Since the Thumb-2 instructions can be 16-bit wide, data in the .text
sections may not be aligned to a 32-bit word and this leads to unaligned
exceptions. This patch does not affect the ARM code generation.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Previous size thresholds were guessed from various user space benchmarks
using a kernel with and without the alternative uaccess option. This
is however not as precise as a kernel based test to measure the real
speed of each method.
This adds a simple test bench to show the time needed for each method.
With this, the optimal size treshold for the alternative implementation
can be determined with more confidence. It appears that the optimal
threshold for both copy_to_user and clear_user is around 64 bytes. This
is not a surprise knowing that the memcpy and memset implementations
need at least 64 bytes to achieve maximum throughput.
One might suggest that such test be used to determine the optimal
threshold at run time instead, but results are near enough to 64 on
tested targets concerned by this alternative copy_to_user implementation,
so adding some overhead associated with a variable threshold is probably
not worth it for now.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Because the alternate copy_to_user implementation has a higher setup cost
than the standard implementation, the size of the memory area to copy
is tested and the standard implementation invoked instead when that size
is too small. Still, that test is made after the processor has preserved
a bunch of registers on the stack which have to be reloaded right away
needlessly in that case, causing a measurable performance regression
compared to plain usage of the standard implementation only.
To make the size test overhead negligible, let's factorize it out of
the alternate copy_to_user function where it is clear to the compiler
that no stack frame is needed. Thanks to CONFIG_ARM_UNWIND allowing
for frame pointers to be disabled and tail call optimization to kick in,
the overhead in the small copy case becomes only 3 assembly instructions.
A similar trick is applied to clear_user as well.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
This implements {copy_to,clear}_user() by faulting in the userland
pages and then using the regular kernel mem{cpy,set}() to copy the
data (while holding the page table lock). This is a win if the regular
mem{cpy,set}() implementations are faster than the user copy functions,
which is the case e.g. on Feroceon, where 8-word STMs (which memcpy()
uses under the right conditions) give significantly higher memory write
throughput than a sequence of individual 32bit stores.
Here are numbers for page sized buffers on some Feroceon cores:
- copy_to_user on Orion5x goes from 51 MB/s to 83 MB/s
- clear_user on Orion5x goes from 89MB/s to 314MB/s
- copy_to_user on Kirkwood goes from 240 MB/s to 356 MB/s
- clear_user on Kirkwood goes from 367 MB/s to 1108 MB/s
- copy_to_user on Disco-Duo goes from 248 MB/s to 398 MB/s
- clear_user on Disco-Duo goes from 328 MB/s to 1741 MB/s
Because the setup cost is non negligible, this is worthwhile only if
the amount of data to copy is large enough. The operation falls back
to the standard implementation when the amount of data is below a certain
threshold. This threshold was determined empirically, however some targets
could benefit from a lower runtime determined value for optimal results
eventually.
In the copy_from_user() case, this technique does not provide any
worthwhile performance gain due to the fact that any kind of read access
allocates the cache and subsequent 32bit loads are just as fast as the
equivalent 8-word LDM.
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Tested-by: Martin Michlmayr <tbm@cyrius.com>
This allows for optional alternative implementations of __copy_to_user
and __clear_user, with a possible runtime fallback to the standard
version when the alternative provides no gain over that standard
version. This is done by making the standard __copy_to_user into a weak
alias for the symbol __copy_to_user_std. Same thing for __clear_user.
Those two functions are particularly good candidates to have alternative
implementations for, since they rely on the STRT instruction which has
lower performances than STM instructions on some CPU cores such as
the ARM1176 and Marvell Feroceon.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Mathieu Desnoyers pointed out that the ARM barriers were lacking:
- cmpxchg, xchg and atomic add return need memory barriers on
architectures which can reorder the relative order in which memory
read/writes can be seen between CPUs, which seems to include recent
ARM architectures. Those barriers are currently missing on ARM.
- test_and_xxx_bit were missing SMP barriers.
So put these barriers in. Provide separate atomic_add/atomic_sub
operations which do not require barriers.
Reported-Reviewed-and-Acked-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
The CLPS7500 platform has not built since 2.6.22-git7 and there
seems to be no interest in fixing it. So, remove the platform
support.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
As suggested by Andrew Morton, remove memzero() - it's not supported
on other architectures so use of it is a potential build breaking bug.
Since the compiler optimizes memset(x,0,n) to __memzero() perfectly
well, we don't miss out on the underlying benefits of memzero().
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Since the other assembly functions do not seem to save the frame
pointer onto the stack, this patch changes the csum_partial_copy_*
functions to behave in the same way.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
The last strnebt instruction has a post-index of 1 but the address
register is set to 0 in the next instruction, so no need for
post-indexing.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
This declaration specifies the "function" type and size for various
assembly functions, mainly needed for generating the correct branch
instructions in Thumb-2.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Remove includes of asm/hardware.h in addition to asm/arch/hardware.h.
Then, since asm/hardware.h only exists to include asm/arch/hardware.h,
update everything to directly include asm/arch/hardware.h and remove
asm/hardware.h.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Move platform independent header files to arch/arm/include/asm, leaving
those in asm/arch* and asm/plat* alone.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
This is a natural extension following the previous patch.
Non Feroceon based targets are unchanged.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
The implementation for memory copy functions on ARM had a (disabled)
provision for aligning the source pointer before loading registers with
data. Turns out that aligning the _destination_ pointer is much more
useful, as the read side is already sufficiently helped with the use of
preload.
So this changes the definition of the CALGN() macro to target the
destination pointer instead, and turns it on for Feroceon processors
where the gain is very noticeable.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
This code is currently disabled, which explains why no one was affected.
Signed-off-by: Nicolas Pitre <nico@marvell.com>
Signed-off-by: Lennert Buytenhek <buytenh@marvell.com>
The code in "1007:" is in the .fixup section, which in the mmuless
case is discarded. Since this code is referenced from the .text
section, it causes an link error. Move this code into the .text
section instead.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
MMUless systems have only one address space for all threads, so
both the usual access_ok() checks, and the exception handling do
not make much sense.
Hence, discard the fixup and exception tables at link time, use
memcpy/memset for the user copy/clearing functions, and define
the permission check macros to be constants.
Some of this patch was derived from the equivalent patch by
Hyok S. Choi.
Signed-off-by: Hyok S. Choi <hyok.choi@samsung.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Back in the days when we had armo (26-bit) and armv (32-bit) combined,
we had an additional layer to the uaccess macros to ensure correct
typing. Since we no longer have 26-bit in this tree, we no longer
need this layer, so eliminate it.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
As for RETINSTR/LOADREGS macros, these were for compatibility
with 26-bit ARMs. No longer required, so remove them.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
RETINSTR is a left-over from the days when we had 26-bit and
32-bit CPU support integrated into the same tree. Since this
is no longer the case, we can now remove RETINSTR.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
Assembly code that calls C code must ensure the C code sees a 64-bit
aligned stack pointer.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Fix a lot of typos. Eyeballed by jmc@ in OpenBSD.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
ifeq ($CONFIG_PREEMPT,y) -> ifeq ($(CONFIG_PREEMPT),y)
Signed-off-by: Hyok S. Choi <hyok.choi@samsung.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Malcolm Parsons
Printking a backtrace requires printk, so disable backtrace code
when printk is disabled.
Without this patch, a kernel with CONFIG_PRINTK disabled does not link:
arch/arm/lib/lib.a(backtrace.o): In function `c_backtrace':
arch/arm/lib/backtrace.S:(.text+0x108): undefined reference to `printk'
arch/arm/lib/backtrace.S:(.text+0x11c): undefined reference to `printk'
arch/arm/lib/lib.a(backtrace.o):(.fixup+0x8): undefined reference to `printk'
Signed-off-by: Malcolm Parsons <malcolm.parsons@gmail.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Peter Teichmann
Currently, if the kernels HZ value is greater than 100, delays with the udelay function are too short. This can cause trouble for instance with the zd1201 usb wlan driver.
This patch suggests a solution that keeps the overhead small and maintains (hopefully) sufficient resolution.
Signed-off-by: Peter Teichmann
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
asm/hardware.h is not required for the majority of processor support
files, ioremap support, mm initialisation, acorn IO support, nor
the debug code (which picks up its machine specific includes via
debug-macros.S)
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
When shifting the low-parts of signed numbers, a logical shift
should be used to avoid sign-extending a bit which isn't a sign
bit.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
The ARM EABI defines new names for GCC helper functions.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
We must make sure that assembly code that modifies the stack pointer
before calling a C function does it so it remains 64-bit aligned.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Catalin Marinas
If the low interrupt latency mode is enabled for the CPU (from ARMv6
onwards), the ldm/stm instructions are no longer atomic. An ldm instruction
restoring the sp and pc registers can be interrupted immediately after sp
was updated but before the pc. If this happens, the CPU restores the base
register to the value before the ldm instruction but if the base register
is not sp, the interrupt routine will corrupt the stack and the restarted
ldm instruction will load garbage.
Note that future ARM cores might always run in the low interrupt latency
mode.
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Unfortunately, later gcc versions error out when our get_user is passed
a const pointer, since we write to a temporary variable declared as
typeof(*(p)) which propagates the const-ness.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
For assembly labels to actually be local they must start with ".L" and
not only "." otherwise they still remain visible in the final link and
clutter kallsyms needlessly, and possibly make for unclear symbolic
backtrace. This patch simply inserts a"L" where appropriate. The code
itself is unchanged.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
For assembly labels to actually be local they must start with ".L" and
not only "." otherwise they still remain visible in the final link and
clutter kallsyms needlessly, and possibly make for unclear symbolic
backtrace. This patch simply inserts a"L" where appropriate. The code
itself is unchanged.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
For assembly labels to actually be local they must start with ".L" and
not only "." otherwise they still remain visible in the final link and
clutter kallsyms needlessly, and possibly make for unclear symbolic
backtrace. This patch simply inserts a"L" where appropriate. The code
itself is unchanged.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Ji-In Park discovered a bug in csumpartial which caused wrong
checksums with misaligned buffers.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
save_and_disable_irqs does not need to use mov + msr (which was
introduced to work around a documentation bug which was propagated
into binutils.) Use msr with an immediate constant, and if we're
building for ARMv6 or later, use the new CPSID instruction.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
ARM processors that have pld instructions are not using those copy_user
implementation anymore. Let's remove the useless PLD lines which were
half wrong anyway.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
The 'K' extension adds several new instructions to the ARMv6 ISA
which are primerily useful for SMP.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
This patch provides a preemption safe implementation of copy_to_user
and copy_from_user based on the copy template also used for memcpy.
It is enabled unconditionally when CONFIG_PREEMPT=y. Otherwise if the
configured architecture is not ARMv3 then it is enabled as well as it
gives better performances at least on StrongARM and XScale cores. If
ARMv3 is not too affected or if it doesn't matter too much then
uaccess.S could be removed altogether.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
This patch provides a new implementation for optimized memory copy
functions on ARM. It is made of two levels: a template that consists of
the core copy code and separate files that define macros to be used with
the core code depending on the type of copy needed. This allows for best
performances while sharing the same core for implementing memcpy(),
copy_from_user() and copy_to_user() for instance.
Two reasons for this work:
1) the current copy_to_user/copy_from_user implementation assumes no
task switch will ever occur in the middle of each copied page making
it completely unsafe with CONFIG_PREEMPT=y.
2) current copy implementations are measurably suboptimal and optimizing
different implementations separately is a pain and more opportunities
for bugs.
The reason for (1) is the fact that copy inside user pages are performed
with the ldm instruction which has no mean for testing user protections
and could possibly race with process preemption bypassing the COW mechanism
for example. This is a longstanding issue that we said ought to be fixed
for about two years now. The solution is to substitute those ldm insns
with a series of ldrt or strt insns to enforce user memory protection.
At least on StrongARM and XScale cores the ldm is not faster than the
equivalent ldr/str insns with a warm i-cache so there is no measurable
performance degradation with that change. The fact that the copy code is
a template makes it pretty easy to reuse the same core code as for memcpy
and benefit from the same performance optimizations.
Now (2) is best demonstrated with actual throughput measurements.
First, here is a summary of memcopy tests performed on a StrongARM core:
PTR alignment buffer size kernel version this version
------------------------------------------------------------
aligned 32 59.73 107.43
unaligned 32 61.31 74.72
aligned 100 132.47 136.15
unaligned 100 103.84 123.76
aligned 4096 130.67 130.80
unaligned 4096 130.68 130.64
aligned 1048576 68.03 68.18
unaligned 1048576 68.03 68.18
The buffer size is in bytes and the measured speed in MB/s. The copy
was performed repeatedly with given buffer and throughput averaged over
3 seconds.
Here we can see that the current kernel version has a higher entry cost
that shows up with small buffers. As buffer size grows both implementation
converge to the same throughput.
Now here's the exact same test performed on an XScale core (PXA255):
PTR alignment buffer size kernel version this version
------------------------------------------------------------
aligned 32 46.99 77.58
unaligned 32 53.61 59.59
aligned 100 107.19 136.59
unaligned 100 83.61 97.58
aligned 4096 129.13 129.98
unaligned 4096 128.36 128.53
aligned 1048576 53.76 59.41
unaligned 1048576 33.67 56.96
Again we can see the entry setup cost being higher for the current kernel
before getting to the main copy loop. Then throughput results converge
as long as the buffer remains in the cache. Then the 1MB case shows more
differences probably due to better pld placement and/or less instruction
interlocks in this proposed implementation.
Disclaimer: The PXA system was running with slower clocks than the
StrongARM system so trying to infer any conclusion by comparing those
separate sets of results side by side would be completely inappropriate.
So... What this patch does is to replace both memcpy and memmove with
an implementation based on the provided copy code template. The memmove
code is kept separate since it is used only if the memory areas involved
do overlap in which case the code is a transposition of the template but
with the copy occurring in the opposite direction (trying to fit that
mode into the template turned it into a mess not worth it for memmove
alone). And obviously both memcpy and memmove were tested with all kinds
of pointer alignments and buffer sizes to exercise all code paths for
correctness.
The next patch will provide the now trivial replacement implementation
copy_to_user and copy_from_user.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
Required for future enhancement patches.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
This patch gets rid of the last C implementations of needed libgcc
functions for the kernel, replacing them with optimized assembly
versions.
Those functions are:
__ashldi3
__ashrdi3
__lshrdi3
__muldi3
__ucmpdi2
The first 3 were lifted from gcc, the other two were written from scratch.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
Here's an ARM assembly SHA1 implementation to replace the default C
version. It is approximately 50% faster than the generic C version. On
an XScale processor running at 400MHz:
generic C version: 9.8 MB/s
my version: 14.5 MB/s
This code is useful to quite a few callers in the tree:
crypto/sha1.c: sha_transform(sctx->state, sctx->buffer, temp);
crypto/sha1.c: sha_transform(sctx->state, &data[i], temp);
drivers/char/random.c: sha_transform(buf, (__u8 *)r->pool+i, buf + 5);
drivers/char/random.c: sha_transform(buf, (__u8 *)data, buf + 5);
net/ipv4/syncookies.c: sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
Signed-off-by: Nicolas Pitre <nico@cam.org>
Seems to work fine on big-endian as well.
Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
We sometimes forgot to check whether the exclusive store succeeded.
Ensure that we always check. Also ensure that we always use the
out of line versions, since the inline versions are not SMP safe.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
If we found that the bit was already in the desired state, we
would skip performing the operation, and write random data back.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Alexander Schulz
This patch brings a new default config file for the shark and
fixes a compilation issue with io addressing and a runtime
problem with the serial ports, where I corrected a wrong
regshift value.
These are all shark specific files so I hope it is ok to
put them in one patch.
Signed-off-by: Alexander Schulz <alex@shark-linux.de>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Patch from Nicolas Pitre
Those are big, slow and generally not recommended for kernel code.
They are even not present on i386. So it should be concluded that
one could as well get away with do_div() alone.
Signed-off-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Convert ARM bitop assembly to a macro. All bitops follow the same
format, so it's silly duplicating the code when only one or two
instructions are different.
Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Catalin Marinas