Song Liu noticed switch_mm_irqs_off() taking a lot of CPU time in recent
kernels,using 1.8% of a 48 CPU system during a netperf to localhost run.
Digging into the profile, we noticed that cpumask_clear_cpu and
cpumask_set_cpu together take about half of the CPU time taken by
switch_mm_irqs_off().
However, the CPUs running netperf end up switching back and forth
between netperf and the idle task, which does not require changes
to the mm_cpumask. Furthermore, the init_mm cpumask ends up being
the most heavily contended one in the system.
Simply skipping changes to mm_cpumask(&init_mm) reduces overhead.
Reported-and-tested-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Rik van Riel <riel@surriel.com>
Acked-by: Dave Hansen <dave.hansen@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: efault@gmx.de
Cc: kernel-team@fb.com
Cc: luto@kernel.org
Link: http://lkml.kernel.org/r/20180716190337.26133-8-riel@surriel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Lazy TLB mode can result in an idle CPU being woken up by a TLB flush,
when all it really needs to do is reload %CR3 at the next context switch,
assuming no page table pages got freed.
Memory ordering is used to prevent race conditions between switch_mm_irqs_off,
which checks whether .tlb_gen changed, and the TLB invalidation code, which
increments .tlb_gen whenever page table entries get invalidated.
The atomic increment in inc_mm_tlb_gen is its own barrier; the context
switch code adds an explicit barrier between reading tlbstate.is_lazy and
next->context.tlb_gen.
Unlike the 2016 version of this patch, CPUs with cpu_tlbstate.is_lazy set
are not removed from the mm_cpumask(mm), since that would prevent the TLB
flush IPIs at page table free time from being sent to all the CPUs
that need them.
This patch reduces total CPU use in the system by about 1-2% for a
memcache workload on two socket systems, and by about 1% for a heavily
multi-process netperf between two systems.
Tested-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Rik van Riel <riel@surriel.com>
Acked-by: Dave Hansen <dave.hansen@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: efault@gmx.de
Cc: kernel-team@fb.com
Cc: luto@kernel.org
Link: http://lkml.kernel.org/r/20180716190337.26133-5-riel@surriel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Andy discovered that speculative memory accesses while in lazy
TLB mode can crash a system, when a CPU tries to dereference a
speculative access using memory contents that used to be valid
page table memory, but have since been reused for something else
and point into la-la land.
The latter problem can be prevented in two ways. The first is to
always send a TLB shootdown IPI to CPUs in lazy TLB mode, while
the second one is to only send the TLB shootdown at page table
freeing time.
The second should result in fewer IPIs, since operationgs like
mprotect and madvise are very common with some workloads, but
do not involve page table freeing. Also, on munmap, batching
of page table freeing covers much larger ranges of virtual
memory than the batching of unmapped user pages.
Tested-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Rik van Riel <riel@surriel.com>
Acked-by: Dave Hansen <dave.hansen@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: efault@gmx.de
Cc: kernel-team@fb.com
Cc: luto@kernel.org
Link: http://lkml.kernel.org/r/20180716190337.26133-3-riel@surriel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
ioremap() calls pud_free_pmd_page() / pmd_free_pte_page() when it creates
a pud / pmd map. The following preconditions are met at their entry.
- All pte entries for a target pud/pmd address range have been cleared.
- System-wide TLB purges have been peformed for a target pud/pmd address
range.
The preconditions assure that there is no stale TLB entry for the range.
Speculation may not cache TLB entries since it requires all levels of page
entries, including ptes, to have P & A-bits set for an associated address.
However, speculation may cache pud/pmd entries (paging-structure caches)
when they have P-bit set.
Add a system-wide TLB purge (INVLPG) to a single page after clearing
pud/pmd entry's P-bit.
SDM 4.10.4.1, Operation that Invalidate TLBs and Paging-Structure Caches,
states that:
INVLPG invalidates all paging-structure caches associated with the
current PCID regardless of the liner addresses to which they correspond.
Fixes: 28ee90fe60 ("x86/mm: implement free pmd/pte page interfaces")
Signed-off-by: Toshi Kani <toshi.kani@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: mhocko@suse.com
Cc: akpm@linux-foundation.org
Cc: hpa@zytor.com
Cc: cpandya@codeaurora.org
Cc: linux-mm@kvack.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: Joerg Roedel <joro@8bytes.org>
Cc: stable@vger.kernel.org
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: <stable@vger.kernel.org>
Link: https://lkml.kernel.org/r/20180627141348.21777-4-toshi.kani@hpe.com
The PAE 3-level paging code currently doesn't mitigate L1TF by flipping the
offset bits, and uses the high PTE word, thus bits 32-36 for type, 37-63 for
offset. The lower word is zeroed, thus systems with less than 4GB memory are
safe. With 4GB to 128GB the swap type selects the memory locations vulnerable
to L1TF; with even more memory, also the swap offfset influences the address.
This might be a problem with 32bit PAE guests running on large 64bit hosts.
By continuing to keep the whole swap entry in either high or low 32bit word of
PTE we would limit the swap size too much. Thus this patch uses the whole PAE
PTE with the same layout as the 64bit version does. The macros just become a
bit tricky since they assume the arch-dependent swp_entry_t to be 32bit.
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Michal Hocko <mhocko@suse.com>
The previous patch has limited swap file size so that large offsets cannot
clear bits above MAX_PA/2 in the pte and interfere with L1TF mitigation.
It assumed that offsets are encoded starting with bit 12, same as pfn. But
on x86_64, offsets are encoded starting with bit 9.
Thus the limit can be raised by 3 bits. That means 16TB with 42bit MAX_PA
and 256TB with 46bit MAX_PA.
Fixes: 377eeaa8e1 ("x86/speculation/l1tf: Limit swap file size to MAX_PA/2")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
For the L1TF workaround its necessary to limit the swap file size to below
MAX_PA/2, so that the higher bits of the swap offset inverted never point
to valid memory.
Add a mechanism for the architecture to override the swap file size check
in swapfile.c and add a x86 specific max swapfile check function that
enforces that limit.
The check is only enabled if the CPU is vulnerable to L1TF.
In VMs with 42bit MAX_PA the typical limit is 2TB now, on a native system
with 46bit PA it is 32TB. The limit is only per individual swap file, so
it's always possible to exceed these limits with multiple swap files or
partitions.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Dave Hansen <dave.hansen@intel.com>
For L1TF PROT_NONE mappings are protected by inverting the PFN in the page
table entry. This sets the high bits in the CPU's address space, thus
making sure to point to not point an unmapped entry to valid cached memory.
Some server system BIOSes put the MMIO mappings high up in the physical
address space. If such an high mapping was mapped to unprivileged users
they could attack low memory by setting such a mapping to PROT_NONE. This
could happen through a special device driver which is not access
protected. Normal /dev/mem is of course access protected.
To avoid this forbid PROT_NONE mappings or mprotect for high MMIO mappings.
Valid page mappings are allowed because the system is then unsafe anyways.
It's not expected that users commonly use PROT_NONE on MMIO. But to
minimize any impact this is only enforced if the mapping actually refers to
a high MMIO address (defined as the MAX_PA-1 bit being set), and also skip
the check for root.
For mmaps this is straight forward and can be handled in vm_insert_pfn and
in remap_pfn_range().
For mprotect it's a bit trickier. At the point where the actual PTEs are
accessed a lot of state has been changed and it would be difficult to undo
on an error. Since this is a uncommon case use a separate early page talk
walk pass for MMIO PROT_NONE mappings that checks for this condition
early. For non MMIO and non PROT_NONE there are no changes.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Acked-by: Dave Hansen <dave.hansen@intel.com>
Hussam reports:
I was poking around and for no real reason, I did cat /dev/mem and
strings /dev/mem. Then I saw the following warning in dmesg. I saved it
and rebooted immediately.
memremap attempted on mixed range 0x000000000009c000 size: 0x1000
------------[ cut here ]------------
WARNING: CPU: 0 PID: 11810 at kernel/memremap.c:98 memremap+0x104/0x170
[..]
Call Trace:
xlate_dev_mem_ptr+0x25/0x40
read_mem+0x89/0x1a0
__vfs_read+0x36/0x170
The memremap() implementation checks for attempts to remap System RAM
with MEMREMAP_WB and instead redirects those mapping attempts to the
linear map. However, that only works if the physical address range
being remapped is page aligned. In low memory we have situations like
the following:
00000000-00000fff : Reserved
00001000-0009fbff : System RAM
0009fc00-0009ffff : Reserved
...where System RAM intersects Reserved ranges on a sub-page page
granularity.
Given that devmem_is_allowed() special cases any attempt to map System
RAM in the first 1MB of memory, replace page_is_ram() with the more
precise region_intersects() to trap attempts to map disallowed ranges.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=199999
Link: http://lkml.kernel.org/r/152856436164.18127.2847888121707136898.stgit@dwillia2-desk3.amr.corp.intel.com
Fixes: 92281dee82 ("arch: introduce memremap()")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: Hussam Al-Tayeb <me@hussam.eu.org>
Tested-by: Hussam Al-Tayeb <me@hussam.eu.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull x86 updates and fixes from Thomas Gleixner:
- Fix the (late) fallout from the vector management rework causing
hlist corruption and irq descriptor reference leaks caused by a
missing sanity check.
The straight forward fix triggered another long standing issue to
surface. The pre rework code hid the issue due to being way slower,
but now the chance that user space sees an EBUSY error return when
updating irq affinities is way higher, though quite a bunch of
userspace tools do not handle it properly despite the fact that EBUSY
could be returned for at least 10 years.
It turned out that the EBUSY return can be avoided completely by
utilizing the existing delayed affinity update mechanism for irq
remapped scenarios as well. That's a bit more error handling in the
kernel, but avoids fruitless fingerpointing discussions with tool
developers.
- Decouple PHYSICAL_MASK from AMD SME as its going to be required for
the upcoming Intel memory encryption support as well.
- Handle legacy device ACPI detection properly for newer platforms
- Fix the wrong argument ordering in the vector allocation tracepoint
- Simplify the IDT setup code for the APIC=n case
- Use the proper string helpers in the MTRR code
- Remove a stale unused VDSO source file
- Convert the microcode update lock to a raw spinlock as its used in
atomic context.
* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/intel_rdt: Enable CMT and MBM on new Skylake stepping
x86/apic/vector: Print APIC control bits in debugfs
genirq/affinity: Defer affinity setting if irq chip is busy
x86/platform/uv: Use apic_ack_irq()
x86/ioapic: Use apic_ack_irq()
irq_remapping: Use apic_ack_irq()
x86/apic: Provide apic_ack_irq()
genirq/migration: Avoid out of line call if pending is not set
genirq/generic_pending: Do not lose pending affinity update
x86/apic/vector: Prevent hlist corruption and leaks
x86/vector: Fix the args of vector_alloc tracepoint
x86/idt: Simplify the idt_setup_apic_and_irq_gates()
x86/platform/uv: Remove extra parentheses
x86/mm: Decouple dynamic __PHYSICAL_MASK from AMD SME
x86: Mark native_set_p4d() as __always_inline
x86/microcode: Make the late update update_lock a raw lock for RT
x86/mtrr: Convert to use strncpy_from_user() helper
x86/mtrr: Convert to use match_string() helper
x86/vdso: Remove unused file
x86/i8237: Register device based on FADT legacy boot flag
AMD SME claims one bit from physical address to indicate whether the
page is encrypted or not. To achieve that we clear out the bit from
__PHYSICAL_MASK.
The capability to adjust __PHYSICAL_MASK is required beyond AMD SME.
For instance for upcoming Intel Multi-Key Total Memory Encryption.
Factor it out into a separate feature with own Kconfig handle.
It also helps with overhead of AMD SME. It saves more than 3k in .text
on defconfig + AMD_MEM_ENCRYPT:
add/remove: 3/2 grow/shrink: 5/110 up/down: 189/-3753 (-3564)
We would need to return to this once we have infrastructure to patch
constants in code. That's good candidate for it.
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>
Cc: linux-mm@kvack.org
Cc: "H. Peter Anvin" <hpa@zytor.com>
Link: https://lkml.kernel.org/r/20180518113028.79825-1-kirill.shutemov@linux.intel.com
Pull x86 debug updates from Ingo Molnar:
"This contains the x86 oops code printing reorganization and cleanups
from Borislav Betkov, with a particular focus in enhancing opcode
dumping all around"
* 'x86-debug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/dumpstack: Explain the reasoning for the prologue and buffer size
x86/dumpstack: Save first regs set for the executive summary
x86/dumpstack: Add a show_ip() function
x86/fault: Dump user opcode bytes on fatal faults
x86/dumpstack: Add loglevel argument to show_opcodes()
x86/dumpstack: Improve opcodes dumping in the code section
x86/dumpstack: Carve out code-dumping into a function
x86/dumpstack: Unexport oops_begin()
x86/dumpstack: Remove code_bytes
Pull x86 cleanups from Ingo Molnar:
"Misc cleanups"
* 'x86-cleanups-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/apm: Fix spelling mistake: "caculate" -> "calculate"
x86/mtrr: Rename main.c to mtrr.c and remove duplicate prefixes
x86: Remove pr_fmt duplicate logging prefixes
x86/early-quirks: Rename duplicate define of dev_err
x86/bpf: Clean up non-standard comments, to make the code more readable
Pull x86 boot updates from Ingo Molnar:
- Centaur CPU updates (David Wang)
- AMD and other CPU topology enumeration improvements and fixes
(Borislav Petkov, Thomas Gleixner, Suravee Suthikulpanit)
- Continued 5-level paging work (Kirill A. Shutemov)
* 'x86-boot-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/mm: Mark __pgtable_l5_enabled __initdata
x86/mm: Mark p4d_offset() __always_inline
x86/mm: Introduce the 'no5lvl' kernel parameter
x86/mm: Stop pretending pgtable_l5_enabled is a variable
x86/mm: Unify pgtable_l5_enabled usage in early boot code
x86/boot/compressed/64: Fix trampoline page table address calculation
x86/CPU: Move x86_cpuinfo::x86_max_cores assignment to detect_num_cpu_cores()
x86/Centaur: Report correct CPU/cache topology
x86/CPU: Move cpu_detect_cache_sizes() into init_intel_cacheinfo()
x86/CPU: Make intel_num_cpu_cores() generic
x86/CPU: Move cpu local function declarations to local header
x86/CPU/AMD: Derive CPU topology from CPUID function 0xB when available
x86/CPU: Modify detect_extended_topology() to return result
x86/CPU/AMD: Calculate last level cache ID from number of sharing threads
x86/CPU: Rename intel_cacheinfo.c to cacheinfo.c
perf/events/amd/uncore: Fix amd_uncore_llc ID to use pre-defined cpu_llc_id
x86/CPU/AMD: Have smp_num_siblings and cpu_llc_id always be present
x86/Centaur: Initialize supported CPU features properly
Pull siginfo updates from Eric Biederman:
"This set of changes close the known issues with setting si_code to an
invalid value, and with not fully initializing struct siginfo. There
remains work to do on nds32, arc, unicore32, powerpc, arm, arm64, ia64
and x86 to get the code that generates siginfo into a simpler and more
maintainable state. Most of that work involves refactoring the signal
handling code and thus careful code review.
Also not included is the work to shrink the in kernel version of
struct siginfo. That depends on getting the number of places that
directly manipulate struct siginfo under control, as it requires the
introduction of struct kernel_siginfo for the in kernel things.
Overall this set of changes looks like it is making good progress, and
with a little luck I will be wrapping up the siginfo work next
development cycle"
* 'siginfo-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: (46 commits)
signal/sh: Stop gcc warning about an impossible case in do_divide_error
signal/mips: Report FPE_FLTUNK for undiagnosed floating point exceptions
signal/um: More carefully relay signals in relay_signal.
signal: Extend siginfo_layout with SIL_FAULT_{MCEERR|BNDERR|PKUERR}
signal: Remove unncessary #ifdef SEGV_PKUERR in 32bit compat code
signal/signalfd: Add support for SIGSYS
signal/signalfd: Remove __put_user from signalfd_copyinfo
signal/xtensa: Use force_sig_fault where appropriate
signal/xtensa: Consistenly use SIGBUS in do_unaligned_user
signal/um: Use force_sig_fault where appropriate
signal/sparc: Use force_sig_fault where appropriate
signal/sparc: Use send_sig_fault where appropriate
signal/sh: Use force_sig_fault where appropriate
signal/s390: Use force_sig_fault where appropriate
signal/riscv: Replace do_trap_siginfo with force_sig_fault
signal/riscv: Use force_sig_fault where appropriate
signal/parisc: Use force_sig_fault where appropriate
signal/parisc: Use force_sig_mceerr where appropriate
signal/openrisc: Use force_sig_fault where appropriate
signal/nios2: Use force_sig_fault where appropriate
...
I got a bug report that the following code (roughly) was
causing a SIGSEGV:
mprotect(ptr, size, PROT_EXEC);
mprotect(ptr, size, PROT_NONE);
mprotect(ptr, size, PROT_READ);
*ptr = 100;
The problem is hit when the mprotect(PROT_EXEC)
is implicitly assigned a protection key to the VMA, and made
that key ACCESS_DENY|WRITE_DENY. The PROT_NONE mprotect()
failed to remove the protection key, and the PROT_NONE->
PROT_READ left the PTE usable, but the pkey still in place
and left the memory inaccessible.
To fix this, we ensure that we always "override" the pkee
at mprotect() if the VMA does not have execute-only
permissions, but the VMA has the execute-only pkey.
We had a check for PROT_READ/WRITE, but it did not work
for PROT_NONE. This entirely removes the PROT_* checks,
which ensures that PROT_NONE now works.
Reported-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Ellermen <mpe@ellerman.id.au>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ram Pai <linuxram@us.ibm.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-mm@kvack.org
Cc: stable@vger.kernel.org
Fixes: 62b5f7d013 ("mm/core, x86/mm/pkeys: Add execute-only protection keys support")
Link: http://lkml.kernel.org/r/20180509171351.084C5A71@viggo.jf.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Sometimes it is useful to see which user opcode bytes RIP points to
when a fault happens: be it to rule out RIP corruption, to dump info
early during boot, when doing core dumps is impossible due to not having
a writable filesystem yet.
Sometimes it is useful if debugging an issue and one doesn't have access
to the executable which caused the fault in order to disassemble it.
That last aspect might have some security implications so
show_unhandled_signals could be revisited for that or a new config option
added.
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Link: https://lkml.kernel.org/r/20180417161124.5294-7-bp@alien8.de
Call clear_siginfo to ensure every stack allocated siginfo is properly
initialized before being passed to the signal sending functions.
Note: It is not safe to depend on C initializers to initialize struct
siginfo on the stack because C is allowed to skip holes when
initializing a structure.
The initialization of struct siginfo in tracehook_report_syscall_exit
was moved from the helper user_single_step_siginfo into
tracehook_report_syscall_exit itself, to make it clear that the local
variable siginfo gets fully initialized.
In a few cases the scope of struct siginfo has been reduced to make it
clear that siginfo siginfo is not used on other paths in the function
in which it is declared.
Instances of using memset to initialize siginfo have been replaced
with calls clear_siginfo for clarity.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
