Commit | Line | Data |
---|---|---|
266921bd MCC |
1 | ========================= |
2 | Dynamic DMA mapping Guide | |
3 | ========================= | |
1da177e4 | 4 | |
266921bd MCC |
5 | :Author: David S. Miller <davem@redhat.com> |
6 | :Author: Richard Henderson <rth@cygnus.com> | |
7 | :Author: Jakub Jelinek <jakub@redhat.com> | |
1da177e4 | 8 | |
216bf58f FT |
9 | This is a guide to device driver writers on how to use the DMA API |
10 | with example pseudo-code. For a concise description of the API, see | |
1da177e4 LT |
11 | DMA-API.txt. |
12 | ||
266921bd MCC |
13 | CPU and DMA addresses |
14 | ===================== | |
77f2ea2f BH |
15 | |
16 | There are several kinds of addresses involved in the DMA API, and it's | |
17 | important to understand the differences. | |
18 | ||
19 | The kernel normally uses virtual addresses. Any address returned by | |
20 | kmalloc(), vmalloc(), and similar interfaces is a virtual address and can | |
266921bd | 21 | be stored in a ``void *``. |
77f2ea2f BH |
22 | |
23 | The virtual memory system (TLB, page tables, etc.) translates virtual | |
24 | addresses to CPU physical addresses, which are stored as "phys_addr_t" or | |
25 | "resource_size_t". The kernel manages device resources like registers as | |
26 | physical addresses. These are the addresses in /proc/iomem. The physical | |
27 | address is not directly useful to a driver; it must use ioremap() to map | |
28 | the space and produce a virtual address. | |
29 | ||
3a9ad0b4 YL |
30 | I/O devices use a third kind of address: a "bus address". If a device has |
31 | registers at an MMIO address, or if it performs DMA to read or write system | |
32 | memory, the addresses used by the device are bus addresses. In some | |
33 | systems, bus addresses are identical to CPU physical addresses, but in | |
34 | general they are not. IOMMUs and host bridges can produce arbitrary | |
77f2ea2f BH |
35 | mappings between physical and bus addresses. |
36 | ||
3a9ad0b4 YL |
37 | From a device's point of view, DMA uses the bus address space, but it may |
38 | be restricted to a subset of that space. For example, even if a system | |
39 | supports 64-bit addresses for main memory and PCI BARs, it may use an IOMMU | |
40 | so devices only need to use 32-bit DMA addresses. | |
41 | ||
266921bd | 42 | Here's a picture and some examples:: |
77f2ea2f BH |
43 | |
44 | CPU CPU Bus | |
45 | Virtual Physical Address | |
46 | Address Address Space | |
47 | Space Space | |
48 | ||
49 | +-------+ +------+ +------+ | |
50 | | | |MMIO | Offset | | | |
51 | | | Virtual |Space | applied | | | |
52 | C +-------+ --------> B +------+ ----------> +------+ A | |
53 | | | mapping | | by host | | | |
54 | +-----+ | | | | bridge | | +--------+ | |
55 | | | | | +------+ | | | | | |
56 | | CPU | | | | RAM | | | | Device | | |
57 | | | | | | | | | | | | |
58 | +-----+ +-------+ +------+ +------+ +--------+ | |
59 | | | Virtual |Buffer| Mapping | | | |
60 | X +-------+ --------> Y +------+ <---------- +------+ Z | |
61 | | | mapping | RAM | by IOMMU | |
62 | | | | | | |
63 | | | | | | |
64 | +-------+ +------+ | |
65 | ||
66 | During the enumeration process, the kernel learns about I/O devices and | |
67 | their MMIO space and the host bridges that connect them to the system. For | |
68 | example, if a PCI device has a BAR, the kernel reads the bus address (A) | |
69 | from the BAR and converts it to a CPU physical address (B). The address B | |
70 | is stored in a struct resource and usually exposed via /proc/iomem. When a | |
71 | driver claims a device, it typically uses ioremap() to map physical address | |
72 | B at a virtual address (C). It can then use, e.g., ioread32(C), to access | |
73 | the device registers at bus address A. | |
74 | ||
75 | If the device supports DMA, the driver sets up a buffer using kmalloc() or | |
76 | a similar interface, which returns a virtual address (X). The virtual | |
77 | memory system maps X to a physical address (Y) in system RAM. The driver | |
78 | can use virtual address X to access the buffer, but the device itself | |
79 | cannot because DMA doesn't go through the CPU virtual memory system. | |
80 | ||
81 | In some simple systems, the device can do DMA directly to physical address | |
3a9ad0b4 | 82 | Y. But in many others, there is IOMMU hardware that translates DMA |
77f2ea2f BH |
83 | addresses to physical addresses, e.g., it translates Z to Y. This is part |
84 | of the reason for the DMA API: the driver can give a virtual address X to | |
85 | an interface like dma_map_single(), which sets up any required IOMMU | |
3a9ad0b4 | 86 | mapping and returns the DMA address Z. The driver then tells the device to |
77f2ea2f BH |
87 | do DMA to Z, and the IOMMU maps it to the buffer at address Y in system |
88 | RAM. | |
1da177e4 LT |
89 | |
90 | So that Linux can use the dynamic DMA mapping, it needs some help from the | |
91 | drivers, namely it has to take into account that DMA addresses should be | |
92 | mapped only for the time they are actually used and unmapped after the DMA | |
93 | transfer. | |
94 | ||
95 | The following API will work of course even on platforms where no such | |
216bf58f FT |
96 | hardware exists. |
97 | ||
98 | Note that the DMA API works with any bus independent of the underlying | |
77f2ea2f BH |
99 | microprocessor architecture. You should use the DMA API rather than the |
100 | bus-specific DMA API, i.e., use the dma_map_*() interfaces rather than the | |
101 | pci_map_*() interfaces. | |
1da177e4 | 102 | |
266921bd | 103 | First of all, you should make sure:: |
1da177e4 | 104 | |
266921bd | 105 | #include <linux/dma-mapping.h> |
1da177e4 | 106 | |
77f2ea2f | 107 | is in your driver, which provides the definition of dma_addr_t. This type |
3a9ad0b4 | 108 | can hold any valid DMA address for the platform and should be used |
77f2ea2f | 109 | everywhere you hold a DMA address returned from the DMA mapping functions. |
1da177e4 | 110 | |
266921bd MCC |
111 | What memory is DMA'able? |
112 | ======================== | |
1da177e4 LT |
113 | |
114 | The first piece of information you must know is what kernel memory can | |
115 | be used with the DMA mapping facilities. There has been an unwritten | |
116 | set of rules regarding this, and this text is an attempt to finally | |
117 | write them down. | |
118 | ||
119 | If you acquired your memory via the page allocator | |
120 | (i.e. __get_free_page*()) or the generic memory allocators | |
121 | (i.e. kmalloc() or kmem_cache_alloc()) then you may DMA to/from | |
122 | that memory using the addresses returned from those routines. | |
123 | ||
124 | This means specifically that you may _not_ use the memory/addresses | |
125 | returned from vmalloc() for DMA. It is possible to DMA to the | |
126 | _underlying_ memory mapped into a vmalloc() area, but this requires | |
127 | walking page tables to get the physical addresses, and then | |
128 | translating each of those pages back to a kernel address using | |
129 | something like __va(). [ EDIT: Update this when we integrate | |
130 | Gerd Knorr's generic code which does this. ] | |
131 | ||
21440d31 DB |
132 | This rule also means that you may use neither kernel image addresses |
133 | (items in data/text/bss segments), nor module image addresses, nor | |
134 | stack addresses for DMA. These could all be mapped somewhere entirely | |
135 | different than the rest of physical memory. Even if those classes of | |
136 | memory could physically work with DMA, you'd need to ensure the I/O | |
137 | buffers were cacheline-aligned. Without that, you'd see cacheline | |
138 | sharing problems (data corruption) on CPUs with DMA-incoherent caches. | |
139 | (The CPU could write to one word, DMA would write to a different one | |
140 | in the same cache line, and one of them could be overwritten.) | |
1da177e4 LT |
141 | |
142 | Also, this means that you cannot take the return of a kmap() | |
143 | call and DMA to/from that. This is similar to vmalloc(). | |
144 | ||
145 | What about block I/O and networking buffers? The block I/O and | |
146 | networking subsystems make sure that the buffers they use are valid | |
147 | for you to DMA from/to. | |
148 | ||
9eb9e96e | 149 | DMA addressing capabilities |
266921bd | 150 | ========================== |
1da177e4 | 151 | |
9eb9e96e CH |
152 | By default, the kernel assumes that your device can address 32-bits of DMA |
153 | addressing. For a 64-bit capable device, this needs to be increased, and for | |
154 | a device with limitations, it needs to be decreased. | |
1da177e4 | 155 | |
9eb9e96e CH |
156 | Special note about PCI: PCI-X specification requires PCI-X devices to support |
157 | 64-bit addressing (DAC) for all transactions. And at least one platform (SGI | |
158 | SN2) requires 64-bit consistent allocations to operate correctly when the IO | |
159 | bus is in PCI-X mode. | |
216bf58f | 160 | |
9eb9e96e CH |
161 | For correct operation, you must set the DMA mask to inform the kernel about |
162 | your devices DMA addressing capabilities. | |
216bf58f | 163 | |
9eb9e96e | 164 | This is performed via a call to dma_set_mask_and_coherent():: |
1da177e4 | 165 | |
4aa806b7 | 166 | int dma_set_mask_and_coherent(struct device *dev, u64 mask); |
1da177e4 | 167 | |
9eb9e96e CH |
168 | which will set the mask for both streaming and coherent APIs together. If you |
169 | have some special requirements, then the following two separate calls can be | |
170 | used instead: | |
1da177e4 | 171 | |
9eb9e96e | 172 | The setup for streaming mappings is performed via a call to |
266921bd | 173 | dma_set_mask():: |
4aa806b7 RK |
174 | |
175 | int dma_set_mask(struct device *dev, u64 mask); | |
176 | ||
9eb9e96e | 177 | The setup for consistent allocations is performed via a call |
266921bd | 178 | to dma_set_coherent_mask():: |
4aa806b7 RK |
179 | |
180 | int dma_set_coherent_mask(struct device *dev, u64 mask); | |
1da177e4 | 181 | |
9eb9e96e CH |
182 | Here, dev is a pointer to the device struct of your device, and mask is a bit |
183 | mask describing which bits of an address your device supports. Often the | |
184 | device struct of your device is embedded in the bus-specific device struct of | |
185 | your device. For example, &pdev->dev is a pointer to the device struct of a | |
186 | PCI device (pdev is a pointer to the PCI device struct of your device). | |
1da177e4 | 187 | |
9eb9e96e CH |
188 | These calls usually return zero to indicated your device can perform DMA |
189 | properly on the machine given the address mask you provided, but they might | |
190 | return an error if the mask is too small to be supportable on the given | |
191 | system. If it returns non-zero, your device cannot perform DMA properly on | |
192 | this platform, and attempting to do so will result in undefined behavior. | |
193 | You must not use DMA on this device unless the dma_set_mask family of | |
194 | functions has returned success. | |
1da177e4 | 195 | |
9eb9e96e | 196 | This means that in the failure case, you have two options: |
1da177e4 | 197 | |
9eb9e96e CH |
198 | 1) Use some non-DMA mode for data transfer, if possible. |
199 | 2) Ignore this device and do not initialize it. | |
1da177e4 | 200 | |
9eb9e96e CH |
201 | It is recommended that your driver print a kernel KERN_WARNING message when |
202 | setting the DMA mask fails. In this manner, if a user of your driver reports | |
203 | that performance is bad or that the device is not even detected, you can ask | |
204 | them for the kernel messages to find out exactly why. | |
1da177e4 | 205 | |
9eb9e96e | 206 | The standard 64-bit addressing device would do something like this:: |
1da177e4 | 207 | |
9eb9e96e | 208 | if (dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64))) { |
77f2ea2f | 209 | dev_warn(dev, "mydev: No suitable DMA available\n"); |
1da177e4 LT |
210 | goto ignore_this_device; |
211 | } | |
212 | ||
9eb9e96e CH |
213 | If the device only supports 32-bit addressing for descriptors in the |
214 | coherent allocations, but supports full 64-bits for streaming mappings | |
215 | it would look like this: | |
1da177e4 | 216 | |
9eb9e96e | 217 | if (dma_set_mask(dev, DMA_BIT_MASK(64))) { |
77f2ea2f | 218 | dev_warn(dev, "mydev: No suitable DMA available\n"); |
1da177e4 LT |
219 | goto ignore_this_device; |
220 | } | |
221 | ||
34c815fb EL |
222 | The coherent mask will always be able to set the same or a smaller mask as |
223 | the streaming mask. However for the rare case that a device driver only | |
224 | uses consistent allocations, one would have to check the return value from | |
225 | dma_set_coherent_mask(). | |
1da177e4 | 226 | |
1da177e4 | 227 | Finally, if your device can only drive the low 24-bits of |
266921bd | 228 | address you might do something like:: |
1da177e4 | 229 | |
216bf58f | 230 | if (dma_set_mask(dev, DMA_BIT_MASK(24))) { |
77f2ea2f | 231 | dev_warn(dev, "mydev: 24-bit DMA addressing not available\n"); |
1da177e4 LT |
232 | goto ignore_this_device; |
233 | } | |
234 | ||
4aa806b7 RK |
235 | When dma_set_mask() or dma_set_mask_and_coherent() is successful, and |
236 | returns zero, the kernel saves away this mask you have provided. The | |
237 | kernel will use this information later when you make DMA mappings. | |
1da177e4 LT |
238 | |
239 | There is a case which we are aware of at this time, which is worth | |
240 | mentioning in this documentation. If your device supports multiple | |
241 | functions (for example a sound card provides playback and record | |
242 | functions) and the various different functions have _different_ | |
243 | DMA addressing limitations, you may wish to probe each mask and | |
244 | only provide the functionality which the machine can handle. It | |
216bf58f | 245 | is important that the last call to dma_set_mask() be for the |
1da177e4 LT |
246 | most specific mask. |
247 | ||
266921bd | 248 | Here is pseudo-code showing how this might be done:: |
1da177e4 | 249 | |
2c5510d4 | 250 | #define PLAYBACK_ADDRESS_BITS DMA_BIT_MASK(32) |
038f7d00 | 251 | #define RECORD_ADDRESS_BITS DMA_BIT_MASK(24) |
1da177e4 LT |
252 | |
253 | struct my_sound_card *card; | |
216bf58f | 254 | struct device *dev; |
1da177e4 LT |
255 | |
256 | ... | |
216bf58f | 257 | if (!dma_set_mask(dev, PLAYBACK_ADDRESS_BITS)) { |
1da177e4 LT |
258 | card->playback_enabled = 1; |
259 | } else { | |
260 | card->playback_enabled = 0; | |
77f2ea2f | 261 | dev_warn(dev, "%s: Playback disabled due to DMA limitations\n", |
1da177e4 LT |
262 | card->name); |
263 | } | |
216bf58f | 264 | if (!dma_set_mask(dev, RECORD_ADDRESS_BITS)) { |
1da177e4 LT |
265 | card->record_enabled = 1; |
266 | } else { | |
267 | card->record_enabled = 0; | |
77f2ea2f | 268 | dev_warn(dev, "%s: Record disabled due to DMA limitations\n", |
1da177e4 LT |
269 | card->name); |
270 | } | |
271 | ||
272 | A sound card was used as an example here because this genre of PCI | |
273 | devices seems to be littered with ISA chips given a PCI front end, | |
274 | and thus retaining the 16MB DMA addressing limitations of ISA. | |
275 | ||
266921bd MCC |
276 | Types of DMA mappings |
277 | ===================== | |
1da177e4 LT |
278 | |
279 | There are two types of DMA mappings: | |
280 | ||
281 | - Consistent DMA mappings which are usually mapped at driver | |
282 | initialization, unmapped at the end and for which the hardware should | |
283 | guarantee that the device and the CPU can access the data | |
284 | in parallel and will see updates made by each other without any | |
285 | explicit software flushing. | |
286 | ||
287 | Think of "consistent" as "synchronous" or "coherent". | |
288 | ||
289 | The current default is to return consistent memory in the low 32 | |
3a9ad0b4 | 290 | bits of the DMA space. However, for future compatibility you should |
216bf58f | 291 | set the consistent mask even if this default is fine for your |
1da177e4 LT |
292 | driver. |
293 | ||
294 | Good examples of what to use consistent mappings for are: | |
295 | ||
296 | - Network card DMA ring descriptors. | |
297 | - SCSI adapter mailbox command data structures. | |
298 | - Device firmware microcode executed out of | |
299 | main memory. | |
300 | ||
301 | The invariant these examples all require is that any CPU store | |
302 | to memory is immediately visible to the device, and vice | |
303 | versa. Consistent mappings guarantee this. | |
304 | ||
266921bd MCC |
305 | .. important:: |
306 | ||
307 | Consistent DMA memory does not preclude the usage of | |
308 | proper memory barriers. The CPU may reorder stores to | |
1da177e4 LT |
309 | consistent memory just as it may normal memory. Example: |
310 | if it is important for the device to see the first word | |
311 | of a descriptor updated before the second, you must do | |
266921bd | 312 | something like:: |
1da177e4 LT |
313 | |
314 | desc->word0 = address; | |
315 | wmb(); | |
316 | desc->word1 = DESC_VALID; | |
317 | ||
318 | in order to get correct behavior on all platforms. | |
319 | ||
21440d31 DB |
320 | Also, on some platforms your driver may need to flush CPU write |
321 | buffers in much the same way as it needs to flush write buffers | |
322 | found in PCI bridges (such as by reading a register's value | |
323 | after writing it). | |
324 | ||
216bf58f FT |
325 | - Streaming DMA mappings which are usually mapped for one DMA |
326 | transfer, unmapped right after it (unless you use dma_sync_* below) | |
327 | and for which hardware can optimize for sequential accesses. | |
1da177e4 | 328 | |
11e285d8 | 329 | Think of "streaming" as "asynchronous" or "outside the coherency |
1da177e4 LT |
330 | domain". |
331 | ||
332 | Good examples of what to use streaming mappings for are: | |
333 | ||
334 | - Networking buffers transmitted/received by a device. | |
335 | - Filesystem buffers written/read by a SCSI device. | |
336 | ||
337 | The interfaces for using this type of mapping were designed in | |
338 | such a way that an implementation can make whatever performance | |
339 | optimizations the hardware allows. To this end, when using | |
340 | such mappings you must be explicit about what you want to happen. | |
341 | ||
216bf58f FT |
342 | Neither type of DMA mapping has alignment restrictions that come from |
343 | the underlying bus, although some devices may have such restrictions. | |
21440d31 DB |
344 | Also, systems with caches that aren't DMA-coherent will work better |
345 | when the underlying buffers don't share cache lines with other data. | |
346 | ||
1da177e4 | 347 | |
266921bd MCC |
348 | Using Consistent DMA mappings |
349 | ============================= | |
1da177e4 LT |
350 | |
351 | To allocate and map large (PAGE_SIZE or so) consistent DMA regions, | |
266921bd | 352 | you should do:: |
1da177e4 LT |
353 | |
354 | dma_addr_t dma_handle; | |
355 | ||
216bf58f | 356 | cpu_addr = dma_alloc_coherent(dev, size, &dma_handle, gfp); |
1da177e4 | 357 | |
266921bd | 358 | where device is a ``struct device *``. This may be called in interrupt |
216bf58f | 359 | context with the GFP_ATOMIC flag. |
1da177e4 LT |
360 | |
361 | Size is the length of the region you want to allocate, in bytes. | |
362 | ||
363 | This routine will allocate RAM for that region, so it acts similarly to | |
77f2ea2f | 364 | __get_free_pages() (but takes size instead of a page order). If your |
1da177e4 | 365 | driver needs regions sized smaller than a page, you may prefer using |
216bf58f FT |
366 | the dma_pool interface, described below. |
367 | ||
368 | The consistent DMA mapping interfaces, for non-NULL dev, will by | |
369 | default return a DMA address which is 32-bit addressable. Even if the | |
370 | device indicates (via DMA mask) that it may address the upper 32-bits, | |
371 | consistent allocation will only return > 32-bit addresses for DMA if | |
372 | the consistent DMA mask has been explicitly changed via | |
373 | dma_set_coherent_mask(). This is true of the dma_pool interface as | |
374 | well. | |
375 | ||
77f2ea2f | 376 | dma_alloc_coherent() returns two values: the virtual address which you |
1da177e4 LT |
377 | can use to access it from the CPU and dma_handle which you pass to the |
378 | card. | |
379 | ||
3a9ad0b4 | 380 | The CPU virtual address and the DMA address are both |
1da177e4 LT |
381 | guaranteed to be aligned to the smallest PAGE_SIZE order which |
382 | is greater than or equal to the requested size. This invariant | |
383 | exists (for example) to guarantee that if you allocate a chunk | |
384 | which is smaller than or equal to 64 kilobytes, the extent of the | |
385 | buffer you receive will not cross a 64K boundary. | |
386 | ||
266921bd | 387 | To unmap and free such a DMA region, you call:: |
1da177e4 | 388 | |
216bf58f | 389 | dma_free_coherent(dev, size, cpu_addr, dma_handle); |
1da177e4 | 390 | |
216bf58f | 391 | where dev, size are the same as in the above call and cpu_addr and |
77f2ea2f | 392 | dma_handle are the values dma_alloc_coherent() returned to you. |
1da177e4 LT |
393 | This function may not be called in interrupt context. |
394 | ||
395 | If your driver needs lots of smaller memory regions, you can write | |
77f2ea2f | 396 | custom code to subdivide pages returned by dma_alloc_coherent(), |
216bf58f | 397 | or you can use the dma_pool API to do that. A dma_pool is like |
77f2ea2f | 398 | a kmem_cache, but it uses dma_alloc_coherent(), not __get_free_pages(). |
1da177e4 LT |
399 | Also, it understands common hardware constraints for alignment, |
400 | like queue heads needing to be aligned on N byte boundaries. | |
401 | ||
266921bd | 402 | Create a dma_pool like this:: |
1da177e4 | 403 | |
216bf58f | 404 | struct dma_pool *pool; |
1da177e4 | 405 | |
2af9da86 | 406 | pool = dma_pool_create(name, dev, size, align, boundary); |
1da177e4 | 407 | |
216bf58f | 408 | The "name" is for diagnostics (like a kmem_cache name); dev and size |
1da177e4 LT |
409 | are as above. The device's hardware alignment requirement for this |
410 | type of data is "align" (which is expressed in bytes, and must be a | |
411 | power of two). If your device has no boundary crossing restrictions, | |
2af9da86 | 412 | pass 0 for boundary; passing 4096 says memory allocated from this pool |
1da177e4 | 413 | must not cross 4KByte boundaries (but at that time it may be better to |
77f2ea2f | 414 | use dma_alloc_coherent() directly instead). |
1da177e4 | 415 | |
266921bd | 416 | Allocate memory from a DMA pool like this:: |
1da177e4 | 417 | |
216bf58f | 418 | cpu_addr = dma_pool_alloc(pool, flags, &dma_handle); |
1da177e4 | 419 | |
2af9da86 GK |
420 | flags are GFP_KERNEL if blocking is permitted (not in_interrupt nor |
421 | holding SMP locks), GFP_ATOMIC otherwise. Like dma_alloc_coherent(), | |
1da177e4 LT |
422 | this returns two values, cpu_addr and dma_handle. |
423 | ||
266921bd | 424 | Free memory that was allocated from a dma_pool like this:: |
1da177e4 | 425 | |
216bf58f | 426 | dma_pool_free(pool, cpu_addr, dma_handle); |
1da177e4 | 427 | |
77f2ea2f BH |
428 | where pool is what you passed to dma_pool_alloc(), and cpu_addr and |
429 | dma_handle are the values dma_pool_alloc() returned. This function | |
1da177e4 LT |
430 | may be called in interrupt context. |
431 | ||
266921bd | 432 | Destroy a dma_pool by calling:: |
1da177e4 | 433 | |
216bf58f | 434 | dma_pool_destroy(pool); |
1da177e4 | 435 | |
77f2ea2f | 436 | Make sure you've called dma_pool_free() for all memory allocated |
1da177e4 LT |
437 | from a pool before you destroy the pool. This function may not |
438 | be called in interrupt context. | |
439 | ||
266921bd MCC |
440 | DMA Direction |
441 | ============= | |
1da177e4 LT |
442 | |
443 | The interfaces described in subsequent portions of this document | |
444 | take a DMA direction argument, which is an integer and takes on | |
266921bd | 445 | one of the following values:: |
1da177e4 | 446 | |
216bf58f FT |
447 | DMA_BIDIRECTIONAL |
448 | DMA_TO_DEVICE | |
449 | DMA_FROM_DEVICE | |
450 | DMA_NONE | |
1da177e4 | 451 | |
77f2ea2f | 452 | You should provide the exact DMA direction if you know it. |
1da177e4 | 453 | |
216bf58f FT |
454 | DMA_TO_DEVICE means "from main memory to the device" |
455 | DMA_FROM_DEVICE means "from the device to main memory" | |
1da177e4 LT |
456 | It is the direction in which the data moves during the DMA |
457 | transfer. | |
458 | ||
459 | You are _strongly_ encouraged to specify this as precisely | |
460 | as you possibly can. | |
461 | ||
462 | If you absolutely cannot know the direction of the DMA transfer, | |
216bf58f | 463 | specify DMA_BIDIRECTIONAL. It means that the DMA can go in |
1da177e4 LT |
464 | either direction. The platform guarantees that you may legally |
465 | specify this, and that it will work, but this may be at the | |
466 | cost of performance for example. | |
467 | ||
216bf58f | 468 | The value DMA_NONE is to be used for debugging. One can |
1da177e4 LT |
469 | hold this in a data structure before you come to know the |
470 | precise direction, and this will help catch cases where your | |
471 | direction tracking logic has failed to set things up properly. | |
472 | ||
473 | Another advantage of specifying this value precisely (outside of | |
474 | potential platform-specific optimizations of such) is for debugging. | |
475 | Some platforms actually have a write permission boolean which DMA | |
476 | mappings can be marked with, much like page protections in the user | |
477 | program address space. Such platforms can and do report errors in the | |
216bf58f | 478 | kernel logs when the DMA controller hardware detects violation of the |
1da177e4 LT |
479 | permission setting. |
480 | ||
481 | Only streaming mappings specify a direction, consistent mappings | |
482 | implicitly have a direction attribute setting of | |
216bf58f | 483 | DMA_BIDIRECTIONAL. |
1da177e4 | 484 | |
be7db055 | 485 | The SCSI subsystem tells you the direction to use in the |
486 | 'sc_data_direction' member of the SCSI command your driver is | |
487 | working on. | |
1da177e4 LT |
488 | |
489 | For Networking drivers, it's a rather simple affair. For transmit | |
216bf58f | 490 | packets, map/unmap them with the DMA_TO_DEVICE direction |
1da177e4 | 491 | specifier. For receive packets, just the opposite, map/unmap them |
216bf58f | 492 | with the DMA_FROM_DEVICE direction specifier. |
1da177e4 | 493 | |
266921bd MCC |
494 | Using Streaming DMA mappings |
495 | ============================ | |
1da177e4 LT |
496 | |
497 | The streaming DMA mapping routines can be called from interrupt | |
498 | context. There are two versions of each map/unmap, one which will | |
499 | map/unmap a single memory region, and one which will map/unmap a | |
500 | scatterlist. | |
501 | ||
266921bd | 502 | To map a single region, you do:: |
1da177e4 | 503 | |
216bf58f | 504 | struct device *dev = &my_dev->dev; |
1da177e4 LT |
505 | dma_addr_t dma_handle; |
506 | void *addr = buffer->ptr; | |
507 | size_t size = buffer->len; | |
508 | ||
216bf58f | 509 | dma_handle = dma_map_single(dev, addr, size, direction); |
b2dd83b3 | 510 | if (dma_mapping_error(dev, dma_handle)) { |
8d7f62e6 SK |
511 | /* |
512 | * reduce current DMA mapping usage, | |
513 | * delay and try again later or | |
514 | * reset driver. | |
515 | */ | |
516 | goto map_error_handling; | |
517 | } | |
1da177e4 | 518 | |
266921bd | 519 | and to unmap it:: |
1da177e4 | 520 | |
216bf58f | 521 | dma_unmap_single(dev, dma_handle, size, direction); |
1da177e4 | 522 | |
8d7f62e6 | 523 | You should call dma_mapping_error() as dma_map_single() could fail and return |
f51f288e CH |
524 | error. Doing so will ensure that the mapping code will work correctly on all |
525 | DMA implementations without any dependency on the specifics of the underlying | |
526 | implementation. Using the returned address without checking for errors could | |
527 | result in failures ranging from panics to silent data corruption. The same | |
528 | applies to dma_map_page() as well. | |
8d7f62e6 | 529 | |
77f2ea2f | 530 | You should call dma_unmap_single() when the DMA activity is finished, e.g., |
1da177e4 LT |
531 | from the interrupt which told you that the DMA transfer is done. |
532 | ||
f311a724 | 533 | Using CPU pointers like this for single mappings has a disadvantage: |
1da177e4 | 534 | you cannot reference HIGHMEM memory in this way. Thus, there is a |
77f2ea2f | 535 | map/unmap interface pair akin to dma_{map,unmap}_single(). These |
f311a724 | 536 | interfaces deal with page/offset pairs instead of CPU pointers. |
266921bd | 537 | Specifically:: |
1da177e4 | 538 | |
216bf58f | 539 | struct device *dev = &my_dev->dev; |
1da177e4 LT |
540 | dma_addr_t dma_handle; |
541 | struct page *page = buffer->page; | |
542 | unsigned long offset = buffer->offset; | |
543 | size_t size = buffer->len; | |
544 | ||
216bf58f | 545 | dma_handle = dma_map_page(dev, page, offset, size, direction); |
b2dd83b3 | 546 | if (dma_mapping_error(dev, dma_handle)) { |
8d7f62e6 SK |
547 | /* |
548 | * reduce current DMA mapping usage, | |
549 | * delay and try again later or | |
550 | * reset driver. | |
551 | */ | |
552 | goto map_error_handling; | |
553 | } | |
1da177e4 LT |
554 | |
555 | ... | |
556 | ||
216bf58f | 557 | dma_unmap_page(dev, dma_handle, size, direction); |
1da177e4 LT |
558 | |
559 | Here, "offset" means byte offset within the given page. | |
560 | ||
8d7f62e6 SK |
561 | You should call dma_mapping_error() as dma_map_page() could fail and return |
562 | error as outlined under the dma_map_single() discussion. | |
563 | ||
77f2ea2f | 564 | You should call dma_unmap_page() when the DMA activity is finished, e.g., |
8d7f62e6 SK |
565 | from the interrupt which told you that the DMA transfer is done. |
566 | ||
266921bd | 567 | With scatterlists, you map a region gathered from several regions by:: |
1da177e4 | 568 | |
216bf58f | 569 | int i, count = dma_map_sg(dev, sglist, nents, direction); |
1da177e4 LT |
570 | struct scatterlist *sg; |
571 | ||
4c2f6d4c | 572 | for_each_sg(sglist, sg, count, i) { |
1da177e4 LT |
573 | hw_address[i] = sg_dma_address(sg); |
574 | hw_len[i] = sg_dma_len(sg); | |
575 | } | |
576 | ||
577 | where nents is the number of entries in the sglist. | |
578 | ||
579 | The implementation is free to merge several consecutive sglist entries | |
580 | into one (e.g. if DMA mapping is done with PAGE_SIZE granularity, any | |
581 | consecutive sglist entries can be merged into one provided the first one | |
582 | ends and the second one starts on a page boundary - in fact this is a huge | |
583 | advantage for cards which either cannot do scatter-gather or have very | |
584 | limited number of scatter-gather entries) and returns the actual number | |
585 | of sg entries it mapped them to. On failure 0 is returned. | |
586 | ||
587 | Then you should loop count times (note: this can be less than nents times) | |
588 | and use sg_dma_address() and sg_dma_len() macros where you previously | |
589 | accessed sg->address and sg->length as shown above. | |
590 | ||
266921bd | 591 | To unmap a scatterlist, just call:: |
1da177e4 | 592 | |
216bf58f | 593 | dma_unmap_sg(dev, sglist, nents, direction); |
1da177e4 LT |
594 | |
595 | Again, make sure DMA activity has already finished. | |
596 | ||
266921bd MCC |
597 | .. note:: |
598 | ||
599 | The 'nents' argument to the dma_unmap_sg call must be | |
600 | the _same_ one you passed into the dma_map_sg call, | |
601 | it should _NOT_ be the 'count' value _returned_ from the | |
602 | dma_map_sg call. | |
1da177e4 | 603 | |
77f2ea2f | 604 | Every dma_map_{single,sg}() call should have its dma_unmap_{single,sg}() |
3a9ad0b4 YL |
605 | counterpart, because the DMA address space is a shared resource and |
606 | you could render the machine unusable by consuming all DMA addresses. | |
1da177e4 LT |
607 | |
608 | If you need to use the same streaming DMA region multiple times and touch | |
609 | the data in between the DMA transfers, the buffer needs to be synced | |
f311a724 | 610 | properly in order for the CPU and device to see the most up-to-date and |
1da177e4 LT |
611 | correct copy of the DMA buffer. |
612 | ||
77f2ea2f | 613 | So, firstly, just map it with dma_map_{single,sg}(), and after each DMA |
266921bd | 614 | transfer call either:: |
1da177e4 | 615 | |
216bf58f | 616 | dma_sync_single_for_cpu(dev, dma_handle, size, direction); |
1da177e4 | 617 | |
266921bd | 618 | or:: |
1da177e4 | 619 | |
216bf58f | 620 | dma_sync_sg_for_cpu(dev, sglist, nents, direction); |
1da177e4 LT |
621 | |
622 | as appropriate. | |
623 | ||
624 | Then, if you wish to let the device get at the DMA area again, | |
f311a724 | 625 | finish accessing the data with the CPU, and then before actually |
266921bd | 626 | giving the buffer to the hardware call either:: |
1da177e4 | 627 | |
216bf58f | 628 | dma_sync_single_for_device(dev, dma_handle, size, direction); |
1da177e4 | 629 | |
266921bd | 630 | or:: |
1da177e4 | 631 | |
216bf58f | 632 | dma_sync_sg_for_device(dev, sglist, nents, direction); |
1da177e4 LT |
633 | |
634 | as appropriate. | |
635 | ||
266921bd MCC |
636 | .. note:: |
637 | ||
638 | The 'nents' argument to dma_sync_sg_for_cpu() and | |
7bc590b2 SA |
639 | dma_sync_sg_for_device() must be the same passed to |
640 | dma_map_sg(). It is _NOT_ the count returned by | |
641 | dma_map_sg(). | |
642 | ||
1da177e4 | 643 | After the last DMA transfer call one of the DMA unmap routines |
77f2ea2f BH |
644 | dma_unmap_{single,sg}(). If you don't touch the data from the first |
645 | dma_map_*() call till dma_unmap_*(), then you don't have to call the | |
646 | dma_sync_*() routines at all. | |
1da177e4 LT |
647 | |
648 | Here is pseudo code which shows a situation in which you would need | |
266921bd | 649 | to use the dma_sync_*() interfaces:: |
1da177e4 LT |
650 | |
651 | my_card_setup_receive_buffer(struct my_card *cp, char *buffer, int len) | |
652 | { | |
653 | dma_addr_t mapping; | |
654 | ||
216bf58f | 655 | mapping = dma_map_single(cp->dev, buffer, len, DMA_FROM_DEVICE); |
be6c3095 | 656 | if (dma_mapping_error(cp->dev, mapping)) { |
8d7f62e6 SK |
657 | /* |
658 | * reduce current DMA mapping usage, | |
659 | * delay and try again later or | |
660 | * reset driver. | |
661 | */ | |
662 | goto map_error_handling; | |
663 | } | |
1da177e4 LT |
664 | |
665 | cp->rx_buf = buffer; | |
666 | cp->rx_len = len; | |
667 | cp->rx_dma = mapping; | |
668 | ||
669 | give_rx_buf_to_card(cp); | |
670 | } | |
671 | ||
672 | ... | |
673 | ||
674 | my_card_interrupt_handler(int irq, void *devid, struct pt_regs *regs) | |
675 | { | |
676 | struct my_card *cp = devid; | |
677 | ||
678 | ... | |
679 | if (read_card_status(cp) == RX_BUF_TRANSFERRED) { | |
680 | struct my_card_header *hp; | |
681 | ||
682 | /* Examine the header to see if we wish | |
683 | * to accept the data. But synchronize | |
684 | * the DMA transfer with the CPU first | |
685 | * so that we see updated contents. | |
686 | */ | |
216bf58f FT |
687 | dma_sync_single_for_cpu(&cp->dev, cp->rx_dma, |
688 | cp->rx_len, | |
689 | DMA_FROM_DEVICE); | |
1da177e4 LT |
690 | |
691 | /* Now it is safe to examine the buffer. */ | |
692 | hp = (struct my_card_header *) cp->rx_buf; | |
693 | if (header_is_ok(hp)) { | |
216bf58f FT |
694 | dma_unmap_single(&cp->dev, cp->rx_dma, cp->rx_len, |
695 | DMA_FROM_DEVICE); | |
1da177e4 LT |
696 | pass_to_upper_layers(cp->rx_buf); |
697 | make_and_setup_new_rx_buf(cp); | |
698 | } else { | |
3f0fb4e8 MM |
699 | /* CPU should not write to |
700 | * DMA_FROM_DEVICE-mapped area, | |
701 | * so dma_sync_single_for_device() is | |
702 | * not needed here. It would be required | |
703 | * for DMA_BIDIRECTIONAL mapping if | |
704 | * the memory was modified. | |
1da177e4 | 705 | */ |
1da177e4 LT |
706 | give_rx_buf_to_card(cp); |
707 | } | |
708 | } | |
709 | } | |
710 | ||
77f2ea2f BH |
711 | Drivers converted fully to this interface should not use virt_to_bus() any |
712 | longer, nor should they use bus_to_virt(). Some drivers have to be changed a | |
713 | little bit, because there is no longer an equivalent to bus_to_virt() in the | |
1da177e4 | 714 | dynamic DMA mapping scheme - you have to always store the DMA addresses |
77f2ea2f BH |
715 | returned by the dma_alloc_coherent(), dma_pool_alloc(), and dma_map_single() |
716 | calls (dma_map_sg() stores them in the scatterlist itself if the platform | |
1da177e4 LT |
717 | supports dynamic DMA mapping in hardware) in your driver structures and/or |
718 | in the card registers. | |
719 | ||
216bf58f FT |
720 | All drivers should be using these interfaces with no exceptions. It |
721 | is planned to completely remove virt_to_bus() and bus_to_virt() as | |
1da177e4 LT |
722 | they are entirely deprecated. Some ports already do not provide these |
723 | as it is impossible to correctly support them. | |
724 | ||
266921bd MCC |
725 | Handling Errors |
726 | =============== | |
4ae9ca82 FT |
727 | |
728 | DMA address space is limited on some architectures and an allocation | |
729 | failure can be determined by: | |
730 | ||
77f2ea2f | 731 | - checking if dma_alloc_coherent() returns NULL or dma_map_sg returns 0 |
4ae9ca82 | 732 | |
77f2ea2f | 733 | - checking the dma_addr_t returned from dma_map_single() and dma_map_page() |
266921bd | 734 | by using dma_mapping_error():: |
4ae9ca82 FT |
735 | |
736 | dma_addr_t dma_handle; | |
737 | ||
738 | dma_handle = dma_map_single(dev, addr, size, direction); | |
739 | if (dma_mapping_error(dev, dma_handle)) { | |
740 | /* | |
741 | * reduce current DMA mapping usage, | |
742 | * delay and try again later or | |
743 | * reset driver. | |
744 | */ | |
8d7f62e6 SK |
745 | goto map_error_handling; |
746 | } | |
747 | ||
748 | - unmap pages that are already mapped, when mapping error occurs in the middle | |
749 | of a multiple page mapping attempt. These example are applicable to | |
750 | dma_map_page() as well. | |
751 | ||
266921bd MCC |
752 | Example 1:: |
753 | ||
8d7f62e6 SK |
754 | dma_addr_t dma_handle1; |
755 | dma_addr_t dma_handle2; | |
756 | ||
757 | dma_handle1 = dma_map_single(dev, addr, size, direction); | |
758 | if (dma_mapping_error(dev, dma_handle1)) { | |
759 | /* | |
760 | * reduce current DMA mapping usage, | |
761 | * delay and try again later or | |
762 | * reset driver. | |
763 | */ | |
764 | goto map_error_handling1; | |
765 | } | |
766 | dma_handle2 = dma_map_single(dev, addr, size, direction); | |
767 | if (dma_mapping_error(dev, dma_handle2)) { | |
768 | /* | |
769 | * reduce current DMA mapping usage, | |
770 | * delay and try again later or | |
771 | * reset driver. | |
772 | */ | |
773 | goto map_error_handling2; | |
774 | } | |
775 | ||
776 | ... | |
777 | ||
778 | map_error_handling2: | |
779 | dma_unmap_single(dma_handle1); | |
780 | map_error_handling1: | |
781 | ||
266921bd MCC |
782 | Example 2:: |
783 | ||
784 | /* | |
785 | * if buffers are allocated in a loop, unmap all mapped buffers when | |
786 | * mapping error is detected in the middle | |
787 | */ | |
8d7f62e6 SK |
788 | |
789 | dma_addr_t dma_addr; | |
790 | dma_addr_t array[DMA_BUFFERS]; | |
791 | int save_index = 0; | |
792 | ||
793 | for (i = 0; i < DMA_BUFFERS; i++) { | |
794 | ||
795 | ... | |
796 | ||
797 | dma_addr = dma_map_single(dev, addr, size, direction); | |
798 | if (dma_mapping_error(dev, dma_addr)) { | |
799 | /* | |
800 | * reduce current DMA mapping usage, | |
801 | * delay and try again later or | |
802 | * reset driver. | |
803 | */ | |
804 | goto map_error_handling; | |
805 | } | |
806 | array[i].dma_addr = dma_addr; | |
807 | save_index++; | |
808 | } | |
809 | ||
810 | ... | |
811 | ||
812 | map_error_handling: | |
813 | ||
814 | for (i = 0; i < save_index; i++) { | |
815 | ||
816 | ... | |
817 | ||
818 | dma_unmap_single(array[i].dma_addr); | |
4ae9ca82 FT |
819 | } |
820 | ||
77f2ea2f | 821 | Networking drivers must call dev_kfree_skb() to free the socket buffer |
4ae9ca82 FT |
822 | and return NETDEV_TX_OK if the DMA mapping fails on the transmit hook |
823 | (ndo_start_xmit). This means that the socket buffer is just dropped in | |
824 | the failure case. | |
825 | ||
826 | SCSI drivers must return SCSI_MLQUEUE_HOST_BUSY if the DMA mapping | |
827 | fails in the queuecommand hook. This means that the SCSI subsystem | |
828 | passes the command to the driver again later. | |
829 | ||
266921bd MCC |
830 | Optimizing Unmap State Space Consumption |
831 | ======================================== | |
1da177e4 | 832 | |
216bf58f | 833 | On many platforms, dma_unmap_{single,page}() is simply a nop. |
1da177e4 LT |
834 | Therefore, keeping track of the mapping address and length is a waste |
835 | of space. Instead of filling your drivers up with ifdefs and the like | |
836 | to "work around" this (which would defeat the whole purpose of a | |
837 | portable API) the following facilities are provided. | |
838 | ||
839 | Actually, instead of describing the macros one by one, we'll | |
840 | transform some example code. | |
841 | ||
216bf58f | 842 | 1) Use DEFINE_DMA_UNMAP_{ADDR,LEN} in state saving structures. |
266921bd | 843 | Example, before:: |
1da177e4 LT |
844 | |
845 | struct ring_state { | |
846 | struct sk_buff *skb; | |
847 | dma_addr_t mapping; | |
848 | __u32 len; | |
849 | }; | |
850 | ||
266921bd | 851 | after:: |
1da177e4 LT |
852 | |
853 | struct ring_state { | |
854 | struct sk_buff *skb; | |
216bf58f FT |
855 | DEFINE_DMA_UNMAP_ADDR(mapping); |
856 | DEFINE_DMA_UNMAP_LEN(len); | |
1da177e4 LT |
857 | }; |
858 | ||
77f2ea2f | 859 | 2) Use dma_unmap_{addr,len}_set() to set these values. |
266921bd | 860 | Example, before:: |
1da177e4 LT |
861 | |
862 | ringp->mapping = FOO; | |
863 | ringp->len = BAR; | |
864 | ||
266921bd | 865 | after:: |
1da177e4 | 866 | |
216bf58f FT |
867 | dma_unmap_addr_set(ringp, mapping, FOO); |
868 | dma_unmap_len_set(ringp, len, BAR); | |
1da177e4 | 869 | |
77f2ea2f | 870 | 3) Use dma_unmap_{addr,len}() to access these values. |
266921bd | 871 | Example, before:: |
1da177e4 | 872 | |
216bf58f FT |
873 | dma_unmap_single(dev, ringp->mapping, ringp->len, |
874 | DMA_FROM_DEVICE); | |
1da177e4 | 875 | |
266921bd | 876 | after:: |
1da177e4 | 877 | |
216bf58f FT |
878 | dma_unmap_single(dev, |
879 | dma_unmap_addr(ringp, mapping), | |
880 | dma_unmap_len(ringp, len), | |
881 | DMA_FROM_DEVICE); | |
1da177e4 LT |
882 | |
883 | It really should be self-explanatory. We treat the ADDR and LEN | |
884 | separately, because it is possible for an implementation to only | |
885 | need the address in order to perform the unmap operation. | |
886 | ||
266921bd MCC |
887 | Platform Issues |
888 | =============== | |
1da177e4 LT |
889 | |
890 | If you are just writing drivers for Linux and do not maintain | |
891 | an architecture port for the kernel, you can safely skip down | |
892 | to "Closing". | |
893 | ||
894 | 1) Struct scatterlist requirements. | |
895 | ||
e92ae527 CH |
896 | You need to enable CONFIG_NEED_SG_DMA_LENGTH if the architecture |
897 | supports IOMMUs (including software IOMMU). | |
1da177e4 | 898 | |
ce00f7fe | 899 | 2) ARCH_DMA_MINALIGN |
2fd74e25 FT |
900 | |
901 | Architectures must ensure that kmalloc'ed buffer is | |
902 | DMA-safe. Drivers and subsystems depend on it. If an architecture | |
903 | isn't fully DMA-coherent (i.e. hardware doesn't ensure that data in | |
904 | the CPU cache is identical to data in main memory), | |
ce00f7fe | 905 | ARCH_DMA_MINALIGN must be set so that the memory allocator |
2fd74e25 FT |
906 | makes sure that kmalloc'ed buffer doesn't share a cache line with |
907 | the others. See arch/arm/include/asm/cache.h as an example. | |
908 | ||
ce00f7fe | 909 | Note that ARCH_DMA_MINALIGN is about DMA memory alignment |
2fd74e25 FT |
910 | constraints. You don't need to worry about the architecture data |
911 | alignment constraints (e.g. the alignment constraints about 64-bit | |
912 | objects). | |
1da177e4 | 913 | |
266921bd MCC |
914 | Closing |
915 | ======= | |
1da177e4 | 916 | |
a33f3224 | 917 | This document, and the API itself, would not be in its current |
1da177e4 LT |
918 | form without the feedback and suggestions from numerous individuals. |
919 | We would like to specifically mention, in no particular order, the | |
266921bd | 920 | following people:: |
1da177e4 LT |
921 | |
922 | Russell King <rmk@arm.linux.org.uk> | |
923 | Leo Dagum <dagum@barrel.engr.sgi.com> | |
924 | Ralf Baechle <ralf@oss.sgi.com> | |
925 | Grant Grundler <grundler@cup.hp.com> | |
926 | Jay Estabrook <Jay.Estabrook@compaq.com> | |
927 | Thomas Sailer <sailer@ife.ee.ethz.ch> | |
928 | Andrea Arcangeli <andrea@suse.de> | |
26bbb29a | 929 | Jens Axboe <jens.axboe@oracle.com> |
1da177e4 | 930 | David Mosberger-Tang <davidm@hpl.hp.com> |