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/misc/mei/debugfs.c | 195 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 195 insertions(+) create mode 100644 drivers/misc/mei/debugfs.c (limited to 'drivers/misc/mei/debugfs.c') diff --git a/drivers/misc/mei/debugfs.c b/drivers/misc/mei/debugfs.c new file mode 100644 index 000000000..3b098d4c8 --- /dev/null +++ b/drivers/misc/mei/debugfs.c @@ -0,0 +1,195 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (c) 2012-2022, Intel Corporation. All rights reserved + * Intel Management Engine Interface (Intel MEI) Linux driver + */ + +#include +#include +#include +#include +#include + +#include + +#include "mei_dev.h" +#include "client.h" +#include "hw.h" + +static int mei_dbgfs_meclients_show(struct seq_file *m, void *unused) +{ + struct mei_device *dev = m->private; + struct mei_me_client *me_cl; + int i = 0; + + if (!dev) + return -ENODEV; + + down_read(&dev->me_clients_rwsem); + + seq_puts(m, " |id|fix| UUID |con|msg len|sb|refc|vt|\n"); + + /* if the driver is not enabled the list won't be consistent */ + if (dev->dev_state != MEI_DEV_ENABLED) + goto out; + + list_for_each_entry(me_cl, &dev->me_clients, list) { + if (!mei_me_cl_get(me_cl)) + continue; + + seq_printf(m, "%2d|%2d|%3d|%pUl|%3d|%7d|%2d|%4d|%2d|\n", + i++, me_cl->client_id, + me_cl->props.fixed_address, + &me_cl->props.protocol_name, + me_cl->props.max_number_of_connections, + me_cl->props.max_msg_length, + me_cl->props.single_recv_buf, + kref_read(&me_cl->refcnt), + me_cl->props.vt_supported); + mei_me_cl_put(me_cl); + } + +out: + up_read(&dev->me_clients_rwsem); + return 0; +} +DEFINE_SHOW_ATTRIBUTE(mei_dbgfs_meclients); + +static int mei_dbgfs_active_show(struct seq_file *m, void *unused) +{ + struct mei_device *dev = m->private; + struct mei_cl *cl; + int i = 0; + + if (!dev) + return -ENODEV; + + mutex_lock(&dev->device_lock); + + seq_puts(m, " |me|host|state|rd|wr|wrq\n"); + + /* if the driver is not enabled the list won't be consistent */ + if (dev->dev_state != MEI_DEV_ENABLED) + goto out; + + list_for_each_entry(cl, &dev->file_list, link) { + + seq_printf(m, "%3d|%2d|%4d|%5d|%2d|%2d|%3u\n", + i, mei_cl_me_id(cl), cl->host_client_id, cl->state, + !list_empty(&cl->rd_completed), cl->writing_state, + cl->tx_cb_queued); + i++; + } +out: + mutex_unlock(&dev->device_lock); + return 0; +} +DEFINE_SHOW_ATTRIBUTE(mei_dbgfs_active); + +static const char *mei_dev_pxp_mode_str(enum mei_dev_pxp_mode state) +{ +#define MEI_PXP_MODE(state) case MEI_DEV_PXP_##state: return #state + switch (state) { + MEI_PXP_MODE(DEFAULT); + MEI_PXP_MODE(INIT); + MEI_PXP_MODE(SETUP); + MEI_PXP_MODE(READY); + default: + return "unknown"; + } +#undef MEI_PXP_MODE +} + +static int mei_dbgfs_devstate_show(struct seq_file *m, void *unused) +{ + struct mei_device *dev = m->private; + + seq_printf(m, "dev: %s\n", mei_dev_state_str(dev->dev_state)); + seq_printf(m, "hbm: %s\n", mei_hbm_state_str(dev->hbm_state)); + + if (dev->hbm_state >= MEI_HBM_ENUM_CLIENTS && + dev->hbm_state <= MEI_HBM_STARTED) { + seq_puts(m, "hbm features:\n"); + seq_printf(m, "\tPG: %01d\n", dev->hbm_f_pg_supported); + seq_printf(m, "\tDC: %01d\n", dev->hbm_f_dc_supported); + seq_printf(m, "\tIE: %01d\n", dev->hbm_f_ie_supported); + seq_printf(m, "\tDOT: %01d\n", dev->hbm_f_dot_supported); + seq_printf(m, "\tEV: %01d\n", dev->hbm_f_ev_supported); + seq_printf(m, "\tFA: %01d\n", dev->hbm_f_fa_supported); + seq_printf(m, "\tOS: %01d\n", dev->hbm_f_os_supported); + seq_printf(m, "\tDR: %01d\n", dev->hbm_f_dr_supported); + seq_printf(m, "\tVT: %01d\n", dev->hbm_f_vt_supported); + seq_printf(m, "\tCAP: %01d\n", dev->hbm_f_cap_supported); + seq_printf(m, "\tCD: %01d\n", dev->hbm_f_cd_supported); + } + + seq_printf(m, "pg: %s, %s\n", + mei_pg_is_enabled(dev) ? "ENABLED" : "DISABLED", + mei_pg_state_str(mei_pg_state(dev))); + + seq_printf(m, "pxp: %s\n", mei_dev_pxp_mode_str(dev->pxp_mode)); + + return 0; +} +DEFINE_SHOW_ATTRIBUTE(mei_dbgfs_devstate); + +static ssize_t mei_dbgfs_write_allow_fa(struct file *file, + const char __user *user_buf, + size_t count, loff_t *ppos) +{ + struct mei_device *dev; + int ret; + + dev = container_of(file->private_data, + struct mei_device, allow_fixed_address); + + ret = debugfs_write_file_bool(file, user_buf, count, ppos); + if (ret < 0) + return ret; + dev->override_fixed_address = true; + return ret; +} + +static const struct file_operations mei_dbgfs_allow_fa_fops = { + .open = simple_open, + .read = debugfs_read_file_bool, + .write = mei_dbgfs_write_allow_fa, + .llseek = generic_file_llseek, +}; + +/** + * mei_dbgfs_deregister - Remove the debugfs files and directories + * + * @dev: the mei device structure + */ +void mei_dbgfs_deregister(struct mei_device *dev) +{ + if (!dev->dbgfs_dir) + return; + debugfs_remove_recursive(dev->dbgfs_dir); + dev->dbgfs_dir = NULL; +} + +/** + * mei_dbgfs_register - Add the debugfs files + * + * @dev: the mei device structure + * @name: the mei device name + */ +void mei_dbgfs_register(struct mei_device *dev, const char *name) +{ + struct dentry *dir; + + dir = debugfs_create_dir(name, NULL); + dev->dbgfs_dir = dir; + + debugfs_create_file("meclients", S_IRUSR, dir, dev, + &mei_dbgfs_meclients_fops); + debugfs_create_file("active", S_IRUSR, dir, dev, + &mei_dbgfs_active_fops); + debugfs_create_file("devstate", S_IRUSR, dir, dev, + &mei_dbgfs_devstate_fops); + debugfs_create_file("allow_fixed_address", S_IRUSR | S_IWUSR, dir, + &dev->allow_fixed_address, + &mei_dbgfs_allow_fa_fops); +} -- cgit v1.3.1