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, 365 insertions, 0 deletions

diff --git a/Documentation/sphinx/rstFlatTable.py b/Documentation/sphinx/rstFlatTable.py
new file mode 100755
index 000000000..3d19569e5
--- /dev/null
+++ b/Documentation/sphinx/rstFlatTable.py
@@ -0,0 +1,365 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8; mode: python -*-
+# SPDX-License-Identifier: GPL-2.0
+# pylint: disable=C0330, R0903, R0912
+
+"""
+    flat-table
+    ~~~~~~~~~~
+
+    Implementation of the ``flat-table`` reST-directive.
+
+    :copyright:  Copyright (C) 2016  Markus Heiser
+    :license:    GPL Version 2, June 1991 see linux/COPYING for details.
+
+    The ``flat-table`` (:py:class:`FlatTable`) is a double-stage list similar to
+    the ``list-table`` with some additional features:
+
+    * *column-span*: with the role ``cspan`` a cell can be extended through
+      additional columns
+
+    * *row-span*: with the role ``rspan`` a cell can be extended through
+      additional rows
+
+    * *auto span* rightmost cell of a table row over the missing cells on the
+      right side of that table-row.  With Option ``:fill-cells:`` this behavior
+      can be changed from *auto span* to *auto fill*, which automatically inserts
+      (empty) cells instead of spanning the last cell.
+
+    Options:
+
+    * header-rows:   [int] count of header rows
+    * stub-columns:  [int] count of stub columns
+    * widths:        [[int] [int] ... ] widths of columns
+    * fill-cells:    instead of autospann missing cells, insert missing cells
+
+    roles:
+
+    * cspan: [int] additionale columns (*morecols*)
+    * rspan: [int] additionale rows (*morerows*)
+"""
+
+# ==============================================================================
+# imports
+# ==============================================================================
+
+from docutils import nodes
+from docutils.parsers.rst import directives, roles
+from docutils.parsers.rst.directives.tables import Table
+from docutils.utils import SystemMessagePropagation
+
+# ==============================================================================
+# common globals
+# ==============================================================================
+
+__version__  = '1.0'
+
+# ==============================================================================
+def setup(app):
+# ==============================================================================
+
+    app.add_directive("flat-table", FlatTable)
+    roles.register_local_role('cspan', c_span)
+    roles.register_local_role('rspan', r_span)
+
+    return dict(
+        version = __version__,
+        parallel_read_safe = True,
+        parallel_write_safe = True
+    )
+
+# ==============================================================================
+def c_span(name, rawtext, text, lineno, inliner, options=None, content=None):
+# ==============================================================================
+    # pylint: disable=W0613
+
+    options  = options if options is not None else {}
+    content  = content if content is not None else []
+    nodelist = [colSpan(span=int(text))]
+    msglist  = []
+    return nodelist, msglist
+
+# ==============================================================================
+def r_span(name, rawtext, text, lineno, inliner, options=None, content=None):
+# ==============================================================================
+    # pylint: disable=W0613
+
+    options  = options if options is not None else {}
+    content  = content if content is not None else []
+    nodelist = [rowSpan(span=int(text))]
+    msglist  = []
+    return nodelist, msglist
+
+
+# ==============================================================================
+class rowSpan(nodes.General, nodes.Element): pass # pylint: disable=C0103,C0321
+class colSpan(nodes.General, nodes.Element): pass # pylint: disable=C0103,C0321
+# ==============================================================================
+
+# ==============================================================================
+class FlatTable(Table):
+# ==============================================================================
+
+    """FlatTable (``flat-table``) directive"""
+
+    option_spec = {
+        'name': directives.unchanged
+        , 'class': directives.class_option
+        , 'header-rows': directives.nonnegative_int
+        , 'stub-columns': directives.nonnegative_int
+        , 'widths': directives.positive_int_list
+        , 'fill-cells' : directives.flag }
+
+    def run(self):
+
+        if not self.content:
+            error = self.state_machine.reporter.error(
+                'The "%s" directive is empty; content required.' % self.name,
+                nodes.literal_block(self.block_text, self.block_text),
+                line=self.lineno)
+            return [error]
+
+        title, messages = self.make_title()
+        node = nodes.Element()          # anonymous container for parsing
+        self.state.nested_parse(self.content, self.content_offset, node)
+
+        tableBuilder = ListTableBuilder(self)
+        tableBuilder.parseFlatTableNode(node)
+        tableNode = tableBuilder.buildTableNode()
+        # SDK.CONSOLE()  # print --> tableNode.asdom().toprettyxml()
+        if title:
+            tableNode.insert(0, title)
+        return [tableNode] + messages
+
+
+# ==============================================================================
+class ListTableBuilder(object):
+# ==============================================================================
+
+    """Builds a table from a double-stage list"""
+
+    def __init__(self, directive):
+        self.directive = directive
+        self.rows      = []
+        self.max_cols  = 0
+
+    def buildTableNode(self):
+
+        colwidths    = self.directive.get_column_widths(self.max_cols)
+        if isinstance(colwidths, tuple):
+            # Since docutils 0.13, get_column_widths returns a (widths,
+            # colwidths) tuple, where widths is a string (i.e. 'auto').
+            # See https://sourceforge.net/p/docutils/patches/120/.
+            colwidths = colwidths[1]
+        stub_columns = self.directive.options.get('stub-columns', 0)
+        header_rows  = self.directive.options.get('header-rows', 0)
+
+        table = nodes.table()
+        tgroup = nodes.tgroup(cols=len(colwidths))
+        table += tgroup
+
+
+        for colwidth in colwidths:
+            colspec = nodes.colspec(colwidth=colwidth)
+            # FIXME: It seems, that the stub method only works well in the
+            # absence of rowspan (observed by the html builder, the docutils-xml
+            # build seems OK).  This is not extraordinary, because there exists
+            # no table directive (except *this* flat-table) which allows to
+            # define coexistent of rowspan and stubs (there was no use-case
+            # before flat-table). This should be reviewed (later).
+            if stub_columns:
+                colspec.attributes['stub'] = 1
+                stub_columns -= 1
+            tgroup += colspec
+        stub_columns = self.directive.options.get('stub-columns', 0)
+
+        if header_rows:
+            thead = nodes.thead()
+            tgroup += thead
+            for row in self.rows[:header_rows]:
+                thead += self.buildTableRowNode(row)
+
+        tbody = nodes.tbody()
+        tgroup += tbody
+
+        for row in self.rows[header_rows:]:
+            tbody += self.buildTableRowNode(row)
+        return table
+
+    def buildTableRowNode(self, row_data, classes=None):
+        classes = [] if classes is None else classes
+        row = nodes.row()
+        for cell in row_data:
+            if cell is None:
+                continue
+            cspan, rspan, cellElements = cell
+
+            attributes = {"classes" : classes}
+            if rspan:
+                attributes['morerows'] = rspan
+            if cspan:
+                attributes['morecols'] = cspan
+            entry = nodes.entry(**attributes)
+            entry.extend(cellElements)
+            row += entry
+        return row
+
+    def raiseError(self, msg):
+        error =  self.directive.state_machine.reporter.error(
+            msg
+            , nodes.literal_block(self.directive.block_text
+                                  , self.directive.block_text)
+            , line = self.directive.lineno )
+        raise SystemMessagePropagation(error)
+
+    def parseFlatTableNode(self, node):
+        """parses the node from a :py:class:`FlatTable` directive's body"""
+
+        if len(node) != 1 or not isinstance(node[0], nodes.bullet_list):
+            self.raiseError(
+                'Error parsing content block for the "%s" directive: '
+                'exactly one bullet list expected.' % self.directive.name )
+
+        for rowNum, rowItem in enumerate(node[0]):
+            row = self.parseRowItem(rowItem, rowNum)
+            self.rows.append(row)
+        self.roundOffTableDefinition()
+
+    def roundOffTableDefinition(self):
+        """Round off the table definition.
+
+        This method rounds off the table definition in :py:member:`rows`.
+
+        * This method inserts the needed ``None`` values for the missing cells
+        arising from spanning cells over rows and/or columns.
+
+        * recount the :py:member:`max_cols`
+
+        * Autospan or fill (option ``fill-cells``) missing cells on the right
+          side of the table-row
+        """
+
+        y = 0
+        while y < len(self.rows):
+            x = 0
+
+            while x < len(self.rows[y]):
+                cell = self.rows[y][x]
+                if cell is None:
+                    x += 1
+                    continue
+                cspan, rspan = cell[:2]
+                # handle colspan in current row
+                for c in range(cspan):
+                    try:
+                        self.rows[y].insert(x+c+1, None)
+                    except: # pylint: disable=W0702
+                        # the user sets ambiguous rowspans
+                        pass # SDK.CONSOLE()
+                # handle colspan in spanned rows
+                for r in range(rspan):
+                    for c in range(cspan + 1):
+                        try:
+                            self.rows[y+r+1].insert(x+c, None)
+                        except: # pylint: disable=W0702
+                            # the user sets ambiguous rowspans
+                            pass # SDK.CONSOLE()
+                x += 1
+            y += 1
+
+        # Insert the missing cells on the right side. For this, first
+        # re-calculate the max columns.
+
+        for row in self.rows:
+            if self.max_cols < len(row):
+                self.max_cols = len(row)
+
+        # fill with empty cells or cellspan?
+
+        fill_cells = False
+        if 'fill-cells' in self.directive.options:
+            fill_cells = True
+
+        for row in self.rows:
+            x =  self.max_cols - len(row)
+            if x and not fill_cells:
+                if row[-1] is None:
+                    row.append( ( x - 1, 0, []) )
+                else:
+                    cspan, rspan, content = row[-1]
+                    row[-1] = (cspan + x, rspan, content)
+            elif x and fill_cells:
+                for i in range(x):
+                    row.append( (0, 0, nodes.comment()) )
+
+    def pprint(self):
+        # for debugging
+        retVal = "[   "
+        for row in self.rows:
+            retVal += "[ "
+            for col in row:
+                if col is None:
+                    retVal += ('%r' % col)
+                    retVal += "\n    , "
+                else:
+                    content = col[2][0].astext()
+                    if len (content) > 30:
+                        content = content[:30] + "..."
+                    retVal += ('(cspan=%s, rspan=%s, %r)'
+                               % (col[0], col[1], content))
+                    retVal += "]\n    , "
+            retVal = retVal[:-2]
+            retVal += "]\n  , "
+        retVal = retVal[:-2]
+        return retVal + "]"
+
+    def parseRowItem(self, rowItem, rowNum):
+        row = []
+        childNo = 0
+        error   = False
+        cell    = None
+        target  = None
+
+        for child in rowItem:
+            if (isinstance(child , nodes.comment)
+                or isinstance(child, nodes.system_message)):
+                pass
+            elif isinstance(child , nodes.target):
+                target = child
+            elif isinstance(child, nodes.bullet_list):
+                childNo += 1
+                cell = child
+            else:
+                error = True
+                break
+
+        if childNo != 1 or error:
+            self.raiseError(
+                'Error parsing content block for the "%s" directive: '
+                'two-level bullet list expected, but row %s does not '
+                'contain a second-level bullet list.'
+                % (self.directive.name, rowNum + 1))
+
+        for cellItem in cell:
+            cspan, rspan, cellElements = self.parseCellItem(cellItem)
+            if target is not None:
+                cellElements.insert(0, target)
+            row.append( (cspan, rspan, cellElements) )
+        return row
+
+    def parseCellItem(self, cellItem):
+        # search and remove cspan, rspan colspec from the first element in
+        # this listItem (field).
+        cspan = rspan = 0
+        if not len(cellItem):
+            return cspan, rspan, []
+        for elem in cellItem[0]:
+            if isinstance(elem, colSpan):
+                cspan = elem.get("span")
+                elem.parent.remove(elem)
+                continue
+            if isinstance(elem, rowSpan):
+                rspan = elem.get("span")
+                elem.parent.remove(elem)
+                continue
+        return cspan, rspan, cellItem[:]