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/scsi/raid_class.c | 267 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 267 insertions(+) create mode 100644 drivers/scsi/raid_class.c (limited to 'drivers/scsi/raid_class.c') diff --git a/drivers/scsi/raid_class.c b/drivers/scsi/raid_class.c new file mode 100644 index 000000000..94f76e358 --- /dev/null +++ b/drivers/scsi/raid_class.c @@ -0,0 +1,267 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * raid_class.c - implementation of a simple raid visualisation class + * + * Copyright (c) 2005 - James Bottomley + * + * This class is designed to allow raid attributes to be visualised and + * manipulated in a form independent of the underlying raid. Ultimately this + * should work for both hardware and software raids. + */ +#include +#include +#include +#include +#include +#include +#include +#include + +#define RAID_NUM_ATTRS 3 + +struct raid_internal { + struct raid_template r; + struct raid_function_template *f; + /* The actual attributes */ + struct device_attribute private_attrs[RAID_NUM_ATTRS]; + /* The array of null terminated pointers to attributes + * needed by scsi_sysfs.c */ + struct device_attribute *attrs[RAID_NUM_ATTRS + 1]; +}; + +struct raid_component { + struct list_head node; + struct device dev; + int num; +}; + +#define to_raid_internal(tmpl) container_of(tmpl, struct raid_internal, r) + +#define tc_to_raid_internal(tcont) ({ \ + struct raid_template *r = \ + container_of(tcont, struct raid_template, raid_attrs); \ + to_raid_internal(r); \ +}) + +#define ac_to_raid_internal(acont) ({ \ + struct transport_container *tc = \ + container_of(acont, struct transport_container, ac); \ + tc_to_raid_internal(tc); \ +}) + +#define device_to_raid_internal(dev) ({ \ + struct attribute_container *ac = \ + attribute_container_classdev_to_container(dev); \ + ac_to_raid_internal(ac); \ +}) + + +static int raid_match(struct attribute_container *cont, struct device *dev) +{ + /* We have to look for every subsystem that could house + * emulated RAID devices, so start with SCSI */ + struct raid_internal *i = ac_to_raid_internal(cont); + + if (IS_ENABLED(CONFIG_SCSI) && scsi_is_sdev_device(dev)) { + struct scsi_device *sdev = to_scsi_device(dev); + + if (i->f->cookie != sdev->host->hostt) + return 0; + + return i->f->is_raid(dev); + } + /* FIXME: look at other subsystems too */ + return 0; +} + +static int raid_setup(struct transport_container *tc, struct device *dev, + struct device *cdev) +{ + struct raid_data *rd; + + BUG_ON(dev_get_drvdata(cdev)); + + rd = kzalloc_obj(*rd); + if (!rd) + return -ENOMEM; + + INIT_LIST_HEAD(&rd->component_list); + dev_set_drvdata(cdev, rd); + + return 0; +} + +static int raid_remove(struct transport_container *tc, struct device *dev, + struct device *cdev) +{ + struct raid_data *rd = dev_get_drvdata(cdev); + struct raid_component *rc, *next; + dev_printk(KERN_ERR, dev, "RAID REMOVE\n"); + dev_set_drvdata(cdev, NULL); + list_for_each_entry_safe(rc, next, &rd->component_list, node) { + list_del(&rc->node); + dev_printk(KERN_ERR, rc->dev.parent, "RAID COMPONENT REMOVE\n"); + device_unregister(&rc->dev); + } + dev_printk(KERN_ERR, dev, "RAID REMOVE DONE\n"); + kfree(rd); + return 0; +} + +static DECLARE_TRANSPORT_CLASS(raid_class, + "raid_devices", + raid_setup, + raid_remove, + NULL); + +static const struct { + enum raid_state value; + char *name; +} raid_states[] = { + { RAID_STATE_UNKNOWN, "unknown" }, + { RAID_STATE_ACTIVE, "active" }, + { RAID_STATE_DEGRADED, "degraded" }, + { RAID_STATE_RESYNCING, "resyncing" }, + { RAID_STATE_OFFLINE, "offline" }, +}; + +static const char *raid_state_name(enum raid_state state) +{ + int i; + char *name = NULL; + + for (i = 0; i < ARRAY_SIZE(raid_states); i++) { + if (raid_states[i].value == state) { + name = raid_states[i].name; + break; + } + } + return name; +} + +static struct { + enum raid_level value; + char *name; +} raid_levels[] = { + { RAID_LEVEL_UNKNOWN, "unknown" }, + { RAID_LEVEL_LINEAR, "linear" }, + { RAID_LEVEL_0, "raid0" }, + { RAID_LEVEL_1, "raid1" }, + { RAID_LEVEL_10, "raid10" }, + { RAID_LEVEL_1E, "raid1e" }, + { RAID_LEVEL_3, "raid3" }, + { RAID_LEVEL_4, "raid4" }, + { RAID_LEVEL_5, "raid5" }, + { RAID_LEVEL_50, "raid50" }, + { RAID_LEVEL_6, "raid6" }, + { RAID_LEVEL_JBOD, "jbod" }, +}; + +static const char *raid_level_name(enum raid_level level) +{ + int i; + char *name = NULL; + + for (i = 0; i < ARRAY_SIZE(raid_levels); i++) { + if (raid_levels[i].value == level) { + name = raid_levels[i].name; + break; + } + } + return name; +} + +#define raid_attr_show_internal(attr, fmt, var, code) \ +static ssize_t raid_show_##attr(struct device *dev, \ + struct device_attribute *attr, \ + char *buf) \ +{ \ + struct raid_data *rd = dev_get_drvdata(dev); \ + code \ + return snprintf(buf, 20, #fmt "\n", var); \ +} + +#define raid_attr_ro_states(attr, states, code) \ +raid_attr_show_internal(attr, %s, name, \ + const char *name; \ + code \ + name = raid_##states##_name(rd->attr); \ +) \ +static DEVICE_ATTR(attr, S_IRUGO, raid_show_##attr, NULL) + + +#define raid_attr_ro_internal(attr, code) \ +raid_attr_show_internal(attr, %d, rd->attr, code) \ +static DEVICE_ATTR(attr, S_IRUGO, raid_show_##attr, NULL) + +#define ATTR_CODE(attr) \ + struct raid_internal *i = device_to_raid_internal(dev); \ + if (i->f->get_##attr) \ + i->f->get_##attr(dev->parent); + +#define raid_attr_ro(attr) raid_attr_ro_internal(attr, ) +#define raid_attr_ro_fn(attr) raid_attr_ro_internal(attr, ATTR_CODE(attr)) +#define raid_attr_ro_state(attr) raid_attr_ro_states(attr, attr, ) +#define raid_attr_ro_state_fn(attr) raid_attr_ro_states(attr, attr, ATTR_CODE(attr)) + + +raid_attr_ro_state(level); +raid_attr_ro_fn(resync); +raid_attr_ro_state_fn(state); + +struct raid_template * +raid_class_attach(struct raid_function_template *ft) +{ + struct raid_internal *i = kzalloc_obj(struct raid_internal); + int count = 0; + + if (unlikely(!i)) + return NULL; + + i->f = ft; + + i->r.raid_attrs.ac.class = &raid_class.class; + i->r.raid_attrs.ac.match = raid_match; + i->r.raid_attrs.ac.attrs = &i->attrs[0]; + + attribute_container_register(&i->r.raid_attrs.ac); + + i->attrs[count++] = &dev_attr_level; + i->attrs[count++] = &dev_attr_resync; + i->attrs[count++] = &dev_attr_state; + + i->attrs[count] = NULL; + BUG_ON(count > RAID_NUM_ATTRS); + + return &i->r; +} +EXPORT_SYMBOL(raid_class_attach); + +void +raid_class_release(struct raid_template *r) +{ + struct raid_internal *i = to_raid_internal(r); + + BUG_ON(attribute_container_unregister(&i->r.raid_attrs.ac)); + + kfree(i); +} +EXPORT_SYMBOL(raid_class_release); + +static __init int raid_init(void) +{ + return transport_class_register(&raid_class); +} + +static __exit void raid_exit(void) +{ + transport_class_unregister(&raid_class); +} + +MODULE_AUTHOR("James Bottomley"); +MODULE_DESCRIPTION("RAID device class"); +MODULE_LICENSE("GPL"); + +module_init(raid_init); +module_exit(raid_exit); + -- cgit v1.3.1