Add support for clobbered blocks

In ext4 filesystems, some blocks might be changed even being mounted
R/O, such as the superblock (block 0). We need to exclude such blocks
from integrity verification. Plus such blocks should always be
written to the target by copying instead of patching.

Bug: http://b/20939131
Change-Id: I657025b7b1ad50d4365e7b18dc39308facfe864e
(cherry picked from commit ff7778166b)
This commit is contained in:
Tao Bao
2015-05-12 11:42:31 -07:00
parent c0ba40b8a1
commit 5ece99d64e
4 changed files with 60 additions and 20 deletions

View File

@@ -21,10 +21,17 @@ import rangelib
class SparseImage(object):
"""Wraps a sparse image file (and optional file map) into an image
object suitable for passing to BlockImageDiff."""
"""Wraps a sparse image file into an image object.
def __init__(self, simg_fn, file_map_fn=None):
Wraps a sparse image file (and optional file map and clobbered_blocks) into
an image object suitable for passing to BlockImageDiff. file_map contains
the mapping between files and their blocks. clobbered_blocks contains the set
of blocks that should be always written to the target regardless of the old
contents (i.e. copying instead of patching). clobbered_blocks should be in
the form of a string like "0" or "0 1-5 8".
"""
def __init__(self, simg_fn, file_map_fn=None, clobbered_blocks=None):
self.simg_f = f = open(simg_fn, "rb")
header_bin = f.read(28)
@@ -57,6 +64,7 @@ class SparseImage(object):
pos = 0 # in blocks
care_data = []
self.offset_map = offset_map = []
self.clobbered_blocks = rangelib.RangeSet(data=clobbered_blocks)
for i in range(total_chunks):
header_bin = f.read(12)
@@ -103,7 +111,7 @@ class SparseImage(object):
self.offset_index = [i[0] for i in offset_map]
if file_map_fn:
self.LoadFileBlockMap(file_map_fn)
self.LoadFileBlockMap(file_map_fn, self.clobbered_blocks)
else:
self.file_map = {"__DATA": self.care_map}
@@ -111,9 +119,10 @@ class SparseImage(object):
return [d for d in self._GetRangeData(ranges)]
def TotalSha1(self):
"""Return the SHA-1 hash of all data in the 'care' regions of this image."""
"""Return the SHA-1 hash of all data in the 'care' regions but not in
clobbered_blocks of this image."""
h = sha1()
for d in self._GetRangeData(self.care_map):
for d in self._GetRangeData(self.care_map.subtract(self.clobbered_blocks)):
h.update(d)
return h.hexdigest()
@@ -156,7 +165,7 @@ class SparseImage(object):
yield fill_data * (this_read * (self.blocksize >> 2))
to_read -= this_read
def LoadFileBlockMap(self, fn):
def LoadFileBlockMap(self, fn, clobbered_blocks):
remaining = self.care_map
self.file_map = out = {}
@@ -166,14 +175,20 @@ class SparseImage(object):
ranges = rangelib.RangeSet.parse(ranges)
out[fn] = ranges
assert ranges.size() == ranges.intersect(remaining).size()
# Currently we assume that blocks in clobbered_blocks are not part of
# any file.
assert not clobbered_blocks.overlaps(ranges)
remaining = remaining.subtract(ranges)
remaining = remaining.subtract(clobbered_blocks)
# For all the remaining blocks in the care_map (ie, those that
# aren't part of the data for any file), divide them into blocks
# that are all zero and blocks that aren't. (Zero blocks are
# handled specially because (1) there are usually a lot of them
# and (2) bsdiff handles files with long sequences of repeated
# bytes especially poorly.)
# aren't part of the data for any file nor part of the clobbered_blocks),
# divide them into blocks that are all zero and blocks that aren't.
# (Zero blocks are handled specially because (1) there are usually
# a lot of them and (2) bsdiff handles files with long sequences of
# repeated bytes especially poorly.)
zero_blocks = []
nonzero_blocks = []
@@ -203,6 +218,7 @@ class SparseImage(object):
out["__ZERO"] = rangelib.RangeSet(data=zero_blocks)
out["__NONZERO"] = rangelib.RangeSet(data=nonzero_blocks)
out["__COPY"] = clobbered_blocks
def ResetFileMap(self):
"""Throw away the file map and treat the entire image as