Roland McGrath [Wed, 30 Jan 2008 12:31:50 +0000 (13:31 +0100)]
x86: x86 i387 user_regset
This revamps the i387 code to be shared across 32-bit, 64-bit,
and 32-on-64. It does so by consolidating the code in one place
based on the user_regset accessor interfaces. This switches
32-bit to using the i387_64.h header and 64-bit to using the
i387.c that was previously i387_32.c, but that's what took the
least cleanup in each file. Here i387.h is stubbed to always
include i387_64.h rather than renaming the file, to keep this
diff smaller and easier to read.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:50 +0000 (13:31 +0100)]
x86: i387 renaming
This renames arch/x86/kernel/{i387_32.c => i387.c}.
This is a pure renaming, but paves the way for merging
the 32-bit and 64-bit versions of this code.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:49 +0000 (13:31 +0100)]
x86: x86 i387 header cleanup
This moves some code into asm-x86/i387_64.h in preparation for
unifying this code between 32 and 64. The 32-bit versions of
some things are copied in some existing names changed to match
32-bit names and share code. For 64, save_i387 is moved into
an inline from i387_64.c; this matches restore_i387, which is
already an inline, and makes sense since there is exactly one
caller (in signal_64.c). The save_i387 function could use more
cosmetic cleanup, but it is just moved verbatim in this patch.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:48 +0000 (13:31 +0100)]
x86: x86 i387 unify structs
The i387_fxsave_struct formats really have the same layout
on 32 and 64, with only some slightly different use of a few
fields. The i387_fsave_struct and i387_soft_struct formats
are never used by 64-bit kernels, but it doesn't hurt to
have the unused types in the union and cuts down on the
amount of #ifdef hair required throughout the i387 code.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:48 +0000 (13:31 +0100)]
x86: x86-64 wired cpu_has macros
This adds hard-wired definitions for the remaining cpu_has_* macros
that correspond to flags required-features.h demands are set for
64-bit. Using these can efficiently avoid some #ifdef's when
merging 32-bit and 64-bit code together.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:48 +0000 (13:31 +0100)]
x86: compat_sys_ptrace
This adds a generic definition of compat_sys_ptrace that calls
compat_arch_ptrace, parallel to sys_ptrace/arch_ptrace. Some
machines needing this already define a function by that name.
The new generic function is defined only on machines that
put #define __ARCH_WANT_COMPAT_SYS_PTRACE into asm/ptrace.h.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:47 +0000 (13:31 +0100)]
x86: compat_ptrace_request
This adds a compat_ptrace_request that is the analogue of ptrace_request
for the things that 32-on-64 ptrace implementations can share in common.
So far there are just a couple of requests handled generically.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:47 +0000 (13:31 +0100)]
x86: ptrace_request peekdata/pokedata
This makes ptrace_request handle {PEEK,POKE}{TEXT,DATA} directly.
Every arch_ptrace that could call generic_ptrace_peekdata already
has a default case calling ptrace_request, so this keeps things
simpler for the arch code.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:47 +0000 (13:31 +0100)]
x86: user_regset user-copy helpers
This defines two new inlines in linux/regset.h, for use in arch_ptrace
implementations and the like. These provide simplified wrappers for using
the user_regset interfaces to copy thread regset data into the caller's
user-space memory. The inlines are trivial, but make the common uses in
places such as ptrace implementation much more concise, easier to read, and
less prone to code-copying errors.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:46 +0000 (13:31 +0100)]
x86: compat_binfmt_elf Kconfig
This adds Kconfig and Makefile bits to build fs/compat_binfmt_elf.c,
just added. Each arch that wants to use this file needs to add a
"select COMPAT_BINFMT_ELF" line in its Kconfig bits that enable COMPAT.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:46 +0000 (13:31 +0100)]
x86: compat_binfmt_elf
This adds fs/compat_binfmt_elf.c, a wrapper around fs/binfmt_elf.c for
32-bit ELF support on 64-bit kernels. It can replace all the hand-rolled
versions of this that each 32/64 arch has, which are all about the same.
To use this, an arch's asm/elf.h has to define at least a few compat_*
macros that parallel the various macros that fs/binfmt_elf.c uses for
native support.
There is no attempt to deal with compat macros for the core dump format
support. To use this file, the arch has to define compat_gregset_t for
linux/elfcore-compat.h and #define CORE_DUMP_USE_REGSET. The 32-bit
compatible formats should come automatically from task_user_regset_view
called on a 32-bit task.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:45 +0000 (13:31 +0100)]
x86: user_regset helpers
This adds some inlines to linux/regset.h intended for arch code to use in
its user_regset get and set functions. These make it pretty easy to deal
with the interface's optional kernel-space or user-space pointers and its
generalized access to a part of the register data at a time.
In simple cases where the internal data structure matches the exported
layout (core dump format), a get function can be nothing but a call to
user_regset_copyout, and a set function a call to user_regset_copyin.
In other cases the exported layout is usually made up of a few pieces each
stored contiguously in a different internal data structure. These helpers
make it straightforward to write a get or set function by processing each
contiguous chunk of the data in order. The start_pos and end_pos arguments
are always constants, so these inlines collapse to a small amount of code.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:45 +0000 (13:31 +0100)]
elf core dump: notes user_regset
This modifies the ELF core dump code under #ifdef CORE_DUMP_USE_REGSET.
It changes nothing when this macro is not defined. When it's #define'd
by some arch header (e.g. asm/elf.h), the arch must support the
user_regset (linux/regset.h) interface for reading thread state.
This provides an alternate version of note segment writing that is based
purely on the user_regset interfaces. When CORE_DUMP_USE_REGSET is set,
the arch need not define macros such as ELF_CORE_COPY_REGS and ELF_ARCH.
All that information is taken from the user_regset data structures.
The core dumps come out exactly the same if arch's definitions for its
user_regset details are correct.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:44 +0000 (13:31 +0100)]
elf core dump: notes reorg
This pulls out the code for writing the notes segment of an ELF core dump
into separate functions. This cleanly isolates into one cluster of
functions everything that deals with the note formats and the hooks into
arch code to fill them. The top-level elf_core_dump function itself now
deals purely with the generic ELF format and the memory segments.
This only moves code around into functions that can be inlined away.
It should not change any behavior at all.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Roland McGrath [Wed, 30 Jan 2008 12:31:44 +0000 (13:31 +0100)]
x86: user_regset header
The new header <linux/regset.h> defines the types struct user_regset and
struct user_regset_view, with some associated declarations. This new set
of interfaces will become the standard way for arch code to expose
user-mode machine-specific state. A single set of entry points into arch
code can do all the low-level work in one place to fill the needs of core
dumps, ptrace, and any other user-mode debugging facilities that might come
along in the future.
For existing arch code to adapt to the user_regset interfaces, each arch
can work from the code it already has to support core files and ptrace.
The formats you want for user_regset are the core file formats. The only
wrinkle in adapting old ptrace implementation code as user_regset get and
set functions is that these functions can be called on current as well as
on another task_struct that is stopped and switched out as for ptrace.
For some kinds of machine state, you may have to load it directly from CPU
registers or otherwise differently for current than for another thread.
(Your core dump support already handles this in elf_core_copy_regs for
current and elf_core_copy_task_regs for other tasks, so just check there.)
The set function should also be made to work on current in case that
entails some special cases, though this was never required before for
ptrace. Adding this flexibility covers the arch needs to open the door to
more sophisticated new debugging facilities that don't always need to
context-switch to do every little thing.
The copyin/copyout helper functions (in a later patch) relieve the arch
code of most of the cumbersome details of the flexible get/set interfaces.
Signed-off-by: Roland McGrath <roland@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Harvey Harrison [Wed, 30 Jan 2008 12:31:43 +0000 (13:31 +0100)]
x86: kprobes leftover cleanups
Eliminate __always_inline, all of these static functions are
only called once. Minor whitespace cleanup. Eliminate one
supefluous return at end of void function. Change the one
#ifndef to #ifdef to match the sense of the rest of the config
tests.
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com> Acked-by: Masami Hiramatsu <mhiramat@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves i387 definitions from processor_32.h and processor_64.h
to processor.h. They are different. Very different. And there's appearently
nothing we can do about it, so they're enclosed inside ifdefs.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
There's only one difference between the NOPs used in asm code for i386 and x86_64:
i386 has a lot more variants. The code is moved to processor.h, and adjusted
accordingly.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves the prefetch[w]? functions to processor.h
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves definitions that are present in only one of the files
(between processor_32.h and processor_64.h), to processor.h. They're mostly
structures and function definitions.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves the mm_segment_t structure definition to processor.h
This makes mmsegment.h file useless, and it is deleted.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch removes definitions and macros that are not used anymore
from processor_64.h
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
x86_cpuinfo is one more to the family of "not fundamentally different"
structs. It's unified in processor.h, with very specific fields enclosed
around ifdefs.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
x86: change bitwise operations to get a void parameter.
This patch changes the bitwise operations in bitops.h to get
a void pointers as a parameter. Before this patch, a lot of warnings
can be seen. They're gone after it.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves the TASK_ALIGN constraints to common header.
The base of it is the same for x86_64 and i386. The only difference
is the presence of vSMP in x86_64. As it's not a worry in i386, we can
safely use the same code for both.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
The thread_struct is not fundamentally different between architectures, and
this patch puts it in the common header. What's really unique for each of
them is enclosed in ifdefs.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Paravirt guests need to inform the underlying hypervisor whenever the sp0
tss field changes. i386 already has such a function, and we use it for
x86_64 too. There's an unnecessary (for 64-bit) msr handling part in the original
version, and it is placed around an ifdef. Making no more sense in
processor_32.h, it is moved to the common header
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Although slighly different, the tss_struct is very similar in x86_64 and
i386. The really different part, which matchs the hardware vision of it, is
now called x86_hw_tss, and each of the architectures provides yours.
It's then used as a field in the outter tss_struct.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
current_text_addr() has a different implementation in x86_64 and
i386, but it is not fundamentally different. I stick to the i386
implementation, that seem to be a common base, and move it to processor.h
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves the pieces of processor_32.h and processor_64 that are
equal to processor.h. Only what's exactly the same is moved around, the rest
not being touched.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
x86: move the definition of set_iopl_mask to common header
This patch moves the definition of set_iopl_mask to processor.h,
instead of letting it at processor_32.h.
For x86_64, nothing is done, as we don't really need such a function.
However, having it on both arches saves us from putting an ifdef in the
pv_cpu_ops struct.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch unifies the paravirt pieces of processor.h
The functionality present in 32 bit, but not (yet) in 64-bit,
like load_sp0 is _not_ done here, and let to a different patch.
With this unification, we get paravirt for free in x86_64 processor.h
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
There are currently two definitions of load_cr3, that essentially do the
same thing. This patch moves them all to processor.h.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch moves the (duplicated) desc_empty implementation to desc.h,
where the descriptor things belong.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
There's no need for the *_MASK flags (TF_MASK, IF_MASK, etc), found in
processor.h (both _32 and _64). They have a one-to-one mapping with the
EFLAGS value. This patch removes the definitions, and use the already
existent X86_EFLAGS_ version when applicable.
[ roland@redhat.com: KVM build fixes. ]
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch wipes out the definitions of tsc_disable from processor_32.h
and move it to tsc.h, were it belongs
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Neil Horman [Wed, 30 Jan 2008 12:31:25 +0000 (13:31 +0100)]
x86, kexec: force x86 arches to boot kdump kernels on boot cpu
Recently a kdump bug was discovered in which a system would hang inside
calibrate_delay during the booting of the kdump kernel. This was caused
by the fact that the jiffies counter was not being incremented during
timer calibration. The root cause of this problem was found to be a
bios misconfiguration of the hypertransport bus. On system affected by
this hang, the bios had assigned APIC ids which used extended apic bits
(more than the nominal 4 bit ids's), but failed to configure bit 17 of
the hypertransport transaction config register, which indicated that the
mask for the destination field of interrupt packets accross the ht bus
(see section 3.3.9 of
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/26094.PDF).
If a crash occurs on a cpu with an APIC id that extends beyond 4 bits,
it will not recieve interrupts during the kdump kernel boot, and this
hang will be the result. The fix is to add this patch, whcih add an
early pci quirk check, to forcibly enable this bit in the httcfg
register. This enables all cpus on a system to receive interrupts, and
allows kdump kernel bootup to procede normally.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Jan Beulich [Wed, 30 Jan 2008 12:31:24 +0000 (13:31 +0100)]
x86: adjust enable_NMI_through_LVT0()
Its previous use in a call to on_each_cpu() was pointless, as at the
time that code gets executed only one CPU is online. Further, the
function can be __cpuinit, and for this to work without
CONFIG_HOTPLUG_CPU setup_nmi() must also get an attribute (this one
can even be __init; on 64-bits check_timer() also was lacking that
attribute).
Signed-off-by: Jan Beulich <jbeulich@novell.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Jan Beulich [Wed, 30 Jan 2008 12:31:23 +0000 (13:31 +0100)]
x86-64: honor notify_die() returning NOTIFY_STOP
This requires making die() return a value, making its callers honor
this (and be prepared that it may return), and making oops_end() have
two additional parameters.
Signed-off-by: Jan Beulich <jbeulich@novell.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Jan Beulich [Wed, 30 Jan 2008 12:31:21 +0000 (13:31 +0100)]
x86-64: make pda's cpunumber and nodenumber unsigned
This generally allows better code to be generated, since the zero-
extension during 32-bit operations comes for free (needed when the
result is used as array index or similar), whereas sign extension must
be done explicitly and frequently requires a one byte larger
instruction due to the necessary rex64 prefix.
Signed-off-by: Jan Beulich <jbeulich@novell.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Masami Hiramatsu [Wed, 30 Jan 2008 12:31:21 +0000 (13:31 +0100)]
x86: kprobes code for x86 unification
This patch unifies kprobes code.
- Unify kprobes_*.h to kprobes.h
- Unify kprobes_*.c to kprobes.c
(Differences are separated by ifdefs)
- Most differences are related to REX prefix and rip relatives.
- Two inline assembly code are different.
- One difference in kprobe_handlre()
- One fixup exception code is different, but it will be unified
if mm/extable_*.c are unified.
- Merge history logs into arch/x86/kernel/kprobes.c.
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com> Signed-off-by: Jim Keniston <jkenisto@us.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Masami Hiramatsu [Wed, 30 Jan 2008 12:31:21 +0000 (13:31 +0100)]
x86: prepare kprobes code for x86 unification
This patch cleanup kprobes code on x86 for unification.
This patch is based on Arjan's previous work.
- Remove spurious whitespace changes
- Add harmless includes
- Make the 32/64 files more identical
- Generalize structure fields' and local variable name.
- Wrap accessing to stack address by macros.
- Modify bitmap making macro.
- Merge fixup code into is_riprel() and change its name to fix_riprel().
- Set MAX_INSN_SIZE to 16 on both arch.
- Use u32 for bitmaps on both architectures.
- Clarify some comments.
Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com> Signed-off-by: Arjan van de Ven <arjan@infradead.org> Signed-off-by: Jim Keniston <jkenisto@us.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Nick Piggin [Wed, 30 Jan 2008 12:31:21 +0000 (13:31 +0100)]
x86: FIFO ticket spinlocks
Introduce ticket lock spinlocks for x86 which are FIFO. The implementation
is described in the comments. The straight-line lock/unlock instruction
sequence is slightly slower than the dec based locks on modern x86 CPUs,
however the difference is quite small on Core2 and Opteron when working out of
cache, and becomes almost insignificant even on P4 when the lock misses cache.
trylock is more significantly slower, but they are relatively rare.
On an 8 core (2 socket) Opteron, spinlock unfairness is extremely noticable,
with a userspace test having a difference of up to 2x runtime per thread, and
some threads are starved or "unfairly" granted the lock up to 1 000 000 (!)
times. After this patch, all threads appear to finish at exactly the same
time.
The memory ordering of the lock does conform to x86 standards, and the
implementation has been reviewed by Intel and AMD engineers.
The algorithm also tells us how many CPUs are contending the lock, so
lockbreak becomes trivial and we no longer have to waste 4 bytes per
spinlock for it.
After this, we can no longer spin on any locks with preempt enabled
and cannot reenable interrupts when spinning on an irq safe lock, because
at that point we have already taken a ticket and the would deadlock if
the same CPU tries to take the lock again. These are questionable anyway:
if the lock happens to be called under a preempt or interrupt disabled section,
then it will just have the same latency problems. The real fix is to keep
critical sections short, and ensure locks are reasonably fair (which this
patch does).
Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Nick Piggin [Wed, 30 Jan 2008 12:31:20 +0000 (13:31 +0100)]
spinlock: lockbreak cleanup
The break_lock data structure and code for spinlocks is quite nasty.
Not only does it double the size of a spinlock but it changes locking to
a potentially less optimal trylock.
Put all of that under CONFIG_GENERIC_LOCKBREAK, and introduce a
__raw_spin_is_contended that uses the lock data itself to determine whether
there are waiters on the lock, to be used if CONFIG_GENERIC_LOCKBREAK is
not set.
Rename need_lockbreak to spin_needbreak, make it use spin_is_contended to
decouple it from the spinlock implementation, and make it typesafe (rwlocks
do not have any need_lockbreak sites -- why do they even get bloated up
with that break_lock then?).
Signed-off-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Markus Metzger [Wed, 30 Jan 2008 12:31:20 +0000 (13:31 +0100)]
x86, ptrace: new ptrace BTS API
Here's the new ptrace BTS API that supports two different overflow handling mechanisms (wrap-around and buffer-full-signal) to support two different use cases (debugging and profiling).
It further combines buffer allocation and configuration.
Opens:
- memory rlimit
- overflow signal
What would be the right signal to use?
Signed-off-by: Markus Metzger <markus.t.metzger@intel.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
x86: avoid ifdefs in desc.h, getting rid of pack_ldt and pack_tss
By Andi Kleen's suggestion, this patch removes pack_ldt() and pack_tss()
wrappers in favour of a general wrapper. It saves us an ifdef and some lines
of code, but more importantly, it's more elegant.
No functional change is made, although the object code is expected to be
different.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Huang, Ying [Wed, 30 Jan 2008 12:31:19 +0000 (13:31 +0100)]
x86: EFI runtime service support
This patch adds basic runtime services support for EFI x86_64 system. The
main file of the patch is the addition of efi_64.c for x86_64. This file is
modeled after the EFI IA32 avatar. EFI runtime services initialization are
implemented in efi_64.c. Some x86_64 specifics are worth noting here. On
x86_64, parameters passed to EFI firmware services need to follow the EFI
calling convention. For this purpose, a set of functions named efi_call<x>
(<x> is the number of parameters) are implemented. EFI function calls are
wrapped before calling the firmware service. The duplicated code between
efi_32.c and efi_64.c is placed in efi.c to remove them from efi_32.c.
Jeff Dike [Wed, 30 Jan 2008 12:31:17 +0000 (13:31 +0100)]
uml: add asm-um/asm.h
git-x86.patch introduces asm-x86/asm.h. UML needs to follow this with
an asm/asm.h of its own.
Signed-off-by: Jeff Dike <jdike@linux.intel.com> 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>
Andrew Morton [Wed, 30 Jan 2008 12:31:17 +0000 (13:31 +0100)]
x86: __vdso_getcpu() warning fix
arch/x86/vdso/vgetcpu.c: In function '__vdso_getcpu':
arch/x86/vdso/vgetcpu.c:22: warning: pointer targets in passing argument 1 of 'native_read_tscp' differ in signedness
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>
This patch unifies the set_tss_desc between i386 and x86_64,
which can now have a common implementation. After the old
functions are removed from desc_{32,64}.h, nothing important is
left, and the files can be removed.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
x86: move _set_gate and its users to a common location
This patch moves _set_gate and its users to desc.h. We can now
use common code for x86_64 and i386.
[ mingo@elte.hu: set_system_gate() fixes for nasty crashes. ]
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
With the types used to access descriptors in x86_64 and i386
now being the same, the code that effectively handles them can
now be easily shared. This patch moves the paravirt part of
desc_32.h into desc.h, and then, we get paravirt support in x86_64
for free.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch makes get_desc_base() receive a struct desc_struct,
and then uses its internal fields to compute the base address.
This is done at both i386 and x86_64, and then it is moved
to common header
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch changes the type of tls_array in x86_64 to
a desc_struct. Now, both i386 and x86_64 tls_array have
the same type, and code accessing it can be shared.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch unifies the non-paravirt part of desc_{32,64}.h into
desc.h. Most of it, is simply common code, that is moved to
the shared header. The only exception is the set_ldt_desc in desc_64.h,
which is changed - included its name - to accomodate for the way
the ldt is set up in i386.
Also, constant definitions used in desc_32.h are moved to desc_defs.h
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
this patch moves constant definitions regarding descriptor types
from desc_32.h to desc_defs.h. The change from defines to enum
to comply with previous versions in desc_defs.h
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch changes the write_gdt_entry function signature.
Instead of the old "a" and "b" parameters, it now receives
a pointer to a desc_struct, and the size of the entry being
handled. This is because x86_64 can have some 16-byte entries
as well as 8-byte ones.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> CC: Zachary Amsden <zach@vmware.com> CC: Jeremy Fitzhardinge <Jeremy.Fitzhardinge.citrix.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Provide a new type, tss_desc, to represent the tss descriptor
in a unified way accross x86_64 and i386
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch introduces fill_ldt(), which populates a ldt descriptor
from a user_desc in once, instead of relying in the LDT_entry_a and
LDT_entry_b macros
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
This patch modifies the write_ldt() function to make use
of the new struct desc_struct instead of entry_1 and entry_2
entries
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
this patch introduces ldt_desc type to account for the differences
in the ldt descriptor in x86_64 and i386
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
this patch changes write_idt_entry signature. It now takes a gate_desc
instead of the a and b parameters. It will allow it to be later unified
between i386 and x86_64.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> CC: Zachary Amsden <zach@vmware.com> CC: Jeremy Fitzhardinge <Jeremy.Fitzhardinge.citrix.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
To account for the differences in gate descriptor in i386 and x86_64
a gate_desc type is introduced.
Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
projects / linux-2.6 / log