Previously, error_entry() and paranoid_entry() saved the GP registers
onto stack space previously allocated by its callers. Combine these two
steps in the callers, and use the generic PUSH_AND_CLEAR_REGS macro
for that.
This adds a significant amount ot text size. However, Ingo Molnar points
out that:
"these numbers also _very_ significantly over-represent the
extra footprint. The assumptions that resulted in
us compressing the IRQ entry code have changed very
significantly with the new x86 IRQ allocation code we
introduced in the last year:
- IRQ vectors are usually populated in tightly clustered
groups.
With our new vector allocator code the typical per CPU
allocation percentage on x86 systems is ~3 device vectors
and ~10 fixed vectors out of ~220 vectors - i.e. a very
low ~6% utilization (!). [...]
The days where we allocated a lot of vectors on every
CPU and the compression of the IRQ entry code text
mattered are over.
- Another issue is that only a small minority of vectors
is frequent enough to actually matter to cache utilization
in practice: 3-4 key IPIs and 1-2 device IRQs at most - and
those vectors tend to be tightly clustered as well into about
two groups, and are probably already on 2-3 cache lines in
practice.
For the common case of 'cache cold' IRQs it's the depth of
the call chain and the fragmentation of the resulting I$
that should be the main performance limit - not the overall
size of it.
- The CPU side cost of IRQ delivery is still very expensive
even in the best, most cached case, as in 'over a thousand
cycles'. So much stuff is done that maybe contemporary x86
IRQ entry microcode already prefetches the IDT entry and its
expected call target address."[*]
[*] http://lkml.kernel.org/r/20180208094710.qnjixhm6hybebdv7@gmail.com
The "testb $3, CS(%rsp)" instruction in the idtentry macro does not need
modification. Previously, %rsp was manually decreased by 15*8; with
this patch, %rsp is decreased by 15 pushq instructions.
[jpoimboe@redhat.com: unwind hint improvements]
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dan.j.williams@intel.com
Link: http://lkml.kernel.org/r/20180211104949.12992-7-linux@dominikbrodowski.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Linus reported that GCC-7.3 generated a switch-table construct that
confused objtool. It turns out that, in particular due to KASAN, it is
possible to have unrelated .rodata usage in between the .rodata setup
for the switch-table and the following indirect jump.
The simple linear reverse search from the indirect jump would hit upon
the KASAN .rodata usage first and fail to find a switch_table,
resulting in a spurious 'sibling call with modified stack frame'
warning.
Fix this by creating a 'jump-stack' which we can 'unwind' during
reversal, thereby skipping over much of the in-between code.
This is not fool proof by any means, but is sufficient to make the
known cases work. Future work would be to construct more comprehensive
flow analysis code.
Reported-and-tested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20180208130232.GF25235@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
At entry userspace may have (maliciously) populated the extra registers
outside the syscall calling convention with arbitrary values that could
be useful in a speculative execution (Spectre style) attack.
Clear these registers to minimize the kernel's attack surface.
Note, this only clears the extra registers and not the unused
registers for syscalls less than 6 arguments, since those registers are
likely to be clobbered well before their values could be put to use
under speculation.
Note, Linus found that the XOR instructions can be executed with
minimized cost if interleaved with the PUSH instructions, and Ingo's
analysis found that R10 and R11 should be included in the register
clearing beyond the typical 'extra' syscall calling convention
registers.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Cc: <stable@vger.kernel.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/151787988577.7847.16733592218894189003.stgit@dwillia2-desk3.amr.corp.intel.com
[ Made small improvements to the changelog and the code comments. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The Indirect Branch Predictor Barrier (IBPB) is an indirect branch
control mechanism. It keeps earlier branches from influencing
later ones.
Unlike IBRS and STIBP, IBPB does not define a new mode of operation.
It's a command that ensures predicted branch targets aren't used after
the barrier. Although IBRS and IBPB are enumerated by the same CPUID
enumeration, IBPB is very different.
IBPB helps mitigate against three potential attacks:
* Mitigate guests from being attacked by other guests.
- This is addressed by issing IBPB when we do a guest switch.
* Mitigate attacks from guest/ring3->host/ring3.
These would require a IBPB during context switch in host, or after
VMEXIT. The host process has two ways to mitigate
- Either it can be compiled with retpoline
- If its going through context switch, and has set !dumpable then
there is a IBPB in that path.
(Tim's patch: https://patchwork.kernel.org/patch/10192871)
- The case where after a VMEXIT you return back to Qemu might make
Qemu attackable from guest when Qemu isn't compiled with retpoline.
There are issues reported when doing IBPB on every VMEXIT that resulted
in some tsc calibration woes in guest.
* Mitigate guest/ring0->host/ring0 attacks.
When host kernel is using retpoline it is safe against these attacks.
If host kernel isn't using retpoline we might need to do a IBPB flush on
every VMEXIT.
Even when using retpoline for indirect calls, in certain conditions 'ret'
can use the BTB on Skylake-era CPUs. There are other mitigations
available like RSB stuffing/clearing.
* IBPB is issued only for SVM during svm_free_vcpu().
VMX has a vmclear and SVM doesn't. Follow discussion here:
https://lkml.org/lkml/2018/1/15/146
Please refer to the following spec for more details on the enumeration
and control.
Refer here to get documentation about mitigations.
https://software.intel.com/en-us/side-channel-security-support
[peterz: rebase and changelog rewrite]
[karahmed: - rebase
- vmx: expose PRED_CMD if guest has it in CPUID
- svm: only pass through IBPB if guest has it in CPUID
- vmx: support !cpu_has_vmx_msr_bitmap()]
- vmx: support nested]
[dwmw2: Expose CPUID bit too (AMD IBPB only for now as we lack IBRS)
PRED_CMD is a write-only MSR]
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: KarimAllah Ahmed <karahmed@amazon.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: kvm@vger.kernel.org
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Arjan Van De Ven <arjan.van.de.ven@intel.com>
Cc: Greg KH <gregkh@linuxfoundation.org>
Cc: Jun Nakajima <jun.nakajima@intel.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/1515720739-43819-6-git-send-email-ashok.raj@intel.com
Link: https://lkml.kernel.org/r/1517522386-18410-3-git-send-email-karahmed@amazon.de
I'm seeing build failures from the two newly introduced arrays that
are marked 'const' and '__initdata', which are mutually exclusive:
arch/x86/kernel/cpu/common.c:882:43: error: 'cpu_no_speculation' causes a section type conflict with 'e820_table_firmware_init'
arch/x86/kernel/cpu/common.c:895:43: error: 'cpu_no_meltdown' causes a section type conflict with 'e820_table_firmware_init'
The correct annotation is __initconst.
Fixes: fec9434a12 ("x86/pti: Do not enable PTI on CPUs which are not vulnerable to Meltdown")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Ricardo Neri <ricardo.neri-calderon@linux.intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Thomas Garnier <thgarnie@google.com>
Cc: David Woodhouse <dwmw@amazon.co.uk>
Link: https://lkml.kernel.org/r/20180202213959.611210-1-arnd@arndb.de
Place the MSR bitmap in struct loaded_vmcs, and update it in place
every time the x2apic or APICv state can change. This is rare and
the loop can handle 64 MSRs per iteration, in a similar fashion as
nested_vmx_prepare_msr_bitmap.
This prepares for choosing, on a per-VM basis, whether to intercept
the SPEC_CTRL and PRED_CMD MSRs.
Cc: stable@vger.kernel.org # prereq for Spectre mitigation
Suggested-by: Jim Mattson <jmattson@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Flush indirect branches when switching into a process that marked itself
non dumpable. This protects high value processes like gpg better,
without having too high performance overhead.
If done naïvely, we could switch to a kernel idle thread and then back
to the original process, such as:
process A -> idle -> process A
In such scenario, we do not have to do IBPB here even though the process
is non-dumpable, as we are switching back to the same process after a
hiatus.
To avoid the redundant IBPB, which is expensive, we track the last mm
user context ID. The cost is to have an extra u64 mm context id to track
the last mm we were using before switching to the init_mm used by idle.
Avoiding the extra IBPB is probably worth the extra memory for this
common scenario.
For those cases where tlb_defer_switch_to_init_mm() returns true (non
PCID), lazy tlb will defer switch to init_mm, so we will not be changing
the mm for the process A -> idle -> process A switch. So IBPB will be
skipped for this case.
Thanks to the reviewers and Andy Lutomirski for the suggestion of
using ctx_id which got rid of the problem of mm pointer recycling.
Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: ak@linux.intel.com
Cc: karahmed@amazon.de
Cc: arjan@linux.intel.com
Cc: torvalds@linux-foundation.org
Cc: linux@dominikbrodowski.net
Cc: peterz@infradead.org
Cc: bp@alien8.de
Cc: luto@kernel.org
Cc: pbonzini@redhat.com
Cc: gregkh@linux-foundation.org
Link: https://lkml.kernel.org/r/1517263487-3708-1-git-send-email-dwmw@amazon.co.uk
Despite the fact that all the other code there seems to be doing it, just
using set_cpu_cap() in early_intel_init() doesn't actually work.
For CPUs with PKU support, setup_pku() calls get_cpu_cap() after
c->c_init() has set those feature bits. That resets those bits back to what
was queried from the hardware.
Turning the bits off for bad microcode is easy to fix. That can just use
setup_clear_cpu_cap() to force them off for all CPUs.
I was less keen on forcing the feature bits *on* that way, just in case
of inconsistencies. I appreciate that the kernel is going to get this
utterly wrong if CPU features are not consistent, because it has already
applied alternatives by the time secondary CPUs are brought up.
But at least if setup_force_cpu_cap() isn't being used, we might have a
chance of *detecting* the lack of the corresponding bit and either
panicking or refusing to bring the offending CPU online.
So ensure that the appropriate feature bits are set within get_cpu_cap()
regardless of how many extra times it's called.
Fixes: 2961298e ("x86/cpufeatures: Clean up Spectre v2 related CPUID flags")
Signed-off-by: David Woodhouse <dwmw@amazon.co.uk>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: karahmed@amazon.de
Cc: peterz@infradead.org
Cc: bp@alien8.de
Link: https://lkml.kernel.org/r/1517322623-15261-1-git-send-email-dwmw@amazon.co.uk
For __get_user() paths, do not allow the kernel to speculate on the value
of a user controlled pointer. In addition to the 'stac' instruction for
Supervisor Mode Access Protection (SMAP), a barrier_nospec() causes the
access_ok() result to resolve in the pipeline before the CPU might take any
speculative action on the pointer value. Given the cost of 'stac' the
speculation barrier is placed after 'stac' to hopefully overlap the cost of
disabling SMAP with the cost of flushing the instruction pipeline.
Since __get_user is a major kernel interface that deals with user
controlled pointers, the __uaccess_begin_nospec() mechanism will prevent
speculative execution past an access_ok() permission check. While
speculative execution past access_ok() is not enough to lead to a kernel
memory leak, it is a necessary precondition.
To be clear, __uaccess_begin_nospec() is addressing a class of potential
problems near __get_user() usages.
Note, that while the barrier_nospec() in __uaccess_begin_nospec() is used
to protect __get_user(), pointer masking similar to array_index_nospec()
will be used for get_user() since it incorporates a bounds check near the
usage.
uaccess_try_nospec provides the same mechanism for get_user_try.
No functional changes.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Suggested-by: Andi Kleen <ak@linux.intel.com>
Suggested-by: Ingo Molnar <mingo@redhat.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-arch@vger.kernel.org
Cc: Tom Lendacky <thomas.lendacky@amd.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: kernel-hardening@lists.openwall.com
Cc: gregkh@linuxfoundation.org
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: alan@linux.intel.com
Link: https://lkml.kernel.org/r/151727415922.33451.5796614273104346583.stgit@dwillia2-desk3.amr.corp.intel.com
The TS_COMPAT bit is very hot and is accessed from code paths that mostly
also touch thread_info::flags. Move it into struct thread_info to improve
cache locality.
The only reason it was in thread_struct is that there was a brief period
during which arch-specific fields were not allowed in struct thread_info.
Linus suggested further changing:
ti->status &= ~(TS_COMPAT|TS_I386_REGS_POKED);
to:
if (unlikely(ti->status & (TS_COMPAT|TS_I386_REGS_POKED)))
ti->status &= ~(TS_COMPAT|TS_I386_REGS_POKED);
on the theory that frequently dirtying the cacheline even in pure 64-bit
code that never needs to modify status hurts performance. That could be a
reasonable followup patch, but I suspect it matters less on top of this
patch.
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Kernel Hardening <kernel-hardening@lists.openwall.com>
Link: https://lkml.kernel.org/r/03148bcc1b217100e6e8ecf6a5468c45cf4304b6.1517164461.git.luto@kernel.org
