| Message ID | 1403493118-7597-7-git-send-email-nathan_lynch@mentor.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Monday 23 June 2014, Nathan Lynch wrote: > +struct vdso_data { > + u32 seq_count; /* sequence count - odd during updates */ > + u16 use_syscall; /* whether to fall back to syscalls */ > + u16 cs_shift; /* clocksource shift */ > + u32 xtime_coarse_sec; /* coarse time */ > + u32 xtime_coarse_nsec; Note that we will at some point have to introduce 64-bit time_t in user space in some form. We will have to add new system calls for anything dealing with time on the kernel boundary, but it would be good to be prepared here. Is vdso_data an ABI-relevant data structure? If so, I think all you need here is to make xtime_coarse_sec a u64 type for now, so we can add the other functions later. The code can of course for now access only the lower half (keeping endianess in mind), so there should not be any performance impact. If the structure is not part of the ABI, there probably isn't much we can do here. Arnd
On 06/30/2014 05:11 AM, Arnd Bergmann wrote: > On Monday 23 June 2014, Nathan Lynch wrote: >> +struct vdso_data { >> + u32 seq_count; /* sequence count - odd during updates */ >> + u16 use_syscall; /* whether to fall back to syscalls */ >> + u16 cs_shift; /* clocksource shift */ >> + u32 xtime_coarse_sec; /* coarse time */ >> + u32 xtime_coarse_nsec; > > Note that we will at some point have to introduce 64-bit time_t in user space > in some form. We will have to add new system calls for anything dealing with > time on the kernel boundary, but it would be good to be prepared here. > > Is vdso_data an ABI-relevant data structure? If so, I think all you need here > is to make xtime_coarse_sec a u64 type for now, so we can add the other functions > later. The code can of course for now access only the lower half (keeping > endianess in mind), so there should not be any performance impact. > > If the structure is not part of the ABI, there probably isn't much we can > do here. In short, no, vdso_data is not ABI-relevant. The vdso_data structure may be accessed legitimately only through the entry points (gettimeofday, clock_gettime) that the VDSO provides, and it is my intent that we should be able to change the structure as needed, just like other architectures. vdso_data should not show up in any exported header and I'll double-check to make sure it does not. So, when the time comes, we should be able to change it to accommodate 64-bit time_t by (I imagine) making any field that stores seconds 64 bits and adding new entry points (clock_gettime64 or whatever).
On Monday 30 June 2014 07:27:59 Nathan Lynch wrote: > On 06/30/2014 05:11 AM, Arnd Bergmann wrote: > > On Monday 23 June 2014, Nathan Lynch wrote: > >> +struct vdso_data { > >> + u32 seq_count; /* sequence count - odd during updates */ > >> + u16 use_syscall; /* whether to fall back to syscalls */ > >> + u16 cs_shift; /* clocksource shift */ > >> + u32 xtime_coarse_sec; /* coarse time */ > >> + u32 xtime_coarse_nsec; > > > > Note that we will at some point have to introduce 64-bit time_t in user space > > in some form. We will have to add new system calls for anything dealing with > > time on the kernel boundary, but it would be good to be prepared here. > > > > Is vdso_data an ABI-relevant data structure? If so, I think all you need here > > is to make xtime_coarse_sec a u64 type for now, so we can add the other functions > > later. The code can of course for now access only the lower half (keeping > > endianess in mind), so there should not be any performance impact. > > > > If the structure is not part of the ABI, there probably isn't much we can > > do here. > > In short, no, vdso_data is not ABI-relevant. > > The vdso_data structure may be accessed legitimately only through the > entry points (gettimeofday, clock_gettime) that the VDSO provides, and > it is my intent that we should be able to change the structure as > needed, just like other architectures. vdso_data should not show up in > any exported header and I'll double-check to make sure it does not. > > So, when the time comes, we should be able to change it to accommodate > 64-bit time_t by (I imagine) making any field that stores seconds 64 > bits and adding new entry points (clock_gettime64 or whatever). Ok, very good. Arnd
diff --git a/arch/arm/include/asm/Kbuild b/arch/arm/include/asm/Kbuild index f5a357601983..0e70f40e04ef 100644 --- a/arch/arm/include/asm/Kbuild +++ b/arch/arm/include/asm/Kbuild @@ -1,6 +1,5 @@ -generic-y += auxvec.h generic-y += bitsperlong.h generic-y += cputime.h generic-y += current.h diff --git a/arch/arm/include/asm/elf.h b/arch/arm/include/asm/elf.h index f4b46d39b9cf..45d2ddff662a 100644 --- a/arch/arm/include/asm/elf.h +++ b/arch/arm/include/asm/elf.h @@ -1,7 +1,9 @@ #ifndef __ASMARM_ELF_H #define __ASMARM_ELF_H +#include <asm/auxvec.h> #include <asm/hwcap.h> +#include <asm/vdso_datapage.h> /* * ELF register definitions.. @@ -129,6 +131,15 @@ extern unsigned long arch_randomize_brk(struct mm_struct *mm); #define arch_randomize_brk arch_randomize_brk #ifdef CONFIG_MMU +#ifdef CONFIG_VDSO +#define ARCH_DLINFO \ +do { \ + /* Account for the data page at the beginning of the [vdso] VMA. */ \ + NEW_AUX_ENT(AT_SYSINFO_EHDR, \ + (elf_addr_t)current->mm->context.vdso + \ + sizeof(union vdso_data_store)); \ +} while (0) +#endif #define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1 struct linux_binprm; int arch_setup_additional_pages(struct linux_binprm *, int); diff --git a/arch/arm/include/asm/mmu.h b/arch/arm/include/asm/mmu.h index 64fd15159b7d..a5b47421059d 100644 --- a/arch/arm/include/asm/mmu.h +++ b/arch/arm/include/asm/mmu.h @@ -11,6 +11,9 @@ typedef struct { #endif unsigned int vmalloc_seq; unsigned long sigpage; +#ifdef CONFIG_VDSO + unsigned long vdso; +#endif } mm_context_t; #ifdef CONFIG_CPU_HAS_ASID diff --git a/arch/arm/include/asm/vdso.h b/arch/arm/include/asm/vdso.h new file mode 100644 index 000000000000..0e2a6154eceb --- /dev/null +++ b/arch/arm/include/asm/vdso.h @@ -0,0 +1,47 @@ +#ifndef __ASM_VDSO_H +#define __ASM_VDSO_H + +#ifdef __KERNEL__ + +#ifndef __ASSEMBLY__ + +#include <linux/mm_types.h> +#include <asm/mmu.h> + +#ifdef CONFIG_VDSO + +static inline bool vma_is_vdso(struct vm_area_struct *vma) +{ + if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso) + return true; + return false; +} + +void arm_install_vdso(struct mm_struct *mm, unsigned long addr); + +extern char vdso_start, vdso_end; + +extern unsigned long vdso_mapping_len; + +#else /* CONFIG_VDSO */ + +static inline bool vma_is_vdso(struct vm_area_struct *vma) +{ + return false; +} + +static inline void arm_install_vdso(struct mm_struct *mm, unsigned long addr) +{ +} + +#define vdso_mapping_len 0 + +#endif /* CONFIG_VDSO */ + +#endif /* __ASSEMBLY__ */ + +#define VDSO_LBASE 0x0 + +#endif /* __KERNEL__ */ + +#endif /* __ASM_VDSO_H */ diff --git a/arch/arm/include/asm/vdso_datapage.h b/arch/arm/include/asm/vdso_datapage.h new file mode 100644 index 000000000000..f08bdb73d3f4 --- /dev/null +++ b/arch/arm/include/asm/vdso_datapage.h @@ -0,0 +1,60 @@ +/* + * Adapted from arm64 version. + * + * Copyright (C) 2012 ARM Limited + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +#ifndef __ASM_VDSO_DATAPAGE_H +#define __ASM_VDSO_DATAPAGE_H + +#ifdef __KERNEL__ + +#ifndef __ASSEMBLY__ + +#include <asm/page.h> + +/* Try to be cache-friendly on systems that don't implement the + * generic timer: fit the unconditionally updated fields in the first + * 32 bytes. + */ +struct vdso_data { + u32 seq_count; /* sequence count - odd during updates */ + u16 use_syscall; /* whether to fall back to syscalls */ + u16 cs_shift; /* clocksource shift */ + u32 xtime_coarse_sec; /* coarse time */ + u32 xtime_coarse_nsec; + + u32 wtm_clock_sec; /* wall to monotonic offset */ + u32 wtm_clock_nsec; + u32 xtime_clock_sec; /* CLOCK_REALTIME - seconds */ + u32 cs_mult; /* clocksource multiplier */ + + u64 cs_cycle_last; /* last cycle value */ + u64 cs_mask; /* clocksource mask */ + + u64 xtime_clock_snsec; /* CLOCK_REALTIME sub-ns base */ + u32 tz_minuteswest; /* timezone info for gettimeofday(2) */ + u32 tz_dsttime; +}; + +union vdso_data_store { + struct vdso_data data; + u8 page[PAGE_SIZE]; +}; + +#endif /* !__ASSEMBLY__ */ + +#endif /* __KERNEL__ */ + +#endif /* __ASM_VDSO_DATAPAGE_H */ diff --git a/arch/arm/include/uapi/asm/Kbuild b/arch/arm/include/uapi/asm/Kbuild index 70a1c9da30ca..a1c05f93d920 100644 --- a/arch/arm/include/uapi/asm/Kbuild +++ b/arch/arm/include/uapi/asm/Kbuild @@ -1,6 +1,7 @@ # UAPI Header export list include include/uapi/asm-generic/Kbuild.asm +header-y += auxvec.h header-y += byteorder.h header-y += fcntl.h header-y += hwcap.h diff --git a/arch/arm/include/uapi/asm/auxvec.h b/arch/arm/include/uapi/asm/auxvec.h new file mode 100644 index 000000000000..f56936b97ec2 --- /dev/null +++ b/arch/arm/include/uapi/asm/auxvec.h @@ -0,0 +1,7 @@ +#ifndef __ASM_AUXVEC_H +#define __ASM_AUXVEC_H + +/* vDSO location */ +#define AT_SYSINFO_EHDR 33 + +#endif diff --git a/arch/arm/kernel/process.c b/arch/arm/kernel/process.c index 219a0d1d10fc..40445fb71ac9 100644 --- a/arch/arm/kernel/process.c +++ b/arch/arm/kernel/process.c @@ -41,6 +41,7 @@ #include <asm/system_misc.h> #include <asm/mach/time.h> #include <asm/tls.h> +#include <asm/vdso.h> #ifdef CONFIG_CC_STACKPROTECTOR #include <linux/stackprotector.h> @@ -472,9 +473,16 @@ int in_gate_area_no_mm(unsigned long addr) const char *arch_vma_name(struct vm_area_struct *vma) { - return is_gate_vma(vma) ? "[vectors]" : - (vma->vm_mm && vma->vm_start == vma->vm_mm->context.sigpage) ? - "[sigpage]" : NULL; + if (is_gate_vma(vma)) + return "[vectors]"; + + if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.sigpage) + return "[sigpage]"; + + if (vma_is_vdso(vma)) + return "[vdso]"; + + return NULL; } static unsigned long vdso_addr(const struct mm_struct *mm, unsigned int npages)
Define the layout of the data structure shared between kernel and userspace. Track the vdso address in the mm_context; needed for reporting [vdso] via /proc/pid/maps and communicating AT_SYSINFO_EHDR to the ELF loader. Add declarations for vma_is_vdso, arm_install_vdso, and !CONFIG_VDSO implementations of same. The CONFIG_VDSO=y implementations of these come later. Make arch_vma_name recognize a vdso VMA, using vma_is_vdso to return "[vdso]" when appropriate. Define AT_SYSINFO_EHDR, and, if CONFIG_VDSO=y, report the vdso shared object address via the ELF auxiliary vector. Note - this adds the AT_SYSINFO_EHDR in a new user-visible header asm/auxvec.h; this is consistent with other architectures. Signed-off-by: Nathan Lynch <nathan_lynch@mentor.com> --- arch/arm/include/asm/Kbuild | 1 - arch/arm/include/asm/elf.h | 11 +++++++ arch/arm/include/asm/mmu.h | 3 ++ arch/arm/include/asm/vdso.h | 47 ++++++++++++++++++++++++++++ arch/arm/include/asm/vdso_datapage.h | 60 ++++++++++++++++++++++++++++++++++++ arch/arm/include/uapi/asm/Kbuild | 1 + arch/arm/include/uapi/asm/auxvec.h | 7 +++++ arch/arm/kernel/process.c | 14 +++++++-- 8 files changed, 140 insertions(+), 4 deletions(-) create mode 100644 arch/arm/include/asm/vdso.h create mode 100644 arch/arm/include/asm/vdso_datapage.h create mode 100644 arch/arm/include/uapi/asm/auxvec.h