linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Dmitrii Bundin <dmitrii.bundin.a@gmail.com>
To: jan.kiszka@siemens.com, kbingham@kernel.org,
	akpm@linux-foundation.org, gregkh@linuxfoundation.org,
	mingo@redhat.com
Cc: x86@kernel.org, linux-mm@kvack.org, linux-kernel@vger.kernel.org,
	dmitrii.bundin.a@gmail.com, vbabka@suse.cz
Subject: [PATCH] scripts/gdb: add mm introspection utils
Date: Fri, 30 Dec 2022 19:35:12 +0300	[thread overview]
Message-ID: <20221230163512.23736-1-dmitrii.bundin.a@gmail.com> (raw)

This command provides a way to traverse the entire page hierarchy by a
given virtual address. In addition to qemu's commands info tlb/info mem
it provides the complete information about the paging structure for an
arbitrary virtual address. It supports 4Kb/2MB/1GB and 5 level paging.

Here is an example output for 2MB success translation:

CR3:
    CR3 BINARY DATA: 0x10c1d2002
    NEXT ENTRY PHYSICALL ADDRESS: 0x10c1d2000
    ---
    PAGE LEVEL WRITE THROUGH(bit 3): False
    PAGE LEVEL CACHE DISABLED(bit 4): False
LEVEL 4:
    ENTRY ADDRESS: 0xffff88810c1d27f0
    PAGE ENTRY BINARY DATA: 0x8000000105dca067
    NEXT ENTRY PHYSICALL ADDRESS: 0x105dca000
    ---
    ENTRY PRESENT(bit 0):      True
    READ/WRITE ACCESS ALLOWED(bit 1): True
    USER ACCESS ALLOWED(bit 2): True
    PAGE LEVEL WRITE THROUGH(bit 3): False
    PAGE LEVEL CACHE DISABLED(bit 4): False
    ENTRY HAS BEEN ACCESSED(bit 5): True
    PAGE SIZE(bit 7): False
    RESTART TO ORDINARY(bit 11): False
    EXECUTE DISABLE(bit 63): True
LEVEL 3:
    ENTRY ADDRESS: 0xffff888105dca9d0
    PAGE ENTRY BINARY DATA: 0x105c87067
    NEXT ENTRY PHYSICALL ADDRESS: 0x105c87000
    ---
    ENTRY PRESENT(bit 0):      True
    READ/WRITE ACCESS ALLOWED(bit 1): True
    USER ACCESS ALLOWED(bit 2): True
    PAGE LEVEL WRITE THROUGH(bit 3): False
    PAGE LEVEL CACHE DISABLED(bit 4): False
    ENTRY HAS BEEN ACCESSED(bit 5): True
    PAGE SIZE(bit 7): False
    RESTART TO ORDINARY(bit 11): False
    EXECUTE DISABLE(bit 63): False
LEVEL 2:
    ENTRY ADDRESS: 0xffff888105c87698
    PAGE ENTRY BINARY DATA: 0x80000001622008e7
    PAGE SIZE: 2MB
    PAGE PHYSICALL ADDRESS: 0x162200000
    ---
    ENTRY PRESENT(bit 0):      True
    READ/WRITE ACCESS ALLOWED(bit 1): True
    USER ACCESS ALLOWED(bit 2): True
    PAGE LEVEL WRITE THROUGH(bit 3): False
    PAGE LEVEL CACHE DISABLED(bit 4): False
    ENTRY HAS BEEN ACCESSED(bit 5): True
    PAGE DIRTY(bit 6): True
    PAGE SIZE(bit 7): True
    GLOBAL TRANSLATION(bit 8): False
    RESTART TO ORDINARY(bit 11): True
    PAT(bit 12): False
    PROTECTION KEY(bits (62, 59)): 0
    EXECUTE DISABLE(bit 63): True

Signed-off-by: Dmitrii Bundin <dmitrii.bundin.a@gmail.com>
---
 scripts/gdb/linux/mm.py    | 220 +++++++++++++++++++++++++++++++++++++
 scripts/gdb/vmlinux-gdb.py |   1 +
 2 files changed, 221 insertions(+)
 create mode 100644 scripts/gdb/linux/mm.py

diff --git a/scripts/gdb/linux/mm.py b/scripts/gdb/linux/mm.py
new file mode 100644
index 000000000000..c6f04e74edbd
--- /dev/null
+++ b/scripts/gdb/linux/mm.py
@@ -0,0 +1,220 @@
+#
+# gdb helper commands and functions for Linux kernel debugging
+#
+#  routines to introspect virtual memory
+#
+# Authors:
+#  Dmitrii Bundin <dmitrii.bundin.a@gmail.com>
+#
+# This work is licensed under the terms of the GNU GPL version 2.
+#
+
+import gdb
+
+from linux import utils
+
+PHYSICAL_ADDRESS_MASK = gdb.parse_and_eval('0xfffffffffffff')
+
+
+def page_mask(level=1):
+    # 4KB
+    if level == 1:
+        return gdb.parse_and_eval('(u64) ~0xfff')
+    # 2MB
+    elif level == 2:
+        return gdb.parse_and_eval('(u64) ~0x1fffff')
+    # 1GB
+    elif level == 3:
+        return gdb.parse_and_eval('(u64) ~0x3fffffff')
+    else:
+        raise Exception(f'Unknown page level: {level}')
+
+
+def _page_offset_base():
+    pob_symbol = gdb.lookup_global_symbol('page_offset_base')
+    pob = pob_symbol.name if pob_symbol else '0xffff888000000000'
+    return gdb.parse_and_eval(pob)
+
+
+def is_bit_defined_tupled(data, offset):
+    return offset, bool(data >> offset & 1)
+
+def content_tupled(data, bit_start, bit_end):
+    return (bit_end, bit_start), data >> bit_start & ((1 << (1 + bit_end - bit_start)) - 1)
+
+def entry_va(level, phys_addr, translating_va):
+        def start_bit(level):
+            if level == 5:
+                return 48
+            elif level == 4:
+                return 39
+            elif level == 3:
+                return 30
+            elif level == 2:
+                return 21
+            elif level == 1:
+                return 12
+            else:
+                raise Exception(f'Unknown level {level}')
+
+        entry_offset =  ((translating_va >> start_bit(level)) & 511) * 8
+        entry_va = _page_offset_base() + phys_addr + entry_offset
+        return entry_va
+
+class Cr3():
+    def __init__(self, cr3, page_levels):
+        self.cr3 = cr3
+        self.page_levels = page_levels
+        self.page_level_write_through = is_bit_defined_tupled(cr3, 3)
+        self.page_level_cache_disabled = is_bit_defined_tupled(cr3, 4)
+        self.next_entry_physical_address = cr3 & PHYSICAL_ADDRESS_MASK & page_mask()
+
+    def next_entry(self, va):
+        next_level = self.page_levels
+        return PageHierarchyEntry(entry_va(next_level, self.next_entry_physical_address, va), next_level)
+
+    def mk_string(self):
+            return f"""\
+CR3:
+    CR3 BINARY DATA: {hex(self.cr3)}
+    NEXT ENTRY PHYSICALL ADDRESS: {hex(self.next_entry_physical_address)}
+    ---
+    PAGE LEVEL WRITE THROUGH(bit {self.page_level_write_through[0]}): {self.page_level_write_through[1]}
+    PAGE LEVEL CACHE DISABLED(bit {self.page_level_cache_disabled[0]}): {self.page_level_cache_disabled[1]}
+"""
+
+
+class PageHierarchyEntry():
+    def __init__(self, address, level):
+        data = int.from_bytes(
+            memoryview(gdb.selected_inferior().read_memory(address, 8)),
+            "little"
+        )
+        if level == 1:
+            self.is_page = True
+            self.entry_present = is_bit_defined_tupled(data, 0)
+            self.read_write = is_bit_defined_tupled(data, 1)
+            self.user_access_allowed = is_bit_defined_tupled(data, 2)
+            self.page_level_write_through = is_bit_defined_tupled(data, 3)
+            self.page_level_cache_disabled = is_bit_defined_tupled(data, 4)
+            self.entry_was_accessed = is_bit_defined_tupled(data, 5)
+            self.dirty = is_bit_defined_tupled(data, 6)
+            self.pat = is_bit_defined_tupled(data, 7)
+            self.global_translation = is_bit_defined_tupled(data, 8)
+            self.page_physical_address = data & PHYSICAL_ADDRESS_MASK & page_mask(level)
+            self.next_entry_physical_address = None
+            self.hlat_restart_with_ordinary = is_bit_defined_tupled(data, 11)
+            self.protection_key = content_tupled(data, 59, 62)
+            self.executed_disable = is_bit_defined_tupled(data, 63)
+        else:
+            page_size = is_bit_defined_tupled(data, 7)
+            page_size_bit = page_size[1]
+            self.is_page = page_size_bit
+            self.entry_present = is_bit_defined_tupled(data, 0)
+            self.read_write = is_bit_defined_tupled(data, 1)
+            self.user_access_allowed = is_bit_defined_tupled(data, 2)
+            self.page_level_write_through = is_bit_defined_tupled(data, 3)
+            self.page_level_cache_disabled = is_bit_defined_tupled(data, 4)
+            self.entry_was_accessed = is_bit_defined_tupled(data, 5)
+            self.page_size = page_size
+            self.dirty = is_bit_defined_tupled(
+                data, 6) if page_size_bit else None
+            self.global_translation = is_bit_defined_tupled(
+                data, 8) if page_size_bit else None
+            self.pat = is_bit_defined_tupled(
+                data, 12) if page_size_bit else None
+            self.page_physical_address = data & PHYSICAL_ADDRESS_MASK & page_mask(level) if page_size_bit else None
+            self.next_entry_physical_address = None if page_size_bit else data & PHYSICAL_ADDRESS_MASK & page_mask()
+            self.hlat_restart_with_ordinary = is_bit_defined_tupled(data, 11)
+            self.protection_key = content_tupled(data, 59, 62) if page_size_bit else None
+            self.executed_disable = is_bit_defined_tupled(data, 63)
+        self.address = address
+        self.page_entry_binary_data = data
+        self.page_hierarchy_level = level
+
+    def next_entry(self, va):
+        if self.is_page or not self.entry_present[1]:
+            return None
+
+        next_level = self.page_hierarchy_level - 1
+        return PageHierarchyEntry(entry_va(next_level, self.next_entry_physical_address, va), next_level)
+
+
+    def mk_string(self):
+        if not self.entry_present[1]:
+            return f"""\
+LEVEL {self.page_hierarchy_level}:
+    ENTRY ADDRESS: {hex(self.address)}
+    PAGE ENTRY BINARY DATA: {hex(self.page_entry_binary_data)}
+    ---
+    PAGE ENTRY IS NOT PRESENT!
+"""
+        elif self.is_page:
+            return f"""\
+LEVEL {self.page_hierarchy_level}:
+    ENTRY ADDRESS: {hex(self.address)}
+    PAGE ENTRY BINARY DATA: {hex(self.page_entry_binary_data)}
+    PAGE SIZE: {'1GB' if self.page_hierarchy_level == 3 else '2MB' if self.page_hierarchy_level == 2 else '4KB' if self.page_hierarchy_level == 1 else 'Unknown page size for level:' + self.page_hierarchy_level}
+    PAGE PHYSICALL ADDRESS: {hex(self.page_physical_address)}
+    ---
+    ENTRY PRESENT(bit {self.entry_present[0]}):      {self.entry_present[1]}
+    READ/WRITE ACCESS ALLOWED(bit {self.read_write[0]}): {self.read_write[1]}
+    USER ACCESS ALLOWED(bit {self.user_access_allowed[0]}): {self.user_access_allowed[1]}
+    PAGE LEVEL WRITE THROUGH(bit {self.page_level_write_through[0]}): {self.page_level_write_through[1]}
+    PAGE LEVEL CACHE DISABLED(bit {self.page_level_cache_disabled[0]}): {self.page_level_cache_disabled[1]}
+    ENTRY HAS BEEN ACCESSED(bit {self.entry_was_accessed[0]}): {self.entry_was_accessed[1]}
+    PAGE DIRTY(bit {self.dirty[0]}): {self.dirty[1]}
+    """ + \
+    ("" if self.page_hierarchy_level == 1 else f"""PAGE SIZE(bit {self.page_size[0]}): {self.page_size[1]}
+    """) + \
+    f"""GLOBAL TRANSLATION(bit {self.global_translation[0]}): {self.global_translation[1]}
+    RESTART TO ORDINARY(bit {self.hlat_restart_with_ordinary[0]}): {self.hlat_restart_with_ordinary[1]}
+    PAT(bit {self.pat[0]}): {self.pat[1]}
+    PROTECTION KEY(bits {self.protection_key[0]}): {self.protection_key[1]}
+    EXECUTE DISABLE(bit {self.executed_disable[0]}): {self.executed_disable[1]}
+"""
+        else:
+            return f"""\
+LEVEL {self.page_hierarchy_level}:
+    ENTRY ADDRESS: {hex(self.address)}
+    PAGE ENTRY BINARY DATA: {hex(self.page_entry_binary_data)}
+    NEXT ENTRY PHYSICALL ADDRESS: {hex(self.next_entry_physical_address)}
+    ---
+    ENTRY PRESENT(bit {self.entry_present[0]}):      {self.entry_present[1]}
+    READ/WRITE ACCESS ALLOWED(bit {self.read_write[0]}): {self.read_write[1]}
+    USER ACCESS ALLOWED(bit {self.user_access_allowed[0]}): {self.user_access_allowed[1]}
+    PAGE LEVEL WRITE THROUGH(bit {self.page_level_write_through[0]}): {self.page_level_write_through[1]}
+    PAGE LEVEL CACHE DISABLED(bit {self.page_level_cache_disabled[0]}): {self.page_level_cache_disabled[1]}
+    ENTRY HAS BEEN ACCESSED(bit {self.entry_was_accessed[0]}): {self.entry_was_accessed[1]}
+    PAGE SIZE(bit {self.page_size[0]}): {self.page_size[1]}
+    RESTART TO ORDINARY(bit {self.hlat_restart_with_ordinary[0]}): {self.hlat_restart_with_ordinary[1]}
+    EXECUTE DISABLE(bit {self.executed_disable[0]}): {self.executed_disable[1]}
+"""
+
+
+class TranslateVM(gdb.Command):
+    """Prints the entire paging structure used to translate a given virtual address.
+
+Having an address space of the currently executed process translates the virtual address
+and prints detailed information of all paging structure levels used for the transaltion."""
+
+    def __init__(self):
+        super(TranslateVM, self).__init__('translate-vm', gdb.COMMAND_USER)
+
+    def invoke(self, arg, from_tty):
+        if utils.is_target_arch("x86"):
+            vm_address = gdb.parse_and_eval(f'{arg}')
+            cr3_data = gdb.parse_and_eval('$cr3')
+            cr4 = gdb.parse_and_eval('$cr4')
+            page_levels = 5 if cr4 & (1 << 12) else 4
+            page_entry = Cr3(cr3_data, page_levels)
+            while page_entry:
+                gdb.write(page_entry.mk_string())
+                page_entry = page_entry.next_entry(vm_address)
+        else:
+            gdb.GdbError("Virtual address translation is not"
+                         "supported for this arch")
+
+
+
+TranslateVM()
diff --git a/scripts/gdb/vmlinux-gdb.py b/scripts/gdb/vmlinux-gdb.py
index 4136dc2c59df..27bd7339bccc 100644
--- a/scripts/gdb/vmlinux-gdb.py
+++ b/scripts/gdb/vmlinux-gdb.py
@@ -37,3 +37,4 @@ else:
     import linux.clk
     import linux.genpd
     import linux.device
+    import linux.mm
-- 
2.17.1


             reply	other threads:[~2022-12-30 16:35 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-12-30 16:35 Dmitrii Bundin [this message]
2022-12-30 20:06 ` [PATCH] scripts/gdb: add mm introspection utils Mike Rapoport
2022-12-31 17:14   ` Dmitrii Bundin
2023-01-01  7:51     ` Mike Rapoport
2023-01-01 17:23       ` [PATCH v3] " Dmitrii Bundin
2023-01-02  8:59         ` Jan Kiszka
2023-01-02 17:10           ` [PATCH v4] " Dmitrii Bundin
2023-01-02 18:28             ` Mike Rapoport
2023-01-12 20:16             ` Greg KH
2023-01-13 17:51               ` [PATCH v5] " Dmitrii Bundin
2023-01-14  4:03                 ` Andrew Morton
2022-12-30 21:37 ` [PATCH] " Andrew Morton
2022-12-31 17:26   ` Dmitrii Bundin
2022-12-31 17:12 ` [PATCH v2] " Dmitrii Bundin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20221230163512.23736-1-dmitrii.bundin.a@gmail.com \
    --to=dmitrii.bundin.a@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=gregkh@linuxfoundation.org \
    --cc=jan.kiszka@siemens.com \
    --cc=kbingham@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mingo@redhat.com \
    --cc=vbabka@suse.cz \
    --cc=x86@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).