x86/kprobes: Fix crash when probing CS CALL instructionsgrafted

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) <mhiramat@kernel.org>
Signed-off-by: Jinke Han <jinkehan@didiglobal.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Borislav Petkov <bp@alien8.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://patch.msgid.link/20260908073742.GA10517@didi-ThinkCentre-M920t-N000

1 files changed, 496 insertions, 0 deletions

diff --git a/drivers/usb/cdns3/cdns3-trace.h b/drivers/usb/cdns3/cdns3-trace.h
new file mode 100644
index 000000000..21a6a2ce7
--- /dev/null
+++ b/drivers/usb/cdns3/cdns3-trace.h
@@ -0,0 +1,496 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * USBSS device controller driver.
+ * Trace support header file.
+ *
+ * Copyright (C) 2018-2019 Cadence.
+ *
+ * Author: Pawel Laszczak <pawell@cadence.com>
+ */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM cdns3
+
+#if !defined(__LINUX_CDNS3_TRACE) || defined(TRACE_HEADER_MULTI_READ)
+#define __LINUX_CDNS3_TRACE
+
+#include <linux/types.h>
+#include <linux/tracepoint.h>
+#include <asm/byteorder.h>
+#include <linux/usb/ch9.h>
+#include "core.h"
+#include "cdns3-gadget.h"
+#include "cdns3-debug.h"
+
+#define CDNS3_MSG_MAX	500
+
+TRACE_EVENT(cdns3_halt,
+	TP_PROTO(struct cdns3_endpoint *ep_priv, u8 halt, u8 flush),
+	TP_ARGS(ep_priv, halt, flush),
+	TP_STRUCT__entry(
+		__string(name, ep_priv->name)
+		__field(u8, halt)
+		__field(u8, flush)
+	),
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->halt = halt;
+		__entry->flush = flush;
+	),
+	TP_printk("Halt %s for %s: %s", __entry->flush ? " and flush" : "",
+		  __get_str(name), __entry->halt ? "set" : "cleared")
+);
+
+TRACE_EVENT(cdns3_wa1,
+	TP_PROTO(struct cdns3_endpoint *ep_priv, char *msg),
+	TP_ARGS(ep_priv, msg),
+	TP_STRUCT__entry(
+		__string(ep_name, ep_priv->name)
+		__string(msg, msg)
+	),
+	TP_fast_assign(
+		__assign_str(ep_name);
+		__assign_str(msg);
+	),
+	TP_printk("WA1: %s %s", __get_str(ep_name), __get_str(msg))
+);
+
+TRACE_EVENT(cdns3_wa2,
+	TP_PROTO(struct cdns3_endpoint *ep_priv, char *msg),
+	TP_ARGS(ep_priv, msg),
+	TP_STRUCT__entry(
+		__string(ep_name, ep_priv->name)
+		__string(msg, msg)
+	),
+	TP_fast_assign(
+		__assign_str(ep_name);
+		__assign_str(msg);
+	),
+	TP_printk("WA2: %s %s", __get_str(ep_name), __get_str(msg))
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_doorbell,
+	TP_PROTO(const char *ep_name, u32 ep_trbaddr),
+	TP_ARGS(ep_name, ep_trbaddr),
+	TP_STRUCT__entry(
+		__string(name, ep_name)
+		__field(u32, ep_trbaddr)
+	),
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->ep_trbaddr = ep_trbaddr;
+	),
+	TP_printk("%s, ep_trbaddr %08x", __get_str(name),
+		  __entry->ep_trbaddr)
+);
+
+DEFINE_EVENT(cdns3_log_doorbell, cdns3_doorbell_ep0,
+	TP_PROTO(const char *ep_name, u32 ep_trbaddr),
+	TP_ARGS(ep_name, ep_trbaddr)
+);
+
+DEFINE_EVENT(cdns3_log_doorbell, cdns3_doorbell_epx,
+	TP_PROTO(const char *ep_name, u32 ep_trbaddr),
+	TP_ARGS(ep_name, ep_trbaddr)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_usb_irq,
+	TP_PROTO(struct cdns3_device *priv_dev, u32 usb_ists),
+	TP_ARGS(priv_dev, usb_ists),
+	TP_STRUCT__entry(
+		__field(enum usb_device_speed, speed)
+		__field(u32, usb_ists)
+	),
+	TP_fast_assign(
+		__entry->speed = cdns3_get_speed(priv_dev);
+		__entry->usb_ists = usb_ists;
+	),
+	TP_printk("%s", cdns3_decode_usb_irq(__get_buf(CDNS3_MSG_MAX), __entry->speed,
+					     __entry->usb_ists))
+);
+
+DEFINE_EVENT(cdns3_log_usb_irq, cdns3_usb_irq,
+	TP_PROTO(struct cdns3_device *priv_dev, u32 usb_ists),
+	TP_ARGS(priv_dev, usb_ists)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_epx_irq,
+	TP_PROTO(struct cdns3_device *priv_dev, struct cdns3_endpoint *priv_ep),
+	TP_ARGS(priv_dev, priv_ep),
+	TP_STRUCT__entry(
+		__string(ep_name, priv_ep->name)
+		__field(u32, ep_sts)
+		__field(u32, ep_traddr)
+		__field(u32, ep_last_sid)
+		__field(u32, use_streams)
+	),
+	TP_fast_assign(
+		__assign_str(ep_name);
+		__entry->ep_sts = readl(&priv_dev->regs->ep_sts);
+		__entry->ep_traddr = readl(&priv_dev->regs->ep_traddr);
+		__entry->ep_last_sid = priv_ep->last_stream_id;
+		__entry->use_streams = priv_ep->use_streams;
+	),
+	TP_printk("%s, ep_traddr: %08x ep_last_sid: %08x use_streams: %d",
+		  cdns3_decode_epx_irq(__get_buf(CDNS3_MSG_MAX),
+				       __get_str(ep_name),
+				       __entry->ep_sts),
+		  __entry->ep_traddr,
+		  __entry->ep_last_sid,
+		  __entry->use_streams)
+);
+
+DEFINE_EVENT(cdns3_log_epx_irq, cdns3_epx_irq,
+	TP_PROTO(struct cdns3_device *priv_dev, struct cdns3_endpoint *priv_ep),
+	TP_ARGS(priv_dev, priv_ep)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_ep0_irq,
+	TP_PROTO(struct cdns3_device *priv_dev,  u32 ep_sts),
+	TP_ARGS(priv_dev, ep_sts),
+	TP_STRUCT__entry(
+		__field(int, ep_dir)
+		__field(u32, ep_sts)
+	),
+	TP_fast_assign(
+		__entry->ep_dir = priv_dev->selected_ep;
+		__entry->ep_sts = ep_sts;
+	),
+	TP_printk("%s", cdns3_decode_ep0_irq(__get_buf(CDNS3_MSG_MAX),
+					     __entry->ep_dir,
+					     __entry->ep_sts))
+);
+
+DEFINE_EVENT(cdns3_log_ep0_irq, cdns3_ep0_irq,
+	TP_PROTO(struct cdns3_device *priv_dev, u32 ep_sts),
+	TP_ARGS(priv_dev, ep_sts)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_ctrl,
+	TP_PROTO(struct usb_ctrlrequest *ctrl),
+	TP_ARGS(ctrl),
+	TP_STRUCT__entry(
+		__field(u8, bRequestType)
+		__field(u8, bRequest)
+		__field(u16, wValue)
+		__field(u16, wIndex)
+		__field(u16, wLength)
+	),
+	TP_fast_assign(
+		__entry->bRequestType = ctrl->bRequestType;
+		__entry->bRequest = ctrl->bRequest;
+		__entry->wValue = le16_to_cpu(ctrl->wValue);
+		__entry->wIndex = le16_to_cpu(ctrl->wIndex);
+		__entry->wLength = le16_to_cpu(ctrl->wLength);
+	),
+	TP_printk("%s", usb_decode_ctrl(__get_buf(CDNS3_MSG_MAX), CDNS3_MSG_MAX,
+					__entry->bRequestType,
+					__entry->bRequest, __entry->wValue,
+					__entry->wIndex, __entry->wLength)
+	)
+);
+
+DEFINE_EVENT(cdns3_log_ctrl, cdns3_ctrl_req,
+	TP_PROTO(struct usb_ctrlrequest *ctrl),
+	TP_ARGS(ctrl)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_request,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req),
+	TP_STRUCT__entry(
+		__string(name, req->priv_ep->name)
+		__field(struct cdns3_request *, req)
+		__field(void *, buf)
+		__field(unsigned int, actual)
+		__field(unsigned int, length)
+		__field(int, status)
+		__field(int, zero)
+		__field(int, short_not_ok)
+		__field(int, no_interrupt)
+		__field(int, start_trb)
+		__field(int, end_trb)
+		__field(int, flags)
+		__field(unsigned int, stream_id)
+	),
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->req = req;
+		__entry->buf = req->request.buf;
+		__entry->actual = req->request.actual;
+		__entry->length = req->request.length;
+		__entry->status = req->request.status;
+		__entry->zero = req->request.zero;
+		__entry->short_not_ok = req->request.short_not_ok;
+		__entry->no_interrupt = req->request.no_interrupt;
+		__entry->start_trb = req->start_trb;
+		__entry->end_trb = req->end_trb;
+		__entry->flags = req->flags;
+		__entry->stream_id = req->request.stream_id;
+	),
+	TP_printk("%s: req: %p, req buff %p, length: %u/%u %s%s%s, status: %d,"
+		  " trb: [start:%d, end:%d], flags:%x SID: %u",
+		__get_str(name), __entry->req, __entry->buf, __entry->actual,
+		__entry->length,
+		__entry->zero ? "Z" : "z",
+		__entry->short_not_ok ? "S" : "s",
+		__entry->no_interrupt ? "I" : "i",
+		__entry->status,
+		__entry->start_trb,
+		__entry->end_trb,
+		__entry->flags,
+		__entry->stream_id
+	)
+);
+
+DEFINE_EVENT(cdns3_log_request, cdns3_alloc_request,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req)
+);
+
+DEFINE_EVENT(cdns3_log_request, cdns3_free_request,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req)
+);
+
+DEFINE_EVENT(cdns3_log_request, cdns3_ep_queue,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req)
+);
+
+DEFINE_EVENT(cdns3_log_request, cdns3_ep_dequeue,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req)
+);
+
+DEFINE_EVENT(cdns3_log_request, cdns3_gadget_giveback,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req)
+);
+
+TRACE_EVENT(cdns3_ep0_queue,
+	TP_PROTO(struct cdns3_device *dev_priv, struct usb_request *request),
+	TP_ARGS(dev_priv, request),
+	TP_STRUCT__entry(
+		__field(int, dir)
+		__field(int, length)
+	),
+	TP_fast_assign(
+		__entry->dir = dev_priv->ep0_data_dir;
+		__entry->length = request->length;
+	),
+	TP_printk("Queue to ep0%s length: %u", __entry->dir ? "in" : "out",
+		  __entry->length)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_aligned_request,
+	TP_PROTO(struct cdns3_request *priv_req),
+	TP_ARGS(priv_req),
+	TP_STRUCT__entry(
+		__string(name, priv_req->priv_ep->name)
+		__field(struct usb_request *, req)
+		__field(void *, buf)
+		__field(dma_addr_t, dma)
+		__field(void *, aligned_buf)
+		__field(dma_addr_t, aligned_dma)
+		__field(u32, aligned_buf_size)
+	),
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->req = &priv_req->request;
+		__entry->buf = priv_req->request.buf;
+		__entry->dma = priv_req->request.dma;
+		__entry->aligned_buf = priv_req->aligned_buf->buf;
+		__entry->aligned_dma = priv_req->aligned_buf->dma;
+		__entry->aligned_buf_size = priv_req->aligned_buf->size;
+	),
+	TP_printk("%s: req: %p, req buf %p, dma %pad a_buf %p a_dma %pad, size %d",
+		__get_str(name), __entry->req, __entry->buf, &__entry->dma,
+		__entry->aligned_buf, &__entry->aligned_dma,
+		__entry->aligned_buf_size
+	)
+);
+
+DEFINE_EVENT(cdns3_log_aligned_request, cdns3_free_aligned_request,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req)
+);
+
+DEFINE_EVENT(cdns3_log_aligned_request, cdns3_prepare_aligned_request,
+	TP_PROTO(struct cdns3_request *req),
+	TP_ARGS(req)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_trb,
+	TP_PROTO(struct cdns3_endpoint *priv_ep, struct cdns3_trb *trb),
+	TP_ARGS(priv_ep, trb),
+	TP_STRUCT__entry(
+		__string(name, priv_ep->name)
+		__field(struct cdns3_trb *, trb)
+		__field(u32, buffer)
+		__field(u32, length)
+		__field(u32, control)
+		__field(u32, type)
+		__field(unsigned int, last_stream_id)
+	),
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->trb = trb;
+		__entry->buffer = le32_to_cpu(trb->buffer);
+		__entry->length = le32_to_cpu(trb->length);
+		__entry->control = le32_to_cpu(trb->control);
+		__entry->type = usb_endpoint_type(priv_ep->endpoint.desc);
+		__entry->last_stream_id = priv_ep->last_stream_id;
+	),
+	TP_printk("%s: trb %p, dma buf: 0x%08x, size: %ld, burst: %d ctrl: 0x%08x (%s%s%s%s%s%s%s) SID:%lu LAST_SID:%u",
+		__get_str(name), __entry->trb, __entry->buffer,
+		TRB_LEN(__entry->length),
+		(u8)TRB_BURST_LEN_GET(__entry->length),
+		__entry->control,
+		__entry->control & TRB_CYCLE ? "C=1, " : "C=0, ",
+		__entry->control & TRB_TOGGLE ? "T=1, " : "T=0, ",
+		__entry->control & TRB_ISP ? "ISP, " : "",
+		__entry->control & TRB_FIFO_MODE ? "FIFO, " : "",
+		__entry->control & TRB_CHAIN ? "CHAIN, " : "",
+		__entry->control & TRB_IOC ? "IOC, " : "",
+		TRB_FIELD_TO_TYPE(__entry->control) == TRB_NORMAL ? "Normal" : "LINK",
+		TRB_FIELD_TO_STREAMID(__entry->control),
+		__entry->last_stream_id
+	)
+);
+
+DEFINE_EVENT(cdns3_log_trb, cdns3_prepare_trb,
+	TP_PROTO(struct cdns3_endpoint *priv_ep, struct cdns3_trb *trb),
+	TP_ARGS(priv_ep, trb)
+);
+
+DEFINE_EVENT(cdns3_log_trb, cdns3_complete_trb,
+	TP_PROTO(struct cdns3_endpoint *priv_ep, struct cdns3_trb *trb),
+	TP_ARGS(priv_ep, trb)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_ring,
+	TP_PROTO(struct cdns3_endpoint *priv_ep),
+	TP_ARGS(priv_ep),
+	TP_STRUCT__entry(
+		__dynamic_array(char, buffer,
+				GET_TRBS_PER_SEGMENT(priv_ep->type) > TRBS_PER_SEGMENT ?
+				CDNS3_MSG_MAX :
+				(GET_TRBS_PER_SEGMENT(priv_ep->type) * 65) + CDNS3_MSG_MAX)
+	),
+	TP_fast_assign(
+		cdns3_dbg_ring(priv_ep, __get_str(buffer));
+	),
+
+	TP_printk("%s", __get_str(buffer))
+);
+
+DEFINE_EVENT(cdns3_log_ring, cdns3_ring,
+	TP_PROTO(struct cdns3_endpoint *priv_ep),
+	TP_ARGS(priv_ep)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_ep,
+	TP_PROTO(struct cdns3_endpoint *priv_ep),
+	TP_ARGS(priv_ep),
+	TP_STRUCT__entry(
+		__string(name, priv_ep->name)
+		__field(unsigned int, maxpacket)
+		__field(unsigned int, maxpacket_limit)
+		__field(unsigned int, max_streams)
+		__field(unsigned int, use_streams)
+		__field(unsigned int, maxburst)
+		__field(unsigned int, flags)
+		__field(unsigned int, dir)
+		__field(u8, enqueue)
+		__field(u8, dequeue)
+	),
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->maxpacket = priv_ep->endpoint.maxpacket;
+		__entry->maxpacket_limit = priv_ep->endpoint.maxpacket_limit;
+		__entry->max_streams = priv_ep->endpoint.max_streams;
+		__entry->use_streams = priv_ep->use_streams;
+		__entry->maxburst = priv_ep->endpoint.maxburst;
+		__entry->flags = priv_ep->flags;
+		__entry->dir = priv_ep->dir;
+		__entry->enqueue = priv_ep->enqueue;
+		__entry->dequeue = priv_ep->dequeue;
+	),
+	TP_printk("%s: mps: %d/%d. streams: %d, stream enable: %d, burst: %d, "
+		  "enq idx: %d, deq idx: %d, flags %s%s%s%s%s%s%s%s, dir: %s",
+		__get_str(name), __entry->maxpacket,
+		__entry->maxpacket_limit, __entry->max_streams,
+		__entry->use_streams,
+		__entry->maxburst, __entry->enqueue,
+		__entry->dequeue,
+		__entry->flags & EP_ENABLED ? "EN | " : "",
+		__entry->flags & EP_STALLED ? "STALLED | " : "",
+		__entry->flags & EP_WEDGE ? "WEDGE | " : "",
+		__entry->flags & EP_TRANSFER_STARTED ? "STARTED | " : "",
+		__entry->flags & EP_UPDATE_EP_TRBADDR ? "UPD TRB | " : "",
+		__entry->flags & EP_PENDING_REQUEST ? "REQ PEN | " : "",
+		__entry->flags & EP_RING_FULL ? "RING FULL |" : "",
+		__entry->flags & EP_CLAIMED ?  "CLAIMED " : "",
+		__entry->dir ? "IN" : "OUT"
+	)
+);
+
+DEFINE_EVENT(cdns3_log_ep, cdns3_gadget_ep_enable,
+	TP_PROTO(struct cdns3_endpoint *priv_ep),
+	TP_ARGS(priv_ep)
+);
+
+DEFINE_EVENT(cdns3_log_ep, cdns3_gadget_ep_disable,
+	TP_PROTO(struct cdns3_endpoint *priv_ep),
+	TP_ARGS(priv_ep)
+);
+
+DECLARE_EVENT_CLASS(cdns3_log_request_handled,
+	TP_PROTO(struct cdns3_request *priv_req, int current_index,
+		 int handled),
+	TP_ARGS(priv_req, current_index, handled),
+	TP_STRUCT__entry(
+		__field(struct cdns3_request *, priv_req)
+		__field(unsigned int, dma_position)
+		__field(unsigned int, handled)
+		__field(unsigned int, dequeue_idx)
+		__field(unsigned int, enqueue_idx)
+		__field(unsigned int, start_trb)
+		__field(unsigned int, end_trb)
+	),
+	TP_fast_assign(
+		__entry->priv_req = priv_req;
+		__entry->dma_position = current_index;
+		__entry->handled = handled;
+		__entry->dequeue_idx = priv_req->priv_ep->dequeue;
+		__entry->enqueue_idx = priv_req->priv_ep->enqueue;
+		__entry->start_trb = priv_req->start_trb;
+		__entry->end_trb = priv_req->end_trb;
+	),
+	TP_printk("Req: %p %s, DMA pos: %d, ep deq: %d, ep enq: %d,"
+		  " start trb: %d, end trb: %d",
+		__entry->priv_req,
+		__entry->handled ? "handled" : "not handled",
+		__entry->dma_position, __entry->dequeue_idx,
+		__entry->enqueue_idx, __entry->start_trb,
+		__entry->end_trb
+	)
+);
+
+DEFINE_EVENT(cdns3_log_request_handled, cdns3_request_handled,
+	TP_PROTO(struct cdns3_request *priv_req, int current_index,
+		 int handled),
+	TP_ARGS(priv_req, current_index, handled)
+);
+#endif /* __LINUX_CDNS3_TRACE */
+
+/* this part must be outside header guard */
+
+#undef TRACE_INCLUDE_PATH
+#define TRACE_INCLUDE_PATH .
+
+#undef TRACE_INCLUDE_FILE
+#define TRACE_INCLUDE_FILE cdns3-trace
+
+#include <trace/define_trace.h>