K 10
svn:author
V 5
gibbs
K 8
svn:date
V 27
2012-05-30T00:27:12.382405Z
K 7
svn:log
V 1809
Properly terminate truncated events with a newline (end of event
token), and discard any data that exceeds zfsd's MAX_EVENT_SIZE
limit until zfsd is resynchronized with the event stream.  This
resolves issues (de)serializing CaseFiles with truncated events.

cddl/sbin/zfsd/case_file.cc:
	Do not assume that deserialization of all events from a
	case file will be successful.  Just skip over unparsable
	events.

cddl/sbin/zfsd/zfsd.cc:
cddl/sbin/zfsd/zfsd.h:
	o In EventBuffer::ExtractEvent(), remove code and a comment
	  dealing with "start tokens" and "end tokens" sharing
	  characters.  In the normal case of untruncated events,
	  the event terminator is present and distinct from the
	  start token.  Devd immediately closes our unix domain
	  socket if it cannot write a complete event, which means
	  a start token for a new event after a truncted event
	  cannot happen.

	o In EventBuffer::ExtractEvent(), remove dead code that
	  tested for end of event tokens other than '\n'.  Due
	  to the way that strcspn() operates, this cannot happen.

	o Add stateful resynchronization to the EventBuffer
	  class.  EventBuffer can get out of synch whenever it
	  truncates an incoming event due to its internal event
	  size limit.  The original code dealt with this issue
	  by looking for start tokens.  Since start tokens may
	  also be valid within an event body, the only fully
	  safe option is to discard data until the end of event
	  token (which can't occur within an event) is read.

	o When truncating an event, properly terminate it so
	  that when EventBuffer is used for CaseFile deserialization
	  the input data is properly formed.

	o EventBuffer::UnParsed() and EventBuffer::NextEventMaxLen()
	  are const.  Enforce this via the const keyword.

Sponsored by:	Spectra Logic Corporation

END