2f6c9bf31a
The current VDSO patch mechanism has several problems: 1) It assumes how gcc will emit a function, with a register window, an initial save instruction and then immediately the %tick read when compiling vread_tick(). There is no such guarantees, code generation could change at any time, gcc could put a nop between the save and the %tick read, etc. So this is extremely fragile and would fail some day. 2) It disallows us to properly inline vread_tick() into the callers and thus get the best possible code sequences. So fix this to patch properly, with location based annotations. We have to be careful because we cannot do it the way we do patches elsewhere in the kernel. Those use a sequence like: 1: insn .section .whatever_patch, "ax" .word 1b replacement_insn .previous This is a dynamic shared object, so that .word cannot be resolved at build time, and thus cannot be used to execute the patches when the kernel initializes the images. Even trying to use label difference equations doesn't work in the above kind of scheme: 1: insn .section .whatever_patch, "ax" .word . - 1b replacement_insn .previous The assembler complains that it cannot resolve that computation. The issue is that this is contained in an executable section. Borrow the sequence used by x86 alternatives, which is: 1: insn .pushsection .whatever_patch, "a" .word . - 1b, . - 1f .popsection .pushsection .whatever_patch_replacements, "ax" 1: replacement_insn .previous This works, allows us to inline vread_tick() as much as we like, and can be used for arbitrary kinds of VDSO patching in the future. Also, reverse the condition for patching. Most systems are %stick based, so if we only patch on %tick systems the patching code will get little or no testing. Signed-off-by: David S. Miller <davem@davemloft.net>
282 lines
6.7 KiB
C
282 lines
6.7 KiB
C
/*
|
|
* Copyright 2006 Andi Kleen, SUSE Labs.
|
|
* Subject to the GNU Public License, v.2
|
|
*
|
|
* Fast user context implementation of clock_gettime, gettimeofday, and time.
|
|
*
|
|
* The code should have no internal unresolved relocations.
|
|
* Check with readelf after changing.
|
|
* Also alternative() doesn't work.
|
|
*/
|
|
/*
|
|
* Copyright (c) 2017 Oracle and/or its affiliates. All rights reserved.
|
|
*/
|
|
|
|
/* Disable profiling for userspace code: */
|
|
#ifndef DISABLE_BRANCH_PROFILING
|
|
#define DISABLE_BRANCH_PROFILING
|
|
#endif
|
|
|
|
#include <linux/kernel.h>
|
|
#include <linux/time.h>
|
|
#include <linux/string.h>
|
|
#include <asm/io.h>
|
|
#include <asm/unistd.h>
|
|
#include <asm/timex.h>
|
|
#include <asm/clocksource.h>
|
|
#include <asm/vvar.h>
|
|
|
|
#undef TICK_PRIV_BIT
|
|
#ifdef CONFIG_SPARC64
|
|
#define TICK_PRIV_BIT (1UL << 63)
|
|
#else
|
|
#define TICK_PRIV_BIT (1ULL << 63)
|
|
#endif
|
|
|
|
#ifdef CONFIG_SPARC64
|
|
#define SYSCALL_STRING \
|
|
"ta 0x6d;" \
|
|
"bcs,a 1f;" \
|
|
" sub %%g0, %%o0, %%o0;" \
|
|
"1:"
|
|
#else
|
|
#define SYSCALL_STRING \
|
|
"ta 0x10;" \
|
|
"bcs,a 1f;" \
|
|
" sub %%g0, %%o0, %%o0;" \
|
|
"1:"
|
|
#endif
|
|
|
|
#define SYSCALL_CLOBBERS \
|
|
"f0", "f1", "f2", "f3", "f4", "f5", "f6", "f7", \
|
|
"f8", "f9", "f10", "f11", "f12", "f13", "f14", "f15", \
|
|
"f16", "f17", "f18", "f19", "f20", "f21", "f22", "f23", \
|
|
"f24", "f25", "f26", "f27", "f28", "f29", "f30", "f31", \
|
|
"f32", "f34", "f36", "f38", "f40", "f42", "f44", "f46", \
|
|
"f48", "f50", "f52", "f54", "f56", "f58", "f60", "f62", \
|
|
"cc", "memory"
|
|
|
|
/*
|
|
* Compute the vvar page's address in the process address space, and return it
|
|
* as a pointer to the vvar_data.
|
|
*/
|
|
static notrace noinline struct vvar_data *
|
|
get_vvar_data(void)
|
|
{
|
|
unsigned long ret;
|
|
|
|
/*
|
|
* vdso data page is the first vDSO page so grab the return address
|
|
* and move up a page to get to the data page.
|
|
*/
|
|
ret = (unsigned long)__builtin_return_address(0);
|
|
ret &= ~(8192 - 1);
|
|
ret -= 8192;
|
|
|
|
return (struct vvar_data *) ret;
|
|
}
|
|
|
|
static notrace long
|
|
vdso_fallback_gettime(long clock, struct timespec *ts)
|
|
{
|
|
register long num __asm__("g1") = __NR_clock_gettime;
|
|
register long o0 __asm__("o0") = clock;
|
|
register long o1 __asm__("o1") = (long) ts;
|
|
|
|
__asm__ __volatile__(SYSCALL_STRING : "=r" (o0) : "r" (num),
|
|
"0" (o0), "r" (o1) : SYSCALL_CLOBBERS);
|
|
return o0;
|
|
}
|
|
|
|
static notrace __always_inline long
|
|
vdso_fallback_gettimeofday(struct timeval *tv, struct timezone *tz)
|
|
{
|
|
register long num __asm__("g1") = __NR_gettimeofday;
|
|
register long o0 __asm__("o0") = (long) tv;
|
|
register long o1 __asm__("o1") = (long) tz;
|
|
|
|
__asm__ __volatile__(SYSCALL_STRING : "=r" (o0) : "r" (num),
|
|
"0" (o0), "r" (o1) : SYSCALL_CLOBBERS);
|
|
return o0;
|
|
}
|
|
|
|
#ifdef CONFIG_SPARC64
|
|
static notrace noinline u64
|
|
vread_tick(void) {
|
|
u64 ret;
|
|
|
|
__asm__ __volatile__("1:\n\t"
|
|
"rd %%tick, %0\n\t"
|
|
".pushsection .tick_patch, \"a\"\n\t"
|
|
".word 1b - ., 1f - .\n\t"
|
|
".popsection\n\t"
|
|
".pushsection .tick_patch_replacement, \"ax\"\n\t"
|
|
"1:\n\t"
|
|
"rd %%asr24, %0\n\t"
|
|
".popsection\n"
|
|
: "=r" (ret));
|
|
return ret & ~TICK_PRIV_BIT;
|
|
}
|
|
#else
|
|
static notrace noinline u64
|
|
vread_tick(void)
|
|
{
|
|
register unsigned long long ret asm("o4");
|
|
|
|
__asm__ __volatile__("1:\n\t"
|
|
"rd %%tick, %L0\n\t"
|
|
"srlx %L0, 32, %H0\n\t"
|
|
".pushsection .tick_patch, \"a\"\n\t"
|
|
".word 1b - ., 1f - .\n\t"
|
|
".popsection\n\t"
|
|
".pushsection .tick_patch_replacement, \"ax\"\n\t"
|
|
"1:\n\t"
|
|
"rd %%asr24, %L0\n\t"
|
|
".popsection\n"
|
|
: "=r" (ret));
|
|
return ret;
|
|
}
|
|
#endif
|
|
|
|
static notrace inline u64
|
|
vgetsns(struct vvar_data *vvar)
|
|
{
|
|
u64 v;
|
|
u64 cycles;
|
|
|
|
cycles = vread_tick();
|
|
v = (cycles - vvar->clock.cycle_last) & vvar->clock.mask;
|
|
return v * vvar->clock.mult;
|
|
}
|
|
|
|
static notrace noinline int
|
|
do_realtime(struct vvar_data *vvar, struct timespec *ts)
|
|
{
|
|
unsigned long seq;
|
|
u64 ns;
|
|
|
|
ts->tv_nsec = 0;
|
|
do {
|
|
seq = vvar_read_begin(vvar);
|
|
ts->tv_sec = vvar->wall_time_sec;
|
|
ns = vvar->wall_time_snsec;
|
|
ns += vgetsns(vvar);
|
|
ns >>= vvar->clock.shift;
|
|
} while (unlikely(vvar_read_retry(vvar, seq)));
|
|
|
|
timespec_add_ns(ts, ns);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static notrace noinline int
|
|
do_monotonic(struct vvar_data *vvar, struct timespec *ts)
|
|
{
|
|
unsigned long seq;
|
|
u64 ns;
|
|
|
|
ts->tv_nsec = 0;
|
|
do {
|
|
seq = vvar_read_begin(vvar);
|
|
ts->tv_sec = vvar->monotonic_time_sec;
|
|
ns = vvar->monotonic_time_snsec;
|
|
ns += vgetsns(vvar);
|
|
ns >>= vvar->clock.shift;
|
|
} while (unlikely(vvar_read_retry(vvar, seq)));
|
|
|
|
timespec_add_ns(ts, ns);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static notrace noinline int
|
|
do_realtime_coarse(struct vvar_data *vvar, struct timespec *ts)
|
|
{
|
|
unsigned long seq;
|
|
|
|
do {
|
|
seq = vvar_read_begin(vvar);
|
|
ts->tv_sec = vvar->wall_time_coarse_sec;
|
|
ts->tv_nsec = vvar->wall_time_coarse_nsec;
|
|
} while (unlikely(vvar_read_retry(vvar, seq)));
|
|
return 0;
|
|
}
|
|
|
|
static notrace noinline int
|
|
do_monotonic_coarse(struct vvar_data *vvar, struct timespec *ts)
|
|
{
|
|
unsigned long seq;
|
|
|
|
do {
|
|
seq = vvar_read_begin(vvar);
|
|
ts->tv_sec = vvar->monotonic_time_coarse_sec;
|
|
ts->tv_nsec = vvar->monotonic_time_coarse_nsec;
|
|
} while (unlikely(vvar_read_retry(vvar, seq)));
|
|
|
|
return 0;
|
|
}
|
|
|
|
notrace int
|
|
__vdso_clock_gettime(clockid_t clock, struct timespec *ts)
|
|
{
|
|
struct vvar_data *vvd = get_vvar_data();
|
|
|
|
switch (clock) {
|
|
case CLOCK_REALTIME:
|
|
if (unlikely(vvd->vclock_mode == VCLOCK_NONE))
|
|
break;
|
|
return do_realtime(vvd, ts);
|
|
case CLOCK_MONOTONIC:
|
|
if (unlikely(vvd->vclock_mode == VCLOCK_NONE))
|
|
break;
|
|
return do_monotonic(vvd, ts);
|
|
case CLOCK_REALTIME_COARSE:
|
|
return do_realtime_coarse(vvd, ts);
|
|
case CLOCK_MONOTONIC_COARSE:
|
|
return do_monotonic_coarse(vvd, ts);
|
|
}
|
|
/*
|
|
* Unknown clock ID ? Fall back to the syscall.
|
|
*/
|
|
return vdso_fallback_gettime(clock, ts);
|
|
}
|
|
int
|
|
clock_gettime(clockid_t, struct timespec *)
|
|
__attribute__((weak, alias("__vdso_clock_gettime")));
|
|
|
|
notrace int
|
|
__vdso_gettimeofday(struct timeval *tv, struct timezone *tz)
|
|
{
|
|
struct vvar_data *vvd = get_vvar_data();
|
|
|
|
if (likely(vvd->vclock_mode != VCLOCK_NONE)) {
|
|
if (likely(tv != NULL)) {
|
|
union tstv_t {
|
|
struct timespec ts;
|
|
struct timeval tv;
|
|
} *tstv = (union tstv_t *) tv;
|
|
do_realtime(vvd, &tstv->ts);
|
|
/*
|
|
* Assign before dividing to ensure that the division is
|
|
* done in the type of tv_usec, not tv_nsec.
|
|
*
|
|
* There cannot be > 1 billion usec in a second:
|
|
* do_realtime() has already distributed such overflow
|
|
* into tv_sec. So we can assign it to an int safely.
|
|
*/
|
|
tstv->tv.tv_usec = tstv->ts.tv_nsec;
|
|
tstv->tv.tv_usec /= 1000;
|
|
}
|
|
if (unlikely(tz != NULL)) {
|
|
/* Avoid memcpy. Some old compilers fail to inline it */
|
|
tz->tz_minuteswest = vvd->tz_minuteswest;
|
|
tz->tz_dsttime = vvd->tz_dsttime;
|
|
}
|
|
return 0;
|
|
}
|
|
return vdso_fallback_gettimeofday(tv, tz);
|
|
}
|
|
int
|
|
gettimeofday(struct timeval *, struct timezone *)
|
|
__attribute__((weak, alias("__vdso_gettimeofday")));
|