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, 335 insertions, 0 deletions

diff --git a/arch/xtensa/mm/cache.c b/arch/xtensa/mm/cache.c
new file mode 100644
index 000000000..5354df52d
--- /dev/null
+++ b/arch/xtensa/mm/cache.c
@@ -0,0 +1,335 @@
+/*
+ * arch/xtensa/mm/cache.c
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2001-2006 Tensilica Inc.
+ *
+ * Chris Zankel	<chris@zankel.net>
+ * Joe Taylor
+ * Marc Gauthier
+ *
+ */
+
+#include <linux/init.h>
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/memblock.h>
+#include <linux/swap.h>
+#include <linux/pagemap.h>
+#include <linux/pgtable.h>
+
+#include <asm/bootparam.h>
+#include <asm/mmu_context.h>
+#include <asm/tlb.h>
+#include <asm/tlbflush.h>
+#include <asm/page.h>
+
+/* 
+ * Note:
+ * The kernel provides one architecture bit PG_arch_1 in the page flags that 
+ * can be used for cache coherency.
+ *
+ * I$-D$ coherency.
+ *
+ * The Xtensa architecture doesn't keep the instruction cache coherent with
+ * the data cache. We use the architecture bit to indicate if the caches
+ * are coherent. The kernel clears this bit whenever a page is added to the
+ * page cache. At that time, the caches might not be in sync. We, therefore,
+ * define this flag as 'clean' if set.
+ *
+ * D-cache aliasing.
+ *
+ * With cache aliasing, we have to always flush the cache when pages are
+ * unmapped (see tlb_start_vma(). So, we use this flag to indicate a dirty
+ * page.
+ * 
+ *
+ *
+ */
+
+#if (DCACHE_WAY_SIZE > PAGE_SIZE)
+static inline void kmap_invalidate_coherent(struct page *page,
+					    unsigned long vaddr)
+{
+	if (!DCACHE_ALIAS_EQ(page_to_phys(page), vaddr)) {
+		unsigned long kvaddr;
+
+		if (!PageHighMem(page)) {
+			kvaddr = (unsigned long)page_to_virt(page);
+
+			__invalidate_dcache_page(kvaddr);
+		} else {
+			kvaddr = TLBTEMP_BASE_1 +
+				(page_to_phys(page) & DCACHE_ALIAS_MASK);
+
+			preempt_disable();
+			__invalidate_dcache_page_alias(kvaddr,
+						       page_to_phys(page));
+			preempt_enable();
+		}
+	}
+}
+
+static inline void *coherent_kvaddr(struct page *page, unsigned long base,
+				    unsigned long vaddr, unsigned long *paddr)
+{
+	*paddr = page_to_phys(page);
+	return (void *)(base + (vaddr & DCACHE_ALIAS_MASK));
+}
+
+void clear_user_highpage(struct page *page, unsigned long vaddr)
+{
+	struct folio *folio = page_folio(page);
+	unsigned long paddr;
+	void *kvaddr = coherent_kvaddr(page, TLBTEMP_BASE_1, vaddr, &paddr);
+
+	preempt_disable();
+	kmap_invalidate_coherent(page, vaddr);
+	set_bit(PG_arch_1, folio_flags(folio, 0));
+	clear_page_alias(kvaddr, paddr);
+	preempt_enable();
+}
+EXPORT_SYMBOL(clear_user_highpage);
+
+void copy_user_highpage(struct page *dst, struct page *src,
+			unsigned long vaddr, struct vm_area_struct *vma)
+{
+	struct folio *folio = page_folio(dst);
+	unsigned long dst_paddr, src_paddr;
+	void *dst_vaddr = coherent_kvaddr(dst, TLBTEMP_BASE_1, vaddr,
+					  &dst_paddr);
+	void *src_vaddr = coherent_kvaddr(src, TLBTEMP_BASE_2, vaddr,
+					  &src_paddr);
+
+	preempt_disable();
+	kmap_invalidate_coherent(dst, vaddr);
+	set_bit(PG_arch_1, folio_flags(folio, 0));
+	copy_page_alias(dst_vaddr, src_vaddr, dst_paddr, src_paddr);
+	preempt_enable();
+}
+EXPORT_SYMBOL(copy_user_highpage);
+
+/*
+ * Any time the kernel writes to a user page cache page, or it is about to
+ * read from a page cache page this routine is called.
+ *
+ */
+
+void flush_dcache_folio(struct folio *folio)
+{
+	struct address_space *mapping = folio_flush_mapping(folio);
+
+	/*
+	 * If we have a mapping but the page is not mapped to user-space
+	 * yet, we simply mark this page dirty and defer flushing the 
+	 * caches until update_mmu().
+	 */
+
+	if (mapping && !mapping_mapped(mapping)) {
+		if (!test_bit(PG_arch_1, &folio->flags.f))
+			set_bit(PG_arch_1, &folio->flags.f);
+		return;
+
+	} else {
+		unsigned long phys = folio_pfn(folio) * PAGE_SIZE;
+		unsigned long temp = folio_pos(folio);
+		unsigned int i, nr = folio_nr_pages(folio);
+		unsigned long alias = !(DCACHE_ALIAS_EQ(temp, phys));
+		unsigned long virt;
+
+		/* 
+		 * Flush the page in kernel space and user space.
+		 * Note that we can omit that step if aliasing is not
+		 * an issue, but we do have to synchronize I$ and D$
+		 * if we have a mapping.
+		 */
+
+		if (!alias && !mapping)
+			return;
+
+		preempt_disable();
+		for (i = 0; i < nr; i++) {
+			virt = TLBTEMP_BASE_1 + (phys & DCACHE_ALIAS_MASK);
+			__flush_invalidate_dcache_page_alias(virt, phys);
+
+			virt = TLBTEMP_BASE_1 + (temp & DCACHE_ALIAS_MASK);
+
+			if (alias)
+				__flush_invalidate_dcache_page_alias(virt, phys);
+
+			if (mapping)
+				__invalidate_icache_page_alias(virt, phys);
+			phys += PAGE_SIZE;
+			temp += PAGE_SIZE;
+		}
+		preempt_enable();
+	}
+
+	/* There shouldn't be an entry in the cache for this page anymore. */
+}
+EXPORT_SYMBOL(flush_dcache_folio);
+
+/*
+ * For now, flush the whole cache. FIXME??
+ */
+
+void local_flush_cache_range(struct vm_area_struct *vma,
+		       unsigned long start, unsigned long end)
+{
+	__flush_invalidate_dcache_all();
+	__invalidate_icache_all();
+}
+EXPORT_SYMBOL(local_flush_cache_range);
+
+/* 
+ * Remove any entry in the cache for this page. 
+ *
+ * Note that this function is only called for user pages, so use the
+ * alias versions of the cache flush functions.
+ */
+
+void local_flush_cache_page(struct vm_area_struct *vma, unsigned long address,
+		      unsigned long pfn)
+{
+	/* Note that we have to use the 'alias' address to avoid multi-hit */
+
+	unsigned long phys = page_to_phys(pfn_to_page(pfn));
+	unsigned long virt = TLBTEMP_BASE_1 + (address & DCACHE_ALIAS_MASK);
+
+	preempt_disable();
+	__flush_invalidate_dcache_page_alias(virt, phys);
+	__invalidate_icache_page_alias(virt, phys);
+	preempt_enable();
+}
+EXPORT_SYMBOL(local_flush_cache_page);
+
+#endif /* DCACHE_WAY_SIZE > PAGE_SIZE */
+
+void update_mmu_cache_range(struct vm_fault *vmf, struct vm_area_struct *vma,
+		unsigned long addr, pte_t *ptep, unsigned int nr)
+{
+	unsigned long pfn = pte_pfn(*ptep);
+	struct folio *folio;
+	unsigned int i;
+
+	if (!pfn_valid(pfn))
+		return;
+
+	folio = page_folio(pfn_to_page(pfn));
+
+	/* Invalidate old entries in TLBs */
+	for (i = 0; i < nr; i++)
+		flush_tlb_page(vma, addr + i * PAGE_SIZE);
+	nr = folio_nr_pages(folio);
+
+#if (DCACHE_WAY_SIZE > PAGE_SIZE)
+
+	if (!folio_test_reserved(folio) && test_bit(PG_arch_1, &folio->flags.f)) {
+		unsigned long phys = folio_pfn(folio) * PAGE_SIZE;
+		unsigned long tmp;
+
+		preempt_disable();
+		for (i = 0; i < nr; i++) {
+			tmp = TLBTEMP_BASE_1 + (phys & DCACHE_ALIAS_MASK);
+			__flush_invalidate_dcache_page_alias(tmp, phys);
+			tmp = TLBTEMP_BASE_1 + (addr & DCACHE_ALIAS_MASK);
+			__flush_invalidate_dcache_page_alias(tmp, phys);
+			__invalidate_icache_page_alias(tmp, phys);
+			phys += PAGE_SIZE;
+		}
+		preempt_enable();
+
+		clear_bit(PG_arch_1, &folio->flags.f);
+	}
+#else
+	if (!folio_test_reserved(folio) && !test_bit(PG_arch_1, &folio->flags.f)
+	    && (vma->vm_flags & VM_EXEC) != 0) {
+		for (i = 0; i < nr; i++) {
+			void *paddr = kmap_local_folio(folio, i * PAGE_SIZE);
+			__flush_dcache_page((unsigned long)paddr);
+			__invalidate_icache_page((unsigned long)paddr);
+			kunmap_local(paddr);
+		}
+		set_bit(PG_arch_1, &folio->flags.f);
+	}
+#endif
+}
+
+/*
+ * access_process_vm() has called get_user_pages(), which has done a
+ * flush_dcache_page() on the page.
+ */
+
+#if (DCACHE_WAY_SIZE > PAGE_SIZE)
+
+void copy_to_user_page(struct vm_area_struct *vma, struct page *page,
+		unsigned long vaddr, void *dst, const void *src,
+		unsigned long len)
+{
+	unsigned long phys = page_to_phys(page);
+	unsigned long alias = !(DCACHE_ALIAS_EQ(vaddr, phys));
+
+	/* Flush and invalidate user page if aliased. */
+
+	if (alias) {
+		unsigned long t = TLBTEMP_BASE_1 + (vaddr & DCACHE_ALIAS_MASK);
+		preempt_disable();
+		__flush_invalidate_dcache_page_alias(t, phys);
+		preempt_enable();
+	}
+
+	/* Copy data */
+	
+	memcpy(dst, src, len);
+
+	/*
+	 * Flush and invalidate kernel page if aliased and synchronize 
+	 * data and instruction caches for executable pages. 
+	 */
+
+	if (alias) {
+		unsigned long t = TLBTEMP_BASE_1 + (vaddr & DCACHE_ALIAS_MASK);
+
+		preempt_disable();
+		__flush_invalidate_dcache_range((unsigned long) dst, len);
+		if ((vma->vm_flags & VM_EXEC) != 0)
+			__invalidate_icache_page_alias(t, phys);
+		preempt_enable();
+
+	} else if ((vma->vm_flags & VM_EXEC) != 0) {
+		__flush_dcache_range((unsigned long)dst,len);
+		__invalidate_icache_range((unsigned long) dst, len);
+	}
+}
+
+extern void copy_from_user_page(struct vm_area_struct *vma, struct page *page,
+		unsigned long vaddr, void *dst, const void *src,
+		unsigned long len)
+{
+	unsigned long phys = page_to_phys(page);
+	unsigned long alias = !(DCACHE_ALIAS_EQ(vaddr, phys));
+
+	/*
+	 * Flush user page if aliased. 
+	 * (Note: a simply flush would be sufficient) 
+	 */
+
+	if (alias) {
+		unsigned long t = TLBTEMP_BASE_1 + (vaddr & DCACHE_ALIAS_MASK);
+		preempt_disable();
+		__flush_invalidate_dcache_page_alias(t, phys);
+		preempt_enable();
+	}
+
+	memcpy(dst, src, len);
+}
+
+#endif