Intel Trust Domain Extensions (TDX) protects guest VMs from malicious
host and certain physical attacks. A CPU-attested software module
called 'the TDX module' runs inside a new isolated memory range as a
trusted hypervisor to manage and run protected VMs.
Pre-TDX Intel hardware has support for a memory encryption architecture
called MKTME. The memory encryption hardware underpinning MKTME is also
used for Intel TDX. TDX ends up "stealing" some of the physical address
space from the MKTME architecture for crypto-protection to VMs. The
BIOS is responsible for partitioning the "KeyID" space between legacy
MKTME and TDX. The KeyIDs reserved for TDX are called 'TDX private
KeyIDs' or 'TDX KeyIDs' for short.
During machine boot, TDX microcode verifies that the BIOS programmed TDX
private KeyIDs consistently and correctly programmed across all CPU
packages. The MSRs are locked in this state after verification. This
is why MSR_IA32_MKTME_KEYID_PARTITIONING gets used for TDX enumeration:
it indicates not just that the hardware supports TDX, but that all the
boot-time security checks passed.
The TDX module is expected to be loaded by the BIOS when it enables TDX,
but the kernel needs to properly initialize it before it can be used to
create and run any TDX guests. The TDX module will be initialized by
the KVM subsystem when KVM wants to use TDX.
Detect platform TDX support by detecting TDX private KeyIDs.
The TDX module itself requires one TDX KeyID as the 'TDX global KeyID'
to protect its metadata. Each TDX guest also needs a TDX KeyID for its
own protection. Just use the first TDX KeyID as the global KeyID and
leave the rest for TDX guests. If no TDX KeyID is left for TDX guests,
disable TDX as initializing the TDX module alone is useless.
[ dhansen: add X86_FEATURE, replace helper function ]
Signed-off-by: Kai Huang <kai.huang@intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Reviewed-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Isaku Yamahata <isaku.yamahata@intel.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Dave Hansen <dave.hansen@linux.intel.com>
Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
Link: https://lore.kernel.org/all/20231208170740.53979-1-dave.hansen%40intel.com
There is no real reason to have a separate ASM entry point implementation
for the legacy INT 0x80 syscall emulation on 64-bit.
IDTENTRY provides all the functionality needed with the only difference
that it does not:
- save the syscall number (AX) into pt_regs::orig_ax
- set pt_regs::ax to -ENOSYS
Both can be done safely in the C code of an IDTENTRY before invoking any of
the syscall related functions which depend on this convention.
Aside of ASM code reduction this prepares for detecting and handling a
local APIC injected vector 0x80.
[ kirill.shutemov: More verbose comments ]
Suggested-by: Linus Torvalds <torvalds@linuxfoundation.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Reviewed-by: Borislav Petkov (AMD) <bp@alien8.de>
Cc: <stable@vger.kernel.org> # v6.0+
Now that GENERIC_CPU_DEVICES calls arch_register_cpu(), which can be
overridden by the arch code, switch over to this to allow common code
to choose when the register_cpu() call is made.
x86's struct cpus come from struct x86_cpu, which has no other members
or users. Remove this and use the version defined by common code.
This is an intermediate step to the logic being moved to drivers/acpi,
where GENERIC_CPU_DEVICES will do the work when booting with acpi=off.
This patch also has the effect of moving the registration of CPUs from
subsys to driver core initialisation, prior to any initcalls running.
----
Changes since RFC:
* Fixed the second copy of arch_register_cpu() used for non-hotplug
Changes since RFC v2:
* Remove duplicate of the weak generic arch_register_cpu(), spotted
by Jonathan Cameron. Add note about initialisation order change.
Changes since RFC v3:
* Adapt to removal of EXPORT_SYMBOL()s
Signed-off-by: James Morse <james.morse@arm.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: "Russell King (Oracle)" <rmk+kernel@armlinux.org.uk>
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/E1r5R3l-00Cszm-UA@rmk-PC.armlinux.org.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
intel_epb_init() is called as a subsys_initcall() to register cpuhp
callbacks. The callbacks make use of get_cpu_device() which will return
NULL unless register_cpu() has been called. register_cpu() is called
from topology_init(), which is also a subsys_initcall().
This is fragile. Moving the register_cpu() to a different
subsys_initcall() leads to a NULL dereference during boot.
Make intel_epb_init() a late_initcall(), user-space can't provide a
policy before this point anyway.
Signed-off-by: James Morse <james.morse@arm.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Signed-off-by: "Russell King (Oracle)" <rmk+kernel@armlinux.org.uk>
Acked-by: "Rafael J. Wysocki" <rafael.j.wysocki@intel.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/E1r5R2m-00Csyb-2S@rmk-PC.armlinux.org.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit in Fixes added an AMD-specific microcode callback. However, it
didn't check the CPU vendor the kernel runs on explicitly.
The only reason the Zenbleed check in it didn't run on other x86 vendors
hardware was pure coincidental luck:
if (!cpu_has_amd_erratum(c, amd_zenbleed))
return;
gives true on other vendors because they don't have those families and
models.
However, with the removal of the cpu_has_amd_erratum() in
05f5f73936 ("x86/CPU/AMD: Drop now unused CPU erratum checking function")
that coincidental condition is gone, leading to the zenbleed check
getting executed on other vendors too.
Add the explicit vendor check for the whole callback as it should've
been done in the first place.
Fixes: 522b1d6921 ("x86/cpu/amd: Add a Zenbleed fix")
Cc: <stable@kernel.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20231201184226.16749-1-bp@alien8.de
After successful update, the late loading routine prints an update
summary similar to:
microcode: load: updated on 128 primary CPUs with 128 siblings
microcode: revision: 0x21000170 -> 0x21000190
Remove the redundant message in the Intel side of the driver.
[ bp: Massage commit message. ]
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/ZWjYhedNfhAUmt0k@a4bf019067fa.jf.intel.com
Use atomic_try_cmpxchg() instead of atomic_cmpxchg(*ptr, old, new) == old.
X86 CMPXCHG instruction returns success in ZF flag, so this change saves a
compare after the CMPXCHG.
Tested by building a native Fedora-38 kernel and rebooting
a 12-way SMP system using "shutdown -r" command some 100 times.
No functional change intended.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20231123203605.3474745-2-ubizjak@gmail.com
Improve code generation in native_stop_other_cpus() a tiny bit:
smp_processor_id() accesses a per-CPU variable, so the compiler
is not able to move the call after the early exit on its own.
Also rename the "cpu" variable to a more descriptive "this_cpu", and
use 'cpu' as a separate iterator variable later in the function.
No functional change intended.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20231123203605.3474745-1-ubizjak@gmail.com
This is a "nice-to-have" change with minor code generation benefits:
- Instruction with %rip-relative address operand is one byte shorter than
its absolute address counterpart,
- it is also compatible with position independent executable (-fpie) builds,
- it is also consistent with what the compiler emits by default when
a symbol is accessed.
No functional changes intended.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Rafael J. Wysocki <rafael@kernel.org>
Link: https://lore.kernel.org/r/20231103104900.409470-1-ubizjak@gmail.com
Contrary to alternatives, relocations are currently not supported in
call thunk templates. Re-use the existing infrastructure from
alternative.c to allow %rip-relative relocations when copying call
thunk template from its storage location.
The patch allows unification of ASM_INCREMENT_CALL_DEPTH, which already
uses PER_CPU_VAR macro, with INCREMENT_CALL_DEPTH, used in call thunk
template, which is currently limited to use absolute address.
Reuse existing relocation infrastructure from alternative.c.,
as suggested by Peter Zijlstra.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20231105213731.1878100-3-ubizjak@gmail.com
A write-access violation page fault kernel crash was observed while running
cpuhotplug LTP testcases on SEV-ES enabled systems. The crash was
observed during hotplug, after the CPU was offlined and the process
was migrated to different CPU. setup_ghcb() is called again which
tries to update ghcb_version in sev_es_negotiate_protocol(). Ideally this
is a read_only variable which is initialised during booting.
Trying to write it results in a pagefault:
BUG: unable to handle page fault for address: ffffffffba556e70
#PF: supervisor write access in kernel mode
#PF: error_code(0x0003) - permissions violation
[ ...]
Call Trace:
<TASK>
? __die_body.cold+0x1a/0x1f
? __die+0x2a/0x35
? page_fault_oops+0x10c/0x270
? setup_ghcb+0x71/0x100
? __x86_return_thunk+0x5/0x6
? search_exception_tables+0x60/0x70
? __x86_return_thunk+0x5/0x6
? fixup_exception+0x27/0x320
? kernelmode_fixup_or_oops+0xa2/0x120
? __bad_area_nosemaphore+0x16a/0x1b0
? kernel_exc_vmm_communication+0x60/0xb0
? bad_area_nosemaphore+0x16/0x20
? do_kern_addr_fault+0x7a/0x90
? exc_page_fault+0xbd/0x160
? asm_exc_page_fault+0x27/0x30
? setup_ghcb+0x71/0x100
? setup_ghcb+0xe/0x100
cpu_init_exception_handling+0x1b9/0x1f0
The fix is to call sev_es_negotiate_protocol() only in the BSP boot phase,
and it only needs to be done once in any case.
[ mingo: Refined the changelog. ]
Fixes: 95d33bfaa3 ("x86/sev: Register GHCB memory when SEV-SNP is active")
Suggested-by: Tom Lendacky <thomas.lendacky@amd.com>
Co-developed-by: Bo Gan <bo.gan@broadcom.com>
Signed-off-by: Bo Gan <bo.gan@broadcom.com>
Signed-off-by: Ashwin Dayanand Kamat <ashwin.kamat@broadcom.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Tom Lendacky <thomas.lendacky@amd.com>
Link: https://lore.kernel.org/r/1701254429-18250-1-git-send-email-kashwindayan@vmware.com
The long names of the SMCA banks are only used by the MCE decoder
module.
Move them out of the arch code and into the decoder module.
[ bp: Name the long names array "smca_long_names", drop local ptr in
decode_smca_error(), constify arrays. ]
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20231118193248.1296798-5-yazen.ghannam@amd.com
Pull x86 microcode fixes from Ingo Molnar:
"Fix/enhance x86 microcode version reporting: fix the bootup log spam,
and remove the driver version announcement to avoid version confusion
when distros backport fixes"
* tag 'x86-urgent-2023-11-26' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/microcode: Rework early revisions reporting
x86/microcode: Remove the driver announcement and version
intel_epb_init() is called as a subsys_initcall() to register cpuhp
callbacks. The callbacks make use of get_cpu_device() which will return
NULL unless register_cpu() has been called. register_cpu() is called
from topology_init(), which is also a subsys_initcall().
This is fragile. Moving the register_cpu() to a different
subsys_initcall() leads to a NULL dereference during boot.
Make intel_epb_init() a late_initcall(), user-space can't provide a
policy before this point anyway.
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Acked-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
some architectures run into a -Wmissing-prototypes warning
for trap_init()
arch/microblaze/kernel/traps.c:21:6: warning: no previous prototype for 'trap_init' [-Wmissing-prototypes]
Include the right header to avoid this consistently, removing
the extra declarations on m68k and x86 that were added as local
workarounds already.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
AMD systems generally allow MCA "simulation" where MCA registers can be
written with valid data and the full MCA handling flow can be tested by
software.
However, the platform on Scalable MCA systems, can prevent software from
writing data to the MCA registers. There is no architectural way to
determine this configuration. Therefore, the MCE injection module will
check for this behavior by writing and reading back a test status value.
This is done during module init, and the check can run on any CPU with
any valid MCA bank.
If MCA_STATUS writes are ignored by the platform, then there are no side
effects on the hardware state.
If the writes are not ignored, then the test status value will remain in
the hardware MCA_STATUS register. It is likely that the value will not
be overwritten by hardware or software, since the tested CPU and bank
are arbitrary. Therefore, the user may see a spurious, synthetic MCA
error reported whenever MCA is polled for this CPU.
Clear the test value immediately after writing it. It is very unlikely
that a valid MCA error is logged by hardware during the test. Errors
that cause an #MC won't be affected.
Fixes: 891e465a1b ("x86/mce: Check whether writes to MCA_STATUS are getting ignored")
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20231118193248.1296798-2-yazen.ghannam@amd.com
Pull hyperv fixes from Wei Liu:
- One fix for the KVP daemon (Ani Sinha)
- Fix for the detection of E820_TYPE_PRAM in a Gen2 VM (Saurabh Sengar)
- Micro-optimization for hv_nmi_unknown() (Uros Bizjak)
* tag 'hyperv-fixes-signed-20231121' of git://git.kernel.org/pub/scm/linux/kernel/git/hyperv/linux:
x86/hyperv: Use atomic_try_cmpxchg() to micro-optimize hv_nmi_unknown()
x86/hyperv: Fix the detection of E820_TYPE_PRAM in a Gen2 VM
hv/hv_kvp_daemon: Some small fixes for handling NM keyfiles
The AMD side of the loader issues the microcode revision for each
logical thread on the system, which can become really noisy on huge
machines. And doing that doesn't make a whole lot of sense - the
microcode revision is already in /proc/cpuinfo.
So in case one is interested in the theoretical support of mixed silicon
steppings on AMD, one can check there.
What is also missing on the AMD side - something which people have
requested before - is showing the microcode revision the CPU had
*before* the early update.
So abstract that up in the main code and have the BSP on each vendor
provide those revision numbers.
Then, dump them only once on driver init.
On Intel, do not dump the patch date - it is not needed.
Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/CAHk-=wg=%2B8rceshMkB4VnKxmRccVLtBLPBawnewZuuqyx5U=3A@mail.gmail.com
First of all, the print is useless. The driver will either load and say
which microcode revision the machine has or issue an error.
Then, the version number is meaningless and actively confusing, as Yazen
mentioned recently: when a subset of patches are backported to a distro
kernel, one can't assume the driver version is the same as the upstream
one. And besides, the version number of the loader hasn't been used and
incremented for a long time. So drop it.
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20231115210212.9981-2-bp@alien8.de
Pull x86 fixes from Borislav Petkov:
- Ignore invalid x2APIC entries in order to not waste per-CPU data
- Fix a back-to-back signals handling scenario when shadow stack is in
use
- A documentation fix
- Add Kirill as TDX maintainer
* tag 'x86_urgent_for_v6.7_rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/acpi: Ignore invalid x2APIC entries
x86/shstk: Delay signal entry SSP write until after user accesses
x86/Documentation: Indent 'note::' directive for protocol version number note
MAINTAINERS: Add Intel TDX entry
mce_device_create() is called only from mce_cpu_online() which in turn
will be called iff MCA support is available. That is, at the time of
mce_device_create() call it's guaranteed that MCA support is available.
No need to duplicate this check so remove it.
[ bp: Massage commit message. ]
Signed-off-by: Nikolay Borisov <nik.borisov@suse.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20231107165529.407349-1-nik.borisov@suse.com
Similar to the alternative patching, use a relative reference for original
instruction offset rather than absolute one, which saves 8 bytes for one
PARA_SITE entry on x86_64. As a result, a R_X86_64_PC32 relocation is
generated instead of an R_X86_64_64 one, which also reduces relocation
metadata on relocatable builds. Hardcode the alignment to 4 now.
[ bp: Massage commit message. ]
Signed-off-by: Hou Wenlong <houwenlong.hwl@antgroup.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Juergen Gross <jgross@suse.com>
Link: https://lore.kernel.org/r/9e6053107fbaabc0d33e5d2865c5af2c67ec9925.1686301237.git.houwenlong.hwl@antgroup.com
AMD does not have the requirement for a synchronization barrier when
acccessing a certain group of MSRs. Do not incur that unnecessary
penalty there.
There will be a CPUID bit which explicitly states that a MFENCE is not
needed. Once that bit is added to the APM, this will be extended with
it.
While at it, move to processor.h to avoid include hell. Untangling that
file properly is a matter for another day.
Some notes on the performance aspect of why this is relevant, courtesy
of Kishon VijayAbraham <Kishon.VijayAbraham@amd.com>:
On a AMD Zen4 system with 96 cores, a modified ipi-bench[1] on a VM
shows x2AVIC IPI rate is 3% to 4% lower than AVIC IPI rate. The
ipi-bench is modified so that the IPIs are sent between two vCPUs in the
same CCX. This also requires to pin the vCPU to a physical core to
prevent any latencies. This simulates the use case of pinning vCPUs to
the thread of a single CCX to avoid interrupt IPI latency.
In order to avoid run-to-run variance (for both x2AVIC and AVIC), the
below configurations are done:
1) Disable Power States in BIOS (to prevent the system from going to
lower power state)
2) Run the system at fixed frequency 2500MHz (to prevent the system
from increasing the frequency when the load is more)
With the above configuration:
*) Performance measured using ipi-bench for AVIC:
Average Latency: 1124.98ns [Time to send IPI from one vCPU to another vCPU]
Cumulative throughput: 42.6759M/s [Total number of IPIs sent in a second from
48 vCPUs simultaneously]
*) Performance measured using ipi-bench for x2AVIC:
Average Latency: 1172.42ns [Time to send IPI from one vCPU to another vCPU]
Cumulative throughput: 40.9432M/s [Total number of IPIs sent in a second from
48 vCPUs simultaneously]
From above, x2AVIC latency is ~4% more than AVIC. However, the expectation is
x2AVIC performance to be better or equivalent to AVIC. Upon analyzing
the perf captures, it is observed significant time is spent in
weak_wrmsr_fence() invoked by x2apic_send_IPI().
With the fix to skip weak_wrmsr_fence()
*) Performance measured using ipi-bench for x2AVIC:
Average Latency: 1117.44ns [Time to send IPI from one vCPU to another vCPU]
Cumulative throughput: 42.9608M/s [Total number of IPIs sent in a second from
48 vCPUs simultaneously]
Comparing the performance of x2AVIC with and without the fix, it can be seen
the performance improves by ~4%.
Performance captured using an unmodified ipi-bench using the 'mesh-ipi' option
with and without weak_wrmsr_fence() on a Zen4 system also showed significant
performance improvement without weak_wrmsr_fence(). The 'mesh-ipi' option ignores
CCX or CCD and just picks random vCPU.
Average throughput (10 iterations) with weak_wrmsr_fence(),
Cumulative throughput: 4933374 IPI/s
Average throughput (10 iterations) without weak_wrmsr_fence(),
Cumulative throughput: 6355156 IPI/s
[1] https://github.com/bytedance/kvm-utils/tree/master/microbenchmark/ipi-bench
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230622095212.20940-1-bp@alien8.de
Memory errors don't happen very often, especially fatal ones. However,
in large-scale scenarios such as data centers, that probability
increases with the amount of machines present.
When a fatal machine check happens, mce_panic() is called based on the
severity grading of that error. The page containing the error is not
marked as poison.
However, when kexec is enabled, tools like makedumpfile understand when
pages are marked as poison and do not touch them so as not to cause
a fatal machine check exception again while dumping the previous
kernel's memory.
Therefore, mark the page containing the error as poisoned so that the
kexec'ed kernel can avoid accessing the page.
[ bp: Rewrite commit message and comment. ]
Co-developed-by: Youquan Song <youquan.song@intel.com>
Signed-off-by: Youquan Song <youquan.song@intel.com>
Signed-off-by: Zhiquan Li <zhiquan1.li@intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Link: https://lore.kernel.org/r/20231014051754.3759099-1-zhiquan1.li@intel.com
