| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Currently the function pwm_imx27_get_state() of enables the clocks once
unconditionally at the start, a second time if the PWM is enabled and
disables unconditionally at the end.
Simplify that to enable once at the start and disable conditionally at
the end.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
The rcar_pwm_get_clock_division() has a loop to calculate the divider,
but the value of div should be calculatable without a loop. So, this
patch improves it.
This algorithm is suggested by Uwe Kleine-König and Laurent Pinchart.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
This patch removes legacy APIs. Since rcar_pwm_{en,dis}able() functions
are reused on "atomic" API, this patch changes the arguments of these
functions. No change in behavior.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
To remove legacy API related functions in the future, this patch
uses "atomic" related function instead. No change in behavior.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
This patch adds support for "atomic" API. This behavior differs with
legacy APIs a little.
Legacy APIs:
The PWMCNT register will be updated in rcar_pwm_config() even if
the PWM state is disabled.
Atomic API:
The PWMCNT register will be updated in rcar_pwm_apply() only if
the PWM state is enabled. Otherwize, if a PWM runs with 30% duty
cycles and the pwm_apply_state() is called with state->enabled = 0,
->duty_cycle = 60 and ->period = 100, this is possible to output
a 60% duty cycle.
Signed-off-by: Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Add support for SAM9X60's PWM controller.
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Add PWM binding for SAM9X60 SoC.
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Rename objects of type atmel_pwm_data to contain chip name instead of
version number.
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
SAM9X60's PWM controller uses 32 bits counters thus it could generate
signals with higher period and duty cycles than the old ones. Prepare
the current driver to be able to work with old controllers (that uses
16 bits counters) and with the new SAM9X60's controller, by providing
counters information based on compatible string.
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Add struct atmel_pwm_data to embed different per controller information.
It prepares adding support for another similar controller that needs
additional information. At this stage, embed a member of type struct
atmel_pwm_registers in it.
Signed-off-by: Claudiu Beznea <claudiu.beznea@microchip.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Use the mtk_pwm_data struction to define different registers
and add MT8183 specific register operations, such as MT8183
doesn't have commit register, needs to disable double buffer
before writing register, and needs to select commit mode
and use PWM_PERIOD/PWM_HIGH_WIDTH.
Signed-off-by: Jitao Shi <jitao.shi@mediatek.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Add support for the hi3559v100-shub-pwm and hisilicon,hi3559v100-pwm
platforms. They require a special quirk: the PWM has to be enabled twice
to force a duty_cycle refresh.
Signed-off-by: Mathieu Othacehe <m.othacehe@gmail.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Add support for hi3559v100-shub-pwm and hisilicon,hi3559v100-pwm
platforms.
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Mathieu Othacehe <m.othacehe@gmail.com>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Split pwm_soc array in one struct per SoC and point to the corresponding
one in of-data.
Signed-off-by: Mathieu Othacehe <m.othacehe@gmail.com>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
"ret" only holds zero and negative error codes. It needs to be signed
for the error handling to work.
Fixes: 9f4c8f9607c3 ("pwm: imx: Add ipg clock operation")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
The two PWM implementations called v1 (for i.MX1 and i.MX21) and v2 (for
i.MX27 and later) have nothing in common apart from needing two clocks
named "per" and "ipg" and being integrated in a SoC named i.MX.
So split the file containing the two disjunct drivers into two files and
two complete separate drivers.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
[thierry.reding@gmail.com: fix a modular build issue]
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
When getting the peripheral clock fails with -EPROBE_DEFER the driver is
usually probed again later and emitting an error message is irritating.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
When ->probe() fails the driver core takes care of unsetting the driver
data.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
The ->remove() callback is only called when probe returned successfully.
In this case the driver data cannot be NULL.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
In pwm_ops there are a few callbacks that are not supposed to be used by
new drivers. Group them at the end of the structure and add a comment.
Similarily for struct pwm_chip group the members that drivers shouldn't
care about at the end and mark them as internal with another comment.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
This callback was introduced in commit 62099abf67a2 ("pwm: Add debugfs
interface") in 2012 and up to now there is not a single user. So drop
this unused code.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
[thierry.reding@gmail.com: remove kerneldoc for ->dbg_show()]
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
Given that struct pwm_state is sparse (at least on some platforms),
variables of this type might represent the same state because all fields
are pairwise identical but still memcmp() returns a difference because
some of the unused bits are different.
To prevent surprises compare member by member instead of the whole
occupied memory.
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Thierry Reding <thierry.reding@gmail.com>
|
|
|
|
The semantics of what "in core" means for the mincore() system call are
somewhat unclear, but Linux has always (since 2.3.52, which is when
mincore() was initially done) treated it as "page is available in page
cache" rather than "page is mapped in the mapping".
The problem with that traditional semantic is that it exposes a lot of
system cache state that it really probably shouldn't, and that users
shouldn't really even care about.
So let's try to avoid that information leak by simply changing the
semantics to be that mincore() counts actual mapped pages, not pages
that might be cheaply mapped if they were faulted (note the "might be"
part of the old semantics: being in the cache doesn't actually guarantee
that you can access them without IO anyway, since things like network
filesystems may have to revalidate the cache before use).
In many ways the old semantics were somewhat insane even aside from the
information leak issue. From the very beginning (and that beginning is
a long time ago: 2.3.52 was released in March 2000, I think), the code
had a comment saying
Later we can get more picky about what "in core" means precisely.
and this is that "later". Admittedly it is much later than is really
comfortable.
NOTE! This is a real semantic change, and it is for example known to
change the output of "fincore", since that program literally does a
mmmap without populating it, and then doing "mincore()" on that mapping
that doesn't actually have any pages in it.
I'm hoping that nobody actually has any workflow that cares, and the
info leak is real.
We may have to do something different if it turns out that people have
valid reasons to want the old semantics, and if we can limit the
information leak sanely.
Cc: Kevin Easton <kevin@guarana.org>
Cc: Jiri Kosina <jikos@kernel.org>
Cc: Masatake YAMATO <yamato@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Commit 594cc251fdd0 ("make 'user_access_begin()' do 'access_ok()'")
broke both alpha and SH booting in qemu, as noticed by Guenter Roeck.
It turns out that the bug wasn't actually in that commit itself (which
would have been surprising: it was mostly a no-op), but in how the
addition of access_ok() to the strncpy_from_user() and strnlen_user()
functions now triggered the case where those functions would test the
access of the very last byte of the user address space.
The string functions actually did that user range test before too, but
they did it manually by just comparing against user_addr_max(). But
with user_access_begin() doing the check (using "access_ok()"), it now
exposed problems in the architecture implementations of that function.
For example, on alpha, the access_ok() helper macro looked like this:
#define __access_ok(addr, size) \
((get_fs().seg & (addr | size | (addr+size))) == 0)
and what it basically tests is of any of the high bits get set (the
USER_DS masking value is 0xfffffc0000000000).
And that's completely wrong for the "addr+size" check. Because it's
off-by-one for the case where we check to the very end of the user
address space, which is exactly what the strn*_user() functions do.
Why? Because "addr+size" will be exactly the size of the address space,
so trying to access the last byte of the user address space will fail
the __access_ok() check, even though it shouldn't. As a result, the
user string accessor functions failed consistently - because they
literally don't know how long the string is going to be, and the max
access is going to be that last byte of the user address space.
Side note: that alpha macro is buggy for another reason too - it re-uses
the arguments twice.
And SH has another version of almost the exact same bug:
#define __addr_ok(addr) \
((unsigned long __force)(addr) < current_thread_info()->addr_limit.seg)
so far so good: yes, a user address must be below the limit. But then:
#define __access_ok(addr, size) \
(__addr_ok((addr) + (size)))
is wrong with the exact same off-by-one case: the case when "addr+size"
is exactly _equal_ to the limit is actually perfectly fine (think "one
byte access at the last address of the user address space")
The SH version is actually seriously buggy in another way: it doesn't
actually check for overflow, even though it did copy the _comment_ that
talks about overflow.
So it turns out that both SH and alpha actually have completely buggy
implementations of access_ok(), but they happened to work in practice
(although the SH overflow one is a serious serious security bug, not
that anybody likely cares about SH security).
This fixes the problems by using a similar macro on both alpha and SH.
It isn't trying to be clever, the end address is based on this logic:
unsigned long __ao_end = __ao_a + __ao_b - !!__ao_b;
which basically says "add start and length, and then subtract one unless
the length was zero". We can't subtract one for a zero length, or we'd
just hit an underflow instead.
For a lot of access_ok() users the length is a constant, so this isn't
actually as expensive as it initially looks.
Reported-and-tested-by: Guenter Roeck <linux@roeck-us.net>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Add support for the Adiantum encryption mode to fscrypt. Adiantum is a
tweakable, length-preserving encryption mode with security provably
reducible to that of XChaCha12 and AES-256, subject to a security bound.
It's also a true wide-block mode, unlike XTS. See the paper
"Adiantum: length-preserving encryption for entry-level processors"
(https://eprint.iacr.org/2018/720.pdf) for more details. Also see
commit 059c2a4d8e16 ("crypto: adiantum - add Adiantum support").
On sufficiently long messages, Adiantum's bottlenecks are XChaCha12 and
the NH hash function. These algorithms are fast even on processors
without dedicated crypto instructions. Adiantum makes it feasible to
enable storage encryption on low-end mobile devices that lack AES
instructions; currently such devices are unencrypted. On ARM Cortex-A7,
on 4096-byte messages Adiantum encryption is about 4 times faster than
AES-256-XTS encryption; decryption is about 5 times faster.
In fscrypt, Adiantum is suitable for encrypting both file contents and
names. With filenames, it fixes a known weakness: when two filenames in
a directory share a common prefix of >= 16 bytes, with CTS-CBC their
encrypted filenames share a common prefix too, leaking information.
Adiantum does not have this problem.
Since Adiantum also accepts long tweaks (IVs), it's also safe to use the
master key directly for Adiantum encryption rather than deriving
per-file keys, provided that the per-file nonce is included in the IVs
and the master key isn't used for any other encryption mode. This
configuration saves memory and improves performance. A new fscrypt
policy flag is added to allow users to opt-in to this configuration.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
|
|
Remove the dot-prefixing since it is just a matter of the
.gitignore file.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Make simply skips a missing rule when it is marked as .PHONY.
Remove the dummy targets.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
You do not have to use define ... endef for filechk_* rules.
For simple cases, the use of assignment looks cleaner, IMHO.
I updated the usage for scripts/Kbuild.include in case somebody
misunderstands the 'define ... endif' is the requirement.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com>
|
|
Now that Kbuild automatically creates asm-generic wrappers for missing
mandatory headers, it is redundant to list the same headers in
generic-y and mandatory-y.
Suggested-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
|
|
Some time ago, Sam pointed out a certain degree of overwrap between
generic-y and mandatory-y. (https://lkml.org/lkml/2017/7/10/121)
I tweaked the meaning of mandatory-y a little bit; now it defines the
minimum set of ASM headers that all architectures must have.
If arch does not have specific implementation of a mandatory header,
Kbuild will let it fallback to the asm-generic one by automatically
generating a wrapper. This will allow to drop lots of redundant
generic-y defines.
Previously, "mandatory" was used in the context of UAPI, but I guess
this can be extended to kernel space ASM headers.
Suggested-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Sam Ravnborg <sam@ravnborg.org>
|
|
These comments are leftovers of commit fcc8487d477a ("uapi: export all
headers under uapi directories").
Prior to that commit, exported headers must be explicitly added to
header-y. Now, all headers under the uapi/ directories are exported.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
This commit removes redundant generic-y defines in
arch/riscv/include/asm/Kbuild.
[1] It is redundant to define the same generic-y in both
arch/$(ARCH)/include/asm/Kbuild and
arch/$(ARCH)/include/uapi/asm/Kbuild.
Remove the following generic-y:
errno.h
fcntl.h
ioctl.h
ioctls.h
ipcbuf.h
mman.h
msgbuf.h
param.h
poll.h
posix_types.h
resource.h
sembuf.h
setup.h
shmbuf.h
signal.h
socket.h
sockios.h
stat.h
statfs.h
swab.h
termbits.h
termios.h
types.h
[2] It is redundant to define generic-y when arch-specific
implementation exists in arch/$(ARCH)/include/asm/*.h
Remove the following generic-y:
cacheflush.h
module.h
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
filechk_* rules often consist of multiple 'echo' lines. They must be
surrounded with { } or ( ) to work correctly. Otherwise, only the
string from the last 'echo' would be written into the target.
Let's take care of that in the 'filechk' in scripts/Kbuild.include
to clean up filechk_* rules.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Since commit 9c2af1c7377a ("kbuild: add .DELETE_ON_ERROR special
target"), the target file is automatically deleted on failure.
The boilerplate code
... || { rm -f $@; false; }
is unneeded.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Commit 3a2429e1faf4 ("kbuild: change if_changed_rule for multi-line
recipe") and commit 4f0e3a57d6eb ("kbuild: Add support for DT binding
schema checks") came in via different sub-systems.
This is a follow-up cleanup.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
The only/last user of UIMAGE_IN/OUT was removed by commit 4722a3e6b716
("microblaze: fix multiple bugs in arch/microblaze/boot/Makefile").
The input and output should always be $< and $@.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Currently, CONFIG_JUMP_LABEL just means "I _want_ to use jump label".
The jump label is controlled by HAVE_JUMP_LABEL, which is defined
like this:
#if defined(CC_HAVE_ASM_GOTO) && defined(CONFIG_JUMP_LABEL)
# define HAVE_JUMP_LABEL
#endif
We can improve this by testing 'asm goto' support in Kconfig, then
make JUMP_LABEL depend on CC_HAS_ASM_GOTO.
Ugly #ifdef HAVE_JUMP_LABEL will go away, and CONFIG_JUMP_LABEL will
match to the real kernel capability.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
|
|
As mentioned in the info pages of gas, the '.align' pseudo op's
interpretation of the alignment value is architecture specific.
It might either be a byte value or taken to the power of two.
On ARM it's actually the latter which leads to unnecessary large
alignments of 16 bytes for 32 bit builds or 256 bytes for 64 bit
builds.
Fix this by switching to '.balign' instead which is consistent
across all architectures.
Signed-off-by: Mathias Krause <minipli@googlemail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Coccinelle doesn't always have access to the values of named
(#define) constants, and they may likely often be bound to true
and false values anyway, resulting in false positives. So stop
warning about them.
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
Avoid reporting on the use of an iterator index variable when
the variable is redeclared.
Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
This has never been used.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
This commit removes redundant generic-y defines in
arch/nds32/include/asm/Kbuild.
[1] It is redundant to define the same generic-y in both
arch/$(ARCH)/include/asm/Kbuild and
arch/$(ARCH)/include/uapi/asm/Kbuild.
Remove the following generic-y:
bitsperlong.h
bpf_perf_event.h
errno.h
fcntl.h
ioctl.h
ioctls.h
mman.h
shmbuf.h
stat.h
[2] It is redundant to define generic-y when arch-specific
implementation exists in arch/$(ARCH)/include/asm/*.h
Remove the following generic-y:
ftrace.h
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
|
|
kernel/dma/Kconfig globally defines HAS_DMA as follows:
config HAS_DMA
bool
depends on !NO_DMA
default y
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
|
|
Fixes build break on most ARM/ARM64 defconfigs:
lib/genalloc.c: In function 'gen_pool_add_virt':
lib/genalloc.c:190:10: error: implicit declaration of function 'vzalloc_node'; did you mean 'kzalloc_node'?
lib/genalloc.c:190:8: warning: assignment to 'struct gen_pool_chunk *' from 'int' makes pointer from integer without a cast [-Wint-conversion]
lib/genalloc.c: In function 'gen_pool_destroy':
lib/genalloc.c:254:3: error: implicit declaration of function 'vfree'; did you mean 'kfree'?
Fixes: 6862d2fc8185 ('lib/genalloc.c: use vzalloc_node() to allocate the bitmap')
Cc: Huang Shijie <sjhuang@iluvatar.ai>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Alexey Skidanov <alexey.skidanov@intel.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
We need to return a dma_addr_t even if we don't have a kernel mapping.
Do so by consolidating the phys_to_dma call in a single place and jump
to it from all the branches that return successfully.
Fixes: bfd56cd60521 ("dma-mapping: support highmem in the generic remap allocator")
Reported-by: Liviu Dudau <liviu@dudau.co.uk
Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Liviu Dudau <liviu@dudau.co.uk>
|
|
In many cases we don't have to create a GART mapping at all, which
also means there is nothing to unmap. Fix the range check that was
incorrectly modified when removing the mapping_error method.
Fixes: 9e8aa6b546 ("x86/amd_gart: remove the mapping_error dma_map_ops method")
Reported-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Tested-by: Michal Kubecek <mkubecek@suse.cz>
|
|
Some non-generic ia64 configs don't build swiotlb, and thus should not
pull in the generic non-coherent DMA infrastructure.
Fixes: 68c608345c ("swiotlb: remove dma_mark_clean")
Reported-by: Tony Luck <tony.luck@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
This has been broken forever, and nobody ever really noticed because
it's purely a performance issue.
Long long ago, in commit 6175ddf06b61 ("x86: Clean up mem*io functions")
Brian Gerst simplified the memory copies to and from iomem, since on
x86, the instructions to access iomem are exactly the same as the
regular instructions.
That is technically true, and things worked, and nobody said anything.
Besides, back then the regular memcpy was pretty simple and worked fine.
Nobody noticed except for David Laight, that is. David has a testing a
TLP monitor he was writing for an FPGA, and has been occasionally
complaining about how memcpy_toio() writes things one byte at a time.
Which is completely unacceptable from a performance standpoint, even if
it happens to technically work.
The reason it's writing one byte at a time is because while it's
technically true that accesses to iomem are the same as accesses to
regular memory on x86, the _granularity_ (and ordering) of accesses
matter to iomem in ways that they don't matter to regular cached memory.
In particular, when ERMS is set, we default to using "rep movsb" for
larger memory copies. That is indeed perfectly fine for real memory,
since the whole point is that the CPU is going to do cacheline
optimizations and executes the memory copy efficiently for cached
memory.
With iomem? Not so much. With iomem, "rep movsb" will indeed work, but
it will copy things one byte at a time. Slowly and ponderously.
Now, originally, back in 2010 when commit 6175ddf06b61 was done, we
didn't use ERMS, and this was much less noticeable.
Our normal memcpy() was simpler in other ways too.
Because in fact, it's not just about using the string instructions. Our
memcpy() these days does things like "read and write overlapping values"
to handle the last bytes of the copy. Again, for normal memory,
overlapping accesses isn't an issue. For iomem? It can be.
So this re-introduces the specialized memcpy_toio(), memcpy_fromio() and
memset_io() functions. It doesn't particularly optimize them, but it
tries to at least not be horrid, or do overlapping accesses. In fact,
this uses the existing __inline_memcpy() function that we still had
lying around that uses our very traditional "rep movsl" loop followed by
movsw/movsb for the final bytes.
Somebody may decide to try to improve on it, but if we've gone almost a
decade with only one person really ever noticing and complaining, maybe
it's not worth worrying about further, once it's not _completely_ broken?
Reported-by: David Laight <David.Laight@aculab.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
This actually enables the __put_user_goto() functionality in
unsafe_put_user().
For an example of the effect of this, this is the code generated for the
unsafe_put_user(signo, &infop->si_signo, Efault);
in the waitid() system call:
movl %ecx,(%rbx) # signo, MEM[(struct __large_struct *)_2]
It's just one single store instruction, along with generating an
exception table entry pointing to the Efault label case in case that
instruction faults.
Before, we would generate this:
xorl %edx, %edx
movl %ecx,(%rbx) # signo, MEM[(struct __large_struct *)_3]
testl %edx, %edx
jne .L309
with the exception table generated for that 'mov' instruction causing us
to jump to a stub that set %edx to -EFAULT and then jumped back to the
'testl' instruction.
So not only do we now get rid of the extra code in the normal sequence,
we also avoid unnecessarily keeping that extra error register live
across it all.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
Uwe Kleine-König
Yoshihiro Shimoda
Claudiu Beznea
Jitao Shi
Mathieu Othacehe
Dan Carpenter
Linus Torvalds
Eric Biggers
Masahiro Yamada
Mathias Krause
Julia Lawall
Olof Johansson
Christoph Hellwig