Pull x86 resource control updates from Borislav Petkov:
- Extend resctrl with the capability of total memory bandwidth
monitoring, thus accomodating systems which support only total but
not local memory bandwidth monitoring. Add the respective new mount
options
- The usual cleanups
* tag 'x86_cache_for_v6.14_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/resctrl: Document the new "mba_MBps_event" file
x86/resctrl: Add write option to "mba_MBps_event" file
x86/resctrl: Add "mba_MBps_event" file to CTRL_MON directories
x86/resctrl: Make mba_sc use total bandwidth if local is not supported
x86/resctrl: Compute memory bandwidth for all supported events
x86/resctrl: Modify update_mba_bw() to use per CTRL_MON group event
x86/resctrl: Prepare for per-CTRL_MON group mba_MBps control
x86/resctrl: Introduce resctrl_file_fflags_init() to initialize fflags
x86/resctrl: Use kthread_run_on_cpu()
Pull x86 CPU speculation update from Borislav Petkov:
- Add support for AMD hardware which is not affected by SRSO on the
user/kernel attack vector and advertise it to guest userspace
* tag 'x86_bugs_for_v6.14_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
KVM: x86: Advertise SRSO_USER_KERNEL_NO to userspace
x86/bugs: Add SRSO_USER_KERNEL_NO support
Pull x86 RAS updates from Borislav Petkov:
- Remove the shared threshold bank hack on AMD and streamline and
simplify it
- Cleanup and sanitize MCA code
* tag 'ras_core_for_v6.14_rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/mce/amd: Remove shared threshold bank plumbing
x86/mce: Remove the redundant mce_hygon_feature_init()
x86/mce: Convert family/model mixed checks to VFM-based checks
x86/mce: Break up __mcheck_cpu_apply_quirks()
x86/mce: Make four functions return bool
x86/mce/threshold: Remove the redundant this_cpu_dec_return()
x86/mce: Make several functions return bool
Instead of marking individual interrupts as safe to be migrated in
arbitrary contexts, mark the interrupt chips, which require the interrupt
to be moved in actual interrupt context, with the new IRQCHIP_MOVE_DEFERRED
flag. This makes more sense because this is a per interrupt chip property
and not restricted to individual interrupts.
That flips the logic from the historical opt-out to a opt-in model. This is
simpler to handle for other architectures, which default to unrestricted
affinity setting. It also allows to cleanup the redundant core logic
significantly.
All interrupt chips, which belong to a top-level domain sitting directly on
top of the x86 vector domain are marked accordingly, unless the related
setup code marks the interrupts with IRQ_MOVE_PCNTXT, i.e. XEN.
No functional change intended.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Steve Wahl <steve.wahl@hpe.com>
Acked-by: Wei Liu <wei.liu@kernel.org>
Link: https://lore.kernel.org/all/20241210103335.563277044@linutronix.de
The FRED RSP0 MSR is only used for delivering events when running
userspace. Linux leverages this property to reduce expensive MSR
writes and optimize context switches. The kernel only writes the
MSR when about to run userspace *and* when the MSR has actually
changed since the last time userspace ran.
This optimization is implemented by maintaining a per-CPU cache of
FRED RSP0 and then checking that against the value for the top of
current task stack before running userspace.
However cpu_init_fred_exceptions() writes the MSR without updating
the per-CPU cache. This means that the kernel might return to
userspace with MSR_IA32_FRED_RSP0==0 when it needed to point to the
top of current task stack. This would induce a double fault (#DF),
which is bad.
A context switch after cpu_init_fred_exceptions() can paper over
the issue since it updates the cached value. That evidently
happens most of the time explaining how this bug got through.
Fix the bug through resynchronizing the FRED RSP0 MSR with its
per-CPU cache in cpu_init_fred_exceptions().
Fixes: fe85ee3919 ("x86/entry: Set FRED RSP0 on return to userspace instead of context switch")
Signed-off-by: Xin Li (Intel) <xin@zytor.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Acked-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc:stable@vger.kernel.org
Link: https://lore.kernel.org/all/20250110174639.1250829-1-xin%40zytor.com
Both i386 and x86_64 now copy the relocate_kernel() function into the control
page and execute it from there, using an open-coded function pointer.
Use a typedef for it instead.
[ bp: Put relocate_kernel_ptr ptr arithmetic on a single line. ]
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250109140757.2841269-10-dwmw2@infradead.org
A few places in the kexec control code page make the assumption that the first
instruction of relocate_kernel is at the very start of the page.
To allow for Clang CFI information to be added to relocate_kernel(), as well
as the general principle of removing unwarranted assumptions, fix them to use
the external __relocate_kernel_start symbol that the linker adds. This means
using a separate addq and subq for calculating offsets, as the assembler can
no longer calculate the delta directly for itself and relocations aren't that
versatile. But those values can at least be used relative to a local label to
avoid absolute relocations.
Turn the jump from relocate_kernel() to identity_mapped() into a real indirect
'jmp *%rsi' too, while touching it. There was no real reason for it to be
a push+ret in the first place, and adding Clang CFI info will also give
objtool enough visibility to start complaining 'return with modified stack
frame' about it.
[ bp: Massage commit message. ]
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250109140757.2841269-9-dwmw2@infradead.org
A recent commit caused the relocate_kernel() function to be invoked through
a function pointer, but it does not have CFI information. The resulting trap
occurs after the IDT and GDT have been invalidated, leading to a triple-fault
if CONFIG_CFI_CLANG is enabled.
Using SYM_TYPED_FUNC_START() to provide the CFI information looks like it will
require a prolonged battle with objtool. And is fairly pointless anyway, as
the actual signature comes from a __kcfi_typeid_… symbol emitted from the
C code based on the function prototype it thinks that relocate_kernel has,
rendering the check somewhat tautological.
The simple fix is just to mark machine_kexec() with __nocfi.
Fixes: eeebbde571 ("x86/kexec: Invoke copy of relocate_kernel() instead of the original")
Reported-by: Nathan Chancellor <nathan@kernel.org>
Suggested-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250109140757.2841269-7-dwmw2@infradead.org
After commit
cb33ff9e06 ("x86/kexec: Move relocate_kernel to kernel .data section"),
kernels configured with an option that uses -ffunction-sections, such as
CONFIG_LTO_CLANG, crash when kexecing because the value of relocate_kernel
does not match the value of __relocate_kernel_start so incorrect code gets
copied via machine_kexec_prepare().
$ llvm-nm good-vmlinux &| rg relocate_kernel
ffffffff83280d41 T __relocate_kernel_end
ffffffff83280b00 T __relocate_kernel_start
ffffffff83280b00 T relocate_kernel
$ llvm-nm bad-vmlinux &| rg relocate_kernel
ffffffff83266100 D __relocate_kernel_end
ffffffff83266100 D __relocate_kernel_start
ffffffff8120b0d8 T relocate_kernel
When -ffunction-sections is enabled, TEXT_MAIN matches on
'.text.[0-9a-zA-Z_]*' to coalesce the function specific functions back
into .text during link time after they have been optimized. Due to the
placement of TEXT_TEXT before KEXEC_RELOCATE_KERNEL in the x86 linker
script, the .text.relocate_kernel section ends up in .text instead of
.data.
Use a second dot in the relocate_kernel section name to avoid matching
on TEXT_MAIN, which matches a similar situation that happened in
commit
79cd2a1122 ("x86/retpoline,kprobes: Fix position of thunk sections with CONFIG_LTO_CLANG"),
which allows kexec to function properly.
While .data.relocate_kernel still ends up in the .data section via
DATA_MAIN -> DATA_DATA, ensure it is located with the
.text.relocate_kernel section as intended by performing the same
transformation.
Fixes: cb33ff9e06 ("x86/kexec: Move relocate_kernel to kernel .data section")
Fixes: 8dbec5c77b ("x86/kexec: Add data section to relocate_kernel")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250109140757.2841269-6-dwmw2@infradead.org
A ::preserve_context kimage can be invoked more than once, and the entry point
can be different every time. When the callee returns to the kernel, it leaves
the address of its entry point for next time on the stack.
That being the case, one might reasonably assume that the caller would
allocate space for it on the stack frame before actually performing the 'call'
into the callee.
Apparently not, though. Ever since the kjump code was first added in 2009, it
has set up a *new* stack at the top of the swap_page scratch page, then just
performed the 'call' without allocating any space for the re-entry address to
be returned. It then reads the re-entry point for next time from 0(%rsp) which
is actually the first qword of the page *after* the swap page, which might not
exist at all! And if the callee has written to that, then it will have
corrupted memory it doesn't own.
Correct this by pushing the entry point of the callee onto the stack before
calling it. The callee may then adjust it, or not, as it sees fit, and
subsequent invocations should work correctly either way.
Remove a stray push of zero to the *relocate_kernel* stack, which may have
been intended for this purpose, but which was actually just noise.
Also, loading the stack for the callee relied on the address of the swap page
being in %r10 without ever documenting that fact. Recent code changes made
that no longer true, so load it directly from the local kexec_pa_swap_page
variable instead.
Fixes: b3adabae8a ("x86/kexec: Drop page_list argument from relocate_kernel()")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250109140757.2841269-5-dwmw2@infradead.org
The swap_pages function expects the swap page to be in %r10, but there
was no documentation to that effect. Once upon a time the setup code
used to load its value from a kernel virtual address and save it to an
address which is accessible in the identity-mapped page tables, and
*happened* to use %r10 to do so, with no comment that it was left there
on *purpose* instead of just being a scratch register. Once that was no
longer necessary, %r10 just holds whatever the kernel happened to leave
in it.
Now that the original value passed by the kernel is accessible via
%rip-relative addressing, load directly from there instead of using %r10
for it. But document the other parameters that the swap_pages function
*does* expect in registers.
Fixes: b3adabae8a ("x86/kexec: Drop page_list argument from relocate_kernel()")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250109140757.2841269-4-dwmw2@infradead.org
The swap_pages() function will only actually *swap*, as its name implies, if
the preserve_context flag in the %r11 register is non-zero. On the way back
from a ::preserve_context kexec, ensure that the %r11 register is non-zero so
that the pages get swapped back.
Fixes: 9e5683e2d0 ("x86/kexec: Only swap pages for ::preserve_context mode")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250109140757.2841269-3-dwmw2@infradead.org
The kernel switches to a new set of page tables during kexec. The global
mappings (_PAGE_GLOBAL==1) can remain in the TLB after this switch. This
is generally not a problem because the new page tables use a different
portion of the virtual address space than the normal kernel mappings.
The critical exception to that generalisation (and the only mapping
which isn't an identity mapping) is the kexec control page itself —
which was ROX in the original kernel mapping, but should be RWX in the
new page tables. If there is a global TLB entry for that in its prior
read-only state, it definitely needs to be flushed before attempting to
write through that virtual mapping.
It would be possible to just avoid writing to the virtual address of the
page and defer all writes until they can be done through the identity
mapping. But there's no good reason to keep the old TLB entries around,
as they can cause nothing but trouble.
Clear the PGE bit in %cr4 early, before storing data in the control page.
Fixes: 5a82223e07 ("x86/kexec: Mark relocate_kernel page as ROX instead of RWX")
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219592
Reported-by: Nathan Chancellor <nathan@kernel.org>
Reported-by: "Ning, Hongyu" <hongyu.ning@linux.intel.com>
Co-developed-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: "Ning, Hongyu" <hongyu.ning@linux.intel.com>
Link: https://lore.kernel.org/r/20250109140757.2841269-2-dwmw2@infradead.org
Now, if CONFIG_MMU_GATHER_RCU_TABLE_FREE is selected, the page table pages
will be freed by semi RCU, that is:
- batch table freeing: asynchronous free by RCU
- single table freeing: IPI + synchronous free
In this way, the page table can be lockless traversed by disabling IRQ in
paths such as fast GUP. But this is not enough to free the empty PTE page
table pages in paths other that munmap and exit_mmap path, because IPI
cannot be synchronized with rcu_read_lock() in pte_offset_map{_lock}().
In preparation for supporting empty PTE page table pages reclaimation, let
single table also be freed by RCU like batch table freeing. Then we can
also use pte_offset_map() etc to prevent PTE page from being freed.
Like pte_free_defer(), we can also safely use ptdesc->pt_rcu_head to free
the page table pages:
- The pt_rcu_head is unioned with pt_list and pmd_huge_pte.
- For pt_list, it is used to manage the PGD page in x86. Fortunately
tlb_remove_table() will not be used for free PGD pages, so it is safe
to use pt_rcu_head.
- For pmd_huge_pte, it is used for THPs, so it is safe.
After applying this patch, if CONFIG_PT_RECLAIM is enabled, the function
call of free_pte() is as follows:
free_pte
pte_free_tlb
__pte_free_tlb
___pte_free_tlb
paravirt_tlb_remove_table
tlb_remove_table [!CONFIG_PARAVIRT, Xen PV, Hyper-V, KVM]
[no-free-memory slowpath:]
tlb_table_invalidate
tlb_remove_table_one
__tlb_remove_table_one [frees via RCU]
[fastpath:]
tlb_table_flush
tlb_remove_table_free [frees via RCU]
native_tlb_remove_table [CONFIG_PARAVIRT on native]
tlb_remove_table [see above]
Link: https://lkml.kernel.org/r/0287d442a973150b0e1019cc406e6322d148277a.1733305182.git.zhengqi.arch@bytedance.com
Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jann Horn <jannh@google.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Muchun Song <muchun.song@linux.dev>
Cc: Peter Xu <peterx@redhat.com>
Cc: Will Deacon <will@kernel.org>
Cc: Zach O'Keefe <zokeefe@google.com>
Cc: Dan Carpenter <dan.carpenter@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
x86_sched_itmt_flags() returns SD_ASYM_PACKING if ITMT support is
enabled by the system. Without ITMT support being enabled, it returns 0
similar to current x86_die_flags() on non-Hybrid systems
(!X86_HYBRID_CPU and !X86_FEATURE_AMD_HETEROGENEOUS_CORES)
On Intel systems that enable ITMT support, either the MC domain
coincides with the PKG domain, or in case of multiple MC groups
within a PKG domain, either Sub-NUMA Cluster (SNC) is enabled or the
processor features Hybrid core layout (X86_HYBRID_CPU) which leads to
three distinct possibilities:
o If PKG and MC domains coincide, PKG domain is degenerated by
sd_parent_degenerate() when building sched domain topology.
o If SNC is enabled, PKG domain is never added since
"x86_has_numa_in_package" is set and the topology will instead contain
NODE and NUMA domains.
o On X86_HYBRID_CPU which contains multiple MC groups within the PKG,
the PKG domain requires x86_sched_itmt_flags().
Thus, on Intel systems that contains multiple MC groups within the PKG
and enables ITMT support, the PKG domain requires
x86_sched_itmt_flags(). In all other cases PKG domain is either never
added or is degenerated. Thus, returning x86_sched_itmt_flags()
unconditionally at PKG domain on Intel systems should not lead to any
functional changes.
On AMD systems with multiple LLCs (MC groups) within a PKG domain,
enabling ITMT support requires setting SD_ASYM_PACKING to the PKG domain
since the core rankings are assigned PKG-wide.
Core rankings on AMD processors is currently set by the amd-pstate
driver when Preferred Core feature is supported. A subset of systems that
support Preferred Core feature can be detected using
X86_FEATURE_AMD_HETEROGENEOUS_CORES however, this does not cover all the
systems that support Preferred Core ranking.
Detecting Preferred Core support on AMD systems requires inspecting CPPC
Highest Perf on all present CPUs and checking if it differs on at least
one CPU. Previous suggestion to use a synthetic feature to detect
Preferred Core support [1] was found to be non-trivial to implement
since BSP alone cannot detect if Preferred Core is supported and by the
time AP comes up, alternatives are patched and setting a X86_FEATURE_*
then is not possible.
Since x86 processors enabling ITMT support that consists multiple
non-NUMA MC groups within a PKG requires SD_ASYM_PACKING flag set at the
PKG domain, return x86_sched_itmt_flags unconditionally for the PKG
domain.
Since x86_die_flags() would have just returned x86_sched_itmt_flags()
after the change, remove the unnecessary wrapper and pass
x86_sched_itmt_flags() directly as the flags function.
Fixes: f3a0523918 ("cpufreq: amd-pstate: Enable amd-pstate preferred core support")
Signed-off-by: K Prateek Nayak <kprateek.nayak@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com>
Link: https://lore.kernel.org/r/20241223043407.1611-6-kprateek.nayak@amd.com
x86_*_flags() wrappers were introduced with commit d3d37d850d
("x86/sched: Add SD_ASYM_PACKING flags to x86 ITMT CPU") to add
x86_sched_itmt_flags() in addition to the default domain flags for SMT
and MC domain.
commit 995998ebde ("x86/sched: Remove SD_ASYM_PACKING from the
SMT domain flags") removed the ITMT flags for SMT domain but not the
x86_smt_flags() wrappers which directly returns cpu_smt_flags().
Remove x86_smt_flags() and directly use cpu_smt_flags() to derive the
flags for SMT domain. No functional changes intended.
Signed-off-by: K Prateek Nayak <kprateek.nayak@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com>
Link: https://lore.kernel.org/r/20241223043407.1611-5-kprateek.nayak@amd.com
"sched_itmt_enabled" was only introduced as a debug toggle for any funky
ITMT behavior. Move the sysctl controlled from
"/proc/sys/kernel/sched_itmt_enabled" to debugfs at
"/sys/kernel/debug/x86/sched_itmt_enabled" with a notable change that a
cat on the file will return "Y" or "N" instead of "1" or "0" to
indicate that feature is enabled or disabled respectively. Either "0" or
"N" (or any string that kstrtobool() interprets as false) can be written
to the file will disable the feature, and writing either "1" or "Y" (or
any string that kstrtobool() interprets as true) will enable it back
when the platform supports ITMT ranking.
Since ITMT is x86 specific (and PowerPC uses SD_ASYM_PACKING too), the
toggle was moved to "/sys/kernel/debug/x86/" as opposed to
"/sys/kernel/debug/sched/"
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: K Prateek Nayak <kprateek.nayak@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com>
Link: https://lore.kernel.org/r/20241223043407.1611-4-kprateek.nayak@amd.com
In preparation to move "sysctl_sched_itmt_enabled" to debugfs, convert
the unsigned int to bool since debugfs readily exposes boolean fops
primitives (debugfs_read_file_bool, debugfs_write_file_bool) which can
streamline the conversion.
Since the current ctl_table initializes extra1 and extra2 to SYSCTL_ZERO
and SYSCTL_ONE respectively, the value of "sysctl_sched_itmt_enabled"
can only be 0 or 1 and this datatype conversion should not cause any
functional changes.
Signed-off-by: K Prateek Nayak <kprateek.nayak@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Tim Chen <tim.c.chen@linux.intel.com>
Link: https://lore.kernel.org/r/20241223043407.1611-2-kprateek.nayak@amd.com
There are more than one SMN index/data pair available for software use.
The register offsets are different, but the protocol is the same.
Use defines for the SMN offset values and allow the index/data offsets
to be passed to the read/write helper function.
This eases code reuse with other SMN users in the kernel.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20241206161210.163701-14-yazen.ghannam@amd.com
Currently, the total AMD node count is determined by searching and counting
CPU/node devices using PCI IDs.
However, AMD node information is already available through topology
CPUID/MSRs. The recent topology rework has made this info easier to access.
Replace the node counting code with a simple product of topology info.
Every node/northbridge is expected to have a 'misc' device. Clear everything
out if a 'misc' device isn't found on a node.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20250107222847.3300430-7-yazen.ghannam@amd.com
The "root" device search was introduced to support SMN access for Zen
systems. This device represents a PCIe root complex. It is not the
same as the "CPU/node" devices found at slots 0x18-0x1F.
There may be multiple PCIe root complexes within an AMD node. Such is
the case with server or High-end Desktop (HEDT) systems, etc. Therefore
it is not enough to assume "root <-> AMD node" is a 1-to-1 association.
Currently, this is handled by skipping "extra" root complexes during the
search. However, the hardware provides the PCI bus number of an AMD
node's root device.
Use the hardware info to get the root device's bus and drop the extra
search code and PCI IDs.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20241206161210.163701-7-yazen.ghannam@amd.com
Use the newly added helper function to look up a CPU/Node function to
find "function 4" devices.
Thus, avoid the need to regularly add new PCI IDs for basic discovery.
The unique PCI IDs are still useful in case of quirks or functional
changes. And they should be used only in such a manner.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20241206161210.163701-6-yazen.ghannam@amd.com
The "AMD Node" concept spans many families of systems and applies to
a number of subsystems and drivers.
Currently, the AMD Northbridge code is overloaded with AMD node
functionality. However, the node concept is broader than just
northbridges.
Start files to host common AMD node functions and definitions. Include
a helper to find an AMD node device function based on the convention
described in AMD documentation.
Anything that needs node functionality should include this rather than
amd_nb.h. The AMD_NB code will be reduced to only northbridge-specific
code needed for legacy systems.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20241206161210.163701-5-yazen.ghannam@amd.com
The x86 shadow stack support has its own set of registers. Those registers
are XSAVE-managed, but they are "supervisor state components" which means
that userspace can not touch them with XSAVE/XRSTOR. It also means that
they are not accessible from the existing ptrace ABI for XSAVE state.
Thus, there is a new ptrace get/set interface for it.
The regset code that ptrace uses provides an ->active() handler in
addition to the get/set ones. For shadow stack this ->active() handler
verifies that shadow stack is enabled via the ARCH_SHSTK_SHSTK bit in the
thread struct. The ->active() handler is checked from some call sites of
the regset get/set handlers, but not the ptrace ones. This was not
understood when shadow stack support was put in place.
As a result, both the set/get handlers can be called with
XFEATURE_CET_USER in its init state, which would cause get_xsave_addr() to
return NULL and trigger a WARN_ON(). The ssp_set() handler luckily has an
ssp_active() check to avoid surprising the kernel with shadow stack
behavior when the kernel is not ready for it (ARCH_SHSTK_SHSTK==0). That
check just happened to avoid the warning.
But the ->get() side wasn't so lucky. It can be called with shadow stacks
disabled, triggering the warning in practice, as reported by Christina
Schimpe:
WARNING: CPU: 5 PID: 1773 at arch/x86/kernel/fpu/regset.c:198 ssp_get+0x89/0xa0
[...]
Call Trace:
<TASK>
? show_regs+0x6e/0x80
? ssp_get+0x89/0xa0
? __warn+0x91/0x150
? ssp_get+0x89/0xa0
? report_bug+0x19d/0x1b0
? handle_bug+0x46/0x80
? exc_invalid_op+0x1d/0x80
? asm_exc_invalid_op+0x1f/0x30
? __pfx_ssp_get+0x10/0x10
? ssp_get+0x89/0xa0
? ssp_get+0x52/0xa0
__regset_get+0xad/0xf0
copy_regset_to_user+0x52/0xc0
ptrace_regset+0x119/0x140
ptrace_request+0x13c/0x850
? wait_task_inactive+0x142/0x1d0
? do_syscall_64+0x6d/0x90
arch_ptrace+0x102/0x300
[...]
Ensure that shadow stacks are active in a thread before looking them up
in the XSAVE buffer. Since ARCH_SHSTK_SHSTK and user_ssp[SHSTK_EN] are
set at the same time, the active check ensures that there will be
something to find in the XSAVE buffer.
[ dhansen: changelog/subject tweaks ]
Fixes: 2fab02b25a ("x86: Add PTRACE interface for shadow stack")
Reported-by: Christina Schimpe <christina.schimpe@intel.com>
Signed-off-by: Rick Edgecombe <rick.p.edgecombe@intel.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Tested-by: Christina Schimpe <christina.schimpe@intel.com>
Cc:stable@vger.kernel.org
Link: https://lore.kernel.org/all/20250107233056.235536-1-rick.p.edgecombe%40intel.com
Legacy AMD systems include an integrated Northbridge that is represented
by MCA bank 4. This is the only non-core MCA bank in legacy systems. The
Northbridge is physically shared by all the CPUs within an AMD "Node".
However, in practice the "shared" MCA bank can only by managed by a
single CPU within that AMD Node. This is known as the "Node Base Core"
(NBC). For example, only the NBC will be able to read the MCA bank 4
registers; they will be Read-as-Zero for other CPUs. Also, the MCA
Thresholding interrupt will only signal the NBC; the other CPUs will not
receive it. This is enforced by hardware, and it should not be managed by
software.
The current AMD Thresholding code attempts to deal with the "shared" MCA
bank by micromanaging the bank's sysfs kobjects. However, this does not
follow the intended kobject use cases. It is also fragile, and it has
caused bugs in the past.
Modern AMD systems do not need this shared MCA bank support, and it
should not be needed on legacy systems either.
Remove the shared threshold bank code. Also, move the threshold struct
definitions to mce/amd.c, since they are no longer needed in amd_nb.c.
Signed-off-by: Yazen Ghannam <yazen.ghannam@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20241206161210.163701-2-yazen.ghannam@amd.com
__static_call_update_early() has a check for early_boot_irqs_disabled, but
is used before early_boot_irqs_disabled is set up in start_kernel().
Xen PV has always special cased early_boot_irqs_disabled, but Xen PVH does
not and falls over the BUG when booting as dom0.
It is very suspect that early_boot_irqs_disabled starts as 0, becomes 1 for
a time, then becomes 0 again, but as this needs backporting to fix a
breakage in a security fix, dropping the BUG_ON() is the far safer option.
Fixes: 0ef8047b73 ("x86/static-call: provide a way to do very early static-call updates")
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=219620
Reported-by: Alex Zenla <alex@edera.dev>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Juergen Gross <jgross@suse.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Alex Zenla <alex@edera.dev>
Link: https://lore.kernel.org/r/20241221211046.6475-1-andrew.cooper3@citrix.com
Instead of open-coding the check for size/data move it inside the
function and make it return a boolean indicating whether data was found
or not.
No functional changes.
[ bp: Write @ret in find_blobs_in_containers() only on success. ]
Signed-off-by: Nikolay Borisov <nik.borisov@suse.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20241018155151.702350-2-nik.borisov@suse.com
The 'storm' variable points to this_cpu_ptr(&storm_desc). Access the
'stormy_bank_count' field through the 'storm' to avoid calling
this_cpu_*() on the same per-CPU variable twice.
This minor optimization reduces the text size by 16 bytes.
$ size threshold.o.*
text data bss dec hex filename
1395 1664 0 3059 bf3 threshold.o.old
1379 1664 0 3043 be3 threshold.o.new
No functional changes intended.
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Tony Luck <tony.luck@intel.com>
Reviewed-by: Nikolay Borisov <nik.borisov@suse.com>
Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>
Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>
Link: https://lore.kernel.org/r/20241212140103.66964-3-qiuxu.zhuo@intel.com
