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, 187 insertions, 0 deletions

diff --git a/drivers/gpu/host1x/context.c b/drivers/gpu/host1x/context.c
new file mode 100644
index 000000000..94ecc8769
--- /dev/null
+++ b/drivers/gpu/host1x/context.c
@@ -0,0 +1,187 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2021, NVIDIA Corporation.
+ */
+
+#include <linux/device.h>
+#include <linux/host1x_context_bus.h>
+#include <linux/kref.h>
+#include <linux/of.h>
+#include <linux/of_device.h>
+#include <linux/pid.h>
+#include <linux/slab.h>
+
+#include "context.h"
+#include "dev.h"
+
+static void host1x_memory_context_release(struct device *dev)
+{
+	/* context device is freed in host1x_memory_context_list_free() */
+}
+
+int host1x_memory_context_list_init(struct host1x *host1x)
+{
+	struct host1x_memory_context_list *cdl = &host1x->context_list;
+	struct device_node *node = host1x->dev->of_node;
+	struct host1x_memory_context *ctx;
+	unsigned int devs, i;
+	int err;
+
+	cdl->devs = NULL;
+	cdl->len = 0;
+	mutex_init(&cdl->lock);
+
+	err = of_property_count_u32_elems(node, "iommu-map");
+	if (err < 0)
+		return 0;
+
+	devs = 0;
+
+	for (i = 0; i < err / 4; i++) {
+		u32 length;
+
+		of_property_read_u32_index(node, "iommu-map", i * 4 + 3, &length);
+		devs += length;
+	}
+
+	cdl->len = devs;
+	cdl->devs = kzalloc_objs(*cdl->devs, cdl->len);
+	if (!cdl->devs)
+		return -ENOMEM;
+
+	for (i = 0; i < cdl->len; i++) {
+		ctx = &cdl->devs[i];
+
+		ctx->host = host1x;
+
+		device_initialize(&ctx->dev);
+
+		/*
+		 * Due to an issue with T194 NVENC, only 38 bits can be used.
+		 * Anyway, 256GiB of IOVA ought to be enough for anyone.
+		 */
+		ctx->dma_mask = DMA_BIT_MASK(38);
+		ctx->dev.dma_mask = &ctx->dma_mask;
+		ctx->dev.coherent_dma_mask = ctx->dma_mask;
+		dev_set_name(&ctx->dev, "host1x-ctx.%d", i);
+		ctx->dev.bus = &host1x_context_device_bus_type;
+		ctx->dev.parent = host1x->dev;
+		ctx->dev.release = host1x_memory_context_release;
+
+		ctx->dev.dma_parms = &ctx->dma_parms;
+		dma_set_max_seg_size(&ctx->dev, UINT_MAX);
+
+		err = device_add(&ctx->dev);
+		if (err) {
+			dev_err(host1x->dev, "could not add context device %d: %d\n", i, err);
+			put_device(&ctx->dev);
+			goto unreg_devices;
+		}
+
+		err = of_dma_configure_id(&ctx->dev, node, true, &i);
+		if (err) {
+			dev_err(host1x->dev, "IOMMU configuration failed for context device %d: %d\n",
+				i, err);
+			device_unregister(&ctx->dev);
+			goto unreg_devices;
+		}
+
+		if (!tegra_dev_iommu_get_stream_id(&ctx->dev, &ctx->stream_id) ||
+		    !device_iommu_mapped(&ctx->dev)) {
+			dev_err(host1x->dev, "Context device %d has no IOMMU!\n", i);
+			device_unregister(&ctx->dev);
+
+			/*
+			 * This means that if IOMMU is disabled but context devices
+			 * are defined in the device tree, Host1x will fail to probe.
+			 * That's probably OK in this time and age.
+			 */
+			err = -EINVAL;
+
+			goto unreg_devices;
+		}
+	}
+
+	return 0;
+
+unreg_devices:
+	while (i--)
+		device_unregister(&cdl->devs[i].dev);
+
+	kfree(cdl->devs);
+	cdl->devs = NULL;
+	cdl->len = 0;
+
+	return err;
+}
+
+void host1x_memory_context_list_free(struct host1x_memory_context_list *cdl)
+{
+	unsigned int i;
+
+	for (i = 0; i < cdl->len; i++)
+		device_unregister(&cdl->devs[i].dev);
+
+	kfree(cdl->devs);
+	cdl->len = 0;
+}
+
+struct host1x_memory_context *host1x_memory_context_alloc(struct host1x *host1x,
+							  struct device *dev,
+							  struct pid *pid)
+{
+	struct host1x_memory_context_list *cdl = &host1x->context_list;
+	struct host1x_memory_context *free = NULL;
+	int i;
+
+	if (!cdl->len)
+		return ERR_PTR(-EOPNOTSUPP);
+
+	mutex_lock(&cdl->lock);
+
+	for (i = 0; i < cdl->len; i++) {
+		struct host1x_memory_context *cd = &cdl->devs[i];
+
+		if (cd->dev.iommu->iommu_dev != dev->iommu->iommu_dev)
+			continue;
+
+		if (cd->owner == pid) {
+			refcount_inc(&cd->ref);
+			mutex_unlock(&cdl->lock);
+			return cd;
+		} else if (!cd->owner && !free) {
+			free = cd;
+		}
+	}
+
+	if (!free) {
+		mutex_unlock(&cdl->lock);
+		return ERR_PTR(-EBUSY);
+	}
+
+	refcount_set(&free->ref, 1);
+	free->owner = get_pid(pid);
+
+	mutex_unlock(&cdl->lock);
+
+	return free;
+}
+EXPORT_SYMBOL_GPL(host1x_memory_context_alloc);
+
+void host1x_memory_context_get(struct host1x_memory_context *cd)
+{
+	refcount_inc(&cd->ref);
+}
+EXPORT_SYMBOL_GPL(host1x_memory_context_get);
+
+void host1x_memory_context_put(struct host1x_memory_context *cd)
+{
+	struct host1x_memory_context_list *cdl = &cd->host->context_list;
+
+	if (refcount_dec_and_mutex_lock(&cd->ref, &cdl->lock)) {
+		put_pid(cd->owner);
+		cd->owner = NULL;
+		mutex_unlock(&cdl->lock);
+	}
+}
+EXPORT_SYMBOL_GPL(host1x_memory_context_put);