tsphinx.addnodesdocument)}( rawsourcechildren]( translations LanguagesNode)}(hhh](h pending_xref)}(hhh]docutils.nodesTextChinese (Simplified)}parenthsba attributes}(ids]classes]names]dupnames]backrefs] refdomainstdreftypedoc reftarget*/translations/zh_CN/core-api/dma-api-howtomodnameN classnameN refexplicitutagnamehhh ubh)}(hhh]hChinese (Traditional)}hh2sbah}(h]h ]h"]h$]h&] refdomainh)reftypeh+ reftarget*/translations/zh_TW/core-api/dma-api-howtomodnameN classnameN refexplicituh1hhh ubh)}(hhh]hItalian}hhFsbah}(h]h ]h"]h$]h&] refdomainh)reftypeh+ reftarget*/translations/it_IT/core-api/dma-api-howtomodnameN classnameN refexplicituh1hhh ubh)}(hhh]hJapanese}hhZsbah}(h]h ]h"]h$]h&] refdomainh)reftypeh+ reftarget*/translations/ja_JP/core-api/dma-api-howtomodnameN classnameN refexplicituh1hhh ubh)}(hhh]hKorean}hhnsbah}(h]h ]h"]h$]h&] refdomainh)reftypeh+ reftarget*/translations/ko_KR/core-api/dma-api-howtomodnameN classnameN refexplicituh1hhh ubh)}(hhh]hSpanish}hhsbah}(h]h ]h"]h$]h&] refdomainh)reftypeh+ reftarget*/translations/sp_SP/core-api/dma-api-howtomodnameN classnameN refexplicituh1hhh ubeh}(h]h ]h"]h$]h&]current_languageEnglishuh1h hh _documenthsourceNlineNubhsection)}(hhh](htitle)}(hDynamic DMA mapping Guideh]hDynamic DMA mapping Guide}(hhhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhhhD/var/lib/git/docbuild/linux/Documentation/core-api/dma-api-howto.rsthKubh field_list)}(hhh](hfield)}(hhh](h field_name)}(hAuthorh]hAuthor}(hhhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhhhKubh field_body)}(h"David S. Miller h]h paragraph)}(hhh](hDavid S. Miller <}(hhhhhNhNubh reference)}(hdavem@redhat.comh]hdavem@redhat.com}(hhhhhNhNubah}(h]h ]h"]h$]h&]refurimailto:davem@redhat.comuh1hhhubh>}(hhhhhNhNubeh}(h]h ]h"]h$]h&]uh1hhhhKhhubah}(h]h ]h"]h$]h&]uh1hhhubeh}(h]h ]h"]h$]h&]uh1hhhhKhhhhubh)}(hhh](h)}(hAuthorh]hAuthor}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhjhhhKubh)}(h"Richard Henderson h]h)}(hjh](hRichard Henderson <}(hjhhhNhNubh)}(hrth@cygnus.comh]hrth@cygnus.com}(hj$hhhNhNubah}(h]h ]h"]h$]h&]refurimailto:rth@cygnus.comuh1hhjubh>}(hjhhhNhNubeh}(h]h ]h"]h$]h&]uh1hhhhKhjubah}(h]h ]h"]h$]h&]uh1hhjubeh}(h]h ]h"]h$]h&]uh1hhhhKhhhhubh)}(hhh](h)}(hAuthorh]hAuthor}(hjMhhhNhNubah}(h]h ]h"]h$]h&]uh1hhjJhhhKubh)}(h!Jakub Jelinek h]h)}(h Jakub Jelinek h](hJakub Jelinek <}(hj_hhhNhNubh)}(hjakub@redhat.comh]hjakub@redhat.com}(hjghhhNhNubah}(h]h ]h"]h$]h&]refurimailto:jakub@redhat.comuh1hhj_ubh>}(hj_hhhNhNubeh}(h]h ]h"]h$]h&]uh1hhhhKhj[ubah}(h]h ]h"]h$]h&]uh1hhjJubeh}(h]h ]h"]h$]h&]uh1hhhhKhhhhubeh}(h]h ]h"]h$]h&]uh1hhhhhhhhKubh)}(hThis is a guide to device driver writers on how to use the DMA API with example pseudo-code. For a concise description of the API, see Documentation/core-api/dma-api.rst.h]hThis is a guide to device driver writers on how to use the DMA API with example pseudo-code. For a concise description of the API, see Documentation/core-api/dma-api.rst.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhK hhhhubh)}(hhh](h)}(hCPU and DMA addressesh]hCPU and DMA addresses}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhjhhhhhKubh)}(hoThere are several kinds of addresses involved in the DMA API, and it's important to understand the differences.h]hqThere are several kinds of addresses involved in the DMA API, and it’s important to understand the differences.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hThe kernel normally uses virtual addresses. Any address returned by kmalloc(), vmalloc(), and similar interfaces is a virtual address and can be stored in a ``void *``.h](hThe kernel normally uses virtual addresses. Any address returned by kmalloc(), vmalloc(), and similar interfaces is a virtual address and can be stored in a }(hjhhhNhNubhliteral)}(h ``void *``h]hvoid *}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1jhjubh.}(hjhhhNhNubeh}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hXThe virtual memory system (TLB, page tables, etc.) translates virtual addresses to CPU physical addresses, which are stored as "phys_addr_t" or "resource_size_t". The kernel manages device resources like registers as physical addresses. These are the addresses in /proc/iomem. The physical address is not directly useful to a driver; it must use ioremap() to map the space and produce a virtual address.h]hXThe virtual memory system (TLB, page tables, etc.) translates virtual addresses to CPU physical addresses, which are stored as “phys_addr_t” or “resource_size_t”. The kernel manages device resources like registers as physical addresses. These are the addresses in /proc/iomem. The physical address is not directly useful to a driver; it must use ioremap() to map the space and produce a virtual address.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hXI/O devices use a third kind of address: a "bus address". If a device has registers at an MMIO address, or if it performs DMA to read or write system memory, the addresses used by the device are bus addresses. In some systems, bus addresses are identical to CPU physical addresses, but in general they are not. IOMMUs and host bridges can produce arbitrary mappings between physical and bus addresses.h]hXI/O devices use a third kind of address: a “bus address”. If a device has registers at an MMIO address, or if it performs DMA to read or write system memory, the addresses used by the device are bus addresses. In some systems, bus addresses are identical to CPU physical addresses, but in general they are not. IOMMUs and host bridges can produce arbitrary mappings between physical and bus addresses.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hXFrom a device's point of view, DMA uses the bus address space, but it may be restricted to a subset of that space. For example, even if a system supports 64-bit addresses for main memory and PCI BARs, it may use an IOMMU so devices only need to use 32-bit DMA addresses.h]hXFrom a device’s point of view, DMA uses the bus address space, but it may be restricted to a subset of that space. For example, even if a system supports 64-bit addresses for main memory and PCI BARs, it may use an IOMMU so devices only need to use 32-bit DMA addresses.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhK%hjhhubh)}(h$Here's a picture and some examples::h]h%Here’s a picture and some examples:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhK*hjhhubh literal_block)}(hX CPU CPU Bus Virtual Physical Address Address Address Space Space Space +-------+ +------+ +------+ | | |MMIO | Offset | | | | Virtual |Space | applied | | C +-------+ --------> B +------+ ----------> +------+ A | | mapping | | by host | | +-----+ | | | | bridge | | +--------+ | | | | +------+ | | | | | CPU | | | | RAM | | | | Device | | | | | | | | | | | +-----+ +-------+ +------+ +------+ +--------+ | | Virtual |Buffer| Mapping | | X +-------+ --------> Y +------+ <---------- +------+ Z | | mapping | RAM | by IOMMU | | | | | | | | +-------+ +------+h]hX CPU CPU Bus Virtual Physical Address Address Address Space Space Space +-------+ +------+ +------+ | | |MMIO | Offset | | | | Virtual |Space | applied | | C +-------+ --------> B +------+ ----------> +------+ A | | mapping | | by host | | +-----+ | | | | bridge | | +--------+ | | | | +------+ | | | | | CPU | | | | RAM | | | | Device | | | | | | | | | | | +-----+ +-------+ +------+ +------+ +--------+ | | Virtual |Buffer| Mapping | | X +-------+ --------> Y +------+ <---------- +------+ Z | | mapping | RAM | by IOMMU | | | | | | | | +-------+ +------+}hjsbah}(h]h ]h"]h$]h&] xml:spacepreserveuh1jhhhK,hjhhubh)}(hX1During the enumeration process, the kernel learns about I/O devices and their MMIO space and the host bridges that connect them to the system. For example, if a PCI device has a BAR, the kernel reads the bus address (A) from the BAR and converts it to a CPU physical address (B). The address B is stored in a struct resource and usually exposed via /proc/iomem. When a driver claims a device, it typically uses ioremap() to map physical address B at a virtual address (C). It can then use, e.g., ioread32(C), to access the device registers at bus address A.h]hX1During the enumeration process, the kernel learns about I/O devices and their MMIO space and the host bridges that connect them to the system. For example, if a PCI device has a BAR, the kernel reads the bus address (A) from the BAR and converts it to a CPU physical address (B). The address B is stored in a struct resource and usually exposed via /proc/iomem. When a driver claims a device, it typically uses ioremap() to map physical address B at a virtual address (C). It can then use, e.g., ioread32(C), to access the device registers at bus address A.}(hj,hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKBhjhhubh)}(hXfIf the device supports DMA, the driver sets up a buffer using kmalloc() or a similar interface, which returns a virtual address (X). The virtual memory system maps X to a physical address (Y) in system RAM. The driver can use virtual address X to access the buffer, but the device itself cannot because DMA doesn't go through the CPU virtual memory system.h]hXhIf the device supports DMA, the driver sets up a buffer using kmalloc() or a similar interface, which returns a virtual address (X). The virtual memory system maps X to a physical address (Y) in system RAM. The driver can use virtual address X to access the buffer, but the device itself cannot because DMA doesn’t go through the CPU virtual memory system.}(hj:hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKKhjhhubh)}(hXIn some simple systems, the device can do DMA directly to physical address Y. But in many others, there is IOMMU hardware that translates DMA addresses to physical addresses, e.g., it translates Z to Y. This is part of the reason for the DMA API: the driver can give a virtual address X to an interface like dma_map_single(), which sets up any required IOMMU mapping and returns the DMA address Z. The driver then tells the device to do DMA to Z, and the IOMMU maps it to the buffer at address Y in system RAM.h]hXIn some simple systems, the device can do DMA directly to physical address Y. But in many others, there is IOMMU hardware that translates DMA addresses to physical addresses, e.g., it translates Z to Y. This is part of the reason for the DMA API: the driver can give a virtual address X to an interface like dma_map_single(), which sets up any required IOMMU mapping and returns the DMA address Z. The driver then tells the device to do DMA to Z, and the IOMMU maps it to the buffer at address Y in system RAM.}(hjHhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKQhjhhubh)}(hSo that Linux can use the dynamic DMA mapping, it needs some help from the drivers, namely it has to take into account that DMA addresses should be mapped only for the time they are actually used and unmapped after the DMA transfer.h]hSo that Linux can use the dynamic DMA mapping, it needs some help from the drivers, namely it has to take into account that DMA addresses should be mapped only for the time they are actually used and unmapped after the DMA transfer.}(hjVhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKZhjhhubh)}(hVThe following API will work of course even on platforms where no such hardware exists.h]hVThe following API will work of course even on platforms where no such hardware exists.}(hjdhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhK_hjhhubh)}(hNote that the DMA API works with any bus independent of the underlying microprocessor architecture. You should use the DMA API rather than the bus-specific DMA API, i.e., use the dma_map_*() interfaces rather than the pci_map_*() interfaces.h]hNote that the DMA API works with any bus independent of the underlying microprocessor architecture. You should use the DMA API rather than the bus-specific DMA API, i.e., use the dma_map_*() interfaces rather than the pci_map_*() interfaces.}(hjrhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKbhjhhubh)}(h$First of all, you should make sure::h]h#First of all, you should make sure:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKghjhhubj)}(h#include h]h#include }hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhKihjhhubh)}(his in your driver, which provides the definition of dma_addr_t. This type can hold any valid DMA address for the platform and should be used everywhere you hold a DMA address returned from the DMA mapping functions.h]his in your driver, which provides the definition of dma_addr_t. This type can hold any valid DMA address for the platform and should be used everywhere you hold a DMA address returned from the DMA mapping functions.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKkhjhhubeh}(h]cpu-and-dma-addressesah ]h"]cpu and dma addressesah$]h&]uh1hhhhhhhhKubh)}(hhh](h)}(hWhat memory is DMA'able?h]hWhat memory is DMA’able?}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhjhhhhhKpubh)}(hThe first piece of information you must know is what kernel memory can be used with the DMA mapping facilities. There has been an unwritten set of rules regarding this, and this text is an attempt to finally write them down.h]hThe first piece of information you must know is what kernel memory can be used with the DMA mapping facilities. There has been an unwritten set of rules regarding this, and this text is an attempt to finally write them down.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKrhjhhubh)}(hIf you acquired your memory via the page allocator (i.e. __get_free_page*()) or the generic memory allocators (i.e. kmalloc() or kmem_cache_alloc()) then you may DMA to/from that memory using the addresses returned from those routines.h]hIf you acquired your memory via the page allocator (i.e. __get_free_page*()) or the generic memory allocators (i.e. kmalloc() or kmem_cache_alloc()) then you may DMA to/from that memory using the addresses returned from those routines.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKwhjhhubh)}(hXThis means specifically that you may _not_ use the memory/addresses returned from vmalloc() for DMA. It is possible to DMA to the _underlying_ memory mapped into a vmalloc() area, but this requires walking page tables to get the physical addresses, and then translating each of those pages back to a kernel address using something like __va(). [ EDIT: Update this when we integrate Gerd Knorr's generic code which does this. ]h]hXThis means specifically that you may _not_ use the memory/addresses returned from vmalloc() for DMA. It is possible to DMA to the _underlying_ memory mapped into a vmalloc() area, but this requires walking page tables to get the physical addresses, and then translating each of those pages back to a kernel address using something like __va(). [ EDIT: Update this when we integrate Gerd Knorr’s generic code which does this. ]}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhK|hjhhubh)}(hXfThis rule also means that you may use neither kernel image addresses (items in data/text/bss segments), nor module image addresses, nor stack addresses for DMA. These could all be mapped somewhere entirely different than the rest of physical memory. Even if those classes of memory could physically work with DMA, you'd need to ensure the I/O buffers were cacheline-aligned. Without that, you'd see cacheline sharing problems (data corruption) on CPUs with DMA-incoherent caches. (The CPU could write to one word, DMA would write to a different one in the same cache line, and one of them could be overwritten.)h]hXjThis rule also means that you may use neither kernel image addresses (items in data/text/bss segments), nor module image addresses, nor stack addresses for DMA. These could all be mapped somewhere entirely different than the rest of physical memory. Even if those classes of memory could physically work with DMA, you’d need to ensure the I/O buffers were cacheline-aligned. Without that, you’d see cacheline sharing problems (data corruption) on CPUs with DMA-incoherent caches. (The CPU could write to one word, DMA would write to a different one in the same cache line, and one of them could be overwritten.)}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hvAlso, this means that you cannot take the return of a kmap() call and DMA to/from that. This is similar to vmalloc().h]hvAlso, this means that you cannot take the return of a kmap() call and DMA to/from that. This is similar to vmalloc().}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hWhat about block I/O and networking buffers? The block I/O and networking subsystems make sure that the buffers they use are valid for you to DMA from/to.h]hWhat about block I/O and networking buffers? The block I/O and networking subsystems make sure that the buffers they use are valid for you to DMA from/to.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubeh}(h]what-memory-is-dma-ableah ]h"]what memory is dma'able?ah$]h&]uh1hhhhhhhhKpubh)}(hhh](h)}(hDMA addressing capabilitiesh]hDMA addressing capabilities}(hj"hhhNhNubah}(h]h ]h"]h$]h&]uh1hhjhhhhhKubh)}(hBy default, the kernel assumes that your device can address 32-bits of DMA addressing. For a 64-bit capable device, this needs to be increased, and for a device with limitations, it needs to be decreased.h]hBy default, the kernel assumes that your device can address 32-bits of DMA addressing. For a 64-bit capable device, this needs to be increased, and for a device with limitations, it needs to be decreased.}(hj0hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hSpecial note about PCI: PCI-X specification requires PCI-X devices to support 64-bit addressing (DAC) for all transactions. And at least one platform (SGI SN2) requires 64-bit consistent allocations to operate correctly when the IO bus is in PCI-X mode.h]hSpecial note about PCI: PCI-X specification requires PCI-X devices to support 64-bit addressing (DAC) for all transactions. And at least one platform (SGI SN2) requires 64-bit consistent allocations to operate correctly when the IO bus is in PCI-X mode.}(hj>hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(huFor correct operation, you must set the DMA mask to inform the kernel about your devices DMA addressing capabilities.h]huFor correct operation, you must set the DMA mask to inform the kernel about your devices DMA addressing capabilities.}(hjLhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(h=This is performed via a call to dma_set_mask_and_coherent()::h]hdev is a pointer to the device struct of a PCI device (pdev is a pointer to the PCI device struct of your device).h]hXzHere, dev is a pointer to the device struct of your device, and mask is a bit mask describing which bits of an address your device supports. Often the device struct of your device is embedded in the bus-specific device struct of your device. For example, &pdev->dev is a pointer to the device struct of a PCI device (pdev is a pointer to the PCI device struct of your device).}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hXThese calls usually return zero to indicate your device can perform DMA properly on the machine given the address mask you provided, but they might return an error if the mask is too small to be supportable on the given system. If it returns non-zero, your device cannot perform DMA properly on this platform, and attempting to do so will result in undefined behavior. You must not use DMA on this device unless the dma_set_mask family of functions has returned success.h]hXThese calls usually return zero to indicate your device can perform DMA properly on the machine given the address mask you provided, but they might return an error if the mask is too small to be supportable on the given system. If it returns non-zero, your device cannot perform DMA properly on this platform, and attempting to do so will result in undefined behavior. You must not use DMA on this device unless the dma_set_mask family of functions has returned success.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(h:This means that in the failure case, you have two options:h]h:This means that in the failure case, you have two options:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubhenumerated_list)}(hhh](h list_item)}(h5Use some non-DMA mode for data transfer, if possible.h]h)}(hjh]h5Use some non-DMA mode for data transfer, if possible.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjubah}(h]h ]h"]h$]h&]uh1jhjhhhhhNubj)}(h-Ignore this device and do not initialize it. h]h)}(h,Ignore this device and do not initialize it.h]h,Ignore this device and do not initialize it.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjubah}(h]h ]h"]h$]h&]uh1jhjhhhhhNubeh}(h]h ]h"]h$]h&]enumtypearabicprefixhsuffix)uh1jhjhhhhhKubh)}(hXIt is recommended that your driver print a kernel KERN_WARNING message when setting the DMA mask fails. In this manner, if a user of your driver reports that performance is bad or that the device is not even detected, you can ask them for the kernel messages to find out exactly why.h]hXIt is recommended that your driver print a kernel KERN_WARNING message when setting the DMA mask fails. In this manner, if a user of your driver reports that performance is bad or that the device is not even detected, you can ask them for the kernel messages to find out exactly why.}(hj3hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(h;The 24-bit addressing device would do something like this::h]h:The 24-bit addressing device would do something like this:}(hjAhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubj)}(hif (dma_set_mask_and_coherent(dev, DMA_BIT_MASK(24))) { dev_warn(dev, "mydev: No suitable DMA available\n"); goto ignore_this_device; }h]hif (dma_set_mask_and_coherent(dev, DMA_BIT_MASK(24))) { dev_warn(dev, "mydev: No suitable DMA available\n"); goto ignore_this_device; }}hjOsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhKhjhhubh)}(hDThe standard 64-bit addressing device would do something like this::h]hCThe standard 64-bit addressing device would do something like this:}(hj]hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubj)}(h0dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64))h]h0dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64))}hjksbah}(h]h ]h"]h$]h&]j*j+uh1jhhhKhjhhubh)}(h^dma_set_mask_and_coherent() never return fail when DMA_BIT_MASK(64). Typical error code like::h]h]dma_set_mask_and_coherent() never return fail when DMA_BIT_MASK(64). Typical error code like:}(hjyhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubj)}(h/* Wrong code */ if (dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64))) dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32))h]h/* Wrong code */ if (dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64))) dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32))}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhKhjhhubh)}(hadma_set_mask_and_coherent() will never return failure when bigger than 32. So typical code like::h]h`dma_set_mask_and_coherent() will never return failure when bigger than 32. So typical code like:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubj)}(h/* Recommended code */ if (support_64bit) dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64)); else dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));h]h/* Recommended code */ if (support_64bit) dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64)); else dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhKhjhhubh)}(hIf the device only supports 32-bit addressing for descriptors in the coherent allocations, but supports full 64-bits for streaming mappings it would look like this::h]hIf the device only supports 32-bit addressing for descriptors in the coherent allocations, but supports full 64-bits for streaming mappings it would look like this:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubj)}(hif (dma_set_mask(dev, DMA_BIT_MASK(64))) { dev_warn(dev, "mydev: No suitable DMA available\n"); goto ignore_this_device; }h]hif (dma_set_mask(dev, DMA_BIT_MASK(64))) { dev_warn(dev, "mydev: No suitable DMA available\n"); goto ignore_this_device; }}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhKhjhhubh)}(hThe coherent mask will always be able to set the same or a smaller mask as the streaming mask. However for the rare case that a device driver only uses consistent allocations, one would have to check the return value from dma_set_coherent_mask().h]hThe coherent mask will always be able to set the same or a smaller mask as the streaming mask. However for the rare case that a device driver only uses consistent allocations, one would have to check the return value from dma_set_coherent_mask().}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(h_Finally, if your device can only drive the low 24-bits of address you might do something like::h]h^Finally, if your device can only drive the low 24-bits of address you might do something like:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubj)}(hif (dma_set_mask(dev, DMA_BIT_MASK(24))) { dev_warn(dev, "mydev: 24-bit DMA addressing not available\n"); goto ignore_this_device; }h]hif (dma_set_mask(dev, DMA_BIT_MASK(24))) { dev_warn(dev, "mydev: 24-bit DMA addressing not available\n"); goto ignore_this_device; }}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhKhjhhubh)}(hWhen dma_set_mask() or dma_set_mask_and_coherent() is successful, and returns zero, the kernel saves away this mask you have provided. The kernel will use this information later when you make DMA mappings.h]hWhen dma_set_mask() or dma_set_mask_and_coherent() is successful, and returns zero, the kernel saves away this mask you have provided. The kernel will use this information later when you make DMA mappings.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhKhjhhubh)}(hXThere is a case which we are aware of at this time, which is worth mentioning in this documentation. If your device supports multiple functions (for example a sound card provides playback and record functions) and the various different functions have _different_ DMA addressing limitations, you may wish to probe each mask and only provide the functionality which the machine can handle. It is important that the last call to dma_set_mask() be for the most specific mask.h]hXThere is a case which we are aware of at this time, which is worth mentioning in this documentation. If your device supports multiple functions (for example a sound card provides playback and record functions) and the various different functions have _different_ DMA addressing limitations, you may wish to probe each mask and only provide the functionality which the machine can handle. It is important that the last call to dma_set_mask() be for the most specific mask.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(h4Here is pseudo-code showing how this might be done::h]h3Here is pseudo-code showing how this might be done:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM hjhhubj)}(hX_#define PLAYBACK_ADDRESS_BITS DMA_BIT_MASK(32) #define RECORD_ADDRESS_BITS DMA_BIT_MASK(24) struct my_sound_card *card; struct device *dev; ... if (!dma_set_mask(dev, PLAYBACK_ADDRESS_BITS)) { card->playback_enabled = 1; } else { card->playback_enabled = 0; dev_warn(dev, "%s: Playback disabled due to DMA limitations\n", card->name); } if (!dma_set_mask(dev, RECORD_ADDRESS_BITS)) { card->record_enabled = 1; } else { card->record_enabled = 0; dev_warn(dev, "%s: Record disabled due to DMA limitations\n", card->name); }h]hX_#define PLAYBACK_ADDRESS_BITS DMA_BIT_MASK(32) #define RECORD_ADDRESS_BITS DMA_BIT_MASK(24) struct my_sound_card *card; struct device *dev; ... if (!dma_set_mask(dev, PLAYBACK_ADDRESS_BITS)) { card->playback_enabled = 1; } else { card->playback_enabled = 0; dev_warn(dev, "%s: Playback disabled due to DMA limitations\n", card->name); } if (!dma_set_mask(dev, RECORD_ADDRESS_BITS)) { card->record_enabled = 1; } else { card->record_enabled = 0; dev_warn(dev, "%s: Record disabled due to DMA limitations\n", card->name); }}hj!sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhjhhubh)}(hA sound card was used as an example here because this genre of PCI devices seems to be littered with ISA chips given a PCI front end, and thus retaining the 16MB DMA addressing limitations of ISA.h]hA sound card was used as an example here because this genre of PCI devices seems to be littered with ISA chips given a PCI front end, and thus retaining the 16MB DMA addressing limitations of ISA.}(hj/hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM$hjhhubeh}(h]dma-addressing-capabilitiesah ]h"]dma addressing capabilitiesah$]h&]uh1hhhhhhhhKubh)}(hhh](h)}(hTypes of DMA mappingsh]hTypes of DMA mappings}(hjHhhhNhNubah}(h]h ]h"]h$]h&]uh1hhjEhhhhhM)ubh)}(h$There are two types of DMA mappings:h]h$There are two types of DMA mappings:}(hjVhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM+hjEhhubh bullet_list)}(hhh](j)}(hXConsistent DMA mappings which are usually mapped at driver initialization, unmapped at the end and for which the hardware should guarantee that the device and the CPU can access the data in parallel and will see updates made by each other without any explicit software flushing. Think of "consistent" as "synchronous" or "coherent". The current default is to return consistent memory in the low 32 bits of the DMA space. However, for future compatibility you should set the consistent mask even if this default is fine for your driver. Good examples of what to use consistent mappings for are: - Network card DMA ring descriptors. - SCSI adapter mailbox command data structures. - Device firmware microcode executed out of main memory. The invariant these examples all require is that any CPU store to memory is immediately visible to the device, and vice versa. Consistent mappings guarantee this. .. important:: Consistent DMA memory does not preclude the usage of proper memory barriers. The CPU may reorder stores to consistent memory just as it may normal memory. Example: if it is important for the device to see the first word of a descriptor updated before the second, you must do something like:: desc->word0 = address; wmb(); desc->word1 = DESC_VALID; in order to get correct behavior on all platforms. Also, on some platforms your driver may need to flush CPU write buffers in much the same way as it needs to flush write buffers found in PCI bridges (such as by reading a register's value after writing it). h](h)}(hXConsistent DMA mappings which are usually mapped at driver initialization, unmapped at the end and for which the hardware should guarantee that the device and the CPU can access the data in parallel and will see updates made by each other without any explicit software flushing.h]hXConsistent DMA mappings which are usually mapped at driver initialization, unmapped at the end and for which the hardware should guarantee that the device and the CPU can access the data in parallel and will see updates made by each other without any explicit software flushing.}(hjmhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM-hjiubh)}(h5Think of "consistent" as "synchronous" or "coherent".h]hAThink of “consistent” as “synchronous” or “coherent”.}(hj{hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM3hjiubh)}(hThe current default is to return consistent memory in the low 32 bits of the DMA space. However, for future compatibility you should set the consistent mask even if this default is fine for your driver.h]hThe current default is to return consistent memory in the low 32 bits of the DMA space. However, for future compatibility you should set the consistent mask even if this default is fine for your driver.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM5hjiubh)}(h9Good examples of what to use consistent mappings for are:h]h9Good examples of what to use consistent mappings for are:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM:hjiubj)}(h- Network card DMA ring descriptors. - SCSI adapter mailbox command data structures. - Device firmware microcode executed out of main memory. h]je)}(hhh](j)}(h"Network card DMA ring descriptors.h]h)}(hjh]h"Network card DMA ring descriptors.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM<hjubah}(h]h ]h"]h$]h&]uh1jhjubj)}(h-SCSI adapter mailbox command data structures.h]h)}(hjh]h-SCSI adapter mailbox command data structures.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM=hjubah}(h]h ]h"]h$]h&]uh1jhjubj)}(h7Device firmware microcode executed out of main memory. h]h)}(h6Device firmware microcode executed out of main memory.h]h6Device firmware microcode executed out of main memory.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM>hjubah}(h]h ]h"]h$]h&]uh1jhjubeh}(h]h ]h"]h$]h&]bullet-uh1jdhhhM<hjubah}(h]h ]h"]h$]h&]uh1jhhhM<hjiubh)}(hThe invariant these examples all require is that any CPU store to memory is immediately visible to the device, and vice versa. Consistent mappings guarantee this.h]hThe invariant these examples all require is that any CPU store to memory is immediately visible to the device, and vice versa. Consistent mappings guarantee this.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMAhjiubh important)}(hXkConsistent DMA memory does not preclude the usage of proper memory barriers. The CPU may reorder stores to consistent memory just as it may normal memory. Example: if it is important for the device to see the first word of a descriptor updated before the second, you must do something like:: desc->word0 = address; wmb(); desc->word1 = DESC_VALID; in order to get correct behavior on all platforms. Also, on some platforms your driver may need to flush CPU write buffers in much the same way as it needs to flush write buffers found in PCI bridges (such as by reading a register's value after writing it).h](h)}(hX%Consistent DMA memory does not preclude the usage of proper memory barriers. The CPU may reorder stores to consistent memory just as it may normal memory. Example: if it is important for the device to see the first word of a descriptor updated before the second, you must do something like::h]hX$Consistent DMA memory does not preclude the usage of proper memory barriers. The CPU may reorder stores to consistent memory just as it may normal memory. Example: if it is important for the device to see the first word of a descriptor updated before the second, you must do something like:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMGhjubj)}(h7desc->word0 = address; wmb(); desc->word1 = DESC_VALID;h]h7desc->word0 = address; wmb(); desc->word1 = DESC_VALID;}hj"sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMNhjubh)}(h2in order to get correct behavior on all platforms.h]h2in order to get correct behavior on all platforms.}(hj0hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMRhjubh)}(hAlso, on some platforms your driver may need to flush CPU write buffers in much the same way as it needs to flush write buffers found in PCI bridges (such as by reading a register's value after writing it).h]hAlso, on some platforms your driver may need to flush CPU write buffers in much the same way as it needs to flush write buffers found in PCI bridges (such as by reading a register’s value after writing it).}(hj>hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMThjubeh}(h]h ]h"]h$]h&]uh1jhjiubeh}(h]h ]h"]h$]h&]uh1jhjfhhhNhNubj)}(hXStreaming DMA mappings which are usually mapped for one DMA transfer, unmapped right after it (unless you use dma_sync_* below) and for which hardware can optimize for sequential accesses. Think of "streaming" as "asynchronous" or "outside the coherency domain". Good examples of what to use streaming mappings for are: - Networking buffers transmitted/received by a device. - Filesystem buffers written/read by a SCSI device. The interfaces for using this type of mapping were designed in such a way that an implementation can make whatever performance optimizations the hardware allows. To this end, when using such mappings you must be explicit about what you want to happen. h](h)}(hStreaming DMA mappings which are usually mapped for one DMA transfer, unmapped right after it (unless you use dma_sync_* below) and for which hardware can optimize for sequential accesses.h]hStreaming DMA mappings which are usually mapped for one DMA transfer, unmapped right after it (unless you use dma_sync_* below) and for which hardware can optimize for sequential accesses.}(hj\hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMYhjXubh)}(hIThink of "streaming" as "asynchronous" or "outside the coherency domain".h]hUThink of “streaming” as “asynchronous” or “outside the coherency domain”.}(hjjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM]hjXubh)}(h8Good examples of what to use streaming mappings for are:h]h8Good examples of what to use streaming mappings for are:}(hjxhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM`hjXubj)}(hk- Networking buffers transmitted/received by a device. - Filesystem buffers written/read by a SCSI device. h]je)}(hhh](j)}(h4Networking buffers transmitted/received by a device.h]h)}(hjh]h4Networking buffers transmitted/received by a device.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMbhjubah}(h]h ]h"]h$]h&]uh1jhjubj)}(h2Filesystem buffers written/read by a SCSI device. h]h)}(h1Filesystem buffers written/read by a SCSI device.h]h1Filesystem buffers written/read by a SCSI device.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMchjubah}(h]h ]h"]h$]h&]uh1jhjubeh}(h]h ]h"]h$]h&]jjuh1jdhhhMbhjubah}(h]h ]h"]h$]h&]uh1jhhhMbhjXubh)}(hThe interfaces for using this type of mapping were designed in such a way that an implementation can make whatever performance optimizations the hardware allows. To this end, when using such mappings you must be explicit about what you want to happen.h]hThe interfaces for using this type of mapping were designed in such a way that an implementation can make whatever performance optimizations the hardware allows. To this end, when using such mappings you must be explicit about what you want to happen.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMehjXubeh}(h]h ]h"]h$]h&]uh1jhjfhhhhhNubeh}(h]h ]h"]h$]h&]jjuh1jdhhhM-hjEhhubh)}(hXNeither type of DMA mapping has alignment restrictions that come from the underlying bus, although some devices may have such restrictions. Also, systems with caches that aren't DMA-coherent will work better when the underlying buffers don't share cache lines with other data.h]hXNeither type of DMA mapping has alignment restrictions that come from the underlying bus, although some devices may have such restrictions. Also, systems with caches that aren’t DMA-coherent will work better when the underlying buffers don’t share cache lines with other data.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMjhjEhhubeh}(h]types-of-dma-mappingsah ]h"]types of dma mappingsah$]h&]uh1hhhhhhhhM)ubh)}(hhh](h)}(hUsing Consistent DMA mappingsh]hUsing Consistent DMA mappings}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhjhhhhhMqubh)}(hSTo allocate and map large (PAGE_SIZE or so) consistent DMA regions, you should do::h]hRTo allocate and map large (PAGE_SIZE or so) consistent DMA regions, you should do:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMshjhhubj)}(hSdma_addr_t dma_handle; cpu_addr = dma_alloc_coherent(dev, size, &dma_handle, gfp);h]hSdma_addr_t dma_handle; cpu_addr = dma_alloc_coherent(dev, size, &dma_handle, gfp);}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMvhjhhubh)}(hhwhere device is a ``struct device *``. This may be called in interrupt context with the GFP_ATOMIC flag.h](hwhere device is a }(hj%hhhNhNubj)}(h``struct device *``h]hstruct device *}(hj-hhhNhNubah}(h]h ]h"]h$]h&]uh1jhj%ubhC. This may be called in interrupt context with the GFP_ATOMIC flag.}(hj%hhhNhNubeh}(h]h ]h"]h$]h&]uh1hhhhMzhjhhubh)}(h@Size is the length of the region you want to allocate, in bytes.h]h@Size is the length of the region you want to allocate, in bytes.}(hjEhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM}hjhhubh)}(hThis routine will allocate RAM for that region, so it acts similarly to __get_free_pages() (but takes size instead of a page order). If your driver needs regions sized smaller than a page, you may prefer using the dma_pool interface, described below.h]hThis routine will allocate RAM for that region, so it acts similarly to __get_free_pages() (but takes size instead of a page order). If your driver needs regions sized smaller than a page, you may prefer using the dma_pool interface, described below.}(hjShhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(hXThe consistent DMA mapping interfaces, will by default return a DMA address which is 32-bit addressable. Even if the device indicates (via the DMA mask) that it may address the upper 32-bits, consistent allocation will only return > 32-bit addresses for DMA if the consistent DMA mask has been explicitly changed via dma_set_coherent_mask(). This is true of the dma_pool interface as well.h]hXThe consistent DMA mapping interfaces, will by default return a DMA address which is 32-bit addressable. Even if the device indicates (via the DMA mask) that it may address the upper 32-bits, consistent allocation will only return > 32-bit addresses for DMA if the consistent DMA mask has been explicitly changed via dma_set_coherent_mask(). This is true of the dma_pool interface as well.}(hjahhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(hdma_alloc_coherent() returns two values: the virtual address which you can use to access it from the CPU and dma_handle which you pass to the card.h]hdma_alloc_coherent() returns two values: the virtual address which you can use to access it from the CPU and dma_handle which you pass to the card.}(hjohhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(hXfThe CPU virtual address and the DMA address are both guaranteed to be aligned to the smallest PAGE_SIZE order which is greater than or equal to the requested size. This invariant exists (for example) to guarantee that if you allocate a chunk which is smaller than or equal to 64 kilobytes, the extent of the buffer you receive will not cross a 64K boundary.h]hXfThe CPU virtual address and the DMA address are both guaranteed to be aligned to the smallest PAGE_SIZE order which is greater than or equal to the requested size. This invariant exists (for example) to guarantee that if you allocate a chunk which is smaller than or equal to 64 kilobytes, the extent of the buffer you receive will not cross a 64K boundary.}(hj}hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(h/To unmap and free such a DMA region, you call::h]h.To unmap and free such a DMA region, you call:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubj)}(h3dma_free_coherent(dev, size, cpu_addr, dma_handle);h]h3dma_free_coherent(dev, size, cpu_addr, dma_handle);}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhjhhubh)}(hwhere dev, size are the same as in the above call and cpu_addr and dma_handle are the values dma_alloc_coherent() returned to you. This function may not be called in interrupt context.h]hwhere dev, size are the same as in the above call and cpu_addr and dma_handle are the values dma_alloc_coherent() returned to you. This function may not be called in interrupt context.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(hXIf your driver needs lots of smaller memory regions, you can write custom code to subdivide pages returned by dma_alloc_coherent(), or you can use the dma_pool API to do that. A dma_pool is like a kmem_cache, but it uses dma_alloc_coherent(), not __get_free_pages(). Also, it understands common hardware constraints for alignment, like queue heads needing to be aligned on N byte boundaries.h]hXIf your driver needs lots of smaller memory regions, you can write custom code to subdivide pages returned by dma_alloc_coherent(), or you can use the dma_pool API to do that. A dma_pool is like a kmem_cache, but it uses dma_alloc_coherent(), not __get_free_pages(). Also, it understands common hardware constraints for alignment, like queue heads needing to be aligned on N byte boundaries.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(hCreate a dma_pool like this::h]hCreate a dma_pool like this:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubj)}(hQstruct dma_pool *pool; pool = dma_pool_create(name, dev, size, align, boundary);h]hQstruct dma_pool *pool; pool = dma_pool_create(name, dev, size, align, boundary);}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhjhhubh)}(hXThe "name" is for diagnostics (like a kmem_cache name); dev and size are as above. The device's hardware alignment requirement for this type of data is "align" (which is expressed in bytes, and must be a power of two). If your device has no boundary crossing restrictions, pass 0 for boundary; passing 4096 says memory allocated from this pool must not cross 4KByte boundaries (but at that time it may be better to use dma_alloc_coherent() directly instead).h]hXThe “name” is for diagnostics (like a kmem_cache name); dev and size are as above. The device’s hardware alignment requirement for this type of data is “align” (which is expressed in bytes, and must be a power of two). If your device has no boundary crossing restrictions, pass 0 for boundary; passing 4096 says memory allocated from this pool must not cross 4KByte boundaries (but at that time it may be better to use dma_alloc_coherent() directly instead).}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(h+Allocate memory from a DMA pool like this::h]h*Allocate memory from a DMA pool like this:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubj)}(h4cpu_addr = dma_pool_alloc(pool, flags, &dma_handle);h]h4cpu_addr = dma_pool_alloc(pool, flags, &dma_handle);}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhjhhubh)}(hflags are GFP_KERNEL if blocking is permitted (not in_interrupt nor holding SMP locks), GFP_ATOMIC otherwise. Like dma_alloc_coherent(), this returns two values, cpu_addr and dma_handle.h]hflags are GFP_KERNEL if blocking is permitted (not in_interrupt nor holding SMP locks), GFP_ATOMIC otherwise. Like dma_alloc_coherent(), this returns two values, cpu_addr and dma_handle.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(h:Free memory that was allocated from a dma_pool like this::h]h9Free memory that was allocated from a dma_pool like this:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubj)}(h*dma_pool_free(pool, cpu_addr, dma_handle);h]h*dma_pool_free(pool, cpu_addr, dma_handle);}hj%sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhjhhubh)}(hwhere pool is what you passed to dma_pool_alloc(), and cpu_addr and dma_handle are the values dma_pool_alloc() returned. This function may be called in interrupt context.h]hwhere pool is what you passed to dma_pool_alloc(), and cpu_addr and dma_handle are the values dma_pool_alloc() returned. This function may be called in interrupt context.}(hj3hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubh)}(hDestroy a dma_pool by calling::h]hDestroy a dma_pool by calling:}(hjAhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubj)}(hdma_pool_destroy(pool);h]hdma_pool_destroy(pool);}hjOsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhjhhubh)}(hMake sure you've called dma_pool_free() for all memory allocated from a pool before you destroy the pool. This function may not be called in interrupt context.h]hMake sure you’ve called dma_pool_free() for all memory allocated from a pool before you destroy the pool. This function may not be called in interrupt context.}(hj]hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjhhubeh}(h]using-consistent-dma-mappingsah ]h"]using consistent dma mappingsah$]h&]uh1hhhhhhhhMqubh)}(hhh](h)}(h DMA Directionh]h DMA Direction}(hjvhhhNhNubah}(h]h ]h"]h$]h&]uh1hhjshhhhhMubh)}(hThe interfaces described in subsequent portions of this document take a DMA direction argument, which is an integer and takes on one of the following values::h]hThe interfaces described in subsequent portions of this document take a DMA direction argument, which is an integer and takes on one of the following values:}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubj)}(h8DMA_BIDIRECTIONAL DMA_TO_DEVICE DMA_FROM_DEVICE DMA_NONEh]h8DMA_BIDIRECTIONAL DMA_TO_DEVICE DMA_FROM_DEVICE DMA_NONE}hjsbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhjshhubh)}(h:You should provide the exact DMA direction if you know it.h]h:You should provide the exact DMA direction if you know it.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hDMA_TO_DEVICE means "from main memory to the device" DMA_FROM_DEVICE means "from the device to main memory" It is the direction in which the data moves during the DMA transfer.h]hDMA_TO_DEVICE means “from main memory to the device” DMA_FROM_DEVICE means “from the device to main memory” It is the direction in which the data moves during the DMA transfer.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hOYou are _strongly_ encouraged to specify this as precisely as you possibly can.h]hOYou are _strongly_ encouraged to specify this as precisely as you possibly can.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hXIf you absolutely cannot know the direction of the DMA transfer, specify DMA_BIDIRECTIONAL. It means that the DMA can go in either direction. The platform guarantees that you may legally specify this, and that it will work, but this may be at the cost of performance for example.h]hXIf you absolutely cannot know the direction of the DMA transfer, specify DMA_BIDIRECTIONAL. It means that the DMA can go in either direction. The platform guarantees that you may legally specify this, and that it will work, but this may be at the cost of performance for example.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hThe value DMA_NONE is to be used for debugging. One can hold this in a data structure before you come to know the precise direction, and this will help catch cases where your direction tracking logic has failed to set things up properly.h]hThe value DMA_NONE is to be used for debugging. One can hold this in a data structure before you come to know the precise direction, and this will help catch cases where your direction tracking logic has failed to set things up properly.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hXAnother advantage of specifying this value precisely (outside of potential platform-specific optimizations of such) is for debugging. Some platforms actually have a write permission boolean which DMA mappings can be marked with, much like page protections in the user program address space. Such platforms can and do report errors in the kernel logs when the DMA controller hardware detects violation of the permission setting.h]hXAnother advantage of specifying this value precisely (outside of potential platform-specific optimizations of such) is for debugging. Some platforms actually have a write permission boolean which DMA mappings can be marked with, much like page protections in the user program address space. Such platforms can and do report errors in the kernel logs when the DMA controller hardware detects violation of the permission setting.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hOnly streaming mappings specify a direction, consistent mappings implicitly have a direction attribute setting of DMA_BIDIRECTIONAL.h]hOnly streaming mappings specify a direction, consistent mappings implicitly have a direction attribute setting of DMA_BIDIRECTIONAL.}(hjhhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hThe SCSI subsystem tells you the direction to use in the 'sc_data_direction' member of the SCSI command your driver is working on.h]hThe SCSI subsystem tells you the direction to use in the ‘sc_data_direction’ member of the SCSI command your driver is working on.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubh)}(hFor Networking drivers, it's a rather simple affair. For transmit packets, map/unmap them with the DMA_TO_DEVICE direction specifier. For receive packets, just the opposite, map/unmap them with the DMA_FROM_DEVICE direction specifier.h]hFor Networking drivers, it’s a rather simple affair. For transmit packets, map/unmap them with the DMA_TO_DEVICE direction specifier. For receive packets, just the opposite, map/unmap them with the DMA_FROM_DEVICE direction specifier.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjshhubeh}(h] dma-directionah ]h"] dma directionah$]h&]uh1hhhhhhhhMubh)}(hhh](h)}(hUsing Streaming DMA mappingsh]hUsing Streaming DMA mappings}(hj) hhhNhNubah}(h]h ]h"]h$]h&]uh1hhj& hhhhhMubh)}(hThe streaming DMA mapping routines can be called from interrupt context. There are two versions of each map/unmap, one which will map/unmap a single memory region, and one which will map/unmap a scatterlist.h]hThe streaming DMA mapping routines can be called from interrupt context. There are two versions of each map/unmap, one which will map/unmap a single memory region, and one which will map/unmap a scatterlist.}(hj7 hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubh)}(h To map a single region, you do::h]hTo map a single region, you do:}(hjE hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM hj& hhubj)}(hXzstruct device *dev = &my_dev->dev; dma_addr_t dma_handle; void *addr = buffer->ptr; size_t size = buffer->len; dma_handle = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; }h]hXzstruct device *dev = &my_dev->dev; dma_addr_t dma_handle; void *addr = buffer->ptr; size_t size = buffer->len; dma_handle = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; }}hjS sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhM hj& hhubh)}(hand to unmap it::h]hand to unmap it:}(hja hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubj)}(h3dma_unmap_single(dev, dma_handle, size, direction);h]h3dma_unmap_single(dev, dma_handle, size, direction);}hjo sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj& hhubh)}(hXYou should call dma_mapping_error() as dma_map_single() could fail and return error. Doing so will ensure that the mapping code will work correctly on all DMA implementations without any dependency on the specifics of the underlying implementation. Using the returned address without checking for errors could result in failures ranging from panics to silent data corruption. The same applies to dma_map_page() as well.h]hXYou should call dma_mapping_error() as dma_map_single() could fail and return error. Doing so will ensure that the mapping code will work correctly on all DMA implementations without any dependency on the specifics of the underlying implementation. Using the returned address without checking for errors could result in failures ranging from panics to silent data corruption. The same applies to dma_map_page() as well.S}(hj} hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubh)}(hYou should call dma_unmap_single() when the DMA activity is finished, e.g., from the interrupt which told you that the DMA transfer is done.h]hYou should call dma_unmap_single() when the DMA activity is finished, e.g., from the interrupt which told you that the DMA transfer is done.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM%hj& hhubh)}(hXUsing CPU pointers like this for single mappings has a disadvantage: you cannot reference HIGHMEM memory in this way. Thus, there is a map/unmap interface pair akin to dma_{map,unmap}_single(). These interfaces deal with page/offset pairs instead of CPU pointers. Specifically::h]hXUsing CPU pointers like this for single mappings has a disadvantage: you cannot reference HIGHMEM memory in this way. Thus, there is a map/unmap interface pair akin to dma_{map,unmap}_single(). These interfaces deal with page/offset pairs instead of CPU pointers. Specifically:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM(hj& hhubj)}(hXstruct device *dev = &my_dev->dev; dma_addr_t dma_handle; struct page *page = buffer->page; unsigned long offset = buffer->offset; size_t size = buffer->len; dma_handle = dma_map_page(dev, page, offset, size, direction); if (dma_mapping_error(dev, dma_handle)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; } ... dma_unmap_page(dev, dma_handle, size, direction);h]hXstruct device *dev = &my_dev->dev; dma_addr_t dma_handle; struct page *page = buffer->page; unsigned long offset = buffer->offset; size_t size = buffer->len; dma_handle = dma_map_page(dev, page, offset, size, direction); if (dma_mapping_error(dev, dma_handle)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; } ... dma_unmap_page(dev, dma_handle, size, direction);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhM.hj& hhubh)}(h7Here, "offset" means byte offset within the given page.h]h;Here, “offset” means byte offset within the given page.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMBhj& hhubh)}(hYou should call dma_mapping_error() as dma_map_page() could fail and return error as outlined under the dma_map_single() discussion.h]hYou should call dma_mapping_error() as dma_map_page() could fail and return error as outlined under the dma_map_single() discussion.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMDhj& hhubh)}(hYou should call dma_unmap_page() when the DMA activity is finished, e.g., from the interrupt which told you that the DMA transfer is done.h]hYou should call dma_unmap_page() when the DMA activity is finished, e.g., from the interrupt which told you that the DMA transfer is done.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMGhj& hhubh)}(hFWith scatterlists, you map a region gathered from several regions by::h]hEWith scatterlists, you map a region gathered from several regions by:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMJhj& hhubj)}(hint i, count = dma_map_sg(dev, sglist, nents, direction); struct scatterlist *sg; for_each_sg(sglist, sg, count, i) { hw_address[i] = sg_dma_address(sg); hw_len[i] = sg_dma_len(sg); }h]hint i, count = dma_map_sg(dev, sglist, nents, direction); struct scatterlist *sg; for_each_sg(sglist, sg, count, i) { hw_address[i] = sg_dma_address(sg); hw_len[i] = sg_dma_len(sg); }}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMLhj& hhubh)}(h3where nents is the number of entries in the sglist.h]h3where nents is the number of entries in the sglist.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMThj& hhubh)}(hXThe implementation is free to merge several consecutive sglist entries into one (e.g. if DMA mapping is done with PAGE_SIZE granularity, any consecutive sglist entries can be merged into one provided the first one ends and the second one starts on a page boundary - in fact this is a huge advantage for cards which either cannot do scatter-gather or have very limited number of scatter-gather entries) and returns the actual number of sg entries it mapped them to. On failure 0 is returned.h]hXThe implementation is free to merge several consecutive sglist entries into one (e.g. if DMA mapping is done with PAGE_SIZE granularity, any consecutive sglist entries can be merged into one provided the first one ends and the second one starts on a page boundary - in fact this is a huge advantage for cards which either cannot do scatter-gather or have very limited number of scatter-gather entries) and returns the actual number of sg entries it mapped them to. On failure 0 is returned.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMVhj& hhubh)}(hThen you should loop count times (note: this can be less than nents times) and use sg_dma_address() and sg_dma_len() macros where you previously accessed sg->address and sg->length as shown above.h]hThen you should loop count times (note: this can be less than nents times) and use sg_dma_address() and sg_dma_len() macros where you previously accessed sg->address and sg->length as shown above.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM^hj& hhubh)}(h#To unmap a scatterlist, just call::h]h"To unmap a scatterlist, just call:}(hj% hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMbhj& hhubj)}(h,dma_unmap_sg(dev, sglist, nents, direction);h]h,dma_unmap_sg(dev, sglist, nents, direction);}hj3 sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMdhj& hhubh)}(h3Again, make sure DMA activity has already finished.h]h3Again, make sure DMA activity has already finished.}(hjA hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMfhj& hhubhnote)}(hThe 'nents' argument to the dma_unmap_sg call must be the _same_ one you passed into the dma_map_sg call, it should _NOT_ be the 'count' value _returned_ from the dma_map_sg call.h]h)}(hThe 'nents' argument to the dma_unmap_sg call must be the _same_ one you passed into the dma_map_sg call, it should _NOT_ be the 'count' value _returned_ from the dma_map_sg call.h]hThe ‘nents’ argument to the dma_unmap_sg call must be the _same_ one you passed into the dma_map_sg call, it should _NOT_ be the ‘count’ value _returned_ from the dma_map_sg call.}(hjU hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMjhjQ ubah}(h]h ]h"]h$]h&]uh1jO hj& hhhhhNubh)}(hEvery dma_map_{single,sg}() call should have its dma_unmap_{single,sg}() counterpart, because the DMA address space is a shared resource and you could render the machine unusable by consuming all DMA addresses.h]hEvery dma_map_{single,sg}() call should have its dma_unmap_{single,sg}() counterpart, because the DMA address space is a shared resource and you could render the machine unusable by consuming all DMA addresses.}(hji hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMohj& hhubh)}(hIf you need to use the same streaming DMA region multiple times and touch the data in between the DMA transfers, the buffer needs to be synced properly in order for the CPU and device to see the most up-to-date and correct copy of the DMA buffer.h]hIf you need to use the same streaming DMA region multiple times and touch the data in between the DMA transfers, the buffer needs to be synced properly in order for the CPU and device to see the most up-to-date and correct copy of the DMA buffer.}(hjw hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMshj& hhubh)}(h^So, firstly, just map it with dma_map_{single,sg}(), and after each DMA transfer call either::h]h]So, firstly, just map it with dma_map_{single,sg}(), and after each DMA transfer call either:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMxhj& hhubj)}(h:dma_sync_single_for_cpu(dev, dma_handle, size, direction);h]h:dma_sync_single_for_cpu(dev, dma_handle, size, direction);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhM{hj& hhubh)}(hor::h]hor:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM}hj& hhubj)}(h3dma_sync_sg_for_cpu(dev, sglist, nents, direction);h]h3dma_sync_sg_for_cpu(dev, sglist, nents, direction);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj& hhubh)}(has appropriate.h]has appropriate.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubh)}(hThen, if you wish to let the device get at the DMA area again, finish accessing the data with the CPU, and then before actually giving the buffer to the hardware call either::h]hThen, if you wish to let the device get at the DMA area again, finish accessing the data with the CPU, and then before actually giving the buffer to the hardware call either:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubj)}(h=dma_sync_single_for_device(dev, dma_handle, size, direction);h]h=dma_sync_single_for_device(dev, dma_handle, size, direction);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj& hhubh)}(hor::h]hor:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubj)}(h6dma_sync_sg_for_device(dev, sglist, nents, direction);h]h6dma_sync_sg_for_device(dev, sglist, nents, direction);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj& hhubh)}(has appropriate.h]has appropriate.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubjP )}(hThe 'nents' argument to dma_sync_sg_for_cpu() and dma_sync_sg_for_device() must be the same passed to dma_map_sg(). It is _NOT_ the count returned by dma_map_sg().h]h)}(hThe 'nents' argument to dma_sync_sg_for_cpu() and dma_sync_sg_for_device() must be the same passed to dma_map_sg(). It is _NOT_ the count returned by dma_map_sg().h]hThe ‘nents’ argument to dma_sync_sg_for_cpu() and dma_sync_sg_for_device() must be the same passed to dma_map_sg(). It is _NOT_ the count returned by dma_map_sg().}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj ubah}(h]h ]h"]h$]h&]uh1jO hj& hhhhhNubh)}(hAfter the last DMA transfer call one of the DMA unmap routines dma_unmap_{single,sg}(). If you don't touch the data from the first dma_map_*() call till dma_unmap_*(), then you don't have to call the dma_sync_*() routines at all.h]hAfter the last DMA transfer call one of the DMA unmap routines dma_unmap_{single,sg}(). If you don’t touch the data from the first dma_map_*() call till dma_unmap_*(), then you don’t have to call the dma_sync_*() routines at all.}(hj) hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubh)}(hhHere is pseudo code which shows a situation in which you would need to use the dma_sync_*() interfaces::h]hgHere is pseudo code which shows a situation in which you would need to use the dma_sync_*() interfaces:}(hj7 hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj& hhubj)}(hX1my_card_setup_receive_buffer(struct my_card *cp, char *buffer, int len) { dma_addr_t mapping; mapping = dma_map_single(cp->dev, buffer, len, DMA_FROM_DEVICE); if (dma_mapping_error(cp->dev, mapping)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; } cp->rx_buf = buffer; cp->rx_len = len; cp->rx_dma = mapping; give_rx_buf_to_card(cp); } ... my_card_interrupt_handler(int irq, void *devid, struct pt_regs *regs) { struct my_card *cp = devid; ... if (read_card_status(cp) == RX_BUF_TRANSFERRED) { struct my_card_header *hp; /* Examine the header to see if we wish * to accept the data. But synchronize * the DMA transfer with the CPU first * so that we see updated contents. */ dma_sync_single_for_cpu(&cp->dev, cp->rx_dma, cp->rx_len, DMA_FROM_DEVICE); /* Now it is safe to examine the buffer. */ hp = (struct my_card_header *) cp->rx_buf; if (header_is_ok(hp)) { dma_unmap_single(&cp->dev, cp->rx_dma, cp->rx_len, DMA_FROM_DEVICE); pass_to_upper_layers(cp->rx_buf); make_and_setup_new_rx_buf(cp); } else { /* CPU should not write to * DMA_FROM_DEVICE-mapped area, * so dma_sync_single_for_device() is * not needed here. It would be required * for DMA_BIDIRECTIONAL mapping if * the memory was modified. */ give_rx_buf_to_card(cp); } } }h]hX1my_card_setup_receive_buffer(struct my_card *cp, char *buffer, int len) { dma_addr_t mapping; mapping = dma_map_single(cp->dev, buffer, len, DMA_FROM_DEVICE); if (dma_mapping_error(cp->dev, mapping)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; } cp->rx_buf = buffer; cp->rx_len = len; cp->rx_dma = mapping; give_rx_buf_to_card(cp); } ... my_card_interrupt_handler(int irq, void *devid, struct pt_regs *regs) { struct my_card *cp = devid; ... if (read_card_status(cp) == RX_BUF_TRANSFERRED) { struct my_card_header *hp; /* Examine the header to see if we wish * to accept the data. But synchronize * the DMA transfer with the CPU first * so that we see updated contents. */ dma_sync_single_for_cpu(&cp->dev, cp->rx_dma, cp->rx_len, DMA_FROM_DEVICE); /* Now it is safe to examine the buffer. */ hp = (struct my_card_header *) cp->rx_buf; if (header_is_ok(hp)) { dma_unmap_single(&cp->dev, cp->rx_dma, cp->rx_len, DMA_FROM_DEVICE); pass_to_upper_layers(cp->rx_buf); make_and_setup_new_rx_buf(cp); } else { /* CPU should not write to * DMA_FROM_DEVICE-mapped area, * so dma_sync_single_for_device() is * not needed here. It would be required * for DMA_BIDIRECTIONAL mapping if * the memory was modified. */ give_rx_buf_to_card(cp); } } }}hjE sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj& hhubeh}(h]using-streaming-dma-mappingsah ]h"]using streaming dma mappingsah$]h&]uh1hhhhhhhhMubh)}(hhh](h)}(hHandling Errorsh]hHandling Errors}(hj^ hhhNhNubah}(h]h ]h"]h$]h&]uh1hhj[ hhhhhMubh)}(hbDMA address space is limited on some architectures and an allocation failure can be determined by:h]hbDMA address space is limited on some architectures and an allocation failure can be determined by:}(hjl hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj[ hhubje)}(hhh](j)}(hFchecking if dma_alloc_coherent() returns NULL or dma_map_sg returns 0 h]h)}(hEchecking if dma_alloc_coherent() returns NULL or dma_map_sg returns 0h]hEchecking if dma_alloc_coherent() returns NULL or dma_map_sg returns 0}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj} ubah}(h]h ]h"]h$]h&]uh1jhjz hhhhhNubj)}(hXchecking the dma_addr_t returned from dma_map_single() and dma_map_page() by using dma_mapping_error():: dma_addr_t dma_handle; dma_handle = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; } h](h)}(hhchecking the dma_addr_t returned from dma_map_single() and dma_map_page() by using dma_mapping_error()::h]hgchecking the dma_addr_t returned from dma_map_single() and dma_map_page() by using dma_mapping_error():}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj ubj)}(hX"dma_addr_t dma_handle; dma_handle = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; }h]hX"dma_addr_t dma_handle; dma_handle = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; }}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj ubeh}(h]h ]h"]h$]h&]uh1jhjz hhhhhNubj)}(hunmap pages that are already mapped, when mapping error occurs in the middle of a multiple page mapping attempt. These example are applicable to dma_map_page() as well. h]h)}(hunmap pages that are already mapped, when mapping error occurs in the middle of a multiple page mapping attempt. These example are applicable to dma_map_page() as well.h]hunmap pages that are already mapped, when mapping error occurs in the middle of a multiple page mapping attempt. These example are applicable to dma_map_page() as well.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj ubah}(h]h ]h"]h$]h&]uh1jhjz hhhhhNubeh}(h]h ]h"]h$]h&]jjuh1jdhhhMhj[ hhubh)}(h Example 1::h]h Example 1:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj[ hhubj)}(hXdma_addr_t dma_handle1; dma_addr_t dma_handle2; dma_handle1 = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle1)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling1; } dma_handle2 = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle2)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling2; } ... map_error_handling2: dma_unmap_single(dma_handle1); map_error_handling1:h]hXdma_addr_t dma_handle1; dma_addr_t dma_handle2; dma_handle1 = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle1)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling1; } dma_handle2 = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_handle2)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling2; } ... map_error_handling2: dma_unmap_single(dma_handle1); map_error_handling1:}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj[ hhubh)}(h Example 2::h]h Example 2:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj[ hhubj)}(hX/* * if buffers are allocated in a loop, unmap all mapped buffers when * mapping error is detected in the middle */ dma_addr_t dma_addr; dma_addr_t array[DMA_BUFFERS]; int save_index = 0; for (i = 0; i < DMA_BUFFERS; i++) { ... dma_addr = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_addr)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; } array[i].dma_addr = dma_addr; save_index++; } ... map_error_handling: for (i = 0; i < save_index; i++) { ... dma_unmap_single(array[i].dma_addr); }h]hX/* * if buffers are allocated in a loop, unmap all mapped buffers when * mapping error is detected in the middle */ dma_addr_t dma_addr; dma_addr_t array[DMA_BUFFERS]; int save_index = 0; for (i = 0; i < DMA_BUFFERS; i++) { ... dma_addr = dma_map_single(dev, addr, size, direction); if (dma_mapping_error(dev, dma_addr)) { /* * reduce current DMA mapping usage, * delay and try again later or * reset driver. */ goto map_error_handling; } array[i].dma_addr = dma_addr; save_index++; } ... map_error_handling: for (i = 0; i < save_index; i++) { ... dma_unmap_single(array[i].dma_addr); }}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj[ hhubh)}(hNetworking drivers must call dev_kfree_skb() to free the socket buffer and return NETDEV_TX_OK if the DMA mapping fails on the transmit hook (ndo_start_xmit). This means that the socket buffer is just dropped in the failure case.h]hNetworking drivers must call dev_kfree_skb() to free the socket buffer and return NETDEV_TX_OK if the DMA mapping fails on the transmit hook (ndo_start_xmit). This means that the socket buffer is just dropped in the failure case.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM:hj[ hhubh)}(hSCSI drivers must return SCSI_MLQUEUE_HOST_BUSY if the DMA mapping fails in the queuecommand hook. This means that the SCSI subsystem passes the command to the driver again later.h]hSCSI drivers must return SCSI_MLQUEUE_HOST_BUSY if the DMA mapping fails in the queuecommand hook. This means that the SCSI subsystem passes the command to the driver again later.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM?hj[ hhubeh}(h]handling-errorsah ]h"]handling errorsah$]h&]uh1hhhhhhhhMubh)}(hhh](h)}(h(Optimizing Unmap State Space Consumptionh]h(Optimizing Unmap State Space Consumption}(hj8 hhhNhNubah}(h]h ]h"]h$]h&]uh1hhj5 hhhhhMDubh)}(hX@On many platforms, dma_unmap_{single,page}() is simply a nop. Therefore, keeping track of the mapping address and length is a waste of space. Instead of filling your drivers up with ifdefs and the like to "work around" this (which would defeat the whole purpose of a portable API) the following facilities are provided.h]hXDOn many platforms, dma_unmap_{single,page}() is simply a nop. Therefore, keeping track of the mapping address and length is a waste of space. Instead of filling your drivers up with ifdefs and the like to “work around” this (which would defeat the whole purpose of a portable API) the following facilities are provided.}(hjF hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMFhj5 hhubh)}(hYActually, instead of describing the macros one by one, we'll transform some example code.h]h[Actually, instead of describing the macros one by one, we’ll transform some example code.}(hjT hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMLhj5 hhubj)}(hhh](j)}(hXmUse DEFINE_DMA_UNMAP_{ADDR,LEN} in state saving structures. Example, before:: struct ring_state { struct sk_buff *skb; dma_addr_t mapping; __u32 len; }; after:: struct ring_state { struct sk_buff *skb; DEFINE_DMA_UNMAP_ADDR(mapping); DEFINE_DMA_UNMAP_LEN(len); }; h](h)}(hMUse DEFINE_DMA_UNMAP_{ADDR,LEN} in state saving structures. Example, before::h]hLUse DEFINE_DMA_UNMAP_{ADDR,LEN} in state saving structures. Example, before:}(hji hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMOhje ubj)}(hbstruct ring_state { struct sk_buff *skb; dma_addr_t mapping; __u32 len; };h]hbstruct ring_state { struct sk_buff *skb; dma_addr_t mapping; __u32 len; };}hjw sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMRhje ubh)}(hafter::h]hafter:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMXhje ubj)}(h~struct ring_state { struct sk_buff *skb; DEFINE_DMA_UNMAP_ADDR(mapping); DEFINE_DMA_UNMAP_LEN(len); };h]h~struct ring_state { struct sk_buff *skb; DEFINE_DMA_UNMAP_ADDR(mapping); DEFINE_DMA_UNMAP_LEN(len); };}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMZhje ubeh}(h]h ]h"]h$]h&]uh1jhjb hhhhhNubj)}(hUse dma_unmap_{addr,len}_set() to set these values. Example, before:: ringp->mapping = FOO; ringp->len = BAR; after:: dma_unmap_addr_set(ringp, mapping, FOO); dma_unmap_len_set(ringp, len, BAR); h](h)}(hEUse dma_unmap_{addr,len}_set() to set these values. Example, before::h]hDUse dma_unmap_{addr,len}_set() to set these values. Example, before:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhM`hj ubj)}(h'ringp->mapping = FOO; ringp->len = BAR;h]h'ringp->mapping = FOO; ringp->len = BAR;}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMchj ubh)}(hafter::h]hafter:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMfhj ubj)}(hLdma_unmap_addr_set(ringp, mapping, FOO); dma_unmap_len_set(ringp, len, BAR);h]hLdma_unmap_addr_set(ringp, mapping, FOO); dma_unmap_len_set(ringp, len, BAR);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhhj ubeh}(h]h ]h"]h$]h&]uh1jhjb hhhhhNubj)}(hXYUse dma_unmap_{addr,len}() to access these values. Example, before:: dma_unmap_single(dev, ringp->mapping, ringp->len, DMA_FROM_DEVICE); after:: dma_unmap_single(dev, dma_unmap_addr(ringp, mapping), dma_unmap_len(ringp, len), DMA_FROM_DEVICE); h](h)}(hDUse dma_unmap_{addr,len}() to access these values. Example, before::h]hCUse dma_unmap_{addr,len}() to access these values. Example, before:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMkhj ubj)}(hTdma_unmap_single(dev, ringp->mapping, ringp->len, DMA_FROM_DEVICE);h]hTdma_unmap_single(dev, ringp->mapping, ringp->len, DMA_FROM_DEVICE);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMnhj ubh)}(hafter::h]hafter:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMqhj ubj)}(hdma_unmap_single(dev, dma_unmap_addr(ringp, mapping), dma_unmap_len(ringp, len), DMA_FROM_DEVICE);h]hdma_unmap_single(dev, dma_unmap_addr(ringp, mapping), dma_unmap_len(ringp, len), DMA_FROM_DEVICE);}hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMshj ubeh}(h]h ]h"]h$]h&]uh1jhjb hhhhhNubeh}(h]h ]h"]h$]h&]j.j/j0hj1j2uh1jhj5 hhhhhMOubh)}(hIt really should be self-explanatory. We treat the ADDR and LEN separately, because it is possible for an implementation to only need the address in order to perform the unmap operation.h]hIt really should be self-explanatory. We treat the ADDR and LEN separately, because it is possible for an implementation to only need the address in order to perform the unmap operation.}(hj1 hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMxhj5 hhubeh}(h](optimizing-unmap-state-space-consumptionah ]h"](optimizing unmap state space consumptionah$]h&]uh1hhhhhhhhMDubh)}(hhh](h)}(hPlatform Issuesh]hPlatform Issues}(hjJ hhhNhNubah}(h]h ]h"]h$]h&]uh1hhjG hhhhhM}ubh)}(hIf you are just writing drivers for Linux and do not maintain an architecture port for the kernel, you can safely skip down to "Closing".h]hIf you are just writing drivers for Linux and do not maintain an architecture port for the kernel, you can safely skip down to “Closing”.}(hjX hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhjG hhubj)}(hhh](j)}(hStruct scatterlist requirements. You need to enable CONFIG_NEED_SG_DMA_LENGTH if the architecture supports IOMMUs (including software IOMMU). h](h)}(h Struct scatterlist requirements.h]h Struct scatterlist requirements.}(hjm hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhji ubh)}(hlYou need to enable CONFIG_NEED_SG_DMA_LENGTH if the architecture supports IOMMUs (including software IOMMU).h]hlYou need to enable CONFIG_NEED_SG_DMA_LENGTH if the architecture supports IOMMUs (including software IOMMU).}(hj{ hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhji ubeh}(h]h ]h"]h$]h&]uh1jhjf hhhhhNubj)}(hXARCH_DMA_MINALIGN Architectures must ensure that kmalloc'ed buffer is DMA-safe. Drivers and subsystems depend on it. If an architecture isn't fully DMA-coherent (i.e. hardware doesn't ensure that data in the CPU cache is identical to data in main memory), ARCH_DMA_MINALIGN must be set so that the memory allocator makes sure that kmalloc'ed buffer doesn't share a cache line with the others. See arch/arm/include/asm/cache.h as an example. Note that ARCH_DMA_MINALIGN is about DMA memory alignment constraints. You don't need to worry about the architecture data alignment constraints (e.g. the alignment constraints about 64-bit objects). h](h)}(hARCH_DMA_MINALIGNh]hARCH_DMA_MINALIGN}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj ubh)}(hXArchitectures must ensure that kmalloc'ed buffer is DMA-safe. Drivers and subsystems depend on it. If an architecture isn't fully DMA-coherent (i.e. hardware doesn't ensure that data in the CPU cache is identical to data in main memory), ARCH_DMA_MINALIGN must be set so that the memory allocator makes sure that kmalloc'ed buffer doesn't share a cache line with the others. See arch/arm/include/asm/cache.h as an example.h]hXArchitectures must ensure that kmalloc’ed buffer is DMA-safe. Drivers and subsystems depend on it. If an architecture isn’t fully DMA-coherent (i.e. hardware doesn’t ensure that data in the CPU cache is identical to data in main memory), ARCH_DMA_MINALIGN must be set so that the memory allocator makes sure that kmalloc’ed buffer doesn’t share a cache line with the others. See arch/arm/include/asm/cache.h as an example.}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj ubh)}(hNote that ARCH_DMA_MINALIGN is about DMA memory alignment constraints. You don't need to worry about the architecture data alignment constraints (e.g. the alignment constraints about 64-bit objects).h]hNote that ARCH_DMA_MINALIGN is about DMA memory alignment constraints. You don’t need to worry about the architecture data alignment constraints (e.g. the alignment constraints about 64-bit objects).}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj ubeh}(h]h ]h"]h$]h&]uh1jhjf hhhhhNubeh}(h]h ]h"]h$]h&]j.j/j0hj1j2uh1jhjG hhhhhMubeh}(h]platform-issuesah ]h"]platform issuesah$]h&]uh1hhhhhhhhM}ubh)}(hhh](h)}(hClosingh]hClosing}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhj hhhhhMubh)}(hThis document, and the API itself, would not be in its current form without the feedback and suggestions from numerous individuals. We would like to specifically mention, in no particular order, the following people::h]hThis document, and the API itself, would not be in its current form without the feedback and suggestions from numerous individuals. We would like to specifically mention, in no particular order, the following people:}(hj hhhNhNubah}(h]h ]h"]h$]h&]uh1hhhhMhj hhubj)}(hXKRussell King Leo Dagum Ralf Baechle Grant Grundler Jay Estabrook Thomas Sailer Andrea Arcangeli Jens Axboe David Mosberger-Tang h]hXKRussell King Leo Dagum Ralf Baechle Grant Grundler Jay Estabrook Thomas Sailer Andrea Arcangeli Jens Axboe David Mosberger-Tang }hj sbah}(h]h ]h"]h$]h&]j*j+uh1jhhhMhj hhubeh}(h]closingah ]h"]closingah$]h&]uh1hhhhhhhhMubeh}(h]dynamic-dma-mapping-guideah ]h"]dynamic dma mapping guideah$]h&]uh1hhhhhhhhKubeh}(h]h ]h"]h$]h&]sourcehuh1hcurrent_sourceN current_lineNsettingsdocutils.frontendValues)}(hN generatorN datestampN source_linkN source_urlN toc_backlinksentryfootnote_backlinksK sectnum_xformKstrip_commentsNstrip_elements_with_classesN strip_classesN report_levelK halt_levelKexit_status_levelKdebugNwarning_streamN tracebackinput_encoding utf-8-siginput_encoding_error_handlerstrictoutput_encodingutf-8output_encoding_error_handlerj1error_encodingutf-8error_encoding_error_handlerbackslashreplace language_codeenrecord_dependenciesNconfigN id_prefixhauto_id_prefixid dump_settingsNdump_internalsNdump_transformsNdump_pseudo_xmlNexpose_internalsNstrict_visitorN_disable_configN_sourceh _destinationN _config_files]7/var/lib/git/docbuild/linux/Documentation/docutils.confafile_insertion_enabled raw_enabledKline_length_limitM'pep_referencesN pep_base_urlhttps://peps.python.org/pep_file_url_templatepep-%04drfc_referencesN rfc_base_url&https://datatracker.ietf.org/doc/html/ tab_widthKtrim_footnote_reference_spacesyntax_highlightlong smart_quotessmartquotes_locales]character_level_inline_markupdoctitle_xform docinfo_xformKsectsubtitle_xform image_loadinglinkembed_stylesheetcloak_email_addressessection_self_linkenvNubreporterNindirect_targets]substitution_defs}substitution_names}refnames}refids}nameids}(j jjjjjjBj?jjjpjmj# j jX jU j2 j/ jD jA j j jju nametypes}(j jjjBjjpj# jX j2 jD j juh}(jhjjjjj?jjjEjmjj jsjU j& j/ j[ jA j5 j jG jj u footnote_refs} citation_refs} autofootnotes]autofootnote_refs]symbol_footnotes]symbol_footnote_refs] footnotes] citations]autofootnote_startKsymbol_footnote_startK id_counter collectionsCounter}Rparse_messages]transform_messages] transformerN include_log] decorationNhhub.