commit edfe63ec97 upstream.
A Xorg failure on qemu32 was reported as a regression [1] caused by
commit 9cd25aac1f ("x86/mm/pat: Emulate PAT when it is disabled").
This patch fixes the Xorg crash.
Negative effects of this regression were the following two failures [2]
in Xorg on QEMU with QEMU CPU model "qemu32" (-cpu qemu32), which were
triggered by the fact that its virtual CPU does not support MTRRs.
#1. copy_process() failed in the check in reserve_pfn_range()
copy_process
copy_mm
dup_mm
dup_mmap
copy_page_range
track_pfn_copy
reserve_pfn_range
A WC map request was tracked as WC in memtype, which set a PTE as
UC (pgprot) per __cachemode2pte_tbl[]. This led to this error in
reserve_pfn_range() called from track_pfn_copy(), which obtained
a pgprot from a PTE. It converts pgprot to page_cache_mode, which
does not necessarily result in the original page_cache_mode since
__cachemode2pte_tbl[] redirects multiple types to UC.
#2. error path in copy_process() then hit WARN_ON_ONCE in
untrack_pfn().
x86/PAT: Xorg:509 map pfn expected mapping type uncached-
minus for [mem 0xfd000000-0xfdffffff], got write-combining
Call Trace:
dump_stack
warn_slowpath_common
? untrack_pfn
? untrack_pfn
warn_slowpath_null
untrack_pfn
? __kunmap_atomic
unmap_single_vma
? pagevec_move_tail_fn
unmap_vmas
exit_mmap
mmput
copy_process.part.47
_do_fork
SyS_clone
do_syscall_32_irqs_on
entry_INT80_32
These negative effects are caused by two separate bugs, but they
can be addressed in separate patches. Fixing the pat_init() issue
described below addresses the root cause, and avoids Xorg to hit
these cases.
When the CPU does not support MTRRs, MTRR does not call pat_init(),
which leaves PAT enabled without initializing PAT. This pat_init()
issue is a long-standing issue, but manifested as issue #1 (and then
hit issue #2) with the above-mentioned commit because the memtype
now tracks cache attribute with 'page_cache_mode'.
This pat_init() issue existed before the commit, but we used pgprot
in memtype. Hence, we did not have issue #1 before. But WC request
resulted in WT in effect because WC pgrot is actually WT when PAT
is not initialized. This is not how it was designed to work. When
PAT is set to disable properly, WC is converted to UC. The use of
WT can result in a system crash if the target range does not support
WT. Fortunately, nobody ran into such issue before.
To fix this pat_init() issue, PAT code has been enhanced to provide
pat_disable() interface. Call this interface when MTRRs are disabled.
By setting PAT to disable properly, PAT bypasses the memtype check,
and avoids issue #1.
[1]: https://lkml.org/lkml/2016/3/3/828
[2]: https://lkml.org/lkml/2016/3/4/775
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Borislav Petkov <bp@suse.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Luis R. Rodriguez <mcgrof@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: elliott@hpe.com
Cc: konrad.wilk@oracle.com
Cc: paul.gortmaker@windriver.com
Cc: xen-devel@lists.xenproject.org
Link: http://lkml.kernel.org/r/1458769323-24491-5-git-send-email-toshi.kani@hpe.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 749d088b8e upstream.
Protocol for the "version" fields is: hypervisor raises it (making it
uneven) before it starts updating the fields and raises it again (making
it even) when it is done. Thus the guest can make sure the time values
it got are consistent by checking the version before and after reading
them.
Add CPU barries after getting version value just like what function
vread_pvclock does, because all of callees in this function is inline.
Fixes: 502dfeff23
Signed-off-by: Minfei Huang <mnghuan@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit abb2bafd29 upstream.
The EFI firmware on Macs contains a full-fledged network stack for
downloading OS X images from osrecovery.apple.com. Unfortunately
on Macs introduced 2011 and 2012, EFI brings up the Broadcom 4331
wireless card on every boot and leaves it enabled even after
ExitBootServices has been called. The card continues to assert its IRQ
line, causing spurious interrupts if the IRQ is shared. It also corrupts
memory by DMAing received packets, allowing for remote code execution
over the air. This only stops when a driver is loaded for the wireless
card, which may be never if the driver is not installed or blacklisted.
The issue seems to be constrained to the Broadcom 4331. Chris Milsted
has verified that the newer Broadcom 4360 built into the MacBookPro11,3
(2013/2014) does not exhibit this behaviour. The chances that Apple will
ever supply a firmware fix for the older machines appear to be zero.
The solution is to reset the card on boot by writing to a reset bit in
its mmio space. This must be done as an early quirk and not as a plain
vanilla PCI quirk to successfully combat memory corruption by DMAed
packets: Matthew Garrett found out in 2012 that the packets are written
to EfiBootServicesData memory (http://mjg59.dreamwidth.org/11235.html).
This type of memory is made available to the page allocator by
efi_free_boot_services(). Plain vanilla PCI quirks run much later, in
subsys initcall level. In-between a time window would be open for memory
corruption. Random crashes occurring in this time window and attributed
to DMAed packets have indeed been observed in the wild by Chris
Bainbridge.
When Matthew Garrett analyzed the memory corruption issue in 2012, he
sought to fix it with a grub quirk which transitions the card to D3hot:
http://git.savannah.gnu.org/cgit/grub.git/commit/?id=9d34bb85da56
This approach does not help users with other bootloaders and while it
may prevent DMAed packets, it does not cure the spurious interrupts
emanating from the card. Unfortunately the card's mmio space is
inaccessible in D3hot, so to reset it, we have to undo the effect of
Matthew's grub patch and transition the card back to D0.
Note that the quirk takes a few shortcuts to reduce the amount of code:
The size of BAR 0 and the location of the PM capability is identical
on all affected machines and therefore hardcoded. Only the address of
BAR 0 differs between models. Also, it is assumed that the BCMA core
currently mapped is the 802.11 core. The EFI driver seems to always take
care of this.
Michael Büsch, Bjorn Helgaas and Matt Fleming contributed feedback
towards finding the best solution to this problem.
The following should be a comprehensive list of affected models:
iMac13,1 2012 21.5" [Root Port 00:1c.3 = 8086:1e16]
iMac13,2 2012 27" [Root Port 00:1c.3 = 8086:1e16]
Macmini5,1 2011 i5 2.3 GHz [Root Port 00:1c.1 = 8086:1c12]
Macmini5,2 2011 i5 2.5 GHz [Root Port 00:1c.1 = 8086:1c12]
Macmini5,3 2011 i7 2.0 GHz [Root Port 00:1c.1 = 8086:1c12]
Macmini6,1 2012 i5 2.5 GHz [Root Port 00:1c.1 = 8086:1e12]
Macmini6,2 2012 i7 2.3 GHz [Root Port 00:1c.1 = 8086:1e12]
MacBookPro8,1 2011 13" [Root Port 00:1c.1 = 8086:1c12]
MacBookPro8,2 2011 15" [Root Port 00:1c.1 = 8086:1c12]
MacBookPro8,3 2011 17" [Root Port 00:1c.1 = 8086:1c12]
MacBookPro9,1 2012 15" [Root Port 00:1c.1 = 8086:1e12]
MacBookPro9,2 2012 13" [Root Port 00:1c.1 = 8086:1e12]
MacBookPro10,1 2012 15" [Root Port 00:1c.1 = 8086:1e12]
MacBookPro10,2 2012 13" [Root Port 00:1c.1 = 8086:1e12]
For posterity, spurious interrupts caused by the Broadcom 4331 wireless
card resulted in splats like this (stacktrace omitted):
irq 17: nobody cared (try booting with the "irqpoll" option)
handlers:
[<ffffffff81374370>] pcie_isr
[<ffffffffc0704550>] sdhci_irq [sdhci] threaded [<ffffffffc07013c0>] sdhci_thread_irq [sdhci]
[<ffffffffc0a0b960>] azx_interrupt [snd_hda_codec]
Disabling IRQ #17
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=79301
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=111781
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=728916
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=895951#c16
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1009819
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1098621
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1149632#c5
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1279130
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1332732
Tested-by: Konstantin Simanov <k.simanov@stlk.ru> # [MacBookPro8,1]
Tested-by: Lukas Wunner <lukas@wunner.de> # [MacBookPro9,1]
Tested-by: Bryan Paradis <bryan.paradis@gmail.com> # [MacBookPro9,2]
Tested-by: Andrew Worsley <amworsley@gmail.com> # [MacBookPro10,1]
Tested-by: Chris Bainbridge <chris.bainbridge@gmail.com> # [MacBookPro10,2]
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Acked-by: Rafał Miłecki <zajec5@gmail.com>
Acked-by: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Chris Milsted <cmilsted@redhat.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Michael Buesch <m@bues.ch>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: b43-dev@lists.infradead.org
Cc: linux-pci@vger.kernel.org
Cc: linux-wireless@vger.kernel.org
Link: http://lkml.kernel.org/r/48d0972ac82a53d460e5fce77a07b2560db95203.1465690253.git.lukas@wunner.de
[ Did minor readability edits. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 850c321027 upstream.
We used to scan secondary buses until the following commit that
was applied in 2009:
8659c406ad ("x86: only scan the root bus in early PCI quirks")
which commit constrained early quirks to the root bus only. Its
motivation was to prevent application of the nvidia_bugs quirk
on secondary buses.
We're about to add a quirk to reset the Broadcom 4331 wireless card on
2011/2012 Macs, which is located on a secondary bus behind a PCIe root
port. To facilitate that, reintroduce scanning of secondary buses.
The commit message of 8659c406ad notes that scanning only the root bus
"saves quite some unnecessary scanning work". The algorithm used prior
to 8659c406ad was particularly time consuming because it scanned
buses 0 to 31 brute force. To avoid lengthening boot time, employ a
recursive strategy which only scans buses that are actually reachable
from the root bus.
Yinghai Lu pointed out that the secondary bus number read from a
bridge's config space may be invalid, in particular a value of 0 would
cause an infinite loop. The PCI core goes beyond that and recurses to a
child bus only if its bus number is greater than the parent bus number
(see pci_scan_bridge()). Since the root bus is numbered 0, this implies
that secondary buses may not be 0. Do the same on early scanning.
If this algorithm is found to significantly impact boot time or cause
infinite loops on broken hardware, it would be possible to limit its
recursion depth: The Broadcom 4331 quirk applies at depth 1, all others
at depth 0, so the bus need not be scanned deeper than that for now. An
alternative approach would be to revert to scanning only the root bus,
and apply the Broadcom 4331 quirk to the root ports 8086:1c12, 8086:1e12
and 8086:1e16. Apple always positioned the card behind either of these
three ports. The quirk would then check presence of the card in slot 0
below the root port and do its deed.
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: linux-pci@vger.kernel.org
Link: http://lkml.kernel.org/r/f0daa70dac1a9b2483abdb31887173eb6ab77bdf.1465690253.git.lukas@wunner.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1ead852dd8 upstream.
Fix boot crash that triggers if this driver is built into a kernel and
run on non-AMD systems.
AMD northbridges users call amd_cache_northbridges() and it returns
a negative value to signal that we weren't able to cache/detect any
northbridges on the system.
At least, it should do so as all its callers expect it to do so. But it
does return a negative value only when kmalloc() fails.
Fix it to return -ENODEV if there are no NBs cached as otherwise, amd_nb
users like amd64_edac, for example, which relies on it to know whether
it should load or not, gets loaded on systems like Intel Xeons where it
shouldn't.
Reported-and-tested-by: Tony Battersby <tonyb@cybernetics.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1466097230-5333-2-git-send-email-bp@alien8.de
Link: https://lkml.kernel.org/r/5761BEB0.9000807@cybernetics.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit dcfc47248d upstream.
Fix kprobe_fault_handler() to clear the TF (trap flag) bit of
the flags register in the case of a fault fixup on single-stepping.
If we put a kprobe on the instruction which caused a
page fault (e.g. actual mov instructions in copy_user_*),
that fault happens on the single-stepping buffer. In this
case, kprobes resets running instance so that the CPU can
retry execution on the original ip address.
However, current code forgets to reset the TF bit. Since this
fault happens with TF bit set for enabling single-stepping,
when it retries, it causes a debug exception and kprobes
can not handle it because it already reset itself.
On the most of x86-64 platform, it can be easily reproduced
by using kprobe tracer. E.g.
# cd /sys/kernel/debug/tracing
# echo p copy_user_enhanced_fast_string+5 > kprobe_events
# echo 1 > events/kprobes/enable
And you'll see a kernel panic on do_debug(), since the debug
trap is not handled by kprobes.
To fix this problem, we just need to clear the TF bit when
resetting running kprobe.
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Reviewed-by: Ananth N Mavinakayanahalli <ananth@linux.vnet.ibm.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: systemtap@sourceware.org
Link: http://lkml.kernel.org/r/20160611140648.25885.37482.stgit@devbox
[ Updated the comments. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit ab92b232ae upstream.
Currently, the PT driver always sets the PMI bit one region (page) before
the STOP region so that we can wake up the consumer before we run out of
room in the buffer and have to disable the event. However, we also need
an interrupt in the last output region, so that we actually get to disable
the event (if no more room from new data is available at that point),
otherwise hardware just quietly refuses to start, but the event is
scheduled in and we end up losing trace data till the event gets removed.
For a cpu-wide event it is even worse since there may not be any
re-scheduling at all and no chance for the ring buffer code to notice
that its buffer is filled up and the event needs to be disabled (so that
the consumer can re-enable it when it finishes reading the data out). In
other words, all the trace data will be lost after the buffer gets filled
up.
This patch makes PT also generate a PMI when the last output region is
full.
Reported-by: Markus Metzger <markus.t.metzger@intel.com>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1462886313-13660-2-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit a21211672c upstream.
There are several reports of freeze on enabling HWP (Hardware PStates)
feature on Skylake-based systems by the Intel P-states driver. The root
cause is identified as the HWP interrupts causing BIOS code to freeze.
HWP interrupts use the thermal LVT which can be handled by Linux
natively, but on the affected Skylake-based systems SMM will respond
to it by default. This is a problem for several reasons:
- On the affected systems the SMM thermal LVT handler is broken (it
will crash when invoked) and a BIOS update is necessary to fix it.
- With thermal interrupt handled in SMM we lose all of the reporting
features of the arch/x86/kernel/cpu/mcheck/therm_throt driver.
- Some thermal drivers like x86-package-temp depend on the thermal
threshold interrupts signaled via the thermal LVT.
- The HWP interrupts are useful for debugging and tuning
performance (if the kernel can handle them).
The native handling of thermal interrupts needs to be enabled
because of that.
This requires some way to tell SMM that the OS can handle thermal
interrupts. That can be done by using _OSC/_PDC in processor
scope very early during ACPI initialization.
The meaning of _OSC/_PDC bit 12 in processor scope is whether or
not the OS supports native handling of interrupts for Collaborative
Processor Performance Control (CPPC) notifications. Since on
HWP-capable systems CPPC is a firmware interface to HWP, setting
this bit effectively tells the firmware that the OS will handle
thermal interrupts natively going forward.
For details on _OSC/_PDC refer to:
http://www.intel.com/content/www/us/en/standards/processor-vendor-specific-acpi-specification.html
To implement the _OSC/_PDC handshake as described, introduce a new
function, acpi_early_processor_osc(), that walks the ACPI
namespace looking for ACPI processor objects and invokes _OSC for
them with bit 12 in the capabilities buffer set and terminates the
namespace walk on the first success.
Also modify intel_thermal_interrupt() to clear HWP status bits in
the HWP_STATUS MSR to acknowledge HWP interrupts (which prevents
them from firing continuously).
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
[ rjw: Subject & changelog, function rename ]
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 886123fb3a upstream.
Currently we read the tsc radio: ratio = (MSR_PLATFORM_INFO >> 8) & 0x1f;
Thus we get bit 8-12 of MSR_PLATFORM_INFO, however according to the SDM
(35.5), the ratio bits are bit 8-15.
Ignoring the upper bits can result in an incorrect tsc ratio, which causes the
TSC calibration and the Local APIC timer frequency to be incorrect.
Fix this problem by masking 0xff instead.
[ tglx: Massaged changelog ]
Fixes: 7da7c15613 "x86, tsc: Add static (MSR) TSC calibration on Intel Atom SoCs"
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Bin Gao <bin.gao@intel.com>
Cc: Len Brown <lenb@kernel.org>
Link: http://lkml.kernel.org/r/1462505619-5516-1-git-send-email-yu.c.chen@intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 1bdb897039 upstream.
If x86_vector_alloc_irq() fails x86_vector_free_irqs() is invoked to cleanup
the already allocated vectors. This subsequently calls clear_vector_irq().
The failed irq has no vector assigned, which triggers the BUG_ON(!vector) in
clear_vector_irq().
We cannot suppress the call to x86_vector_free_irqs() for the failed
interrupt, because the other data related to this irq must be cleaned up as
well. So calling clear_vector_irq() with vector == 0 is legitimate.
Remove the BUG_ON and return if vector is zero,
[ tglx: Massaged changelog ]
Fixes: b5dc8e6c21 "x86/irq: Use hierarchical irqdomain to manage CPU interrupt vectors"
Signed-off-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit c3d266c8a9 upstream.
This patch tries to fix a PEBS warning found in my stress test. The
following perf command can easily trigger the pebs warning or spurious
NMI error on Skylake/Broadwell/Haswell platforms:
sudo perf record -e 'cpu/umask=0x04,event=0xc4/pp,cycles,branches,ref-cycles,cache-misses,cache-references' --call-graph fp -b -c1000 -a
Also the NMI watchdog must be enabled.
For this case, the events number is larger than counter number. So
perf has to do multiplexing.
In perf_mux_hrtimer_handler, it does perf_pmu_disable(), schedule out
old events, rotate_ctx, schedule in new events and finally
perf_pmu_enable().
If the old events include precise event, the MSR_IA32_PEBS_ENABLE
should be cleared when perf_pmu_disable(). The MSR_IA32_PEBS_ENABLE
should keep 0 until the perf_pmu_enable() is called and the new event is
precise event.
However, there is a corner case which could restore PEBS_ENABLE to
stale value during the above period. In perf_pmu_disable(), GLOBAL_CTRL
will be set to 0 to stop overflow and followed PMI. But there may be
pending PMI from an earlier overflow, which cannot be stopped. So even
GLOBAL_CTRL is cleared, the kernel still be possible to get PMI. At
the end of the PMI handler, __intel_pmu_enable_all() will be called,
which will restore the stale values if old events haven't scheduled
out.
Once the stale pebs value is set, it's impossible to be corrected if
the new events are non-precise. Because the pebs_enabled will be set
to 0. x86_pmu.enable_all() will ignore the MSR_IA32_PEBS_ENABLE
setting. As a result, the following NMI with stale PEBS_ENABLE
trigger pebs warning.
The pending PMI after enabled=0 will become harmless if the NMI handler
does not change the state. This patch checks cpuc->enabled in pmi and
only restore the state when PMU is active.
Here is the dump:
Call Trace:
<NMI> [<ffffffff813c3a2e>] dump_stack+0x63/0x85
[<ffffffff810a46f2>] warn_slowpath_common+0x82/0xc0
[<ffffffff810a483a>] warn_slowpath_null+0x1a/0x20
[<ffffffff8100fe2e>] intel_pmu_drain_pebs_nhm+0x2be/0x320
[<ffffffff8100caa9>] intel_pmu_handle_irq+0x279/0x460
[<ffffffff810639b6>] ? native_write_msr_safe+0x6/0x40
[<ffffffff811f290d>] ? vunmap_page_range+0x20d/0x330
[<ffffffff811f2f11>] ? unmap_kernel_range_noflush+0x11/0x20
[<ffffffff8148379f>] ? ghes_copy_tofrom_phys+0x10f/0x2a0
[<ffffffff814839c8>] ? ghes_read_estatus+0x98/0x170
[<ffffffff81005a7d>] perf_event_nmi_handler+0x2d/0x50
[<ffffffff810310b9>] nmi_handle+0x69/0x120
[<ffffffff810316f6>] default_do_nmi+0xe6/0x100
[<ffffffff810317f2>] do_nmi+0xe2/0x130
[<ffffffff817aea71>] end_repeat_nmi+0x1a/0x1e
[<ffffffff810639b6>] ? native_write_msr_safe+0x6/0x40
[<ffffffff810639b6>] ? native_write_msr_safe+0x6/0x40
[<ffffffff810639b6>] ? native_write_msr_safe+0x6/0x40
<<EOE>> <IRQ> [<ffffffff81006df8>] ? x86_perf_event_set_period+0xd8/0x180
[<ffffffff81006eec>] x86_pmu_start+0x4c/0x100
[<ffffffff8100722d>] x86_pmu_enable+0x28d/0x300
[<ffffffff811994d7>] perf_pmu_enable.part.81+0x7/0x10
[<ffffffff8119cb70>] perf_mux_hrtimer_handler+0x200/0x280
[<ffffffff8119c970>] ? __perf_install_in_context+0xc0/0xc0
[<ffffffff8110f92d>] __hrtimer_run_queues+0xfd/0x280
[<ffffffff811100d8>] hrtimer_interrupt+0xa8/0x190
[<ffffffff81199080>] ? __perf_read_group_add.part.61+0x1a0/0x1a0
[<ffffffff81051bd8>] local_apic_timer_interrupt+0x38/0x60
[<ffffffff817af01d>] smp_apic_timer_interrupt+0x3d/0x50
[<ffffffff817ad15c>] apic_timer_interrupt+0x8c/0xa0
<EOI> [<ffffffff81199080>] ? __perf_read_group_add.part.61+0x1a0/0x1a0
[<ffffffff81123de5>] ? smp_call_function_single+0xd5/0x130
[<ffffffff81123ddb>] ? smp_call_function_single+0xcb/0x130
[<ffffffff81199080>] ? __perf_read_group_add.part.61+0x1a0/0x1a0
[<ffffffff8119765a>] event_function_call+0x10a/0x120
[<ffffffff8119c660>] ? ctx_resched+0x90/0x90
[<ffffffff811971e0>] ? cpu_clock_event_read+0x30/0x30
[<ffffffff811976d0>] ? _perf_event_disable+0x60/0x60
[<ffffffff8119772b>] _perf_event_enable+0x5b/0x70
[<ffffffff81197388>] perf_event_for_each_child+0x38/0xa0
[<ffffffff811976d0>] ? _perf_event_disable+0x60/0x60
[<ffffffff811a0ffd>] perf_ioctl+0x12d/0x3c0
[<ffffffff8134d855>] ? selinux_file_ioctl+0x95/0x1e0
[<ffffffff8124a3a1>] do_vfs_ioctl+0xa1/0x5a0
[<ffffffff81036d29>] ? sched_clock+0x9/0x10
[<ffffffff8124a919>] SyS_ioctl+0x79/0x90
[<ffffffff817ac4b2>] entry_SYSCALL_64_fastpath+0x1a/0xa4
---[ end trace aef202839fe9a71d ]---
Uhhuh. NMI received for unknown reason 2d on CPU 2.
Do you have a strange power saving mode enabled?
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Link: http://lkml.kernel.org/r/1457046448-6184-1-git-send-email-kan.liang@intel.com
[ Fixed various typos and other small details. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 8077eca079 upstream.
This patch fixes an issue with the GLOBAL_OVERFLOW_STATUS bits on
Haswell, Broadwell and Skylake processors when using PEBS.
The SDM stipulates that when the PEBS iterrupt threshold is crossed,
an interrupt is posted and the kernel is interrupted. The kernel will
find GLOBAL_OVF_SATUS bit 62 set indicating there are PEBS records to
drain. But the bits corresponding to the actual counters should NOT be
set. The kernel follows the SDM and assumes that all PEBS events are
processed in the drain_pebs() callback. The kernel then checks for
remaining overflows on any other (non-PEBS) events and processes these
in the for_each_bit_set(&status) loop.
As it turns out, under certain conditions on HSW and later processors,
on PEBS buffer interrupt, bit 62 is set but the counter bits may be
set as well. In that case, the kernel drains PEBS and generates
SAMPLES with the EXACT tag, then it processes the counter bits, and
generates normal (non-EXACT) SAMPLES.
I ran into this problem by trying to understand why on HSW sampling on
a PEBS event was sometimes returning SAMPLES without the EXACT tag.
This should not happen on user level code because HSW has the
eventing_ip which always point to the instruction that caused the
event.
The workaround in this patch simply ensures that the bits for the
counters used for PEBS events are cleared after the PEBS buffer has
been drained. With this fix 100% of the PEBS samples on my user code
report the EXACT tag.
Before:
$ perf record -e cpu/event=0xd0,umask=0x81/upp ./multichase
$ perf report -D | fgrep SAMPLES
PERF_RECORD_SAMPLE(IP, 0x2): 11775/11775: 0x406de5 period: 73469 addr: 0 exact=Y
\--- EXACT tag is missing
After:
$ perf record -e cpu/event=0xd0,umask=0x81/upp ./multichase
$ perf report -D | fgrep SAMPLES
PERF_RECORD_SAMPLE(IP, 0x4002): 11775/11775: 0x406de5 period: 73469 addr: 0 exact=Y
\--- EXACT tag is set
The problem tends to appear more often when multiple PEBS events are used.
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: adrian.hunter@intel.com
Cc: kan.liang@intel.com
Cc: namhyung@kernel.org
Link: http://lkml.kernel.org/r/1457034642-21837-3-git-send-email-eranian@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 551adc6057 upstream.
Harry reported, that he's able to trigger a system freeze with cpu hot
unplug. The freeze turned out to be a live lock caused by recent changes in
irq_force_complete_move().
When fixup_irqs() and from there irq_force_complete_move() is called on the
dying cpu, then all other cpus are in stop machine an wait for the dying cpu
to complete the teardown. If there is a move of an interrupt pending then
irq_force_complete_move() sends the cleanup IPI to the cpus in the old_domain
mask and waits for them to clear the mask. That's obviously impossible as
those cpus are firmly stuck in stop machine with interrupts disabled.
I should have known that, but I completely overlooked it being concentrated on
the locking issues around the vectors. And the existance of the call to
__irq_complete_move() in the code, which actually sends the cleanup IPI made
it reasonable to wait for that cleanup to complete. That call was bogus even
before the recent changes as it was just a pointless distraction.
We have to look at two cases:
1) The move_in_progress flag of the interrupt is set
This means the ioapic has been updated with the new vector, but it has not
fired yet. In theory there is a race:
set_ioapic(new_vector) <-- Interrupt is raised before update is effective,
i.e. it's raised on the old vector.
So if the target cpu cannot handle that interrupt before the old vector is
cleaned up, we get a spurious interrupt and in the worst case the ioapic
irq line becomes stale, but my experiments so far have only resulted in
spurious interrupts.
But in case of cpu hotplug this should be a non issue because if the
affinity update happens right before all cpus rendevouz in stop machine,
there is no way that the interrupt can be blocked on the target cpu because
all cpus loops first with interrupts enabled in stop machine, so the old
vector is not yet cleaned up when the interrupt fires.
So the only way to run into this issue is if the delivery of the interrupt
on the apic/system bus would be delayed beyond the point where the target
cpu disables interrupts in stop machine. I doubt that it can happen, but at
least there is a theroretical chance. Virtualization might be able to
expose this, but AFAICT the IOAPIC emulation is not as stupid as the real
hardware.
I've spent quite some time over the weekend to enforce that situation,
though I was not able to trigger the delayed case.
2) The move_in_progress flag is not set and the old_domain cpu mask is not
empty.
That means, that an interrupt was delivered after the change and the
cleanup IPI has been sent to the cpus in old_domain, but not all CPUs have
responded to it yet.
In both cases we can assume that the next interrupt will arrive on the new
vector, so we can cleanup the old vectors on the cpus in the old_domain cpu
mask.
Fixes: 98229aa36c "x86/irq: Plug vector cleanup race"
Reported-by: Harry Junior <harryjr@outlook.fr>
Tested-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Joe Lawrence <joe.lawrence@stratus.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Ben Hutchings <ben@decadent.org.uk>
Link: http://lkml.kernel.org/r/alpine.DEB.2.11.1603140931430.3657@nanos
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 92f9e179a7 upstream.
Pause/unpause graph tracing around do_suspend_lowlevel as it has
inconsistent call/return info after it jumps to the wakeup vector.
The graph trace buffer will otherwise become misaligned and
may eventually crash and hang on suspend.
To reproduce the issue and test the fix:
Run a function_graph trace over suspend/resume and set the graph
function to suspend_devices_and_enter. This consistently hangs the
system without this fix.
Signed-off-by: Todd Brandt <todd.e.brandt@linux.intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 98229aa36c upstream.
We still can end up with a stale vector due to the following:
CPU0 CPU1 CPU2
lock_vector()
data->move_in_progress=0
sendIPI()
unlock_vector()
set_affinity()
assign_irq_vector()
lock_vector() handle_IPI
move_in_progress = 1 lock_vector()
unlock_vector()
move_in_progress == 1
So we need to serialize the vector assignment against a pending cleanup. The
solution is rather simple now. We not only check for the move_in_progress flag
in assign_irq_vector(), we also check whether there is still a cleanup pending
in the old_domain cpumask. If so, we return -EBUSY to the caller and let him
deal with it. Though we have to be careful in the cpu unplug case. If the
cleanout has not yet completed then the following setaffinity() call would
return -EBUSY. Add code which prevents this.
Full context is here: http://lkml.kernel.org/r/5653B688.4050809@stratus.com
Reported-and-tested-by: Joe Lawrence <joe.lawrence@stratus.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Borislav Petkov <bp@alien8.de>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Jeremiah Mahler <jmmahler@gmail.com>
Cc: andy.shevchenko@gmail.com
Cc: Guenter Roeck <linux@roeck-us.net>
Link: http://lkml.kernel.org/r/20151231160107.207265407@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit c1684f5035 upstream.
send_cleanup_vector() fiddles with the old_domain mask unprotected because it
relies on the protection by the move_in_progress flag. But this is fatal, as
the flag is reset after the IPI has been sent. So a cpu which receives the IPI
can still see the flag set and therefor ignores the cleanup request. If no
other cleanup request happens then the vector stays stale on that cpu and in
case of an irq removal the vector still persists. That can lead to use after
free when the next cleanup IPI happens.
Protect the code with vector_lock and clear move_in_progress before sending
the IPI.
This does not plug the race which Joe reported because:
CPU0 CPU1 CPU2
lock_vector()
data->move_in_progress=0
sendIPI()
unlock_vector()
set_affinity()
assign_irq_vector()
lock_vector() handle_IPI
move_in_progress = 1 lock_vector()
unlock_vector()
move_in_progress == 1
The full fix comes with a later patch.
Reported-and-tested-by: Joe Lawrence <joe.lawrence@stratus.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Borislav Petkov <bp@alien8.de>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Jeremiah Mahler <jmmahler@gmail.com>
Cc: andy.shevchenko@gmail.com
Cc: Guenter Roeck <linux@roeck-us.net>
Link: http://lkml.kernel.org/r/20151231160106.892412198@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 3716fd27a6 upstream.
__assign_irq_vector() uses the vector_cpumask which is assigned by
apic->vector_allocation_domain() without doing basic sanity checks. That can
result in a situation where the final assignement of a newly found vector
fails in apic->cpu_mask_to_apicid_and(). So we have to do rollbacks for no
reason.
apic->cpu_mask_to_apicid_and() only fails if
vector_cpumask & requested_cpumask & cpu_online_mask
is empty.
Check for this condition right away and if the result is empty try immediately
the next possible cpu in the requested mask. So in case of a failure the old
setting is unchanged and we can remove the rollback code.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Borislav Petkov <bp@alien8.de>
Tested-by: Joe Lawrence <joe.lawrence@stratus.com>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Jeremiah Mahler <jmmahler@gmail.com>
Cc: andy.shevchenko@gmail.com
Cc: Guenter Roeck <linux@roeck-us.net>
Link: http://lkml.kernel.org/r/20151231160106.561877324@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
commit 111abeba67 upstream.
There's a race condition between
x86_vector_free_irqs()
{
free_apic_chip_data(irq_data->chip_data);
xxxxx //irq_data->chip_data has been freed, but the pointer
//hasn't been reset yet
irq_domain_reset_irq_data(irq_data);
}
and
smp_irq_move_cleanup_interrupt()
{
raw_spin_lock(&vector_lock);
data = apic_chip_data(irq_desc_get_irq_data(desc));
access data->xxxx // may access freed memory
raw_spin_unlock(&desc->lock);
}
which may cause smp_irq_move_cleanup_interrupt() to access freed memory.
Call irq_domain_reset_irq_data(), which clears the pointer with vector lock
held.
[ tglx: Free memory outside of lock held region. ]
Signed-off-by: Jiang Liu <jiang.liu@linux.intel.com>
Tested-by: Borislav Petkov <bp@alien8.de>
Tested-by: Joe Lawrence <joe.lawrence@stratus.com>
Cc: Jeremiah Mahler <jmmahler@gmail.com>
Cc: andy.shevchenko@gmail.com
Cc: Guenter Roeck <linux@roeck-us.net>
Link: http://lkml.kernel.org/r/1450880014-11741-3-git-send-email-jiang.liu@linux.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Adding the rtc platform device in non-privileged Xen PV guests causes
an IRQ conflict because these guests do not have legacy PIC and may
allocate irqs in the legacy range.
In a single VCPU Xen PV guest we should have:
/proc/interrupts:
CPU0
0: 4934 xen-percpu-virq timer0
1: 0 xen-percpu-ipi spinlock0
2: 0 xen-percpu-ipi resched0
3: 0 xen-percpu-ipi callfunc0
4: 0 xen-percpu-virq debug0
5: 0 xen-percpu-ipi callfuncsingle0
6: 0 xen-percpu-ipi irqwork0
7: 321 xen-dyn-event xenbus
8: 90 xen-dyn-event hvc_console
...
But hvc_console cannot get its interrupt because it is already in use
by rtc0 and the console does not work.
genirq: Flags mismatch irq 8. 00000000 (hvc_console) vs. 00000000 (rtc0)
We can avoid this problem by realizing that unprivileged PV guests (both
Xen and lguests) are not supposed to have rtc_cmos device and so
adding it is not necessary.
Privileged guests (i.e. Xen's dom0) do use it but they should not have
irq conflicts since they allocate irqs above legacy range (above
gsi_top, in fact).
Instead of explicitly testing whether the guest is privileged we can
extend pv_info structure to include information about guest's RTC
support.
Reported-and-tested-by: Sander Eikelenboom <linux@eikelenboom.it>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: vkuznets@redhat.com
Cc: xen-devel@lists.xenproject.org
Cc: konrad.wilk@oracle.com
Cc: stable@vger.kernel.org # 4.2+
Link: http://lkml.kernel.org/r/1449842873-2613-1-git-send-email-boris.ostrovsky@oracle.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Wanpeng Li