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

5 files changed, 483 insertions, 0 deletions

diff --git a/drivers/gpu/drm/vgem/Kconfig b/drivers/gpu/drm/vgem/Kconfig
new file mode 100644
index 000000000..c419cdadd
--- /dev/null
+++ b/drivers/gpu/drm/vgem/Kconfig
@@ -0,0 +1,9 @@
+# SPDX-License-Identifier: GPL-2.0-only
+config DRM_VGEM
+	tristate "Virtual GEM provider"
+	depends on DRM && MMU
+	select DRM_GEM_SHMEM_HELPER
+	help
+	  Choose this option to get a virtual graphics memory manager,
+	  as used by Mesa's software renderer for enhanced performance.
+	  If M is selected the module will be called vgem.
diff --git a/drivers/gpu/drm/vgem/Makefile b/drivers/gpu/drm/vgem/Makefile
new file mode 100644
index 000000000..55eb5be04
--- /dev/null
+++ b/drivers/gpu/drm/vgem/Makefile
@@ -0,0 +1,4 @@
+# SPDX-License-Identifier: GPL-2.0-only
+vgem-y := vgem_drv.o vgem_fence.o
+
+obj-$(CONFIG_DRM_VGEM)	+= vgem.o
diff --git a/drivers/gpu/drm/vgem/vgem_drv.c b/drivers/gpu/drm/vgem/vgem_drv.c
new file mode 100644
index 000000000..a9f5c247d
--- /dev/null
+++ b/drivers/gpu/drm/vgem/vgem_drv.c
@@ -0,0 +1,181 @@
+/*
+ * Copyright 2011 Red Hat, Inc.
+ * Copyright © 2014 The Chromium OS Authors
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software")
+ * to deal in the software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * them Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTIBILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES, OR OTHER LIABILITY, WHETHER
+ * IN AN ACTION OF CONTRACT, TORT, OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors:
+ *	Adam Jackson <ajax@redhat.com>
+ *	Ben Widawsky <ben@bwidawsk.net>
+ */
+
+/*
+ * This is vgem, a (non-hardware-backed) GEM service.  This is used by Mesa's
+ * software renderer and the X server for efficient buffer sharing.
+ */
+
+#include <linux/dma-buf.h>
+#include <linux/module.h>
+#include <linux/device/faux.h>
+#include <linux/shmem_fs.h>
+#include <linux/vmalloc.h>
+
+#include <drm/drm_drv.h>
+#include <drm/drm_file.h>
+#include <drm/drm_gem_shmem_helper.h>
+#include <drm/drm_ioctl.h>
+#include <drm/drm_managed.h>
+#include <drm/drm_prime.h>
+
+#include "vgem_drv.h"
+
+#define DRIVER_NAME	"vgem"
+#define DRIVER_DESC	"Virtual GEM provider"
+#define DRIVER_MAJOR	1
+#define DRIVER_MINOR	0
+
+static struct vgem_device {
+	struct drm_device drm;
+	struct faux_device *faux_dev;
+} *vgem_device;
+
+static int vgem_open(struct drm_device *dev, struct drm_file *file)
+{
+	struct vgem_file *vfile;
+	int ret;
+
+	vfile = kzalloc_obj(*vfile);
+	if (!vfile)
+		return -ENOMEM;
+
+	file->driver_priv = vfile;
+
+	ret = vgem_fence_open(vfile);
+	if (ret) {
+		kfree(vfile);
+		return ret;
+	}
+
+	return 0;
+}
+
+static void vgem_postclose(struct drm_device *dev, struct drm_file *file)
+{
+	struct vgem_file *vfile = file->driver_priv;
+
+	vgem_fence_close(vfile);
+	kfree(vfile);
+}
+
+static struct drm_ioctl_desc vgem_ioctls[] = {
+	DRM_IOCTL_DEF_DRV(VGEM_FENCE_ATTACH, vgem_fence_attach_ioctl, DRM_RENDER_ALLOW),
+	DRM_IOCTL_DEF_DRV(VGEM_FENCE_SIGNAL, vgem_fence_signal_ioctl, DRM_RENDER_ALLOW),
+};
+
+DEFINE_DRM_GEM_FOPS(vgem_driver_fops);
+
+static struct drm_gem_object *vgem_gem_create_object(struct drm_device *dev, size_t size)
+{
+	struct drm_gem_shmem_object *obj;
+
+	obj = kzalloc_obj(*obj);
+	if (!obj)
+		return ERR_PTR(-ENOMEM);
+
+	/*
+	 * vgem doesn't have any begin/end cpu access ioctls, therefore must use
+	 * coherent memory or dma-buf sharing just wont work.
+	 */
+	obj->map_wc = true;
+
+	return &obj->base;
+}
+
+static const struct drm_driver vgem_driver = {
+	.driver_features		= DRIVER_GEM | DRIVER_RENDER,
+	.open				= vgem_open,
+	.postclose			= vgem_postclose,
+	.ioctls				= vgem_ioctls,
+	.num_ioctls 			= ARRAY_SIZE(vgem_ioctls),
+	.fops				= &vgem_driver_fops,
+
+	DRM_GEM_SHMEM_DRIVER_OPS,
+	.gem_create_object		= vgem_gem_create_object,
+
+	.name	= DRIVER_NAME,
+	.desc	= DRIVER_DESC,
+	.major	= DRIVER_MAJOR,
+	.minor	= DRIVER_MINOR,
+};
+
+static int __init vgem_init(void)
+{
+	int ret;
+	struct faux_device *fdev;
+
+	fdev = faux_device_create("vgem", NULL, NULL);
+	if (!fdev)
+		return -ENODEV;
+
+	if (!devres_open_group(&fdev->dev, NULL, GFP_KERNEL)) {
+		ret = -ENOMEM;
+		goto out_unregister;
+	}
+
+	dma_coerce_mask_and_coherent(&fdev->dev,
+				     DMA_BIT_MASK(64));
+
+	vgem_device = devm_drm_dev_alloc(&fdev->dev, &vgem_driver,
+					 struct vgem_device, drm);
+	if (IS_ERR(vgem_device)) {
+		ret = PTR_ERR(vgem_device);
+		goto out_devres;
+	}
+	vgem_device->faux_dev = fdev;
+
+	/* Final step: expose the device/driver to userspace */
+	ret = drm_dev_register(&vgem_device->drm, 0);
+	if (ret)
+		goto out_devres;
+
+	return 0;
+
+out_devres:
+	devres_release_group(&fdev->dev, NULL);
+out_unregister:
+	faux_device_destroy(fdev);
+	return ret;
+}
+
+static void __exit vgem_exit(void)
+{
+	struct faux_device *fdev = vgem_device->faux_dev;
+
+	drm_dev_unregister(&vgem_device->drm);
+	devres_release_group(&fdev->dev, NULL);
+	faux_device_destroy(fdev);
+}
+
+module_init(vgem_init);
+module_exit(vgem_exit);
+
+MODULE_AUTHOR("Red Hat, Inc.");
+MODULE_AUTHOR("Intel Corporation");
+MODULE_DESCRIPTION(DRIVER_DESC);
+MODULE_LICENSE("GPL and additional rights");
diff --git a/drivers/gpu/drm/vgem/vgem_drv.h b/drivers/gpu/drm/vgem/vgem_drv.h
new file mode 100644
index 000000000..34cf63e6f
--- /dev/null
+++ b/drivers/gpu/drm/vgem/vgem_drv.h
@@ -0,0 +1,51 @@
+/*
+ * Copyright © 2012 Intel Corporation
+ * Copyright © 2014 The Chromium OS Authors
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Ben Widawsky <ben@bwidawsk.net>
+ *
+ */
+
+#ifndef _VGEM_DRV_H_
+#define _VGEM_DRV_H_
+
+#include <drm/drm_gem.h>
+#include <drm/drm_cache.h>
+
+#include <uapi/drm/vgem_drm.h>
+
+struct vgem_file {
+	struct idr fence_idr;
+	struct mutex fence_mutex;
+};
+
+int vgem_fence_open(struct vgem_file *file);
+int vgem_fence_attach_ioctl(struct drm_device *dev,
+			    void *data,
+			    struct drm_file *file);
+int vgem_fence_signal_ioctl(struct drm_device *dev,
+			    void *data,
+			    struct drm_file *file);
+void vgem_fence_close(struct vgem_file *file);
+
+#endif
diff --git a/drivers/gpu/drm/vgem/vgem_fence.c b/drivers/gpu/drm/vgem/vgem_fence.c
new file mode 100644
index 000000000..0e53ba7db
--- /dev/null
+++ b/drivers/gpu/drm/vgem/vgem_fence.c
@@ -0,0 +1,238 @@
+/*
+ * Copyright 2016 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software")
+ * to deal in the software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * them Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTIBILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES, OR OTHER LIABILITY, WHETHER
+ * IN AN ACTION OF CONTRACT, TORT, OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <linux/dma-buf.h>
+#include <linux/dma-resv.h>
+
+#include <drm/drm_file.h>
+
+#include "vgem_drv.h"
+
+#define VGEM_FENCE_TIMEOUT (10*HZ)
+
+struct vgem_fence {
+	struct dma_fence base;
+	struct spinlock lock;
+	struct timer_list timer;
+};
+
+static const char *vgem_fence_get_driver_name(struct dma_fence *fence)
+{
+	return "vgem";
+}
+
+static const char *vgem_fence_get_timeline_name(struct dma_fence *fence)
+{
+	return "unbound";
+}
+
+static void vgem_fence_release(struct dma_fence *base)
+{
+	struct vgem_fence *fence = container_of(base, typeof(*fence), base);
+
+	timer_delete_sync(&fence->timer);
+	dma_fence_free(&fence->base);
+}
+
+static const struct dma_fence_ops vgem_fence_ops = {
+	.get_driver_name = vgem_fence_get_driver_name,
+	.get_timeline_name = vgem_fence_get_timeline_name,
+	.release = vgem_fence_release,
+};
+
+static void vgem_fence_timeout(struct timer_list *t)
+{
+	struct vgem_fence *fence = timer_container_of(fence, t, timer);
+
+	dma_fence_signal(&fence->base);
+}
+
+static struct dma_fence *vgem_fence_create(struct vgem_file *vfile,
+					   unsigned int flags)
+{
+	struct vgem_fence *fence;
+
+	fence = kzalloc_obj(*fence);
+	if (!fence)
+		return NULL;
+
+	spin_lock_init(&fence->lock);
+	dma_fence_init(&fence->base, &vgem_fence_ops, &fence->lock,
+		       dma_fence_context_alloc(1), 1);
+
+	timer_setup(&fence->timer, vgem_fence_timeout, TIMER_IRQSAFE);
+
+	/* We force the fence to expire within 10s to prevent driver hangs */
+	mod_timer(&fence->timer, jiffies + VGEM_FENCE_TIMEOUT);
+
+	return &fence->base;
+}
+
+/*
+ * vgem_fence_attach_ioctl (DRM_IOCTL_VGEM_FENCE_ATTACH):
+ *
+ * Create and attach a fence to the vGEM handle. This fence is then exposed
+ * via the dma-buf reservation object and visible to consumers of the exported
+ * dma-buf. If the flags contain VGEM_FENCE_WRITE, the fence indicates the
+ * vGEM buffer is being written to by the client and is exposed as an exclusive
+ * fence, otherwise the fence indicates the client is current reading from the
+ * buffer and all future writes should wait for the client to signal its
+ * completion. Note that if a conflicting fence is already on the dma-buf (i.e.
+ * an exclusive fence when adding a read, or any fence when adding a write),
+ * -EBUSY is reported. Serialisation between operations should be handled
+ * by waiting upon the dma-buf.
+ *
+ * This returns the handle for the new fence that must be signaled within 10
+ * seconds (or otherwise it will automatically expire). See
+ * vgem_fence_signal_ioctl (DRM_IOCTL_VGEM_FENCE_SIGNAL).
+ *
+ * If the vGEM handle does not exist, vgem_fence_attach_ioctl returns -ENOENT.
+ */
+int vgem_fence_attach_ioctl(struct drm_device *dev,
+			    void *data,
+			    struct drm_file *file)
+{
+	struct drm_vgem_fence_attach *arg = data;
+	struct vgem_file *vfile = file->driver_priv;
+	struct dma_resv *resv;
+	struct drm_gem_object *obj;
+	enum dma_resv_usage usage;
+	struct dma_fence *fence;
+	int ret;
+
+	if (arg->flags & ~VGEM_FENCE_WRITE)
+		return -EINVAL;
+
+	if (arg->pad)
+		return -EINVAL;
+
+	obj = drm_gem_object_lookup(file, arg->handle);
+	if (!obj)
+		return -ENOENT;
+
+	fence = vgem_fence_create(vfile, arg->flags);
+	if (!fence) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
+	/* Check for a conflicting fence */
+	resv = obj->resv;
+	usage = dma_resv_usage_rw(arg->flags & VGEM_FENCE_WRITE);
+	if (!dma_resv_test_signaled(resv, usage)) {
+		ret = -EBUSY;
+		goto err_fence;
+	}
+
+	/* Expose the fence via the dma-buf */
+	dma_resv_lock(resv, NULL);
+	ret = dma_resv_reserve_fences(resv, 1);
+	if (!ret)
+		dma_resv_add_fence(resv, fence, arg->flags & VGEM_FENCE_WRITE ?
+				   DMA_RESV_USAGE_WRITE : DMA_RESV_USAGE_READ);
+	dma_resv_unlock(resv);
+
+	/* Record the fence in our idr for later signaling */
+	if (ret == 0) {
+		mutex_lock(&vfile->fence_mutex);
+		ret = idr_alloc(&vfile->fence_idr, fence, 1, 0, GFP_KERNEL);
+		mutex_unlock(&vfile->fence_mutex);
+		if (ret > 0) {
+			arg->out_fence = ret;
+			ret = 0;
+		}
+	}
+err_fence:
+	if (ret) {
+		dma_fence_signal(fence);
+		dma_fence_put(fence);
+	}
+err:
+	drm_gem_object_put(obj);
+	return ret;
+}
+
+/*
+ * vgem_fence_signal_ioctl (DRM_IOCTL_VGEM_FENCE_SIGNAL):
+ *
+ * Signal and consume a fence ealier attached to a vGEM handle using
+ * vgem_fence_attach_ioctl (DRM_IOCTL_VGEM_FENCE_ATTACH).
+ *
+ * All fences must be signaled within 10s of attachment or otherwise they
+ * will automatically expire (and a vgem_fence_signal_ioctl returns -ETIMEDOUT).
+ *
+ * Signaling a fence indicates to all consumers of the dma-buf that the
+ * client has completed the operation associated with the fence, and that the
+ * buffer is then ready for consumption.
+ *
+ * If the fence does not exist (or has already been signaled by the client),
+ * vgem_fence_signal_ioctl returns -ENOENT.
+ */
+int vgem_fence_signal_ioctl(struct drm_device *dev,
+			    void *data,
+			    struct drm_file *file)
+{
+	struct vgem_file *vfile = file->driver_priv;
+	struct drm_vgem_fence_signal *arg = data;
+	struct dma_fence *fence;
+	int ret = 0;
+
+	if (arg->flags)
+		return -EINVAL;
+
+	mutex_lock(&vfile->fence_mutex);
+	fence = idr_replace(&vfile->fence_idr, NULL, arg->fence);
+	mutex_unlock(&vfile->fence_mutex);
+	if (!fence)
+		return -ENOENT;
+	if (IS_ERR(fence))
+		return PTR_ERR(fence);
+
+	if (dma_fence_is_signaled(fence))
+		ret = -ETIMEDOUT;
+
+	dma_fence_signal(fence);
+	dma_fence_put(fence);
+	return ret;
+}
+
+int vgem_fence_open(struct vgem_file *vfile)
+{
+	mutex_init(&vfile->fence_mutex);
+	idr_init_base(&vfile->fence_idr, 1);
+
+	return 0;
+}
+
+static int __vgem_fence_idr_fini(int id, void *p, void *data)
+{
+	dma_fence_signal(p);
+	dma_fence_put(p);
+	return 0;
+}
+
+void vgem_fence_close(struct vgem_file *vfile)
+{
+	idr_for_each(&vfile->fence_idr, __vgem_fence_idr_fini, vfile);
+	idr_destroy(&vfile->fence_idr);
+	mutex_destroy(&vfile->fence_mutex);
+}