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, 218 insertions, 0 deletions

diff --git a/drivers/gpu/host1x/debug.c b/drivers/gpu/host1x/debug.c
new file mode 100644
index 000000000..b828f773f
--- /dev/null
+++ b/drivers/gpu/host1x/debug.c
@@ -0,0 +1,218 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2010 Google, Inc.
+ * Author: Erik Gilling <konkers@android.com>
+ *
+ * Copyright (C) 2011-2013 NVIDIA Corporation
+ */
+
+#include <linux/debugfs.h>
+#include <linux/pm_runtime.h>
+#include <linux/seq_file.h>
+#include <linux/uaccess.h>
+
+#include <linux/io.h>
+
+#include "dev.h"
+#include "debug.h"
+#include "channel.h"
+
+static DEFINE_MUTEX(debug_lock);
+
+unsigned int host1x_debug_trace_cmdbuf;
+
+static pid_t host1x_debug_force_timeout_pid;
+static u32 host1x_debug_force_timeout_val;
+static u32 host1x_debug_force_timeout_channel;
+
+void host1x_debug_output(struct output *o, const char *fmt, ...)
+{
+	va_list args;
+	int len;
+
+	va_start(args, fmt);
+	len = vscnprintf(o->buf, sizeof(o->buf), fmt, args);
+	va_end(args);
+
+	o->fn(o->ctx, o->buf, len, false);
+}
+
+void host1x_debug_cont(struct output *o, const char *fmt, ...)
+{
+	va_list args;
+	int len;
+
+	va_start(args, fmt);
+	len = vscnprintf(o->buf, sizeof(o->buf), fmt, args);
+	va_end(args);
+
+	o->fn(o->ctx, o->buf, len, true);
+}
+
+static int show_channel(struct host1x_channel *ch, void *data, bool show_fifo)
+{
+	struct host1x *m = dev_get_drvdata(ch->dev->parent);
+	struct output *o = data;
+	int err;
+
+	err = pm_runtime_resume_and_get(m->dev);
+	if (err < 0)
+		return err;
+
+	mutex_lock(&ch->cdma.lock);
+	mutex_lock(&debug_lock);
+
+	if (show_fifo)
+		host1x_hw_show_channel_fifo(m, ch, o);
+
+	host1x_hw_show_channel_cdma(m, ch, o);
+
+	mutex_unlock(&debug_lock);
+	mutex_unlock(&ch->cdma.lock);
+
+	pm_runtime_put(m->dev);
+
+	return 0;
+}
+
+static void show_syncpts(struct host1x *m, struct output *o, bool show_all)
+{
+	unsigned long irqflags;
+	struct list_head *pos;
+	unsigned int i;
+	int err;
+
+	host1x_debug_output(o, "---- syncpts ----\n");
+
+	err = pm_runtime_resume_and_get(m->dev);
+	if (err < 0)
+		return;
+
+	for (i = 0; i < host1x_syncpt_nb_pts(m); i++) {
+		u32 max = host1x_syncpt_read_max(m->syncpt + i);
+		u32 min = host1x_syncpt_load(m->syncpt + i);
+		unsigned int waiters = 0;
+
+		spin_lock_irqsave(&m->syncpt[i].fences.lock, irqflags);
+		list_for_each(pos, &m->syncpt[i].fences.list)
+			waiters++;
+		spin_unlock_irqrestore(&m->syncpt[i].fences.lock, irqflags);
+
+		if (!kref_read(&m->syncpt[i].ref))
+			continue;
+
+		if (!show_all && !min && !max && !waiters)
+			continue;
+
+		host1x_debug_output(o,
+				    "id %u (%s) min %d max %d (%d waiters)\n",
+				    i, m->syncpt[i].name, min, max, waiters);
+	}
+
+	for (i = 0; i < host1x_syncpt_nb_bases(m); i++) {
+		u32 base_val;
+
+		base_val = host1x_syncpt_load_wait_base(m->syncpt + i);
+		if (base_val)
+			host1x_debug_output(o, "waitbase id %u val %d\n", i,
+					    base_val);
+	}
+
+	pm_runtime_put(m->dev);
+
+	host1x_debug_output(o, "\n");
+}
+
+static void show_all(struct host1x *m, struct output *o, bool show_fifo)
+{
+	unsigned int i;
+
+	host1x_hw_show_mlocks(m, o);
+	show_syncpts(m, o, true);
+	host1x_debug_output(o, "---- channels ----\n");
+
+	for (i = 0; i < m->info->nb_channels; ++i) {
+		struct host1x_channel *ch = host1x_channel_get_index(m, i);
+
+		if (ch) {
+			show_channel(ch, o, show_fifo);
+			host1x_channel_put(ch);
+		}
+	}
+}
+
+static int host1x_debug_all_show(struct seq_file *s, void *unused)
+{
+	struct output o = {
+		.fn = write_to_seqfile,
+		.ctx = s
+	};
+
+	show_all(s->private, &o, true);
+
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(host1x_debug_all);
+
+static int host1x_debug_show(struct seq_file *s, void *unused)
+{
+	struct output o = {
+		.fn = write_to_seqfile,
+		.ctx = s
+	};
+
+	show_all(s->private, &o, false);
+
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(host1x_debug);
+
+static void host1x_debugfs_init(struct host1x *host1x)
+{
+	struct dentry *de = debugfs_create_dir("tegra-host1x", NULL);
+
+	/* Store the created entry */
+	host1x->debugfs = de;
+
+	debugfs_create_file("status", S_IRUGO, de, host1x, &host1x_debug_fops);
+	debugfs_create_file("status_all", S_IRUGO, de, host1x,
+			    &host1x_debug_all_fops);
+
+	debugfs_create_u32("trace_cmdbuf", S_IRUGO|S_IWUSR, de,
+			   &host1x_debug_trace_cmdbuf);
+
+	host1x_hw_debug_init(host1x, de);
+
+	debugfs_create_u32("force_timeout_pid", S_IRUGO|S_IWUSR, de,
+			   &host1x_debug_force_timeout_pid);
+	debugfs_create_u32("force_timeout_val", S_IRUGO|S_IWUSR, de,
+			   &host1x_debug_force_timeout_val);
+	debugfs_create_u32("force_timeout_channel", S_IRUGO|S_IWUSR, de,
+			   &host1x_debug_force_timeout_channel);
+}
+
+static void host1x_debugfs_exit(struct host1x *host1x)
+{
+	debugfs_remove_recursive(host1x->debugfs);
+}
+
+void host1x_debug_init(struct host1x *host1x)
+{
+	if (IS_ENABLED(CONFIG_DEBUG_FS))
+		host1x_debugfs_init(host1x);
+}
+
+void host1x_debug_deinit(struct host1x *host1x)
+{
+	if (IS_ENABLED(CONFIG_DEBUG_FS))
+		host1x_debugfs_exit(host1x);
+}
+
+void host1x_debug_dump(struct host1x *host1x)
+{
+	struct output o = {
+		.fn = write_to_printk
+	};
+
+	show_all(host1x, &o, true);
+}