When kernel doesn't support X2APIC but BIOS has enabled X2APIC, system
may panic or hang without useful messages. On the other hand, it's
hard to dynamically disable X2APIC when CONFIG_X86_X2APIC is disabled.
So panic with a clear message in such a case.
Now system panics as below when X2APIC is disabled and interrupt remapping
is enabled:
[ 0.316118] LAPIC pending interrupts after 512 EOI
[ 0.322126] ..TIMER: vector=0x30 apic1=0 pin1=2 apic2=-1 pin2=-1
[ 0.368655] Kernel panic - not syncing: timer doesn't work through Interrupt-remapped IO-APIC
[ 0.378300] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.18.0+ #340
[ 0.385300] Hardware name: Intel Corporation BRICKLAND/BRICKLAND, BIOS BRIVTIN1.86B.0051.L05.1406240953 06/24/2014
[ 0.396997] ffff88046dc03000 ffff88046c307dd8 ffffffff8179dada 00000000000043f2
[ 0.405629] ffffffff81a92158 ffff88046c307e58 ffffffff8179b757 0000000000000002
[ 0.414261] 0000000000000008 ffff88046c307e68 ffff88046c307e08 ffffffff813ad82b
[ 0.422890] Call Trace:
[ 0.425711] [<ffffffff8179dada>] dump_stack+0x45/0x57
[ 0.431533] [<ffffffff8179b757>] panic+0xc1/0x1f5
[ 0.436978] [<ffffffff813ad82b>] ? delay_tsc+0x3b/0x70
[ 0.442910] [<ffffffff8166fa2c>] panic_if_irq_remap+0x1c/0x20
[ 0.449524] [<ffffffff81d73645>] setup_IO_APIC+0x405/0x82e
[ 0.464979] [<ffffffff81d6fcc2>] native_smp_prepare_cpus+0x2d9/0x31c
[ 0.472274] [<ffffffff81d5d0ac>] kernel_init_freeable+0xd6/0x223
[ 0.479170] [<ffffffff81792ad0>] ? rest_init+0x80/0x80
[ 0.485099] [<ffffffff81792ade>] kernel_init+0xe/0xf0
[ 0.490932] [<ffffffff817a537c>] ret_from_fork+0x7c/0xb0
[ 0.497054] [<ffffffff81792ad0>] ? rest_init+0x80/0x80
[ 0.502983] ---[ end Kernel panic - not syncing: timer doesn't work through Interrupt-remapped IO-APIC
System hangs as below when X2APIC and interrupt remapping are both disabled:
[ 1.102782] pci 0000:00:02.0: System wakeup disabled by ACPI
[ 1.109351] pci 0000:00:03.0: System wakeup disabled by ACPI
[ 1.115915] pci 0000:00:03.2: System wakeup disabled by ACPI
[ 1.122479] pci 0000:00:03.3: System wakeup disabled by ACPI
[ 1.132274] pci 0000:00:1c.0: Enabling MPC IRBNCE
[ 1.137620] pci 0000:00:1c.0: Intel PCH root port ACS workaround enabled
[ 1.145239] pci 0000:00:1c.0: System wakeup disabled by ACPI
[ 1.151790] pci 0000:00:1c.7: Enabling MPC IRBNCE
[ 1.157128] pci 0000:00:1c.7: Intel PCH root port ACS workaround enabled
[ 1.164748] pci 0000:00:1c.7: System wakeup disabled by ACPI
[ 1.171447] pci 0000:00:1e.0: System wakeup disabled by ACPI
[ 1.178612] acpiphp: Slot [8] registered
[ 1.183095] pci 0000:00:02.0: PCI bridge to [bus 01]
[ 1.188867] acpiphp: Slot [2] registered
With this patch applied, the system panics in both cases with a proper
panic message.
Signed-off-by: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: iommu@lists.linux-foundation.org
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: David Rientjes <rientjes@google.com>
Cc: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>
Cc: Jan Beulich <JBeulich@suse.com>
Cc: Richard Weinberger <richard@nod.at>
Cc: Oren Twaig <oren@scalemp.com>
Link: http://lkml.kernel.org/r/1420615903-28253-5-git-send-email-jiang.liu@linux.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
If x2apic got disabled on the kernel command line, then the following
issue can happen:
enable_IR_x2apic()
....
x2apic_mode = 1;
enable_x2apic();
if (x2apic_disabled) {
__disable_x2apic();
return;
}
That leaves X2APIC disabled in hardware, but x2apic_mode stays 1. So
all other code which checks x2apic_mode gets the wrong information.
Set x2apic_mode to 0 after disabling it in hardware.
This is just a hotfix. The proper solution is to rework this code so
it has seperate functions for the initial setup on the boot processor
and the secondary cpus, but that's beyond the scope of this fix.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: iommu@lists.linux-foundation.org
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: David Rientjes <rientjes@google.com>
Cc: HATAYAMA Daisuke <d.hatayama@jp.fujitsu.com>
Cc: Jan Beulich <JBeulich@suse.com>
Cc: Richard Weinberger <richard@nod.at>
Cc: Oren Twaig <oren@scalemp.com>
Pull x86 apic updates from Thomas Gleixner:
"After stopping the full x86/apic branch, I took some time to go
through the first block of patches again, which are mostly cleanups
and preparatory work for the irqdomain conversion and ioapic hotplug
support.
Unfortunaly one of the real problematic commits was right at the
beginning, so I rebased this portion of the pending patches without
the offenders.
It would be great to get this into 3.19. That makes reworking the
problematic parts simpler. The usual tip testing did not unearth any
issues and it is fully bisectible now.
I'm pretty confident that this wont affect the calmness of the xmas
season.
Changes:
- Split the convoluted io_apic.c code into domain specific parts
(vector, ioapic, msi, htirq)
- Introduce proper helper functions to retrieve irq specific data
instead of open coded dereferencing of pointers
- Preparatory work for ioapic hotplug and irqdomain conversion
- Removal of the non functional pci-ioapic driver
- Removal of unused irq entry stubs
- Make native_smp_prepare_cpus() preemtible to avoid GFP_ATOMIC
allocations for everything which is called from there.
- Small cleanups and fixes"
* 'x86-apic-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (36 commits)
iommu/amd: Use helpers to access irq_cfg data structure associated with IRQ
iommu/vt-d: Use helpers to access irq_cfg data structure associated with IRQ
x86: irq_remapping: Use helpers to access irq_cfg data structure associated with IRQ
x86, irq: Use helpers to access irq_cfg data structure associated with IRQ
x86, irq: Make MSI and HT_IRQ indepenent of X86_IO_APIC
x86, irq: Move IRQ initialization routines from io_apic.c into vector.c
x86, irq: Move IOAPIC related declarations from hw_irq.h into io_apic.h
x86, irq: Move HT IRQ related code from io_apic.c into htirq.c
x86, irq: Move PCI MSI related code from io_apic.c into msi.c
x86, irq: Replace printk(KERN_LVL) with pr_lvl() utilities
x86, irq: Make UP version of irq_complete_move() an inline stub
x86, irq: Move local APIC related code from io_apic.c into vector.c
x86, irq: Introduce helpers to access struct irq_cfg
x86, irq: Protect __clear_irq_vector() with vector_lock
x86, irq: Rename local APIC related functions in io_apic.c as apic_xxx()
x86, irq: Refine hw_irq.h to prepare for irqdomain support
x86, irq: Convert irq_2_pin list to generic list
x86, irq: Kill useless parameter 'irq_attr' of IO_APIC_get_PCI_irq_vector()
x86, irq, acpi: Get rid of special handling of GSI for ACPI SCI
x86, irq: Introduce helper to check whether an IOAPIC has been registered
...
When X86_LOCAL_APIC (i.e. unconditionally on x86-64),
first_system_vector will never end up being higher than
LOCAL_TIMER_VECTOR (0xef), and hence building stubs for vectors
0xef...0xff is pointlessly reducing code density. Deal with this at
build time already.
Taking into consideration that X86_64 implies X86_LOCAL_APIC, also
simplify (and hence make easier to read and more consistent with the
change done here) some #if-s in arch/x86/kernel/irqinit.c.
While we could further improve the packing of the IRQ entry stubs (the
four ones now left in the last set could be fit into the four padding
bytes each of the final four sets have) this doesn't seem to provide
any real benefit: Both irq_entries_start and common_interrupt getting
cache line aligned, eliminating the 30th set would just produce 32
bytes of padding between the 29th and common_interrupt.
[ tglx: Folded lguest fix from Dan Carpenter ]
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: lguest@lists.ozlabs.org
Cc: Rusty Russell <rusty@rustcorp.com.au>
Link: http://lkml.kernel.org/r/54574D5F0200007800044389@mail.emea.novell.com
Link: http://lkml.kernel.org/r/20141115185718.GB6530@mwanda
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
After commit b2b49ccbdd (PM: Kconfig: Set PM_RUNTIME if PM_SLEEP is
selected) PM_RUNTIME is always set if PM is set, so #ifdef blocks
depending on CONFIG_PM_RUNTIME may now be changed to depend on
CONFIG_PM.
Replace CONFIG_PM_RUNTIME with CONFIG_PM in x86/kernel/apic/io_apic.c.
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Pull nmi-safe seq_buf printk update from Steven Rostedt:
"This code is a fork from the trace-3.19 pull as it needed the
trace_seq clean ups from that branch.
This code solves the issue of performing stack dumps from NMI context.
The issue is that printk() is not safe from NMI context as if the NMI
were to trigger when a printk() was being performed, the NMI could
deadlock from the printk() internal locks. This has been seen in
practice.
With lots of review from Petr Mladek, this code went through several
iterations, and we feel that it is now at a point of quality to be
accepted into mainline.
Here's what is contained in this patch set:
- Creates a "seq_buf" generic buffer utility that allows a descriptor
to be passed around where functions can write their own "printk()"
formatted strings into it. The generic version was pulled out of
the trace_seq() code that was made specifically for tracing.
- The seq_buf code was change to model the seq_file code. I have a
patch (not included for 3.19) that converts the seq_file.c code
over to use seq_buf.c like the trace_seq.c code does. This was
done to make sure that seq_buf.c is compatible with seq_file.c. I
may try to get that patch in for 3.20.
- The seq_buf.c file was moved to lib/ to remove it from being
dependent on CONFIG_TRACING.
- The printk() was updated to allow for a per_cpu "override" of the
internal calls. That is, instead of writing to the console, a call
to printk() may do something else. This made it easier to allow
the NMI to change what printk() does in order to call dump_stack()
without needing to update that code as well.
- Finally, the dump_stack from all CPUs via NMI code was converted to
use the seq_buf code. The caller to trigger the NMI code would
wait till all the NMIs finished, and then it would print the
seq_buf data to the console safely from a non NMI context
One added bonus is that this code also makes the NMI dump stack work
on PREEMPT_RT kernels. As printk() includes sleeping locks on
PREEMPT_RT, printk() only writes to console if the console does not
use any rt_mutex converted spin locks. Which a lot do"
* tag 'trace-seq-buf-3.19' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
x86/nmi: Fix use of unallocated cpumask_var_t
printk/percpu: Define printk_func when printk is not defined
x86/nmi: Perform a safe NMI stack trace on all CPUs
printk: Add per_cpu printk func to allow printk to be diverted
seq_buf: Move the seq_buf code to lib/
seq-buf: Make seq_buf_bprintf() conditional on CONFIG_BINARY_PRINTF
tracing: Add seq_buf_get_buf() and seq_buf_commit() helper functions
tracing: Have seq_buf use full buffer
seq_buf: Add seq_buf_can_fit() helper function
tracing: Add paranoid size check in trace_printk_seq()
tracing: Use trace_seq_used() and seq_buf_used() instead of len
tracing: Clean up tracing_fill_pipe_page()
seq_buf: Create seq_buf_used() to find out how much was written
tracing: Add a seq_buf_clear() helper and clear len and readpos in init
tracing: Convert seq_buf fields to be like seq_file fields
tracing: Convert seq_buf_path() to be like seq_path()
tracing: Create seq_buf layer in trace_seq
Pull x86 platform changes from Ingo Molnar:
"A handful of numachip APIC driver updates/fixes, and two small SGI/UV
fixes"
* 'x86-platform-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86: numachip: APIC driver cleanups
x86: numachip: Elide self-IPI ICR polling
x86: numachip: Fix 16-bit APIC ID truncation
* 'x86-uv-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86: UV BAU: Increase maximum CPUs per socket/hub
x86: UV BAU: Avoid NULL pointer reference in ptc_seq_show
Pull irq domain updates from Thomas Gleixner:
"The real interesting irq updates:
- Support for hierarchical irq domains:
For complex interrupt routing scenarios where more than one
interrupt related chip is involved we had no proper representation
in the generic interrupt infrastructure so far. That made people
implement rather ugly constructs in their nested irq chip
implementations. The main offenders are x86 and arm/gic.
To distangle that mess we have now hierarchical irqdomains which
seperate the various interrupt chips and connect them via the
hierarchical domains. That keeps the domain specific details
internal to the particular hierarchy level and removes the
criss/cross referencing of chip internals. The resulting hierarchy
for a complex x86 system will look like this:
vector mapped: 74
msi-0 mapped: 2
dmar-ir-1 mapped: 69
ioapic-1 mapped: 4
ioapic-0 mapped: 20
pci-msi-2 mapped: 45
dmar-ir-0 mapped: 3
ioapic-2 mapped: 1
pci-msi-1 mapped: 2
htirq mapped: 0
Neither ioapic nor pci-msi know about the dmar interrupt remapping
between themself and the vector domain. If interrupt remapping is
disabled ioapic and pci-msi become direct childs of the vector
domain.
In hindsight we should have done that years ago, but in hindsight
we always know better :)
- Support for generic MSI interrupt domain handling
We have more and more non PCI related MSI interrupts, so providing
a generic infrastructure for this is better than having all
affected architectures implementing their own private hacks.
- Support for PCI-MSI interrupt domain handling, based on the generic
MSI support.
This part carries the pci/msi branch from Bjorn Helgaas pci tree to
avoid a massive conflict. The PCI/MSI parts are acked by Bjorn.
I have two more branches on top of this. The full conversion of x86
to hierarchical domains and a partial conversion of arm/gic"
* 'irq-irqdomain-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (41 commits)
genirq: Move irq_chip_write_msi_msg() helper to core
PCI/MSI: Allow an msi_controller to be associated to an irq domain
PCI/MSI: Provide mechanism to alloc/free MSI/MSIX interrupt from irqdomain
PCI/MSI: Enhance core to support hierarchy irqdomain
PCI/MSI: Move cached entry functions to irq core
genirq: Provide default callbacks for msi_domain_ops
genirq: Introduce msi_domain_alloc/free_irqs()
asm-generic: Add msi.h
genirq: Add generic msi irq domain support
genirq: Introduce callback irq_chip.irq_write_msi_msg
genirq: Work around __irq_set_handler vs stacked domains ordering issues
irqdomain: Introduce helper function irq_domain_add_hierarchy()
irqdomain: Implement a method to automatically call parent domains alloc/free
genirq: Introduce helper irq_domain_set_info() to reduce duplicated code
genirq: Split out flow handler typedefs into seperate header file
genirq: Add IRQ_SET_MASK_OK_DONE to support stacked irqchip
genirq: Introduce irq_chip.irq_compose_msi_msg() to support stacked irqchip
genirq: Add more helper functions to support stacked irq_chip
genirq: Introduce helper functions to support stacked irq_chip
irqdomain: Do irq_find_mapping and set_type for hierarchy irqdomain in case OF
...
If the TSC is unusable or disabled, then this patch fixes:
- Confusion while trying to clear old APIC interrupts.
- Division by zero and incorrect programming of the TSC deadline
timer.
This fixes boot if the CPU has a TSC deadline timer but a missing or
broken TSC. The failure to boot can be observed with qemu using
-cpu qemu64,-tsc,+tsc-deadline
This also happens to me in nested KVM for unknown reasons.
With this patch, I can boot cleanly (although without a TSC).
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Cc: Bandan Das <bsd@redhat.com>
Cc: stable@vger.kernel.org
Link: http://lkml.kernel.org/r/e2fa274e498c33988efac0ba8b7e3120f7f92d78.1413393027.git.luto@amacapital.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Pull percpu consistent-ops changes from Tejun Heo:
"Way back, before the current percpu allocator was implemented, static
and dynamic percpu memory areas were allocated and handled separately
and had their own accessors. The distinction has been gone for many
years now; however, the now duplicate two sets of accessors remained
with the pointer based ones - this_cpu_*() - evolving various other
operations over time. During the process, we also accumulated other
inconsistent operations.
This pull request contains Christoph's patches to clean up the
duplicate accessor situation. __get_cpu_var() uses are replaced with
with this_cpu_ptr() and __this_cpu_ptr() with raw_cpu_ptr().
Unfortunately, the former sometimes is tricky thanks to C being a bit
messy with the distinction between lvalues and pointers, which led to
a rather ugly solution for cpumask_var_t involving the introduction of
this_cpu_cpumask_var_ptr().
This converts most of the uses but not all. Christoph will follow up
with the remaining conversions in this merge window and hopefully
remove the obsolete accessors"
* 'for-3.18-consistent-ops' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu: (38 commits)
irqchip: Properly fetch the per cpu offset
percpu: Resolve ambiguities in __get_cpu_var/cpumask_var_t -fix
ia64: sn_nodepda cannot be assigned to after this_cpu conversion. Use __this_cpu_write.
percpu: Resolve ambiguities in __get_cpu_var/cpumask_var_t
Revert "powerpc: Replace __get_cpu_var uses"
percpu: Remove __this_cpu_ptr
clocksource: Replace __this_cpu_ptr with raw_cpu_ptr
sparc: Replace __get_cpu_var uses
avr32: Replace __get_cpu_var with __this_cpu_write
blackfin: Replace __get_cpu_var uses
tile: Use this_cpu_ptr() for hardware counters
tile: Replace __get_cpu_var uses
powerpc: Replace __get_cpu_var uses
alpha: Replace __get_cpu_var
ia64: Replace __get_cpu_var uses
s390: cio driver &__get_cpu_var replacements
s390: Replace __get_cpu_var uses
mips: Replace __get_cpu_var uses
MIPS: Replace __get_cpu_var uses in FPU emulator.
arm: Replace __this_cpu_ptr with raw_cpu_ptr
...
Pull x86 ras, uv and vdso fixlets from Ingo Molnar:
"ras: tone down a kernel message to only occur during initial bootup,
not during suspend/resume cycles.
uv: a cleanup commit
vdso: a fix to error checking"
* 'x86-ras-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/mce: Avoid showing repetitive message from intel_init_thermal()
* 'x86-uv-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/apic/uv: Remove unnecessary #ifdef
* 'x86-vdso-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/vdso: Fix vdso2c's special_pages[] error checking
Pull x86 fixes from Ingo Molnar:
"Misc smaller fixes that missed the v3.17 cycle"
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/build: Add arch/x86/purgatory/ make generated files to gitignore
x86: Fix section conflict for numachip
x86: Reject x32 executables if x32 ABI not supported
x86_64, entry: Filter RFLAGS.NT on entry from userspace
x86, boot, kaslr: Fix nuisance warning on 32-bit builds
A commit in linux-next was causing boot to fail and bisection
identified the patch 4ba2968420 ("percpu: Resolve ambiguities in
__get_cpu_var/cpumask_var_"). One of the changes in that patch looks
very suspicious. Reverting the full patch fixes boot as does this
fixlet.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux.com>