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, 231 insertions, 0 deletions

diff --git a/drivers/iommu/intel/svm.c b/drivers/iommu/intel/svm.c
new file mode 100644
index 000000000..432e35723
--- /dev/null
+++ b/drivers/iommu/intel/svm.c
@@ -0,0 +1,231 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright © 2015 Intel Corporation.
+ *
+ * Authors: David Woodhouse <dwmw2@infradead.org>
+ */
+
+#include <linux/mmu_notifier.h>
+#include <linux/sched.h>
+#include <linux/sched/mm.h>
+#include <linux/slab.h>
+#include <linux/rculist.h>
+#include <linux/pci.h>
+#include <linux/pci-ats.h>
+#include <linux/dmar.h>
+#include <linux/interrupt.h>
+#include <linux/mm_types.h>
+#include <linux/xarray.h>
+#include <asm/page.h>
+#include <asm/fpu/api.h>
+
+#include "iommu.h"
+#include "pasid.h"
+#include "perf.h"
+#include "../iommu-pages.h"
+#include "trace.h"
+
+void intel_svm_check(struct intel_iommu *iommu)
+{
+	if (!pasid_supported(iommu) || !ecap_smpwc(iommu->ecap))
+		return;
+
+	if (cpu_feature_enabled(X86_FEATURE_GBPAGES) &&
+	    !cap_fl1gp_support(iommu->cap)) {
+		pr_err("%s SVM disabled, incompatible 1GB page capability\n",
+		       iommu->name);
+		return;
+	}
+
+	if (cpu_feature_enabled(X86_FEATURE_LA57) &&
+	    !cap_fl5lp_support(iommu->cap)) {
+		pr_err("%s SVM disabled, incompatible paging mode\n",
+		       iommu->name);
+		return;
+	}
+
+	iommu->flags |= VTD_FLAG_SVM_CAPABLE;
+}
+
+/* Pages have been freed at this point */
+static void intel_arch_invalidate_secondary_tlbs(struct mmu_notifier *mn,
+					struct mm_struct *mm,
+					unsigned long start, unsigned long end)
+{
+	struct dmar_domain *domain = container_of(mn, struct dmar_domain, notifier);
+
+	if (start == 0 && end == ULONG_MAX) {
+		cache_tag_flush_all(domain);
+		return;
+	}
+
+	/*
+	 * The mm_types defines vm_end as the first byte after the end address,
+	 * different from IOMMU subsystem using the last address of an address
+	 * range.
+	 */
+	cache_tag_flush_range(domain, start, end - 1, 0);
+}
+
+static void intel_mm_release(struct mmu_notifier *mn, struct mm_struct *mm)
+{
+	struct dmar_domain *domain = container_of(mn, struct dmar_domain, notifier);
+	struct dev_pasid_info *dev_pasid;
+	struct device_domain_info *info;
+	unsigned long flags;
+
+	/* This might end up being called from exit_mmap(), *before* the page
+	 * tables are cleared. And __mmu_notifier_release() will delete us from
+	 * the list of notifiers so that our invalidate_range() callback doesn't
+	 * get called when the page tables are cleared. So we need to protect
+	 * against hardware accessing those page tables.
+	 *
+	 * We do it by clearing the entry in the PASID table and then flushing
+	 * the IOTLB and the PASID table caches. This might upset hardware;
+	 * perhaps we'll want to point the PASID to a dummy PGD (like the zero
+	 * page) so that we end up taking a fault that the hardware really
+	 * *has* to handle gracefully without affecting other processes.
+	 */
+	spin_lock_irqsave(&domain->lock, flags);
+	list_for_each_entry(dev_pasid, &domain->dev_pasids, link_domain) {
+		info = dev_iommu_priv_get(dev_pasid->dev);
+		intel_pasid_tear_down_entry(info->iommu, dev_pasid->dev,
+					    dev_pasid->pasid, true);
+	}
+	spin_unlock_irqrestore(&domain->lock, flags);
+
+}
+
+static void intel_mm_free_notifier(struct mmu_notifier *mn)
+{
+	struct dmar_domain *domain = container_of(mn, struct dmar_domain, notifier);
+
+	kfree(domain->qi_batch);
+	kfree(domain);
+}
+
+static const struct mmu_notifier_ops intel_mmuops = {
+	.release = intel_mm_release,
+	.arch_invalidate_secondary_tlbs = intel_arch_invalidate_secondary_tlbs,
+	.free_notifier = intel_mm_free_notifier,
+};
+
+static int intel_iommu_sva_supported(struct device *dev)
+{
+	struct device_domain_info *info = dev_iommu_priv_get(dev);
+	struct intel_iommu *iommu;
+
+	if (!info || dmar_policy_off())
+		return -EINVAL;
+
+	iommu = info->iommu;
+	if (!iommu)
+		return -EINVAL;
+
+	if (!(iommu->flags & VTD_FLAG_SVM_CAPABLE))
+		return -ENODEV;
+
+	if (!info->pasid_enabled || !info->ats_enabled)
+		return -EINVAL;
+
+	/*
+	 * Devices having device-specific I/O fault handling should not
+	 * support PCI/PRI. The IOMMU side has no means to check the
+	 * capability of device-specific IOPF.  Therefore, IOMMU can only
+	 * default that if the device driver enables SVA on a non-PRI
+	 * device, it will handle IOPF in its own way.
+	 */
+	if (!info->pri_supported)
+		return 0;
+
+	/* Devices supporting PRI should have it enabled. */
+	if (!info->pri_enabled)
+		return -EINVAL;
+
+	return 0;
+}
+
+static int intel_svm_set_dev_pasid(struct iommu_domain *domain,
+				   struct device *dev, ioasid_t pasid,
+				   struct iommu_domain *old)
+{
+	struct device_domain_info *info = dev_iommu_priv_get(dev);
+	struct intel_iommu *iommu = info->iommu;
+	struct mm_struct *mm = domain->mm;
+	struct dev_pasid_info *dev_pasid;
+	unsigned long sflags;
+	int ret = 0;
+
+	ret = intel_iommu_sva_supported(dev);
+	if (ret)
+		return ret;
+
+	dev_pasid = domain_add_dev_pasid(domain, dev, pasid);
+	if (IS_ERR(dev_pasid))
+		return PTR_ERR(dev_pasid);
+
+	ret = iopf_for_domain_replace(domain, old, dev);
+	if (ret)
+		goto out_remove_dev_pasid;
+
+	/* Setup the pasid table: */
+	sflags = cpu_feature_enabled(X86_FEATURE_LA57) ? PASID_FLAG_FL5LP : 0;
+	sflags |= PASID_FLAG_PWSNP;
+	ret = __domain_setup_first_level(iommu, dev, pasid,
+					 FLPT_DEFAULT_DID, __pa(mm->pgd),
+					 sflags, old);
+	if (ret)
+		goto out_unwind_iopf;
+
+	domain_remove_dev_pasid(old, dev, pasid);
+
+	return 0;
+out_unwind_iopf:
+	iopf_for_domain_replace(old, domain, dev);
+out_remove_dev_pasid:
+	domain_remove_dev_pasid(domain, dev, pasid);
+	return ret;
+}
+
+static void intel_svm_domain_free(struct iommu_domain *domain)
+{
+	struct dmar_domain *dmar_domain = to_dmar_domain(domain);
+
+	/* dmar_domain free is deferred to the mmu free_notifier callback. */
+	mmu_notifier_put(&dmar_domain->notifier);
+}
+
+static const struct iommu_domain_ops intel_svm_domain_ops = {
+	.set_dev_pasid		= intel_svm_set_dev_pasid,
+	.free			= intel_svm_domain_free
+};
+
+struct iommu_domain *intel_svm_domain_alloc(struct device *dev,
+					    struct mm_struct *mm)
+{
+	struct dmar_domain *domain;
+	int ret;
+
+	ret = intel_iommu_sva_supported(dev);
+	if (ret)
+		return ERR_PTR(ret);
+
+	domain = kzalloc_obj(*domain);
+	if (!domain)
+		return ERR_PTR(-ENOMEM);
+
+	domain->domain.ops = &intel_svm_domain_ops;
+	INIT_LIST_HEAD(&domain->dev_pasids);
+	INIT_LIST_HEAD(&domain->cache_tags);
+	spin_lock_init(&domain->cache_lock);
+	spin_lock_init(&domain->lock);
+
+	domain->notifier.ops = &intel_mmuops;
+	ret = mmu_notifier_register(&domain->notifier, mm);
+	if (ret) {
+		kfree(domain);
+		return ERR_PTR(ret);
+	}
+
+	return &domain->domain;
+}