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 --- fs/afs/callback.c | 262 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 262 insertions(+) create mode 100644 fs/afs/callback.c (limited to 'fs/afs/callback.c') diff --git a/fs/afs/callback.c b/fs/afs/callback.c new file mode 100644 index 000000000..61354003c --- /dev/null +++ b/fs/afs/callback.c @@ -0,0 +1,262 @@ +/* + * Copyright (c) 2002, 2007 Red Hat, Inc. All rights reserved. + * + * This software may be freely redistributed under the terms of the + * GNU General Public License. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * + * Authors: David Woodhouse + * David Howells + * + */ + +#include +#include +#include +#include +#include +#include "internal.h" + +/* + * Handle invalidation of an mmap'd file. We invalidate all the PTEs referring + * to the pages in this file's pagecache, forcing the kernel to go through + * ->fault() or ->page_mkwrite() - at which point we can handle invalidation + * more fully. + */ +void afs_invalidate_mmap_work(struct work_struct *work) +{ + struct afs_vnode *vnode = container_of(work, struct afs_vnode, cb_work); + + unmap_mapping_pages(vnode->netfs.inode.i_mapping, 0, 0, false); +} + +static void afs_volume_init_callback(struct afs_volume *volume) +{ + struct afs_vnode *vnode; + + down_read(&volume->open_mmaps_lock); + + list_for_each_entry(vnode, &volume->open_mmaps, cb_mmap_link) { + if (vnode->cb_v_check != atomic_read(&volume->cb_v_break)) { + afs_clear_cb_promise(vnode, afs_cb_promise_clear_vol_init_cb); + queue_work(system_dfl_wq, &vnode->cb_work); + } + } + + up_read(&volume->open_mmaps_lock); +} + +/* + * Allow the fileserver to request callback state (re-)initialisation. + * Unfortunately, UUIDs are not guaranteed unique. + */ +void afs_init_callback_state(struct afs_server *server) +{ + struct afs_server_entry *se; + + down_read(&server->cell->vs_lock); + + list_for_each_entry(se, &server->volumes, slink) { + se->cb_expires_at = AFS_NO_CB_PROMISE; + se->volume->cb_expires_at = AFS_NO_CB_PROMISE; + trace_afs_cb_v_break(se->volume->vid, atomic_read(&se->volume->cb_v_break), + afs_cb_break_for_s_reinit); + if (!list_empty(&se->volume->open_mmaps)) + afs_volume_init_callback(se->volume); + } + + up_read(&server->cell->vs_lock); +} + +/* + * actually break a callback + */ +void __afs_break_callback(struct afs_vnode *vnode, enum afs_cb_break_reason reason) +{ + _enter(""); + + clear_bit(AFS_VNODE_NEW_CONTENT, &vnode->flags); + if (afs_clear_cb_promise(vnode, afs_cb_promise_clear_cb_break)) { + vnode->cb_break++; + vnode->cb_v_check = atomic_read(&vnode->volume->cb_v_break); + afs_clear_permits(vnode); + + if (vnode->lock_state == AFS_VNODE_LOCK_WAITING_FOR_CB) + afs_lock_may_be_available(vnode); + + if (reason != afs_cb_break_for_deleted && + vnode->status.type == AFS_FTYPE_FILE && + atomic_read(&vnode->cb_nr_mmap)) + queue_work(system_dfl_wq, &vnode->cb_work); + + trace_afs_cb_break(&vnode->fid, vnode->cb_break, reason, true); + } else { + trace_afs_cb_break(&vnode->fid, vnode->cb_break, reason, false); + } +} + +void afs_break_callback(struct afs_vnode *vnode, enum afs_cb_break_reason reason) +{ + write_seqlock(&vnode->cb_lock); + __afs_break_callback(vnode, reason); + write_sequnlock(&vnode->cb_lock); +} + +/* + * Look up a volume by volume ID under RCU conditions. + */ +static struct afs_volume *afs_lookup_volume_rcu(struct afs_cell *cell, + afs_volid_t vid) +{ + struct afs_volume *volume = NULL; + struct rb_node *p; + + scoped_seqlock_read(&cell->volume_lock, ss_lock) { + /* Unfortunately, rbtree walking doesn't give reliable results + * under just the RCU read lock, so we have to check for + * changes. + */ + p = rcu_dereference_raw(cell->volumes.rb_node); + while (p) { + volume = rb_entry(p, struct afs_volume, cell_node); + + if (volume->vid < vid) + p = rcu_dereference_raw(p->rb_left); + else if (volume->vid > vid) + p = rcu_dereference_raw(p->rb_right); + else + break; + volume = NULL; + } + + if (volume && afs_try_get_volume(volume, afs_volume_trace_get_callback)) + break; + volume = NULL; + } + + return volume; +} + +/* + * Allow the fileserver to break callbacks at the volume-level. This is + * typically done when, for example, a R/W volume is snapshotted to a R/O + * volume (the only way to change an R/O volume). It may also, however, happen + * when a volserver takes control of a volume (offlining it, moving it, etc.). + * + * Every file in that volume will need to be reevaluated. + */ +static void afs_break_volume_callback(struct afs_server *server, + struct afs_volume *volume) + __releases(RCU) +{ + struct afs_server_list *slist = rcu_dereference(volume->servers); + unsigned int i, cb_v_break; + + write_lock(&volume->cb_v_break_lock); + + for (i = 0; i < slist->nr_servers; i++) + if (slist->servers[i].server == server) + slist->servers[i].cb_expires_at = AFS_NO_CB_PROMISE; + volume->cb_expires_at = AFS_NO_CB_PROMISE; + + cb_v_break = atomic_inc_return_release(&volume->cb_v_break); + trace_afs_cb_v_break(volume->vid, cb_v_break, afs_cb_break_for_volume_callback); + + write_unlock(&volume->cb_v_break_lock); + rcu_read_unlock(); + + if (!list_empty(&volume->open_mmaps)) + afs_volume_init_callback(volume); +} + +/* + * allow the fileserver to explicitly break one callback + * - happens when + * - the backing file is changed + * - a lock is released + */ +static void afs_break_one_callback(struct afs_server *server, + struct afs_volume *volume, + struct afs_fid *fid) +{ + struct super_block *sb; + struct afs_vnode *vnode; + struct inode *inode; + + /* See if we can find a matching inode - even an I_NEW inode needs to + * be marked as it can have its callback broken before we finish + * setting up the local inode. + */ + sb = rcu_dereference(volume->sb); + if (!sb) + return; + + inode = find_inode_rcu(sb, fid->vnode, afs_ilookup5_test_by_fid, fid); + if (inode) { + vnode = AFS_FS_I(inode); + afs_break_callback(vnode, afs_cb_break_for_callback); + } else { + trace_afs_cb_miss(fid, afs_cb_break_for_callback); + } +} + +static void afs_break_some_callbacks(struct afs_server *server, + struct afs_callback_break *cbb, + size_t *_count) +{ + struct afs_callback_break *residue = cbb; + struct afs_volume *volume; + afs_volid_t vid = cbb->fid.vid; + size_t i; + + rcu_read_lock(); + volume = afs_lookup_volume_rcu(server->cell, vid); + if (!volume) { + /* Ignore breaks on unknown volumes. */ + rcu_read_unlock(); + *_count = 0; + } else if (cbb->fid.vnode == 0 && cbb->fid.unique == 0) { + afs_break_volume_callback(server, volume); + *_count -= 1; + if (*_count) + memmove(cbb, cbb + 1, sizeof(*cbb) * *_count); + } else { + /* TODO: Find all matching volumes if we couldn't match the server and + * break them anyway. + */ + + for (i = *_count; i > 0; cbb++, i--) { + if (cbb->fid.vid == vid) { + _debug("- Fid { vl=%08llx n=%llu u=%u }", + cbb->fid.vid, + cbb->fid.vnode, + cbb->fid.unique); + --*_count; + if (volume) + afs_break_one_callback(server, volume, &cbb->fid); + } else { + *residue++ = *cbb; + } + } + rcu_read_unlock(); + } + + afs_put_volume(volume, afs_volume_trace_put_callback); +} + +/* + * allow the fileserver to break callback promises + */ +void afs_break_callbacks(struct afs_server *server, size_t count, + struct afs_callback_break *callbacks) +{ + _enter("%p,%zu,", server, count); + + ASSERT(server != NULL); + + while (count > 0) + afs_break_some_callbacks(server, callbacks, &count); +} -- cgit v1.3.1