Re: [PATCH 16/31] nds32: VDSO support

From: Arnd Bergmann
Date: Wed Nov 08 2017 - 04:38:03 EST


On Wed, Nov 8, 2017 at 6:55 AM, Greentime Hu <green.hu@xxxxxxxxx> wrote:

> --- /dev/null
> +++ b/arch/nds32/include/asm/vdso_datapage.h
> @@ -0,0 +1,51 @@

> +#ifndef __ASM_VDSO_DATAPAGE_H
> +#define __ASM_VDSO_DATAPAGE_H
> +
> +#ifdef __KERNEL__
> +
> +#ifndef __ASSEMBLY__
> +
> +struct vdso_data {
> + bool cycle_count_down; /* timer cyclye counter is decrease with time */
> + u32 cycle_count_offset; /* offset of timer cycle counter register */
> + u32 seq_count; /* sequence count - odd during updates */
> + 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 */
> + u32 cs_shift; /* Cycle to nanosecond divisor (power of two) */
> +
> + u64 cs_cycle_last; /* last cycle value */
> + u64 cs_mask; /* clocksource mask */
> +
> + u64 xtime_clock_nsec; /* CLOCK_REALTIME sub-ns base */
> + u32 tz_minuteswest; /* timezone info for gettimeofday(2) */
> + u32 tz_dsttime;
> +};

I need some insight from Deepa and Palmer here: to prepare for 64-bit
time_t in the
future, would it make sense to define the vdso to use 64-bit seconds numbers
consistently, and provide vdso symbols that return 64-bit times, having the
glibc convert that to normal timespec values, or should we leave it for now?

For the normal syscalls I think we are better off keeping things consistent
between architectures, but the vdso is architecture specific by definition, so
we may as well use 64-bit times there now (same for risc-v, which still
has time to modify this before the 4.15 release and glibc merge).

> +/*
> + * This controls what symbols we export from the DSO.
> + */
> +VERSION
> +{
> + LINUX_2.6 {
> + global:
> + __kernel_rt_sigreturn;
> + __vdso_gettimeofday;
> + __vdso_clock_getres;
> + __vdso_clock_gettime;
> + local: *;
> + };
> +}

I still struggle to understand how symbol versioning is supposed to work
in a vdso (as opposed to a library you compile against), but I think this should
use the version from the kernel that you plan to merge into, i.e. LINUX_4
or LINUX_4_16.

Arnd