Root/Documentation/feature-removal-schedule.txt

1The following is a list of files and features that are going to be
2removed in the kernel source tree. Every entry should contain what
3exactly is going away, why it is happening, and who is going to be doing
4the work. When the feature is removed from the kernel, it should also
5be removed from this file.
6
7---------------------------
8
9What: PRISM54
10When: 2.6.34
11
12Why: prism54 FullMAC PCI / Cardbus devices used to be supported only by the
13    prism54 wireless driver. After Intersil stopped selling these
14    devices in preference for the newer more flexible SoftMAC devices
15    a SoftMAC device driver was required and prism54 did not support
16    them. The p54pci driver now exists and has been present in the kernel for
17    a while. This driver supports both SoftMAC devices and FullMAC devices.
18    The main difference between these devices was the amount of memory which
19    could be used for the firmware. The SoftMAC devices support a smaller
20    amount of memory. Because of this the SoftMAC firmware fits into FullMAC
21    devices's memory. p54pci supports not only PCI / Cardbus but also USB
22    and SPI. Since p54pci supports all devices prism54 supports
23    you will have a conflict. I'm not quite sure how distributions are
24    handling this conflict right now. prism54 was kept around due to
25    claims users may experience issues when using the SoftMAC driver.
26    Time has passed users have not reported issues. If you use prism54
27    and for whatever reason you cannot use p54pci please let us know!
28    E-mail us at: linux-wireless@vger.kernel.org
29
30    For more information see the p54 wiki page:
31
32    http://wireless.kernel.org/en/users/Drivers/p54
33
34Who: Luis R. Rodriguez <lrodriguez@atheros.com>
35
36---------------------------
37
38What: IRQF_SAMPLE_RANDOM
39Check: IRQF_SAMPLE_RANDOM
40When: July 2009
41
42Why: Many of IRQF_SAMPLE_RANDOM users are technically bogus as entropy
43    sources in the kernel's current entropy model. To resolve this, every
44    input point to the kernel's entropy pool needs to better document the
45    type of entropy source it actually is. This will be replaced with
46    additional add_*_randomness functions in drivers/char/random.c
47
48Who: Robin Getz <rgetz@blackfin.uclinux.org> & Matt Mackall <mpm@selenic.com>
49
50---------------------------
51
52What: Deprecated snapshot ioctls
53When: 2.6.36
54
55Why: The ioctls in kernel/power/user.c were marked as deprecated long time
56    ago. Now they notify users about that so that they need to replace
57    their userspace. After some more time, remove them completely.
58
59Who: Jiri Slaby <jirislaby@gmail.com>
60
61---------------------------
62
63What: The ieee80211_regdom module parameter
64When: March 2010 / desktop catchup
65
66Why: This was inherited by the CONFIG_WIRELESS_OLD_REGULATORY code,
67    and currently serves as an option for users to define an
68    ISO / IEC 3166 alpha2 code for the country they are currently
69    present in. Although there are userspace API replacements for this
70    through nl80211 distributions haven't yet caught up with implementing
71    decent alternatives through standard GUIs. Although available as an
72    option through iw or wpa_supplicant its just a matter of time before
73    distributions pick up good GUI options for this. The ideal solution
74    would actually consist of intelligent designs which would do this for
75    the user automatically even when travelling through different countries.
76    Until then we leave this module parameter as a compromise.
77
78    When userspace improves with reasonable widely-available alternatives for
79    this we will no longer need this module parameter. This entry hopes that
80    by the super-futuristically looking date of "March 2010" we will have
81    such replacements widely available.
82
83Who: Luis R. Rodriguez <lrodriguez@atheros.com>
84
85---------------------------
86
87What: dev->power.power_state
88When: July 2007
89Why: Broken design for runtime control over driver power states, confusing
90    driver-internal runtime power management with: mechanisms to support
91    system-wide sleep state transitions; event codes that distinguish
92    different phases of swsusp "sleep" transitions; and userspace policy
93    inputs. This framework was never widely used, and most attempts to
94    use it were broken. Drivers should instead be exposing domain-specific
95    interfaces either to kernel or to userspace.
96Who: Pavel Machek <pavel@ucw.cz>
97
98---------------------------
99
100What: Video4Linux API 1 ioctls and from Video devices.
101When: kernel 2.6.38
102Files: include/linux/videodev.h
103Check: include/linux/videodev.h
104Why: V4L1 AP1 was replaced by V4L2 API during migration from 2.4 to 2.6
105    series. The old API have lots of drawbacks and don't provide enough
106    means to work with all video and audio standards. The newer API is
107    already available on the main drivers and should be used instead.
108    Newer drivers should use v4l_compat_translate_ioctl function to handle
109    old calls, replacing to newer ones.
110    Decoder iocts are using internally to allow video drivers to
111    communicate with video decoders. This should also be improved to allow
112    V4L2 calls being translated into compatible internal ioctls.
113    Compatibility ioctls will be provided, for a while, via
114    v4l1-compat module.
115Who: Mauro Carvalho Chehab <mchehab@infradead.org>
116
117---------------------------
118
119What: Video4Linux obsolete drivers using V4L1 API
120When: kernel 2.6.38
121Files: drivers/staging/cpia/* drivers/staging/stradis/*
122Check: drivers/staging/cpia/cpia.c drivers/staging/stradis/stradis.c
123Why: There are some drivers still using V4L1 API, despite all efforts we've done
124    to migrate. Those drivers are for obsolete hardware that the old maintainer
125    didn't care (or not have the hardware anymore), and that no other developer
126    could find any hardware to buy. They probably have no practical usage today,
127    and people with such old hardware could probably keep using an older version
128    of the kernel. Those drivers will be moved to staging on 2.6.37 and, if nobody
129    care enough to port and test them with V4L2 API, they'll be removed on 2.6.38.
130Who: Mauro Carvalho Chehab <mchehab@infradead.org>
131
132---------------------------
133
134What: sys_sysctl
135When: September 2010
136Option: CONFIG_SYSCTL_SYSCALL
137Why: The same information is available in a more convenient from
138    /proc/sys, and none of the sysctl variables appear to be
139    important performance wise.
140
141    Binary sysctls are a long standing source of subtle kernel
142    bugs and security issues.
143
144    When I looked several months ago all I could find after
145    searching several distributions were 5 user space programs and
146    glibc (which falls back to /proc/sys) using this syscall.
147
148    The man page for sysctl(2) documents it as unusable for user
149    space programs.
150
151    sysctl(2) is not generally ABI compatible to a 32bit user
152    space application on a 64bit and a 32bit kernel.
153
154    For the last several months the policy has been no new binary
155    sysctls and no one has put forward an argument to use them.
156
157    Binary sysctls issues seem to keep happening appearing so
158    properly deprecating them (with a warning to user space) and a
159    2 year grace warning period will mean eventually we can kill
160    them and end the pain.
161
162    In the mean time individual binary sysctls can be dealt with
163    in a piecewise fashion.
164
165Who: Eric Biederman <ebiederm@xmission.com>
166
167---------------------------
168
169What: /proc/<pid>/oom_adj
170When: August 2012
171Why: /proc/<pid>/oom_adj allows userspace to influence the oom killer's
172    badness heuristic used to determine which task to kill when the kernel
173    is out of memory.
174
175    The badness heuristic has since been rewritten since the introduction of
176    this tunable such that its meaning is deprecated. The value was
177    implemented as a bitshift on a score generated by the badness()
178    function that did not have any precise units of measure. With the
179    rewrite, the score is given as a proportion of available memory to the
180    task allocating pages, so using a bitshift which grows the score
181    exponentially is, thus, impossible to tune with fine granularity.
182
183    A much more powerful interface, /proc/<pid>/oom_score_adj, was
184    introduced with the oom killer rewrite that allows users to increase or
185    decrease the badness() score linearly. This interface will replace
186    /proc/<pid>/oom_adj.
187
188    A warning will be emitted to the kernel log if an application uses this
189    deprecated interface. After it is printed once, future warnings will be
190    suppressed until the kernel is rebooted.
191
192---------------------------
193
194What: remove EXPORT_SYMBOL(kernel_thread)
195When: August 2006
196Files: arch/*/kernel/*_ksyms.c
197Check: kernel_thread
198Why: kernel_thread is a low-level implementation detail. Drivers should
199        use the <linux/kthread.h> API instead which shields them from
200    implementation details and provides a higherlevel interface that
201    prevents bugs and code duplication
202Who: Christoph Hellwig <hch@lst.de>
203
204---------------------------
205
206What: Unused EXPORT_SYMBOL/EXPORT_SYMBOL_GPL exports
207    (temporary transition config option provided until then)
208    The transition config option will also be removed at the same time.
209When: before 2.6.19
210Why: Unused symbols are both increasing the size of the kernel binary
211    and are often a sign of "wrong API"
212Who: Arjan van de Ven <arjan@linux.intel.com>
213
214---------------------------
215
216What: PHYSDEVPATH, PHYSDEVBUS, PHYSDEVDRIVER in the uevent environment
217When: October 2008
218Why: The stacking of class devices makes these values misleading and
219    inconsistent.
220    Class devices should not carry any of these properties, and bus
221    devices have SUBSYTEM and DRIVER as a replacement.
222Who: Kay Sievers <kay.sievers@suse.de>
223
224---------------------------
225
226What: ACPI procfs interface
227When: July 2008
228Why: ACPI sysfs conversion should be finished by January 2008.
229    ACPI procfs interface will be removed in July 2008 so that
230    there is enough time for the user space to catch up.
231Who: Zhang Rui <rui.zhang@intel.com>
232
233---------------------------
234
235What: /proc/acpi/button
236When: August 2007
237Why: /proc/acpi/button has been replaced by events to the input layer
238    since 2.6.20.
239Who: Len Brown <len.brown@intel.com>
240
241---------------------------
242
243What: /proc/acpi/event
244When: February 2008
245Why: /proc/acpi/event has been replaced by events via the input layer
246    and netlink since 2.6.23.
247Who: Len Brown <len.brown@intel.com>
248
249---------------------------
250
251What: i386/x86_64 bzImage symlinks
252When: April 2010
253
254Why: The i386/x86_64 merge provides a symlink to the old bzImage
255    location so not yet updated user space tools, e.g. package
256    scripts, do not break.
257Who: Thomas Gleixner <tglx@linutronix.de>
258
259---------------------------
260
261What: GPIO autorequest on gpio_direction_{input,output}() in gpiolib
262When: February 2010
263Why: All callers should use explicit gpio_request()/gpio_free().
264    The autorequest mechanism in gpiolib was provided mostly as a
265    migration aid for legacy GPIO interfaces (for SOC based GPIOs).
266    Those users have now largely migrated. Platforms implementing
267    the GPIO interfaces without using gpiolib will see no changes.
268Who: David Brownell <dbrownell@users.sourceforge.net>
269---------------------------
270
271What: b43 support for firmware revision < 410
272When: The schedule was July 2008, but it was decided that we are going to keep the
273        code as long as there are no major maintanance headaches.
274    So it _could_ be removed _any_ time now, if it conflicts with something new.
275Why: The support code for the old firmware hurts code readability/maintainability
276    and slightly hurts runtime performance. Bugfixes for the old firmware
277    are not provided by Broadcom anymore.
278Who: Michael Buesch <mb@bu3sch.de>
279
280---------------------------
281
282What: /sys/o2cb symlink
283When: January 2010
284Why: /sys/fs/o2cb is the proper location for this information - /sys/o2cb
285    exists as a symlink for backwards compatibility for old versions of
286    ocfs2-tools. 2 years should be sufficient time to phase in new versions
287    which know to look in /sys/fs/o2cb.
288Who: ocfs2-devel@oss.oracle.com
289
290---------------------------
291
292What: Ability for non root users to shm_get hugetlb pages based on mlock
293    resource limits
294When: 2.6.31
295Why: Non root users need to be part of /proc/sys/vm/hugetlb_shm_group or
296    have CAP_IPC_LOCK to be able to allocate shm segments backed by
297    huge pages. The mlock based rlimit check to allow shm hugetlb is
298    inconsistent with mmap based allocations. Hence it is being
299    deprecated.
300Who: Ravikiran Thirumalai <kiran@scalex86.org>
301
302---------------------------
303
304What: CONFIG_THERMAL_HWMON
305When: January 2009
306Why: This option was introduced just to allow older lm-sensors userspace
307    to keep working over the upgrade to 2.6.26. At the scheduled time of
308    removal fixed lm-sensors (2.x or 3.x) should be readily available.
309Who: Rene Herman <rene.herman@gmail.com>
310
311---------------------------
312
313What: Code that is now under CONFIG_WIRELESS_EXT_SYSFS
314    (in net/core/net-sysfs.c)
315When: After the only user (hal) has seen a release with the patches
316    for enough time, probably some time in 2010.
317Why: Over 1K .text/.data size reduction, data is available in other
318    ways (ioctls)
319Who: Johannes Berg <johannes@sipsolutions.net>
320
321---------------------------
322
323What: sysfs ui for changing p4-clockmod parameters
324When: September 2009
325Why: See commits 129f8ae9b1b5be94517da76009ea956e89104ce8 and
326    e088e4c9cdb618675874becb91b2fd581ee707e6.
327    Removal is subject to fixing any remaining bugs in ACPI which may
328    cause the thermal throttling not to happen at the right time.
329Who: Dave Jones <davej@redhat.com>, Matthew Garrett <mjg@redhat.com>
330
331-----------------------------
332
333What: __do_IRQ all in one fits nothing interrupt handler
334When: 2.6.32
335Why: __do_IRQ was kept for easy migration to the type flow handlers.
336    More than two years of migration time is enough.
337Who: Thomas Gleixner <tglx@linutronix.de>
338
339-----------------------------
340
341What: fakephp and associated sysfs files in /sys/bus/pci/slots/
342When: 2011
343Why: In 2.6.27, the semantics of /sys/bus/pci/slots was redefined to
344    represent a machine's physical PCI slots. The change in semantics
345    had userspace implications, as the hotplug core no longer allowed
346    drivers to create multiple sysfs files per physical slot (required
347    for multi-function devices, e.g.). fakephp was seen as a developer's
348    tool only, and its interface changed. Too late, we learned that
349    there were some users of the fakephp interface.
350
351    In 2.6.30, the original fakephp interface was restored. At the same
352    time, the PCI core gained the ability that fakephp provided, namely
353    function-level hot-remove and hot-add.
354
355    Since the PCI core now provides the same functionality, exposed in:
356
357        /sys/bus/pci/rescan
358        /sys/bus/pci/devices/.../remove
359        /sys/bus/pci/devices/.../rescan
360
361    there is no functional reason to maintain fakephp as well.
362
363    We will keep the existing module so that 'modprobe fakephp' will
364    present the old /sys/bus/pci/slots/... interface for compatibility,
365    but users are urged to migrate their applications to the API above.
366
367    After a reasonable transition period, we will remove the legacy
368    fakephp interface.
369Who: Alex Chiang <achiang@hp.com>
370
371---------------------------
372
373What: CONFIG_RFKILL_INPUT
374When: 2.6.33
375Why: Should be implemented in userspace, policy daemon.
376Who: Johannes Berg <johannes@sipsolutions.net>
377
378----------------------------
379
380What: sound-slot/service-* module aliases and related clutters in
381    sound/sound_core.c
382When: August 2010
383Why: OSS sound_core grabs all legacy minors (0-255) of SOUND_MAJOR
384    (14) and requests modules using custom sound-slot/service-*
385    module aliases. The only benefit of doing this is allowing
386    use of custom module aliases which might as well be considered
387    a bug at this point. This preemptive claiming prevents
388    alternative OSS implementations.
389
390    Till the feature is removed, the kernel will be requesting
391    both sound-slot/service-* and the standard char-major-* module
392    aliases and allow turning off the pre-claiming selectively via
393    CONFIG_SOUND_OSS_CORE_PRECLAIM and soundcore.preclaim_oss
394    kernel parameter.
395
396    After the transition phase is complete, both the custom module
397    aliases and switches to disable it will go away. This removal
398    will also allow making ALSA OSS emulation independent of
399    sound_core. The dependency will be broken then too.
400Who: Tejun Heo <tj@kernel.org>
401
402----------------------------
403
404What: Support for lcd_switch and display_get in asus-laptop driver
405When: March 2010
406Why: These two features use non-standard interfaces. There are the
407    only features that really need multiple path to guess what's
408    the right method name on a specific laptop.
409
410    Removing them will allow to remove a lot of code an significantly
411    clean the drivers.
412
413    This will affect the backlight code which won't be able to know
414    if the backlight is on or off. The platform display file will also be
415    write only (like the one in eeepc-laptop).
416
417    This should'nt affect a lot of user because they usually know
418    when their display is on or off.
419
420Who: Corentin Chary <corentin.chary@gmail.com>
421
422----------------------------
423
424What: sysfs-class-rfkill state file
425When: Feb 2014
426Files: net/rfkill/core.c
427Why: Documented as obsolete since Feb 2010. This file is limited to 3
428    states while the rfkill drivers can have 4 states.
429Who: anybody or Florian Mickler <florian@mickler.org>
430
431----------------------------
432
433What: sysfs-class-rfkill claim file
434When: Feb 2012
435Files: net/rfkill/core.c
436Why: It is not possible to claim an rfkill driver since 2007. This is
437    Documented as obsolete since Feb 2010.
438Who: anybody or Florian Mickler <florian@mickler.org>
439
440----------------------------
441
442What: capifs
443When: February 2011
444Files: drivers/isdn/capi/capifs.*
445Why: udev fully replaces this special file system that only contains CAPI
446    NCCI TTY device nodes. User space (pppdcapiplugin) works without
447    noticing the difference.
448Who: Jan Kiszka <jan.kiszka@web.de>
449
450----------------------------
451
452What: KVM paravirt mmu host support
453When: January 2011
454Why: The paravirt mmu host support is slower than non-paravirt mmu, both
455    on newer and older hardware. It is already not exposed to the guest,
456    and kept only for live migration purposes.
457Who: Avi Kivity <avi@redhat.com>
458
459----------------------------
460
461What: iwlwifi 50XX module parameters
462When: 2.6.40
463Why: The "..50" modules parameters were used to configure 5000 series and
464    up devices; different set of module parameters also available for 4965
465    with same functionalities. Consolidate both set into single place
466    in drivers/net/wireless/iwlwifi/iwl-agn.c
467
468Who: Wey-Yi Guy <wey-yi.w.guy@intel.com>
469
470----------------------------
471
472What: iwl4965 alias support
473When: 2.6.40
474Why: Internal alias support has been present in module-init-tools for some
475    time, the MODULE_ALIAS("iwl4965") boilerplate aliases can be removed
476    with no impact.
477
478Who: Wey-Yi Guy <wey-yi.w.guy@intel.com>
479
480---------------------------
481
482What: xt_NOTRACK
483Files: net/netfilter/xt_NOTRACK.c
484When: April 2011
485Why: Superseded by xt_CT
486Who: Netfilter developer team <netfilter-devel@vger.kernel.org>
487
488----------------------------
489
490What: IRQF_DISABLED
491When: 2.6.36
492Why: The flag is a NOOP as we run interrupt handlers with interrupts disabled
493Who: Thomas Gleixner <tglx@linutronix.de>
494
495----------------------------
496
497What: The acpi_sleep=s4_nonvs command line option
498When: 2.6.37
499Files: arch/x86/kernel/acpi/sleep.c
500Why: superseded by acpi_sleep=nonvs
501Who: Rafael J. Wysocki <rjw@sisk.pl>
502
503----------------------------
504
505What: PCI DMA unmap state API
506When: August 2012
507Why: PCI DMA unmap state API (include/linux/pci-dma.h) was replaced
508    with DMA unmap state API (DMA unmap state API can be used for
509    any bus).
510Who: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
511
512----------------------------
513
514What: DMA_xxBIT_MASK macros
515When: Jun 2011
516Why: DMA_xxBIT_MASK macros were replaced with DMA_BIT_MASK() macros.
517Who: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
518
519----------------------------
520
521What: namespace cgroup (ns_cgroup)
522When: 2.6.38
523Why: The ns_cgroup leads to some problems:
524    * cgroup creation is out-of-control
525    * cgroup name can conflict when pids are looping
526    * it is not possible to have a single process handling
527    a lot of namespaces without falling in a exponential creation time
528    * we may want to create a namespace without creating a cgroup
529
530    The ns_cgroup is replaced by a compatibility flag 'clone_children',
531    where a newly created cgroup will copy the parent cgroup values.
532    The userspace has to manually create a cgroup and add a task to
533    the 'tasks' file.
534Who: Daniel Lezcano <daniel.lezcano@free.fr>
535
536----------------------------
537
538What: iwlwifi disable_hw_scan module parameters
539When: 2.6.40
540Why: Hareware scan is the prefer method for iwlwifi devices for
541    scanning operation. Remove software scan support for all the
542    iwlwifi devices.
543
544Who: Wey-Yi Guy <wey-yi.w.guy@intel.com>
545
546----------------------------
547
548What: access to nfsd auth cache through sys_nfsservctl or '.' files
549        in the 'nfsd' filesystem.
550When: 2.6.40
551Why: This is a legacy interface which have been replaced by a more
552        dynamic cache. Continuing to maintain this interface is an
553        unnecessary burden.
554Who: NeilBrown <neilb@suse.de>
555
556----------------------------
557
558What: i2c_adapter.id
559When: June 2011
560Why: This field is deprecated. I2C device drivers shouldn't change their
561    behavior based on the underlying I2C adapter. Instead, the I2C
562    adapter driver should instantiate the I2C devices and provide the
563    needed platform-specific information.
564Who: Jean Delvare <khali@linux-fr.org>
565
566----------------------------
567

Archive Download this file



interactive