pm24.git/include/linux/syscalls.h, branch v4.18-rc2

rseq: Introduce restartable sequences system call

2018-06-06T09:58:31Z

Expose a new system call allowing each thread to register one userspace memory area to be used as an ABI between kernel and user-space for two purposes: user-space restartable sequences and quick access to read the current CPU number value from user-space. * Restartable sequences (per-cpu atomics) Restartables sequences allow user-space to perform update operations on per-cpu data without requiring heavy-weight atomic operations. The restartable critical sections (percpu atomics) work has been started by Paul Turner and Andrew Hunter. It lets the kernel handle restart of critical sections. [1] [2] The re-implementation proposed here brings a few simplifications to the ABI which facilitates porting to other architectures and speeds up the user-space fast path. Here are benchmarks of various rseq use-cases. Test hardware: arm32: ARMv7 Processor rev 4 (v7l) "Cubietruck", 2-core x86-64: Intel E5-2630 v3@2.40GHz, 16-core, hyperthreading The following benchmarks were all performed on a single thread. * Per-CPU statistic counter increment getcpu+atomic (ns/op) rseq (ns/op) speedup arm32: 344.0 31.4 11.0 x86-64: 15.3 2.0 7.7 * LTTng-UST: write event 32-bit header, 32-bit payload into tracer per-cpu buffer getcpu+atomic (ns/op) rseq (ns/op) speedup arm32: 2502.0 2250.0 1.1 x86-64: 117.4 98.0 1.2 * liburcu percpu: lock-unlock pair, dereference, read/compare word getcpu+atomic (ns/op) rseq (ns/op) speedup arm32: 751.0 128.5 5.8 x86-64: 53.4 28.6 1.9 * jemalloc memory allocator adapted to use rseq Using rseq with per-cpu memory pools in jemalloc at Facebook (based on rseq 2016 implementation): The production workload response-time has 1-2% gain avg. latency, and the P99 overall latency drops by 2-3%. * Reading the current CPU number Speeding up reading the current CPU number on which the caller thread is running is done by keeping the current CPU number up do date within the cpu_id field of the memory area registered by the thread. This is done by making scheduler preemption set the TIF_NOTIFY_RESUME flag on the current thread. Upon return to user-space, a notify-resume handler updates the current CPU value within the registered user-space memory area. User-space can then read the current CPU number directly from memory. Keeping the current cpu id in a memory area shared between kernel and user-space is an improvement over current mechanisms available to read the current CPU number, which has the following benefits over alternative approaches: - 35x speedup on ARM vs system call through glibc - 20x speedup on x86 compared to calling glibc, which calls vdso executing a "lsl" instruction, - 14x speedup on x86 compared to inlined "lsl" instruction, - Unlike vdso approaches, this cpu_id value can be read from an inline assembly, which makes it a useful building block for restartable sequences. - The approach of reading the cpu id through memory mapping shared between kernel and user-space is portable (e.g. ARM), which is not the case for the lsl-based x86 vdso. On x86, yet another possible approach would be to use the gs segment selector to point to user-space per-cpu data. This approach performs similarly to the cpu id cache, but it has two disadvantages: it is not portable, and it is incompatible with existing applications already using the gs segment selector for other purposes. Benchmarking various approaches for reading the current CPU number: ARMv7 Processor rev 4 (v7l) Machine model: Cubietruck - Baseline (empty loop): 8.4 ns - Read CPU from rseq cpu_id: 16.7 ns - Read CPU from rseq cpu_id (lazy register): 19.8 ns - glibc 2.19-0ubuntu6.6 getcpu: 301.8 ns - getcpu system call: 234.9 ns x86-64 Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz: - Baseline (empty loop): 0.8 ns - Read CPU from rseq cpu_id: 0.8 ns - Read CPU from rseq cpu_id (lazy register): 0.8 ns - Read using gs segment selector: 0.8 ns - "lsl" inline assembly: 13.0 ns - glibc 2.19-0ubuntu6 getcpu: 16.6 ns - getcpu system call: 53.9 ns - Speed (benchmark taken on v8 of patchset) Running 10 runs of hackbench -l 100000 seems to indicate, contrary to expectations, that enabling CONFIG_RSEQ slightly accelerates the scheduler: Configuration: 2 sockets * 8-core Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz (directly on hardware, hyperthreading disabled in BIOS, energy saving disabled in BIOS, turboboost disabled in BIOS, cpuidle.off=1 kernel parameter), with a Linux v4.6 defconfig+localyesconfig, restartable sequences series applied. * CONFIG_RSEQ=n avg.: 41.37 s std.dev.: 0.36 s * CONFIG_RSEQ=y avg.: 40.46 s std.dev.: 0.33 s - Size On x86-64, between CONFIG_RSEQ=n/y, the text size increase of vmlinux is 567 bytes, and the data size increase of vmlinux is 5696 bytes. [1] https://lwn.net/Articles/650333/ [2] http://www.linuxplumbersconf.org/2013/ocw/system/presentations/1695/original/LPC%20-%20PerCpu%20Atomics.pdf Signed-off-by: Mathieu Desnoyers Signed-off-by: Thomas Gleixner Acked-by: Peter Zijlstra (Intel) Cc: Joel Fernandes Cc: Catalin Marinas Cc: Dave Watson Cc: Will Deacon Cc: Andi Kleen Cc: "H . Peter Anvin" Cc: Chris Lameter Cc: Russell King Cc: Andrew Hunter Cc: Michael Kerrisk Cc: "Paul E . McKenney" Cc: Paul Turner Cc: Boqun Feng Cc: Josh Triplett Cc: Steven Rostedt Cc: Ben Maurer Cc: Alexander Viro Cc: linux-api@vger.kernel.org Cc: Andy Lutomirski Cc: Andrew Morton Cc: Linus Torvalds Link: http://lkml.kernel.org/r/20151027235635.16059.11630.stgit@pjt-glaptop.roam.corp.google.com Link: http://lkml.kernel.org/r/20150624222609.6116.86035.stgit@kitami.mtv.corp.google.com Link: https://lkml.kernel.org/r/20180602124408.8430-3-mathieu.desnoyers@efficios.com

Merge branch 'timers-2038-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2018-06-05T04:02:18Z

Pull time/Y2038 updates from Thomas Gleixner: - Consolidate SySV IPC UAPI headers - Convert SySV IPC to the new COMPAT_32BIT_TIME mechanism - Cleanup the core interfaces and standardize on the ktime_get_* naming convention. - Convert the X86 platform ops to timespec64 - Remove the ugly temporary timespec64 hack * 'timers-2038-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (22 commits) x86: Convert x86_platform_ops to timespec64 timekeeping: Add more coarse clocktai/boottime interfaces timekeeping: Add ktime_get_coarse_with_offset timekeeping: Standardize on ktime_get_*() naming timekeeping: Clean up ktime_get_real_ts64 timekeeping: Remove timespec64 hack y2038: ipc: Redirect ipc(SEMTIMEDOP, ...) to compat_ksys_semtimedop y2038: ipc: Enable COMPAT_32BIT_TIME y2038: ipc: Use __kernel_timespec y2038: ipc: Report long times to user space y2038: ipc: Use ktime_get_real_seconds consistently y2038: xtensa: Extend sysvipc data structures y2038: powerpc: Extend sysvipc data structures y2038: sparc: Extend sysvipc data structures y2038: parisc: Extend sysvipc data structures y2038: mips: Extend sysvipc data structures y2038: arm64: Extend sysvipc compat data structures y2038: s390: Remove unneeded ipc uapi header files y2038: ia64: Remove unneeded ipc uapi header files y2038: alpha: Remove unneeded ipc uapi header files ...

Merge branch 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

2018-06-05T03:27:54Z

Pull timers and timekeeping updates from Thomas Gleixner: - Core infrastucture work for Y2038 to address the COMPAT interfaces: + Add a new Y2038 safe __kernel_timespec and use it in the core code + Introduce config switches which allow to control the various compat mechanisms + Use the new config switch in the posix timer code to control the 32bit compat syscall implementation. - Prevent bogus selection of CPU local clocksources which causes an endless reselection loop - Remove the extra kthread in the clocksource code which has no value and just adds another level of indirection - The usual bunch of trivial updates, cleanups and fixlets all over the place - More SPDX conversions * 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (24 commits) clocksource/drivers/mxs_timer: Switch to SPDX identifier clocksource/drivers/timer-imx-tpm: Switch to SPDX identifier clocksource/drivers/timer-imx-gpt: Switch to SPDX identifier clocksource/drivers/timer-imx-gpt: Remove outdated file path clocksource/drivers/arc_timer: Add comments about locking while read GFRC clocksource/drivers/mips-gic-timer: Add pr_fmt and reword pr_* messages clocksource/drivers/sprd: Fix Kconfig dependency clocksource: Move inline keyword to the beginning of function declarations timer_list: Remove unused function pointer typedef timers: Adjust a kernel-doc comment tick: Prefer a lower rating device only if it's CPU local device clocksource: Remove kthread time: Change nanosleep to safe __kernel_* types time: Change types to new y2038 safe __kernel_* types time: Fix get_timespec64() for y2038 safe compat interfaces time: Add new y2038 safe __kernel_timespec posix-timers: Make compat syscalls depend on CONFIG_COMPAT_32BIT_TIME time: Introduce CONFIG_COMPAT_32BIT_TIME time: Introduce CONFIG_64BIT_TIME in architectures compat: Enable compat_get/put_timespec64 always ...

aio: implement io_pgetevents

2018-05-02T17:57:24Z

This is the io_getevents equivalent of ppoll/pselect and allows to properly mix signals and aio completions (especially with IOCB_CMD_POLL) and atomically executes the following sequence: sigset_t origmask; pthread_sigmask(SIG_SETMASK, &sigmask, &origmask); ret = io_getevents(ctx, min_nr, nr, events, timeout); pthread_sigmask(SIG_SETMASK, &origmask, NULL); Note that unlike many other signal related calls we do not pass a sigmask size, as that would get us to 7 arguments, which aren't easily supported by the syscall infrastructure. It seems a lot less painful to just add a new syscall variant in the unlikely case we're going to increase the sigset size. Signed-off-by: Christoph Hellwig Reviewed-by: Greg Kroah-Hartman Reviewed-by: Darrick J. Wong

y2038: ipc: Use __kernel_timespec

2018-04-20T14:20:24Z

This is a preparatation for changing over __kernel_timespec to 64-bit times, which involves assigning new system call numbers for mq_timedsend(), mq_timedreceive() and semtimedop() for compatibility with future y2038 proof user space. The existing ABIs will remain available through compat code. Signed-off-by: Arnd Bergmann

time: Change nanosleep to safe __kernel_* types

2018-04-19T11:32:03Z

Change over clock_nanosleep syscalls to use y2038 safe __kernel_timespec times. This will enable changing over of these syscalls to use new y2038 safe syscalls when the architectures define the CONFIG_64BIT_TIME. Note that nanosleep syscall is deprecated and does not have a plan for making it y2038 safe. But, the syscall should work as before on 64 bit machines and on 32 bit machines, the syscall works correctly until y2038 as before using the existing compat syscall version. There is no new syscall for supporting 64 bit time_t on 32 bit architectures. Cc: linux-api@vger.kernel.org Signed-off-by: Deepa Dinamani Signed-off-by: Arnd Bergmann

time: Change types to new y2038 safe __kernel_* types

2018-04-19T11:31:49Z

Change over clock_settime, clock_gettime and clock_getres syscalls to use __kernel_timespec times. This will enable changing over of these syscalls to use new y2038 safe syscalls when the architectures define the CONFIG_64BIT_TIME. Cc: linux-api@vger.kernel.org Signed-off-by: Deepa Dinamani Signed-off-by: Arnd Bergmann

syscalls/core, syscalls/x86: Clean up syscall stub naming convention

2018-04-09T14:47:27Z

Tidy the naming convention for compat syscall subs. Hints which describe the purpose of the stub go in front and receive a double underscore to denote that they are generated on-the-fly by the SYSCALL_DEFINEx() macro. For the generic case, this means (0xffffffff prefix removed): 810f08d0 t kernel_waitid # common C function (see kernel/exit.c) __do_sys_waitid # inlined helper doing the actual work # (takes original parameters as declared) 810f1aa0 T __se_sys_waitid # sign-extending C function calling inlined # helper (takes parameters of type long; # casts them to the declared type) 810f1aa0 T sys_waitid # alias to __se_sys_waitid() (taking # parameters as declared), to be included # in syscall table For x86, the naming is as follows: 810efc70 t kernel_waitid # common C function (see kernel/exit.c) __do_sys_waitid # inlined helper doing the actual work # (takes original parameters as declared) 810efd60 t __se_sys_waitid # sign-extending C function calling inlined # helper (takes parameters of type long; # casts them to the declared type) 810f1140 T __ia32_sys_waitid # IA32_EMULATION 32-bit-ptregs -> C stub, # calls __se_sys_waitid(); to be included # in syscall table 810f1110 T sys_waitid # x86 64-bit-ptregs -> C stub, calls # __se_sys_waitid(); to be included in # syscall table For x86, sys_waitid() will be re-named to __x64_sys_waitid in a follow-up patch. Suggested-by: Ingo Molnar Signed-off-by: Dominik Brodowski Cc: Al Viro Cc: Andrew Morton Cc: Andy Lutomirski Cc: Borislav Petkov Cc: Brian Gerst Cc: Denys Vlasenko Cc: Josh Poimboeuf Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20180409105145.5364-2-linux@dominikbrodowski.net Signed-off-by: Ingo Molnar

syscalls/x86: Use 'struct pt_regs' based syscall calling convention for 64-bit syscalls

2018-04-05T14:59:26Z

Let's make use of ARCH_HAS_SYSCALL_WRAPPER=y on pure 64-bit x86-64 systems: Each syscall defines a stub which takes struct pt_regs as its only argument. It decodes just those parameters it needs, e.g: asmlinkage long sys_xyzzy(const struct pt_regs *regs) { return SyS_xyzzy(regs->di, regs->si, regs->dx); } This approach avoids leaking random user-provided register content down the call chain. For example, for sys_recv() which is a 4-parameter syscall, the assembly now is (in slightly reordered fashion): : callq <__fentry__> /* decode regs->di, ->si, ->dx and ->r10 */ mov 0x70(%rdi),%rdi mov 0x68(%rdi),%rsi mov 0x60(%rdi),%rdx mov 0x38(%rdi),%rcx [ SyS_recv() is automatically inlined by the compiler, as it is not [yet] used anywhere else ] /* clear %r9 and %r8, the 5th and 6th args */ xor %r9d,%r9d xor %r8d,%r8d /* do the actual work */ callq __sys_recvfrom /* cleanup and return */ cltq retq The only valid place in an x86-64 kernel which rightfully calls a syscall function on its own -- vsyscall -- needs to be modified to pass struct pt_regs onwards as well. To keep the syscall table generation working independent of SYSCALL_PTREGS being enabled, the stubs are named the same as the "original" syscall stubs, i.e. sys_*(). This patch is based on an original proof-of-concept | From: Linus Torvalds | Signed-off-by: Linus Torvalds and was split up and heavily modified by me, in particular to base it on ARCH_HAS_SYSCALL_WRAPPER, to limit it to 64-bit-only for the time being, and to update the vsyscall to the new calling convention. Signed-off-by: Dominik Brodowski Acked-by: Linus Torvalds Cc: Al Viro Cc: Andrew Morton Cc: Andy Lutomirski Cc: Borislav Petkov Cc: Brian Gerst Cc: Denys Vlasenko Cc: H. Peter Anvin Cc: Josh Poimboeuf Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20180405095307.3730-4-linux@dominikbrodowski.net Signed-off-by: Ingo Molnar

syscalls/core: Introduce CONFIG_ARCH_HAS_SYSCALL_WRAPPER=y

2018-04-05T14:59:25Z

It may be useful for an architecture to override the definitions of the SYSCALL_DEFINE0() and __SYSCALL_DEFINEx() macros in , in particular to use a different calling convention for syscalls. This patch provides a mechanism to do so: It introduces CONFIG_ARCH_HAS_SYSCALL_WRAPPER. If it is enabled, is included in and may be used to define the macros mentioned above. Moreover, as the syscall calling convention may be different if CONFIG_ARCH_HAS_SYSCALL_WRAPPER is set, the syscall function prototypes in are #ifndef'd out in that case. Signed-off-by: Dominik Brodowski Acked-by: Linus Torvalds Cc: Al Viro Cc: Andrew Morton Cc: Andy Lutomirski Cc: Borislav Petkov Cc: Brian Gerst Cc: Denys Vlasenko Cc: H. Peter Anvin Cc: Josh Poimboeuf Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20180405095307.3730-3-linux@dominikbrodowski.net Signed-off-by: Ingo Molnar