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/nfsd/flexfilelayout.c | 144 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 144 insertions(+) create mode 100644 fs/nfsd/flexfilelayout.c (limited to 'fs/nfsd/flexfilelayout.c') diff --git a/fs/nfsd/flexfilelayout.c b/fs/nfsd/flexfilelayout.c new file mode 100644 index 000000000..6d531285a --- /dev/null +++ b/fs/nfsd/flexfilelayout.c @@ -0,0 +1,144 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (c) 2016 Tom Haynes + * + * The following implements a super-simple flex-file server + * where the NFSv4.1 mds is also the ds. And the storage is + * the same. I.e., writing to the mds via a NFSv4.1 WRITE + * goes to the same location as the NFSv3 WRITE. + */ +#include + +#include + +#include + +#include "flexfilelayoutxdr.h" +#include "pnfs.h" +#include "vfs.h" + +#define NFSDDBG_FACILITY NFSDDBG_PNFS + +static __be32 +nfsd4_ff_proc_layoutget(struct svc_rqst *rqstp, struct inode *inode, + const struct svc_fh *fhp, struct nfsd4_layoutget *args) +{ + struct nfsd4_layout_seg *seg = &args->lg_seg; + u32 device_generation = 0; + int error; + uid_t u; + + struct pnfs_ff_layout *fl; + + /* + * The super simple flex file server has 1 mirror, 1 data server, + * and 1 file handle. So instead of 4 allocs, do 1 for now. + * Zero it out for the stateid - don't want junk in there! + */ + error = -ENOMEM; + fl = kzalloc_obj(*fl); + if (!fl) + goto out_error; + args->lg_content = fl; + + /* + * Avoid layout commit, try to force the I/O to the DS, + * and for fun, cause all IOMODE_RW layout segments to + * effectively be WRITE only. + */ + fl->flags = FF_FLAGS_NO_LAYOUTCOMMIT | FF_FLAGS_NO_IO_THRU_MDS | + FF_FLAGS_NO_READ_IO; + + /* Do not allow a IOMODE_READ segment to have write pemissions */ + if (seg->iomode == IOMODE_READ) { + u = from_kuid(&init_user_ns, inode->i_uid) + 1; + fl->uid = make_kuid(&init_user_ns, u); + } else + fl->uid = inode->i_uid; + fl->gid = inode->i_gid; + + error = nfsd4_set_deviceid(&fl->deviceid, fhp, device_generation); + if (error) + goto out_error; + + fl->fh.size = fhp->fh_handle.fh_size; + memcpy(fl->fh.data, &fhp->fh_handle.fh_raw, fl->fh.size); + + /* Give whole file layout segments */ + seg->offset = 0; + seg->length = NFS4_MAX_UINT64; + + dprintk("GET: 0x%llx:0x%llx %d\n", seg->offset, seg->length, + seg->iomode); + return 0; + +out_error: + seg->length = 0; + return nfserrno(error); +} + +static __be32 +nfsd4_ff_proc_getdeviceinfo(struct super_block *sb, struct svc_rqst *rqstp, + struct nfs4_client *clp, struct nfsd4_getdeviceinfo *gdp) +{ + struct pnfs_ff_device_addr *da; + + u16 port; + char addr[INET6_ADDRSTRLEN]; + + da = kzalloc_obj(struct pnfs_ff_device_addr); + if (!da) + return nfserrno(-ENOMEM); + + gdp->gd_device = da; + + da->version = 3; + da->minor_version = 0; + + da->rsize = svc_max_payload(rqstp); + da->wsize = da->rsize; + + rpc_ntop((struct sockaddr *)&rqstp->rq_daddr, + addr, INET6_ADDRSTRLEN); + if (rqstp->rq_daddr.ss_family == AF_INET) { + struct sockaddr_in *sin; + + sin = (struct sockaddr_in *)&rqstp->rq_daddr; + port = ntohs(sin->sin_port); + snprintf(da->netaddr.netid, FF_NETID_LEN + 1, "tcp"); + da->netaddr.netid_len = 3; + } else { + struct sockaddr_in6 *sin6; + + sin6 = (struct sockaddr_in6 *)&rqstp->rq_daddr; + port = ntohs(sin6->sin6_port); + snprintf(da->netaddr.netid, FF_NETID_LEN + 1, "tcp6"); + da->netaddr.netid_len = 4; + } + + da->netaddr.addr_len = + snprintf(da->netaddr.addr, FF_ADDR_LEN + 1, + "%s.%d.%d", addr, port >> 8, port & 0xff); + + da->tightly_coupled = false; + + return 0; +} + +static __be32 +nfsd4_ff_proc_layoutcommit(struct inode *inode, struct svc_rqst *rqstp, + struct nfsd4_layoutcommit *lcp) +{ + return nfs_ok; +} + +const struct nfsd4_layout_ops ff_layout_ops = { + .notify_types = + NOTIFY_DEVICEID4_DELETE | NOTIFY_DEVICEID4_CHANGE, + .disable_recalls = true, + .proc_getdeviceinfo = nfsd4_ff_proc_getdeviceinfo, + .encode_getdeviceinfo = nfsd4_ff_encode_getdeviceinfo, + .proc_layoutget = nfsd4_ff_proc_layoutget, + .encode_layoutget = nfsd4_ff_encode_layoutget, + .proc_layoutcommit = nfsd4_ff_proc_layoutcommit, +}; -- cgit v1.3.1