Per the discussion about FRED MSR writes with WRMSRNS instruction [1],
use the alternatives mechanism to choose WRMSRNS when it's available,
otherwise fallback to WRMSR.
Remove the dependency on X86_FEATURE_WRMSRNS as WRMSRNS is no longer
dependent on FRED.
[1] https://lore.kernel.org/lkml/15f56e6a-6edd-43d0-8e83-bb6430096514@citrix.com/
Use DS prefix to pad WRMSR instead of a NOP. The prefix is ignored. At
least that's the current information from the hardware folks.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Xin Li (Intel) <xin@zytor.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240822073906.2176342-3-xin@zytor.com
with 32-bit guests, seeing stale instruction bytes, to one working on
a buffer, like the rest of the alternatives code does
- Add a long overdue check to the X86_FEATURE flag modifying functions to warn
when former get changed in a non-compatible way after alternatives have been
patched because those changes will be already wrong
- Other cleanups
-----BEGIN PGP SIGNATURE-----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=gaPh
-----END PGP SIGNATURE-----
Merge tag 'x86_alternatives_for_v6.10_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 asm alternatives updates from Borislav Petkov:
- Switch the in-place instruction patching which lead to at least one
weird bug with 32-bit guests, seeing stale instruction bytes, to one
working on a buffer, like the rest of the alternatives code does
- Add a long overdue check to the X86_FEATURE flag modifying functions
to warn when former get changed in a non-compatible way after
alternatives have been patched because those changes will be already
wrong
- Other cleanups
* tag 'x86_alternatives_for_v6.10_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/alternatives: Remove alternative_input_2()
x86/alternatives: Sort local vars in apply_alternatives()
x86/alternatives: Optimize optimize_nops()
x86/alternatives: Get rid of __optimize_nops()
x86/alternatives: Use a temporary buffer when optimizing NOPs
x86/alternatives: Catch late X86_FEATURE modifiers
Fix cpuid_deps[] to list the correct dependencies for GFNI, VAES, and
VPCLMULQDQ. These features don't depend on AVX512, and there exist CPUs
that support these features but not AVX512. GFNI actually doesn't even
depend on AVX.
This prevents GFNI from being unnecessarily disabled if AVX is disabled
to mitigate the GDS vulnerability.
This also prevents all three features from being unnecessarily disabled
if AVX512VL (or its dependency AVX512F) were to be disabled, but it
looks like there isn't any case where this happens anyway.
Fixes: c128dbfa0f ("x86/cpufeatures: Enable new SSE/AVX/AVX512 CPU features")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Acked-by: Dave Hansen <dave.hansen@linux.intel.com>
Link: https://lore.kernel.org/r/20240417060434.47101-1-ebiggers@kernel.org
After alternatives have been patched, changes to the X86_FEATURE flags
won't take effect and could potentially even be wrong.
Warn about it.
This is something which has been long overdue.
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Srikanth Aithal <sraithal@amd.com>
Link: https://lore.kernel.org/r/20240327154317.29909-3-bp@alien8.de
Any FRED enabled CPU will always have the following features as its
baseline:
1) LKGS, load attributes of the GS segment but the base address into
the IA32_KERNEL_GS_BASE MSR instead of the GS segment’s descriptor
cache.
2) WRMSRNS, non-serializing WRMSR for faster MSR writes.
Signed-off-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Signed-off-by: Xin Li <xin3.li@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Shan Kang <shan.kang@intel.com>
Link: https://lore.kernel.org/r/20231205105030.8698-7-xin3.li@intel.com
The Control-Flow Enforcement Technology contains two related features,
one of which is Shadow Stacks. Future patches will utilize this feature
for shadow stack support in KVM, so add a CPU feature flags for Shadow
Stacks (CPUID.(EAX=7,ECX=0):ECX[bit 7]).
To protect shadow stack state from malicious modification, the registers
are only accessible in supervisor mode. This implementation
context-switches the registers with XSAVES. Make X86_FEATURE_SHSTK depend
on XSAVES.
The shadow stack feature, enumerated by the CPUID bit described above,
encompasses both supervisor and userspace support for shadow stack. In
near future patches, only userspace shadow stack will be enabled. In
expectation of future supervisor shadow stack support, create a software
CPU capability to enumerate kernel utilization of userspace shadow stack
support. This user shadow stack bit should depend on the HW "shstk"
capability and that logic will be implemented in future patches.
Co-developed-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
Signed-off-by: Yu-cheng Yu <yu-cheng.yu@intel.com>
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Reviewed-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Mike Rapoport (IBM) <rppt@kernel.org>
Tested-by: Pengfei Xu <pengfei.xu@intel.com>
Tested-by: John Allen <john.allen@amd.com>
Tested-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/all/20230613001108.3040476-9-rick.p.edgecombe%40intel.com
Newer AMD processors support the new feature Bandwidth Monitoring Event
Configuration (BMEC).
The feature support is identified via CPUID Fn8000_0020_EBX_x0[3]: EVT_CFG -
Bandwidth Monitoring Event Configuration (BMEC)
The bandwidth monitoring events mbm_total_bytes and mbm_local_bytes are set to
count all the total and local reads/writes, respectively. With the introduction
of slow memory, the two counters are not enough to count all the different types
of memory events. Therefore, BMEC provides the option to configure
mbm_total_bytes and mbm_local_bytes to count the specific type of events.
Each BMEC event has a configuration MSR which contains one field for each
bandwidth type that can be used to configure the bandwidth event to track any
combination of supported bandwidth types. The event will count requests from
every bandwidth type bit that is set in the corresponding configuration
register.
Following are the types of events supported:
==== ========================================================
Bits Description
==== ========================================================
6 Dirty Victims from the QOS domain to all types of memory
5 Reads to slow memory in the non-local NUMA domain
4 Reads to slow memory in the local NUMA domain
3 Non-temporal writes to non-local NUMA domain
2 Non-temporal writes to local NUMA domain
1 Reads to memory in the non-local NUMA domain
0 Reads to memory in the local NUMA domain
==== ========================================================
By default, the mbm_total_bytes configuration is set to 0x7F to count
all the event types and the mbm_local_bytes configuration is set to 0x15 to
count all the local memory events.
Feature description is available in the specification, "AMD64 Technology
Platform Quality of Service Extensions, Revision: 1.03 Publication" at
https://bugzilla.kernel.org/attachment.cgi?id=301365
Signed-off-by: Babu Moger <babu.moger@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Reinette Chatre <reinette.chatre@intel.com>
Link: https://lore.kernel.org/r/20230113152039.770054-5-babu.moger@amd.com
The new Asynchronous Exit (AEX) notification mechanism (AEX-notify)
allows one enclave to receive a notification in the ERESUME after the
enclave exit due to an AEX. EDECCSSA is a new SGX user leaf function
(ENCLU[EDECCSSA]) to facilitate the AEX notification handling. The new
EDECCSSA is enumerated via CPUID(EAX=0x12,ECX=0x0):EAX[11].
Besides Allowing reporting the new AEX-notify attribute to KVM guests,
also allow reporting the new EDECCSSA user leaf function to KVM guests
so the guest can fully utilize the AEX-notify mechanism.
Similar to existing X86_FEATURE_SGX1 and X86_FEATURE_SGX2, introduce a
new scattered X86_FEATURE_SGX_EDECCSSA bit for the new EDECCSSA, and
report it in KVM's supported CPUIDs.
Note, no additional KVM enabling is required to allow the guest to use
EDECCSSA. It's impossible to trap ENCLU (without completely preventing
the guest from using SGX). Advertise EDECCSSA as supported purely so
that userspace doesn't need to special case EDECCSSA, i.e. doesn't need
to manually check host CPUID.
The inability to trap ENCLU also means that KVM can't prevent the guest
from using EDECCSSA, but that virtualization hole is benign as far as
KVM is concerned. EDECCSSA is simply a fancy way to modify internal
enclave state.
More background about how do AEX-notify and EDECCSSA work:
SGX maintains a Current State Save Area Frame (CSSA) for each enclave
thread. When AEX happens, the enclave thread context is saved to the
CSSA and the CSSA is increased by 1. For a normal ERESUME which doesn't
deliver AEX notification, it restores the saved thread context from the
previously saved SSA and decreases the CSSA. If AEX-notify is enabled
for one enclave, the ERESUME acts differently. Instead of restoring the
saved thread context and decreasing the CSSA, it acts like EENTER which
doesn't decrease the CSSA but establishes a clean slate thread context
using the CSSA for the enclave to handle the notification. After some
handling, the enclave must discard the "new-established" SSA and switch
back to the previously saved SSA (upon AEX). Otherwise, the enclave
will run out of SSA space upon further AEXs and eventually fail to run.
To solve this problem, the new EDECCSSA essentially decreases the CSSA.
It can be used by the enclave notification handler to switch back to the
previous saved SSA when needed, i.e. after it handles the notification.
Signed-off-by: Kai Huang <kai.huang@intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Acked-by: Sean Christopherson <seanjc@google.com>
Acked-by: Jarkko Sakkinen <jarkko@kernel.org>
Link: https://lore.kernel.org/all/20221101022422.858944-1-kai.huang%40intel.com
tl;dr: AMX state is ~8k. Signal frames can have space for this
~8k and each signal entry writes out all 8k even if it is zeros.
Skip writing zeros for AMX to speed up signal delivery by about
4% overall when AMX is in its init state.
This is a user-visible change to the sigframe ABI.
== Hardware XSAVE Background ==
XSAVE state components may be tracked by the processor as being
in their initial configuration. Software can detect which
features are in this configuration by looking at the XSTATE_BV
field in an XSAVE buffer or with the XGETBV(1) instruction.
Both the XSAVE and XSAVEOPT instructions enumerate features s
being in the initial configuration via the XSTATE_BV field in the
XSAVE header, However, XSAVEOPT declines to actually write
features in their initial configuration to the buffer. XSAVE
writes the feature unconditionally, regardless of whether it is
in the initial configuration or not.
Basically, XSAVE users never need to inspect XSTATE_BV to
determine if the feature has been written to the buffer.
XSAVEOPT users *do* need to inspect XSTATE_BV. They might also
need to clear out the buffer if they want to make an isolated
change to the state, like modifying one register.
== Software Signal / XSAVE Background ==
Signal frames have historically been written with XSAVE itself.
Each state is written in its entirety, regardless of being in its
initial configuration.
In other words, the signal frame ABI uses the XSAVE behavior, not
the XSAVEOPT behavior.
== Problem ==
This means that any application which has acquired permission to
use AMX via ARCH_REQ_XCOMP_PERM will write 8k of state to the
signal frame. This 8k write will occur even when AMX was in its
initial configuration and software *knows* this because of
XSTATE_BV.
This problem also exists to a lesser degree with AVX-512 and its
2k of state. However, AVX-512 use does not require
ARCH_REQ_XCOMP_PERM and is more likely to have existing users
which would be impacted by any change in behavior.
== Solution ==
Stop writing out AMX xfeatures which are in their initial state
to the signal frame. This effectively makes the signal frame
XSAVE buffer look as if it were written with a combination of
XSAVEOPT and XSAVE behavior. Userspace which handles XSAVEOPT-
style buffers should be able to handle this naturally.
For now, include only the AMX xfeatures: XTILE and XTILEDATA in
this new behavior. These require new ABI to use anyway, which
makes their users very unlikely to be broken. This XSAVEOPT-like
behavior should be expected for all future dynamic xfeatures. It
may also be extended to legacy features like AVX-512 in the
future.
Only attempt this optimization on systems with dynamic features.
Disable dynamic feature support (XFD) if XGETBV1 is unavailable
by adding a CPUID dependency.
This has been measured to reduce the *overall* cycle cost of
signal delivery by about 4%.
Fixes: 2308ee57d9 ("x86/fpu/amx: Enable the AMX feature in 64-bit mode")
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: "Chang S. Bae" <chang.seok.bae@intel.com>
Link: https://lore.kernel.org/r/20211102224750.FA412E26@davehans-spike.ostc.intel.com
Add the AMX state components in XFEATURE_MASK_USER_SUPPORTED and the
TILE_DATA component to the dynamic states and update the permission check
table accordingly.
This is only effective on 64 bit kernels as for 32bit kernels
XFEATURE_MASK_TILE is defined as 0.
TILE_DATA is caller-saved state and the only dynamic state. Add build time
sanity check to ensure the assumption that every dynamic feature is caller-
saved.
Make AMX state depend on XFD as it is dynamic feature.
Signed-off-by: Chang S. Bae <chang.seok.bae@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lore.kernel.org/r/20211021225527.10184-24-chang.seok.bae@intel.com
Intel's eXtended Feature Disable (XFD) feature is an extension of the XSAVE
architecture. XFD allows the kernel to enable a feature state in XCR0 and
to receive a #NM trap when a task uses instructions accessing that state.
This is going to be used to postpone the allocation of a larger XSTATE
buffer for a task to the point where it is actually using a related
instruction after the permission to use that facility has been granted.
XFD is not used by the kernel, but only applied to userspace. This is a
matter of policy as the kernel knows how a fpstate is reallocated and the
XFD state.
The compacted XSAVE format is adjustable for dynamic features. Make XFD
depend on XSAVES.
Signed-off-by: Chang S. Bae <chang.seok.bae@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Chang S. Bae <chang.seok.bae@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20211021225527.10184-13-chang.seok.bae@intel.com
Add SGX1 and SGX2 feature flags, via CPUID.0x12.0x0.EAX, as scattered
features, since adding a new leaf for only two bits would be wasteful.
As part of virtualizing SGX, KVM will expose the SGX CPUID leafs to its
guest, and to do so correctly needs to query hardware and kernel support
for SGX1 and SGX2.
Suppress both SGX1 and SGX2 from /proc/cpuinfo. SGX1 basically means
SGX, and for SGX2 there is no concrete use case of using it in
/proc/cpuinfo.
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Kai Huang <kai.huang@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Acked-by: Dave Hansen <dave.hansen@intel.com>
Acked-by: Jarkko Sakkinen <jarkko@kernel.org>
Link: https://lkml.kernel.org/r/d787827dbfca6b3210ac3e432e3ac1202727e786.1616136308.git.kai.huang@intel.com
Move SGX_LC feature bit to CPUID dependency table to make clearing all
SGX feature bits easier. Also remove clear_sgx_caps() since it is just
a wrapper of setup_clear_cpu_cap(X86_FEATURE_SGX) now.
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Kai Huang <kai.huang@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Sean Christopherson <seanjc@google.com>
Acked-by: Dave Hansen <dave.hansen@intel.com>
Acked-by: Jarkko Sakkinen <jarkko@kernel.org>
Link: https://lkml.kernel.org/r/5d4220fd0a39f52af024d3fa166231c1d498dd10.1616136308.git.kai.huang@intel.com
Enumerate AVX512 Half-precision floating point (FP16) CPUID feature
flag. Compared with using FP32, using FP16 cut the number of bits
required for storage in half, reducing the exponent from 8 bits to 5,
and the mantissa from 23 bits to 10. Using FP16 also enables developers
to train and run inference on deep learning models fast when all
precision or magnitude (FP32) is not needed.
A processor supports AVX512 FP16 if CPUID.(EAX=7,ECX=0):EDX[bit 23]
is present. The AVX512 FP16 requires AVX512BW feature be implemented
since the instructions for manipulating 32bit masks are associated with
AVX512BW.
The only in-kernel usage of this is kvm passthrough. The CPU feature
flag is shown as "avx512_fp16" in /proc/cpuinfo.
Signed-off-by: Kyung Min Park <kyung.min.park@intel.com>
Acked-by: Dave Hansen <dave.hansen@intel.com>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Message-Id: <20201208033441.28207-2-kyung.min.park@intel.com>
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
James Morse.
* Add support for controlling per-thread memory bandwidth throttling
delay values on hw which supports it, by Fenghua Yu.
-----BEGIN PGP SIGNATURE-----
iQIzBAABCgAdFiEEzv7L6UO9uDPlPSfHEsHwGGHeVUoFAl+ENo0ACgkQEsHwGGHe
VUpIAw/+JtO9mP/OxLUUQEkYGMlYWxiJKGxHdI0cnw6gN02TGakVPZS3RAhdrDPP
Oahfl8g2EiC2sXSo0QEMFfZyEc/eOWo17wL1B+wgPfIIxy6KfGe6WtkHMNlOkWOS
zKxUvR93PjSs7e1vS+AMGbqQVFcL4RTSZN5H/QDaBnkxd3O5uLEvUm4pOxPs9FtX
etnK3eM4Uk6qfH9Pa0XZowp2RU0okRsatu+VREkEBplEplA1tusw3u//SlGgi266
Jsy2Pa2S7D0PGaP2D2+eziNmff319AT1mLtZ/0PKjkeZtqq/Sz0MJ9TxkesyEQPH
iv7IWzp+Dfc8Ui5rDNDvOIY+uJxQPMC0qwpU6sZdAgpsCcI5/xiSqTbBz6mxZeql
vTINIs7Lg/FBfkUn52LxbWkl8QA6aLXYr3PwdcFJzyTYmQitYzdEKxn1i+teWKr2
16QHR2GnXIEfc87JuHJpwiToUYZg+5UlVPkFTLNk/2n0gSiJzWMGecuHdS9spToR
vtpt5vmcAJKUptJLwKId+oEHbMLrvDGjXLApD4x3ROeiKGY7Cf1OwNhAmn8QZ8K5
S7wv9hbPZvkByQSsaNgDzzFUuYTP7cR9ILbwkHDixlpLyESnPzAsip5H4rq8gxLn
OwRKFGRvGid72EaapEY3yMA++EfzPfnebUmiLakSfWLHquh+0XQ=
=u3qb
-----END PGP SIGNATURE-----
Merge tag 'x86_cache_for_v5.10' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86 cache resource control updates from Borislav Petkov:
- Misc cleanups to the resctrl code in preparation for the ARM side
(James Morse)
- Add support for controlling per-thread memory bandwidth throttling
delay values on hw which supports it (Fenghua Yu)
* tag 'x86_cache_for_v5.10' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/resctrl: Enable user to view thread or core throttling mode
x86/resctrl: Enumerate per-thread MBA controls
cacheinfo: Move resctrl's get_cache_id() to the cacheinfo header file
x86/resctrl: Add struct rdt_cache::arch_has_{sparse, empty}_bitmaps
x86/resctrl: Merge AMD/Intel parse_bw() calls
x86/resctrl: Add struct rdt_membw::arch_needs_linear to explain AMD/Intel MBA difference
x86/resctrl: Use is_closid_match() in more places
x86/resctrl: Include pid.h
x86/resctrl: Use container_of() in delayed_work handlers
x86/resctrl: Fix stale comment
x86/resctrl: Remove struct rdt_membw::max_delay
x86/resctrl: Remove unused struct mbm_state::chunks_bw
Work submission instruction comes in two flavors. ENQCMD can be called
both in ring 3 and ring 0 and always uses the contents of a PASID MSR
when shipping the command to the device. ENQCMDS allows a kernel driver
to submit commands on behalf of a user process. The driver supplies the
PASID value in ENQCMDS. There isn't any usage of ENQCMD in the kernel as
of now.
The CPU feature flag is shown as "enqcmd" in /proc/cpuinfo.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/1600187413-163670-5-git-send-email-fenghua.yu@intel.com
Some systems support per-thread Memory Bandwidth Allocation (MBA) which
applies a throttling delay value to each hardware thread instead of to
a core. Per-thread MBA is enumerated by CPUID.
No feature flag is shown in /proc/cpuinfo. User applications need to
check a resctrl throttling mode info file to know if the feature is
supported.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Babu Moger <babu.moger@amd.com>
Reviewed-by: Reinette Chatre <reinette.chatre@intel.com>
Link: https://lkml.kernel.org/r/1598296281-127595-2-git-send-email-fenghua.yu@intel.com
Add a new AVX512 instruction group/feature for enumeration in
/proc/cpuinfo: AVX512_VP2INTERSECT.
CPUID.(EAX=7,ECX=0):EDX[bit 8] AVX512_VP2INTERSECT
Detailed information of CPUID bits for this feature can be found in
the Intel Architecture Intsruction Set Extensions Programming Reference
document (refer to Table 1-2). A copy of this document is available at
https://bugzilla.kernel.org/show_bug.cgi?id=204215.
Signed-off-by: Gayatri Kammela <gayatri.kammela@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190717234632.32673-3-gayatri.kammela@intel.com
Improve code readability by adding a tab between the elements of each
structure in an array of cpuid-dep struct so longer feature names will fit.
Signed-off-by: Gayatri Kammela <gayatri.kammela@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190717234632.32673-2-gayatri.kammela@intel.com
Pull x86 CPU feature updates from Thomas Gleixner:
"Updates for x86 CPU features:
- Support for UMWAIT/UMONITOR, which allows to use MWAIT and MONITOR
instructions in user space to save power e.g. in HPC workloads
which spin wait on synchronization points.
The maximum time a MWAIT can halt in userspace is controlled by the
kernel and can be adjusted by the sysadmin.
- Speed up the MTRR handling code on CPUs which support cache
self-snooping correctly.
On those CPUs the wbinvd() invocations can be omitted which speeds
up the MTRR setup by a factor of 50.
- Support for the new x86 vendor Zhaoxin who develops processors
based on the VIA Centaur technology.
- Prevent 'cat /proc/cpuinfo' from affecting isolated NOHZ_FULL CPUs
by sending IPIs to retrieve the CPU frequency and use the cached
values instead.
- The addition and late revert of the FSGSBASE support. The revert
was required as it turned out that the code still has hard to
diagnose issues. Yet another engineering trainwreck...
- Small fixes, cleanups, improvements and the usual new Intel CPU
family/model addons"
* 'x86-cpu-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (41 commits)
x86/fsgsbase: Revert FSGSBASE support
selftests/x86/fsgsbase: Fix some test case bugs
x86/entry/64: Fix and clean up paranoid_exit
x86/entry/64: Don't compile ignore_sysret if 32-bit emulation is enabled
selftests/x86: Test SYSCALL and SYSENTER manually with TF set
x86/mtrr: Skip cache flushes on CPUs with cache self-snooping
x86/cpu/intel: Clear cache self-snoop capability in CPUs with known errata
Documentation/ABI: Document umwait control sysfs interfaces
x86/umwait: Add sysfs interface to control umwait maximum time
x86/umwait: Add sysfs interface to control umwait C0.2 state
x86/umwait: Initialize umwait control values
x86/cpufeatures: Enumerate user wait instructions
x86/cpu: Disable frequency requests via aperfmperf IPI for nohz_full CPUs
x86/acpi/cstate: Add Zhaoxin processors support for cache flush policy in C3
ACPI, x86: Add Zhaoxin processors support for NONSTOP TSC
x86/cpu: Create Zhaoxin processors architecture support file
x86/cpu: Split Tremont based Atoms from the rest
Documentation/x86/64: Add documentation for GS/FS addressing mode
x86/elf: Enumerate kernel FSGSBASE capability in AT_HWCAP2
x86/cpu: Enable FSGSBASE on 64bit by default and add a chicken bit
...
The command line option `no387' is designed to disable the FPU
entirely. This only 'works' with CONFIG_MATH_EMULATION enabled.
But on 64bit this cannot work because user space expects SSE to work which
required basic FPU support. MATH_EMULATION does not help because SSE is not
emulated.
The command line option `nofxsr' should also be limited to 32bit because
FXSR is part of the required flags on 64bit so turning it off is not
possible.
Clearing X86_FEATURE_FPU without emulation enabled will not work anyway and
hang in fpu__init_system_early_generic() before the console is enabled.
Setting additioal dependencies, ensures that the CPU still boots on a
modern CPU. Otherwise, dropping FPU will leave FXSR enabled causing the
kernel to crash early in fpu__init_system_mxcsr().
With XSAVE support it will crash in fpu__init_cpu_xstate(). The problem is
that xsetbv() with XMM set and SSE cleared is not allowed. That means
XSAVE has to be disabled. The XSAVE support is disabled in
fpu__init_system_xstate_size_legacy() but it is too late. It can be
removed, it has been added in commit
1f999ab5a1 ("x86, xsave: Disable xsave in i387 emulation mode")
to use `no387' on a CPU with XSAVE support.
All this happens before console output.
After hat, the next possible crash is in RAID6 detect code because MMX
remained enabled. With a 3DNOW enabled config it will explode in memcpy()
for instance due to kernel_fpu_begin() but this is unconditionally enabled.
This is enough to boot a Debian Wheezy on a 32bit qemu "host" CPU which
supports everything up to XSAVES, AVX2 without 3DNOW. Later, Debian
increased the minimum requirements to i686 which means it does not boot
userland atleast due to CMOV.
After masking the additional features it still keeps SSE4A and 3DNOW*
enabled (if present on the host) but those are unused in the kernel.
Restrict `no387' and `nofxsr' otions to 32bit only. Add dependencies for
FPU, FXSR to additionaly mask CMOV, MMX, XSAVE if FXSR or FPU is cleared.
Reported-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190703083247.57kjrmlxkai3vpw3@linutronix.de
AVX512 BFLOAT16 instructions support 16-bit BFLOAT16 floating-point
format (BF16) for deep learning optimization.
BF16 is a short version of 32-bit single-precision floating-point
format (FP32) and has several advantages over 16-bit half-precision
floating-point format (FP16). BF16 keeps FP32 accumulation after
multiplication without loss of precision, offers more than enough
range for deep learning training tasks, and doesn't need to handle
hardware exception.
AVX512 BFLOAT16 instructions are enumerated in CPUID.7.1:EAX[bit 5]
AVX512_BF16.
CPUID.7.1:EAX contains only feature bits. Reuse the currently empty
word 12 as a pure features word to hold the feature bits including
AVX512_BF16.
Detailed information of the CPUID bit and AVX512 BFLOAT16 instructions
can be found in the latest Intel Architecture Instruction Set Extensions
and Future Features Programming Reference.
[ bp: Check CPUID(7) subleaf validity before accessing subleaf 1. ]
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: "Chang S. Bae" <chang.seok.bae@intel.com>
Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: Nadav Amit <namit@vmware.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: Radim Krcmar <rkrcmar@redhat.com>
Cc: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com>
Cc: "Ravi V Shankar" <ravi.v.shankar@intel.com>
Cc: Robert Hoo <robert.hu@linux.intel.com>
Cc: "Sean J Christopherson" <sean.j.christopherson@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Thomas Lendacky <Thomas.Lendacky@amd.com>
Cc: x86 <x86@kernel.org>
Link: https://lkml.kernel.org/r/1560794416-217638-3-git-send-email-fenghua.yu@intel.com
It's a waste for the four X86_FEATURE_CQM_* feature bits to occupy two
whole feature bits words. To better utilize feature words, re-define
word 11 to host scattered features and move the four X86_FEATURE_CQM_*
features into Linux defined word 11. More scattered features can be
added in word 11 in the future.
Rename leaf 11 in cpuid_leafs to CPUID_LNX_4 to reflect it's a
Linux-defined leaf.
Rename leaf 12 as CPUID_DUMMY which will be replaced by a meaningful
name in the next patch when CPUID.7.1:EAX occupies world 12.
Maximum number of RMID and cache occupancy scale are retrieved from
CPUID.0xf.1 after scattered CQM features are enumerated. Carve out the
code into a separate function.
KVM doesn't support resctrl now. So it's safe to move the
X86_FEATURE_CQM_* features to scattered features word 11 for KVM.
Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: Aaron Lewis <aaronlewis@google.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Babu Moger <babu.moger@amd.com>
Cc: "Chang S. Bae" <chang.seok.bae@intel.com>
Cc: "Sean J Christopherson" <sean.j.christopherson@intel.com>
Cc: Frederic Weisbecker <frederic@kernel.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: kvm ML <kvm@vger.kernel.org>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Peter Feiner <pfeiner@google.com>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com>
Cc: Ravi V Shankar <ravi.v.shankar@intel.com>
Cc: Sherry Hurwitz <sherry.hurwitz@amd.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Thomas Lendacky <Thomas.Lendacky@amd.com>
Cc: x86 <x86@kernel.org>
Link: https://lkml.kernel.org/r/1560794416-217638-2-git-send-email-fenghua.yu@intel.com
This is the only spot where the 'const static' specifier is used;
everywhere else 'static const' is preferred, as static should be the
first specifier.
This is just a cosmetic fix that aligns this, no functional change.
Signed-off-by: Ralf Ramsauer <ralf.ramsauer@oth-regensburg.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Gayatri Kammela <gayatri.kammela@intel.com>
Link: https://lkml.kernel.org/r/20180307160734.6691-1-ralf.ramsauer@oth-regensburg.de
Peter pointed out that the set/clear_bit32() variants are broken in various
aspects.
Replace them with open coded set/clear_bit() and type cast
cpu_info::x86_capability as it's done in all other places throughout x86.
Fixes: 0b00de857a ("x86/cpuid: Add generic table for CPUID dependencies")
Reported-by: Peter Ziljstra <peterz@infradead.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andi Kleen <ak@linux.intel.com>
Add a few new SSE/AVX/AVX512 instruction groups/features for enumeration
in /proc/cpuinfo: AVX512_VBMI2, GFNI, VAES, VPCLMULQDQ, AVX512_VNNI,
AVX512_BITALG.
CPUID.(EAX=7,ECX=0):ECX[bit 6] AVX512_VBMI2
CPUID.(EAX=7,ECX=0):ECX[bit 8] GFNI
CPUID.(EAX=7,ECX=0):ECX[bit 9] VAES
CPUID.(EAX=7,ECX=0):ECX[bit 10] VPCLMULQDQ
CPUID.(EAX=7,ECX=0):ECX[bit 11] AVX512_VNNI
CPUID.(EAX=7,ECX=0):ECX[bit 12] AVX512_BITALG
Detailed information of CPUID bits for these features can be found
in the Intel Architecture Instruction Set Extensions and Future Features
Programming Interface document (refer to Table 1-1. and Table 1-2.).
A copy of this document is available at
https://bugzilla.kernel.org/show_bug.cgi?id=197239
Signed-off-by: Gayatri Kammela <gayatri.kammela@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ravi Shankar <ravi.v.shankar@intel.com>
Cc: Ricardo Neri <ricardo.neri@intel.com>
Cc: Yang Zhong <yang.zhong@intel.com>
Cc: bp@alien8.de
Link: http://lkml.kernel.org/r/1509412829-23380-1-git-send-email-gayatri.kammela@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
do_clear_cpu_cap() allocates a bitmap to keep track of disabled feature
dependencies. That bitmap is sized NCAPINTS * BITS_PER_INIT. The possible
'features' which can be handed in are larger than this, because after the
capabilities the bug 'feature' bits occupy another 32bit. Not really
obvious...
So clearing any of the misfeature bits, as 32bit does for the F00F bug,
accesses that bitmap out of bounds thereby corrupting the stack.
Size the bitmap proper and add a sanity check to catch accidental out of
bound access.
Fixes: 0b00de857a ("x86/cpuid: Add generic table for CPUID dependencies")
Reported-by: kernel test robot <xiaolong.ye@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Borislav Petkov <bp@alien8.de>
Link: https://lkml.kernel.org/r/20171018022023.GA12058@yexl-desktop
Some CPUID features depend on other features. Currently it's
possible to to clear dependent features, but not clear the base features,
which can cause various interesting problems.
This patch implements a generic table to describe dependencies
between CPUID features, to be used by all code that clears
CPUID.
Some subsystems (like XSAVE) had an own implementation of this,
but it's better to do it all in a single place for everyone.
Then clear_cpu_cap and setup_clear_cpu_cap always look up
this table and clear all dependencies too.
This is intended to be a practical table: only for features
that make sense to clear. If someone for example clears FPU,
or other features that are essentially part of the required
base feature set, not much is going to work. Handling
that is right now out of scope. We're only handling
features which can be usefully cleared.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Jonathan McDowell <noodles@earth.li>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20171013215645.23166-3-andi@firstfloor.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
