Skip to content

Commit

Permalink
Merge branch 'feature/nvs_tool_integrity_check_refactor' into 'master'
Browse files Browse the repository at this point in the history
feat(nvs): NVS tool add key consistency check

Closes IDF-8146

See merge request espressif/esp-idf!25638
  • Loading branch information
rrtandler committed Oct 11, 2023
2 parents b73e462 + d42bde0 commit 6c83e63
Show file tree
Hide file tree
Showing 3 changed files with 259 additions and 211 deletions.
257 changes: 257 additions & 0 deletions components/nvs_flash/nvs_partition_tool/nvs_check.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,257 @@
#!/usr/bin/env python3
# SPDX-FileCopyrightText: 2023 Espressif Systems (Shanghai) CO LTD
# SPDX-License-Identifier: Apache-2.0
from typing import Dict, List

from nvs_logger import NVS_Logger
from nvs_parser import NVS_Entry, NVS_Partition, nvs_const


def integrity_check(nvs_partition: NVS_Partition, nvs_log: NVS_Logger) -> None:
used_namespaces: Dict[int, None] = {}
found_namespaces: Dict[int, str] = {}
blobs: Dict = {}
blob_chunks: List[NVS_Entry] = []
empty_entry = NVS_Entry(-1, bytearray(32), 'Erased')

# Partition size check
if len(nvs_partition.pages) < 3:
nvs_log.info(
nvs_log.yellow(
'NVS Partition must contain 3 pages (sectors) at least to function properly!'
)
)

# Free/empty page check
if not any(page.header['status'] == 'Empty' for page in nvs_partition.pages):
nvs_log.info(
nvs_log.red(
'''No free (empty) page found in the NVS partition,
at least one free page is required for proper function!'''
)
)
nvs_log.info(nvs_log.red('NVS partition possibly truncated?\n'))

for page in nvs_partition.pages:
# page: NVS_Page

# Print page header
if page.header['status'] == 'Empty':
nvs_log.info(nvs_log.cyan('Page Empty'))

# Check if page is truly empty
if page.raw_entry_state_bitmap != bytearray({0xFF}) * nvs_const.entry_size:
nvs_log.info(
nvs_log.red(
'The page is reported as Empty but its entry state bitmap is not empty!'
)
)
if any([not e.is_empty for e in page.entries]):
nvs_log.info(
nvs_log.red('The page is reported as Empty but there are data written!')
)
else:
# Check page header CRC32
if page.header['crc']['original'] == page.header['crc']['computed']:
nvs_log.info(
nvs_log.cyan(f'Page no. {page.header["page_index"]}'), '\tCRC32: OK'
)
else:
nvs_log.info(
nvs_log.cyan(f'Page no. {page.header["page_index"]}'),
f'Original CRC32:',
nvs_log.red(f'{page.header["crc"]["original"]:x}'),
f'Generated CRC32:',
nvs_log.green(f'{page.header["crc"]["computed"]:x}'),
)

# Check all entries
seen_written_entires: Dict[str, list[NVS_Entry]] = {}
for entry in page.entries:
# entry: NVS_Entry

# Entries stored in 'page.entries' are primitive data types, blob indexes or string/blob data

# Variable length values themselves occupy whole 32 bytes (therefore metadata values are meaningless)
# and are stored in as entries inside string/blob data entry 'entry.children' list

# Duplicate entry check (1) - same key, different index - find duplicates
if entry.state == 'Written':
if entry.key in seen_written_entires:
seen_written_entires[entry.key].append(entry)
else:
seen_written_entires[entry.key] = [entry]

# Entry state check - doesn't check variable length values (metadata such as state are meaningless as all 32 bytes are pure data)
if entry.is_empty:
if entry.state == 'Written':
nvs_log.info(
nvs_log.red(
f' Entry #{entry.index:03d} is reported as Written but it is empty!'
)
)
continue
elif entry.state == 'Erased':
nvs_log.info(
nvs_log.yellow(
f' Entry #{entry.index:03d} is reported as Erased but it is empty! (Only entries reported as Empty should be empty)'
)
)

if entry.state == 'Written':
# Entry CRC32 check
if (
entry.metadata['crc']['original']
!= entry.metadata['crc']['computed']
):
nvs_log.info(
nvs_log.red(
f' Entry #{entry.index:03d} {entry.key} has wrong CRC32!{"": <5}'
),
f'Written:',
nvs_log.red(f'{entry.metadata["crc"]["original"]:x}'),
f'Generated:',
nvs_log.green(f'{entry.metadata["crc"]["computed"]:x}'),
)

# Entry children CRC32 check
if (
entry.metadata['span'] > 1
and (entry.metadata['crc']['data_original'] != entry.metadata['crc']['data_computed'])
):
nvs_log.info(
nvs_log.red(
f' Entry #{entry.index:03d} {entry.key} data (string, blob) has wrong CRC32!'
),
f'Written:',
nvs_log.red(f'{entry.metadata["crc"]["data_original"]:x}'),
f'Generated:',
nvs_log.green(f'{entry.metadata["crc"]["data_computed"]:x}'),
)

# Entry type check
if entry.metadata['type'] not in [
nvs_const.item_type[key] for key in nvs_const.item_type
]:
nvs_log.info(
nvs_log.yellow(
f' Type of entry #{entry.index:03d} {entry.key} is unrecognized!'
),
f'Type: {entry.metadata["type"]}',
)

# Span check
if (
entry.index + entry.metadata['span'] - 1
>= int(nvs_const.page_size / nvs_const.entry_size) - 2
):
nvs_log.info(
nvs_log.red(
f' Variable length entry #{entry.index:03d} {entry.key} is out of bounds!'
)
)
# Spanned entry state checks
elif entry.metadata['span'] > 1:
parent_state = entry.state
for kid in entry.children:
if parent_state != kid.state:
nvs_log.info(
nvs_log.yellow(' Inconsistent data state!'),
f'Entry #{entry.index:03d} {entry.key} state: {parent_state},',
f'Data entry #{kid.index:03d} {entry.key} state: {kid.state}',
)

# Gather blobs & namespaces
if entry.metadata['type'] == 'blob_index':
blobs[f'{entry.metadata["namespace"]:03d}{entry.key}'] = [entry] + [
empty_entry
] * entry.data['chunk_count']
elif entry.metadata['type'] == 'blob_data':
blob_chunks.append(entry)

if entry.metadata['namespace'] == 0:
found_namespaces[entry.data['value']] = entry.key
else:
used_namespaces[entry.metadata['namespace']] = None

# Duplicate entry check (2) - same key, different index - print duplicates
duplicate_entries_list = [seen_written_entires[key] for key in seen_written_entires if len(seen_written_entires[key]) > 1]
for duplicate_entries in duplicate_entries_list:
# duplicate_entries: list[NVS_Entry]
nvs_log.info(
nvs_log.red(
f'''Entry key {duplicate_entries[0].key} on page no. {page.header["page_index"]}
with status {page.header["status"]} is used by the following entries:'''
)
)
for entry in duplicate_entries:
nvs_log.info(
nvs_log.red(
f'Entry #{entry.index:03d} {entry.key} is a duplicate!'
)
)

nvs_log.info()

# Blob checks
# Assemble blobs
for chunk in blob_chunks:
parent = blobs.get(
f'{chunk.metadata["namespace"]:03d}{chunk.key}', [empty_entry]
)[0]
# Blob chunk without blob index check
if parent is empty_entry:
nvs_log.info(
nvs_log.red(f'Blob {chunk.key} chunk has no blob index!'),
f'Namespace index: {chunk.metadata["namespace"]:03d}',
f'[{found_namespaces.get(chunk.metadata["namespace"], "undefined")}],',
f'Chunk Index: {chunk.metadata["chunk_index"]:03d}',
)
else:
blob_key = f'{chunk.metadata["namespace"]:03d}{chunk.key}'
chunk_index = chunk.metadata['chunk_index'] - parent.data['chunk_start']
blobs[blob_key][chunk_index + 1] = chunk

# Blob data check
for blob_key in blobs:
blob_index = blobs[blob_key][0]
blob_chunks = blobs[blob_key][1:]
blob_size = blob_index.data['size']

for i, chunk in enumerate(blob_chunks):
# Blob missing chunk check
if chunk is empty_entry:
nvs_log.info(
nvs_log.red(f'Blob {blob_index.key} is missing a chunk!'),
f'Namespace index: {blob_index.metadata["namespace"]:03d}',
f'[{found_namespaces.get(blob_index.metadata["namespace"], "undefined")}],',
f'Chunk Index: {i:03d}',
)
else:
blob_size -= len(chunk.children) * nvs_const.entry_size

# Blob missing data check
if blob_size > 0:
nvs_log.info(
nvs_log.red(f'Blob {blob_index.key} is missing {blob_size} B of data!'),
f'Namespace index: {blob_index.metadata["namespace"]:03d}',
)

# Namespace checks
# Undefined namespace index check
for used_ns in used_namespaces:
key = found_namespaces.pop(used_ns, '')
if key == '':
nvs_log.info(
nvs_log.red('Undefined namespace index!'),
f'Namespace index: {used_ns:03d}',
f'[undefined]',
)

# Unused namespace index check
for unused_ns in found_namespaces:
nvs_log.info(
nvs_log.yellow('Found unused namespace.'),
f'Namespace index: {unused_ns:03d}',
f'[{found_namespaces[unused_ns]}]',
)
Loading

0 comments on commit 6c83e63

Please sign in to comment.