]> err.no Git - linux-2.6/log
linux-2.6
17 years ago[PATCH] x86-64: Remove mk_pte_phys()
Andi Kleen [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] x86-64: Remove mk_pte_phys()

- Convert last user to pfn_pte
- Remove mk_pte_phys

Suggested by Jan Beulich

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Fix broken CONFIG_COMPAT_VDSO on i386
Jan Beulich [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] i386: Fix broken CONFIG_COMPAT_VDSO on i386

After updating several machines to 2.6.20, I can't boot  anymore the single
one of them that supports the NX bit and is configured as a 32-bit system.

My understanding is that the VDSO changes in 2.6.20-rc7 were not fully
cooked, in that with that config option enabled VDSO_SYM(x) now equals
x, meaning that an address in the fixmap area is now being passed to
apps via AT_SYSINFO. However, the page is mapped with PAGE_READONLY
rather than PAGE_READONLY_EXEC.

I'm not certain whether having app code go through the fixmap area is
intended, but in case it is here is the simple patch that makes things work
again.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: fix 32-bit ioctls on x64_32
Giuliano Procida [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] i386: fix 32-bit ioctls on x64_32

[MTRR] fix 32-bit ioctls on x64_32

Signed-off-by: Giuliano Procida <giuliano.procida@googlemail.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86: Unify pcspeaker platform device code between i386/x86-64
Andi Kleen [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] x86: Unify pcspeaker platform device code between i386/x86-64

Trivial cleanup.

Only change is that it is always compiled in now on x86-64 like on i386.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Remove extern declaration from mm/discontig.c, put in header.
Rusty Russell [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] i386: Remove extern declaration from mm/discontig.c, put in header.

Extern declarations belong in headers.  Times, they are a'changin.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andi Kleen <ak@suse.de>
===================================================================

17 years ago[PATCH] i386: Rename cpu_gdt_descr and remove extern declaration from smpboot.c
Rusty Russell [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] i386: Rename cpu_gdt_descr and remove extern declaration from smpboot.c

When I implemented the DECLARE_PER_CPU(var) macros, I was careful that
people couldn't use "var" in a non-percpu context, by prepending
percpu__.  I never considered that this would allow them to overload
the same name for a per-cpu and a non-percpu variable.

It is only one of many horrors in the i386 boot code, but let's rename
the non-perpcu cpu_gdt_descr to early_gdt_descr (not boot_gdt_descr,
that's something else...)

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andi Kleen <ak@suse.de>
===================================================================

17 years ago[PATCH] i386: Move mce_disabled to asm/mce.h
Rusty Russell [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] i386: Move mce_disabled to asm/mce.h

Allows external actors to disable mce.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andi Kleen <ak@suse.de>
===================================================================

17 years ago[PATCH] i386: paravirt unhandled fallthrough
Rusty Russell [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] i386: paravirt unhandled fallthrough

The current code simply calls "start_kernel" directly if we're under a
hypervisor and no paravirt_ops backend wants us, because paravirt.c
registers that as a backend.

This was always a vain hope; start_kernel won't get far without setup.
It's also impossible for paravirt_ops backends which don't sit in the
arch/i386/kernel directory: they can't link before paravirt.o anyway.

Keep it simple: if we pass all the registered paravirt probes, BUG().

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86_64: Wire up compat epoll_pwait
Ralf Baechle [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] x86_64: Wire up compat epoll_pwait

> Which remembers me that I think that MIPS is using the non-compat version
> of sys_epoll_pwait for compat syscalls. But maybe MIPS doesn't need a compat
> syscall for some reason. Dunno.

Which reminds me that x86_64 i386 compat doesn't wire up sys_epoll_pwait ;-)

Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86: Don't require the vDSO for handling a.out signals
Andi Kleen [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] x86: Don't require the vDSO for handling a.out signals

and in other strange binfmts. vDSO is not necessarily mapped there.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Fix Cyrix MediaGX detection
Alan [Tue, 13 Feb 2007 12:26:26 +0000 (13:26 +0100)]
[PATCH] i386: Fix Cyrix MediaGX detection

The old Cyrix 5520 CPU detection code relied upon the PCI layer setup being
done earlier than the CPU setup, which is no longer true.  Fortunately we
know that if the processor is a MediaGX we can do type 1 pci config
accesses to check the companion chip.  We thus do those directly and from
this find the 5520 and implement the workarounds for the timer problem

Original report from takada@mbf.nifty.com, I sent a proposed patch which
Takara then corrected, tested and sent back to the list on 10th January.

Submitting for merging as it seems to have been missed

AK: Changed to use pci-direct.h and fix warning for !CONFIG_PCI (later
AK: originally from akpm)

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: <takada@mbf.nifty.com>
Cc: Jordan Crouse <jordan.crouse@amd.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] i386: Fix warning in cpu initialization
Andi Kleen [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] i386: Fix warning in cpu initialization

Fix bogus warning

linux/arch/i386/kernel/cpu/transmeta.c:12: warning: ‘cpu_freq’ may be used uninitialized in this function

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Fix warning in microcode.c
Andi Kleen [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] i386: Fix warning in microcode.c

Fix bogus gcc warning

linux/arch/i386/kernel/microcode.c:387: warning: ‘new_mc’ may be used uninitialized in this function

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86: Enable NMI watchdog for AMD Family 0x10 CPUs
Andi Kleen [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86: Enable NMI watchdog for AMD Family 0x10 CPUs

For i386/x86-64.

Straight forward -- just reuse the Family 0xf code.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86: Add new CPUID bits for AMD Family 10 CPUs in /proc/cpuinfo
Andi Kleen [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86: Add new CPUID bits for AMD Family 10 CPUs in /proc/cpuinfo

Just various new acronyms. The new popcnt bit is in the middle
of Intel space. This looks a little weird, but I've been assured
it's ok.

Also I fixed RDTSCP for i386 which was at the wrong place.

For i386 and x86-64.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Remove fastcall in paravirt.[ch]
Andi Kleen [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] i386: Remove fastcall in paravirt.[ch]

Not needed because fastcall is always default now

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Fix wrong gcc check in bitops.h
Andi Kleen [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86-64: Fix wrong gcc check in bitops.h

gcc 5.0 will likely not have the constraint problem
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: survive having no irq mapping for a vector
Eric W. Biederman [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86-64: survive having no irq mapping for a vector

Occasionally the kernel has bugs that result in no irq being found for a
given cpu vector.  If we acknowledge the irq the system has a good chance
of continuing even though we dropped an irq message.  If we continue to
simply print a message and not acknowledge the irq the system is likely to
become non-responsive shortly there after.

AK: Fixed compilation for UP kernels

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: "Luigi Genoni" <luigi.genoni@pirelli.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] i386: geode configuration fixes
TAKADA Yoshihito [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] i386: geode configuration fixes

Original code doesn't write back to CCR4 register.  This patch reflects a
value of a register.

Cc: Jordan Crouse <jordan.crouse@amd.com>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: add option to show more code in oops reports
Chuck Ebbert [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] i386: add option to show more code in oops reports

Sometimes developers need to see more object code in an oops report,
e.g. when kernel may be corrupted at runtime.

Add the "code_bytes" option for this.

Signed-off-by: Chuck Ebbert <cebbert@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] x86-64: Minor patch for compilation warning in x86_64 signal code
Evgeniy Polyakov [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86-64: Minor patch for compilation warning in x86_64 signal code

If DEBUG_SIG is enbaled in source code, ia32_signal.c compiles with warning
due to wrong format string.  Attached patch fixes that.  It is quite minor
update, since by default DEBUG_SIG is not enabled and can not be turned on
without code modification.

Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] x86-64: avoid warning message livelock
Roland Dreier [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86-64: avoid warning message livelock

I've seen my box paralyzed by an endless spew of

    rtc: lost some interrupts at 1024Hz.

messages on the serial console.  What seems to be happening is that
something real causes an interrupt to be lost and triggers the
message.  But then printing the message to the serial console (from
the hpet interrupt handler) takes more than 1/1024th of a second, and
then some more interrupts are lost, so the message triggers again....

Fix this by adding a printk_ratelimit() before printing the warning.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: update IO-APIC dest field to 8-bit for xAPIC
Benjamin Romer [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86-64: update IO-APIC dest field to 8-bit for xAPIC

On the Unisys ES7000/ONE system, we encountered a problem where performing
a kexec reboot or dump on any cell other than cell 0 causes the system
timer to stop working, resulting in a hang during timer calibration in the
new kernel.

We traced the problem to one line of code in disable_IO_APIC(), which needs
to restore the timer's IO-APIC configuration before rebooting.  The code is
currently using the 4-bit physical destination field, rather than using the
8-bit logical destination field, and it cuts off the upper 4 bits of the
timer's APIC ID.  If we change this to use the logical destination field,
the timer works and we can kexec on the upper cells.  This was tested on
two different cells (0 and 2) in an ES7000/ONE system.

For reference, the relevant Intel xAPIC spec is kept at
ftp://download.intel.com/design/chipsets/e8501/datashts/30962001.pdf,
specifically on page 334.

Signed-off-by: Benjamin M Romer <benjamin.romer@unisys.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] i386: Remove unused kernel config option X86_XADD
Robert P. J. Day [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] i386: Remove unused kernel config option X86_XADD

Remove the unused kernel config option X86_XADD, which is unused in any
source or header file.

Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] x86-64: clean up sparsemem memory_present call
Bob Picco [Tue, 13 Feb 2007 12:26:25 +0000 (13:26 +0100)]
[PATCH] x86-64: clean up sparsemem memory_present call

Eliminate arch specific memory_present call x86_64 NUMA by utilizing
sparse_memory_present_with_active_regions.

Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Bob Picco <bob.picco@hp.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] i386: entry.S END/ENDPROC annotations
Jan Beulich [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] i386: entry.S END/ENDPROC annotations

Annotate i386/kernel/entry.S with END/ENDPROC to assist disassemblers and
other analysis tools.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] i386: support Classic MediaGXm
takada [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] i386: support Classic MediaGXm

I hope to support "classic" MediaGXm in kernel.

The DIR1 register of MediaGXm( or Geode) shows the following values for
identify CPU.  For example, My MediaGXm shows 0x42.

We can read National Semiconductor's datasheet without any NDAs.
  http://www.national.com/pf/GX/GXLV.html

from datasheets:
DIR1
0x30 - 0x33 GXm rev. 1.0 - 2.3
0x34 - 0x4f GXm rev. 2.4 - 3.x
0x5x        GXm rev. 5.0 - 5.4
0x6x        GXLV
0x7x         (unknow)
0x8x     Gx1

In nsc driver of X, accept 0x30 through 0x82. What will 0x7x mean?

Cc: Jordan Crouse <jordan.crouse@amd.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: avoid gcc extension
Randy Dunlap [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] i386: avoid gcc extension

setcc() in math-emu is written as a gcc extension statement expression
macro that returns a value.  However, it's not used that way and it's not
needed like that, so just make it a inline function so that we
don't use an extension when it's not needed.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Segher Boessenkool <segher@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] i386: All Transmeta CPUs have constant TSCs
H. Peter Anvin [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] i386: All Transmeta CPUs have constant TSCs

All Transmeta CPUs ever produced have constant-rate TSCs.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] x86: fix laptop bootup hang in init_acpi()
Ingo Molnar [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86: fix laptop bootup hang in init_acpi()

During kernel bootup, a new T60 laptop (CoreDuo, 32-bit) hangs about
10%-20% of the time in acpi_init():

 Calling initcall 0xc055ce1a: topology_init+0x0/0x2f()
 Calling initcall 0xc055d75e: mtrr_init_finialize+0x0/0x2c()
 Calling initcall 0xc05664f3: param_sysfs_init+0x0/0x175()
 Calling initcall 0xc014cb65: pm_sysrq_init+0x0/0x17()
 Calling initcall 0xc0569f99: init_bio+0x0/0xf4()
 Calling initcall 0xc056b865: genhd_device_init+0x0/0x50()
 Calling initcall 0xc056c4bd: fbmem_init+0x0/0x87()
 Calling initcall 0xc056dd74: acpi_init+0x0/0x1ee()

It's a hard hang that not even an NMI could punch through!  Frustratingly,
adding printks or function tracing to the ACPI code made the hangs go away
...

After some time an additional detail emerged: disabling the NMI watchdog
made these occasional hangs go away.

So i spent the better part of today trying to debug this and trying out
various theories when i finally found the likely reason for the hang: if
acpi_ns_initialize_devices() executes an _INI AML method and an NMI
happens to hit that AML execution in the wrong moment, the machine would
hang.  (my theory is that this must be some sort of chipset setup method
doing stores to chipset mmio registers?)

Unfortunately given the characteristics of the hang it was sheer
impossible to figure out which of the numerous AML methods is impacted
by this problem.

As a workaround i wrote an interface to disable chipset-based NMIs while
executing _INI sections - and indeed this fixed the hang.  I did a
boot-loop of 100 separate reboots and none hung - while without the patch
it would hang every 5-10 attempts.  Out of caution i did not touch the
nmi_watchdog=2 case (it's not related to the chipset anyway and didnt
hang).

I implemented this for both x86_64 and i686, tested the i686 laptop both
with nmi_watchdog=1 [which triggered the hangs] and nmi_watchdog=2, and
tested an Athlon64 box with the 64-bit kernel as well. Everything builds
and works with the patch applied.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
17 years ago[PATCH] x86-64: robustify bad_dma_address handling
Muli Ben-Yehuda [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: robustify bad_dma_address handling

- set bad_dma_address explicitly to 0x0
- reserve 32 pages from bad_dma_address and up
- WARN_ON() a driver feeding us bad_dma_address

Thanks to Leo Duran <leo.duran@amd.com> for the suggestion.

Signed-off-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Leo Duran <leo.duran@amd.com>
Cc: Job Mason <jdmason@kudzu.us>
17 years ago[PATCH] x86-64: define dma noncoherent API functions
Jeff Garzik [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: define dma noncoherent API functions

x86-64 is missing these:

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Don't reserve ROMs
Andi Kleen [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: Don't reserve ROMs

We trust the e820 table, so explicitely reserving ROMs shouldn't
be needed.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Fix off by one error in IOMMU boundary checking
Andi Kleen [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: Fix off by one error in IOMMU boundary checking

Should be harmless because there is normally no memory there, but
technically it was incorrect.

Pointed out by Leo Duran

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: x86_64 - Fix FS/GS registers for VT execution
Zachary Amsden [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: x86_64 - Fix FS/GS registers for VT execution

Initialize FS and GS to __KERNEL_DS as well.  The actual value of them is not
important, but it is important to reload them in protected mode.  At this time,
they still retain the real mode values from initial boot.  VT disallows
execution of code under such conditions, which means hardware virtualization
can not be used to boot the kernel on Intel platforms, making the boot time
painfully slow.

This requires moving the GS load before the load of GS_BASE, so just move
all the segments loads there to keep them together in the code.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Unexport __supported_pte_mask
Andi Kleen [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: Unexport __supported_pte_mask

The symbol is needed to manipulate page tables, and modules shouldn't
do that.

Leftover from 2.4, but no in tree module should need it now.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Check return value of putreg in PTRACE_SETREGS
Andi Kleen [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: Check return value of putreg in PTRACE_SETREGS

This means if an illegal value is set for the segment registers there
ptrace will error out now with an errno instead of silently ignoring
it.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: - Ignore long SMI interrupts in clock calibration code - update 1
Jack Steiner [Tue, 13 Feb 2007 12:26:24 +0000 (13:26 +0100)]
[PATCH] x86-64: - Ignore long SMI interrupts in clock calibration code - update 1

Add failsafe mechanism to HPET/TSC clock calibration.

Signed-off-by: Jack Steiner <steiner@sgi.com>
Updated to include failsafe mechanism & additional community feedback.
Patch built on latest 2.6.20-rc4-mm1 tree.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: fix size_or_mask and size_and_mask
Andreas Herrmann [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] i386: fix size_or_mask and size_and_mask

mtrr: fix size_or_mask and size_and_mask

This fixes two bugs in /proc/mtrr interface:
o If physical address size crosses the 44 bit boundary
  size_or_mask is evaluated wrong.
o size_and_mask limits width of physical base
  address for an MTRR to be less than 44 bits.

TBD: later patch had one more change, but I think that was bogus.
TBD: need to double check

Signed-off-by: Andreas Herrmann <andreas.herrmann3@amd.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Convert /proc/apm to seqfile
Alexey Dobriyan [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] i386: Convert /proc/apm to seqfile

Byte-to-byte identical /proc/apm here.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Fix preprocessor condition
Josef 'Jeff' Sipek [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86-64: Fix preprocessor condition

Old code was legal standard C, but apparently not sparse-C.

Signed-off-by: Josef 'Jeff' Sipek <jsipek@cs.sunysb.edu>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: use smp_call_function_single()
Alexey Dobriyan [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] i386: use smp_call_function_single()

It will execure cpuid only on the cpu we need.

Signed-off-by: Alexey Dobriyan <adobriyan@openvz.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: use smp_call_function_single()
Alexey Dobriyan [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] i386: use smp_call_function_single()

It will execute rdmsr and wrmsr only on the cpu we need.

Signed-off-by: Alexey Dobriyan <adobriyan@openvz.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Kconfig typos
Nicolas Kaiser [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86-64: Kconfig typos

Some typos in Kconfig.

Signed-off-by: Nicolas Kaiser <nikai@nikai.net>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Small cleanup to TLB flush code
Andi Kleen [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] i386: Small cleanup to TLB flush code

- Remove outdated comment
- Use cpu_relax() in a busy loop

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: remove get_pmd()
Jan Beulich [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86-64: remove get_pmd()

Function is dead.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Allow to run a program when a machine check event is detected
Andi Kleen [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86-64: Allow to run a program when a machine check event is detected

When a machine check event is detected (including a AMD RevF threshold
overflow event) allow to run a "trigger" program. This allows user space
to react to such events sooner.

The trigger is configured using a new trigger entry in the
machinecheck sysfs interface. It is currently shared between
all CPUs.

I also fixed the AMD threshold handler to run the machine
check polling code immediately to actually log any events
that might have caused the threshold interrupt.

Also added some documentation for the mce sysfs interface.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Tighten mce_amd driver MSR reads
Jan Beulich [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86-64: Tighten mce_amd driver MSR reads

while debugging an unrelated problem in Xen, I noticed odd reads from
non-existent MSRs. Having now found time to look why these happen, I
came up with below patch, which
- prevents accessing MCi_MISCj with j > 0 when the block pointer in
MCi_MISC0 is zero
- accesses only contiguous MCi_MISCj until a non-implemented one is
found
- doesn't touch unimplemented blocks in mce_threshold_interrupt at all
- gives names to two bits previously derived from MASK_VALID_HI (it
took me some time to understand the code without this)

The first three items, besides being apparently closer to the spec, should
namely help cutting down on the time mce_threshold_interrupt() takes.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86: simplify notify_page_fault()
Jan Beulich [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86: simplify notify_page_fault()

Remove all parameters from this function that aren't really variable.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: list x86_64 quilt tree
Randy Dunlap [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86-64: list x86_64 quilt tree

List x86_64 quilt tree in MAINTAINERS.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: cleanup Doc/x86_64/ files
Randy Dunlap [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] x86-64: cleanup Doc/x86_64/ files

Fix typos.
Lots of whitespace changes for readability and consistency.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Handle 32 bit PerfMon Counter writes cleanly in oprofile
Venkatesh Pallipadi [Tue, 13 Feb 2007 12:26:23 +0000 (13:26 +0100)]
[PATCH] i386: Handle 32 bit PerfMon Counter writes cleanly in oprofile

Handle these 32 bit perfmon counter MSR writes cleanly in oprofile.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Handle 32 bit PerfMon Counter writes cleanly in i386 nmi_watchdog
Venkatesh Pallipadi [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] i386: Handle 32 bit PerfMon Counter writes cleanly in i386 nmi_watchdog

Change i386 nmi handler to handle 32 bit perfmon counter MSR writes cleanly.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Handle 32 bit PerfMon Counter writes cleanly in x86_64 nmi_watchdog
Venkatesh Pallipadi [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] x86-64: Handle 32 bit PerfMon Counter writes cleanly in x86_64 nmi_watchdog

P6 CPUs and Core/Core 2 CPUs which has 'architectural perf mon' feature,
only supports write of low 32 bits in Performance Monitoring Counters.
Bits 32..39 are sign extended based on bit 31 and bits 40..63 are reserved
and should be zero.

This patch:

Change x86_64 nmi handler to handle this case cleanly.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Use constant instead of raw number in x86_64 ioperm.c
Glauber de Oliveira Costa [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] x86-64: Use constant instead of raw number in x86_64 ioperm.c

This is a tiny cleanup to increase readability

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: Remove fastcall references in x86_64 code
Glauber de Oliveira Costa [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] x86-64: Remove fastcall references in x86_64 code

Unlike x86, x86_64 already passes arguments in registers.  The use of
regparm attribute makes no difference in produced code, and the use of
fastcall just bloats the code.

Signed-off-by: Glauber de Oliveira Costa <gcosta@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: Fix fake numa for x86_64 machines with big IO hole
Rohit Seth [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] x86-64: Fix fake numa for x86_64 machines with big IO hole

This patch resolves the issue of running with numa=fake=X on kernel command
line on x86_64 machines that have big IO hole.  While calculating the size
of each node now we look at the total hole size in that range.

Previously there were nodes that only had IO holes in them causing kernel
boot problems.  We now use the NODE_MIN_SIZE (64MB) as the minimum size of
memory that any node must have.  We reduce the number of allocated nodes if
the number of nodes specified on kernel command line results in any node
getting memory smaller than NODE_MIN_SIZE.

This change allows the extra memory to be incremented in NODE_MIN_SIZE
granule and uniformly distribute among as many nodes (called big nodes) as
possible.

[akpm@osdl.org: build fix]
Signed-off-by: David Rientjes <reintjes@google.com>
Signed-off-by: Paul Menage <menage@google.com>
Signed-off-by: Rohit Seth <rohitseth@google.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: romsignature/checksum cleanup
Rene Herman [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] i386: romsignature/checksum cleanup

Use adding __init to romsignature() (it's only called from probe_roms()
which is itself __init) as an excuse to submit a pedantic cleanup.

Signed-off-by: Rene Herman <rene.herman@gmail.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: improve sched_clock() on i686
Ingo Molnar [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] i386: improve sched_clock() on i686

Clean up sched_clock() on i686: it will use the TSC if available and falls
back to jiffies only if the user asked for it to be disabled via notsc or
the CPU calibration code didnt figure out the right cpu_khz.

This generally makes the scheduler timestamps more finegrained, on all
hardware.  (the current scheduler is pretty resistant against asynchronous
sched_clock() values on different CPUs, it will allow at most up to a jiffy
of jitter.)

Also simplify sched_clock()'s check for TSC availability: propagate the
desire and ability to use the TSC into the tsc_disable flag, previously
this flag only indicated whether the notsc option was passed.  This makes
the rare low-res sched_clock() codepath a single branch off a read-mostly
flag.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: add idle notifier
Stephane Eranian [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] i386: add idle notifier

Add a notifier mechanism to the low level idle loop.  You can register a
callback function which gets invoked on entry and exit from the low level idle
loop.  The low level idle loop is defined as the polling loop, low-power call,
or the mwait instruction.  Interrupts processed by the idle thread are not
considered part of the low level loop.

The notifier can be used to measure precisely how much is spent in useless
execution (or low power mode).  The perfmon subsystem uses it to turn on/off
monitoring.

Signed-off-by: stephane eranian <eranian@hpl.hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: arch/i386/kernel/cpu/mcheck/mce.c should #include <asm/mce.h>
Adrian Bunk [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] i386: arch/i386/kernel/cpu/mcheck/mce.c should #include <asm/mce.h>

Every file should include the headers containing the prototypes for
it's global functions.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] generic: Break init() in two parts to avoid MODPOST warnings
Vivek Goyal [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] generic: Break init() in two parts to avoid MODPOST warnings

o init() is a non __init function in .text section but it calls many
  functions which are in .init.text section. Hence MODPOST generates lots
  of cross reference warnings on i386 if compiled with CONFIG_RELOCATABLE=y

WARNING: vmlinux - Section mismatch: reference to .init.text:smp_prepare_cpus from .text between 'init' (at offset 0xc0101049) and 'rest_init'
WARNING: vmlinux - Section mismatch: reference to .init.text:migration_init from .text between 'init' (at offset 0xc010104e) and 'rest_init'
WARNING: vmlinux - Section mismatch: reference to .init.text:spawn_ksoftirqd from .text between 'init' (at offset 0xc0101053) and 'rest_init'

o This patch breaks down init() in two parts. One part which can go
  in .init.text section and can be freed and other part which has to
  be non __init(init_post()). Now init() calls init_post() and init_post()
  does not call any functions present in .init sections. Hence getting
  rid of warnings.

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: move startup_32() in text.head section
Vivek Goyal [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] i386: move startup_32() in text.head section

o Entry startup_32 was in .text section but it was accessing some init
  data too and it prompts MODPOST to generate compilation warnings.

WARNING: vmlinux - Section mismatch: reference to .init.data:boot_params from
.text between '_text' (at offset 0xc0100029) and 'startup_32_smp'
WARNING: vmlinux - Section mismatch: reference to .init.data:boot_params from
.text between '_text' (at offset 0xc0100037) and 'startup_32_smp'
WARNING: vmlinux - Section mismatch: reference to
.init.data:init_pg_tables_end from .text between '_text' (at offset
0xc0100099) and 'startup_32_smp'

o Can't move startup_32 to .init.text as this entry point has to be at the
  start of bzImage. Hence moved startup_32 to a new section .text.head and
  instructed MODPOST to not to generate warnings if init data is being
  accessed from .text.head section. This code has been audited.

o SMP boot up code (startup_32_smp) can go into .init.text if CPU hotplug
  is not supported. Otherwise it generates more warnings

WARNING: vmlinux - Section mismatch: reference to .init.data:new_cpu_data from
.text between 'checkCPUtype' (at offset 0xc0100126) and 'is486'
WARNING: vmlinux - Section mismatch: reference to .init.data:new_cpu_data from
.text between 'checkCPUtype' (at offset 0xc0100130) and 'is486'

Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Paravirt debug defaults off
Zachary Amsden [Tue, 13 Feb 2007 12:26:22 +0000 (13:26 +0100)]
[PATCH] i386: Paravirt debug defaults off

Deliberate register clobber around performance critical inline code is great for
testing, bad to leave on by default.  Many people ship with DEBUG_KERNEL turned
on, so stop making DEBUG_PARAVIRT default on.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Vmi timer race
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: Vmi timer race

Because timer code moves around, and we might eventually move our init to a
late_time_init hook, save and restore IRQs around this code because it is
definitely not interrupt safe.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Kprobe rpl fix
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: Kprobe rpl fix

Kprobes bugfix for paravirt compatibility - RPL on the CS when inserting
BPs must match running kernel.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
CC: Eric Biederman <ebiederm@xmission.com>
17 years ago[PATCH] i386: Profile pc badness
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: Profile pc badness

Profile_pc was broken when using paravirtualization because the
assumption the kernel was running at CPL 0 was violated, causing
bad logic to read a random value off the stack.

The only way to be in kernel lock functions is to be in kernel
code, so validate that assumption explicitly by checking the CS
value.  We don't want to be fooled by BIOS / APM segments and
try to read those stacks, so only match KERNEL_CS.

I moved some stuff in segment.h to make it prettier.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: vMI timer patches
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: vMI timer patches

VMI timer code.  It works by taking over the local APIC clock when APIC is
configured, which requires a couple hooks into the APIC code.  The backend
timer code could be commonized into the timer infrastructure, but there are
some pieces missing (stolen time, in particular), and the exact semantics of
when to do accounting for NO_IDLE need to be shared between different
hypervisors as well.  So for now, VMI timer is a separate module.

[Adrian Bunk: cleanups]

Subject: VMI timer patches
Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: vMI backend for paravirt-ops
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: vMI backend for paravirt-ops

Fairly straightforward implementation of VMI backend for paravirt-ops.

[Adrian Bunk: some cleanups]

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: SMP boot hook for paravirt
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: SMP boot hook for paravirt

Add VMI SMP boot hook.  We emulate a regular boot sequence and use the same
APIC IPI initiation, we just poke magic values to load into the CPU state when
the startup IPI is received, rather than having to jump through a real mode
trampoline.

This is all that was needed to get SMP to work.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: iOPL handling for paravirt guests
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: iOPL handling for paravirt guests

I found a clever way to make the extra IOPL switching invisible to
non-paravirt compiles - since kernel_rpl is statically defined to be zero
there, and only non-zero rpl kernel have a problem restoring IOPL, as popf
does not restore IOPL flags unless run at CPL-0.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: paravirt CPU hypercall batching mode
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: paravirt CPU hypercall batching mode

The VMI ROM has a mode where hypercalls can be queued and batched.  This turns
out to be a significant win during context switch, but must be done at a
specific point before side effects to CPU state are visible to subsequent
instructions.  This is similar to the MMU batching hooks already provided.
The same hooks could be used by the Xen backend to implement a context switch
multicall.

To explain a bit more about lazy modes in the paravirt patches, basically, the
idea is that only one of lazy CPU or MMU mode can be active at any given time.
 Lazy MMU mode is similar to this lazy CPU mode, and allows for batching of
multiple PTE updates (say, inside a remap loop), but to avoid keeping some
kind of state machine about when to flush cpu or mmu updates, we just allow
one or the other to be active.  Although there is no real reason a more
comprehensive scheme could not be implemented, there is also no demonstrated
need for this extra complexity.

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] MM: page allocation hooks for VMI backend
Zachary Amsden [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] MM: page allocation hooks for VMI backend

The VMI backend uses explicit page type notification to track shadow page
tables.  The allocation of page table roots is especially tricky.  We need to
clone the root for non-PAE mode while it is protected under the pgd lock to
correctly copy the shadow.

We don't need to allocate pgds in PAE mode, (PDPs in Intel terminology) as
they only have 4 entries, and are cached entirely by the processor, which
makes shadowing them rather simple.

For base page table level allocation, pmd_populate provides the exact hook
point we need.  Also, we need to allocate pages when splitting a large page,
and we must release pages before returning the page to any free pool.

Despite being required with these slightly odd semantics for VMI, Xen also
uses these hooks to determine the exact moment when page tables are created or
released.

AK: All nops for other architectures

Signed-off-by: Zachary Amsden <zach@vmware.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: arch/i386/kernel/e820.c should #include <asm/setup.h
Adrian Bunk [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] i386: arch/i386/kernel/e820.c should #include <asm/setup.h

Every file should #include the headers containing the prototypes for
its global functions.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: do not always end the stack trace with ULONG_MAX
Catalin Marinas [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] x86-64: do not always end the stack trace with ULONG_MAX

It makes more sense to end the stack trace with ULONG_MAX only if
nr_entries < max_entries.  Otherwise, we lose one entry in the long stack
traces and cannot know whether the trace was complete or not.

Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: improved iommu documentation
Karsten Weiss [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] x86-64: improved iommu documentation

- add SWIOTLB config help text
- mention Documentation/x86_64/boot-options.txt in
  Documentation/kernel-parameters.txt
- remove the duplication of the iommu kernel parameter documentation.
- Better explanation of some of the iommu kernel parameter options.
- "32MB<<order" instead of "32MB^order".
- Mention the default "order" value.
- list the four existing PCI-DMA mapping implementations of arch x86_64
- group the iommu= option keywords by PCI-DMA mapping implementation.
- Distinguish iommu= option keywords from number arguments.
- Explain the meaning of DAC and SAC.

Signed-off-by: Karsten Weiss <knweiss@science-computing.de>
Signed-off-by: Andi Kleen <ak@suse.de>
Acked-by: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: get rid of ARCH_HAVE_XTIME_LOCK
Eric Dumazet [Tue, 13 Feb 2007 12:26:21 +0000 (13:26 +0100)]
[PATCH] x86-64: get rid of ARCH_HAVE_XTIME_LOCK

ARCH_HAVE_XTIME_LOCK is used by x86_64 arch .  This arch needs to place a
read only copy of xtime_lock into vsyscall page.  This read only copy is
named __xtime_lock, and xtime_lock is defined in
arch/x86_64/kernel/vmlinux.lds.S as an alias.  So the declaration of
xtime_lock in kernel/timer.c was guarded by ARCH_HAVE_XTIME_LOCK define,
defined to true on x86_64.

We can get same result with _attribute__((weak)) in the declaration. linker
should do the job.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] mmconfig: Move e820 check into pci_mmcfg_reject_broken()
OGAWA Hirofumi [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: Move e820 check into pci_mmcfg_reject_broken()

This is just cleanup. It moves to e820 check into pci_mmcfg_reject_broken().

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] mmconfig: fix unreachable_devices()
OGAWA Hirofumi [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: fix unreachable_devices()

Currently, unreachable_devices() compares value of mmconfig and value
of conf1. But it doesn't check the device is reachable or not.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] mmconfig: minor cleanup in mmconfig code
OGAWA Hirofumi [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: minor cleanup in mmconfig code

This just cleans up.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] mmconfig: remove #define MMCONFIG_APER_XXX
OGAWA Hirofumi [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: remove #define MMCONFIG_APER_XXX

MMCONFIG_APER_XXX is unneeded in arch/x86_64/pci/mmconfig.c.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] mmconfig: Reject a broken MCFG tables on Asus etc
OGAWA Hirofumi [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: Reject a broken MCFG tables on Asus etc

This rejects broken MCFG tables on Asus. When the table
looks bogus just disable mmconfig

Arjan and Andi suggested this.

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] mmconfig: Fix x86_64 ioremap base_address
OGAWA Hirofumi [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: Fix x86_64 ioremap base_address

Current mmconfig has some problems of remapped range.

a) In the case of broken MCFG tables on Asus etc., we need to remap 256M
   range, but currently only remap 1M.

b) The base address always corresponds to bus number 0, but currently we
   are assuming it corresponds to start bus number.

This patch fixes the above problems.

(akpm: Arjan suggests that if the MCFG table is broken we just shouldn't use
it, rather than try to work around things).

Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] mmconfig: Reserve resources but only when we're sure about them.
Olivier Galibert [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: Reserve resources but only when we're sure about them.

Put back the resource reservation as per
4c6e052adfe285ede5884e4e8c4d33af33932c13 but use it *only* when the range(s)
come from a chipset probe instead of the bios.

Signed-off-by: Olivier Galibert <galibert@pobox.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] mmconfig: Detect and support the E7520 and the 945G/GZ/P/PL
Olivier Galibert [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: Detect and support the E7520 and the 945G/GZ/P/PL

It seems that the only way to reliably support mmconfig in the presence of
funky biosen is to detect the hostbridge and read where the window is mapped
from its registers.  Do that for the E7520 and the 945G/GZ/P/PL for a start.

Signed-off-by: Olivier Galibert <galibert@pobox.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: Only call unreachable_devices() when type 1 is available.
Olivier Galibert [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] i386: Only call unreachable_devices() when type 1 is available.

unreachable_devices compares between the results of pci configuration accesses
through type1 and mmconfig, so it should be called only if type1 actually
works in the first place.

Signed-off-by: Olivier Galibert <galibert@pobox.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] mmconfig: Share parts of mmconfig code between i386 and x86-64
Olivier Galibert [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] mmconfig: Share parts of mmconfig code between i386 and x86-64

i386 and x86-64 pci mmconfig code have a lot in common.  So share what's
shareable between the two.

Signed-off-by: Olivier Galibert <galibert@pobox.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: Fix a typo in an IRQ handler name
Maciej W. Rozycki [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] i386: Fix a typo in an IRQ handler name

The "fasteoi" IRQ handler is named "fasteio" incorrectly.  This is a fix.

Signed-off-by: Maciej W. Rozycki <macro@linux-mips.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] i386: Convert i386 PDA code to use %fs
Jeremy Fitzhardinge [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] i386: Convert i386 PDA code to use %fs

Convert the PDA code to use %fs rather than %gs as the segment for
per-processor data.  This is because some processors show a small but
measurable performance gain for reloading a NULL segment selector (as %fs
generally is in user-space) versus a non-NULL one (as %gs generally is).

On modern processors the difference is very small, perhaps undetectable.
Some old AMD "K6 3D+" processors are noticably slower when %fs is used
rather than %gs; I have no idea why this might be, but I think they're
sufficiently rare that it doesn't matter much.

This patch also fixes the math emulator, which had not been adjusted to
match the changed struct pt_regs.

[frederik.deweerdt@gmail.com: fixit with gdb]
[mingo@elte.hu: Fix KVM too]

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Ian Campbell <Ian.Campbell@XenSource.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Zachary Amsden <zach@vmware.com>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Frederik Deweerdt <frederik.deweerdt@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: x86_64-make-the-numa-hash-function-nodemap-allocation fix fix
Amul Shah [Tue, 13 Feb 2007 12:26:20 +0000 (13:26 +0100)]
[PATCH] x86-64: x86_64-make-the-numa-hash-function-nodemap-allocation fix fix

- Removed an extraneous debug message from allocate_cachealigned_map

- Changed extract_lsb_from_nodes to return 63 for the case where there was
  only one memory node.  The prevents the creation of the dynamic hashmap.

- Changed extract_lsb_from_nodes to use only the starting memory address of
  a node.  On an ES7000, our nodes overlap the starting and ending address,
  meaning, that we see nodes like

00000 - 10000
10000 - 20000

  But other systems have nodes whose start and end addresses do not overlap.
   For example:

00000 - 0FFFF
10000 - 1FFFF

  In this case, using the ending address will result in an LSB much lower
  than what is possible.  In this case an LSB of 1 when in reality it should
  be 16.

Cc: Andi Kleen <ak@suse.de>
Cc: Rohit Seth <rohitseth@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Allocate the NUMA hash function nodemap dynamically
Amul Shah [Tue, 13 Feb 2007 12:26:19 +0000 (13:26 +0100)]
[PATCH] x86-64: Allocate the NUMA hash function nodemap dynamically

Remove the statically allocated memory to NUMA node hash map in favor of a
dynamically allocated memory to node hash map (it is cache aligned).

This patch has the nice side effect in that it allows the hash map to grow
for systems with large amounts of memory (256GB - 1TB), but suffer from
having small PCI space tacked onto the boot node (which is somewhere
between 192MB to 512MB on the ES7000).

Signed-off-by: Amul Shah <amul.shah@unisys.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Rohit Seth <rohitseth@google.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
17 years ago[PATCH] x86-64: Add __copy_from_user_nocache
Andi Kleen [Tue, 13 Feb 2007 12:26:19 +0000 (13:26 +0100)]
[PATCH] x86-64: Add __copy_from_user_nocache

This does user copies in fs write() into the page cache with write combining.
This pushes the destination out of the CPU's cache, but allows higher bandwidth
in some case.

The theory is that the page cache data is usually not touched by the
CPU again and it's better to not pollute the cache with it. Also it is a little
faster.

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] i386: Update defconfig
Andi Kleen [Tue, 13 Feb 2007 12:26:19 +0000 (13:26 +0100)]
[PATCH] i386: Update defconfig

Signed-off-by: Andi Kleen <ak@suse.de>
17 years ago[PATCH] x86-64: Update defconfig
Andi Kleen [Tue, 13 Feb 2007 12:26:19 +0000 (13:26 +0100)]
[PATCH] x86-64: Update defconfig

Signed-off-by: Andi Kleen <ak@suse.de>
17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6
Linus Torvalds [Mon, 12 Feb 2007 23:35:30 +0000 (15:35 -0800)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6:
  [SPARC]: Re-export saved_command_line to modules.
  [SPARC64]: Increase command line size to 2048 like other arches.
  [SPARC64]: We do not need ZONE_DMA.

17 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Mon, 12 Feb 2007 23:34:17 +0000 (15:34 -0800)]
Merge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6: (25 commits)
  [XFRM]: Fix OOPSes in xfrm_audit_log().
  [TCP]: cleanup of htcp (resend)
  [TCP]: Use read mostly for CUBIC parameters.
  [NETFILTER]: nf_conntrack_tcp: make sysctl variables static
  [NETFILTER]: ip6t_mh: drop piggyback payload packet on MH packets
  [NETFILTER]: Fix whitespace errors
  [NETFILTER]: Kconfig: improve dependency handling
  [NETFILTER]: xt_mac/xt_CLASSIFY: use IPv6 hook names for IPv6 registration
  [NETFILTER]: nf_conntrack: change nf_conntrack_l[34]proto_unregister to void
  [NETFILTER]: nf_conntrack: properly use RCU for nf_conntrack_destroyed callback
  [NETFILTER]: ip_conntrack: properly use RCU for ip_conntrack_destroyed callback
  [NETFILTER]: nf_conntrack: fix invalid conntrack statistics RCU assumption
  [NETFILTER]: ip_conntrack: fix invalid conntrack statistics RCU assumption
  [NETFILTER]: nf_conntrack: properly use RCU API for nf_ct_protos/nf_ct_l3protos arrays
  [NETFILTER]: ip_conntrack: properly use RCU API for ip_ct_protos array
  [NETFILTER]: nf_nat: properly use RCU API for nf_nat_protos array
  [NETFILTER]: ip_nat: properly use RCU API for ip_nat_protos array
  [NETFILTER]: nf_log: minor cleanups
  [NETFILTER]: nf_log: switch logger registration/unregistration to mutex
  [NETFILTER]: nf_log: make nf_log_unregister_pf return void
  ...

17 years ago[SPARC]: Re-export saved_command_line to modules.
David S. Miller [Mon, 12 Feb 2007 23:10:56 +0000 (15:10 -0800)]
[SPARC]: Re-export saved_command_line to modules.

This reverts some bogosity from the dynamic command-line
changes made on sparc32 and sparc64.

Drivers such as drivers/sbus/char/openprom.c reference
saved_command_line, and can be modular.

The boot_command_line is __initdata, yet the dynamic command-line
changes add modular exports of that symbol, obviously wrong.

Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[SPARC64]: Increase command line size to 2048 like other arches.
David S. Miller [Mon, 12 Feb 2007 19:01:21 +0000 (11:01 -0800)]
[SPARC64]: Increase command line size to 2048 like other arches.

Signed-off-by: David S. Miller <davem@davemloft.net>
17 years ago[SPARC64]: We do not need ZONE_DMA.
David S. Miller [Mon, 12 Feb 2007 08:13:31 +0000 (00:13 -0800)]
[SPARC64]: We do not need ZONE_DMA.

Signed-off-by: David S. Miller <davem@davemloft.net>
17 years agodon't use 'localversion*' files twice
Linus Torvalds [Mon, 12 Feb 2007 23:05:15 +0000 (15:05 -0800)]
don't use 'localversion*' files twice

Since we look in both source and object directories for localversion*
files, we accidentally ended up getting them twice.  Use 'sort -u' to
avoid that.

Reported-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>