serialize :properties, Hash
before_validation :check_encoding
+ before_validation :log_invalid_manifest_format
before_validation :check_signatures
- before_validation :strip_manifest_text_and_clear_replication_confirmed
- before_validation :set_portable_data_hash
- validate :ensure_hash_matches_manifest_text
+ before_validation :strip_signatures_and_update_replication_confirmed
+ validate :ensure_pdh_matches_manifest_text
before_save :set_file_names
# Query only undeleted collections by default.
t.add :replication_confirmed_at
end
- LOCATOR_REGEXP = /^([[:xdigit:]]{32})(\+([[:digit:]]+))?(\+([[:upper:]][[:alnum:]+@_-]*))?$/
-
def self.attributes_required_columns
super.merge(
# If we don't list manifest_text explicitly, the
)
end
+ FILE_TOKEN = /^[[:digit:]]+:[[:digit:]]+:/
def check_signatures
return false if self.manifest_text.nil?
# subsequent passes without checking any signatures. This is
# important because the signatures have probably been stripped off
# by the time we get to a second validation pass!
- computed_pdh = compute_pdh
return true if @signatures_checked and @signatures_checked == computed_pdh
if self.manifest_text_changed?
api_token: api_token,
now: db_current_time.to_i,
}
- self.manifest_text.lines.each do |entry|
- entry.split[1..-1].each do |tok|
- if /^[[:digit:]]+:[[:digit:]]+:/.match tok
+ self.manifest_text.each_line do |entry|
+ entry.split.each do |tok|
+ if tok == '.' or tok.starts_with? './'
+ # Stream name token.
+ elsif tok =~ FILE_TOKEN
# This is a filename token, not a blob locator. Note that we
# keep checking tokens after this, even though manifest
# format dictates that all subsequent tokens will also be
@signatures_checked = computed_pdh
end
- def strip_manifest_text_and_clear_replication_confirmed
+ def strip_signatures_and_update_replication_confirmed
if self.manifest_text_changed?
in_old_manifest = {}
- self.class.munge_manifest_locators!(manifest_text_was) do |match|
- in_old_manifest[match[1]] = true
+ if not self.replication_confirmed.nil?
+ self.class.each_manifest_locator(manifest_text_was) do |match|
+ in_old_manifest[match[1]] = true
+ end
end
- cleared_replication_confirmed = false
-
- # Remove any permission signatures from the manifest.
- self[:manifest_text] = self.class.munge_manifest_locators!(self[:manifest_text]) do |match|
- if not cleared_replication_confirmed and not in_old_manifest[match[1]]
+ stripped_manifest = self.class.munge_manifest_locators(manifest_text) do |match|
+ if not self.replication_confirmed.nil? and not in_old_manifest[match[1]]
+ # If the new manifest_text contains locators whose hashes
+ # weren't in the old manifest_text, storage replication is no
+ # longer confirmed.
self.replication_confirmed_at = nil
self.replication_confirmed = nil
- cleared_replication_confirmed = true
end
- self.class.locator_without_signature(match)
- end
- end
- true
- end
- def self.locator_without_signature match
- without_signature = match[1]
- without_signature += match[2] if match[2]
- if match[4]
- hints = match[4].split('+').reject { |hint| hint.start_with?("A") }
- without_signature += hints.join('+')
- end
- without_signature
- end
+ # Return the locator with all permission signatures removed,
+ # but otherwise intact.
+ match[0].gsub(/\+A[^+]*/, '')
+ end
- def set_portable_data_hash
- if (portable_data_hash.nil? or
- portable_data_hash == "" or
- (manifest_text_changed? and !portable_data_hash_changed?))
- @need_pdh_validation = false
- self.portable_data_hash = compute_pdh
- elsif portable_data_hash_changed?
- @need_pdh_validation = true
- begin
- loc = Keep::Locator.parse!(self.portable_data_hash)
- loc.strip_hints!
- if loc.size
- self.portable_data_hash = loc.to_s
- else
- self.portable_data_hash = "#{loc.hash}+#{portable_manifest_text.bytesize}"
- end
- rescue ArgumentError => e
- errors.add(:portable_data_hash, "#{e}")
- return false
+ if @computed_pdh_for_manifest_text == manifest_text
+ # If the cached PDH was valid before stripping, it is still
+ # valid after stripping.
+ @computed_pdh_for_manifest_text = stripped_manifest.dup
end
+
+ self[:manifest_text] = stripped_manifest
end
true
end
- def ensure_hash_matches_manifest_text
- return true unless manifest_text_changed? or portable_data_hash_changed?
- # No need verify it if :set_portable_data_hash just computed it!
- return true if not @need_pdh_validation
- expect_pdh = compute_pdh
- if expect_pdh != portable_data_hash
+ def ensure_pdh_matches_manifest_text
+ if not manifest_text_changed? and not portable_data_hash_changed?
+ true
+ elsif portable_data_hash.nil? or not portable_data_hash_changed?
+ self.portable_data_hash = computed_pdh
+ elsif portable_data_hash !~ Keep::Locator::LOCATOR_REGEXP
+ errors.add(:portable_data_hash, "is not a valid locator")
+ false
+ elsif portable_data_hash[0..31] != computed_pdh[0..31]
errors.add(:portable_data_hash,
- "does not match computed hash #{expect_pdh}")
- return false
+ "does not match computed hash #{computed_pdh}")
+ false
+ else
+ # Ignore the client-provided size part: always store
+ # computed_pdh in the database.
+ self.portable_data_hash = computed_pdh
end
end
end
end
+ def log_invalid_manifest_format
+ begin
+ Keep::Manifest.validate! manifest_text if manifest_text
+ rescue => e
+ logger.warn e
+ end
+ true
+ end
+
def signed_manifest_text
if has_attribute? :manifest_text
token = current_api_client_authorization.andand.api_token
api_token: token,
expire: db_current_time.to_i + Rails.configuration.blob_signature_ttl,
}
- m = manifest.dup
- m = munge_manifest_locators!(m) do |match|
- Blob.sign_locator(locator_without_signature(match), signing_opts)
+ m = munge_manifest_locators(manifest) do |match|
+ Blob.sign_locator(match[0], signing_opts)
end
return m
end
- def self.munge_manifest_locators! manifest
- # Given a manifest text and a block, yield each locator,
- # and replace it with whatever the block returns.
+ def self.munge_manifest_locators manifest
+ # Given a manifest text and a block, yield the regexp MatchData
+ # for each locator. Return a new manifest in which each locator
+ # has been replaced by the block's return value.
+ return nil if !manifest
+ return '' if manifest == ''
+
new_lines = []
- lines = manifest.andand.split("\n")
- lines.andand.each do |line|
- words = line.split(' ')
+ manifest.each_line do |line|
+ line.rstrip!
new_words = []
- words.each do |word|
- if match = LOCATOR_REGEXP.match(word.strip)
+ line.split(' ').each do |word|
+ if new_words.empty?
+ new_words << word
+ elsif match = Keep::Locator::LOCATOR_REGEXP.match(word)
new_words << yield(match)
else
- new_words << word.strip
+ new_words << word
end
end
new_lines << new_words.join(' ')
end
+ new_lines.join("\n") + "\n"
+ end
- if !new_lines.empty?
- ends_with_newline = manifest.end_with?("\n")
- manifest = new_lines.join("\n")
- manifest += "\n" if ends_with_newline
+ def self.each_manifest_locator manifest
+ # Given a manifest text and a block, yield the regexp match object
+ # for each locator.
+ manifest.each_line do |line|
+ # line will have a trailing newline, but the last token is never
+ # a locator, so it's harmless here.
+ line.split(' ').each do |word|
+ if match = Keep::Locator::LOCATOR_REGEXP.match(word)
+ yield(match)
+ end
+ end
end
-
- manifest
end
def self.normalize_uuid uuid
protected
def portable_manifest_text
- portable_manifest = self[:manifest_text].dup
- portable_manifest = self.class.munge_manifest_locators!(portable_manifest) do |match|
+ self.class.munge_manifest_locators(manifest_text) do |match|
if match[2] # size
match[1] + match[2]
else
match[1]
end
end
- portable_manifest
end
def compute_pdh
- return @computed_pdh if @computed_pdh
portable_manifest = portable_manifest_text
- @computed_pdh = (Digest::MD5.hexdigest(portable_manifest) +
- '+' +
- portable_manifest.bytesize.to_s)
+ (Digest::MD5.hexdigest(portable_manifest) +
+ '+' +
+ portable_manifest.bytesize.to_s)
+ end
+
+ def computed_pdh
+ if @computed_pdh_for_manifest_text == manifest_text
+ return @computed_pdh
+ end
+ @computed_pdh = compute_pdh
+ @computed_pdh_for_manifest_text = manifest_text.dup
@computed_pdh
end