| Message ID | 20181109123730.8743-1-vincenzo.frascino@arm.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Unify vDSOs across more architectures | expand |
On 11/9/18, Vincenzo Frascino <vincenzo.frascino@arm.com> wrote: > vDSO (virtual dynamic shared object) is a mechanism that the Linux > kernel provides as an alternative to system calls to reduce where > possible the costs in terms of cycles. > This is possible because certain syscalls like gettimeofday() do > not write any data and return one or more values that are stored > in the kernel, which makes relatively safe calling them directly > as a library function. > > Even if the mechanism is pretty much standard, every architecture > in the last few years ended up implementing their own vDSO library > in the architectural code. > > The purpose of this patch-set is to identify the commonalities in > between the architectures and try to consolidate the common code > paths, starting with gettimeofday(). I'm very happy to see the generalization overall, as this is going to make it much easier to add the 64-bit time_t on the 32-bit vdso implementations, as well as other extensions we may want in the future. > This implementation contains the following design choices: > * Every architecture defines the arch specific code in an header in > "asm/vdso/". > * The generic implementation includes the arch specific one and lives > in "lib/vdso". Makes sense. > * The arch specific code for gettimeofday lives in > "<arch path>/vdso/gettimeofday.c" and includes the generic code only. > * This approach allows to consolidate the common code in a single place > with the benefit of avoiding code duplication. I would prefer to have that named clock_gettime.c and only implement that one system call there. We can easily have the gettimeofday(), time() and clock_getres() in the common code where that is needed for backwards compatibility. As I already mentioned in my comment for one of the patches, we won't have time() or gettimeofday() system calls on new architectures, and 32-bit architectures won't have them with 64-bit time_t. I suppose we can have the time() and gettimeofday() calls implemented in the generic vdso for everyone, but then only hook those up on architectures that already had them. The implementation of course is trivial based on clock_gettime() with CLOCK_REALTIME or CLOCK_REALTIME_COARSE respectively. > This implementation contains the portings to the common library for: arm64, > compat mode for arm64, arm and mips. > > The mips porting has been tested on qemu for mips32el. A configuration to > repeat the tests can be found at [4]. Ah, I missed the fact that you didn't do this for x86, so I assumed that this work was triggered by Thomas' recent cleanup of the x86 clock_gettime vdso code. I think it would be best to get x86 to use the same code and get the recent improvements ported to the others as well that way. Arnd
On 11/10/18 4:33 PM, Arnd Bergmann wrote: > On 11/9/18, Vincenzo Frascino <vincenzo.frascino@arm.com> wrote: >> vDSO (virtual dynamic shared object) is a mechanism that the Linux >> kernel provides as an alternative to system calls to reduce where >> possible the costs in terms of cycles. >> This is possible because certain syscalls like gettimeofday() do >> not write any data and return one or more values that are stored >> in the kernel, which makes relatively safe calling them directly >> as a library function. >> >> Even if the mechanism is pretty much standard, every architecture >> in the last few years ended up implementing their own vDSO library >> in the architectural code. >> >> The purpose of this patch-set is to identify the commonalities in >> between the architectures and try to consolidate the common code >> paths, starting with gettimeofday(). > > I'm very happy to see the generalization overall, as this is going > to make it much easier to add the 64-bit time_t on the 32-bit > vdso implementations, as well as other extensions we may want > in the future. > Thanks, >> This implementation contains the following design choices: >> * Every architecture defines the arch specific code in an header in >> "asm/vdso/". >> * The generic implementation includes the arch specific one and lives >> in "lib/vdso". > > Makes sense. > >> * The arch specific code for gettimeofday lives in >> "<arch path>/vdso/gettimeofday.c" and includes the generic code only. >> * This approach allows to consolidate the common code in a single place >> with the benefit of avoiding code duplication. > > I would prefer to have that named clock_gettime.c and only implement > that one system call there. We can easily have the gettimeofday(), > time() and clock_getres() in the common code where that is needed > for backwards compatibility. As I already mentioned in my comment > for one of the patches, we won't have time() or gettimeofday() system > calls on new architectures, and 32-bit architectures won't have them > with 64-bit time_t. > The reason why I used gettimeofday.c as a filename is because it is the naming convention that I could see more often in the kernel and I feel that, if we have even the implementations of gettimeofday(), time() and clock_getres() in the same file is probably the most representative. > I suppose we can have the time() and gettimeofday() calls > implemented in the generic vdso for everyone, but then > only hook those up on architectures that already had > them. The implementation of course is trivial based on > clock_gettime() with CLOCK_REALTIME or CLOCK_REALTIME_COARSE > respectively. > I agree, and in the long term would be my preferred approach as well, but I would prefer to keep this "optimization" in a separate patchset for what I mentioned in my previous email. >> This implementation contains the portings to the common library for: arm64, >> compat mode for arm64, arm and mips. >> >> The mips porting has been tested on qemu for mips32el. A configuration to >> repeat the tests can be found at [4]. > > Ah, I missed the fact that you didn't do this for x86, so I assumed that > this work was triggered by Thomas' recent cleanup of the x86 clock_gettime > vdso code. I think it would be best to get x86 to use the same code and > get the recent improvements ported to the others as well that way. > Yes, I agree it would be great to have have the other architectures to join the effort because I believe that the whole ecosystem would benefit from it. > Arnd >
vDSO (virtual dynamic shared object) is a mechanism that the Linux kernel provides as an alternative to system calls to reduce where possible the costs in terms of cycles. This is possible because certain syscalls like gettimeofday() do not write any data and return one or more values that are stored in the kernel, which makes relatively safe calling them directly as a library function. Even if the mechanism is pretty much standard, every architecture in the last few years ended up implementing their own vDSO library in the architectural code. The purpose of this patch-set is to identify the commonalities in between the architectures and try to consolidate the common code paths, starting with gettimeofday(). This implementation contains the following design choices: * Every architecture defines the arch specific code in an header in "asm/vdso/". * The generic implementation includes the arch specific one and lives in "lib/vdso". * The arch specific code for gettimeofday lives in "<arch path>/vdso/gettimeofday.c" and includes the generic code only. * This approach allows to consolidate the common code in a single place with the benefit of avoiding code duplication. This implementation contains the portings to the common library for: arm64, compat mode for arm64, arm and mips. The mips porting has been tested on qemu for mips32el. A configuration to repeat the tests can be found at [4]. Last but not least from this porting arm64, compat arm64, arm and mips gain the support for: * CLOCK_BOOTTIME that can be useful in certain scenarios since it keeps track of the time during sleep as well. * CLOCK_TAI that is like CLOCK_REALTIME, but uses the International Atomic Time (TAI) reference instead of UTC to avoid jumping on leap second updates. for both clock_gettime and clock_getres. The porting has been validated using the vdsotest test-suite [1] extended to cover all the clock ids [2]. To simplify the testing, a copy of the patchset on top of a recent linux tree can be found at [3]. [1] https://github.com/nlynch-mentor/vdsotest [2] https://github.com/fvincenzo/vdsotest [3] git://linux-arm.org/linux-vf.git vdso/v1 [4] git://linux-arm.org/linux-vf.git vdso-mips/v1 Cc: Catalin Marinas <catalin.marinas@arm.com> Cc: Will Deacon <will.deacon@arm.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Russell King <linux@armlinux.org.uk> Cc: Ralf Baechle <ralf@linux-mips.org> Cc: Paul Burton <paul.burton@mips.com> Cc: Daniel Lezcano <daniel.lezcano@linaro.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Mark Salyzyn <salyzyn@android.com> Cc: Peter Collingbourne <pcc@google.com> Signed-off-by: Vincenzo Frascino <vincenzo.frascino@arm.com> Vincenzo Frascino (27): kernel: Standardize vdso_datapage kernel: Add Monotonic boot time support kernel: Add International Atomic Time support kernel: Add masks support for Raw and NTP time kernel: Add clock_mode support kernel: Define gettimeofday vdso common code arm64: Substitute gettimeofday with C implementation arm64: compat: Alloc separate pages for vectors and sigpage arm64: compat: Split kuser32 arm64: compat: Refactor aarch32_alloc_vdso_pages() arm64: compat: Add KUSER_HELPERS config option arm64: compat: Add missing syscall numbers arm64: compat: Expose signal related structures arm64: compat: Generate asm offsets for signals lib: vdso: Add compat support arm64: compat: Add vDSO arm64: Refactor vDSO code arm64: compat: vDSO setup for compat layer arm64: elf: vDSO code page discovery arm64: compat: Get sigreturn trampolines from vDSO arm64: Add vDSO compat support arm64: Enable compat vDSO support arm: Add support for generic vDSO mips: Introduce vdso_direct clock: csrc-4k: Add support for vdso_direct clock: gic-timer: Add support for vdso_direct mips: Add support for generic vDSO arch/arm/Kconfig | 4 + arch/arm/include/asm/vdso/gettimeofday.h | 79 ++++ arch/arm/include/asm/vdso_datapage.h | 25 +- arch/arm/kernel/vdso.c | 28 +- arch/arm/vdso/Makefile | 13 +- arch/arm/vdso/note.c | 15 + arch/arm/vdso/vdso.lds.S | 2 + arch/arm/vdso/vgettimeofday.c | 263 +---------- arch/arm64/Kconfig | 25 ++ arch/arm64/Makefile | 21 +- arch/arm64/configs/defconfig | 1 + arch/arm64/include/asm/elf.h | 20 +- arch/arm64/include/asm/processor.h | 4 +- arch/arm64/include/asm/signal32.h | 58 ++- arch/arm64/include/asm/unistd.h | 3 + arch/arm64/include/asm/vdso.h | 3 + arch/arm64/include/asm/vdso/compat_barrier.h | 36 ++ .../include/asm/vdso/compat_gettimeofday.h | 89 ++++ arch/arm64/include/asm/vdso/gettimeofday.h | 82 ++++ arch/arm64/include/asm/vdso_datapage.h | 47 -- arch/arm64/kernel/Makefile | 7 +- arch/arm64/kernel/asm-offsets.c | 13 +- arch/arm64/kernel/kuser32.S | 65 +-- arch/arm64/kernel/signal32.c | 76 ++-- arch/arm64/kernel/sigreturn32.S | 46 ++ arch/arm64/kernel/vdso.c | 413 ++++++++++++----- arch/arm64/kernel/vdso/Makefile | 24 +- arch/arm64/kernel/vdso/gettimeofday.S | 328 -------------- arch/arm64/kernel/vdso/vdso.lds.S | 1 + arch/arm64/kernel/vdso/vgettimeofday.c | 30 ++ arch/arm64/kernel/vdso32/.gitignore | 2 + arch/arm64/kernel/vdso32/Makefile | 184 ++++++++ arch/arm64/kernel/vdso32/note.c | 15 + arch/arm64/kernel/vdso32/sigreturn.S | 62 +++ arch/arm64/kernel/vdso32/vdso.S | 19 + arch/arm64/kernel/vdso32/vdso.lds.S | 82 ++++ arch/arm64/kernel/vdso32/vgettimeofday.c | 43 ++ arch/mips/Kconfig | 3 + arch/mips/include/asm/clocksource.h | 1 + arch/mips/include/asm/vdso.h | 53 +-- arch/mips/include/asm/vdso/gettimeofday.h | 161 +++++++ arch/mips/{ => include/asm}/vdso/vdso.h | 6 +- arch/mips/kernel/csrc-r4k.c | 4 +- arch/mips/kernel/vdso.c | 56 ++- arch/mips/vdso/Makefile | 23 +- arch/mips/vdso/elf.S | 2 +- arch/mips/vdso/gettimeofday.c | 273 ------------ arch/mips/vdso/sigreturn.S | 2 +- arch/mips/vdso/vdso.lds.S | 2 + arch/mips/vdso/vgettimeofday.c | 30 ++ drivers/clocksource/mips-gic-timer.c | 5 +- include/vdso/datapage.h | 65 +++ lib/Kconfig | 5 + lib/vdso/Kconfig | 41 ++ lib/vdso/Makefile | 22 + lib/vdso/gettimeofday.c | 418 ++++++++++++++++++ 56 files changed, 2167 insertions(+), 1233 deletions(-) create mode 100644 arch/arm/include/asm/vdso/gettimeofday.h create mode 100644 arch/arm/vdso/note.c create mode 100644 arch/arm64/include/asm/vdso/compat_barrier.h create mode 100644 arch/arm64/include/asm/vdso/compat_gettimeofday.h create mode 100644 arch/arm64/include/asm/vdso/gettimeofday.h delete mode 100644 arch/arm64/include/asm/vdso_datapage.h create mode 100644 arch/arm64/kernel/sigreturn32.S delete mode 100644 arch/arm64/kernel/vdso/gettimeofday.S create mode 100644 arch/arm64/kernel/vdso/vgettimeofday.c create mode 100644 arch/arm64/kernel/vdso32/.gitignore create mode 100644 arch/arm64/kernel/vdso32/Makefile create mode 100644 arch/arm64/kernel/vdso32/note.c create mode 100644 arch/arm64/kernel/vdso32/sigreturn.S create mode 100644 arch/arm64/kernel/vdso32/vdso.S create mode 100644 arch/arm64/kernel/vdso32/vdso.lds.S create mode 100644 arch/arm64/kernel/vdso32/vgettimeofday.c create mode 100644 arch/mips/include/asm/vdso/gettimeofday.h rename arch/mips/{ => include/asm}/vdso/vdso.h (90%) delete mode 100644 arch/mips/vdso/gettimeofday.c create mode 100644 arch/mips/vdso/vgettimeofday.c create mode 100644 include/vdso/datapage.h create mode 100644 lib/vdso/Kconfig create mode 100644 lib/vdso/Makefile create mode 100644 lib/vdso/gettimeofday.c