The principles of operation states reads are in order, writes are in
order, writes can be reordered after reads, but no reads can be
reordered after writes.
The atomic and bitops variantes for z196 use the interlocked-access
facility instructions with a memory barrier before and after the
instruction. Because of the memory ordering the first barrier is
unnecessary and can be removed.
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The generic implementation for test_and_set_bit_lock in include/asm-generic
uses the standard test_and_set_bit operation. This is done with either a
'csg' or a 'loag' instruction. For both version the cache line is fetched
exclusively, even if the bit is already set. The result is an increase in
cache traffic, for a contented lock this is a bad idea.
Acked-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove the 31 bit support in order to reduce maintenance cost and
effectively remove dead code. Since a couple of years there is no
distribution left that comes with a 31 bit kernel.
The 31 bit kernel also has been broken since more than a year before
anybody noticed. In addition I added a removal warning to the kernel
shown at ipl for 5 minutes: a960062e58 ("s390: add 31 bit warning
message") which let everybody know about the plan to remove 31 bit
code. We didn't get any response.
Given that the last 31 bit only machine was introduced in 1999 let's
remove the code.
Anybody with 31 bit user space code can still use the compat mode.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
When reworking the bitops and atomic ops I missed that those instructions
that got atomic behaviour only perform a "specific-operand-serialization"
instead of a full "serialization".
The compare-and-swap instruction used before performs a full serialization
before and after the instruction is executed, which means it has full
memory barrier semantics.
In order to give the new bitops and atomic ops functions also full memory
barrier semantics add a "bcr 14,0" before and after each of those new
instructions which performs full serialization as well.
This restores memory barrier semantics for bitops and atomic ops functions
which return values, like e.g. atomic_add_return(), but not for functions
which do not return a value, like e.g. atomic_add().
This is consistent to other architectures and what common code requires.
Cc: stable@vger.kernel.org # v3.13+
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Fix some numbers in the comments describing the layout of the bit maps.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
find_first_bit_left() and friends have nothing to do with the normal
LSB0 bit numbering for big endian machines used in Linux (least
significant bit has bit number 0).
Instead they use MSB0 bit numbering, where the most signficant bit has
bit number 0. So rename find_first_bit_left() and friends to
find_first_bit_inv(), to avoid any confusion.
Also provide inv versions of set_bit, clear_bit and test_bit.
This also removes the confusing use of e.g. set_bit() in airq.c which
uses a "be_to_le" bit number conversion, which could imply that instead
set_bit_le() could be used. But that is entirely wrong since the _le
bitops variant uses yet another bit numbering scheme.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Since z9 109 we have the flogr instruction which can be used to implement
optimized versions of __ffs, ffs, __fls, fls and fls64.
So implement and use them, instead of the generic variants.
This reduces the size of the kernel image (defconfig, -march=z9-109)
by 19,648 bytes.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Just like all other architectures we should use out-of-line find bit
operations, since the inline variant bloat the size of the kernel image.
And also like all other architecures we should only supply optimized
variants of the __ffs, ffs, etc. primitives.
Therefore this patch removes the inlined s390 find bit functions and uses
the generic out-of-line variants instead.
The optimization of the primitives follows with the next patch.
With this patch also the functions find_first_bit_left() and
find_next_bit_left() have been reimplemented, since logically, they are
nothing else but a find_first_bit()/find_next_bit() implementation that
use an inverted __fls() instead of __ffs().
Also the restriction that these functions only work on machines which
support the "flogr" instruction is gone now.
This reduces the size of the kernel image (defconfig, -march=z9-109)
by 144,482 bytes.
Alone the size of the function build_sched_domains() gets reduced from
7 KB to 3,5 KB.
We also git rid of unused functions like find_first_bit_le()...
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Since zEC12 we have the interlocked-access facility 2 which allows to
use the instructions ni/oi/xi to update a single byte in storage with
compare-and-swap semantics.
So change set_bit(), clear_bit() and change_bit() to generate such code
instead of a compare-and-swap loop (or using the load-and-* instruction
family), if possible.
This reduces the text segment by yet another 8KB (defconfig).
Alternatively the long displacement variants niy/oiy/xiy could have
been used, but the extended displacement field is usually not needed
and therefore would only increase the size of the text segment again.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove CONFIG_SMP from bitops code. This reduces the C code significantly
but also generates better code for the SMP case.
This means that for !CONFIG_SMP set_bit() and friends now also have
compare and swap semantics (read: more code). However nobody really cares
for !CONFIG_SMP and this is the trade-off to simplify the SMP code which we
do care about.
The non-atomic bitops like __set_bit() now generate also better code
because the old code did not have a __builtin_contant_p() check for the
CONFIG_SMP case and therefore always generated the inline assembly variant.
However the inline assemblies for the non-atomic case now got completely
removed since gcc can produce better code, which accesses less memory
operands.
test_bit() got also a bit simplified since it did have a
__builtin_constant_p() check, however two identical code pathes for each
case (written differently).
In result this mainly reduces the to be maintained code but is not very
relevant for code generation, since there are not many non-atomic bitops
usages that we care about.
(code reduction defconfig kernel image before/after: 560 bytes).
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Make use of the interlocked-access facility 1 that got added with the
z196 architecure.
This facilility added new instructions which can atomically update a
storage location without a compare-and-swap loop. E.g. setting a bit
within a "long" can be done with a single instruction.
The size of the kernel image gets ~30kb smaller. Considering that there
are appr. 1900 bitops call sites this means that each one saves about
15-16 bytes per call site which is expected.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Fix inline assembly contraints for non atomic bitops functions.
This is broken since 2.6.34 987bcdac "[S390] use inline assembly
contraints available with gcc 3.3.3".
Reported-by: Andreas Krebbel <krebbel@linux.vnet.ibm.com>
Reported-by: Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The find_next_bit_left function is broken if used with an offset which
is not a multiple of 64. The shift to mask the bits of a 64-bit word
not to search is in the wrong direction, the result can be either a
bit found smaller than the offset or failure to find a set bit.
Cc: <stable@vger.kernel.org> # v3.8+
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove unused __BITOPS_ALIGN, and replace __BITOPS_WORDSIZE with
BITS_PER_LONG.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The flogr instruction scans a bitmap starting from the leftmost bit.
Implement support for these bitops. This could be useful to scan
bitmaps like an interrupt vector set by the hardware starting
at the leftmost bit.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove the file name from the comment at top of many files. In most
cases the file name was wrong anyway, so it's rather pointless.
Also unify the IBM copyright statement. We did have a lot of sightly
different statements and wanted to change them one after another
whenever a file gets touched. However that never happened. Instead
people start to take the old/"wrong" statements to use as a template
for new files.
So unify all of them in one go.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Replace __s390x__ with CONFIG_64BIT in all places that are not exported
to userspace or guarded with #ifdef __KERNEL__.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The majority of architectures implement ext2 atomic bitops as
test_and_{set,clear}_bit() without spinlock.
This adds this type of generic implementation in ext2-atomic-setbit.h and
use it wherever possible.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Suggested-by: Andreas Dilger <adilger@dilger.ca>
Suggested-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The previous style change enables to use asm-generic/bitops/le.h on s390.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The style that we normally use in asm-generic is to test the macro itself
for existence, so in asm-generic, do:
#ifndef find_next_zero_bit_le
extern unsigned long find_next_zero_bit_le(const void *addr,
unsigned long size, unsigned long offset);
#endif
and in the architectures, write
static inline unsigned long find_next_zero_bit_le(const void *addr,
unsigned long size, unsigned long offset)
#define find_next_zero_bit_le find_next_zero_bit_le
This adds the #define for each of the optimized find bitops in the
architectures.
Suggested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
Acked-by: Russell King <rmk+kernel@arm.linux.org.uk>
Acked-by: Greg Ungerer <gerg@uclinux.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
minix bit operations are only used by minix filesystem and useless by
other modules. Because byte order of inode and block bitmaps is different
on each architecture like below:
m68k:
big-endian 16bit indexed bitmaps
h8300, microblaze, s390, sparc, m68knommu:
big-endian 32 or 64bit indexed bitmaps
m32r, mips, sh, xtensa:
big-endian 32 or 64bit indexed bitmaps for big-endian mode
little-endian bitmaps for little-endian mode
Others:
little-endian bitmaps
In order to move minix bit operations from asm/bitops.h to architecture
independent code in minix filesystem, this provides two config options.
CONFIG_MINIX_FS_BIG_ENDIAN_16BIT_INDEXED is only selected by m68k.
CONFIG_MINIX_FS_NATIVE_ENDIAN is selected by the architectures which use
native byte order bitmaps (h8300, microblaze, s390, sparc, m68knommu,
m32r, mips, sh, xtensa). The architectures which always use little-endian
bitmaps do not select these options.
Finally, we can remove minix bit operations from asm/bitops.h for all
architectures.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Greg Ungerer <gerg@uclinux.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Andreas Schwab <schwab@linux-m68k.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Michal Simek <monstr@monstr.eu>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Paul Mundt <lethal@linux-sh.org>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
As the result of conversions, there are no users of ext2 non-atomic bit
operations except for ext2 filesystem itself. Now we can put them into
architecture independent code in ext2 filesystem, and remove from
asm/bitops.h for all architectures.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Introduce little-endian bit operations by renaming native ext2 bit
operations. The ext2 bit operations are kept as wrapper macros using
little-endian bit operations to maintain bisectability until the
conversions are finished.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Drop support to compile the kernel with gcc versions older than 3.3.3.
This allows us to use the "Q" inline assembly contraint on some more
inline assemblies without duplicating a lot of complex code (e.g. __xchg
and __cmpxchg). The distinction for older gcc versions can be removed
which saves a few lines and simplifies the code.
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
likely/unlikely profiling revealed that none of the branches in bitops
is taken likely or unlikely. So remove the annotations.
In addition the generated code is shorter.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
ext4 does not work on s390 because ext2_find_next_bit is broken. Fortunately
this function is only used by ext4. The function uses ffs which does not work
analog to ffz. The result of ffs has an offset of 1 which is not taken into
account. To fix this use the low level __ffs_word function directly instead
of the ill defined ffs.
In addition the patch improves find_next_zero_bit and ext2_find_next_zero_bit
by passing the bit offset into __ffz_word instead of adding it after the
function call returned.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Heiko Carstens