From a5f7a5bb3b7f28ba7e4fa246775b29a0e5537255 Mon Sep 17 00:00:00 2001 From: Jinke Han Date: Tue, 8 Sep 2026 15:37:42 +0800 Subject: x86/kprobes: Fix crash when probing CS CALL instructions When using eBPF to probe CS CALL instructions within a function, a crash can be triggered. The eBPF tool probes offset 257 of the __hrtimer_run_queues() function: <__hrtimer_run_queues+249>: nopl 0x0(%rax,%rax,1) <__hrtimer_run_queues+254>: mov %r14,%rdi <__hrtimer_run_queues+257>: cs call <__x86_indirect_thunk_r12> <__hrtimer_run_queues+263>: mov %eax,%r12d <__hrtimer_run_queues+266>: xchg %ax,%ax <__hrtimer_run_queues+268>: mov %r13,%rdi Which triggers this crash: BUG: unable to handle page fault for address: 00000000000f41c9 #PF: supervisor write access in kernel mode #PF: error_code(0x0002) - not-present page PGD 0 P4D 0 Oops: 0002 [#1] SMP NOPTI CPU: 1 PID: 0 Comm: swapper/1 Kdump: loaded Tainted: P RIP: 0010:__hrtimer_run_queues+0x106/0x230 Note that __hrtimer_run_queues+0x106 is __hrtimer_run_queues+262, which is at the 6th byte of the above CS CALL instruction. Since the CS CALL instruction occupies 6 bytes, the exception occurred in the middle of that call instruction. The root cause is that when using eBPF tools to probe in the middle of a function, a kprobe with INT3 is used as the underlying implementation. During single-step emulation of the original CALL instruction, int3_emulate_call() assumes that the probed CALL instruction is 5 bytes long. However, the actual CS-prefixed CALL instruction occupies 6 bytes, so it constructs an incorrect exception return address. When the CPU returns from the kprobe handler, the next instruction to be executed is at the address of the last byte of that CS CALL instruction. Coincidentally, starting from that address, the CPU fetches and decodes a completely different instruction, which ultimately triggers a kernel crash. Fix the issue by using the actual instruction length obtained from the instruction decoder when constructing the exception return address, rather than relying on the hardcoded CALL_INSN_SIZE macro. [ mingo: Refined the changelog ] Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step") Suggested-by: Masami Hiramatsu (Google) Signed-off-by: Jinke Han Signed-off-by: Ingo Molnar Reviewed-by: Masami Hiramatsu (Google) Acked-by: Yafang Shao Acked-by: Borislav Petkov Cc: Peter Zijlstra Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000 --- lib/vdso/gettimeofday.c | 513 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 513 insertions(+) create mode 100644 lib/vdso/gettimeofday.c (limited to 'lib/vdso/gettimeofday.c') diff --git a/lib/vdso/gettimeofday.c b/lib/vdso/gettimeofday.c new file mode 100644 index 000000000..f7a591aba --- /dev/null +++ b/lib/vdso/gettimeofday.c @@ -0,0 +1,513 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Generic userspace implementations of gettimeofday() and similar. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +/* + * The generic vDSO implementation requires that gettimeofday.h + * provides: + * - __arch_get_hw_counter(): to get the hw counter based on the + * clock_mode. + * - gettimeofday_fallback(): fallback for gettimeofday. + * - clock_gettime_fallback(): fallback for clock_gettime. + * - clock_getres_fallback(): fallback for clock_getres. + */ +#include + +#include + +/* Bring in default accessors */ +#include + +#ifndef vdso_calc_ns + +#ifdef VDSO_DELTA_NOMASK +# define VDSO_DELTA_MASK(vd) ULLONG_MAX +#else +# define VDSO_DELTA_MASK(vd) (vd->mask) +#endif + +#ifdef CONFIG_GENERIC_VDSO_OVERFLOW_PROTECT +static __always_inline bool vdso_delta_ok(const struct vdso_clock *vc, u64 delta) +{ + return delta < vc->max_cycles; +} +#else +static __always_inline bool vdso_delta_ok(const struct vdso_clock *vc, u64 delta) +{ + return true; +} +#endif + +#ifndef vdso_shift_ns +static __always_inline u64 vdso_shift_ns(u64 ns, u32 shift) +{ + return ns >> shift; +} +#endif + +/* + * Default implementation which works for all sane clocksources. That + * obviously excludes x86/TSC. + */ +static __always_inline u64 vdso_calc_ns(const struct vdso_clock *vc, u64 cycles, u64 base) +{ + u64 delta = (cycles - vc->cycle_last) & VDSO_DELTA_MASK(vc); + + if (likely(vdso_delta_ok(vc, delta))) + return vdso_shift_ns((delta * vc->mult) + base, vc->shift); + + return mul_u64_u32_add_u64_shr(delta, vc->mult, base, vc->shift); +} +#endif /* vdso_calc_ns */ + +#ifndef __arch_vdso_hres_capable +static inline bool __arch_vdso_hres_capable(void) +{ + return true; +} +#endif + +#ifndef vdso_clocksource_ok +static inline bool vdso_clocksource_ok(const struct vdso_clock *vc) +{ + return vc->clock_mode != VDSO_CLOCKMODE_NONE; +} +#endif + +#ifndef vdso_cycles_ok +static inline bool vdso_cycles_ok(u64 cycles) +{ + return true; +} +#endif + +static __always_inline bool vdso_clockid_valid(clockid_t clock) +{ + /* Check for negative values or invalid clocks */ + return likely((u32) clock <= CLOCK_AUX_LAST); +} + +/* + * Must not be invoked within the sequence read section as a race inside + * that loop could result in __iter_div_u64_rem() being extremely slow. + */ +static __always_inline void vdso_set_timespec(struct __kernel_timespec *ts, u64 sec, u64 ns) +{ + ts->tv_sec = sec + __iter_div_u64_rem(ns, NSEC_PER_SEC, &ns); + ts->tv_nsec = ns; +} + +static __always_inline +bool vdso_get_timestamp(const struct vdso_time_data *vd, const struct vdso_clock *vc, + unsigned int clkidx, u64 *sec, u64 *ns) +{ + const struct vdso_timestamp *vdso_ts = &vc->basetime[clkidx]; + u64 cycles; + + if (unlikely(!vdso_clocksource_ok(vc))) + return false; + + cycles = __arch_get_hw_counter(vc->clock_mode, vd); + if (unlikely(!vdso_cycles_ok(cycles))) + return false; + + *ns = vdso_calc_ns(vc, cycles, vdso_ts->nsec); + *sec = vdso_ts->sec; + + return true; +} + +static __always_inline +const struct vdso_time_data *vdso_timens_data(const struct vdso_time_data *vd) +{ + return (void *)vd + PAGE_SIZE; +} + +static __always_inline +bool do_hres_timens(const struct vdso_time_data *vdns, const struct vdso_clock *vcns, + clockid_t clk, struct __kernel_timespec *ts) +{ + const struct vdso_time_data *vd = vdso_timens_data(vdns); + const struct timens_offset *offs = &vcns->offset[clk]; + const struct vdso_clock *vc = vd->clock_data; + u32 seq; + s64 sec; + u64 ns; + + if (clk != CLOCK_MONOTONIC_RAW) + vc = &vc[CS_HRES_COARSE]; + else + vc = &vc[CS_RAW]; + + do { + seq = vdso_read_begin(vc); + + if (!vdso_get_timestamp(vd, vc, clk, &sec, &ns)) + return false; + } while (vdso_read_retry(vc, seq)); + + /* Add the namespace offset */ + sec += offs->sec; + ns += offs->nsec; + + vdso_set_timespec(ts, sec, ns); + + return true; +} + +static __always_inline +bool do_hres(const struct vdso_time_data *vd, const struct vdso_clock *vc, + clockid_t clk, struct __kernel_timespec *ts) +{ + u64 sec, ns; + u32 seq; + + /* Allows to compile the high resolution parts out */ + if (!__arch_vdso_hres_capable()) + return false; + + do { + if (vdso_read_begin_timens(vc, &seq)) + return do_hres_timens(vd, vc, clk, ts); + + if (!vdso_get_timestamp(vd, vc, clk, &sec, &ns)) + return false; + } while (vdso_read_retry(vc, seq)); + + vdso_set_timespec(ts, sec, ns); + + return true; +} + +static __always_inline +bool do_coarse_timens(const struct vdso_time_data *vdns, const struct vdso_clock *vcns, + clockid_t clk, struct __kernel_timespec *ts) +{ + const struct vdso_time_data *vd = vdso_timens_data(vdns); + const struct timens_offset *offs = &vcns->offset[clk]; + const struct vdso_clock *vc = vd->clock_data; + const struct vdso_timestamp *vdso_ts; + u64 nsec; + s64 sec; + s32 seq; + + vdso_ts = &vc->basetime[clk]; + + do { + seq = vdso_read_begin(vc); + sec = vdso_ts->sec; + nsec = vdso_ts->nsec; + } while (vdso_read_retry(vc, seq)); + + /* Add the namespace offset */ + sec += offs->sec; + nsec += offs->nsec; + + vdso_set_timespec(ts, sec, nsec); + + return true; +} + +static __always_inline +bool do_coarse(const struct vdso_time_data *vd, const struct vdso_clock *vc, + clockid_t clk, struct __kernel_timespec *ts) +{ + const struct vdso_timestamp *vdso_ts = &vc->basetime[clk]; + u32 seq; + + do { + if (vdso_read_begin_timens(vc, &seq)) + return do_coarse_timens(vd, vc, clk, ts); + + ts->tv_sec = vdso_ts->sec; + ts->tv_nsec = vdso_ts->nsec; + } while (vdso_read_retry(vc, seq)); + + return true; +} + +static __always_inline +bool do_aux(const struct vdso_time_data *vd, clockid_t clock, struct __kernel_timespec *ts) +{ + const struct vdso_clock *vc; + u32 seq, idx; + u64 sec, ns; + + if (!IS_ENABLED(CONFIG_POSIX_AUX_CLOCKS)) + return false; + + idx = clock - CLOCK_AUX; + vc = &vd->aux_clock_data[idx]; + + do { + while (vdso_read_begin_timens(vc, &seq)) { + /* Re-read from the real time data page, reload seq by looping */ + vd = vdso_timens_data(vd); + vc = &vd->aux_clock_data[idx]; + } + + /* Auxclock disabled? */ + if (vc->clock_mode == VDSO_CLOCKMODE_NONE) + return false; + + if (!vdso_get_timestamp(vd, vc, VDSO_BASE_AUX, &sec, &ns)) + return false; + } while (vdso_read_retry(vc, seq)); + + vdso_set_timespec(ts, sec, ns); + + return true; +} + +static __always_inline bool +__cvdso_clock_gettime_common(const struct vdso_time_data *vd, clockid_t clock, + struct __kernel_timespec *ts) +{ + const struct vdso_clock *vc = vd->clock_data; + u32 msk; + + if (!vdso_clockid_valid(clock)) + return false; + + /* + * Convert the clockid to a bitmask and use it to check which + * clocks are handled in the VDSO directly. + */ + msk = 1U << clock; + if (likely(msk & VDSO_HRES)) + vc = &vc[CS_HRES_COARSE]; + else if (msk & VDSO_COARSE) + return do_coarse(vd, &vc[CS_HRES_COARSE], clock, ts); + else if (msk & VDSO_RAW) + vc = &vc[CS_RAW]; + else if (msk & VDSO_AUX) + return do_aux(vd, clock, ts); + else + return false; + + return do_hres(vd, vc, clock, ts); +} + +static int +__cvdso_clock_gettime_data(const struct vdso_time_data *vd, clockid_t clock, + struct __kernel_timespec *ts) +{ + bool ok; + + ok = __cvdso_clock_gettime_common(vd, clock, ts); + + if (unlikely(!ok)) + return clock_gettime_fallback(clock, ts); + return 0; +} + +static __maybe_unused int +__cvdso_clock_gettime(clockid_t clock, struct __kernel_timespec *ts) +{ + return __cvdso_clock_gettime_data(__arch_get_vdso_u_time_data(), clock, ts); +} + +#ifdef BUILD_VDSO32 +static int +__cvdso_clock_gettime32_data(const struct vdso_time_data *vd, clockid_t clock, + struct old_timespec32 *res) +{ + struct __kernel_timespec ts; + bool ok; + + BUILD_BUG_ON(!IS_ENABLED(CONFIG_COMPAT_32BIT_TIME)); + + ok = __cvdso_clock_gettime_common(vd, clock, &ts); + + if (unlikely(!ok)) + return clock_gettime32_fallback(clock, res); + + /* For ok == true */ + res->tv_sec = ts.tv_sec; + res->tv_nsec = ts.tv_nsec; + + return 0; +} + +static __maybe_unused int +__cvdso_clock_gettime32(clockid_t clock, struct old_timespec32 *res) +{ + return __cvdso_clock_gettime32_data(__arch_get_vdso_u_time_data(), clock, res); +} +#endif /* BUILD_VDSO32 */ + +static int +__cvdso_gettimeofday_data(const struct vdso_time_data *vd, + struct __kernel_old_timeval *tv, struct timezone *tz) +{ + const struct vdso_clock *vc = vd->clock_data; + +#ifndef __NR_gettimeofday + BUILD_BUG(); +#endif + + BUILD_BUG_ON(sizeof(tv->tv_sec) != 8 && !IS_ENABLED(CONFIG_COMPAT_32BIT_TIME)); + + if (likely(tv != NULL)) { + struct __kernel_timespec ts; + + if (!do_hres(vd, &vc[CS_HRES_COARSE], CLOCK_REALTIME, &ts)) + return gettimeofday_fallback(tv, tz); + + tv->tv_sec = ts.tv_sec; + tv->tv_usec = (u32)ts.tv_nsec / NSEC_PER_USEC; + } + + if (unlikely(tz != NULL)) { + if (vdso_is_timens_clock(vc)) + vd = vdso_timens_data(vd); + + tz->tz_minuteswest = vd[CS_HRES_COARSE].tz_minuteswest; + tz->tz_dsttime = vd[CS_HRES_COARSE].tz_dsttime; + } + + return 0; +} + +static __maybe_unused int +__cvdso_gettimeofday(struct __kernel_old_timeval *tv, struct timezone *tz) +{ + return __cvdso_gettimeofday_data(__arch_get_vdso_u_time_data(), tv, tz); +} + +#ifdef VDSO_HAS_TIME +static __kernel_old_time_t +__cvdso_time_data(const struct vdso_time_data *vd, __kernel_old_time_t *time) +{ + const struct vdso_clock *vc = vd->clock_data; + __kernel_old_time_t t; + +#ifndef __NR_time + BUILD_BUG(); +#endif + + BUILD_BUG_ON(sizeof(*time) != 8 && !IS_ENABLED(CONFIG_COMPAT_32BIT_TIME)); + + if (vdso_is_timens_clock(vc)) { + vd = vdso_timens_data(vd); + vc = vd->clock_data; + } + + t = READ_ONCE(vc[CS_HRES_COARSE].basetime[CLOCK_REALTIME].sec); + + if (time) + *time = t; + + return t; +} + +static __maybe_unused __kernel_old_time_t __cvdso_time(__kernel_old_time_t *time) +{ + return __cvdso_time_data(__arch_get_vdso_u_time_data(), time); +} +#endif /* VDSO_HAS_TIME */ + +#ifdef VDSO_HAS_CLOCK_GETRES +static __always_inline +bool __cvdso_clock_getres_common(const struct vdso_time_data *vd, clockid_t clock, + struct __kernel_timespec *res) +{ + const struct vdso_clock *vc = vd->clock_data; + u32 msk; + u64 ns; + + if (!vdso_clockid_valid(clock)) + return false; + + if (vdso_is_timens_clock(vc)) + vd = vdso_timens_data(vd); + + /* + * Convert the clockid to a bitmask and use it to check which + * clocks are handled in the VDSO directly. + */ + msk = 1U << clock; + if (msk & (VDSO_HRES | VDSO_RAW)) { + /* + * Preserves the behaviour of posix_get_hrtimer_res(). + */ + ns = READ_ONCE(vd->hrtimer_res); + } else if (msk & VDSO_COARSE) { + /* + * Preserves the behaviour of posix_get_coarse_res(). + */ + ns = LOW_RES_NSEC; + } else if (msk & VDSO_AUX) { + ns = aux_clock_resolution_ns(); + } else { + return false; + } + + if (likely(res)) { + res->tv_sec = 0; + res->tv_nsec = ns; + } + return true; +} + +static +int __cvdso_clock_getres_data(const struct vdso_time_data *vd, clockid_t clock, + struct __kernel_timespec *res) +{ + bool ok; + + ok = __cvdso_clock_getres_common(vd, clock, res); + + if (unlikely(!ok)) + return clock_getres_fallback(clock, res); + return 0; +} + +static __maybe_unused +int __cvdso_clock_getres(clockid_t clock, struct __kernel_timespec *res) +{ + return __cvdso_clock_getres_data(__arch_get_vdso_u_time_data(), clock, res); +} + +#ifdef BUILD_VDSO32 +static int +__cvdso_clock_getres_time32_data(const struct vdso_time_data *vd, clockid_t clock, + struct old_timespec32 *res) +{ + struct __kernel_timespec ts; + bool ok; + + BUILD_BUG_ON(!IS_ENABLED(CONFIG_COMPAT_32BIT_TIME)); + + ok = __cvdso_clock_getres_common(vd, clock, &ts); + + if (unlikely(!ok)) + return clock_getres32_fallback(clock, res); + + if (likely(res)) { + res->tv_sec = ts.tv_sec; + res->tv_nsec = ts.tv_nsec; + } + return 0; +} + +static __maybe_unused int +__cvdso_clock_getres_time32(clockid_t clock, struct old_timespec32 *res) +{ + return __cvdso_clock_getres_time32_data(__arch_get_vdso_u_time_data(), + clock, res); +} +#endif /* BUILD_VDSO32 */ +#endif /* VDSO_HAS_CLOCK_GETRES */ -- cgit v1.3.1