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 --- drivers/net/phy/swphy.c | 175 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 175 insertions(+) create mode 100644 drivers/net/phy/swphy.c (limited to 'drivers/net/phy/swphy.c') diff --git a/drivers/net/phy/swphy.c b/drivers/net/phy/swphy.c new file mode 100644 index 000000000..59f1ba4d4 --- /dev/null +++ b/drivers/net/phy/swphy.c @@ -0,0 +1,175 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Software PHY emulation + * + * Code taken from fixed_phy.c by Russell King. + * + * Author: Vitaly Bordug + * Anton Vorontsov + * + * Copyright (c) 2006-2007 MontaVista Software, Inc. + */ +#include +#include +#include +#include + +#include "swphy.h" + +#define MII_REGS_NUM 29 + +struct swmii_regs { + u16 bmsr; + u16 lpa; + u16 lpagb; + u16 estat; +}; + +enum { + SWMII_SPEED_10 = 0, + SWMII_SPEED_100, + SWMII_SPEED_1000, + SWMII_DUPLEX_HALF = 0, + SWMII_DUPLEX_FULL, +}; + +/* + * These two tables get bitwise-anded together to produce the final result. + * This means the speed table must contain both duplex settings, and the + * duplex table must contain all speed settings. + */ +static const struct swmii_regs speed[] = { + [SWMII_SPEED_10] = { + .lpa = LPA_10FULL | LPA_10HALF, + }, + [SWMII_SPEED_100] = { + .bmsr = BMSR_100FULL | BMSR_100HALF, + .lpa = LPA_100FULL | LPA_100HALF, + }, + [SWMII_SPEED_1000] = { + .bmsr = BMSR_ESTATEN, + .lpagb = LPA_1000FULL | LPA_1000HALF, + .estat = ESTATUS_1000_TFULL | ESTATUS_1000_THALF, + }, +}; + +static const struct swmii_regs duplex[] = { + [SWMII_DUPLEX_HALF] = { + .bmsr = BMSR_ESTATEN | BMSR_100HALF, + .lpa = LPA_10HALF | LPA_100HALF, + .lpagb = LPA_1000HALF, + .estat = ESTATUS_1000_THALF, + }, + [SWMII_DUPLEX_FULL] = { + .bmsr = BMSR_ESTATEN | BMSR_100FULL, + .lpa = LPA_10FULL | LPA_100FULL, + .lpagb = LPA_1000FULL, + .estat = ESTATUS_1000_TFULL, + }, +}; + +static int swphy_decode_speed(int speed) +{ + switch (speed) { + case 1000: + return SWMII_SPEED_1000; + case 100: + return SWMII_SPEED_100; + case 10: + return SWMII_SPEED_10; + default: + return -EINVAL; + } +} + +/** + * swphy_validate_state - validate the software phy status + * @state: software phy status + * + * This checks that we can represent the state stored in @state can be + * represented in the emulated MII registers. Returns 0 if it can, + * otherwise returns -EINVAL. + */ +int swphy_validate_state(const struct fixed_phy_status *state) +{ + int err; + + if (state->link) { + err = swphy_decode_speed(state->speed); + if (err < 0) { + pr_warn("swphy: unknown speed\n"); + return -EINVAL; + } + } + return 0; +} +EXPORT_SYMBOL_GPL(swphy_validate_state); + +/** + * swphy_read_reg - return a MII register from the fixed phy state + * @reg: MII register + * @state: fixed phy status + * + * Return the MII @reg register generated from the fixed phy state @state. + */ +int swphy_read_reg(int reg, const struct fixed_phy_status *state) +{ + int speed_index, duplex_index; + u16 bmsr = BMSR_ANEGCAPABLE; + u16 estat = 0; + u16 lpagb = 0; + u16 lpa = 0; + + if (reg > MII_REGS_NUM) + return -1; + + speed_index = swphy_decode_speed(state->speed); + if (WARN_ON(speed_index < 0)) + return 0; + + duplex_index = state->duplex ? SWMII_DUPLEX_FULL : SWMII_DUPLEX_HALF; + + bmsr |= speed[speed_index].bmsr & duplex[duplex_index].bmsr; + estat |= speed[speed_index].estat & duplex[duplex_index].estat; + + if (state->link) { + bmsr |= BMSR_LSTATUS | BMSR_ANEGCOMPLETE; + + lpa |= speed[speed_index].lpa & duplex[duplex_index].lpa; + lpagb |= speed[speed_index].lpagb & duplex[duplex_index].lpagb; + + if (state->pause) + lpa |= LPA_PAUSE_CAP; + + if (state->asym_pause) + lpa |= LPA_PAUSE_ASYM; + } + + switch (reg) { + case MII_BMCR: + return BMCR_ANENABLE; + case MII_BMSR: + return bmsr; + case MII_PHYSID1: + case MII_PHYSID2: + return 0; + case MII_LPA: + return lpa; + case MII_STAT1000: + return lpagb; + case MII_ESTATUS: + return estat; + + /* + * We do not support emulating Clause 45 over Clause 22 register + * reads. Return an error instead of bogus data. + */ + case MII_MMD_CTRL: + case MII_MMD_DATA: + return -1; + + default: + return 0xffff; + } +} +EXPORT_SYMBOL_GPL(swphy_read_reg); -- cgit v1.3.1