When using resctrl on systems with Sub-NUMA Clustering enabled, monitoring
groups may be allocated RMID values which would overrun the
arch_mbm_{local,total} arrays.
This is due to inconsistencies in whether the SNC-adjusted num_rmid value or
the unadjusted value in resctrl_arch_system_num_rmid_idx() is used. The
num_rmid value for the L3 resource is currently:
resctrl_arch_system_num_rmid_idx() / snc_nodes_per_l3_cache
As a simple fix, make resctrl_arch_system_num_rmid_idx() return the
SNC-adjusted, L3 num_rmid value on x86.
Fixes: e13db55b5a ("x86/resctrl: Introduce snc_nodes_per_l3_cache")
Signed-off-by: Peter Newman <peternewman@google.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Reinette Chatre <reinette.chatre@intel.com>
Link: https://lore.kernel.org/r/20240822190212.1848788-1-peternewman@google.com
The FRED RSP0 MSR points to the top of the kernel stack for user level
event delivery. As this is the task stack it needs to be updated when a
task is scheduled in.
The update is done at context switch. That means it's also done when
switching to kernel threads, which is pointless as those never go out to
user space. For KVM threads this means there are two writes to FRED_RSP0 as
KVM has to switch to the guest value before VMENTER.
Defer the update to the exit to user space path and cache the per CPU
FRED_RSP0 value, so redundant writes can be avoided.
Provide fred_sync_rsp0() for KVM to keep the cache in sync with the actual
MSR value after returning from guest to host mode.
[ tglx: Massage change log ]
Suggested-by: Sean Christopherson <seanjc@google.com>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Xin Li (Intel) <xin@zytor.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240822073906.2176342-4-xin@zytor.com
SS is initialized to NULL during boot time and not explicitly set to
__KERNEL_DS.
With FRED enabled, if a kernel event is delivered before a CPU goes to
user level for the first time, its SS is NULL thus NULL is pushed into
the SS field of the FRED stack frame. But before ERETS is executed,
the CPU may context switch to another task and go to user level. Then
when the CPU comes back to kernel mode, SS is changed to __KERNEL_DS.
Later when ERETS is executed to return from the kernel event handler,
a #GP fault is generated because SS doesn't match the SS saved in the
FRED stack frame.
Initialize SS to __KERNEL_DS when enabling FRED to prevent that.
Note, IRET doesn't check if SS matches the SS saved in its stack frame,
thus IDT doesn't have this problem. For IDT it doesn't matter whether
SS is set to __KERNEL_DS or not, because it's set to NULL upon interrupt
or exception delivery and __KERNEL_DS upon SYSCALL. Thus it's pointless
to initialize SS for IDT.
Signed-off-by: Xin Li (Intel) <xin@zytor.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240816104316.2276968-1-xin@zytor.com
init_per_cpu_var() returns a pointer in the percpu address space while
rip_rel_ptr() expects a pointer in the generic address space.
When strict address space checks are enabled, GCC's named address space
checks fail:
asm.h:124:63: error: passing argument 1 of 'rip_rel_ptr' from
pointer to non-enclosed address space
Add a explicit cast to remove address space of the returned pointer.
Fixes: 11e36b0f7c ("x86/boot/64: Load the final kernel GDT during early boot directly, remove startup_gdt[]")
Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240819083334.148536-1-ubizjak@gmail.com
Building the SGX code with W=1 generates below warning:
arch/x86/kernel/cpu/sgx/main.c:741: warning: Function parameter or
struct member 'low' not described in 'sgx_calc_section_metric'
arch/x86/kernel/cpu/sgx/main.c:741: warning: Function parameter or
struct member 'high' not described in 'sgx_calc_section_metric'
...
The function sgx_calc_section_metric() is a simple helper which is only
used in sgx/main.c. There's no need to use kernel-doc style comment for
it.
Downgrade to a normal comment.
Signed-off-by: Kai Huang <kai.huang@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240825080649.145250-1-kai.huang@intel.com
Add defines for the architectural memory types that can be shoved into
various MSRs and registers, e.g. MTRRs, PAT, VMX capabilities MSRs, EPTPs,
etc. While most MSRs/registers support only a subset of all memory types,
the values themselves are architectural and identical across all users.
Leave the goofy MTRR_TYPE_* definitions as-is since they are in a uapi
header, but add compile-time assertions to connect the dots (and sanity
check that the msr-index.h values didn't get fat-fingered).
Keep the VMX_EPTP_MT_* defines so that it's slightly more obvious that the
EPTP holds a single memory type in 3 of its 64 bits; those bits just
happen to be 2:0, i.e. don't need to be shifted.
Opportunistically use X86_MEMTYPE_WB instead of an open coded '6' in
setup_vmcs_config().
No functional change intended.
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Kai Huang <kai.huang@intel.com>
Reviewed-by: Xiaoyao Li <xiaoyao.li@intel.com>
Reviewed-by: Kai Huang <kai.huang@intel.com>
Link: https://lore.kernel.org/r/20240605231918.2915961-2-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Refactor the list of constant variables into a macro.
This should make it easier to add more constants in the future.
Signed-off-by: Jann Horn <jannh@google.com>
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
There are two distinct CPU features related to the use of XSAVES and LBR:
whether LBR is itself supported and whether XSAVES supports LBR. The LBR
subsystem correctly checks both in intel_pmu_arch_lbr_init(), but the
XSTATE subsystem does not.
The LBR bit is only removed from xfeatures_mask_independent when LBR is not
supported by the CPU, but there is no validation of XSTATE support.
If XSAVES does not support LBR the write to IA32_XSS causes a #GP fault,
leaving the state of IA32_XSS unchanged, i.e. zero. The fault is handled
with a warning and the boot continues.
Consequently the next XRSTORS which tries to restore supervisor state fails
with #GP because the RFBM has zero for all supervisor features, which does
not match the XCOMP_BV field.
As XFEATURE_MASK_FPSTATE includes supervisor features setting up the FPU
causes a #GP, which ends up in fpu_reset_from_exception_fixup(). That fails
due to the same problem resulting in recursive #GPs until the kernel runs
out of stack space and double faults.
Prevent this by storing the supported independent features in
fpu_kernel_cfg during XSTATE initialization and use that cached value for
retrieving the independent feature bits to be written into IA32_XSS.
[ tglx: Massaged change log ]
Fixes: f0dccc9da4 ("x86/fpu/xstate: Support dynamic supervisor feature for LBR")
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Mitchell Levy <levymitchell0@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20240812-xsave-lbr-fix-v3-1-95bac1bf62f4@gmail.com
On 64-bit init_mem_mapping() relies on the minimal page fault handler
provided by the early IDT mechanism. The real page fault handler is
installed right afterwards into the IDT.
This is problematic on CPUs which have X86_FEATURE_FRED set because the
real page fault handler retrieves the faulting address from the FRED
exception stack frame and not from CR2, but that does obviously not work
when FRED is not yet enabled in the CPU.
To prevent this enable FRED right after init_mem_mapping() without
interrupt stacks. Those are enabled later in trap_init() after the CPU
entry area is set up.
[ tglx: Encapsulate the FRED details ]
Fixes: 14619d912b ("x86/fred: FRED entry/exit and dispatch code")
Reported-by: Hou Wenlong <houwenlong.hwl@antgroup.com>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Xin Li (Intel) <xin@zytor.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240709154048.3543361-4-xin@zytor.com
To enable FRED earlier, move the RSP initialization out of
cpu_init_fred_exceptions() into cpu_init_fred_rsps().
This is required as the FRED RSP initialization depends on the availability
of the CPU entry areas which are set up late in trap_init(),
No functional change intended. Marked with Fixes as it's a depedency for
the real fix.
Fixes: 14619d912b ("x86/fred: FRED entry/exit and dispatch code")
Signed-off-by: Xin Li (Intel) <xin@zytor.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240709154048.3543361-3-xin@zytor.com
Depending on whether FRED is enabled, sysvec_install() installs a system
interrupt handler into either into FRED's system vector dispatch table or
into the IDT.
However FRED can be disabled later in trap_init(), after sysvec_install()
has been invoked already; e.g., the HYPERVISOR_CALLBACK_VECTOR handler is
registered with sysvec_install() in kvm_guest_init(), which is called in
setup_arch() but way before trap_init().
IOW, there is a gap between FRED is available and available but disabled.
As a result, when FRED is available but disabled, early sysvec_install()
invocations fail to install the IDT handler resulting in spurious
interrupts.
Fix it by parsing cmdline param "fred=" in cpu_parse_early_param() to
ensure that FRED is disabled before the first sysvec_install() incovations.
Fixes: 3810da1271 ("x86/fred: Add a fred= cmdline param")
Reported-by: Hou Wenlong <houwenlong.hwl@antgroup.com>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Xin Li (Intel) <xin@zytor.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240709154048.3543361-2-xin@zytor.com
x2apic_disable() clears x2apic_state and x2apic_mode unconditionally, even
when the state is X2APIC_ON_LOCKED, which prevents the kernel to disable
it thereby creating inconsistent state.
Due to the early state check for X2APIC_ON, the code path which warns about
a locked X2APIC cannot be reached.
Test for state < X2APIC_ON instead and move the clearing of the state and
mode variables to the place which actually disables X2APIC.
[ tglx: Massaged change log. Added Fixes tag. Moved clearing so it's at the
right place for back ports ]
Fixes: a57e456a7b ("x86/apic: Fix fallout from x2apic cleanup")
Signed-off-by: Yuntao Wang <yuntao.wang@linux.dev>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20240813014827.895381-1-yuntao.wang@linux.dev
For any changes of struct fd representation we need to
turn existing accesses to fields into calls of wrappers.
Accesses to struct fd::flags are very few (3 in linux/file.h,
1 in net/socket.c, 3 in fs/overlayfs/file.c and 3 more in
explicit initializers).
Those can be dealt with in the commit converting to
new layout; accesses to struct fd::file are too many for that.
This commit converts (almost) all of f.file to
fd_file(f). It's not entirely mechanical ('file' is used as
a member name more than just in struct fd) and it does not
even attempt to distinguish the uses in pointer context from
those in boolean context; the latter will be eventually turned
into a separate helper (fd_empty()).
NOTE: mass conversion to fd_empty(), tempting as it
might be, is a bad idea; better do that piecewise in commit
that convert from fdget...() to CLASS(...).
[conflicts in fs/fhandle.c, kernel/bpf/syscall.c, mm/memcontrol.c
caught by git; fs/stat.c one got caught by git grep]
[fs/xattr.c conflict]
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Logical destination mode of the local APIC is used for systems with up to
8 CPUs. It has an advantage over physical destination mode as it allows to
target multiple CPUs at once with IPIs.
That advantage was definitely worth it when systems with up to 8 CPUs
were state of the art for servers and workstations, but that's history.
Aside of that there are systems which fail to work with logical destination
mode as the ACPI/DMI quirks show and there are AMD Zen1 systems out there
which fail when interrupt remapping is enabled as reported by Rob and
Christian. The latter problem can be cured by firmware updates, but not all
OEMs distribute the required changes.
Physical destination mode is guaranteed to work because it is the only way
to get a CPU up and running via the INIT/INIT/STARTUP sequence.
As the number of CPUs keeps increasing, logical destination mode becomes a
less used code path so there is no real good reason to keep it around.
Therefore remove logical destination mode support for 64-bit and default to
physical destination mode.
Reported-by: Rob Newcater <rob@durendal.co.uk>
Reported-by: Christian Heusel <christian@heusel.eu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Borislav Petkov (AMD) <bp@alien8.de>
Tested-by: Rob Newcater <rob@durendal.co.uk>
Link: https://lore.kernel.org/all/877cd5u671.ffs@tglx
Add Bus Lock Detect (called Bus Lock Trap in AMD docs) support for AMD
platforms. Bus Lock Detect is enumerated with CPUID Fn0000_0007_ECX_x0
bit [24 / BUSLOCKTRAP]. It can be enabled through MSR_IA32_DEBUGCTLMSR.
When enabled, hardware clears DR6[11] and raises a #DB exception on
occurrence of Bus Lock if CPL > 0. More detail about the feature can be
found in AMD APM[1].
[1]: AMD64 Architecture Programmer's Manual Pub. 40332, Rev. 4.07 - June
2023, Vol 2, 13.1.3.6 Bus Lock Trap
https://bugzilla.kernel.org/attachment.cgi?id=304653
Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>
Link: https://lore.kernel.org/all/20240808062937.1149-3-ravi.bangoria@amd.com
Bus Lock Detect functionality on AMD platforms works identical to Intel.
Move split_lock and bus_lock specific code from intel.c to a dedicated
file so that it can be compiled and supported on non-Intel platforms.
Also, introduce CONFIG_X86_BUS_LOCK_DETECT, make it dependent on
CONFIG_CPU_SUP_INTEL and add compilation dependency of the new bus_lock.c
file on CONFIG_X86_BUS_LOCK_DETECT.
Signed-off-by: Ravi Bangoria <ravi.bangoria@amd.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>
Link: https://lore.kernel.org/all/20240808062937.1149-2-ravi.bangoria@amd.com
MTRRs have an obsolete fixed variant for fine grained caching control
of the 640K-1MB region that uses separate MSRs. This fixed variant has
a separate capability bit in the MTRR capability MSR.
So far all x86 CPUs which support MTRR have this separate bit set, so it
went unnoticed that mtrr_save_state() does not check the capability bit
before accessing the fixed MTRR MSRs.
Though on a CPU that does not support the fixed MTRR capability this
results in a #GP. The #GP itself is harmless because the RDMSR fault is
handled gracefully, but results in a WARN_ON().
Add the missing capability check to prevent this.
Fixes: 2b1f6278d7 ("[PATCH] x86: Save the MTRRs of the BSP before booting an AP")
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20240808000244.946864-1-ak@linux.intel.com
The kernel can change spinlock behavior when running as a guest. But this
guest-friendly behavior causes performance problems on bare metal.
The kernel uses a static key to switch between the two modes.
In theory, the static key is enabled by default (run in guest mode) and
should be disabled for bare metal (and in some guests that want native
behavior or paravirt spinlock).
A performance drop is reported when running encode/decode workload and
BenchSEE cache sub-workload.
Bisect points to commit ce0a1b608b ("x86/paravirt: Silence unused
native_pv_lock_init() function warning"). When CONFIG_PARAVIRT_SPINLOCKS is
disabled the virt_spin_lock_key is incorrectly set to true on bare
metal. The qspinlock degenerates to test-and-set spinlock, which decreases
the performance on bare metal.
Set the default value of virt_spin_lock_key to false. If booting in a VM,
enable this key. Later during the VM initialization, if other
high-efficient spinlock is preferred (e.g. paravirt-spinlock), or the user
wants the native qspinlock (via nopvspin boot commandline), the
virt_spin_lock_key is disabled accordingly.
This results in the following decision matrix:
X86_FEATURE_HYPERVISOR Y Y Y N
CONFIG_PARAVIRT_SPINLOCKS Y Y N Y/N
PV spinlock Y N N Y/N
virt_spin_lock_key N Y/N Y N
Fixes: ce0a1b608b ("x86/paravirt: Silence unused native_pv_lock_init() function warning")
Reported-by: Prem Nath Dey <prem.nath.dey@intel.com>
Reported-by: Xiaoping Zhou <xiaoping.zhou@intel.com>
Suggested-by: Dave Hansen <dave.hansen@linux.intel.com>
Suggested-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Suggested-by: Nikolay Borisov <nik.borisov@suse.com>
Signed-off-by: Chen Yu <yu.c.chen@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Nikolay Borisov <nik.borisov@suse.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20240806112207.29792-1-yu.c.chen@intel.com
On a platform using the "Multiprocessor Wakeup Structure"[1] to startup
secondary CPUs the control processor needs to memremap() the physical
address of the MP Wakeup Structure mailbox to the variable
acpi_mp_wake_mailbox, which holds the virtual address of mailbox.
To wake up the AP the control processor writes the APIC ID of AP, the
wakeup vector and the ACPI_MP_WAKE_COMMAND_WAKEUP command into the mailbox.
Current implementation doesn't consider the case which restricts boot time
CPU bringup to 1 with the kernel parameter "maxcpus=1" and brings other
CPUs online later from user space as it sets acpi_mp_wake_mailbox to
read-only after init. So when the first AP is tried to brought online
after init, the attempt to update the variable results in a kernel panic.
The memremap() call that initializes the variable cannot be moved into
acpi_parse_mp_wake() because memremap() is not functional at that point in
the boot process. Also as the APs might never be brought up, keep the
memremap() call in acpi_wakeup_cpu() so that the operation only takes place
when needed.
Fixes: 24dd05da8c ("x86/apic: Mark acpi_mp_wake_* variables as __ro_after_init")
Signed-off-by: Zhiquan Li <zhiquan1.li@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Link: https://lore.kernel.org/all/20240805103531.1230635-1-zhiquan1.li@intel.com
Breno observed panics when using failslab under certain conditions during
runtime:
can not alloc irq_pin_list (-1,0,20)
Kernel panic - not syncing: IO-APIC: failed to add irq-pin. Can not proceed
panic+0x4e9/0x590
mp_irqdomain_alloc+0x9ab/0xa80
irq_domain_alloc_irqs_locked+0x25d/0x8d0
__irq_domain_alloc_irqs+0x80/0x110
mp_map_pin_to_irq+0x645/0x890
acpi_register_gsi_ioapic+0xe6/0x150
hpet_open+0x313/0x480
That's a pointless panic which is a leftover of the historic IO/APIC code
which panic'ed during early boot when the interrupt allocation failed.
The only place which might justify panic is the PIT/HPET timer_check() code
which tries to figure out whether the timer interrupt is delivered through
the IO/APIC. But that code does not require to handle interrupt allocation
failures. If the interrupt cannot be allocated then timer delivery fails
and it either panics due to that or falls back to legacy mode.
Cure this by removing the panic wrapper around __add_pin_to_irq_node() and
making mp_irqdomain_alloc() aware of the failure condition and handle it as
any other failure in this function gracefully.
Reported-by: Breno Leitao <leitao@debian.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Breno Leitao <leitao@debian.org>
Tested-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Link: https://lore.kernel.org/all/ZqfJmUF8sXIyuSHN@gmail.com
Link: https://lore.kernel.org/all/20240802155440.275200843@linutronix.de
A kexec kernel boot failure is sometimes observed on AMD CPUs due to an
unmapped EFI config table array. This can be seen when "nogbpages" is on
the kernel command line, and has been observed as a full BIOS reboot rather
than a successful kexec.
This was also the cause of reported regressions attributed to Commit
7143c5f4cf ("x86/mm/ident_map: Use gbpages only where full GB page should
be mapped.") which was subsequently reverted.
To avoid this page fault, explicitly include the EFI config table array in
the kexec identity map.
Further explanation:
The following 2 commits caused the EFI config table array to be
accessed when enabling sev at kernel startup.
commit ec1c66af3a ("x86/compressed/64: Detect/setup SEV/SME features
earlier during boot")
commit c01fce9cef ("x86/compressed: Add SEV-SNP feature
detection/setup")
This is in the code that examines whether SEV should be enabled or not, so
it can even affect systems that are not SEV capable.
This may result in a page fault if the EFI config table array's address is
unmapped. Since the page fault occurs before the new kernel establishes its
own identity map and page fault routines, it is unrecoverable and kexec
fails.
Most often, this problem is not seen because the EFI config table array
gets included in the map by the luck of being placed at a memory address
close enough to other memory areas that *are* included in the map created
by kexec.
Both the "nogbpages" command line option and the "use gpbages only where
full GB page should be mapped" change greatly reduce the chance of being
included in the map by luck, which is why the problem appears.
Signed-off-by: Tao Liu <ltao@redhat.com>
Signed-off-by: Steve Wahl <steve.wahl@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Pavin Joseph <me@pavinjoseph.com>
Tested-by: Sarah Brofeldt <srhb@dbc.dk>
Tested-by: Eric Hagberg <ehagberg@gmail.com>
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/all/20240717213121.3064030-2-steve.wahl@hpe.com
A Linux guest on Hyper-V gets the TSC frequency from a synthetic MSR, if
available. In this case, set X86_FEATURE_TSC_KNOWN_FREQ so that Linux
doesn't unnecessarily do refined TSC calibration when setting up the TSC
clocksource.
With this change, a message such as this is no longer output during boot
when the TSC is used as the clocksource:
[ 1.115141] tsc: Refined TSC clocksource calibration: 2918.408 MHz
Furthermore, the guest and host will have exactly the same view of the
TSC frequency, which is important for features such as the TSC deadline
timer that are emulated by the Hyper-V host.
Signed-off-by: Michael Kelley <mhklinux@outlook.com>
Reviewed-by: Roman Kisel <romank@linux.microsoft.com>
Link: https://lore.kernel.org/r/20240606025559.1631-1-mhklinux@outlook.com
Signed-off-by: Wei Liu <wei.liu@kernel.org>
Message-ID: <20240606025559.1631-1-mhklinux@outlook.com>
The unsynchronized_tsc() eventually checks num_possible_cpus(), and if the
system is non-Intel and the number of possible CPUs is greater than one,
assumes that TSCs are unsynchronized. This despite the comment saying
"assume multi socket systems are not synchronized", that is, socket rather
than CPU. This behavior was preserved by commit 8fbbc4b45c ("x86: merge
tsc_init and clocksource code") and by the previous relevant commit
7e69f2b1ea ("clocksource: Remove the update callback").
The clocksource drivers were added by commit 5d0cf410e9 ("Time: i386
Clocksource Drivers") back in 2006, and the comment still said "socket"
rather than "CPU".
Therefore, bravely (and perhaps foolishly) make the code match the
comment.
Note that it is possible to bypass both code and comment by booting
with tsc=reliable, but this also disables the clocksource watchdog,
which is undesirable when trust in the TSC is strictly limited.
Reported-by: Zhengxu Chen <zhxchen17@meta.com>
Reported-by: Danielle Costantino <dcostantino@meta.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240802154618.4149953-5-paulmck@kernel.org
According to the data sheet, writing the MODE register should stop the
counter (and thus the interrupts). This appears to work on real hardware,
at least modern Intel and AMD systems. It should also work on Hyper-V.
However, on some buggy virtual machines the mode change doesn't have any
effect until the counter is subsequently loaded (or perhaps when the IRQ
next fires).
So, set MODE 0 and then load the counter, to ensure that those buggy VMs
do the right thing and the interrupts stop. And then write MODE 0 *again*
to stop the counter on compliant implementations too.
Apparently, Hyper-V keeps firing the IRQ *repeatedly* even in mode zero
when it should only happen once, but the second MODE write stops that too.
Userspace test program (mostly written by tglx):
=====
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <stdint.h>
#include <sys/io.h>
static __always_inline void __out##bwl(type value, uint16_t port) \
{ \
asm volatile("out" #bwl " %" #bw "0, %w1" \
: : "a"(value), "Nd"(port)); \
} \
\
static __always_inline type __in##bwl(uint16_t port) \
{ \
type value; \
asm volatile("in" #bwl " %w1, %" #bw "0" \
: "=a"(value) : "Nd"(port)); \
return value; \
}
BUILDIO(b, b, uint8_t)
#define inb __inb
#define outb __outb
#define PIT_MODE 0x43
#define PIT_CH0 0x40
#define PIT_CH2 0x42
static int is8254;
static void dump_pit(void)
{
if (is8254) {
// Latch and output counter and status
outb(0xC2, PIT_MODE);
printf("%02x %02x %02x\n", inb(PIT_CH0), inb(PIT_CH0), inb(PIT_CH0));
} else {
// Latch and output counter
outb(0x0, PIT_MODE);
printf("%02x %02x\n", inb(PIT_CH0), inb(PIT_CH0));
}
}
int main(int argc, char* argv[])
{
int nr_counts = 2;
if (argc > 1)
nr_counts = atoi(argv[1]);
if (argc > 2)
is8254 = 1;
if (ioperm(0x40, 4, 1) != 0)
return 1;
dump_pit();
printf("Set oneshot\n");
outb(0x38, PIT_MODE);
outb(0x00, PIT_CH0);
outb(0x0F, PIT_CH0);
dump_pit();
usleep(1000);
dump_pit();
printf("Set periodic\n");
outb(0x34, PIT_MODE);
outb(0x00, PIT_CH0);
outb(0x0F, PIT_CH0);
dump_pit();
usleep(1000);
dump_pit();
dump_pit();
usleep(100000);
dump_pit();
usleep(100000);
dump_pit();
printf("Set stop (%d counter writes)\n", nr_counts);
outb(0x30, PIT_MODE);
while (nr_counts--)
outb(0xFF, PIT_CH0);
dump_pit();
usleep(100000);
dump_pit();
usleep(100000);
dump_pit();
printf("Set MODE 0\n");
outb(0x30, PIT_MODE);
dump_pit();
usleep(100000);
dump_pit();
usleep(100000);
dump_pit();
return 0;
}
=====
Suggested-by: Sean Christopherson <seanjc@google.com>
Co-developed-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: Li RongQing <lirongqing@baidu.com>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Michael Kelley <mhkelley@outlook.com>
Link: https://lore.kernel.org/all/20240802135555.564941-2-dwmw2@infradead.org
Leaving the PIT interrupt running can cause noticeable steal time for
virtual guests. The VMM generally has a timer which toggles the IRQ input
to the PIC and I/O APIC, which takes CPU time away from the guest. Even
on real hardware, running the counter may use power needlessly (albeit
not much).
Make sure it's turned off if it isn't going to be used.
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Michael Kelley <mhkelley@outlook.com>
Link: https://lore.kernel.org/all/20240802135555.564941-1-dwmw2@infradead.org
A process can disable access to the alternate signal stack by not
enabling the altstack's PKEY in the PKRU register.
Nevertheless, the kernel updates the PKRU temporarily for signal
handling. However, in sigreturn(), restore_sigcontext() will restore the
PKRU to the user-defined PKRU value.
This will cause restore_altstack() to fail with a SIGSEGV as it needs read
access to the altstack which is prohibited by the user-defined PKRU value.
Fix this by restoring altstack before restoring PKRU.
Signed-off-by: Aruna Ramakrishna <aruna.ramakrishna@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240802061318.2140081-5-aruna.ramakrishna@oracle.com
If the alternate signal stack is protected by a different PKEY than the
current execution stack, copying XSAVE data to the sigaltstack will fail
if its PKEY is not enabled in the PKRU register.
It's unknown which pkey was used by the application for the altstack, so
enable all PKEYS before XSAVE.
But this updated PKRU value is also pushed onto the sigframe, which
means the register value restored from sigcontext will be different from
the user-defined one, which is incorrect.
Fix that by overwriting the PKRU value on the sigframe with the original,
user-defined PKRU.
Signed-off-by: Aruna Ramakrishna <aruna.ramakrishna@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240802061318.2140081-4-aruna.ramakrishna@oracle.com
In the case where a user thread sets up an alternate signal stack protected
by the default PKEY (i.e. PKEY 0), while the thread's stack is protected by
a non-zero PKEY, both these PKEYS have to be enabled in the PKRU register
for the signal to be delivered to the application correctly. However, the
PKRU value restored after handling the signal must not enable this extra
PKEY (i.e. PKEY 0) - i.e., the PKRU value in the sigframe has to be
overwritten with the user-defined value.
Add helper functions that will update PKRU value in the sigframe after
XSAVE.
Note that sig_prepare_pkru() makes no assumption about which PKEY could
be used to protect the altstack (i.e. it may not be part of init_pkru),
and so enables all PKEYS.
No functional change.
Signed-off-by: Aruna Ramakrishna <aruna.ramakrishna@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240802061318.2140081-3-aruna.ramakrishna@oracle.com
Assume there's a multithreaded application that runs untrusted user
code. Each thread has its stack/code protected by a non-zero PKEY, and the
PKRU register is set up such that only that particular non-zero PKEY is
enabled. Each thread also sets up an alternate signal stack to handle
signals, which is protected by PKEY zero. The PKEYs man page documents that
the PKRU will be reset to init_pkru when the signal handler is invoked,
which means that PKEY zero access will be enabled. But this reset happens
after the kernel attempts to push fpu state to the alternate stack, which
is not (yet) accessible by the kernel, which leads to a new SIGSEGV being
sent to the application, terminating it.
Enabling both the non-zero PKEY (for the thread) and PKEY zero in
userspace will not work for this use case. It cannot have the alt stack
writeable by all - the rationale here is that the code running in that
thread (using a non-zero PKEY) is untrusted and should not have access
to the alternate signal stack (that uses PKEY zero), to prevent the
return address of a function from being changed. The expectation is that
kernel should be able to set up the alternate signal stack and deliver
the signal to the application even if PKEY zero is explicitly disabled
by the application. The signal handler accessibility should not be
dictated by whatever PKRU value the thread sets up.
The PKRU register is managed by XSAVE, which means the sigframe contents
must match the register contents - which is not the case here. It's
required that the signal frame contains the user-defined PKRU value (so
that it is restored correctly from sigcontext) but the actual register must
be reset to init_pkru so that the alt stack is accessible and the signal
can be delivered to the application. It seems that the proper fix here
would be to remove PKRU from the XSAVE framework and manage it separately,
which is quite complicated. As a workaround, do this:
orig_pkru = rdpkru();
wrpkru(orig_pkru & init_pkru_value);
xsave_to_user_sigframe();
put_user(pkru_sigframe_addr, orig_pkru)
In preparation for writing PKRU to sigframe, pass PKRU as an additional
parameter down the call chain from get_sigframe().
No functional change.
Signed-off-by: Aruna Ramakrishna <aruna.ramakrishna@oracle.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/all/20240802061318.2140081-2-aruna.ramakrishna@oracle.com
