task is therefore required to stay on that CPU until it has 118 finished, lest some other task displace its mappings. 119 120 kmap_atomic() may also be used by interrupt contexts, since it does not 121 sleep and the callers too may not sleep until after kunmap_atomic() is 122 called. 123 124 Each call of kmap_atomic() in the kernel creates a non-preemptible section 125 and disable pagefaults. This could be a source of unwanted latency. Therefore 126 users should prefer kmap_local_page() instead of kmap_atomic(). 127 128 It is assumed that k[un]map_atomic() won't fail. 129 130* kmap(). This function has been deprecated; use kmap_local_page(). 131 132 NOTE: Conversions to kmap_local_page() must take care to follow the mapping 133 restrictions imposed on kmap_local_page(). In particular, it is necessary to 134 make sure that the kernel virtual memory pointer is only valid in the thread 135 that obtained it. 136 137 [Legacy documentation] 138 139 This should be used to make short duration mapping of a single page with no 140 restrictions on preemption or migration. It comes with an overhead as mapping 141 space is restricted and protected by a global lock for synchronization. When 142 mapping is no longer needed, the address that the page was mapped to must be 143 released with kunmap(). 144 145 Mapping changes must be propagated across all the CPUs. kmap() also 146 requires global TLB invalidation when the kmap's pool wraps and it might 147 block when the mapping space is fully utilized until a slot becomes 148 available. Therefore, kmap() is only callable from preemptible context. 149 150 All the above work is necessary if a mapping must last for a relatively 151 long time but the bulk of high-memory mappings in the kernel are 152 short-lived and only used in one place. This means that the cost of 153 kmap() is mostly wasted in such cases. kmap() was not intended for long 154 term mappings but it has morphed in that direction and its use is 155 strongly discouraged in newer code and the set of the preceding functions 156 should be preferred. 157 158 On 64-bit systems, calls to kmap_local_page(), kmap_atomic() and kmap() have 159 no real work to do because a 64-bit address space is more than sufficient to 160 address all the physical memory whose pages are permanently mapped. 161 162* vmap(). This can be used to make a long duration mapping of multiple 163 physical pages into a contiguous virtual space. It needs global 164 synchronization to unmap. 165 166 167Cost of Temporary Mappings 168========================== 169 170The cost of creating temporary mappings can be quite high. The arch has to 171manipulate the kernel's page tables, the data TLB and/or the MMU's registers. 172 173If CONFIG_HIGHMEM is not set, then the kernel will try and create a mapping 174simply with a bit of arithmetic that will convert the page struct address into 175a pointer to the page contents rather than juggling mappings about. In such a 176case, the unmap operation may be a null operation. 177 178If CONFIG_MMU is not set, then there can be no temporary mappings and no 179highmem. In such a case, the arithmetic approach will also be used. 180 181 182i386 PAE 183======== 184 185The i386 arch, under some circumstances, will permit you to stick up to 64GiB 186of RAM into your 32-bit machine. This has a number of consequences: 187 188* Linux needs a page-frame structure for each page in the system and the 189 pageframes need to live in the permanent mapping, which means: 190 191* you can have 896M/sizeof(struct page) page-frames at most; with struct 192 page being 32-bytes that would end up being something in the order of 112G 193 worth of pages; the kernel, however, needs to store more than just 194 page-frames in that memory... 195 196* PAE makes your page tables larger - which slows the system down as more 197 data has to be accessed to traverse in TLB fills and the like. One 198 advantage is that PAE has more PTE bits and can provide advanced features 199 like NX and PAT. 200 201The general recommendation is that you don't use more than 8GiB on a 32-bit 202machine - although more might work for you and your workload, you're pretty 203much on your own - don't expect kernel developers to really care much if things 204come apart. 205 206 207Functions 208========= 209 210.. kernel-doc:: include/linux/highmem.h 211.. kernel-doc:: include/linux/highmem-internal.h 212