After programming the HPET, we do a readback as a workaround for
ATI/SBx00 chipsets as a synchronization. Unfortunately this triggers
an erratum in newer ICH chipsets (ICH9+) where reading the comparator
immediately after the write returns the old value. Furthermore, as
always, I/O reads are bad for performance.
Therefore, restrict the readback to the chipsets that need it, or, for
debugging purposes, when we are running with hpet=verbose.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Venkatesh Pallipadi <venki@google.com>
LKML-Reference: <20100225185348.GA9674@linux-os.sc.intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
PCI: Ensure we re-enable devices on resume
x86/PCI: parse additional host bridge window resource types
PCI: revert broken device warning
PCI aerdrv: use correct bit defines and add 2ms delay to aer_root_reset
x86/PCI: ignore Consumer/Producer bit in ACPI window descriptions
This is a standalone version of VMware Balloon driver. Ballooning is a
technique that allows hypervisor dynamically limit the amount of memory
available to the guest (with guest cooperation). In the overcommit
scenario, when hypervisor set detects that it needs to shuffle some
memory, it instructs the driver to allocate certain number of pages, and
the underlying memory gets returned to the hypervisor. Later hypervisor
may return memory to the guest by reattaching memory to the pageframes and
instructing the driver to "deflate" balloon.
We are submitting a standalone driver because KVM maintainer (Avi Kivity)
expressed opinion (rightly) that our transport does not fit well into
virtqueue paradigm and thus it does not make much sense to integrate with
virtio.
There were also some concerns whether current ballooning technique is the
right thing. If there appears a better framework to achieve this we are
prepared to evaluate and switch to using it, but in the meantime we'd like
to get this driver upstream.
We want to get the driver accepted in distributions so that users do not
have to deal with an out-of-tree module and many distributions have
"upstream first" requirement.
The driver has been shipping for a number of years and users running on
VMware platform will have it installed as part of VMware Tools even if it
will not come from a distribution, thus there should not be additional
risk in pulling the driver into mainline. The driver will only activate
if host is VMware so everyone else should not be affected at all.
Signed-off-by: Dmitry Torokhov <dtor@vmware.com>
Cc: Avi Kivity <avi@redhat.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This adds support for Memory24, Memory32, and Memory32Fixed descriptors in
PCI host bridge _CRS.
I experimentally determined that Windows (2008 R2) accepts these descriptors
and treats them as windows that are forwarded to the PCI bus, e.g., if
it finds any PCI devices with BARs outside the windows, it moves them into
the windows.
I don't know whether any machines actually use these descriptors in PCI
host bridge _CRS methods, but if any exist and they're new enough that we
automatically turn on "pci=use_crs", they will work with Windows but not
with Linux.
Here are the details: https://bugzilla.kernel.org/show_bug.cgi?id=15817
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
* 'kvm-updates/2.6.34' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
KVM: x86: Fix TSS size check for 16-bit tasks
KVM: Add missing srcu_read_lock() for kvm_mmu_notifier_release()
KVM: Increase NR_IOBUS_DEVS limit to 200
KVM: fix the handling of dirty bitmaps to avoid overflows
KVM: MMU: fix kvm_mmu_zap_page() and its calling path
KVM: VMX: Save/restore rflags.vm correctly in real mode
KVM: allow bit 10 to be cleared in MSR_IA32_MC4_CTL
KVM: Don't spam kernel log when injecting exceptions due to bad cr writes
KVM: SVM: Fix memory leaks that happen when svm_create_vcpu() fails
KVM: take srcu lock before call to complete_pio()
A 16-bit TSS is only 44 bytes long. So make sure to test for the correct
size on task switch.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
Before commit e28cbf2293 ("improve
sys_newuname() for compat architectures") 64-bit x86 had a private
implementation of sys_uname which was just called sys_uname, which other
architectures used for the old uname.
Due to some merge issues with the uname refactoring patches we ended up
calling the old uname version for both the old and new system call
slots, which lead to the domainname filed never be set which caused
failures with libnss_nis.
Reported-and-tested-by: Andy Isaacson <adi@hexapodia.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Int is not long enough to store the size of a dirty bitmap.
This patch fixes this problem with the introduction of a wrapper
function to calculate the sizes of dirty bitmaps.
Note: in mark_page_dirty(), we have to consider the fact that
__set_bit() takes the offset as int, not long.
Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Currently we set eflags.vm unconditionally when entering real mode emulation
through virtual-8086 mode, and clear it unconditionally when we enter protected
mode. The means that the following sequence
KVM_SET_REGS (rflags.vm=1)
KVM_SET_SREGS (cr0.pe=1)
Ends up with rflags.vm clear due to KVM_SET_SREGS triggering enter_pmode().
Fix by shadowing rflags.vm (and rflags.iopl) correctly while in real mode:
reads and writes to those bits access a shadow register instead of the actual
register.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
There is a quirk for AMD K8 CPUs in many Linux kernels (see
arch/x86/kernel/cpu/mcheck/mce.c:__mcheck_cpu_apply_quirks()) that
clears bit 10 in that MCE related MSR. KVM can only cope with all
zeros or all ones, so it will inject a #GP into the guest, which
will let it panic.
So lets add a quirk to the quirk and ignore this single cleared bit.
This fixes -cpu kvm64 on all machines and -cpu host on K8 machines
with some guest Linux kernels.
Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
svm_create_vcpu() does not free the pages allocated during the creation
when it fails to complete the allocations. This patch fixes it.
Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Signed-off-by: Avi Kivity <avi@redhat.com>
complete_pio() may use slot table which is protected by srcu.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Cc: stable@kernel.org
Signed-off-by: Avi Kivity <avi@redhat.com>
This is a partial revert of 4cd8b5e2a1 "lguest: use KVM hypercalls";
we revert to using (just as questionable but more reliable) int $15 for
hypercalls. I didn't revert the register mapping, so we still use the
same calling convention as kvm.
KVM in more recent incarnations stopped injecting a fault when a guest
tried to use the VMCALL instruction from ring 1, so lguest under kvm
fails to make hypercalls. It was nice to share code with our KVM
cousins, but this was overreach.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Matias Zabaljauregui <zabaljauregui@gmail.com>
Cc: Avi Kivity <avi@redhat.com>
When we fetch the hot regs and rewind to the nth caller, it
might happen that we dereference a frame pointer outside the
kernel stack boundaries, like in this example:
perf_trace_sched_switch+0xd5/0x120
schedule+0x6b5/0x860
retint_careful+0xd/0x21
Since we directly dereference a userspace frame pointer here while
rewinding behind retint_careful, this may end up in a crash.
Fix this by simply using probe_kernel_address() when we rewind the
frame pointer.
This issue will have a much more proper fix in the next version of the
perf_arch_fetch_caller_regs() API that will only need to rewind to the
first caller.
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Tested-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: David Miller <davem@davemloft.net>
Cc: Archs <linux-arch@vger.kernel.org>
ACPI Address Space Descriptors (used in _CRS) have a Consumer/Producer
bit that is supposed to distinguish regions that are consumed directly
by a device from those that are forwarded ("produced") by a bridge.
But BIOSes have apparently not used this consistently, and Windows
seems to ignore it, so I think Linux should ignore it as well.
I can't point to any of these supposed broken BIOSes, but since we
now rely on _CRS by default, I think it's safer to ignore this bit
from the start.
Here are details of my experiments with how Windows handles it:
https://bugzilla.kernel.org/show_bug.cgi?id=15701
Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-tip:
x86: Fix double enable_IR_x2apic() call on SMP kernel on !SMP boards
x86: Increase CONFIG_NODES_SHIFT max to 10
ibft, x86: Change reserve_ibft_region() to find_ibft_region()
x86, hpet: Fix bug in RTC emulation
x86, hpet: Erratum workaround for read after write of HPET comparator
bootmem, x86: Fix 32bit numa system without RAM on node 0
nobootmem, x86: Fix 32bit numa system without RAM on node 0
x86: Handle overlapping mptables
x86: Make e820_remove_range to handle all covered case
x86-32, resume: do a global tlb flush in S4 resume
If we boot into a crash-kernel the gart might still be
enabled and its caches might be dirty. This can result in
undefined behavior later. Fix it by explicitly disabling the
gart hardware before initialization and flushing the caches
after enablement.
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
This effectively reverts commit 61d047be99.
Disabling the IOMMU can potetially allow DMA transactions to
complete without being translated. Leave it enabled, and allow
crash kernel to do the IOMMU reinitialization properly.
Cc: stable@kernel.org
Cc: Joerg Roedel <joerg.roedel@amd.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
To catch future potential issues we can add a warning whenever we issue
a command before the command buffer is fully initialized.
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
Hit another kdump problem as reported by Neil Horman. When initializaing
the IOMMU, we attach devices to their domains before the IOMMU is
fully (re)initialized. Attaching a device will issue some important
invalidations. In the context of the newly kexec'd kdump kernel, the
IOMMU may have stale cached data from the original kernel. Because we
do the attach too early, the invalidation commands are placed in the new
command buffer before the IOMMU is updated w/ that buffer. This leaves
the stale entries in the kdump context and can renders device unusable.
Simply enable the IOMMU before we do the attach.
Cc: stable@kernel.org
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
According to Intel Software Devel Manual Volume 3B, the
Nehalem-EX PMU is just like regular Nehalem (except for the
uncore support, which is completely different).
Signed-off-by: Vince Weaver <vweaver1@eecs.utk.edu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Lin Ming <ming.m.lin@intel.com>
LKML-Reference: <alpine.DEB.2.00.1004060956580.1417@cl320.eecs.utk.edu>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Jan Grossmann reported kernel boot panic while booting SMP
kernel on his system with a single core cpu. SMP kernels call
enable_IR_x2apic() from native_smp_prepare_cpus() and on
platforms where the kernel doesn't find SMP configuration we
ended up again calling enable_IR_x2apic() from the
APIC_init_uniprocessor() call in the smp_sanity_check(). Thus
leading to kernel panic.
Don't call enable_IR_x2apic() and default_setup_apic_routing()
from APIC_init_uniprocessor() in CONFIG_SMP case.
NOTE: this kind of non-idempotent and assymetric initialization
sequence is rather fragile and unclean, we'll clean that up
in v2.6.35. This is the minimal fix for v2.6.34.
Reported-by: Jan.Grossmann@kielnet.net
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: <jbarnes@virtuousgeek.org>
Cc: <david.woodhouse@intel.com>
Cc: <weidong.han@intel.com>
Cc: <youquan.song@intel.com>
Cc: <Jan.Grossmann@kielnet.net>
Cc: <stable@kernel.org> # [v2.6.32.x, v2.6.33.x]
LKML-Reference: <1270083887.7835.78.camel@sbs-t61.sc.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
When profiling a 32-bit process on a 64-bit kernel, callgraph tracing
stopped after the first function, because it has seen a garbage memory
address (tried to interpret the frame pointer, and return address as a
64-bit pointer).
Fix this by using a struct stack_frame with 32-bit pointers when the
TIF_IA32 flag is set.
Note that TIF_IA32 flag must be used, and not is_compat_task(), because
the latter is only set when the 32-bit process is executing a syscall,
which may not always be the case (when tracing page fault events for
example).
Signed-off-by: Török Edwin <edwintorok@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: x86@kernel.org
Cc: linux-kernel@vger.kernel.org
LKML-Reference: <1268820436-13145-1-git-send-email-edwintorok@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Commit 3f6da39 ("perf: Rework and fix the arch CPU-hotplug hooks") moved
the amd northbridge allocation from CPUS_ONLINE to CPUS_PREPARE_UP
however amd_nb_id() doesn't work yet on prepare so it would simply bail
basically reverting to a state where we do not properly track node wide
constraints - causing weird perf results.
Fix up the AMD NorthBridge initialization code by allocating from
CPU_UP_PREPARE and installing it from CPU_STARTING once we have the
proper nb_id. It also properly deals with the allocation failing.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
[ robustify using amd_has_nb() ]
Signed-off-by: Stephane Eranian <eranian@google.com>
LKML-Reference: <1269353485.5109.48.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Because we need to have cpu identification things done by the time we run
CPU_STARTING notifiers.
( This init ordering will be relied on by the next fix. )
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1269353485.5109.48.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Some larger systems require more than 512 nodes, so increase the
maximum CONFIG_NODES_SHIFT to 10 for a new max of 1024 nodes.
This was tested with numa=fake=64M on systems with more than
64GB of RAM. A total of 1022 nodes were initialized.
Successfully builds with no additional warnings on x86_64
allyesconfig.
( No effect on any existing config. Newly enabled CONFIG_MAXSMP=y
will see the new default. )
Signed-off-by: David Rientjes <rientjes@google.com>
LKML-Reference: <alpine.DEB.2.00.1003251538060.8589@chino.kir.corp.google.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
This allows arch code could decide the way to reserve the ibft.
And we should reserve ibft as early as possible, instead of BOOTMEM
stage, in case the table is in RAM range and is not reserved by BIOS
(this will often be the case.)
Move to just after find_smp_config().
Also when CONFIG_NO_BOOTMEM=y, We will not have reserve_bootmem() anymore.
-v2: fix typo about ibft pointed by Konrad Rzeszutek Wilk <konrad@darnok.org>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <4BB510FB.80601@kernel.org>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Peter Jones <pjones@redhat.com>
Cc: Konrad Rzeszutek Wilk <konrad@kernel.org>
CC: Jan Beulich <jbeulich@novell.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
We think there exists a bug in the HPET code that emulates the RTC.
In the normal case, when the RTC frequency is set, the rtc driver tells
the hpet code about it here:
int hpet_set_periodic_freq(unsigned long freq)
{
uint64_t clc;
if (!is_hpet_enabled())
return 0;
if (freq <= DEFAULT_RTC_INT_FREQ)
hpet_pie_limit = DEFAULT_RTC_INT_FREQ / freq;
else {
clc = (uint64_t) hpet_clockevent.mult * NSEC_PER_SEC;
do_div(clc, freq);
clc >>= hpet_clockevent.shift;
hpet_pie_delta = (unsigned long) clc;
}
return 1;
}
If freq is set to 64Hz (DEFAULT_RTC_INT_FREQ) or lower, then
hpet_pie_limit (a static) is set to non-zero. Then, on every one-shot
HPET interrupt, hpet_rtc_timer_reinit is called to compute the next
timeout. Well, that function has this logic:
if (!(hpet_rtc_flags & RTC_PIE) || hpet_pie_limit)
delta = hpet_default_delta;
else
delta = hpet_pie_delta;
Since hpet_pie_limit is not 0, hpet_default_delta is used. That
corresponds to 64Hz.
Now, if you set a different rtc frequency, you'll take the else path
through hpet_set_periodic_freq, but unfortunately no one resets
hpet_pie_limit back to 0.
Boom....now you are stuck with 64Hz RTC interrupts forever.
The patch below just resets the hpet_pie_limit value when requested freq
is greater than DEFAULT_RTC_INT_FREQ, which we think fixes this problem.
Signed-off-by: Alok N Kataria <akataria@vmware.com>
LKML-Reference: <201003112200.o2BM0Hre012875@imap1.linux-foundation.org>
Signed-off-by: Daniel Hecht <dhecht@vmware.com>
Cc: Venkatesh Pallipadi <venkatesh.pallipadi@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
On Wed, Feb 24, 2010 at 03:37:04PM -0800, Justin Piszcz wrote:
> Hello,
>
> Again, on the Intel DP55KG board:
>
> # uname -a
> Linux host 2.6.33 #1 SMP Wed Feb 24 18:31:00 EST 2010 x86_64 GNU/Linux
>
> [ 1.237600] ------------[ cut here ]------------
> [ 1.237890] WARNING: at arch/x86/kernel/hpet.c:404 hpet_next_event+0x70/0x80()
> [ 1.238221] Hardware name:
> [ 1.238504] hpet: compare register read back failed.
> [ 1.238793] Modules linked in:
> [ 1.239315] Pid: 0, comm: swapper Not tainted 2.6.33 #1
> [ 1.239605] Call Trace:
> [ 1.239886] <IRQ> [<ffffffff81056c13>] ? warn_slowpath_common+0x73/0xb0
> [ 1.240409] [<ffffffff81079608>] ? tick_dev_program_event+0x38/0xc0
> [ 1.240699] [<ffffffff81056cb0>] ? warn_slowpath_fmt+0x40/0x50
> [ 1.240992] [<ffffffff81079608>] ? tick_dev_program_event+0x38/0xc0
> [ 1.241281] [<ffffffff81041ad0>] ? hpet_next_event+0x70/0x80
> [ 1.241573] [<ffffffff81079608>] ? tick_dev_program_event+0x38/0xc0
> [ 1.241859] [<ffffffff81078e32>] ? tick_handle_oneshot_broadcast+0xe2/0x100
> [ 1.246533] [<ffffffff8102a67a>] ? timer_interrupt+0x1a/0x30
> [ 1.246826] [<ffffffff81085499>] ? handle_IRQ_event+0x39/0xd0
> [ 1.247118] [<ffffffff81087368>] ? handle_edge_irq+0xb8/0x160
> [ 1.247407] [<ffffffff81029f55>] ? handle_irq+0x15/0x20
> [ 1.247689] [<ffffffff810294a2>] ? do_IRQ+0x62/0xe0
> [ 1.247976] [<ffffffff8146be53>] ? ret_from_intr+0x0/0xa
> [ 1.248262] <EOI> [<ffffffff8102f277>] ? mwait_idle+0x57/0x80
> [ 1.248796] [<ffffffff8102645c>] ? cpu_idle+0x5c/0xb0
> [ 1.249080] ---[ end trace db7f668fb6fef4e1 ]---
>
> Is this something Intel has to fix or is it a bug in the kernel?
This is a chipset erratum.
Thomas: You mentioned we can retain this check only for known-buggy and
hpet debug kind of options. But here is the simple workaround patch for
this particular erratum.
Some chipsets have a erratum due to which read immediately following a
write of HPET comparator returns old comparator value instead of most
recently written value.
Erratum 15 in
"Intel I/O Controller Hub 9 (ICH9) Family Specification Update"
(http://www.intel.com/assets/pdf/specupdate/316973.pdf)
Workaround for the errata is to read the comparator twice if the first
one fails.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
LKML-Reference: <20100225185348.GA9674@linux-os.sc.intel.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Venkatesh Pallipadi <venkatesh.pallipadi@gmail.com>
Cc: <stable@kernel.org>
We found a system where the MP table MPC and MPF structures overlap.
That doesn't really matter because the mptable is not used anyways with ACPI,
but it leads to a panic in the early allocator due to the overlapping
reservations in 2.6.33.
Earlier kernels handled this without problems.
Simply change these reservations to reserve_early_overlap_ok to avoid
the panic.
Reported-by: Thomas Renninger <trenn@suse.de>
Tested-by: Thomas Renninger <trenn@suse.de>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
LKML-Reference: <20100329074111.GA22821@basil.fritz.box>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: <stable@kernel.org>
It is required to call hw_breakpoint_init() on an attr before using it
in any other calls. This fixes the problem where kgdb will sometimes
fail to initialize on x86_64.
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: 2.6.33 <stable@kernel.org>
LKML-Reference: <1269975907-27602-1-git-send-email-jason.wessel@windriver.com>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Scheduler's task migration events don't work because they always
pass NULL regs perf_sw_event(). The event hence gets filtered
in perf_swevent_add().
Scheduler's context switches events use task_pt_regs() to get
the context when the event occured which is a wrong thing to
do as this won't give us the place in the kernel where we went
to sleep but the place where we left userspace. The result is
even more wrong if we switch from a kernel thread.
Use the hot regs snapshot for both events as they belong to the
non-interrupt/exception based events family. Unlike page faults
or so that provide the regs matching the exact origin of the event,
we need to save the current context.
This makes the task migration event working and fix the context
switch callchains and origin ip.
Example: perf record -a -e cs
Before:
10.91% ksoftirqd/0 0 [k] 0000000000000000
|
--- (nil)
perf_callchain
perf_prepare_sample
__perf_event_overflow
perf_swevent_overflow
perf_swevent_add
perf_swevent_ctx_event
do_perf_sw_event
__perf_sw_event
perf_event_task_sched_out
schedule
run_ksoftirqd
kthread
kernel_thread_helper
After:
23.77% hald-addon-stor [kernel.kallsyms] [k] schedule
|
--- schedule
|
|--60.00%-- schedule_timeout
| wait_for_common
| wait_for_completion
| blk_execute_rq
| scsi_execute
| scsi_execute_req
| sr_test_unit_ready
| |
| |--66.67%-- sr_media_change
| | media_changed
| | cdrom_media_changed
| | sr_block_media_changed
| | check_disk_change
| | cdrom_open
v2: Always build perf_arch_fetch_caller_regs() now that software
events need that too. They don't need it from modules, unlike trace
events, so we keep the EXPORT_SYMBOL in trace_event_perf.c
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Rusty found on lguest with trim_bios_range, max_pfn is not right anymore, and
looks e820_remove_range does not work right.
[ 0.000000] BIOS-provided physical RAM map:
[ 0.000000] LGUEST: 0000000000000000 - 0000000004000000 (usable)
[ 0.000000] Notice: NX (Execute Disable) protection missing in CPU or disabled in BIOS!
[ 0.000000] DMI not present or invalid.
[ 0.000000] last_pfn = 0x3fa0 max_arch_pfn = 0x100000
[ 0.000000] init_memory_mapping: 0000000000000000-0000000003fa0000
root cause is: the e820_remove_range doesn't handle the all covered
case. e820_remove_range(BIOS_START, BIOS_END - BIOS_START, ...)
produces a bogus range as a result.
Make it match e820_update_range() by handling that case too.
Reported-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Tested-by: Rusty Russell <rusty@rustcorp.com.au>
LKML-Reference: <4BB18E55.6090903@kernel.org>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Colin King reported a strange oops in S4 resume code path (see below). The test
system has i5/i7 CPU. The kernel doesn't open PAE, so 4M page table is used.
The oops always happen a virtual address 0xc03ff000, which is mapped to the
last 4k of first 4M memory. Doing a global tlb flush fixes the issue.
EIP: 0060:[<c0493a01>] EFLAGS: 00010086 CPU: 0
EIP is at copy_loop+0xe/0x15
EAX: 36aeb000 EBX: 00000000 ECX: 00000400 EDX: f55ad46c
ESI: 0f800000 EDI: c03ff000 EBP: f67fbec4 ESP: f67fbea8
DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
...
...
CR2: 00000000c03ff000
Tested-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
LKML-Reference: <20100305005932.GA22675@sli10-desk.sh.intel.com>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: <stable@kernel.org>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Including slab.h from x86 pgtable_32.h creates a troublesome
dependency chain w/ ftrace enabled. The following chain leads to
inclusion of pgtable_32.h from define_trace.h.
trace/define_trace.h
trace/ftrace.h
linux/ftrace_event.h
linux/ring_buffer.h
linux/mm.h
asm/pgtable.h
asm/pgtable_32.h
slab.h itself defines trace hooks via
linux/sl[aou]b_def.h
linux/kmemtrace.h
trace/events/kmem.h
If slab.h is not included before define_trace.h is included, this
leads to duplicate definitions of kmemtrace hooks or other include
dependency problems.
pgtable_32.h doesn't need slab.h to begin with. Don't include it from
there.
Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: H. Peter Anvin <hpa@zytor.com>
When CONFIG_NO_BOOTMEM=y, it could use memory more effiently, or
in a more compact fashion.
Example:
Allocated new RAMDISK: 00ec2000 - 0248ce57
Move RAMDISK from 000000002ea04000 - 000000002ffcee56 to 00ec2000 - 0248ce56
The new RAMDISK's end is not page aligned.
Last page could be shared with other users.
When free_init_pages are called for initrd or .init, the page
could be freed and we could corrupt other data.
code segment in free_init_pages():
| for (; addr < end; addr += PAGE_SIZE) {
| ClearPageReserved(virt_to_page(addr));
| init_page_count(virt_to_page(addr));
| memset((void *)(addr & ~(PAGE_SIZE-1)),
| POISON_FREE_INITMEM, PAGE_SIZE);
| free_page(addr);
| totalram_pages++;
| }
last half page could be used as one whole free page.
So page align the boundaries.
-v2: make the original initramdisk to be aligned, according to
Johannes, otherwise we have the chance to lose one page.
we still need to keep initrd_end not aligned, otherwise it could
confuse decompressor.
-v3: change to WARN_ON instead, suggested by Johannes.
-v4: use PAGE_ALIGN, suggested by Johannes.
We may fix that macro name later to PAGE_ALIGN_UP, and PAGE_ALIGN_DOWN
Add comments about assuming ramdisk start is aligned
in relocate_initrd(), change to re get ramdisk_image instead of save it
to make diff smaller. Add warning for wrong range, suggested by Johannes.
-v6: remove one WARN()
We need to align beginning in free_init_pages()
do not copy more than ramdisk_size, noticed by Johannes
Reported-by: Stanislaw Gruszka <sgruszka@redhat.com>
Tested-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: David Miller <davem@davemloft.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
LKML-Reference: <1269830604-26214-3-git-send-email-yinghai@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
x86/PCI: truncate _CRS windows with _LEN > _MAX - _MIN + 1
x86/PCI: for host bridge address space collisions, show conflicting resource
frv/PCI: remove redundant warnings
x86/PCI: remove redundant warnings
PCI: don't say we claimed a resource if we failed
PCI quirk: Disable MSI on VIA K8T890 systems
PCI quirk: RS780/RS880: work around missing MSI initialization
PCI quirk: only apply CX700 PCI bus parking quirk if external VT6212L is present
PCI: complain about devices that seem to be broken
PCI: print resources consistently with %pR
PCI: make disabled window printk style match the enabled ones
PCI: break out primary/secondary/subordinate for readability
PCI: for address space collisions, show conflicting resource
resources: add interfaces that return conflict information
PCI: cleanup error return for pcix get and set mmrbc functions
PCI: fix access of PCI_X_CMD by pcix get and set mmrbc functions
PCI: kill off pci_register_set_vga_state() symbol export.
PCI: fix return value from pcix_get_max_mmrbc()
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86, amd: Restrict usage of c1e_idle()
x86: Fix placement of FIX_OHCI1394_BASE
x86: Handle legacy PIC interrupts on all the cpu's
