As we will not have a cpu_x86_find_by_name() function anymore,
move the KVM default-vendor hack to instance_init.
Unfortunately we can't move that code to class_init because it depends
on KVM being initialized.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
As eventually the x86_def_t data is going to be provided by the CPU
class, it's better to not touch it, and handle the special cases on the
X86CPU object itself.
Current behavior of the code should stay exactly the same.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The compiler is already able to eliminate the kvm_arch_get_supported_cpuid()
calls in kvm_cpu_fill_host() and filter_features_for_kvm(), so we can
eliminate the CONFIG_KVM #ifdefs there.
Also, kvm_cpu_fill_host() and host_cpuid() don't need to check
CONFIG_KVM, as they don't have any KVM-specific function calls.
Tested to build successfully with CONFIG_KVM disabled, using the
following CFLAGS combinations: "-DNDEBUG", "-DNDEBUG -O', "-DNDEBUG
-O0", "-DNDEBUG -O1", "-DNDEBUG -O2".
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
http://msdn.microsoft.com/en-us/library/windows/hardware/ff541625%28v=vs.85%29.aspx
This code is generic for activating reference time counter or virtual reference time stamp counter
Signed-off-by: Vadim Rozenfeld <vrozenfe@redhat.com>
Reviewed-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The MS docs specify HV_X64_MSR_HYPERCALL as a mandatory interface,
thus we must provide the MSRs even if the user only specified
features that, like relaxed timing, in principle don't require them.
And the MSRs are only there if the hypervisor has KVM_CAP_HYPERV.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
kvm_arch_init_vcpu's initialization of the KVM leaves at 0x40000100
is broken, because KVM_CPUID_FEATURES is left at 0x40000001. Move
it to 0x40000101 if Hyper-V is enabled.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
We don't need the ft[] array on kvm_check_features_against_host()
anymore, as we can simply use the feature_word_info[] array, that has
everything we need.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that the kvm_cpu_fill_host() code is simplified, we can simply set
the feature word array using a simple loop.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reorder the code so all the code that sets x86_cpu_def->features is at
the end of the function.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
There's no need to check CPU xlevel2 before calling
kvm_arch_get_supported_cpuid(s, 0xC0000001, 0, R_EDX), because:
* The kernel won't return any entry for 0xC0000000 if host CPU vendor
is not Centaur (See kvm_dev_ioctl_get_supported_cpuid() on the kernel
code)
* Similarly, the kernel won't return any entry for 0xC0000001 if
CPUID[0xC0000000].EAX is < 0xC0000001
* kvm_arch_get_supported_cpuid() will return 0 if no entry is returned
by the kernel for the requested leaf
For similar reasons, we can simply set x86_cpu_def->xlevel2 directly
instead of making it conditional, because it will be set to 0 CPU vendor
is not Centaur.
This will simplify the kvm_cpu_fill_host() code a little.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
[Remove unparseable comment. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
There's no need to check CPU vendor before calling
kvm_arch_get_supported_cpuid(s, 0xC0000000, 0, R_EAX), because:
* The kernel won't return any entry for 0xC0000000 if host CPU vendor
is not Centaur (See kvm_dev_ioctl_get_cpuid() on the kernel code);
* kvm_arch_get_supported_cpuid() will return 0 if no entry is returned
by the kernel for the requested leaf.
This will simplify the kvm_cpu_fill_host() code a little.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
There's no need to check level (CPUID[0].EAX) before calling
kvm_arch_get_supported_cpuid(s, 0x7, 0, R_EBX), because:
* The kernel won't return any entry for CPUID 7 if CPUID[0].EAX is < 7
on the host (See kvm_dev_ioctl_get_cpuid() on the kernel code);
* kvm_arch_get_supported_cpuid() will return 0 if no entry is returned
by the kernel for the requested leaf.
This will simplify the kvm_cpu_fill_host() code a little.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Those host_cpuid() calls are useless. They are leftovers from when the
old code using host_cpuid() was removed.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
* qemu-kvm/uq/master:
kvm: always update the MPX model specific register
KVM: fix addr type for KVM_IOEVENTFD
KVM: Retry KVM_CREATE_VM on EINTR
mempath prefault: fix off-by-one error
kvm: x86: Separately write feature control MSR on reset
roms: Flush icache when writing roms to guest memory
target-i386: clear guest TSC on reset
target-i386: do not special case TSC writeback
target-i386: Intel MPX
Conflicts:
exec.c
aliguori: fix trivial merge conflict in exec.c
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
The original patch from Liu Jinsong restricted them to reset or full
state updates, but that's unnecessary (and wrong) since the BNDCFGS
MSR has no side effects.
Cc: Liu Jinsong <jinsong.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reduce ifdefs, share more code between paths, reduce the number of TCG
ops generated. Avoid re-computing the size of the operation across
gen_pop_T0 and gen_pop_update.
Add forgotten zero-extension in the TARGET_X86_64, !CODE64, ss32 case.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Reduce ifdefs, share more code between paths, reduce the number of TCG
ops generated.
Add forgotten zero-extension in the TARGET_X86_64, !CODE64, ss32 case.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Unlike the addr32, there was no bug. But we can use the same
technique to reduce the number of TCG ops.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Changing the domain to TCGMemOp makes it easier to interoperate
with other portions of the rest of the translator.
We now only have one domain for size operands inside the translator,
which makes things less confusing all the way around. There are
still a number of helpers that continue to use the log2-1 domain.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Change the domain of the parameter and update all callers.
Which lets us defer completely to gen_op_mov_reg_v.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Changing the domain to TCGMemOp makes it easier to interoperate
with other portions of the rest of the translator.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Change the domain of the parameter and update all callers.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
These functions used the aflags/dflags domain, which is log2-1
of the byte size. Confusingly, they used enumeration values
from the log2 domain.
Change the domain of the parameter and update all callers.
Since we're now in a common domain, defer the deposit/extend/mov
decision to gen_op_mov_reg_v.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
The 'ot' variables (operand type?) hold the log2(byte size) of
the operand being manipulated. This is the same as the MO_SIZE
subset of the TCGMemOp. Indeed, we often pass 'ot' to the
tcg_gen_qemu_ld/st functions.
Changing the type from 'int' makes it easier to see what domain
the variable should be.
This does require adding some default cases to some switch statements,
to avoid the 'unhandled enumeration value' warning that would result
from the change of type.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Replace it with tcg_gen_ext16u_tl, and in two cases merge with a
previous move from cpu_regs.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Replace it with tcg_gen_ext16u_tl. In four places we can combine that
with a previous move into cpu_T[0], and in one place we can infer that
the zero-extension has already happened via the previous load.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Propagate the definitions into all users. In two cases, this allows
us to share code between the 32-bit and 64-bit immediate moves.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Propagate the definitions into all users. The only time that
gen_op_movl_T1_imu was used, the input was type 'unsigned',
so the replacement works identically.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Propagate the definition of gen_op_movl_T0_im to all users.
The function gen_op_movl_T0_imu was unused.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
For the known MO_32/MO_64 cases, we don't need to extend a 32-bit temp
into a 64-bit temp before storing into the hardware register.
We do need the extension for the MO_8/MO_16 cases, in order for the
deposit_tl operation to work, so leave those alone.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
We can now use tcg_gen_qemu_st_i32 directly to avoid the extension.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
We can now use tcg_gen_qemu_ld_i32 directly to avoid the truncation.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
For the 16 and 32-bit cases, we don't need to truncate via
a temporary register.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
The reg_ptr and offset_ptr outputs are universally unused.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Always perform a sign-extending load. In the extremely unlikely
case that we've used an 0x66 prefix, the extension to 64-bits is
unnecessary but not wrong; the store will still examine only 16 bits.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
We can use the MO_SIGN bit to tidy the reg-reg switch statement
as well as pass it on to gen_op_ld_v, eliminating one call.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
By inspection, obviously we should be storing T[1] not T[0].
This could only happen for x86_64 in 64-bit mode with 0x66
prefix to call insn -- i.e. never.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Too many places have the same test vs OR_TMP0 to indicate
a write back to memory. Hoist that to a subroutine.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Replace its users by gen_op_ld_v with the MO_SIGN bit set.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
The MO_8/16/32/64 constants have the same encoding and meaning
as the OT_BYTE/WORD/LONG/QUAD. Since we rely on them being the
same, for the qemu_ld/st helpers, standardize on the common names.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
In preference to the older helpers. Stores only in this patch.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
In preference to the older helpers. Loads only in this patch.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Now that we don't combine mem_index with operand size info,
we don't need to encode it. Which tidies many places that
access it.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Rather than add s->mem_index into a combined size+mem_index
argument, pass the context down. This will allow cleaning
up s->mem_index later.
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <rth@twiddle.net>
Replace an assert_no_error() usage with the error_abort system.
Signed-off-by: Peter Crosthwaite <peter.crosthwaite@xilinx.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Features family, model, stepping, level, hv_spinlocks are treated similarly
when passed from command line, so it's not necessary to handle each of them
individually. Collapse them to one catch-all branch which will treat
any not explicitly handled feature in format 'foo=val'.
Any unknown feature will be rejected by property setter so there is no
need to check for unknown feature in cpu_x86_parse_featurestr(), therefore
it's replaced by above mentioned catch-all handler.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
Features check, enforce, hv_relaxed and hv_vapic are treated as boolean
set to 'on' when passed from command line, so it's not necessary to
handle each of them separately. Collapse them to one catch-all branch
which will treat any feature in format 'foo' as boolean set to 'on'.
Any unknown feature will be rejected by CPU property setter so there is no
need to check for unknown feature in cpu_x86_parse_featurestr(), therefore
it's replaced by above mentioned catch-all handler.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
* Additionally convert check_cpuid & enforce_cpuid to bool and make them
members of X86CPU
* Make 'enforce' feature independent from 'check'
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
This motion is preparing for refactoring vCPU APIC subsequently.
Signed-off-by: Chen Fan <chen.fan.fnst@cn.fujitsu.com>
Signed-off-by: Andreas Färber <afaerber@suse.de>
When we're running in non-64bit mode with qemu-system-x86_64 we can
still end up with virtual addresses that are above the 32bit boundary
if a segment offset is set up.
GNU Hurd does exactly that. It sets the segment offset to 0x80000000 and
puts its EIP value to 0x8xxxxxxx to access low memory.
This doesn't hit us when we enable paging, as there we just mask away the
unused bits. But with real mode, we assume that vaddr == paddr which is
wrong in this case. Real hardware wraps the virtual address around at the
32bit boundary. So let's do the same.
This fixes booting GNU Hurd in qemu-system-x86_64 for me.
Reported-by: Michael Tokarev <mjt@tls.msk.ru>
Signed-off-by: Alexander Graf <agraf@suse.de>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
If the guest is running in nested mode on system reset, clearing the
feature MSR signals the kernel to leave this mode. Recent kernels
processes this properly, but leave the VCPU state undefined behind. It
is the job of userspace to bring it to a proper shape. Therefore, write
this specific MSR first so that no state transfer gets lost.
This allows to cleanly reset a guest with VMX in use.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
VCPU TSC is not cleared by a warm reset (*), which leaves some types of Linux
guests (non-pvops guests and those with the kernel parameter no-kvmclock set)
vulnerable to the overflow in cyc2ns_offset fixed by upstream commit
9993bc635d01a6ee7f6b833b4ee65ce7c06350b1 ("sched/x86: Fix overflow in
cyc2ns_offset").
To put it in a nutshell, if such a Linux guest without the patch above applied
has been up more than 208 days and attempts a warm reset chances are that
the newly booted kernel will panic or hang.
(*) Intel Xeon E5 processors show the same broken behavior due to
the errata "TSC is Not Affected by Warm Reset" (Intel® Xeon®
Processor E5 Family Specification Update - August 2013): "The
TSC (Time Stamp Counter MSR 10H) should be cleared on
reset. Due to this erratum the TSC is not affected by warm
reset."
Cc: Will Auld <will.auld@intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fernando Luis Vázquez Cao <fernando_b1@lab.ntt.co.jp>
Newer kernels are capable of synchronizing TSC values of multiple VCPUs
on writeback, but we were excluding the power up case, which is not needed
anymore.
Signed-off-by: Fernando Luis Vazquez Cao <fernando@oss.ntt.co.jp>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fernando Luis Vázquez Cao <fernando_b1@lab.ntt.co.jp>
Add some MPX related definiation, and hardcode sizes and offsets
of xsave features 3 and 4. It also add corresponding part to
kvm_get/put_xsave, and vmstate.
Signed-off-by: Liu Jinsong <jinsong.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
# By Alex Williamson (1) and others
# Via Paolo Bonzini
* qemu-kvm/uq/master:
target-i386: fix cpuid leaf 0x0d
qemu: mempath: prefault pages manually (v4)
kvm: Query KVM for available memory slots
Message-id: 1386345276-9803-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
Fix cpuid leaf 0x0d which incorrectly parsed eax and ebx.
However, before this patch the CPUID worked fine -- the .offset
field contained the size _and_ was stored in the register that
is supposed to hold the size (eax), and likewise the .size field
contained the offset _and_ was stored in the register trhat is
supposed to hold the offset (ebx).
Signed-off-by: Liu Jinsong <jinsong.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
were getting forgotten or that did not have a clear maintainer responsible
for making a pull request.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)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=c4Hz
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'bonzini/tags/for-anthony' into staging
Here are a bunch of 1.7-tagged patches that I was afraid
were getting forgotten or that did not have a clear maintainer responsible
for making a pull request.
# gpg: Signature made Thu 21 Nov 2013 08:40:59 AM PST using RSA key ID 9B4D86F2
# gpg: Can't check signature: public key not found
# By Peter Maydell (3) and others
# Via Paolo Bonzini
* bonzini/tags/for-anthony:
qga: Fix compiler warnings (missing format attribute, wrong format strings)
mips jazz: do not raise data bus exception when accessing invalid addresses
target-i386: yield to another VCPU on PAUSE
rng-egd: offset the point when repeatedly read from the buffer
rng-egd: remove redundant free
target-i386: Fix build by providing stub kvm_arch_get_supported_cpuid()
vfio-pci: Fix multifunction=on
atomic.h: Fix build with clang
pc: get rid of builtin pvpanic for "-M pc-1.5"
configure: Explicitly set ARFLAGS so we can build with GNU Make 4.0
sun4m: Add FCode ROM for TCX framebuffer
Message-id: 1385052578-32352-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
After commit b1bbfe7 (aio / timers: On timer modification, qemu_notify
or aio_notify, 2013-08-21) FreeBSD guests report a huge slowdown.
The problem shows up as soon as FreeBSD turns out its periodic (~1 ms)
tick, but the timers are only the trigger for a pre-existing problem.
Before the offending patch, setting a timer did a timer_settime system call.
After, setting the timer exits the event loop (which uses poll) and
reenters it with a new deadline. This does not cause any slowdown; the
difference is between one system call (timer_settime and a signal
delivery (SIGALRM) before the patch, and two system calls afterwards
(write to a pipe or eventfd + calling poll again when re-entering the
event loop).
Unfortunately, the exit/enter causes the main loop to grab the iothread
lock, which in turns kicks the VCPU thread out of execution. This
causes TCG to execute the next VCPU in its round-robin scheduling of
VCPUS. When the second VCPU is mostly unused, FreeBSD runs a "pause"
instruction in its idle loop which only burns cycles without any
progress. As soon as the timer tick expires, the first VCPU runs
the interrupt handler but very soon it sets it again---and QEMU
then goes back doing nothing in the second VCPU.
The fix is to make the pause instruction do "cpu_loop_exit".
Reported-by: Luigi Rizzo <rizzo@iet.unipi.it>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Fix build failures with clang when KVM is not enabled by
providing a stub version of kvm_arch_get_supported_cpuid().
We retain the compile time check that this function isn't
called when CONFIG_KVM is not set by guarding the stub with
ifndef __OPTIMIZE__ (we assume that an optimizing build will
do sufficient constant folding and dead code elimination to
remove the calls before linking).
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Commit 787aaf5 (target-i386: forward CPUID cache leaves when -cpu host is
used, 2013-09-02) brings bits 31..26 of CPUID leaf 04h out of sync with
the APIC IDs that QEMU reserves for each package. This number must come
from "-smp" options rather than from the host CPUID.
It also turns out that this unsyncing makes Windows Server 2012R2 fail
to boot.
Tested-by: Peter Lieven <pl@kamp.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Benoit Canet <benoit@irqsave.net>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Message-id: 1384879786-6721-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
After commit b1bbfe7 (aio / timers: On timer modification, qemu_notify
or aio_notify, 2013-08-21) FreeBSD guests report a huge slowdown.
The problem shows up as soon as FreeBSD turns out its periodic (~1 ms)
tick, but the timers are only the trigger for a pre-existing problem.
Before the offending patch, setting a timer did a timer_settime system call.
After, setting the timer exits the event loop (which uses poll) and
reenters it with a new deadline. This does not cause any slowdown; the
difference is between one system call (timer_settime and a signal
delivery (SIGALRM) before the patch, and two system calls afterwards
(write to a pipe or eventfd + calling poll again when re-entering the
event loop).
Unfortunately, the exit/enter causes the main loop to grab the iothread
lock, which in turns kicks the VCPU thread out of execution. This
causes TCG to execute the next VCPU in its round-robin scheduling of
VCPUS. When the second VCPU is mostly unused, FreeBSD runs a "pause"
instruction in its idle loop which only burns cycles without any
progress. As soon as the timer tick expires, the first VCPU runs
the interrupt handler but very soon it sets it again---and QEMU
then goes back doing nothing in the second VCPU.
The fix is to make the pause instruction do "cpu_loop_exit".
Cc: Richard Henderson <rth@twiddle.net>
Reported-by: Luigi Rizzo <rizzo@iet.unipi.it>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Message-id: 1384948442-24217-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
This error was reported by valgrind when running qemu-system-x86_64
with kvm:
==975== Conditional jump or move depends on uninitialised value(s)
==975== at 0x521C38: cpuid_find_entry (kvm.c:176)
==975== by 0x5235BA: kvm_arch_init_vcpu (kvm.c:686)
==975== by 0x4D5175: kvm_init_vcpu (kvm-all.c:267)
==975== by 0x45035B: qemu_kvm_cpu_thread_fn (cpus.c:858)
==975== by 0xD361E0D: start_thread (pthread_create.c:311)
==975== by 0xD65E9EC: clone (clone.S:113)
==975== Uninitialised value was created by a stack allocation
==975== at 0x5226E4: kvm_arch_init_vcpu (kvm.c:446)
Instead of adding more memset calls for parts of cpuid_data, the existing
calls were removed and cpuid_data is now initialized completely in one
call.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Only the first item of the array was ever looked at. No
practical effect, but still worth fixing.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
# By Paolo Bonzini (2) and Jan Kiszka (1)
# Via Gleb Natapov
* qemu-kvm/uq/master:
kvmvapic: Prevent reading beyond the end of guest RAM
x86: cpuid: reconstruct leaf 0Dh data
x86: fix migration from pre-version 12
Message-id: 1382108641-4862-1-git-send-email-pbonzini@redhat.com
Signed-off-by: Anthony Liguori <aliguori@amazon.com>
