Implement the ability to individually get and set registers for kdb
and kgdb for x86.
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
CC: Ingo Molnar <mingo@redhat.com>
CC: x86@kernel.org
HW breakpoints events stopped working correctly with kgdb
as a result of commit: 018cbffe68
(Merge commit 'v2.6.33' into perf/core).
The regression occurred because the behavior changed for setting
NOTIFY_STOP as the return value to the die notifier if the breakpoint
was known to the HW breakpoint API. Because kgdb is using the HW
breakpoint API to register HW breakpoints slots, it must also now
implement the overflow_handler call back else kgdb does not get to see
the events from the die notifier.
The kgdb_ll_trap function will be changed to be general purpose code
which can allow an easy way to implement the hw_breakpoint API
overflow call back.
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Dongdong Deng <dongdong.deng@windriver.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86: Do not try to disable hpet if it hasn't been initialized before
x86, i8259: Only register sysdev if we have a real 8259 PIC
The Pstate transition latency check was added for broken F10h BIOSen
which wrongly contain a value of 0 for transition and bus master
latency. Fam11h and later, however, (will) have similar transition
latency so extend that behavior for them too.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Signed-off-by: Dave Jones <davej@redhat.com>
The PCC cpufreq driver unmaps the mailbox address range if any CPUs fail to
initialise, but doesn't do anything to remove the registered CPUs from the
cpufreq core resulting in failures further down the line. We're better off
simply returning a failure - the cpufreq core will unregister us cleanly if
we end up with no successfully registered CPUs. Tidy up the failure path
and also add a sanity check to ensure that the firmware gives us a realistic
frequency - the core deals badly with that being set to 0.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
Signed-off-by: Dave Jones <davej@redhat.com>
The pcc specification documents an _OSC method that's incompatible with the
one defined as part of the ACPI spec. This shouldn't be a problem as both
are supposed to be guarded with a UUID. Unfortunately approximately nobody
(including HP, who wrote this spec) properly check the UUID on entry to the
_OSC call. Right now this could result in surprising behaviour if the pcc
driver performs an _OSC call on a machine that doesn't implement the pcc
specification. Check whether the PCCH method exists first in order to reduce
this probability.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: Naga Chumbalkar <nagananda.chumbalkar@hp.com>
Signed-off-by: Dave Jones <davej@redhat.com>
Commit 2a6b69765a
(ACPI: Store NVS state even when entering suspend to RAM) caused the
ACPI suspend code save the NVS area during suspend and restore it
during resume unconditionally, although it is known that some systems
need to use acpi_sleep=s4_nonvs for hibernation to work. To allow
the affected systems to avoid saving and restoring the NVS area
during suspend to RAM and resume, introduce kernel command line
option acpi_sleep=nonvs and make acpi_sleep=s4_nonvs work as its
alias temporarily (add acpi_sleep=s4_nonvs to the feature removal
file).
Addresses https://bugzilla.kernel.org/show_bug.cgi?id=16396 .
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reported-and-tested-by: tomas m <tmezzadra@gmail.com>
Signed-off-by: Len Brown <len.brown@intel.com>
hpet_disable is called unconditionally on machine reboot if hpet support
is compiled in the kernel.
hpet_disable only checks if the machine is hpet capable but doesn't make
sure that hpet has been initialized.
[ tglx: Made it a one liner and removed the redundant hpet_address check ]
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Venkatesh Pallipadi <venki@google.com>
LKML-Reference: <alpine.DEB.2.00.1007211726240.22235@kaball-desktop>
Cc: stable@kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
It turns out that there is a bit in the _CST for Intel FFH C3
that tells the OS if we should be checking BM_STS or not.
Linux has been unconditionally checking BM_STS.
If the chip-set is configured to enable BM_STS,
it can retard or completely prevent entry into
deep C-states -- as illustrated by turbostat:
http://userweb.kernel.org/~lenb/acpi/utils/pmtools/turbostat/
ref: Intel Processor Vendor-Specific ACPI Interface Specification
table 4 "_CST FFH GAS Field Encoding"
Bit 1: Set to 1 if OSPM should use Bus Master avoidance for this C-state
https://bugzilla.kernel.org/show_bug.cgi?id=15886
Signed-off-by: Len Brown <len.brown@intel.com>
In the CONFIG_AUDITSYSCALL fast-path for x86 64-bit system calls,
we can pass a bad return value and/or error indication for the
system call to audit_syscall_exit(). This happens when
TIF_NEED_RESCHED was set as the system call returned, so we went
out to schedule() and came back to the exit-audit fast-path. The
fix is to reload the user return value register from the pt_regs
before using it for audit_syscall_exit().
Both the 32-bit kernel's fast path and the 64-bit kernel's 32-bit
system call fast paths work slightly differently, so that they
always leave the fast path entirely to reschedule and don't return
there, so they don't have the analogous bugs.
Reported-by: Alexander Viro <aviro@redhat.com>
Signed-off-by: Roland McGrath <roland@redhat.com>
Pointed out by Lucas who found the new one in a comment in
setup_percpu.c. And then I fixed the others that I grepped
for.
Reported-by: Lucas <canolucas@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit e534c7c5f8 ("numa: x86_64: use generic percpu var
numa_node_id() implementation") broke numa systems that don't have ram
on node0 when MEMORY_HOTPLUG is enabled, because cpu_up() will call
cpu_to_node() before per_cpu(numa_node) is setup for APs.
When Node0 doesn't have RAM, on x86, cpus already round it to nearest
node with RAM in x86_cpu_to_node_map. and per_cpu(numa_node) is not set
up until in c_init for APs.
When later cpu_up() calling cpu_to_node() will get 0 again, and make it
online even there is no RAM on node0. so later all APs can not booted up,
and later will have panic.
[ 1.611101] On node 0 totalpages: 0
.........
[ 2.608558] On node 0 totalpages: 0
[ 2.612065] Brought up 1 CPUs
[ 2.615199] Total of 1 processors activated (3990.31 BogoMIPS).
...
93.225341] calling loop_init+0x0/0x1a4 @ 1
[ 93.229314] PERCPU: allocation failed, size=80 align=8, failed to populate
[ 93.246539] Pid: 1, comm: swapper Tainted: G W 2.6.35-rc4-tip-yh-04371-gd64e6c4-dirty #354
[ 93.264621] Call Trace:
[ 93.266533] [<ffffffff81125e43>] pcpu_alloc+0x83a/0x8e7
[ 93.270710] [<ffffffff81125f15>] __alloc_percpu+0x10/0x12
[ 93.285849] [<ffffffff8140786c>] alloc_disk_node+0x94/0x16d
[ 93.291811] [<ffffffff81407956>] alloc_disk+0x11/0x13
[ 93.306157] [<ffffffff81503e51>] loop_alloc+0xa7/0x180
[ 93.310538] [<ffffffff8277ef48>] loop_init+0x9b/0x1a4
[ 93.324909] [<ffffffff8277eead>] ? loop_init+0x0/0x1a4
[ 93.329650] [<ffffffff810001f2>] do_one_initcall+0x57/0x136
[ 93.345197] [<ffffffff827486d0>] kernel_init+0x184/0x20e
[ 93.348146] [<ffffffff81034954>] kernel_thread_helper+0x4/0x10
[ 93.365194] [<ffffffff81c7cc3c>] ? restore_args+0x0/0x30
[ 93.369305] [<ffffffff8274854c>] ? kernel_init+0x0/0x20e
[ 93.386011] [<ffffffff81034950>] ? kernel_thread_helper+0x0/0x10
[ 93.392047] loop: out of memory
...
Try to assign per_cpu(numa_node) early
[akpm@linux-foundation.org: tidy up code comment]
Signed-off-by: Yinghai <yinghai@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
My platform makes use of the null_legacy_pic choice and oopses when doing
a shutdown as the shutdown code goes through all the registered sysdevs
and calls their shutdown method which in my case poke on a non-existing
i8259. Imho the i8259 specific sysdev should only be registered if the
i8259 is actually there.
Do not register the sysdev function when the null_legacy_pic is used so
that the i8259 resume, suspend and shutdown functions are not called.
Signed-off-by: Adam Lackorzynski <adam@os.inf.tu-dresden.de>
LKML-Reference: <201007202218.o6KMIJ3m020955@imap1.linux-foundation.org>
Cc: Jacob Pan <jacob.jun.pan@intel.com>
Cc: <stable@kernel.org> 2.6.34
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86, pci, mrst: Add extra sanity check in walking the PCI extended cap chain
x86: Fix x2apic preenabled system with kexec
x86: Force HPET readback_cmp for all ATI chipsets
In commit f007ea26, the order of the %es and %ds segment registers
got accidentally swapped, so synthesized 'struct pt_regs' frames
have the two values inverted. It's almost sure that these values
never matter, and that they also never differ. But wrong is wrong.
Signed-off-by: Roland McGrath <roland@redhat.com>
Found one x2apic system kexec loop test failed
when CONFIG_NMI_WATCHDOG=y (old) or CONFIG_LOCKUP_DETECTOR=y (current tip)
first kernel can kexec second kernel, but second kernel can not kexec third one.
it can be duplicated on another system with BIOS preenabled x2apic.
First kernel can not kexec second kernel.
It turns out, when kernel boot with pre-enabled x2apic, it will not execute
disable_local_APIC on shutdown path.
when init_apic_mappings() is called in setup_arch, it will skip setting of
apic_phys when x2apic_mode is set. ( x2apic_mode is much early check_x2apic())
Then later, disable_local_APIC() will bail out early because !apic_phys.
So check !x2apic_mode in x2apic_mode in disable_local_APIC with !apic_phys.
another solution could be updating init_apic_mappings() to set apic_phys even
for preenabled x2apic system. Actually even for x2apic system, that lapic
address is mapped already in early stage.
BTW: is there any x2apic preenabled system with apicid of boot cpu > 255?
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <4C3EB22B.3000701@kernel.org>
Acked-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: stable@kernel.org
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
commit 30a564be (x86, hpet: Restrict read back to affected ATI
chipset) restricted the workaround for the HPET bug to SMX00
chipsets. This was reasonable as those were the only ones against
which we ever got a bug report.
Stephan Wolf reported now that this patch breaks his IXP400 based
machine. Though it's confirmed to work on other IXP400 based systems.
To error out on the safe side, we force the HPET readback workaround
for all ATI SMbus class chipsets.
Reported-by: Stephan Wolf <stephan@letzte-bankreihe.de>
LKML-Reference: <alpine.LFD.2.00.1007142134140.3321@localhost.localdomain>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Stephan Wolf <stephan@letzte-bankreihe.de>
Acked-by: Borislav Petkov <borislav.petkov@amd.com>
It will just return 0 as there is no i8042 controller
Signed-off-by: Feng Tang <feng.tang@intel.com>
LKML-Reference: <1278342202-10973-3-git-send-email-feng.tang@intel.com>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Some x86 platforms like Intel MID platforms don't have i8042 controllers,
and i8042 driver's probe to some legacy IO ports may hang the MID
processor. With this hook, i8042 driver can runtime check and skip the
probe when the pretection fail which also saves some probe time
[ hpa note: this is currently a compile-time check, which breaks the
i386 allyesconfig build. This patch series thus does fix a regression. ]
Signed-off-by: Feng Tang <feng.tang@intel.com>
LKML-Reference: <1278342202-10973-2-git-send-email-feng.tang@intel.com>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Export x86_platform to modules in preparation of using it for i8042
discovery control.
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
LKML-Reference: <1278342202-10973-1-git-send-email-feng.tang@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Feng Tang <feng.tang@intel.com>
Cc: Dmitry Torokhov <dtor@mail.ru>
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
rbtree: Undo augmented trees performance damage and regression
x86, Calgary: Limit the max PHB number to 256
* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
perf, x86: Fix incorrect branches event on AMD CPUs
perf tools: Fix find tids routine by excluding "." and ".."
x86: Send a SIGTRAP for user icebp traps
While doing some performance counter validation tests on some
assembly language programs I noticed that the "branches:u"
count was very wrong on AMD machines.
It looks like the wrong event was selected.
Signed-off-by: Vince Weaver <vweaver1@eecs.utk.edu>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: <stable@kernel.org>
LKML-Reference: <alpine.DEB.2.00.1007011526010.23160@cl320.eecs.utk.edu>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
The x3950 family can have as many as 256 PCI buses in a single system, so
change the limits to the maximum. Since there can only be 256 PCI buses in one
domain, we no longer need the BUG_ON check.
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
LKML-Reference: <20100701004519.GQ15515@tux1.beaverton.ibm.com>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Before we had a generic breakpoint layer, x86 used to send a
sigtrap for any debug event that happened in userspace,
except if it was caused by lazy dr7 switches.
Currently we only send such signal for single step or breakpoint
events.
However, there are three other kind of debug exceptions:
- debug register access detected: trigger an exception if the
next instruction touches the debug registers. We don't use
it.
- task switch, but we don't use tss.
- icebp/int01 trap. This instruction (0xf1) is undocumented and
generates an int 1 exception. Unlike single step through TF
flag, it doesn't set the single step origin of the exception
in dr6.
icebp then used to be reported in userspace using trap signals
but this have been incidentally broken with the new breakpoint
code. Reenable this. Since this is the only debug event that
doesn't set anything in dr6, this is all we have to check.
This fixes a regression in Wine where World Of Warcraft got broken
as it uses this for software protection checks purposes. And
probably other apps do.
Reported-and-tested-by: Alexandre Julliard <julliard@winehq.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Prasad <prasad@linux.vnet.ibm.com>
Cc: 2.6.33.x 2.6.34.x <stable@kernel.org>
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86, Calgary: Increase max PHB number
x86: Fix rebooting on Dell Precision WorkStation T7400
x86: Fix vsyscall on gcc 4.5 with -Os
x86, pat: Proper init of memtype subtree_max_end
um, hweight: Fix UML boot crash due to x86 optimized hweight
x86, setup: Set ax register in boot vga query
percpu, x86: Avoid warnings of unused variables in per cpu
x86, irq: Rename gsi_end gsi_top, and fix off by one errors
x86: use __ASSEMBLY__ rather than __ASSEMBLER__
Newer systems (x3950M2) can have 48 PHBs per chassis and 8
chassis, so bump the limits up and provide an explanation
of the requirements for each class.
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Acked-by: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: Corinna Schultz <cschultz@linux.vnet.ibm.com>
Cc: <stable@kernel.org>
LKML-Reference: <20100624212647.GI15515@tux1.beaverton.ibm.com>
[ v2: Fixed build bug, added back PHBS_PER_CALGARY == 4 ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Saving platform non-volatile state may be required for suspend to RAM as
well as hibernation. Move it to more generic code.
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Acked-by: Rafael J. Wysocki <rjw@sisk.pl>
Tested-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Len Brown <len.brown@intel.com>
When I introduced the global variable gsi_end I thought gsi_end on
io_apics was one past the end of the gsi range for the io_apic. After
it was pointed out the the range on io_apics was inclusive I changed
my global variable to match. That was a big mistake. Inclusive
semantics without a range start cannot describe the case when no gsi's
are allocated. Describing the case where no gsi's are allocated is
important in sfi.c and mpparse.c so that we can assign gsi numbers
instead of blindly copying the gsi assignments the BIOS has done as we
do in the acpi case.
To keep from getting the global variable confused with the gsi range
end rename it gsi_top.
To allow describing the case where no gsi's are allocated have gsi_top
be one place the highest gsi number seen in the system.
This fixes an off by one bug in sfi.c:
Reported-by: jacob pan <jacob.jun.pan@linux.intel.com>
This fixes the same off by one bug in mpparse.c:
This fixes an off unreachable by one bug in acpi/boot.c:irq_to_gsi
Reported-by: Yinghai <yinghai.lu@oracle.com>
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
LKML-Reference: <m17hm9jre7.fsf_-_@fess.ebiederm.org>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
* 'linux_next' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/i7core: (83 commits)
i7core_edac: Better describe the supported devices
Add support for Westmere to i7core_edac driver
i7core_edac: don't free on success
i7core_edac: Add support for X5670
Always call i7core_[ur]dimm_check_mc_ecc_err
i7core_edac: fix memory leak of i7core_dev
EDAC: add __init to i7core_xeon_pci_fixup
i7core_edac: Fix wrong device id for channel 1 devices
i7core: add support for Lynnfield alternate address
i7core_edac: Add initial support for Lynnfield
i7core_edac: do not export static functions
edac: fix i7core build
edac: i7core_edac produces undefined behaviour on 32bit
i7core_edac: Use a more generic approach for probing PCI devices
i7core_edac: PCI device is called NONCORE, instead of NOCORE
i7core_edac: Fix ringbuffer maxsize
i7core_edac: First store, then increment
i7core_edac: Better parse "any" addrmask
i7core_edac: Use a lockless ringbuffer
edac: Create an unique instance for each kobj
...
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86, smpboot: Fix cores per node printing on boot
x86/amd-iommu: Fall back to GART if initialization fails
x86/amd-iommu: Fix crash when request_mem_region fails
x86/mm: Remove unused DBG() macro
arch/x86/kernel: Add missing spin_unlock
* 'perf-fixes-for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
perf: Fix crash in swevents
perf buildid-list: Fix --with-hits event processing
perf scripts python: Give field dict to unhandled callback
perf hist: fix objdump output parsing
perf-record: Check correct pid when forking
perf: Do the comm inheritance per thread in event__process_task
perf: Use event__process_task from perf sched
perf: Process comm events by tid
blktrace: Fix new kernel-doc warnings
perf_events: Fix unincremented buffer base on partial copy
perf_events: Fix event scheduling issues introduced by transactional API
perf_events, trace: Fix perf_trace_destroy(), mutex went missing
perf_events, trace: Fix probe unregister race
perf_events: Fix races in group composition
perf_events: Fix races and clean up perf_event and perf_mmap_data interaction
Percpu initialization happens now after booting the cores on the
machine and this causes them all to be displayed as belonging to
node 0:
Jun 8 05:57:21 kepek kernel: [ 0.106999] Booting Node 0,
Processors #1#2#3#4#5#6#7#8#9#10#11#12#13#14#15#16#17#18#19#20#21#22#23 Ok.
Use early_cpu_to_node() to get the correct node of each core
instead.
Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>
Cc: Mike Travis <travis@sgi.com>
LKML-Reference: <20100601190455.GA14237@aftab>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'for-35' of git://repo.or.cz/linux-kbuild: (81 commits)
kbuild: Revert part of e8d400a to resolve a conflict
kbuild: Fix checking of scm-identifier variable
gconfig: add support to show hidden options that have prompts
menuconfig: add support to show hidden options which have prompts
gconfig: remove show_debug option
gconfig: remove dbg_print_ptype() and dbg_print_stype()
kconfig: fix zconfdump()
kconfig: some small fixes
add random binaries to .gitignore
kbuild: Include gen_initramfs_list.sh and the file list in the .d file
kconfig: recalc symbol value before showing search results
.gitignore: ignore *.lzo files
headerdep: perlcritic warning
scripts/Makefile.lib: Align the output of LZO
kbuild: Generate modules.builtin in make modules_install
Revert "kbuild: specify absolute paths for cscope"
kbuild: Do not unnecessarily regenerate modules.builtin
headers_install: use local file handles
headers_check: fix perl warnings
export_report: fix perl warnings
...
This patch implements a fallback to the GART IOMMU if this
is possible and the AMD IOMMU initialization failed.
Otherwise the fallback would be nommu which is very
problematic on machines with more than 4GB of memory or
swiotlb which hurts io-performance.
Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
When request_mem_region fails the error path tries to
disable the IOMMUs. This accesses the mmio-region which was
not allocated leading to a kernel crash. This patch fixes
the issue.
Cc: stable@kernel.org
Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
The transactional API patch between the generic and model-specific
code introduced several important bugs with event scheduling, at
least on X86. If you had pinned events, e.g., watchdog, and were
over-committing the PMU, you would get bogus counts. The bug was
showing up on Intel CPU because events would move around more
often that on AMD. But the problem also existed on AMD, though
harder to expose.
The issues were:
- group_sched_in() was missing a cancel_txn() in the error path
- cpuc->n_added was not properly maintained, leading to missing
actions in hw_perf_enable(), i.e., n_running being 0. You cannot
update n_added until you know the transaction has succeeded. In
case of failed transaction n_added was not adjusted back.
- in case of failed transactions, event_sched_out() was called
and eventually invoked x86_disable_event() to touch the HW reg.
But with transactions, on X86, event_sched_in() does not touch
HW registers, it simply collects events into a list. Thus, you
could end up calling x86_disable_event() on a counter which
did not correspond to the current event when idx != -1.
The patch modifies the generic and X86 code to avoid all those problems.
First, we keep track of the number of events added last. In case the
transaction fails, we substract them from n_added. This approach is
necessary (as opposed to delaying updates to n_added) because not all
event updates use the transaction API, e.g., single events.
Second, we encapsulate the event_sched_in() and event_sched_out() in
group_sched_in() inside the transaction. That makes the operations
symmetrical and you can also detect that you are inside a transaction
and skip the HW reg access by checking cpuc->group_flag.
With this patch, you can now overcommit the PMU even with pinned
system-wide events present and still get valid counts.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1274796225.5882.1389.camel@twins>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86, cpufeature: Unbreak compile with gcc 3.x
x86, pat: Fix memory leak in free_memtype
x86, k8: Fix section mismatch for powernowk8_exit()
lib/atomic64_test: fix missing include of linux/kernel.h
x86: remove last traces of quicklist usage
x86, setup: Phoenix BIOS fixup is needed on Dell Inspiron Mini 1012
x86: "nosmp" command line option should force the system into UP mode
arch/x86/pci: use kasprintf
x86, apic: ack all pending irqs when crashed/on kexec
