Background:
The MCE handler has several paths that it can take, depending on various
conditions of the MCE status and the value of the 'tolerant' knob. The
exact semantics are not well defined and the code is a bit twisty.
Description:
This patch makes the MCE handler's behavior more clear by documenting the
behavior for various 'tolerant' levels. It also fixes or enhances
several small things in the handler. Specifically:
* If RIPV is set it is not safe to restart, so set the 'no way out'
flag rather than the 'kill it' flag.
* Don't panic() on correctable MCEs.
* If the _OVER bit is set *and* the _UC bit is set (meaning possibly
dropped uncorrected errors), set the 'no way out' flag.
* Use EIPV for testing whether an app can be killed (SIGBUS) rather
than RIPV. According to docs, EIPV indicates that the error is
related to the IP, while RIPV simply means the IP is valid to
restart from.
* Don't clear the MCi_STATUS registers until after the panic() path.
This leaves the status bits set after the panic() so clever BIOSes
can find them (and dumb BIOSes can do nothing).
This patch also calls nonseekable_open() in mce_open (as suggested by akpm).
Result:
Tolerant levels behave almost identically to how they always have, but
not it's well defined. There's a slightly higher chance of panic()ing
when multiple errors happen (a good thing, IMHO). If you take an MBE and
panic(), the error status bits are not cleared.
Alternatives:
None.
Testing:
I used software to inject correctable and uncorrectable errors. With
tolerant = 3, the system usually survives. With tolerant = 2, the system
usually panic()s (PCC) but not always. With tolerant = 1, the system
always panic()s. When the system panic()s, the BIOS is able to detect
that the cause of death was an MC4. I was not able to reproduce the
case of a non-PCC error in userspace, with EIPV, with (tolerant < 3).
That will be rare at best.
Signed-off-by: Tim Hockin <thockin@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
.. and adjust documentation to properly reflect options that are
x86-64 specific.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Background:
We've found that MCEs (specifically DRAM SBEs) tend to come in bunches,
especially when we are trying really hard to stress the system out. The
current MCE poller uses a static interval which does not care whether it
has or has not found MCEs recently.
Description:
This patch makes the MCE poller adjust the polling interval dynamically.
If we find an MCE, poll 2x faster (down to 10 ms). When we stop finding
MCEs, poll 2x slower (up to check_interval seconds). The check_interval
tunable becomes the max polling interval. The "Machine check events
logged" printk() is rate limited to the check_interval, which should be
identical behavior to the old functionality.
Result:
If you start to take a lot of correctable errors (not exceptions), you
log them faster and more accurately (less chance of overflowing the MCA
registers). If you don't take a lot of errors, you will see no change.
Alternatives:
I considered simply reducing the polling interval to 10 ms immediately
and keeping it there as long as we continue to find errors. This felt a
bit heavy handed, but does perform significantly better for the default
check_interval of 5 minutes (we're using a few seconds when testing for
DRAM errors). I could be convinced to go with this, if anyone felt it
was not too aggressive.
Testing:
I used an error-injecting DIMM to create lots of correctable DRAM errors
and verified that the polling interval accelerates. The printk() only
happens once per check_interval seconds.
Patch:
This patch is against 2.6.21-rc7.
Signed-Off-By: Tim Hockin <thockin@google.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Create a document to explain how to use numa=fake in conjunction with cpusets
for coarse memory resource management.
An attempt to get more awareness and testing for this feature.
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Extends the numa=fake x86_64 command-line option to split the remaining system
memory into nodes of fixed size. Any leftover memory is allocated to a final
node unless the command-line ends with a comma.
For example:
numa=fake=2*512,*128 gives two 512M nodes and the remaining system
memory is split into nodes of 128M each.
This is beneficial for systems where the exact size of RAM is unknown or not
necessarily relevant, but the size of the remaining nodes to be allocated is
known based on their capacity for resource management.
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Extends the numa=fake x86_64 command-line option to split the remaining
system memory into equal-sized nodes.
For example:
numa=fake=2*512,4* gives two 512M nodes and the remaining system
memory is split into four approximately equal
chunks.
This is beneficial for systems where the exact size of RAM is unknown or not
necessarily relevant, but the granularity with which nodes shall be allocated
is known.
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Extends the numa=fake x86_64 command-line option to allow for configurable
node sizes. These nodes can be used in conjunction with cpusets for coarse
memory resource management.
The old command-line option is still supported:
numa=fake=32 gives 32 fake NUMA nodes, ignoring the NUMA setup of the
actual machine.
But now you may configure your system for the node sizes of your choice:
numa=fake=2*512,1024,2*256
gives two 512M nodes, one 1024M node, two 256M nodes, and
the rest of system memory to a sixth node.
The existing hash function is maintained to support the various node sizes
that are possible with this implementation.
Each node of the same size receives roughly the same amount of available
pages, regardless of any reserved memory with its address range. The total
available pages on the system is calculated and divided by the number of equal
nodes to allocate. These nodes are then dynamically allocated and their
borders extended until such time as their number of available pages reaches
the required size.
Configurable node sizes are recommended when used in conjunction with cpusets
for memory control because it eliminates the overhead associated with scanning
the zonelists of many smaller full nodes on page_alloc().
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
noreplacement is dangerous on modern systems because it will not replace the
context switch FNSAVE with SSE aware FXSAVE. But other places in the kernel still assume
SSE and do FXSAVE and the CPU will then access FXSAVE information with
FNSAVE and cause corruption.
Easiest way to avoid this is to remove the option. It was mostly for paranoia
reasons anyways and alternative()s have been stable for some time.
Thanks to Jeremy F. for reporting and helping debug it.
Signed-off-by: Andi Kleen <ak@suse.de>
When a machine check event is detected (including a AMD RevF threshold
overflow event) allow to run a "trigger" program. This allows user space
to react to such events sooner.
The trigger is configured using a new trigger entry in the
machinecheck sysfs interface. It is currently shared between
all CPUs.
I also fixed the AMD threshold handler to run the machine
check polling code immediately to actually log any events
that might have caused the threshold interrupt.
Also added some documentation for the mce sysfs interface.
Signed-off-by: Andi Kleen <ak@suse.de>
Fix typos.
Lots of whitespace changes for readability and consistency.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andi Kleen <ak@suse.de>
- add SWIOTLB config help text
- mention Documentation/x86_64/boot-options.txt in
Documentation/kernel-parameters.txt
- remove the duplication of the iommu kernel parameter documentation.
- Better explanation of some of the iommu kernel parameter options.
- "32MB<<order" instead of "32MB^order".
- Mention the default "order" value.
- list the four existing PCI-DMA mapping implementations of arch x86_64
- group the iommu= option keywords by PCI-DMA mapping implementation.
- Distinguish iommu= option keywords from number arguments.
- Explain the meaning of DAC and SAC.
Signed-off-by: Karsten Weiss <knweiss@science-computing.de>
Signed-off-by: Andi Kleen <ak@suse.de>
Acked-by: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
This reverts commit b026872601, which has
been linked to several problem reports with IO-APIC and the timer.
Machines either don't boot because the timer doesn't happen, or we get
double timer interrupts because we end up double-routing the timer irq
through multiple interfaces.
See for example
http://lkml.org/lkml/2006/12/16/101http://lkml.org/lkml/2007/1/3/9http://bugzilla.kernel.org/show_bug.cgi?id=7789
about some of the discussion.
Patches to fix this cleanup exist (and have been confirmed to work fine
at least for some of the affected cases) and we'll revisit it for
2.6.21, but this late in the -rc series we're better off just reverting
the incomplete commit that caused the problems.
Suggested-by: Adrian Bunk <bunk@stusta.de>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Yinghai Lu <yinghai.lu@amd.com>
Cc: Andrew Morton <akpm@osdl.org>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch makes it possible to compile Calgary in but not use it by
default. In this mode, use 'iommu=calgary' to activate it.
Signed-off-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: Andi Kleen <ak@suse.de>
Instead of adding all kinds of more quirks try various timer
routing variants in check_timer.
In particular this tries to handle quirks from:
- Nvidia NF2-4 reference BIOS: wrong timer override
- Asus: Wrong timer override but no HPET table
- ATI: require timer disabled in 8259
- Some boards: require timer enabled in 8259
We just try many of the the known variants in the hopefully right order
in check_timer.
Trying pin 0/2 on Nvidia suggested by Tim Hockin.
TBD Experimental. Needs a lot of testing
Signed-off-by: Andi Kleen <ak@suse.de>
This patch fixes typos in various Documentation txts. The patch addresses
some words starting with the letters 'Q'-'R'.
Signed-off-by: Matt LaPlante <kernel1@cyberdogtech.com>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Describes the stack organization on x86-64.
I changed it a bit and removed some obsolete information and the
questions.
Cc: kaos@sgi.com
Signed-off-by: Andi Kleen <ak@suse.de>
This patch hooks Calgary into the build, the x86-64 IOMMU
initialization paths, and introduces the Calgary specific bits. The
implementation draws inspiration from both PPC (which has support for
the same chip but requires firmware support which we don't have on
x86-64) and gart. Calgary is different from gart in that it support a
translation table per PHB, as opposed to the single gart aperture.
Changes from previous version:
* Addition of boot-time disablement for bus-level translation/isolation
(e.g, enable userspace DMA for things like X)
* Usage of newer IOMMU abstraction functions
Signed-off-by: Muli Ben-Yehuda <muli@il.ibm.com>
Signed-off-by: Jon Mason <jdmason@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
From: Keith Mannthey, Andi Kleen
Implement memory hotadd without sparsemem. The memory in the SRAT
hotadd area is just preserved instead and can be activated later.
There are a few restrictions:
- Only one continuous hotadd area allowed per node
The main problem is dealing with the many buggy SRAT tables
that are out there. The strategy here is to reject anything
suspicious.
Originally from Keith Mannthey, with several hacks and changes by AK
and also contributions from Andrew Morton
[ TBD: Problems pointed out by KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>:
1) Goto's rebuild_zonelist patch will not work if CONFIG_MEMORY_HOTPLUG=n.
Rebuilding zonelist is necessary when the system has just memory <
4G at boot, and hot add memory > 4G. because x86_64 has DMA32,
ZONE_NORAML is not included into zonelist at boot time if system
doesn't have memory >4G at boot.
[AK: should just force the higher zones at boot time when SRAT tells us]
2) zone and node's spanned_pages and present_pages are not incremented.
They should be.
For example, our server (ia64/Fujitsu PrimeQuest) can equip memory
from 4G to 1T(maybe 2T in future), and SRAT will *always* say we have
possible 1T +memory. (Microsoft requires "write all possible memory
in SRAT") When we reserve memmap for possible 1T memory, Linux will
not work well in +minimum 4G configuraion ;)
[AK: needs limiting to 5-10% of max memory]
]
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The previous experiment for using apicmaintimer on ATI systems didn't
work out very well. In particular laptops with C2/C3 support often
don't let it tick during idle, which makes it useless. There were also
some other bugs that made the apicmaintimer often not used at all.
I tried some other experiments - running timer over RTC and some other
things but they didn't really work well neither.
I rechecked the specs now and it turns out this simple change is
actually enough to avoid the double ticks on the ATI systems. We just
turn off IRQ 0 in the 8254 and only route it directly using the IO-APIC.
I tested it on a few ATI systems and it worked there. In fact it worked
on all chipsets (NVidia, Intel, AMD, ATI) I tried it on.
According to the ACPI spec routing should always work through the
IO-APIC so I think it's the correct thing to do anyways (and most of the
old gunk in check_timer should be thrown away for x86-64).
But for 2.6.16 it's best to do a fairly minimal change:
- Use the known to be working everywhere-but-ATI IRQ0 both over 8254
and IO-APIC setup everywhere
- Except on ATI disable IRQ0 in the 8254
- Remove the code to select apicmaintimer on ATI chipsets
- Add some boot options to allow to override this (just paranoia)
In 2.6.17 I hope to switch the default over to this for everybody.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
On some broken motherboards (at least one NForce3 based AMD64 laptop)
the PIT timer runs at a incorrect frequency. This patch adds a new
option "apicpmtimer" that allows to use the APIC timer and calibrate it
using the PMTimer. It requires the earlier patch that allows to run the
main timer from the APIC.
Specifying apicpmtimer implies apicmaintimer.
The option defaults to off for now.
I tested it on a few systems and the resulting APIC timer frequencies
were usually a bit off, but always <1%, which should be tolerable.
TBD figure out heuristic to enable this automatically on the affected
systems TBD perhaps do it on all NForce3s or using DMI?
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Another piece from the no-idle-tick patch.
This can be enabled with the "apicmaintimer" option.
This is mainly useful when the PIT/HPET interrupt is unreliable.
Note there are some systems that are known to stop the APIC
timer in C3. For those it will never work, but this case
should be automatically detected.
It also only works with PM timer right now. When HPET is used
the way the main timer handler computes the delay doesn't work.
It should be a bit more efficient because there is one less
regular interrupt to process on the boot processor.
Requires earlier bugfix from Venkatesh
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Most users don't need it so no need to waste memory.
This means an user has to specify the appropiate number of
hotplug CPUs on the command line with additional_cpus=...
or fix their BIOS to follow the convention in
Documentation/x86-64/cpu-hotplug-spec
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
CONFIG_CHECKING covered some debugging code used in the early times
of the port. But it wasn't even SMP safe for quite some time
and the bugs it checked for seem to be gone.
This patch removes all the code to verify GS at kernel entry. There
haven't been any new bugs in this area for a long time.
Previously it also covered the sysctl for the page fault tracing.
That didn't make much sense because that code was unconditionally
compiled in. I made that a boot option now because it is typically
only useful at boot.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The logging for boot errors was turned off because it was broken
on some AMD systems. But give Intel EM64T systems a chance because they are
supposed to be correct there.
The advantage is that there is a chance to actually log uncorrected
machine checks after the reset.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
With a NR_CPUS==128 kernel with CPU hotplug enabled we would waste 4MB
on per CPU data of all possible CPUs. The reason was that HOTPLUG
always set up possible map to NR_CPUS cpus and then we need to allocate
that much (each per CPU data is roughly ~32k now)
The underlying problem is that ACPI didn't tell us how many hotplug CPUs
the platform supports. So the old code just assumed all, which would
lead to this memory wastage.
This implements some new heuristics:
- If the BIOS specified disabled CPUs in the ACPI/mptables assume they
can be enabled later (this is bending the ACPI specification a bit,
but seems like a obvious extension)
- The user can overwrite it with a new additionals_cpus=NUM option
- Otherwise use half of the available CPUs or 2, whatever is more.
Cc: ashok.raj@intel.com
Cc: len.brown@intel.com
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I got some questions on this, so just fix up the documentation.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Don't log machine check events left over from boot. Too many BIOSes leave
bogus events in there.
This unfortunately also makes it impossible to log events that caused a
reboot. For people with non broken BIOS there is mce=bootlog
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This works around the too fast timer seen on some ATI boards.
I don't feel confident enough about it yet to enable it by default, but give
users the option.
Patch and debugging from Christopher Allen Wing <wingc@engin.umich.edu>, with
minor tweaks (renamed the option and documented it)
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Tim Hockin