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, 367 insertions, 0 deletions

diff --git a/drivers/s390/cio/ccwreq.c b/drivers/s390/cio/ccwreq.c
new file mode 100644
index 000000000..763f477cc
--- /dev/null
+++ b/drivers/s390/cio/ccwreq.c
@@ -0,0 +1,367 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ *  Handling of internal CCW device requests.
+ *
+ *    Copyright IBM Corp. 2009, 2011
+ *    Author(s): Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
+ */
+
+#define pr_fmt(fmt) "cio: " fmt
+
+#include <linux/types.h>
+#include <linux/err.h>
+#include <asm/ccwdev.h>
+#include <asm/cio.h>
+
+#include "io_sch.h"
+#include "cio.h"
+#include "device.h"
+#include "cio_debug.h"
+
+/**
+ * lpm_adjust - adjust path mask
+ * @lpm: path mask to adjust
+ * @mask: mask of available paths
+ *
+ * Shift @lpm right until @lpm and @mask have at least one bit in common or
+ * until @lpm is zero. Return the resulting lpm.
+ */
+int lpm_adjust(int lpm, int mask)
+{
+	while (lpm && ((lpm & mask) == 0))
+		lpm >>= 1;
+	return lpm;
+}
+
+/*
+ * Adjust path mask to use next path and reset retry count. Return resulting
+ * path mask.
+ */
+static u16 ccwreq_next_path(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+
+	if (!req->singlepath) {
+		req->mask = 0;
+		goto out;
+	}
+	req->retries	= req->maxretries;
+	req->mask	= lpm_adjust(req->mask >> 1, req->lpm);
+out:
+	return req->mask;
+}
+
+/*
+ * Clean up device state and report to callback.
+ */
+static void ccwreq_stop(struct ccw_device *cdev, int rc)
+{
+	struct ccw_request *req = &cdev->private->req;
+
+	if (req->done)
+		return;
+	req->done = 1;
+	ccw_device_set_timeout(cdev, 0);
+	memset(&cdev->private->dma_area->irb, 0, sizeof(struct irb));
+	if (rc && rc != -ENODEV && req->drc)
+		rc = req->drc;
+	req->callback(cdev, req->data, rc);
+}
+
+/*
+ * (Re-)Start the operation until retries and paths are exhausted.
+ */
+static void ccwreq_do(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw1 *cp = req->cp;
+	int rc = -EACCES;
+
+	while (req->mask) {
+		if (req->retries-- == 0) {
+			/* Retries exhausted, try next path. */
+			ccwreq_next_path(cdev);
+			continue;
+		}
+		/* Perform start function. */
+		memset(&cdev->private->dma_area->irb, 0, sizeof(struct irb));
+		rc = cio_start(sch, cp, (u8) req->mask);
+		if (rc == 0) {
+			/* I/O started successfully. */
+			ccw_device_set_timeout(cdev, req->timeout);
+			return;
+		}
+		if (rc == -ENODEV) {
+			/* Permanent device error. */
+			break;
+		}
+		if (rc == -EACCES) {
+			/* Permant path error. */
+			ccwreq_next_path(cdev);
+			continue;
+		}
+		/* Temporary improper status. */
+		rc = cio_clear(sch);
+		if (rc)
+			break;
+		return;
+	}
+	ccwreq_stop(cdev, rc);
+}
+
+/**
+ * ccw_request_start - perform I/O request
+ * @cdev: ccw device
+ *
+ * Perform the I/O request specified by cdev->req.
+ */
+void ccw_request_start(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+
+	if (req->singlepath) {
+		/* Try all paths twice to counter link flapping. */
+		req->mask = 0x8080;
+	} else
+		req->mask = req->lpm;
+
+	req->retries	= req->maxretries;
+	req->mask	= lpm_adjust(req->mask, req->lpm);
+	req->drc	= 0;
+	req->done	= 0;
+	req->cancel	= 0;
+	if (!req->mask)
+		goto out_nopath;
+	ccwreq_do(cdev);
+	return;
+
+out_nopath:
+	ccwreq_stop(cdev, -EACCES);
+}
+
+/**
+ * ccw_request_cancel - cancel running I/O request
+ * @cdev: ccw device
+ *
+ * Cancel the I/O request specified by cdev->req. Return non-zero if request
+ * has already finished, zero otherwise.
+ */
+int ccw_request_cancel(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	int rc;
+
+	if (req->done)
+		return 1;
+	req->cancel = 1;
+	rc = cio_clear(sch);
+	if (rc)
+		ccwreq_stop(cdev, rc);
+	return 0;
+}
+
+/*
+ * Return the status of the internal I/O started on the specified ccw device.
+ * Perform BASIC SENSE if required.
+ */
+static enum io_status ccwreq_status(struct ccw_device *cdev, struct irb *lcirb)
+{
+	struct irb *irb = &cdev->private->dma_area->irb;
+	struct cmd_scsw *scsw = &irb->scsw.cmd;
+	enum uc_todo todo;
+
+	/* Perform BASIC SENSE if needed. */
+	if (ccw_device_accumulate_and_sense(cdev, lcirb))
+		return IO_RUNNING;
+	/* Check for halt/clear interrupt. */
+	if (scsw->fctl & (SCSW_FCTL_HALT_FUNC | SCSW_FCTL_CLEAR_FUNC))
+		return IO_KILLED;
+	/* Check for path error. */
+	if (scsw->cc == 3 || scsw->pno)
+		return IO_PATH_ERROR;
+	/* Handle BASIC SENSE data. */
+	if (irb->esw.esw0.erw.cons) {
+		CIO_TRACE_EVENT(2, "sensedata");
+		CIO_HEX_EVENT(2, &cdev->private->dev_id,
+			      sizeof(struct ccw_dev_id));
+		CIO_HEX_EVENT(2, &cdev->private->dma_area->irb.ecw,
+			      SENSE_MAX_COUNT);
+		/* Check for command reject. */
+		if (irb->ecw[0] & SNS0_CMD_REJECT)
+			return IO_REJECTED;
+		/* Ask the driver what to do */
+		if (cdev->drv && cdev->drv->uc_handler) {
+			todo = cdev->drv->uc_handler(cdev, lcirb);
+			CIO_TRACE_EVENT(2, "uc_response");
+			CIO_HEX_EVENT(2, &todo, sizeof(todo));
+			switch (todo) {
+			case UC_TODO_RETRY:
+				return IO_STATUS_ERROR;
+			case UC_TODO_RETRY_ON_NEW_PATH:
+				return IO_PATH_ERROR;
+			case UC_TODO_STOP:
+				return IO_REJECTED;
+			default:
+				return IO_STATUS_ERROR;
+			}
+		}
+		/* Assume that unexpected SENSE data implies an error. */
+		return IO_STATUS_ERROR;
+	}
+	/* Check for channel errors. */
+	if (scsw->cstat != 0)
+		return IO_STATUS_ERROR;
+	/* Check for device errors. */
+	if (scsw->dstat & ~(DEV_STAT_CHN_END | DEV_STAT_DEV_END))
+		return IO_STATUS_ERROR;
+	/* Check for final state. */
+	if (!(scsw->dstat & DEV_STAT_DEV_END))
+		return IO_RUNNING;
+	/* Check for other improper status. */
+	if (scsw->cc == 1 && (scsw->stctl & SCSW_STCTL_ALERT_STATUS))
+		return IO_STATUS_ERROR;
+	return IO_DONE;
+}
+
+/*
+ * Log ccw request status.
+ */
+static void ccwreq_log_status(struct ccw_device *cdev, enum io_status status)
+{
+	struct ccw_request *req = &cdev->private->req;
+	struct {
+		struct ccw_dev_id dev_id;
+		u16 retries;
+		u8 lpm;
+		u8 status;
+	}  __attribute__ ((packed)) data;
+	data.dev_id	= cdev->private->dev_id;
+	data.retries	= req->retries;
+	data.lpm	= (u8) req->mask;
+	data.status	= (u8) status;
+	CIO_TRACE_EVENT(2, "reqstat");
+	CIO_HEX_EVENT(2, &data, sizeof(data));
+}
+
+/**
+ * ccw_request_handler - interrupt handler for I/O request procedure.
+ * @cdev: ccw device
+ *
+ * Handle interrupt during I/O request procedure.
+ */
+void ccw_request_handler(struct ccw_device *cdev)
+{
+	struct irb *irb = this_cpu_ptr(&cio_irb);
+	struct ccw_request *req = &cdev->private->req;
+	enum io_status status;
+	int rc = -EOPNOTSUPP;
+
+	/* Check status of I/O request. */
+	status = ccwreq_status(cdev, irb);
+	if (req->filter)
+		status = req->filter(cdev, req->data, irb, status);
+	if (status != IO_RUNNING)
+		ccw_device_set_timeout(cdev, 0);
+	if (status != IO_DONE && status != IO_RUNNING)
+		ccwreq_log_status(cdev, status);
+	switch (status) {
+	case IO_DONE:
+		break;
+	case IO_RUNNING:
+		return;
+	case IO_REJECTED:
+		goto err;
+	case IO_PATH_ERROR:
+		goto out_next_path;
+	case IO_STATUS_ERROR:
+		goto out_restart;
+	case IO_KILLED:
+		/* Check if request was cancelled on purpose. */
+		if (req->cancel) {
+			rc = -EIO;
+			goto err;
+		}
+		goto out_restart;
+	}
+	/* Check back with request initiator. */
+	if (!req->check)
+		goto out;
+	switch (req->check(cdev, req->data)) {
+	case 0:
+		break;
+	case -EAGAIN:
+		goto out_restart;
+	case -EACCES:
+		goto out_next_path;
+	default:
+		goto err;
+	}
+out:
+	ccwreq_stop(cdev, 0);
+	return;
+
+out_next_path:
+	/* Try next path and restart I/O. */
+	if (!ccwreq_next_path(cdev)) {
+		rc = -EACCES;
+		goto err;
+	}
+out_restart:
+	/* Restart. */
+	ccwreq_do(cdev);
+	return;
+err:
+	ccwreq_stop(cdev, rc);
+}
+
+
+/**
+ * ccw_request_timeout - timeout handler for I/O request procedure
+ * @cdev: ccw device
+ *
+ * Handle timeout during I/O request procedure.
+ */
+void ccw_request_timeout(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	int rc = -ENODEV, chp;
+
+	if (cio_update_schib(sch))
+		goto err;
+
+	for (chp = 0; chp < 8; chp++) {
+		if ((0x80 >> chp) & sch->schib.pmcw.lpum)
+			pr_warn("%s: No interrupt was received within %lus (CS=%02x, DS=%02x, CHPID=%x.%02x)\n",
+				dev_name(&cdev->dev), req->timeout / HZ,
+				scsw_cstat(&sch->schib.scsw),
+				scsw_dstat(&sch->schib.scsw),
+				sch->schid.cssid,
+				sch->schib.pmcw.chpid[chp]);
+	}
+
+	if (!ccwreq_next_path(cdev)) {
+		/* set the final return code for this request */
+		req->drc = -ETIME;
+	}
+	rc = cio_clear(sch);
+	if (rc)
+		goto err;
+	return;
+
+err:
+	ccwreq_stop(cdev, rc);
+}
+
+/**
+ * ccw_request_notoper - notoper handler for I/O request procedure
+ * @cdev: ccw device
+ *
+ * Handle notoper during I/O request procedure.
+ */
+void ccw_request_notoper(struct ccw_device *cdev)
+{
+	ccwreq_stop(cdev, -ENODEV);
+}