Subject: [PATCH v2 0/8] Add GDB memory helper commands

Hi all,

I've created some GDB commands I think useful when I debug
some memory issues and kernel module issue.

For memory issue, we would like to get slabinfo, slabtrace,
page_owner and vmallocinfo to debug the memory issues.

For module issue, we would like to query kernel module name
when we get a module text address and load module
symbol by specific path.

Patch 1-2:
- Add kernel module related command.
Patch 3-5:
- Prepares for the memory-related command.
Patch 6-8:
- Add memory-related commands.

V1->V2:
- Fix s390 build error with defconfig (Stephen)
- Use freeptr_t structure for dereference freepointer
in slab
- Add CONFIG_STACKDEPOT statement to handle stackdepot
is disable
- Add CONFIG_PAGE_OWNER statement to handle page owner
is disable
- Rebase to linux-next/next-20230808

Kuan-Ying Lee (8):
scripts/gdb/symbols: add specific ko module load command
scripts/gdb/modules: add get module text support
scripts/gdb/utils: add common type usage
scripts/gdb/aarch64: add aarch64 page operation helper commands and
configs
scripts/gdb/stackdepot: Add stackdepot support
scripts/gdb/page_owner: add page owner support
scripts/gdb/slab: Add slab support
scripts/gdb/vmalloc: add vmallocinfo support

scripts/gdb/linux/constants.py.in | 52 +++
scripts/gdb/linux/mm.py | 582 +++++++++++++++++++-----------
scripts/gdb/linux/modules.py | 32 +-
scripts/gdb/linux/page_owner.py | 190 ++++++++++
scripts/gdb/linux/pgtable.py | 222 ++++++++++++
scripts/gdb/linux/slab.py | 326 +++++++++++++++++
scripts/gdb/linux/stackdepot.py | 55 +++
scripts/gdb/linux/symbols.py | 23 +-
scripts/gdb/linux/utils.py | 20 +
scripts/gdb/linux/vmalloc.py | 56 +++
scripts/gdb/vmlinux-gdb.py | 7 +-
11 files changed, 1358 insertions(+), 207 deletions(-)
create mode 100644 scripts/gdb/linux/page_owner.py
create mode 100644 scripts/gdb/linux/pgtable.py
create mode 100644 scripts/gdb/linux/slab.py
create mode 100644 scripts/gdb/linux/stackdepot.py
create mode 100644 scripts/gdb/linux/vmalloc.py

--
2.18.0



Subject: [PATCH v2 7/8] scripts/gdb/slab: Add slab support

Add 'lx-slabinfo' and 'lx-slabtrace' support.

This GDB scripts print slabinfo and slabtrace for user
to analyze slab memory usage.

Example output like below:
(gdb) lx-slabinfo
Pointer | name | active_objs | num_objs | objsize | objperslab | pagesperslab
------------------ | -------------------- | ------------ | ------------ | -------- | ----------- | -------------
0xffff0000c59df480 | p9_req_t | 0 | 0 | 280 | 29 | 2
0xffff0000c59df280 | isp1760_qh | 0 | 0 | 160 | 25 | 1
0xffff0000c59df080 | isp1760_qtd | 0 | 0 | 184 | 22 | 1
0xffff0000c59dee80 | isp1760_urb_listite | 0 | 0 | 136 | 30 | 1
0xffff0000c59dec80 | asd_sas_event | 0 | 0 | 256 | 32 | 2
0xffff0000c59dea80 | sas_task | 0 | 0 | 448 | 36 | 4
0xffff0000c59de880 | bio-120 | 18 | 21 | 384 | 21 | 2
0xffff0000c59de680 | io_kiocb | 0 | 0 | 448 | 36 | 4
0xffff0000c59de480 | bfq_io_cq | 0 | 0 | 1504 | 21 | 8
0xffff0000c59de280 | bfq_queue | 0 | 0 | 720 | 22 | 4
0xffff0000c59de080 | mqueue_inode_cache | 1 | 28 | 1152 | 28 | 8
0xffff0000c59dde80 | v9fs_inode_cache | 0 | 0 | 832 | 39 | 8
...

(gdb) lx-slabtrace --cache_name kmalloc-1k
63 <tty_register_device_attr+508> waste=16632/264 age=46856/46871/46888 pid=1 cpus=6,
0xffff800008720240 <__kmem_cache_alloc_node+236>: mov x22, x0
0xffff80000862a4fc <kmalloc_trace+64>: mov x21, x0
0xffff8000095d086c <tty_register_device_attr+508>: mov x19, x0
0xffff8000095d0f98 <tty_register_driver+704>: cmn x0, #0x1, lsl #12
0xffff80000c2677e8 <vty_init+620>: Cannot access memory at address 0xffff80000c2677e8
0xffff80000c265a10 <tty_init+276>: Cannot access memory at address 0xffff80000c265a10
0xffff80000c26d3c4 <chr_dev_init+204>: Cannot access memory at address 0xffff80000c26d3c4
0xffff8000080161d4 <do_one_initcall+176>: mov w21, w0
0xffff80000c1c1b58 <kernel_init_freeable+956>: Cannot access memory at address 0xffff80000c1c1b58
0xffff80000acf1334 <kernel_init+36>: bl 0xffff8000081ac040 <async_synchronize_full>
0xffff800008018d00 <ret_from_fork+16>: mrs x28, sp_el0

(gdb) lx-slabtrace --cache_name kmalloc-1k --free
428 <not-available> age=4294958600 pid=0 cpus=0,

Signed-off-by: Kuan-Ying Lee <[email protected]>
---
scripts/gdb/linux/constants.py.in | 13 ++
scripts/gdb/linux/slab.py | 326 ++++++++++++++++++++++++++++++
scripts/gdb/vmlinux-gdb.py | 1 +
3 files changed, 340 insertions(+)
create mode 100644 scripts/gdb/linux/slab.py

diff --git a/scripts/gdb/linux/constants.py.in b/scripts/gdb/linux/constants.py.in
index 52f61d65f430..03fa6d2cfe01 100644
--- a/scripts/gdb/linux/constants.py.in
+++ b/scripts/gdb/linux/constants.py.in
@@ -20,6 +20,7 @@
#include <linux/of_fdt.h>
#include <linux/page_ext.h>
#include <linux/radix-tree.h>
+#include <linux/slab.h>
#include <linux/threads.h>

/* We need to stringify expanded macros so that they can be parsed */
@@ -95,6 +96,16 @@ if IS_BUILTIN(CONFIG_PAGE_OWNER):
LX_GDBPARSED(PAGE_EXT_OWNER)
LX_GDBPARSED(PAGE_EXT_OWNER_ALLOCATED)

+/* linux/slab.h */
+LX_GDBPARSED(SLAB_RED_ZONE)
+LX_GDBPARSED(SLAB_POISON)
+LX_GDBPARSED(SLAB_KMALLOC)
+LX_GDBPARSED(SLAB_HWCACHE_ALIGN)
+LX_GDBPARSED(SLAB_CACHE_DMA)
+LX_GDBPARSED(SLAB_CACHE_DMA32)
+LX_GDBPARSED(SLAB_STORE_USER)
+LX_GDBPARSED(SLAB_PANIC)
+
/* Kernel Configs */
LX_CONFIG(CONFIG_GENERIC_CLOCKEVENTS)
LX_CONFIG(CONFIG_GENERIC_CLOCKEVENTS_BROADCAST)
@@ -136,3 +147,5 @@ if IS_BUILTIN(CONFIG_NUMA):
LX_CONFIG(CONFIG_DEBUG_VIRTUAL)
LX_CONFIG(CONFIG_STACKDEPOT)
LX_CONFIG(CONFIG_PAGE_OWNER)
+LX_CONFIG(CONFIG_SLUB_DEBUG)
+LX_CONFIG(CONFIG_SLAB_FREELIST_HARDENED)
diff --git a/scripts/gdb/linux/slab.py b/scripts/gdb/linux/slab.py
new file mode 100644
index 000000000000..f012ba38c7d9
--- /dev/null
+++ b/scripts/gdb/linux/slab.py
@@ -0,0 +1,326 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Copyright (c) 2023 MediaTek Inc.
+#
+# Authors:
+# Kuan-Ying Lee <[email protected]>
+#
+
+import gdb
+import re
+import traceback
+from linux import lists, utils, stackdepot, constants, mm
+
+SLAB_RED_ZONE = constants.LX_SLAB_RED_ZONE
+SLAB_POISON = constants.LX_SLAB_POISON
+SLAB_KMALLOC = constants.LX_SLAB_KMALLOC
+SLAB_HWCACHE_ALIGN = constants.LX_SLAB_HWCACHE_ALIGN
+SLAB_CACHE_DMA = constants.LX_SLAB_CACHE_DMA
+SLAB_CACHE_DMA32 = constants.LX_SLAB_CACHE_DMA32
+SLAB_STORE_USER = constants.LX_SLAB_STORE_USER
+SLAB_PANIC = constants.LX_SLAB_PANIC
+
+OO_SHIFT = 16
+OO_MASK = (1 << OO_SHIFT) - 1
+
+if constants.LX_CONFIG_SLUB_DEBUG:
+ slab_type = utils.CachedType("struct slab")
+ slab_ptr_type = slab_type.get_type().pointer()
+ kmem_cache_type = utils.CachedType("struct kmem_cache")
+ kmem_cache_ptr_type = kmem_cache_type.get_type().pointer()
+ freeptr_t = utils.CachedType("freeptr_t")
+ freeptr_t_ptr = freeptr_t.get_type().pointer()
+
+ track_type = gdb.lookup_type('struct track')
+ track_alloc = int(gdb.parse_and_eval('TRACK_ALLOC'))
+ track_free = int(gdb.parse_and_eval('TRACK_FREE'))
+
+def slab_folio(slab):
+ return slab.cast(gdb.lookup_type("struct folio").pointer())
+
+def slab_address(slab):
+ p_ops = mm.page_ops().ops
+ folio = slab_folio(slab)
+ return p_ops.folio_address(folio)
+
+def for_each_object(cache, addr, slab_objects):
+ p = addr
+ if cache['flags'] & SLAB_RED_ZONE:
+ p += int(cache['red_left_pad'])
+ while p < addr + (slab_objects * cache['size']):
+ yield p
+ p = p + int(cache['size'])
+
+def get_info_end(cache):
+ if (cache['offset'] >= cache['inuse']):
+ return cache['inuse'] + gdb.lookup_type("void").pointer().sizeof
+ else:
+ return cache['inuse']
+
+def get_orig_size(cache, obj):
+ if cache['flags'] & SLAB_STORE_USER and cache['flags'] & SLAB_KMALLOC:
+ p = mm.page_ops().ops.kasan_reset_tag(obj)
+ p += get_info_end(cache)
+ p += gdb.lookup_type('struct track').sizeof * 2
+ p = p.cast(utils.get_uint_type().pointer())
+ return p.dereference()
+ else:
+ return cache['object_size']
+
+def get_track(cache, object_pointer, alloc):
+ p = object_pointer + get_info_end(cache)
+ p += (alloc * track_type.sizeof)
+ return p
+
+def oo_objects(x):
+ return int(x['x']) & OO_MASK
+
+def oo_order(x):
+ return int(x['x']) >> OO_SHIFT
+
+def reciprocal_divide(a, R):
+ t = (a * int(R['m'])) >> 32
+ return (t + ((a - t) >> int(R['sh1']))) >> int(R['sh2'])
+
+def __obj_to_index(cache, addr, obj):
+ return reciprocal_divide(int(mm.page_ops().ops.kasan_reset_tag(obj)) - addr, cache['reciprocal_size'])
+
+def swab64(x):
+ result = (((x & 0x00000000000000ff) << 56) | \
+ ((x & 0x000000000000ff00) << 40) | \
+ ((x & 0x0000000000ff0000) << 24) | \
+ ((x & 0x00000000ff000000) << 8) | \
+ ((x & 0x000000ff00000000) >> 8) | \
+ ((x & 0x0000ff0000000000) >> 24) | \
+ ((x & 0x00ff000000000000) >> 40) | \
+ ((x & 0xff00000000000000) >> 56))
+ return result
+
+def freelist_ptr_decode(cache, ptr, ptr_addr):
+ if constants.LX_CONFIG_SLAB_FREELIST_HARDENED:
+ return ptr['v'] ^ cache['random'] ^ swab64(int(ptr_addr))
+ else:
+ return ptr['v']
+
+def get_freepointer(cache, obj):
+ obj = mm.page_ops().ops.kasan_reset_tag(obj)
+ ptr_addr = obj + cache['offset']
+ p = ptr_addr.cast(freeptr_t_ptr).dereference()
+ return freelist_ptr_decode(cache, p, ptr_addr)
+
+def loc_exist(loc_track, addr, handle, waste):
+ for loc in loc_track:
+ if loc['addr'] == addr and loc['handle'] == handle and loc['waste'] == waste:
+ return loc
+ return None
+
+def add_location(loc_track, cache, track, orig_size):
+ jiffies = gdb.parse_and_eval("jiffies_64")
+ age = jiffies - track['when']
+ handle = 0
+ waste = cache['object_size'] - int(orig_size)
+ pid = int(track['pid'])
+ cpuid = int(track['cpu'])
+ addr = track['addr']
+ if constants.LX_CONFIG_STACKDEPOT:
+ handle = track['handle']
+
+ loc = loc_exist(loc_track, addr, handle, waste)
+ if loc:
+ loc['count'] += 1
+ if track['when']:
+ loc['sum_time'] += age
+ loc['min_time'] = min(loc['min_time'], age)
+ loc['max_time'] = max(loc['max_time'], age)
+ loc['min_pid'] = min(loc['min_pid'], pid)
+ loc['max_pid'] = max(loc['max_pid'], pid)
+ loc['cpus'].add(cpuid)
+ else:
+ loc_track.append({
+ 'count' : 1,
+ 'addr' : addr,
+ 'sum_time' : age,
+ 'min_time' : age,
+ 'max_time' : age,
+ 'min_pid' : pid,
+ 'max_pid' : pid,
+ 'handle' : handle,
+ 'waste' : waste,
+ 'cpus' : {cpuid}
+ }
+ )
+
+def slabtrace(alloc, cache_name):
+
+ def __fill_map(obj_map, cache, slab):
+ p = slab['freelist']
+ addr = slab_address(slab)
+ while p != gdb.Value(0):
+ index = __obj_to_index(cache, addr, p)
+ obj_map[index] = True # free objects
+ p = get_freepointer(cache, p)
+
+ # process every slab page on the slab_list (partial and full list)
+ def process_slab(loc_track, slab_list, alloc, cache):
+ for slab in lists.list_for_each_entry(slab_list, slab_ptr_type, "slab_list"):
+ obj_map[:] = [False] * oo_objects(cache['oo'])
+ __fill_map(obj_map, cache, slab)
+ addr = slab_address(slab)
+ for object_pointer in for_each_object(cache, addr, slab['objects']):
+ if obj_map[__obj_to_index(cache, addr, object_pointer)] == True:
+ continue
+ p = get_track(cache, object_pointer, alloc)
+ track = gdb.Value(p).cast(track_type.pointer())
+ if alloc == track_alloc:
+ size = get_orig_size(cache, object_pointer)
+ else:
+ size = cache['object_size']
+ add_location(loc_track, cache, track, size)
+ continue
+
+ slab_caches = gdb.parse_and_eval("slab_caches")
+ if mm.page_ops().ops.MAX_NUMNODES > 1:
+ nr_node_ids = int(gdb.parse_and_eval("nr_node_ids"))
+ else:
+ nr_node_ids = 1
+
+ target_cache = None
+ loc_track = []
+
+ for cache in lists.list_for_each_entry(slab_caches, kmem_cache_ptr_type, 'list'):
+ if cache['name'].string() == cache_name:
+ target_cache = cache
+ break
+
+ obj_map = [False] * oo_objects(target_cache['oo'])
+
+ if target_cache['flags'] & SLAB_STORE_USER:
+ for i in range(0, nr_node_ids):
+ cache_node = target_cache['node'][i]
+ if cache_node['nr_slabs']['counter'] == 0:
+ continue
+ process_slab(loc_track, cache_node['partial'], alloc, target_cache)
+ process_slab(loc_track, cache_node['full'], alloc, target_cache)
+ else:
+ raise gdb.GdbError("SLAB_STORE_USER is not set in %s" % target_cache['name'].string())
+
+ for loc in sorted(loc_track, key=lambda x:x['count'], reverse=True):
+ if loc['addr']:
+ addr = loc['addr'].cast(utils.get_ulong_type().pointer())
+ gdb.write("%d %s" % (loc['count'], str(addr).split(' ')[-1]))
+ else:
+ gdb.write("%d <not-available>" % loc['count'])
+
+ if loc['waste']:
+ gdb.write(" waste=%d/%d" % (loc['count'] * loc['waste'], loc['waste']))
+
+ if loc['sum_time'] != loc['min_time']:
+ gdb.write(" age=%d/%d/%d" % (loc['min_time'], loc['sum_time']/loc['count'], loc['max_time']))
+ else:
+ gdb.write(" age=%d" % loc['min_time'])
+
+ if loc['min_pid'] != loc['max_pid']:
+ gdb.write(" pid=%d-%d" % (loc['min_pid'], loc['max_pid']))
+ else:
+ gdb.write(" pid=%d" % loc['min_pid'])
+
+ if constants.LX_NR_CPUS > 1:
+ nr_cpu = gdb.parse_and_eval('__num_online_cpus')['counter']
+ if nr_cpu > 1:
+ gdb.write(" cpus=")
+ for i in loc['cpus']:
+ gdb.write("%d," % i)
+ gdb.write("\n")
+ if constants.LX_CONFIG_STACKDEPOT:
+ if loc['handle']:
+ stackdepot.stack_depot_print(loc['handle'])
+ gdb.write("\n")
+
+def help():
+ t = """Usage: lx-slabtrace --cache_name [cache_name] [Options]
+ Options:
+ --alloc
+ print information of allocation trace of the allocated objects
+ --free
+ print information of freeing trace of the allocated objects
+ Example:
+ lx-slabtrace --cache_name kmalloc-1k --alloc
+ lx-slabtrace --cache_name kmalloc-1k --free\n"""
+ gdb.write("Unrecognized command\n")
+ raise gdb.GdbError(t)
+
+class LxSlabTrace(gdb.Command):
+ """Show specific cache slabtrace"""
+
+ def __init__(self):
+ super(LxSlabTrace, self).__init__("lx-slabtrace", gdb.COMMAND_DATA)
+
+ def invoke(self, arg, from_tty):
+ if not constants.LX_CONFIG_SLUB_DEBUG:
+ raise gdb.GdbError("CONFIG_SLUB_DEBUG is not enabled")
+
+ argv = gdb.string_to_argv(arg)
+ alloc = track_alloc # default show alloc_traces
+
+ if len(argv) == 3:
+ if argv[2] == '--alloc':
+ alloc = track_alloc
+ elif argv[2] == '--free':
+ alloc = track_free
+ else:
+ help()
+ if len(argv) >= 2 and argv[0] == '--cache_name':
+ slabtrace(alloc, argv[1])
+ else:
+ help()
+LxSlabTrace()
+
+def slabinfo():
+ nr_node_ids = None
+
+ if not constants.LX_CONFIG_SLUB_DEBUG:
+ raise gdb.GdbError("CONFIG_SLUB_DEBUG is not enabled")
+
+ def count_free(slab):
+ total_free = 0
+ for slab in lists.list_for_each_entry(slab, slab_ptr_type, 'slab_list'):
+ total_free += int(slab['objects'] - slab['inuse'])
+ return total_free
+
+ gdb.write("{:^18} | {:^20} | {:^12} | {:^12} | {:^8} | {:^11} | {:^13}\n".format('Pointer', 'name', 'active_objs', 'num_objs', 'objsize', 'objperslab', 'pagesperslab'))
+ gdb.write("{:-^18} | {:-^20} | {:-^12} | {:-^12} | {:-^8} | {:-^11} | {:-^13}\n".format('', '', '', '', '', '', ''))
+
+ slab_caches = gdb.parse_and_eval("slab_caches")
+ if mm.page_ops().ops.MAX_NUMNODES > 1:
+ nr_node_ids = int(gdb.parse_and_eval("nr_node_ids"))
+ else:
+ nr_node_ids = 1
+
+ for cache in lists.list_for_each_entry(slab_caches, kmem_cache_ptr_type, 'list'):
+ nr_objs = 0
+ nr_free = 0
+ nr_slabs = 0
+ for i in range(0, nr_node_ids):
+ cache_node = cache['node'][i]
+ try:
+ nr_slabs += cache_node['nr_slabs']['counter']
+ nr_objs = int(cache_node['total_objects']['counter'])
+ nr_free = count_free(cache_node['partial'])
+ except:
+ raise gdb.GdbError(traceback.format_exc())
+ active_objs = nr_objs - nr_free
+ num_objs = nr_objs
+ active_slabs = nr_slabs
+ objects_per_slab = oo_objects(cache['oo'])
+ cache_order = oo_order(cache['oo'])
+ gdb.write("{:18s} | {:20.19s} | {:12} | {:12} | {:8} | {:11} | {:13}\n".format(hex(cache), cache['name'].string(), str(active_objs), str(num_objs), str(cache['size']), str(objects_per_slab), str(1 << cache_order)))
+
+class LxSlabInfo(gdb.Command):
+ """Show slabinfo"""
+
+ def __init__(self):
+ super(LxSlabInfo, self).__init__("lx-slabinfo", gdb.COMMAND_DATA)
+
+ def invoke(self, arg, from_tty):
+ slabinfo()
+LxSlabInfo()
diff --git a/scripts/gdb/vmlinux-gdb.py b/scripts/gdb/vmlinux-gdb.py
index 89465f0de548..2526364f31fd 100644
--- a/scripts/gdb/vmlinux-gdb.py
+++ b/scripts/gdb/vmlinux-gdb.py
@@ -47,3 +47,4 @@ else:
import linux.mm
import linux.stackdepot
import linux.page_owner
+ import linux.slab
--
2.18.0


Subject: [PATCH v2 1/8] scripts/gdb/symbols: add specific ko module load command

Add lx-symbols <ko_path> command to support add specific
ko module.

Example output like below:
(gdb) lx-symbols mm/kasan/kasan_test.ko
loading @0xffff800002d30000: mm/kasan/kasan_test.ko

Signed-off-by: Kuan-Ying Lee <[email protected]>
---
scripts/gdb/linux/symbols.py | 23 +++++++++++++++++++++--
1 file changed, 21 insertions(+), 2 deletions(-)

diff --git a/scripts/gdb/linux/symbols.py b/scripts/gdb/linux/symbols.py
index c8047f4441e6..5179edd1b627 100644
--- a/scripts/gdb/linux/symbols.py
+++ b/scripts/gdb/linux/symbols.py
@@ -111,11 +111,12 @@ lx-symbols command."""
return "{textaddr} {sections}".format(
textaddr=textaddr, sections="".join(args))

- def load_module_symbols(self, module):
+ def load_module_symbols(self, module, module_file=None):
module_name = module['name'].string()
module_addr = str(module['mem'][constants.LX_MOD_TEXT]['base']).split()[0]

- module_file = self._get_module_file(module_name)
+ if not module_file:
+ module_file = self._get_module_file(module_name)
if not module_file and not self.module_files_updated:
self._update_module_files()
module_file = self._get_module_file(module_name)
@@ -138,6 +139,19 @@ lx-symbols command."""
else:
gdb.write("no module object found for '{0}'\n".format(module_name))

+ def load_ko_symbols(self, mod_path):
+ self.loaded_modules = []
+ module_list = modules.module_list()
+
+ for module in module_list:
+ module_name = module['name'].string()
+ module_pattern = ".*/{0}\.ko(?:.debug)?$".format(
+ module_name.replace("_", r"[_\-]"))
+ if re.match(module_pattern, mod_path) and os.path.exists(mod_path):
+ self.load_module_symbols(module, mod_path)
+ return
+ raise gdb.GdbError("%s is not a valid .ko\n" % mod_path)
+
def load_all_symbols(self):
gdb.write("loading vmlinux\n")

@@ -176,6 +190,11 @@ lx-symbols command."""
self.module_files = []
self.module_files_updated = False

+ argv = gdb.string_to_argv(arg)
+ if len(argv) == 1:
+ self.load_ko_symbols(argv[0])
+ return
+
self.load_all_symbols()

if hasattr(gdb, 'Breakpoint'):
--
2.18.0


Subject: [PATCH v2 8/8] scripts/gdb/vmalloc: add vmallocinfo support

This GDB script shows the vmallocinfo for user to
analyze the vmalloc memory usage.

Example output:
0xffff800008000000-0xffff800008009000 36864 <start_kernel+372> pages=8 vmalloc
0xffff800008009000-0xffff80000800b000 8192 <gicv2m_init_one+400> phys=0x8020000 ioremap
0xffff80000800b000-0xffff80000800d000 8192 <bpf_prog_alloc_no_stats+72> pages=1 vmalloc
0xffff80000800d000-0xffff80000800f000 8192 <bpf_jit_alloc_exec+16> pages=1 vmalloc
0xffff800008010000-0xffff80000ad30000 47316992 <paging_init+452> phys=0x40210000 vmap
0xffff80000ad30000-0xffff80000c1c0000 21561344 <paging_init+556> phys=0x42f30000 vmap
0xffff80000c1c0000-0xffff80000c370000 1769472 <paging_init+592> phys=0x443c0000 vmap
0xffff80000c370000-0xffff80000de90000 28442624 <paging_init+692> phys=0x44570000 vmap
0xffff80000de90000-0xffff80000f4c1000 23269376 <paging_init+788> phys=0x46090000 vmap
0xffff80000f4c1000-0xffff80000f4c3000 8192 <gen_pool_add_owner+112> pages=1 vmalloc
0xffff80000f4c3000-0xffff80000f4c5000 8192 <gen_pool_add_owner+112> pages=1 vmalloc
0xffff80000f4c5000-0xffff80000f4c7000 8192 <gen_pool_add_owner+112> pages=1 vmalloc

Signed-off-by: Kuan-Ying Lee <[email protected]>
---
scripts/gdb/linux/constants.py.in | 8 +++++
scripts/gdb/linux/vmalloc.py | 56 +++++++++++++++++++++++++++++++
scripts/gdb/vmlinux-gdb.py | 1 +
3 files changed, 65 insertions(+)
create mode 100644 scripts/gdb/linux/vmalloc.py

diff --git a/scripts/gdb/linux/constants.py.in b/scripts/gdb/linux/constants.py.in
index 03fa6d2cfe01..e3517d4ab8ec 100644
--- a/scripts/gdb/linux/constants.py.in
+++ b/scripts/gdb/linux/constants.py.in
@@ -22,6 +22,7 @@
#include <linux/radix-tree.h>
#include <linux/slab.h>
#include <linux/threads.h>
+#include <linux/vmalloc.h>

/* We need to stringify expanded macros so that they can be parsed */

@@ -91,6 +92,13 @@ LX_GDBPARSED(RADIX_TREE_MAP_SIZE)
LX_GDBPARSED(RADIX_TREE_MAP_SHIFT)
LX_GDBPARSED(RADIX_TREE_MAP_MASK)

+/* linux/vmalloc.h */
+LX_VALUE(VM_IOREMAP)
+LX_VALUE(VM_ALLOC)
+LX_VALUE(VM_MAP)
+LX_VALUE(VM_USERMAP)
+LX_VALUE(VM_DMA_COHERENT)
+
/* linux/page_ext.h */
if IS_BUILTIN(CONFIG_PAGE_OWNER):
LX_GDBPARSED(PAGE_EXT_OWNER)
diff --git a/scripts/gdb/linux/vmalloc.py b/scripts/gdb/linux/vmalloc.py
new file mode 100644
index 000000000000..48e4a4fae7bb
--- /dev/null
+++ b/scripts/gdb/linux/vmalloc.py
@@ -0,0 +1,56 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# Copyright (c) 2023 MediaTek Inc.
+#
+# Authors:
+# Kuan-Ying Lee <[email protected]>
+#
+
+import gdb
+import re
+from linux import lists, utils, stackdepot, constants, mm
+
+vmap_area_type = utils.CachedType('struct vmap_area')
+vmap_area_ptr_type = vmap_area_type.get_type().pointer()
+
+def is_vmalloc_addr(x):
+ pg_ops = mm.page_ops().ops
+ addr = pg_ops.kasan_reset_tag(x)
+ return addr >= pg_ops.VMALLOC_START and addr < pg_ops.VMALLOC_END
+
+class LxVmallocInfo(gdb.Command):
+ """Show vmallocinfo"""
+
+ def __init__(self):
+ super(LxVmallocInfo, self).__init__("lx-vmallocinfo", gdb.COMMAND_DATA)
+
+ def invoke(self, arg, from_tty):
+ vmap_area_list = gdb.parse_and_eval('vmap_area_list')
+ for vmap_area in lists.list_for_each_entry(vmap_area_list, vmap_area_ptr_type, "list"):
+ if not vmap_area['vm']:
+ gdb.write("0x%x-0x%x %10d vm_map_ram\n" % (vmap_area['va_start'], vmap_area['va_end'],
+ vmap_area['va_end'] - vmap_area['va_start']))
+ continue
+ v = vmap_area['vm']
+ gdb.write("0x%x-0x%x %10d" % (v['addr'], v['addr'] + v['size'], v['size']))
+ if v['caller']:
+ gdb.write(" %s" % str(v['caller']).split(' ')[-1])
+ if v['nr_pages']:
+ gdb.write(" pages=%d" % v['nr_pages'])
+ if v['phys_addr']:
+ gdb.write(" phys=0x%x" % v['phys_addr'])
+ if v['flags'] & constants.LX_VM_IOREMAP:
+ gdb.write(" ioremap")
+ if v['flags'] & constants.LX_VM_ALLOC:
+ gdb.write(" vmalloc")
+ if v['flags'] & constants.LX_VM_MAP:
+ gdb.write(" vmap")
+ if v['flags'] & constants.LX_VM_USERMAP:
+ gdb.write(" user")
+ if v['flags'] & constants.LX_VM_DMA_COHERENT:
+ gdb.write(" dma-coherent")
+ if is_vmalloc_addr(v['pages']):
+ gdb.write(" vpages")
+ gdb.write("\n")
+
+LxVmallocInfo()
diff --git a/scripts/gdb/vmlinux-gdb.py b/scripts/gdb/vmlinux-gdb.py
index 2526364f31fd..fc53cdf286f1 100644
--- a/scripts/gdb/vmlinux-gdb.py
+++ b/scripts/gdb/vmlinux-gdb.py
@@ -48,3 +48,4 @@ else:
import linux.stackdepot
import linux.page_owner
import linux.slab
+ import linux.vmalloc
--
2.18.0


2023-09-12 22:28:33

by Johannes Berg

[permalink] [raw]
Subject: Re: [PATCH v2 1/8] scripts/gdb/symbols: add specific ko module load command

On Tue, 2023-08-08 at 16:30 +0800, Kuan-Ying Lee wrote:
> Add lx-symbols <ko_path> command to support add specific
> ko module.

I'm not sure how this was supposed to work? It should have updated the
documentation, but more importantly, it shouldn't have broken the
documented usage of this command:

The kernel (vmlinux) is taken from the current working directly. Modules (.ko)
are scanned recursively, starting in the same directory. Optionally, the module
search path can be extended by a space separated list of paths passed to the
lx-symbols command.

Note how that talks about a "space separated list of paths" for which
clearly a single path seems like it should be accepted?

> @@ -138,6 +139,19 @@ lx-symbols command."""
> else:
> gdb.write("no module object found for '{0}'\n".format(module_name))
>
> + def load_ko_symbols(self, mod_path):
> + self.loaded_modules = []
> + module_list = modules.module_list()
> +
> + for module in module_list:
> + module_name = module['name'].string()
> + module_pattern = ".*/{0}\.ko(?:.debug)?$".format(
> + module_name.replace("_", r"[_\-]"))
> + if re.match(module_pattern, mod_path) and os.path.exists(mod_path):
> + self.load_module_symbols(module, mod_path)
> + return
> + raise gdb.GdbError("%s is not a valid .ko\n" % mod_path)
> +
> def load_all_symbols(self):
> gdb.write("loading vmlinux\n")
>
> @@ -176,6 +190,11 @@ lx-symbols command."""
> self.module_files = []
> self.module_files_updated = False
>
> + argv = gdb.string_to_argv(arg)
> + if len(argv) == 1:
> + self.load_ko_symbols(argv[0])
> + return

But this obviously breaks it, since passing a single path will go into
the if, then complain "some/folder/ is not a valid .ko" and exit.


But I'm not even sure how you intended this to work _at all_, because in
the context before this if, we have:

self.module_paths = [os.path.abspath(os.path.expanduser(p))
for p in arg.split()]
self.module_paths.append(os.getcwd())

so you first add the (file!) to the list of paths, and then try to load
the file by finding modules in the paths, and filtering by the specified
file? That seems ... very roundabout, and can even only work if the file
can be found via os.getcwd(), so you could never specify an absolute
filename?

All that seems counter to what the patch was meant to do.

I suspect that really you need to individually check "is it a file or a
directory" before handling any of this, and if it's actually a file,
don't use it as a _filter_ as you do in load_ko_symbols() now but
directly use it as is with load_module_symbols()?


@Jan, can we revert this? I came up with the following trivial fix that
makes it at least not break the original use case of passing a single-
entry directory list, but it really doesn't seem right one way or the
other.


--- a/scripts/gdb/linux/symbols.py
+++ b/scripts/gdb/linux/symbols.py
@@ -191,7 +191,7 @@ lx-symbols command."""
self.module_files_updated = False

argv = gdb.string_to_argv(arg)
- if len(argv) == 1:
+ if len(argv) == 1 and os.path.isfile(argv[0]):
self.load_ko_symbols(argv[0])
return



johannes

2023-09-13 02:34:22

by Andrew Morton

[permalink] [raw]
Subject: Re: [PATCH v2 1/8] scripts/gdb/symbols: add specific ko module load command

On Tue, 12 Sep 2023 11:41:29 +0200 Johannes Berg <[email protected]> wrote:

> On Tue, 2023-08-08 at 16:30 +0800, Kuan-Ying Lee wrote:
> > Add lx-symbols <ko_path> command to support add specific
> > ko module.
>
> I'm not sure how this was supposed to work? It should have updated the
> documentation, but more importantly, it shouldn't have broken the
> documented usage of this command:
>
> The kernel (vmlinux) is taken from the current working directly. Modules (.ko)
> are scanned recursively, starting in the same directory. Optionally, the module
> search path can be extended by a space separated list of paths passed to the
> lx-symbols command.
>
> Note how that talks about a "space separated list of paths" for which
> clearly a single path seems like it should be accepted?
>
> > @@ -138,6 +139,19 @@ lx-symbols command."""

Thanks, I queued a revert.

From: Andrew Morton <[email protected]>
Subject: revert "scripts/gdb/symbols: add specific ko module load command"
Date: Tue Sep 12 09:19:10 AM PDT 2023

Revert 11f956538c07 ("scripts/gdb/symbols: add specific ko module load
command") due to breakage identified by Johannes Berg in [1].

Fixes: 11f956538c07 ("scripts/gdb/symbols: add specific ko module load command")
Reported-by: Johannes Berg <[email protected]>
Closes: https://lkml.kernel.org/r/[email protected] [1]
Cc: AngeloGioacchino Del Regno <[email protected]>
Cc: Chinwen Chang <[email protected]>
Cc: Jan Kiszka <[email protected]>
Cc: Kieran Bingham <[email protected]>
Cc: Kuan-Ying Lee <[email protected]>
Cc: Matthias Brugger <[email protected]>
Cc: Qun-Wei Lin <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
---

scripts/gdb/linux/symbols.py | 23 ++---------------------
1 file changed, 2 insertions(+), 21 deletions(-)

--- a/scripts/gdb/linux/symbols.py~revert-scripts-gdb-symbols-add-specific-ko-module-load-command
+++ a/scripts/gdb/linux/symbols.py
@@ -111,12 +111,11 @@ lx-symbols command."""
return "{textaddr} {sections}".format(
textaddr=textaddr, sections="".join(args))

- def load_module_symbols(self, module, module_file=None):
+ def load_module_symbols(self, module):
module_name = module['name'].string()
module_addr = str(module['mem'][constants.LX_MOD_TEXT]['base']).split()[0]

- if not module_file:
- module_file = self._get_module_file(module_name)
+ module_file = self._get_module_file(module_name)
if not module_file and not self.module_files_updated:
self._update_module_files()
module_file = self._get_module_file(module_name)
@@ -139,19 +138,6 @@ lx-symbols command."""
else:
gdb.write("no module object found for '{0}'\n".format(module_name))

- def load_ko_symbols(self, mod_path):
- self.loaded_modules = []
- module_list = modules.module_list()
-
- for module in module_list:
- module_name = module['name'].string()
- module_pattern = ".*/{0}\.ko(?:.debug)?$".format(
- module_name.replace("_", r"[_\-]"))
- if re.match(module_pattern, mod_path) and os.path.exists(mod_path):
- self.load_module_symbols(module, mod_path)
- return
- raise gdb.GdbError("%s is not a valid .ko\n" % mod_path)
-
def load_all_symbols(self):
gdb.write("loading vmlinux\n")

@@ -190,11 +176,6 @@ lx-symbols command."""
self.module_files = []
self.module_files_updated = False

- argv = gdb.string_to_argv(arg)
- if len(argv) == 1:
- self.load_ko_symbols(argv[0])
- return
-
self.load_all_symbols()

if hasattr(gdb, 'Breakpoint'):
_