1 ================================
2 Devres - Managed Device Resource
3 ================================
5 Tejun Heo <teheo@suse.de>
7 First draft 10 January 2007
11 1. Intro : Huh? Devres?
12 2. Devres : Devres in a nutshell
13 3. Devres Group : Group devres'es and release them together
14 4. Details : Life time rules, calling context, ...
15 5. Overhead : How much do we have to pay for this?
16 6. List of managed interfaces: Currently implemented managed interfaces
22 devres came up while trying to convert libata to use iomap. Each
23 iomapped address should be kept and unmapped on driver detach. For
24 example, a plain SFF ATA controller (that is, good old PCI IDE) in
25 native mode makes use of 5 PCI BARs and all of them should be
28 As with many other device drivers, libata low level drivers have
29 sufficient bugs in ->remove and ->probe failure path. Well, yes,
30 that's probably because libata low level driver developers are lazy
31 bunch, but aren't all low level driver developers? After spending a
32 day fiddling with braindamaged hardware with no document or
33 braindamaged document, if it's finally working, well, it's working.
35 For one reason or another, low level drivers don't receive as much
36 attention or testing as core code, and bugs on driver detach or
37 initialization failure don't happen often enough to be noticeable.
38 Init failure path is worse because it's much less travelled while
39 needs to handle multiple entry points.
41 So, many low level drivers end up leaking resources on driver detach
42 and having half broken failure path implementation in ->probe() which
43 would leak resources or even cause oops when failure occurs. iomap
44 adds more to this mix. So do msi and msix.
50 devres is basically linked list of arbitrarily sized memory areas
51 associated with a struct device. Each devres entry is associated with
52 a release function. A devres can be released in several ways. No
53 matter what, all devres entries are released on driver detach. On
54 release, the associated release function is invoked and then the
55 devres entry is freed.
57 Managed interface is created for resources commonly used by device
58 drivers using devres. For example, coherent DMA memory is acquired
59 using dma_alloc_coherent(). The managed version is called
60 dmam_alloc_coherent(). It is identical to dma_alloc_coherent() except
61 for the DMA memory allocated using it is managed and will be
62 automatically released on driver detach. Implementation looks like
68 dma_addr_t dma_handle;
71 static void dmam_coherent_release(struct device *dev, void *res)
73 struct dma_devres *this = res;
75 dma_free_coherent(dev, this->size, this->vaddr, this->dma_handle);
78 dmam_alloc_coherent(dev, size, dma_handle, gfp)
80 struct dma_devres *dr;
83 dr = devres_alloc(dmam_coherent_release, sizeof(*dr), gfp);
86 /* alloc DMA memory as usual */
87 vaddr = dma_alloc_coherent(...);
90 /* record size, vaddr, dma_handle in dr */
99 If a driver uses dmam_alloc_coherent(), the area is guaranteed to be
100 freed whether initialization fails half-way or the device gets
101 detached. If most resources are acquired using managed interface, a
102 driver can have much simpler init and exit code. Init path basically
103 looks like the following::
109 d = devm_kzalloc(dev, sizeof(*d), GFP_KERNEL);
113 d->ring = dmam_alloc_coherent(...);
121 return register_to_upper_layer(d);
128 unregister_from_upper_layer(d);
129 shutdown_my_hardware();
132 As shown above, low level drivers can be simplified a lot by using
133 devres. Complexity is shifted from less maintained low level drivers
134 to better maintained higher layer. Also, as init failure path is
135 shared with exit path, both can get more testing.
137 Note though that when converting current calls or assignments to
138 managed devm_* versions it is up to you to check if internal operations
139 like allocating memory, have failed. Managed resources pertains to the
140 freeing of these resources *only* - all other checks needed are still
141 on you. In some cases this may mean introducing checks that were not
142 necessary before moving to the managed devm_* calls.
148 Devres entries can be grouped using devres group. When a group is
149 released, all contained normal devres entries and properly nested
150 groups are released. One usage is to rollback series of acquired
151 resources on failure. For example::
153 if (!devres_open_group(dev, NULL, GFP_KERNEL))
165 devres_remove_group(dev, NULL);
169 devres_release_group(dev, NULL);
172 As resource acquisition failure usually means probe failure, constructs
173 like above are usually useful in midlayer driver (e.g. libata core
174 layer) where interface function shouldn't have side effect on failure.
175 For LLDs, just returning error code suffices in most cases.
177 Each group is identified by `void *id`. It can either be explicitly
178 specified by @id argument to devres_open_group() or automatically
179 created by passing NULL as @id as in the above example. In both
180 cases, devres_open_group() returns the group's id. The returned id
181 can be passed to other devres functions to select the target group.
182 If NULL is given to those functions, the latest open group is
185 For example, you can do something like the following::
187 int my_midlayer_create_something()
189 if (!devres_open_group(dev, my_midlayer_create_something, GFP_KERNEL))
194 devres_close_group(dev, my_midlayer_create_something);
198 void my_midlayer_destroy_something()
200 devres_release_group(dev, my_midlayer_create_something);
207 Lifetime of a devres entry begins on devres allocation and finishes
208 when it is released or destroyed (removed and freed) - no reference
211 devres core guarantees atomicity to all basic devres operations and
212 has support for single-instance devres types (atomic
213 lookup-and-add-if-not-found). Other than that, synchronizing
214 concurrent accesses to allocated devres data is caller's
215 responsibility. This is usually non-issue because bus ops and
216 resource allocations already do the job.
218 For an example of single-instance devres type, read pcim_iomap_table()
221 All devres interface functions can be called without context if the
222 right gfp mask is given.
228 Each devres bookkeeping info is allocated together with requested data
229 area. With debug option turned off, bookkeeping info occupies 16
230 bytes on 32bit machines and 24 bytes on 64bit (three pointers rounded
231 up to ull alignment). If singly linked list is used, it can be
232 reduced to two pointers (8 bytes on 32bit, 16 bytes on 64bit).
234 Each devres group occupies 8 pointers. It can be reduced to 6 if
235 singly linked list is used.
237 Memory space overhead on ahci controller with two ports is between 300
238 and 400 bytes on 32bit machine after naive conversion (we can
239 certainly invest a bit more effort into libata core layer).
242 6. List of managed interfaces
243 -----------------------------
247 devm_clk_get_optional()
250 devm_clk_bulk_get_all()
251 devm_clk_bulk_get_optional()
252 devm_get_clk_from_childl()
253 devm_clk_hw_register()
254 devm_of_clk_add_hw_provider()
255 devm_clk_hw_register_clkdev()
258 dmaenginem_async_device_register()
259 dmam_alloc_coherent()
270 devm_gpiod_get_array()
271 devm_gpiod_get_array_optional()
272 devm_gpiod_get_index()
273 devm_gpiod_get_index_optional()
274 devm_gpiod_get_optional()
277 devm_gpiochip_add_data()
279 devm_gpio_request_one()
283 devm_i2c_new_dummy_device()
286 devm_iio_device_alloc()
287 devm_iio_device_free()
288 devm_iio_device_register()
289 devm_iio_device_unregister()
290 devm_iio_kfifo_allocate()
291 devm_iio_kfifo_free()
292 devm_iio_triggered_buffer_setup()
293 devm_iio_triggered_buffer_cleanup()
294 devm_iio_trigger_alloc()
295 devm_iio_trigger_free()
296 devm_iio_trigger_register()
297 devm_iio_trigger_unregister()
298 devm_iio_channel_get()
299 devm_iio_channel_release()
300 devm_iio_channel_get_all()
301 devm_iio_channel_release_all()
304 devm_input_allocate_device()
307 devm_release_mem_region()
308 devm_release_region()
309 devm_release_resource()
310 devm_request_mem_region()
311 devm_request_region()
312 devm_request_resource()
320 devm_ioremap_resource() : checks resource, requests memory region, ioremaps
321 devm_ioremap_resource_wc()
322 devm_platform_ioremap_resource() : calls devm_ioremap_resource() for platform device
323 devm_platform_ioremap_resource_wc()
324 devm_platform_ioremap_resource_byname()
327 pcim_iomap_regions() : do request_region() and iomap() on multiple BARs
328 pcim_iomap_table() : array of mapped addresses indexed by BAR
333 devm_request_any_context_irq()
335 devm_request_threaded_irq()
336 devm_irq_alloc_descs()
337 devm_irq_alloc_desc()
338 devm_irq_alloc_desc_at()
339 devm_irq_alloc_desc_from()
340 devm_irq_alloc_descs_from()
341 devm_irq_alloc_generic_chip()
342 devm_irq_setup_generic_chip()
346 devm_led_classdev_register()
347 devm_led_classdev_unregister()
351 devm_mdiobus_alloc_size()
356 devm_get_free_pages()
368 devm_mfd_add_devices()
371 devm_mux_chip_alloc()
372 devm_mux_chip_register()
373 devm_mux_control_get()
380 devm_pci_alloc_host_bridge() : managed PCI host bridge allocation
381 devm_pci_remap_cfgspace() : ioremap PCI configuration space
382 devm_pci_remap_cfg_resource() : ioremap PCI configuration space resource
383 pcim_enable_device() : after success, all PCI ops become managed
384 pcim_pin_device() : keep PCI device enabled after release
393 devm_pinctrl_register()
394 devm_pinctrl_unregister()
397 devm_reboot_mode_register()
398 devm_reboot_mode_unregister()
405 devm_regulator_bulk_get()
408 devm_regulator_register()
411 devm_reset_control_get()
412 devm_reset_controller_register()
415 devm_serdev_device_open()
418 devm_acpi_dma_controller_register()
421 devm_spi_register_master()
424 devm_watchdog_register_device()