This patch will fix a boot memory reservation bug that trashes memory on
the ES7000 when loading the kdump crash kernel.
The code in arch/x86_64/kernel/setup.c to reserve boot memory for the crash
kernel uses the non-numa aware "reserve_bootmem" function instead of the
NUMA aware "reserve_bootmem_generic". I checked to make sure that no other
function was using "reserve_bootmem" and found none, except the ones that
had NUMA ifdef'ed out.
I have tested this patch only on an ES7000 with NUMA on and off (numa=off)
in a single (non-NUMA) and multi-cell (NUMA) configurations.
Signed-off-by: Amul Shah <amul.shah@unisys.com>
Looks-good-to: Vivek Goyal <vgoyal@in.ibm.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
* 'release' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6: (65 commits)
ACPI: suppress power button event on S3 resume
ACPI: resolve merge conflict between sem2mutex and processor_perflib.c
ACPI: use for_each_possible_cpu() instead of for_each_cpu()
ACPI: delete newly added debugging macros in processor_perflib.c
ACPI: UP build fix for bugzilla-5737
Enable P-state software coordination via _PDC
P-state software coordination for speedstep-centrino
P-state software coordination for acpi-cpufreq
P-state software coordination for ACPI core
ACPI: create acpi_thermal_resume()
ACPI: create acpi_fan_suspend()/acpi_fan_resume()
ACPI: pass pm_message_t from acpi_device_suspend() to root_suspend()
ACPI: create acpi_device_suspend()/acpi_device_resume()
ACPI: replace spin_lock_irq with mutex for ec poll mode
ACPI: Allow a WAN module enable/disable on a Thinkpad X60.
sem2mutex: acpi, acpi_link_lock
ACPI: delete unused acpi_bus_drivers_lock
sem2mutex: drivers/acpi/processor_perflib.c
ACPI add ia64 exports to build acpi_memhotplug as a module
ACPI: asus_acpi_init(): propagate correct return value
...
Manual resolve of conflicts in:
arch/i386/kernel/cpu/cpufreq/acpi-cpufreq.c
arch/i386/kernel/cpu/cpufreq/speedstep-centrino.c
include/acpi/processor.h
flush_tlb_all uses on_each_cpu, which will disable/enable interrupt.
In suspend/resume time, this will make interrupt wrongly enabled.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Pages (Reserved/ACPI NVS/ACPI Data) below end_pfn will be saved/restored by S4
currently. We should mark 'Reserved' pages not saveable.
Pages (Reserved/ACPI NVS/ACPI Data) above end_pfn will not be saved/restored
by S4 currently. We should save the 'ACPI NVS/ACPI Data' pages.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Nigel Cunningham <nigel@suspend2.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: "Andy Currid" <ACurrid@nvidia.com>
This patch fixes a kernel panic during boot that occurs on NVIDIA platforms
that have HPET enabled.
When HPET is enabled, the standard timer IRQ is routed to IOAPIC pin 2 and is
advertised as such in the ACPI APIC table - but an earlier workaround in the
kernel was ignoring this override. The fix is to honor timer IRQ overrides
from ACPI when HPET is detected on an NVIDIA platform.
Signed-off-by: Andy Currid <acurrid@nvidia.com>
Cc: "Brown, Len" <len.brown@intel.com>
Cc: "Yu, Luming" <luming.yu@intel.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
int_ret_from_syscall already does syscall exit tracing, so
no need to do it again in the caller.
This caused problems for UML and some other special programs doing
syscall interception.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: Robert Hentosh <robert_hentosh@dell.com>
Actually, we just stumbled on a different bug found in find_e820_area() in
e820.c. The following code does not handle the edge condition correctly:
while (bad_addr(&addr, size) && addr+size < ei->addr + ei->size)
;
last = addr + size;
if ( last > ei->addr + ei->size )
continue;
The second statement in the while loop needs to be a <= b so that it is the
logical negavite of the if (a > b) outside it. It needs to read:
while (bad_addr(&addr, size) && addr+size <= ei->addr + ei->size)
;
In the case that failed bad_addr was returning an address that is exactly size
bellow the end of the e820 range.
AK: Again together with the earlier avoid edma fix this fixes
boot on a Dell PE6850/16GB
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: Daniel Yeisley <dan.yeisley@unisys.com>
It is possible to boot a Unisys ES7000 with CPUs from multiple cells, and not
also include the memory from those cells. This can create a scenario where
node 0 has cpus, but no associated memory. The system will boot fine in a
configuration where node 0 has memory, but nodes 2 and 3 do not.
[AK: I rechecked the code and generic code seems to indeed handle that already.
Dan's original patch had a change for mm/slab.c that seems to be already in now.]
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: "Jan Beulich" <jbeulich@novell.com>
The PM timer code updates vxtime.last_tsc, but this update was done
incorrectly in two ways:
- offset_delay being in microseconds requires multiplying with cpu_mhz
rather than cpu_khz
- the multiplication of offset_delay and cpu_khz (both being 32-bit
values) on most current CPUs would overflow (observed value of the
delay was approximately 4000us, yielding an overflow for frequencies
starting a little above 1GHz)
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Complaining about the IOMMU not compiled in doesn't make sense
here because it is clearly compiled in.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Problem:
If we put a probe onto a callq instruction and the probe is executed,
kernel panic of Bad RIP value occurs.
Root cause:
If resume_execution() found 0xff at first byte of p->ainsn.insn, it must
check the _second_ byte. But current resume_execution check _first_ byte
again.
I changed it checks second byte of p->ainsn.insn.
Kprobes on i386 don't have this problem, because the implementation is a
little bit different from x86_64.
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Satoshi Oshima <soshima@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Extends an earlier patch from John Blackwood to more exception handlers
that also run on the exception stacks.
Expand the use of preempt_conditional_{sti,cli} to all cases where
interrupts are to be re-enabled during exception handling while running
on an IST stack.
Based on original patch from Jan Beulich.
Cc: John Blackwood <john.blackwood@ccur.com>
Cc: jbeulich@novell.com
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This triggers for b44's 1GB DMA workaround which tries to map
first and then bounces.
The 32bit heuristic is reasonable because the IOMMU doesn't attempt
to handle < 32bit masks anyways.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Based on analysis&patch from Robert Hentosch
Observed on a Dell PE6850 with 16GB
The problem occurs very early on, when the kernel allocates space for the
temporary memory map called bootmap. The bootmap overlaps the EBDA region.
EBDA region is not historically reserved in the e820 mapping. When the
bootmap is freed it marks the EBDA region as usable.
If you notice in setup.c there is already code to work around the EBDA
in reserve_ebda_region(), this check however occurs after the bootmap
is allocated and doesn't prevent the bootmap from using this range.
AK: I redid the original patch. Thanks also to Jan Beulich for
spotting some mistakes.
Cc: Robert_Hentosch@dell.com
Cc: jbeulich@novell.com
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Playing with NMI watchdog on x86_64, I discovered that it didn't
do what I expected. It always panic-ed, even when it didn't
happen from interrupt context. This patch solves that
problem for me. Also, in this case, do_exit() will be called
with interrupts disabled, I believe. Would it be wise to also
call local_irq_enable() after nmi_exit()?
[Yes I added it -AK]
Currently, on x86_64, any NMI watchdog timeout will cause a panic
because the irq count will always be set to be in an interrupt
when do_exit() is called from die_nmi(). If we add nmi_exit() to
the die_nmi() call (since the nmi will never exit "normally")
it seems to solve this problem. The following small program
can be used to trigger the NMI watchdog to reproduce this:
main ()
{
iopl(3);
for (;;) asm("cli");
}
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I noticed this when poking around in this area.
The oops_begin() function in x86_64 would only conditionally claim
the die_lock if the call is nested, but oops_end() would always
release the spinlock. This patch adds a nest count for the die lock
so that the release of the lock is only done on the final oops_end().
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The IOMMU code can only deal with 8 northbridges. Error out when
more are found.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The patch addresses a problem with ACPI SCI interrupt entry, which gets
re-used, and the IRQ is assigned to another unrelated device. The patch
corrects the code such that SCI IRQ is skipped and duplicate entry is
avoided. Second issue came up with VIA chipset, the problem was caused by
original patch assigning IRQs starting 16 and up. The VIA chipset uses
4-bit IRQ register for internal interrupt routing, and therefore cannot
handle IRQ numbers assigned to its devices. The patch corrects this
problem by allowing PCI IRQs below 16.
Cc: len.brown@intel.com
Signed-off by: Natalie Protasevich <Natalie.Protasevich@unisys.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
* 'audit.b10' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current:
[PATCH] Audit Filter Performance
[PATCH] Rework of IPC auditing
[PATCH] More user space subject labels
[PATCH] Reworked patch for labels on user space messages
[PATCH] change lspp ipc auditing
[PATCH] audit inode patch
[PATCH] support for context based audit filtering, part 2
[PATCH] support for context based audit filtering
[PATCH] no need to wank with task_lock() and pinning task down in audit_syscall_exit()
[PATCH] drop task argument of audit_syscall_{entry,exit}
[PATCH] drop gfp_mask in audit_log_exit()
[PATCH] move call of audit_free() into do_exit()
[PATCH] sockaddr patch
[PATCH] deal with deadlocks in audit_free()
The PC Speaker driver's ->probe() routine doesn't even get called in the
64-bit kernels. The reason for that is that the arch code apparently has
to explictly add a "pcspkr" platform device in order for the driver core to
call the ->probe() routine. arch/i386/kernel/setup.c unconditionally adds
a "pcspkr" device, but the x86_64 kernel has no code at all related to the
PC Speaker.
The patch below copies the relevant code from i386 to x86_64, which makes
the PC Speaker work for me on x86_64.
Cc: Dmitry Torokhov <dtor_core@ameritech.net>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Few of the notifier_chain_register() callers use __init in the definition
of notifier_call. It is incorrect as the function definition should be
available after the initializations (they do not unregister them during
initializations).
This patch fixes all such usages to _not_ have the notifier_call __init
section.
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
We do this by removing a micro-optimization that tries to avoid grabbing
the iommu_bitmap_lock spinlock and using a bus-locked operation.
This still races with other simultaneous alloc_iommu or free_iommu(size >
1) which both use bus-unlocked operations.
The end result of this race is eventually ending up with an
iommu_gart_bitmap that has bits errornously set all over, making large
contiguous iommu space allocations fail with 'PCI-DMA: Out of IOMMU space'.
Signed-off-by: Mike Waychison <mikew@google.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
AMD K7/K8 CPUs only save/restore the FOP/FIP/FDP x87 registers in FXSAVE
when an exception is pending. This means the value leak through
context switches and allow processes to observe some x87 instruction
state of other processes.
This was actually documented by AMD, but nobody recognized it as
being different from Intel before.
The fix first adds an optimization: instead of unconditionally
calling FNCLEX after each FXSAVE test if ES is pending and skip
it when not needed. Then do a x87 load from a kernel variable to
clear FOP/FIP/FDP.
This means other processes always will only see a constant value
defined by the kernel in their FP state.
I took some pain to make sure to chose a variable that's already
in L1 during context switch to make the overhead of this low.
Also alternative() is used to patch away the new code on CPUs
who don't need it.
Patch for both i386/x86-64.
The problem was discovered originally by Jan Beulich. Richard
Brunner provided the basic code for the workarounds, with contribution
from Jan.
This is CVE-2006-1056
Cc: richard.brunner@amd.com
Cc: jbeulich@novell.com
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Andrew Morton pointed out that compiler might not inline the functions
marked for inline in kprobes. There-by allowing the insertion of probes
on these kprobes routines, which might cause recursion.
This patch removes all such inline and adds them to kprobes section
there by disallowing probes on all such routines. Some of the routines
can even still be inlined, since these routines gets executed after the
kprobes had done necessay setup for reentrancy.
Signed-off-by: Prasanna S Panchamukhi <prasanna@in.ibm.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
o Start booting into the capture kernel after an Oops if system is in a
unrecoverable state. System will boot into the capture kernel, if one is
pre-loaded by the user, and capture the kernel core dump.
o One of the following conditions should be true to trigger the booting of
capture kernel.
- panic_on_oops is set.
- pid of current thread is 0
- pid of current thread is 1
- Oops happened inside interrupt context.
Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
dmi_scan.c is arch-independent and is used by i386, x86_64, and ia64.
Currently all three arches compile it from arch/i386, which means that ia64
and x86_64 depend on things in arch/i386 that they wouldn't otherwise care
about.
This is simply "mv arch/i386/kernel/dmi_scan.c drivers/firmware/" (removing
trailing whitespace) and the associated Makefile changes. All three
architectures already set CONFIG_DMI in their top-level Kconfig files.
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Andi Kleen <ak@muc.de>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Andrey Panin <pazke@orbita1.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As pointed out by Linus it is useless now because entry.S should
handle it correctly in all cases.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
While cleaning up parisc_ksyms.c earlier, I noticed that strpbrk wasn't
being exported from lib/string.c. Investigating further, I noticed a
changeset that removed its export and added it to _ksyms.c on a few more
architectures. The justification was that "other arches do it."
I think this is wrong, since no architecture currently defines
__HAVE_ARCH_STRPBRK, there's no reason for any of them to be exporting it
themselves. Therefore, consolidate the export to lib/string.c.
Signed-off-by: Kyle McMartin <kyle@parisc-linux.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
In linux-2.6.16, we have noticed a problem where the gs base value
returned from an arch_prtcl(ARCH_GET_GS, ...) call will be incorrect if:
- the current/calling task has NOT set its own gs base yet to a
non-zero value,
- some other task that ran on the same processor previously set their
own gs base to a non-zero value.
In this situation, the ARCH_GET_GS code will read and return the
MSR_KERNEL_GS_BASE msr register.
However, since the __switch_to() code does NOT load/zero the
MSR_KERNEL_GS_BASE register when the task that is switched IN has a zero
next->gs value, the caller of arch_prctl(ARCH_GET_GS, ...) will get back
the value of some previous tasks's gs base value instead of 0.
Change the arch_prctl() ARCH_GET_GS code to only read and return
the MSR_KERNEL_GS_BASE msr register if the 'gs' register of the calling
task is non-zero.
Side note: Since in addition to using arch_prctl(ARCH_SET_GS, ...),
a task can also setup a gs base value by using modify_ldt() and write
an index value into 'gs' from user space, the patch below reads
'gs' instead of using thread.gs, since in the modify_ldt() case,
the thread.gs value will be 0, and incorrect value would be returned
(the task->thread.gs value).
When the user has not set its own gs base value and the 'gs'
register is zero, then the MSR_KERNEL_GS_BASE register will not be
read and a value of zero will be returned by reading and returning
'task->thread.gs'.
The first patch shown below is an attempt at implementing this
approach.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
If the HPET timer is enabled, the clock can drift by ~3 seconds a day.
This is due to the HPET timer not being initialized with the correct
setting (still using PIT count).
If HZ changes, this drift can become even more pronounced.
HPET patch initializes tick_nsec with correct tick_nsec settings for
HPET timer.
Vojtech comments:
"It's not entirely correct (it assumes the HPET ticks totally
exactly), but it's significantly better than assuming the PIT error
there."
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Intel EM64T CPUs handle uncanonical return addresses differently
from AMD CPUs.
The exception is reported in the SYSRET, not the next instruction.
This leads to the kernel exception handler running on the user stack
with the wrong GS because the kernel didn't expect exceptions
on this instruction.
This version of the patch has the teething problems that plagued an earlier
version fixed.
This is CVE-2006-0744
Thanks to Ernie Petrides and Asit B. Mallick for analysis and initial
patches.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Machine checks can stall the machine for a long time and
it's not good to trigger the nmi watchdog during that.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Needed for other checks later in ACPI.
Pointed out by Len Brown
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Introduce a e820_all_mapped() function which checks if the entire range
<start,end> is mapped with type.
This is done by moving the local start variable to the end of each
known-good region; if at the end of the function the start address is
still before end, there must be a part that's not of the correct type;
otherwise it's a good region.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Rename e820_mapped to e820_any_mapped since it tests if any part of the
range is mapped according to the type.
Later steps will introduce e820_all_mapped which will check if the
entire range is mapped with the type. Both have their merit.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Memory hotadd doesn't need SPARSEMEM, but can be handled by just preallocating
mem_maps. This only needs some untangling of ifdefs to enable the necessary
code even without SPARSEMEM.
Originally from Keith Mannthey, hacked by AK.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Just call IRET always, no need for any special cases.
Needed for the next bug fix.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The only user of get_wchan is the proc fs - and proc can't be built modular.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The boot cmdline is parsed in parse_early_param() and
parse_args(,unknown_bootoption).
And __setup() is used in obsolete_checksetup().
start_kernel()
-> parse_args()
-> unknown_bootoption()
-> obsolete_checksetup()
If __setup()'s callback (->setup_func()) returns 1 in
obsolete_checksetup(), obsolete_checksetup() thinks a parameter was
handled.
If ->setup_func() returns 0, obsolete_checksetup() tries other
->setup_func(). If all ->setup_func() that matched a parameter returns 0,
a parameter is seted to argv_init[].
Then, when runing /sbin/init or init=app, argv_init[] is passed to the app.
If the app doesn't ignore those arguments, it will warning and exit.
This patch fixes a wrong usage of it, however fixes obvious one only.
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The kernel's implementation of notifier chains is unsafe. There is no
protection against entries being added to or removed from a chain while the
chain is in use. The issues were discussed in this thread:
http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2
We noticed that notifier chains in the kernel fall into two basic usage
classes:
"Blocking" chains are always called from a process context
and the callout routines are allowed to sleep;
"Atomic" chains can be called from an atomic context and
the callout routines are not allowed to sleep.
We decided to codify this distinction and make it part of the API. Therefore
this set of patches introduces three new, parallel APIs: one for blocking
notifiers, one for atomic notifiers, and one for "raw" notifiers (which is
really just the old API under a new name). New kinds of data structures are
used for the heads of the chains, and new routines are defined for
registration, unregistration, and calling a chain. The three APIs are
explained in include/linux/notifier.h and their implementation is in
kernel/sys.c.
With atomic and blocking chains, the implementation guarantees that the chain
links will not be corrupted and that chain callers will not get messed up by
entries being added or removed. For raw chains the implementation provides no
guarantees at all; users of this API must provide their own protections. (The
idea was that situations may come up where the assumptions of the atomic and
blocking APIs are not appropriate, so it should be possible for users to
handle these things in their own way.)
There are some limitations, which should not be too hard to live with. For
atomic/blocking chains, registration and unregistration must always be done in
a process context since the chain is protected by a mutex/rwsem. Also, a
callout routine for a non-raw chain must not try to register or unregister
entries on its own chain. (This did happen in a couple of places and the code
had to be changed to avoid it.)
Since atomic chains may be called from within an NMI handler, they cannot use
spinlocks for synchronization. Instead we use RCU. The overhead falls almost
entirely in the unregister routine, which is okay since unregistration is much
less frequent that calling a chain.
Here is the list of chains that we adjusted and their classifications. None
of them use the raw API, so for the moment it is only a placeholder.
ATOMIC CHAINS
-------------
arch/i386/kernel/traps.c: i386die_chain
arch/ia64/kernel/traps.c: ia64die_chain
arch/powerpc/kernel/traps.c: powerpc_die_chain
arch/sparc64/kernel/traps.c: sparc64die_chain
arch/x86_64/kernel/traps.c: die_chain
drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list
kernel/panic.c: panic_notifier_list
kernel/profile.c: task_free_notifier
net/bluetooth/hci_core.c: hci_notifier
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain
net/ipv6/addrconf.c: inet6addr_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain
net/netlink/af_netlink.c: netlink_chain
BLOCKING CHAINS
---------------
arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain
arch/s390/kernel/process.c: idle_chain
arch/x86_64/kernel/process.c idle_notifier
drivers/base/memory.c: memory_chain
drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list
drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list
drivers/macintosh/adb.c: adb_client_list
drivers/macintosh/via-pmu.c sleep_notifier_list
drivers/macintosh/via-pmu68k.c sleep_notifier_list
drivers/macintosh/windfarm_core.c wf_client_list
drivers/usb/core/notify.c usb_notifier_list
drivers/video/fbmem.c fb_notifier_list
kernel/cpu.c cpu_chain
kernel/module.c module_notify_list
kernel/profile.c munmap_notifier
kernel/profile.c task_exit_notifier
kernel/sys.c reboot_notifier_list
net/core/dev.c netdev_chain
net/decnet/dn_dev.c: dnaddr_chain
net/ipv4/devinet.c: inetaddr_chain
It's possible that some of these classifications are wrong. If they are,
please let us know or submit a patch to fix them. Note that any chain that
gets called very frequently should be atomic, because the rwsem read-locking
used for blocking chains is very likely to incur cache misses on SMP systems.
(However, if the chain's callout routines may sleep then the chain cannot be
atomic.)
The patch set was written by Alan Stern and Chandra Seetharaman, incorporating
material written by Keith Owens and suggestions from Paul McKenney and Andrew
Morton.
[jes@sgi.com: restructure the notifier chain initialization macros]
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add a new sched domain for representing multi-core with shared caches
between cores. Consider a dual package system, each package containing two
cores and with last level cache shared between cores with in a package. If
there are two runnable processes, with this appended patch those two
processes will be scheduled on different packages.
On such systems, with this patch we have observed 8% perf improvement with
specJBB(2 warehouse) benchmark and 35% improvement with CFP2000 rate(with 2
users).
This new domain will come into play only on multi-core systems with shared
caches. On other systems, this sched domain will be removed by domain
degeneration code. This new domain can be also used for implementing power
savings policy (see OLS 2005 CMP kernel scheduler paper for more details..
I will post another patch for power savings policy soon)
Most of the arch/* file changes are for cpu_coregroup_map() implementation.
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
