Existing PRSTATUS note contains only si_signo, si_code, si_errno fields
from the siginfo of the signal which caused core to be dumped.
There are tools which try to analyze crashes for possible security
implications, and they want to use, among other data, si_addr field from
the SIGSEGV.
This patch adds a new elf note, NT_SIGINFO, which contains the complete
siginfo_t of the signal which killed the process.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Amerigo Wang <amwang@redhat.com>
Cc: "Jonathan M. Foote" <jmfoote@cert.org>
Cc: Roland McGrath <roland@hack.frob.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a preparatory patch for the introduction of NT_SIGINFO elf note.
Make the location of compat_siginfo_t uniform across eight architectures
which have it. Now it can be pulled in by including asm/compat.h or
linux/compat.h.
Most of the copies are verbatim. compat_uid[32]_t had to be replaced by
__compat_uid[32]_t. compat_uptr_t had to be moved up before
compat_siginfo_t in asm/compat.h on a several architectures (tile already
had it moved up). compat_sigval_t had to be relocated from linux/compat.h
to asm/compat.h.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Amerigo Wang <amwang@redhat.com>
Cc: "Jonathan M. Foote" <jmfoote@cert.org>
Cc: Roland McGrath <roland@hack.frob.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In !CORE_DUMP_USE_REGSET case, if elf_note_info_init fails to allocate
memory for info->fields, it frees already allocated stuff and returns
error to its caller, fill_note_info. Which in turn returns error to its
caller, elf_core_dump. Which jumps to cleanup label and calls
free_note_info, which will happily try to free all info->fields again.
BOOM.
This is the fix.
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Venu Byravarasu <vbyravarasu@nvidia.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Previous code was using optimizations which were developed to work well
even on narrow-word CPUs (by today's standards). But Linux runs only on
32-bit and wider CPUs. We can use that.
First: using 32x32->64 multiply and trivial 32-bit shift, we can correctly
divide by 10 much larger numbers, and thus we can print groups of 9 digits
instead of groups of 5 digits.
Next: there are two algorithms to print larger numbers. One is generic:
divide by 1000000000 and repeatedly print groups of (up to) 9 digits.
It's conceptually simple, but requires an (unsigned long long) /
1000000000 division.
Second algorithm splits 64-bit unsigned long long into 16-bit chunks,
manipulates them cleverly and generates groups of 4 decimal digits. It so
happens that it does NOT require long long division.
If long is > 32 bits, division of 64-bit values is relatively easy, and we
will use the first algorithm. If long long is > 64 bits (strange
architecture with VERY large long long), second algorithm can't be used,
and we again use the first one.
Else (if long is 32 bits and long long is 64 bits) we use second one.
And third: there is a simple optimization which takes fast path not only
for zero as was done before, but for all one-digit numbers.
In all tested cases new code is faster than old one, in many cases by 30%,
in few cases by more than 50% (for example, on x86-32, conversion of
12345678). Code growth is ~0 in 32-bit case and ~130 bytes in 64-bit
case.
This patch is based upon an original from Michal Nazarewicz.
[akpm@linux-foundation.org: checkpatch fixes]
Signed-off-by: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Douglas W Jones <jones@cs.uiowa.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
PTRACE_SEIZE code is tested and ready for production use, remove the
code which requires special bit in data argument to make PTRACE_SEIZE
work.
Strace team prepares for a new release of strace, and we would like to
ship the code which uses PTRACE_SEIZE, preferably after this change goes
into released kernel.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
PTRACE_EVENT_foo and PTRACE_O_TRACEfoo used to match.
New PTRACE_EVENT_STOP is the first event which has no corresponding
PTRACE_O_TRACE option. If we will ever want to add another such option,
its PTRACE_EVENT's value will collide with PTRACE_EVENT_STOP's value.
This patch changes PTRACE_EVENT_STOP value to prevent this.
While at it, added a comment - the one atop PTRACE_EVENT block, saying
"Wait extended result codes for the above trace options", is not true
for PTRACE_EVENT_STOP.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Tejun Heo <tj@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This can be used to close a few corner cases in strace where we get
unwanted racy behavior after attach, but before we have a chance to set
options (the notorious post-execve SIGTRAP comes to mind), and removes
the need to track "did we set opts for this task" state in strace
internals.
While we are at it:
Make it possible to extend SEIZE in the future with more functionality
by passing non-zero 'addr' parameter. To that end, error out if 'addr'
is non-zero. PTRACE_ATTACH did not (and still does not) have such
check, and users (strace) do pass garbage there... let's avoid
repeating this mistake with SEIZE.
Set all task->ptrace bits in one operation - before this change, we were
adding PT_SEIZED and PT_PTRACE_CAP with task->ptrace |= BIT ops. This
was probably ok (not a bug), but let's be on a safer side.
Changes since v2: use (unsigned long) casts instead of (long) ones, move
PTRACE_SEIZE_DEVEL-related code to separate lines of code.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Pedro Alves <palves@redhat.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Exchange PT_TRACESYSGOOD and PT_PTRACE_CAP bit positions, which makes
PT_option bits contiguous and therefore makes code in
ptrace_setoptions() much simpler.
Every PTRACE_O_TRACEevent is defined to (1 << PTRACE_EVENT_event)
instead of using explicit numeric constants, to ensure we don't mess up
relationship between bit positions and event ids.
PT_EVENT_FLAG_SHIFT was not particularly useful, PT_OPT_FLAG_SHIFT with
value of PT_EVENT_FLAG_SHIFT-1 is easier to use.
PT_TRACE_MASK constant is nuked, the only its use is replaced by
(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT).
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
On ptrace(PTRACE_SETOPTIONS, pid, 0, <opts>), we used to set those
option bits which are known, and then fail with -EINVAL if there are
some unknown bits in <opts>.
This is inconsistent with typical error handling, which does not change
any state if input is invalid.
This patch changes PTRACE_SETOPTIONS behavior so that in this case, we
return -EINVAL and don't change any bits in task->ptrace.
It's very unlikely that there is userspace code in the wild which will
be affected by this change: it should have the form
ptrace(PTRACE_SETOPTIONS, pid, 0, PTRACE_O_BOGUSOPT)
where PTRACE_O_BOGUSOPT is a constant unknown to the kernel. But kernel
headers, naturally, don't contain any PTRACE_O_BOGUSOPTs, thus the only
way userspace can use one if it defines one itself. I can't see why
anyone would do such a thing deliberately.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
I just received another user's pleas for help when their init
mysteriously died. I again explained that they need to check whether it
died because of bad instruction, a segv, or something else. Which was
an annoying detour into writing a trivial C program to spawn his init
and print its exit code:
http://lists.busybox.net/pipermail/busybox/2012-January/077172.html
I hear you saying "just test it under /bin/sh". Well, the crashing init
_was_ /bin/sh.
Which prompted me to make kernel do this first step automatically. We can
print exit code, which makes it possible to see that death was from e.g.
SIGILL without writing test programs.
[akpm@linux-foundation.org: add 0x to hex number output]
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
text data bss dec hex filename
8455963 532732 1810804 10799499 a4c98b vmlinux.o.before
8448899 532732 1810804 10792435 a4adf3 vmlinux.o
This change also removes commented-out copy of __nlmsg_put
which was last touched in 2005 with "Enable once all users
have been converted" comment on top.
Changes in v2: rediffed against net-next.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When multithreaded program execs under ptrace,
all traced threads report WIFEXITED status, except for
thread group leader and the thread which execs.
Unless tracer tracks thread group relationship between tracees,
which is a nontrivial task, it will not detect that
execed thread no longer exists.
This patch allows tracer to figure out which thread
performed this exec, by requesting PTRACE_GETEVENTMSG
in PTRACE_EVENT_EXEC stop.
Another, samller problem which is solved by this patch
is that tracer now can figure out which of the several
concurrent execs in multithreaded program succeeded.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
This patch uninlines four similar functions, foo_update_attr(), in four
fbcon-related files.
These functions contain loops, two of theam have _nested_ loops, and they
have more than one callsite each. I think they should not be inlined.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The STACK_GROWSUP case of stack expansion was missing a test for 'prev',
which got removed by commit cb8f488c33
("mmap.c: deinline a few functions") by mistake.
I found my original email in "sent" folder. The patch in that mail
does NOT remove !prev. That change had beed added by someone else.
Ok, I think we are not much interested in who did it, let's
fix it for good.
[ "It looks like this was caused by me fixing rejects. That was the
fancy include-lots-of-context-so-it-wont-apply patch." - akpm ]
Reported-and-bisected-by: Helge Deller <deller@gmx.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__vma_link_file and expand_downwards functions are not small, yeat they
are marked inline. They probably had one callsite sometime in the past,
but now they have more. In order to prevent similar thing, I also
deinlined expand_upwards, despite it having only pne callsite. Nowadays
gcc auto-inlines such static functions anyway. In find_extend_vma, I
removed one extra level of indirection.
Patch is deliberately generated with -U $BIGNUM to make
it easier to see that functions are big.
Result:
# size */*/mmap.o */vmlinux
text data bss dec hex filename
9514 188 16 9718 25f6 0.org/mm/mmap.o
9237 188 16 9441 24e1 deinline/mm/mmap.o
6124402 858996 389480 7372878 70804e 0.org/vmlinux
6124113 858996 389480 7372589 707f2d deinline/vmlinux
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kmem_free() function takes (ptr, size) arguments but doesn't actually use
second one.
This patch removes size argument from all callsites.
SGI-PV: 981498
SGI-Modid: xfs-linux-melb:xfs-kern:31050a
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
This shrinks module.o and each *.ko file.
And finally, structure members which hold length of module
code (four such members there) and count of symbols
are converted from longs to ints.
We cannot possibly have a module where 32 bits won't
be enough to hold such counts.
For one, module loading checks module size for sanity
before loading, so such insanely big module will fail
that test first.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
module.c and module.h conatains code for finding
exported symbols which are declared with EXPORT_UNUSED_SYMBOL,
and this code is compiled in even if CONFIG_UNUSED_SYMBOLS is not set
and thus there can be no EXPORT_UNUSED_SYMBOLs in modules anyway
(because EXPORT_UNUSED_SYMBOL(x) are compiled out to nothing then).
This patch adds required #ifdefs.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Deinlines and moves big functions from .h to .c files.
Adds prototypes for ahc_lookup_scb and ahd_lookup_scb to .h files.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
sound/pci/rme9652/hdspm.c has unusually large number of static inline
functions - 22.
I looked through them and some of them seem to be too big to warrant inlining.
This patch removes "inline" from these static functions (regardless of number
of callsites - gcc nowadays auto-inlines statics with one callsite).
Size difference on 32bit x86:
text data bss dec hex filename
20437 2160 516 23113 5a49 linux-2.6-ALLYES/sound/pci/rme9652/hdspm.o
18036 2160 516 20712 50e8 linux-2.6.inline-ALLYES/sound/pci/rme9652/hdspm.o
[coding fix by Takashi Iwai <tiwai@suse.de>]
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
drivers/net/sb1000.c has lots of inlined static functions.
Mst of them are used at initialization, wait for some
hardware register to change (wait using yield, sleep etc),
or do slow port-based I/O. Inlining thse "for speed" makes no sense.
This patch removes "inline" from biggest static function
(regardless of number of callsites - gcc nowadays auto-inlines
statics with one callsite).
Size difference for 32bit x86:
text data bss dec hex filename
6299 129 0 6428 191c linux-2.6-ALLYES/drivers/net/sb1000.o
5418 129 0 5547 15ab linux-2.6.inline-ALLYES/drivers/net/sb1000.o
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
drivers/atm/horizon.c has unusually large number
of static inline functions - 36.
I looked through them. Most of them seems to be small enough,
but a few are big, others are using udelay or busy loop,
and as such are better not be inlined.
This patch removes "inline" from these static functions
(regardless of number of callsites - gcc nowadays auto-inlines
statics with one callsite).
Size difference for 32bit x86:
text data bss dec hex filename
8201 180 6 8387 20c3 linux-2.6-ALLYES/drivers/atm/horizon.o
7840 180 6 8026 1f5a linux-2.6.inline-ALLYES/drivers/atm/horizon.o
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Hi John,
Can you please take a look at this patch?
drivers/net/wireless/wavelan_cs.c has unusually large number
of static inline functions - 27.
I looked through them and 20 of them do not seem to warrant inlining.
Some are really big; others call mdelay(1) or busy-wait for a bit
to be set in a hardware register - it's pointless
to optimize such functions for speed.
This patch removes "inline" from these static function
(regardless of number of callsites - gcc nowadays auto-inlines
statics with one callsite).
Size difference for 32bit x86:
text data bss dec hex filename
17020 372 8 17400 43f8 linux-2.6-ALLYES/drivers/net/wireless/wavelan_cs.o
14032 356 8 14396 383c linux-2.6.inline-ALLYES/drivers/net/wireless/wavelan_cs.o
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
--
vda
Signed-off-by: John W. Linville <linville@tuxdriver.com>
drivers/atm/ambassador.c has unusually large number
of static inline functions - 22.
I looked through them and half of them seem to be too big
to warrant inlining.
This patch removes "inline" from these static functions
(regardless of number of callsites - gcc nowadays auto-inlines
statics with one callsite).
Size difference for 32bit x86:
text data bss dec hex filename
10209 8488 4 18701 490d linux-2.6-ALLYES/drivers/atm/ambassador.o
9462 8488 4 17954 4622 linux-2.6.inline-ALLYES/drivers/atm/ambassador.o
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In arch/x86/boot/printf.c gets rid of unused tail of digits: const char
*digits = "0123456789abcdefghijklmnopqrstuvwxyz"; (we are using 0-9a-f
only)
Uses smaller/faster lowercasing (by ORing with 0x20)
if we know that we work on numbers/digits. Makes
strtoul smaller, and also we are getting rid of
static const char small_digits[] = "0123456789abcdefx";
static const char large_digits[] = "0123456789ABCDEFX";
since this works equally well:
static const char digits[16] = "0123456789ABCDEF";
Size savings:
$ size vmlinux.org vmlinux
text data bss dec hex filename
877320 112252 90112 1079684 107984 vmlinux.org
877048 112252 90112 1079412 107874 vmlinux
It may be also a tiny bit faster because code has less
branches now, but I doubt it is measurable.
[ hugh@veritas.com: uppercase pointers fix ]
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Stop using unsigned _longs_ for printk buffer indexes. Log buffer is way
smaller than 2 gigabytes and unsigned ints will work too . Indeed, they do
work nicely on all 32-bit platforms where longs and ints are the same.
With this patch, we have following size savings on amd64:
text data bss dec hex filename
5997 313 17736 24046 5dee 2.6.23.1.t64/kernel/printk.o
5858 313 17700 23871 5d3f 2.6.23.1.printk.t64/kernel/printk.o
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The current TCP RST construction reuses the old packet and can't
deal with IP options as a consequence of that. Construct the
RST from scratch instead.
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes inlines except those which are used
by packet matching code and thus are performance-critical.
Before:
$ size */*/*/ip*tables*.o
text data bss dec hex filename
6402 500 16 6918 1b06 net/ipv4/netfilter/ip_tables.o
7130 500 16 7646 1dde net/ipv6/netfilter/ip6_tables.o
After:
$ size */*/*/ip*tables*.o
text data bss dec hex filename
6307 500 16 6823 1aa7 net/ipv4/netfilter/ip_tables.o
7010 500 16 7526 1d66 net/ipv6/netfilter/ip6_tables.o
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Analogously to camellia7 patch, move
"absorb kw2 to other subkeys" and "absorb kw4 to other subkeys"
code parts into camellia_setup_tail(). This further reduces
source and object code size at the cost of two brances
in key setup code.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Move "key XOR is end of F-function" code part into
camellia_setup_tail(), it is sufficiently similar
between camellia_setup128 and camellia_setup256.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
unifies encrypt/decrypt routines for different key lengths.
This reduces module size by ~25%, with tiny (less than 1%)
speed impact.
Also collapses encrypt/decrypt into more readable
(visually shorter) form using macros.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Remove unused macro params.
Use (u8)(expr) instead of (expr) & 0xff,
helps gcc to realize how to use simpler commands.
Move CAMELLIA_FLS macro closer to encrypt/decrypt routines.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Move huge unrolled pieces of code (3 screenfuls) at the end of
128/256 key setup routines into common camellia_setup_tail(),
convert it to loop there.
Loop is still unrolled six times, so performance hit is very small,
code size win is big.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Noriaki TAKAMIYA <takamiya@po.ntts.co.jp>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Optimize GETU32 to use 4-byte memcpy (modern gcc will convert
such memcpy to single move instruction on i386).
Original GETU32 did four byte fetches, and shifted/XORed those.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Noriaki TAKAMIYA <takamiya@po.ntts.co.jp>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Rename some macros to shorter names: CAMELLIA_RR8 -> ROR8,
making it easier to understand that it is just a right rotation,
nothing camellia-specific in it.
CAMELLIA_SUBKEY_L() -> SUBKEY_L() - just shorter.
Move be32 <-> cpu conversions out of en/decrypt128/256 and into
camellia_en/decrypt - no reason to have that code duplicated twice.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Noriaki TAKAMIYA <takamiya@po.ntts.co.jp>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Move code blocks around so that related pieces are closer together:
e.g. CAMELLIA_ROUNDSM macro does not need to be separated
from the rest of the code by huge array of constants.
Remove unused macros (COPY4WORD, SWAP4WORD, XOR4WORD[2])
Drop SUBL(), SUBR() macros which only obscure things.
Same for CAMELLIA_SP1110() macro and KEY_TABLE_TYPE typedef.
Remove useless comments:
/* encryption */ -- well it's obvious enough already!
void camellia_encrypt128(...)
Combine swap with copying at the beginning/end of encrypt/decrypt.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Noriaki TAKAMIYA <takamiya@po.ntts.co.jp>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Currently twofish cipher key setup code
has unrolled loops - approximately 70-100
instructions are repeated 40 times.
As a result, twofish module is the biggest module
in crypto/*.
Unrolling produces x2.5 more code (+18k on i386), and speeds up key
setup by 7%:
unrolled: twofish_setkey/sec: 41128
loop: twofish_setkey/sec: 38148
CALC_K256: ~100 insns each
CALC_K192: ~90 insns
CALC_K: ~70 insns
Attached patch removes this unrolling.
$ size */twofish_common.o
text data bss dec hex filename
37920 0 0 37920 9420 crypto.org/twofish_common.o
13209 0 0 13209 3399 crypto/twofish_common.o
Run tested (modprobe tcrypt reports ok). Please apply.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add const modifiers to a few struct nls_table's member pointers in
include/linux/nls.h and adds a lot of const's in fs/nls/*.c files.
Resulting changes as visible by size:
text data bss dec hex filename
113612 481216 2368 597196 91ccc nls.org/built-in.o
593548 3296 288 597132 91c8c nls/built-in.o
Apparently compiler managed to optimize code a bit better
because of const-ness.
No other changes are made.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch modifies gzip unpacking code in bnx2 driver so that
it does not depend on bnx2 internals. I will move this code
out of the driver and into zlib in follow-on patch.
It can be useful in other drivers which need to store firmwares
or any other relatively big binary blobs - fonts, cursor bitmaps,
whatever.
Patch is run tested by Michael Chan (driver author).
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Optimize integer-to-string conversion in vsprintf.c for base 10. This is
by far the most used conversion, and in some use cases it impacts
performance. For example, top reads /proc/$PID/stat for every process, and
with 4000 processes decimal conversion alone takes noticeable time.
Using code from
http://www.cs.uiowa.edu/~jones/bcd/decimal.html
(with permission from the author, Douglas W. Jones)
binary-to-decimal-string conversion is done in groups of five digits at
once, using only additions/subtractions/shifts (with -O2; -Os throws in
some multiply instructions).
On i386 arch gcc 4.1.2 -O2 generates ~500 bytes of code.
This patch is run tested. Userspace benchmark/test is also attached.
I tested it on PIII and AMD64 and new code is generally ~2.5 times
faster. On AMD64:
# ./vsprintf_verify-O2
Original decimal conv: .......... 151 ns per iteration
Patched decimal conv: .......... 62 ns per iteration
Testing correctness
12895992590592 ok... [Ctrl-C]
# ./vsprintf_verify-O2
Original decimal conv: .......... 151 ns per iteration
Patched decimal conv: .......... 62 ns per iteration
Testing correctness
26025406464 ok... [Ctrl-C]
More realistic test: top from busybox project was modified to
report how many us it took to scan /proc (this does not account
any processing done after that, like sorting process list),
and then I test it with 4000 processes:
#!/bin/sh
i=4000
while test $i != 0; do
sleep 30 &
let i--
done
busybox top -b -n3 >/dev/null
on unpatched kernel:
top: 4120 processes took 102864 microseconds to scan
top: 4120 processes took 91757 microseconds to scan
top: 4120 processes took 92517 microseconds to scan
top: 4120 processes took 92581 microseconds to scan
on patched kernel:
top: 4120 processes took 75460 microseconds to scan
top: 4120 processes took 66451 microseconds to scan
top: 4120 processes took 67267 microseconds to scan
top: 4120 processes took 67618 microseconds to scan
The speedup comes from much faster generation of /proc/PID/stat
by sprintf() calls inside the kernel.
Signed-off-by: Douglas W Jones <jones@cs.uiowa.edu>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* There is no point in having full "0...9a...z" constant vector,
if we use only "0...9a...f" (and "x" for "0x").
* Post-decrement usually needs a few more instructions, so use
pre decrement instead where makes sense:
- while (i < precision--) {
+ while (i <= --precision) {
* if base != 10 (=> base 8 or 16), we can avoid using division
in a loop and use mask/shift, obtaining much faster conversion.
(More complex optimization for base 10 case is in the second patch).
Overall, size vsprintf.o shows ~80 bytes smaller text section
with this patch applied.
Signed-off-by: Douglas W Jones <jones@cs.uiowa.edu>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Comment says "Read high byte first as some registers increment..."
but code doesn't guarantee that, I think:
return ((ahd_inb(ahd, port+1) << 8) | ahd_inb(ahd, port));
Compiler can reorder it.
Make the order explicit.
Signed-off-by: Denis Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Fixed rejections and added aic7xxx code
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
First warning result from open-coded PTR_ERR,
the rest is caused by code like this:
*(u32 *) ((u32) buf_priv->kernel_virtual + used)
I've also fixed a missing PTR_ERR in i830_dma.c
Signed-off-by: Dave Airlie <airlied@linux.ie>
Denys Vlasenko