When building x86 with Clang LTO and CFI, CFI jump regions are
automatically added to the end of the .text section late in linking. As a
result, the _etext position was being labelled before the appended jump
regions, causing confusion about where the boundaries of the executable
region actually are in the running kernel, and broke at least the fault
injection code. This moves the _etext mark to outside (and immediately
after) the .text area, as it already the case on other architectures
(e.g. arm64, arm).
Reported-and-tested-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20190423183827.GA4012@beast
Signed-off-by: Ingo Molnar <mingo@kernel.org>
In-NMI warnings have been added to vmalloc_fault() via:
ebc8827f75 ("x86: Barf when vmalloc and kmemcheck faults happen in NMI")
back in the time when our NMI entry code could not cope with nested NMIs.
These days, it's perfectly fine to take a fault in NMI context and we
don't have to care about the fact that IRET from the fault handler might
cause NMI nesting.
This warning has already been removed from 32-bit implementation of
vmalloc_fault() in:
6863ea0cda ("x86/mm: Remove in_nmi() warning from vmalloc_fault()")
but the 64-bit version was omitted.
Remove the bogus warning also from 64-bit implementation of vmalloc_fault().
Reported-by: Nicolai Stange <nstange@suse.de>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Joerg Roedel <jroedel@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 6863ea0cda ("x86/mm: Remove in_nmi() warning from vmalloc_fault()")
Link: http://lkml.kernel.org/r/nycvar.YFH.7.76.1904240902280.9803@cbobk.fhfr.pm
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The __put_user() macro evaluates it's @ptr argument inside the
__uaccess_begin() / __uaccess_end() region. While this would normally
not be expected to be an issue, an UBSAN bug (it ignored -fwrapv,
fixed in GCC 8+) would transform the @ptr evaluation for:
drivers/gpu/drm/i915/i915_gem_execbuffer.c: if (unlikely(__put_user(offset, &urelocs[r-stack].presumed_offset))) {
into a signed-overflow-UB check and trigger the objtool AC validation.
Finish this commit:
2a418cf3f5 ("x86/uaccess: Don't leak the AC flag into __put_user() value evaluation")
and explicitly evaluate all 3 arguments early.
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org> # build-tested
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: luto@kernel.org
Fixes: 2a418cf3f5 ("x86/uaccess: Don't leak the AC flag into __put_user() value evaluation")
Link: http://lkml.kernel.org/r/20190424072208.695962771@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The first kmemleak_scan() call after boot would trigger the crash below
because this callpath:
kernel_init
free_initmem
mem_encrypt_free_decrypted_mem
free_init_pages
unmaps memory inside the .bss when DEBUG_PAGEALLOC=y.
kmemleak_init() will register the .data/.bss sections and then
kmemleak_scan() will scan those addresses and dereference them looking
for pointer references. If free_init_pages() frees and unmaps pages in
those sections, kmemleak_scan() will crash if referencing one of those
addresses:
BUG: unable to handle kernel paging request at ffffffffbd402000
CPU: 12 PID: 325 Comm: kmemleak Not tainted 5.1.0-rc4+ #4
RIP: 0010:scan_block
Call Trace:
scan_gray_list
kmemleak_scan
kmemleak_scan_thread
kthread
ret_from_fork
Since kmemleak_free_part() is tolerant to unknown objects (not tracked
by kmemleak), it is fine to call it from free_init_pages() even if not
all address ranges passed to this function are known to kmemleak.
[ bp: Massage. ]
Fixes: b3f0907c71 ("x86/mm: Add .bss..decrypted section to hold shared variables")
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Brijesh Singh <brijesh.singh@amd.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: x86-ml <x86@kernel.org>
Link: https://lkml.kernel.org/r/20190423165811.36699-1-cai@lca.pw
UAPI Changes:
- uAPI "Fixes:" patch for the upcoming kernel 5.1, included here too
We have an Ack from the media folks (only current user) for this
late tweak
Cross-subsystem Changes:
- ALSA: hda: Fix racy display power access (Takashi, Chris)
Driver Changes:
- DDI and MIPI-DSI clocks fixes for Icelake (Vandita)
- Fix Icelake frequency change/locking (RPS) (Mika)
- Temporarily disable ppGTT read-only bit on Icelake (Mika)
- Add missing Icelake W/As (Mika)
- Enable 12 deep CSB status FIFO on Icelake (Mika)
- Inherit more Icelake code for Elkhartlake (Bob, Jani)
- Handle catastrophic error on engine reset (Mika)
- Shortcut readiness to reset check (Mika)
- Regression fix for GEM_BUSY causing us to report a mixed uabi-class request as not busy (Chris)
- Revert back to max link rate and lane count on eDP (Jani)
- Fix pipe BPP readout for BXT/GLK DSI (Ville)
- Set DP min_bpp to 8*3 for non-RGB output formats (Ville)
- Enable coarse preemption boundaries for Gen8 (Chris)
- Do not enable FEC without DSC (Ville)
- Restore correct BXT DDI latency optim setting calculation (Ville)
- Always reset context's RING registers to avoid running workload twice during reset (Chris)
- Set GPU wedged on driver unload (Janusz)
- Consolidate two similar barries from timeline into one (Chris)
- Only reset the pinned kernel contexts on resume (Chris)
- Wakeref tracking improvements (Chris, Imre)
- Lockdep fixes for shrinker interactions (Chris)
- Bump ready tasks ahead of busywaits in prep of semaphore use (Chris)
- Huge step in splitting display code into fine grained files (Jani)
- Refactor the IRQ init/reset macros for code saving (Paulo)
- Convert IRQ initialization code to uncore MMIO access (Paulo)
- Convert workarounds code to use uncore MMIO access (Chris)
- Nuke drm_crtc_state and use intel_atomic_state instead (Manasi)
- Update SKL clock-gating WA (Radhakrishna, Ville)
- Isolate GuC reset code flow (Chris)
- Expose force_dsc_enable through debugfs (Manasi)
- Header standalone compile testing framework (Jani)
- Code cleanups to reduce driver footprint (Chris)
- PSR code fixes and cleanups (Jose)
- Sparse and kerneldoc updates (Chris)
- Suppress spurious combo PHY B warning (Vile)
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190418080426.GA6409@jlahtine-desk.ger.corp.intel.com
Add a new command-line option "xen_timer_slop=<INT>" that sets the
minimum delta of virtual Xen timers. This commit does not change the
default timer slop value for virtual Xen timers.
Lowering the timer slop value should improve the accuracy of virtual
timers (e.g., better process dispatch latency), but it will likely
increase the number of virtual timer interrupts (relative to the
original slop setting).
The original timer slop value has not changed since the introduction
of the Xen-aware Linux kernel code. This commit provides users an
opportunity to tune timer performance given the refinements to
hardware and the Xen event channel processing. It also mirrors
a feature in the Xen hypervisor - the "timer_slop" Xen command line
option.
[boris: updated comment describing TIMER_SLOP]
Signed-off-by: Ryan Thibodeaux <ryan.thibodeaux@starlab.io>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
The kdump crashkernel low reservation is limited to under 896M even for
X86_64. This obscure and miserable limitation exists for compatibility
with old kexec-tools but the reason is not documented anywhere.
Some more tests/investigations about the background:
a) Previously, old kexec-tools could only load purgatory to memory under
2G. Eric removed that limitation in 2012 in kexec-tools:
b4f9f8599679 ("kexec x86_64: Make purgatory relocatable anywhere
in the 64bit address space.")
b) Back in 2013 Yinghai removed all the limitations in new kexec-tools,
bzImage64 can be loaded anywhere:
82c3dd2280d2 ("kexec, x86_64: Load bzImage64 above 4G")
c) Test results with old kexec-tools with old and latest kernels:
1. Old kexec-tools can not build with modern toolchain anymore,
I built it in a RHEL6 vm.
2. 2.0.0 kexec-tools does not work with the latest kernel even with
memory under 896M and gives an error:
"ELF core (kcore) parse failed"
For that it needs below kexec-tools fix:
ed15ba1b9977 ("build_mem_phdrs(): check if p_paddr is invalid")
3. Even with patched kexec-tools which fixes 2), it still needs some
other fixes to work correctly for KASLR-enabled kernels.
So the situation is:
* Old kexec-tools is already broken with latest kernels.
* We can not keep these limitations forever just for compatibility with very
old kexec-tools.
* If one must use old tools then he/she can choose crashkernel=X@Y.
* People have reported bugs where crashkernel=384M failed because KASLR
makes the 0-896M space sparse.
* Crashkernel can reserve in low or high area, it is natural to understand
low as memory under 4G.
Hence drop the 896M limitation and change crashkernel low reservation to
reserve under 4G by default.
Signed-off-by: Dave Young <dyoung@redhat.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Baoquan He <bhe@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Eric Biederman <ebiederm@xmission.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Petr Tesarik <ptesarik@suse.cz>
Cc: piliu@redhat.com
Cc: Ram Pai <linuxram@us.ibm.com>
Cc: Sinan Kaya <okaya@codeaurora.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: vgoyal@redhat.com
Cc: x86-ml <x86@kernel.org>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Zhimin Gu <kookoo.gu@intel.com>
Link: https://lkml.kernel.org/r/20190421035058.943630505@redhat.com
Pull perf fixes from Ingo Molnar:
"Misc fixes:
- various tooling fixes
- kretprobe fixes
- kprobes annotation fixes
- kprobes error checking fix
- fix the default events for AMD Family 17h CPUs
- PEBS fix
- AUX record fix
- address filtering fix"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/kprobes: Avoid kretprobe recursion bug
kprobes: Mark ftrace mcount handler functions nokprobe
x86/kprobes: Verify stack frame on kretprobe
perf/x86/amd: Add event map for AMD Family 17h
perf bpf: Return NULL when RB tree lookup fails in perf_env__find_btf()
perf tools: Fix map reference counting
perf evlist: Fix side band thread draining
perf tools: Check maps for bpf programs
perf bpf: Return NULL when RB tree lookup fails in perf_env__find_bpf_prog_info()
tools include uapi: Sync sound/asound.h copy
perf top: Always sample time to satisfy needs of use of ordered queuing
perf evsel: Use hweight64() instead of hweight_long(attr.sample_regs_user)
tools lib traceevent: Fix missing equality check for strcmp
perf stat: Disable DIR_FORMAT feature for 'perf stat record'
perf scripts python: export-to-sqlite.py: Fix use of parent_id in calls_view
perf header: Fix lock/unlock imbalances when processing BPF/BTF info
perf/x86: Fix incorrect PEBS_REGS
perf/ring_buffer: Fix AUX record suppression
perf/core: Fix the address filtering fix
kprobes: Fix error check when reusing optimized probes
Pull x86 fixes from Ingo Molnar:
"Misc fixes all over the place: a console spam fix, section attributes
fixes, a KASLR fix, a TLB stack-variable alignment fix, a reboot
quirk, boot options related warnings fix, an LTO fix, a deadlock fix
and an RDT fix"
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/cpu/intel: Lower the "ENERGY_PERF_BIAS: Set to normal" message's log priority
x86/cpu/bugs: Use __initconst for 'const' init data
x86/mm/KASLR: Fix the size of the direct mapping section
x86/Kconfig: Fix spelling mistake "effectivness" -> "effectiveness"
x86/mm/tlb: Revert "x86/mm: Align TLB invalidation info"
x86/reboot, efi: Use EFI reboot for Acer TravelMate X514-51T
x86/mm: Prevent bogus warnings with "noexec=off"
x86/build/lto: Fix truncated .bss with -fdata-sections
x86/speculation: Prevent deadlock on ssb_state::lock
x86/resctrl: Do not repeat rdtgroup mode initialization
ia64, parisc and sparc just use a copy of the generic version
of asm/sockios.h, and x86 is a redirect to the same file, so we
can just let the header file be generated.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case when the number of entries in the section header table is larger
then or equal to SHN_LORESERVE the size of the table is held in the sh_size
member of the initial entry in section header table instead of e_shnum.
Same with the string table index which is located in sh_link instead of
e_shstrndx.
This case is easily reproducible with KCFLAGS="-ffunction-sections",
bzImage build fails with "String table index out of bounds" error.
Signed-off-by: Artem Savkov <asavkov@redhat.com>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Eric W . Biederman <ebiederm@xmission.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20181129155615.2594-1-asavkov@redhat.com
[ Simplify the die() lines. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
DEBUG_HOTPLUG_CPU0 debug feature offlines a CPU as early as possible
allowing userspace to boot up without that CPU (so that it is possible
to check for unwanted dependencies towards the offlined CPU). After
doing so it emits a "CPU %u is now offline" pr_info, which is not enough
descriptive of why the CPU was offlined (e.g., one might be running with
a config that triggered some problem, not being aware that CONFIG_DEBUG_
HOTPLUG_CPU0 is set).
Add a bit more of informative text to the pr_info, so that it is
immediately obvious why a CPU has been offlined in early boot stages.
Background:
Got to scratch my head a bit while debugging a WARNING splat related to
the offlining of CPU0. Without being aware yet of this debug option it
wasn't immediately obvious why CPU0 was being offlined by the kernel.
Signed-off-by: Juri Lelli <juri.lelli@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: fenghua.yu@intel.com
Link: http://lkml.kernel.org/r/20181219151647.15073-1-juri.lelli@redhat.com
[ Merge line-broken line. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Linus pointed out that deciphering the raw #PF error code and printing
a more human readable message are two different things, and also that
printing the negative cases is mostly just noise[1]. For example, the
USER bit doesn't mean the fault originated in user code and stating
that an oops wasn't due to a protection keys violation isn't interesting
since an oops on a keys violation is a one-in-a-million scenario.
Remove the per-bit decoding of the error code and instead print:
- the raw error code
- why the fault occurred
- the effective privilege level of the access
- the type of access
- whether the fault originated in user code or kernel code
This provides the user with the information needed to triage 99.9% of
oopses without polluting the log with useless information or conflating
the error_code with the CPL.
Sample output:
BUG: kernel NULL pointer dereference, address = 0000000000000008
#PF: supervisor-privileged instruction fetch from kernel code
#PF: error_code(0x0010) - not-present page
BUG: unable to handle page fault for address = ffffbeef00000000
#PF: supervisor-privileged instruction fetch from kernel code
#PF: error_code(0x0010) - not-present page
BUG: unable to handle page fault for address = ffffc90000230000
#PF: supervisor-privileged write access from kernel code
#PF: error_code(0x000b) - reserved bit violation
[1] https://lkml.kernel.org/r/CAHk-=whk_fsnxVMvF1T2fFCaP2WrvSybABrLQCWLJyCvHw6NKA@mail.gmail.com
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20181221213657.27628-3-sean.j.christopherson@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reword the NULL pointer dereference case to simply state that a NULL
pointer was dereferenced, i.e. drop "unable to handle" as that implies
that there are instances where the kernel actual does handle NULL
pointer dereferences, which is not true barring funky exception fixup.
For the non-NULL case, replace "kernel paging request" with "page fault"
as the kernel can technically oops on faults that originated in user
code. Dropping "kernel" also allows future patches to provide detailed
information on where the fault occurred, e.g. user vs. kernel, without
conflicting with the initial BUG message.
In both cases, replace "at address=" with wording more appropriate to
the oops, as "at" may be interpreted as stating that the address is the
RIP of the instruction that faulted.
Last, and probably least, further qualify the NULL-pointer path by
checking that the fault actually originated in kernel code. It's
technically possible for userspace to map address 0, and not printing
a super specific message is the least of our worries if the kernel does
manage to oops on an actual NULL pointer dereference from userspace.
Before:
BUG: unable to handle kernel NULL pointer dereference at ffffbeef00000000
BUG: unable to handle kernel paging request at ffffbeef00000000
After:
BUG: kernel NULL pointer dereference, address = 0000000000000008
BUG: unable to handle page fault for address = ffffbeef00000000
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Yu-cheng Yu <yu-cheng.yu@intel.com>
Link: http://lkml.kernel.org/r/20181221213657.27628-2-sean.j.christopherson@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
LTO will happily inline __const_udelay() everywhere it is used. Forcing it
noinline saves ~44k text in a LTO build.
13999560 1740864 1499136 17239560 1070e08 vmlinux-with-udelay-inline
13954764 1736768 1499136 17190668 1064f0c vmlinux-wo-udelay-inline
Even without LTO this function should never be inlined.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190330004743.29541-4-andi@firstfloor.org
With gcc toplevel assembler statements that do not mark themselves as .text
may end up in other sections. This causes LTO boot crashes because various
assembler statements ended up in the middle of the initcall section. It's
also a latent problem without LTO, although it's currently not known to
cause any real problems.
According to the gcc team it's expected behavior.
Always mark all the top level assembler statements as text so that they
switch to the right section.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190330004743.29541-1-andi@firstfloor.org
Avoid kretprobe recursion loop bg by setting a dummy
kprobes to current_kprobe per-CPU variable.
This bug has been introduced with the asm-coded trampoline
code, since previously it used another kprobe for hooking
the function return placeholder (which only has a nop) and
trampoline handler was called from that kprobe.
This revives the old lost kprobe again.
With this fix, we don't see deadlock anymore.
And you can see that all inner-called kretprobe are skipped.
event_1 235 0
event_2 19375 19612
The 1st column is recorded count and the 2nd is missed count.
Above shows (event_1 rec) + (event_2 rec) ~= (event_2 missed)
(some difference are here because the counter is racy)
Reported-by: Andrea Righi <righi.andrea@gmail.com>
Tested-by: Andrea Righi <righi.andrea@gmail.com>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Fixes: c9becf58d9 ("[PATCH] kretprobe: kretprobe-booster")
Link: http://lkml.kernel.org/r/155094064889.6137.972160690963039.stgit@devbox
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Make the anon_inodes facility unconditional so that it can be used by core
VFS code and pidfd code.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
[christian@brauner.io: adapt commit message to mention pidfds]
Signed-off-by: Christian Brauner <christian@brauner.io>
The XO-1 and XO-1.5 batteries apparently differ in an ability to report
ambient temperature. We need to use a different compatible string for the
XO-1.5 battery.
Previously olpc_dt_fixup() used the presence of the battery node's
compatible property to decide whether the DT is up to date. Now we need
to look for a particular value in the compatible string, to decide
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Acked-by: Pavel Machek <pavel@ucw.cz>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
It was pointed out in a review, and checkpatch.pl complains about this.
Breaking it down into multiple ofw evaluations works just as well and
reads better.
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
To minimize the latency of timer interrupts as observed by the guest,
KVM adjusts the values it programs into the host timers to account for
the host's overhead of programming and handling the timer event. In
the event that the adjustments are too aggressive, i.e. the timer fires
earlier than the guest expects, KVM busy waits immediately prior to
entering the guest.
Currently, KVM manually converts the delay from nanoseconds to clock
cycles. But, the conversion is done in the guest's time domain, while
the delay occurs in the host's time domain. This is perfectly ok when
the guest and host are using the same TSC ratio, but if the guest is
using a different ratio then the delay may not be accurate and could
wait too little or too long.
When the guest is not using the host's ratio, convert the delay from
guest clock cycles to host nanoseconds and use ndelay() instead of
__delay() to provide more accurate timing. Because converting to
nanoseconds is relatively expensive, e.g. requires division and more
multiplication ops, continue using __delay() directly when guest and
host TSCs are running at the same ratio.
Cc: Liran Alon <liran.alon@oracle.com>
Cc: Wanpeng Li <wanpengli@tencent.com>
Cc: stable@vger.kernel.org
Fixes: 3b8a5df6c4 ("KVM: LAPIC: Tune lapic_timer_advance_ns automatically")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The introduction of adaptive tuning of lapic timer advancement did not
allow for the scenario where userspace would want to disable adaptive
tuning but still employ timer advancement, e.g. for testing purposes or
to handle a use case where adaptive tuning is unable to settle on a
suitable time. This is epecially pertinent now that KVM places a hard
threshold on the maximum advancment time.
Rework the timer semantics to accept signed values, with a value of '-1'
being interpreted as "use adaptive tuning with KVM's internal default",
and any other value being used as an explicit advancement time, e.g. a
time of '0' effectively disables advancement.
Note, this does not completely restore the original behavior of
lapic_timer_advance_ns. Prior to tracking the advancement per vCPU,
which is necessary to support autotuning, userspace could adjust
lapic_timer_advance_ns for *running* vCPU. With per-vCPU tracking, the
module params are snapshotted at vCPU creation, i.e. applying a new
advancement effectively requires restarting a VM.
Dynamically updating a running vCPU is possible, e.g. a helper could be
added to retrieve the desired delay, choosing between the global module
param and the per-VCPU value depending on whether or not auto-tuning is
(globally) enabled, but introduces a great deal of complexity. The
wrapper itself is not complex, but understanding and documenting the
effects of dynamically toggling auto-tuning and/or adjusting the timer
advancement is nigh impossible since the behavior would be dependent on
KVM's implementation as well as compiler optimizations. In other words,
providing stable behavior would require extremely careful consideration
now and in the future.
Given that the expected use of a manually-tuned timer advancement is to
"tune once, run many", use the vastly simpler approach of recognizing
changes to the module params only when creating a new vCPU.
Cc: Liran Alon <liran.alon@oracle.com>
Cc: Wanpeng Li <wanpengli@tencent.com>
Reviewed-by: Liran Alon <liran.alon@oracle.com>
Cc: stable@vger.kernel.org
Fixes: 3b8a5df6c4 ("KVM: LAPIC: Tune lapic_timer_advance_ns automatically")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Automatically adjusting the globally-shared timer advancement could
corrupt the timer, e.g. if multiple vCPUs are concurrently adjusting
the advancement value. That could be partially fixed by using a local
variable for the arithmetic, but it would still be susceptible to a
race when setting timer_advance_adjust_done.
And because virtual_tsc_khz and tsc_scaling_ratio are per-vCPU, the
correct calibration for a given vCPU may not apply to all vCPUs.
Furthermore, lapic_timer_advance_ns is marked __read_mostly, which is
effectively violated when finding a stable advancement takes an extended
amount of timer.
Opportunistically change the definition of lapic_timer_advance_ns to
a u32 so that it matches the style of struct kvm_timer. Explicitly
pass the param to kvm_create_lapic() so that it doesn't have to be
exposed to lapic.c, thus reducing the probability of unintentionally
using the global value instead of the per-vCPU value.
Cc: Liran Alon <liran.alon@oracle.com>
Cc: Wanpeng Li <wanpengli@tencent.com>
Reviewed-by: Liran Alon <liran.alon@oracle.com>
Cc: stable@vger.kernel.org
Fixes: 3b8a5df6c4 ("KVM: LAPIC: Tune lapic_timer_advance_ns automatically")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
To minimize the latency of timer interrupts as observed by the guest,
KVM adjusts the values it programs into the host timers to account for
the host's overhead of programming and handling the timer event. Now
that the timer advancement is automatically tuned during runtime, it's
effectively unbounded by default, e.g. if KVM is running as L1 the
advancement can measure in hundreds of milliseconds.
Disable timer advancement if adaptive tuning yields an advancement of
more than 5000ns, as large advancements can break reasonable assumptions
of the guest, e.g. that a timer configured to fire after 1ms won't
arrive on the next instruction. Although KVM busy waits to mitigate the
case of a timer event arriving too early, complications can arise when
shifting the interrupt too far, e.g. kvm-unit-test's vmx.interrupt test
will fail when its "host" exits on interrupts as KVM may inject the INTR
before the guest executes STI+HLT. Arguably the unit test is "broken"
in the sense that delaying a timer interrupt by 1ms doesn't technically
guarantee the interrupt will arrive after STI+HLT, but it's a reasonable
assumption that KVM should support.
Furthermore, an unbounded advancement also effectively unbounds the time
spent busy waiting, e.g. if the guest programs a timer with a very large
delay.
5000ns is a somewhat arbitrary threshold. When running on bare metal,
which is the intended use case, timer advancement is expected to be in
the general vicinity of 1000ns. 5000ns is high enough that false
positives are unlikely, while not being so high as to negatively affect
the host's performance/stability.
Note, a future patch will enable userspace to disable KVM's adaptive
tuning, which will allow priveleged userspace will to specifying an
advancement value in excess of this arbitrary threshold in order to
satisfy an abnormal use case.
Cc: Liran Alon <liran.alon@oracle.com>
Cc: Wanpeng Li <wanpengli@tencent.com>
Cc: stable@vger.kernel.org
Fixes: 3b8a5df6c4 ("KVM: LAPIC: Tune lapic_timer_advance_ns automatically")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
It was reported that with some special Multi Processor Group configuration,
e.g:
bcdedit.exe /set groupsize 1
bcdedit.exe /set maxgroup on
bcdedit.exe /set groupaware on
for a 16-vCPU guest WS2012 shows BSOD on boot when PV TLB flush mechanism
is in use.
Tracing kvm_hv_flush_tlb immediately reveals the issue:
kvm_hv_flush_tlb: processor_mask 0x0 address_space 0x0 flags 0x2
The only flag set in this request is HV_FLUSH_ALL_VIRTUAL_ADDRESS_SPACES,
however, processor_mask is 0x0 and no HV_FLUSH_ALL_PROCESSORS is specified.
We don't flush anything and apparently it's not what Windows expects.
TLFS doesn't say anything about such requests and newer Windows versions
seem to be unaffected. This all feels like a WS2012 bug, which is, however,
easy to workaround in KVM: let's flush everything when we see an empty
flush request, over-flushing doesn't hurt.
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
If guest sets MSR_IA32_TSCDEADLINE to value such that in host
time-domain it's shorter than lapic_timer_advance_ns, we can
reach a case that we call hrtimer_start() with expiration time set at
the past.
Because lapic_timer.timer is init with HRTIMER_MODE_ABS_PINNED, it
is not allowed to run in softirq and therefore will never expire.
To avoid such a scenario, verify that deadline expiration time is set on
host time-domain further than (now + lapic_timer_advance_ns).
A future patch can also consider adding a min_timer_deadline_ns module parameter,
similar to min_timer_period_us to avoid races that amount of ns it takes
to run logic could still call hrtimer_start() with expiration timer set
at the past.
Reviewed-by: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Liran Alon <liran.alon@oracle.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
