stress: Remove essentially unused library.

The "stress" library was introduced years ago.  We intended at the time to
start using it to provoke errors in testing, to make sure that Open vSwitch
was resilient against those errors.  The intention was good, but there were
few actual implementations of stress options, and the testing never
materialized.

Rather than adapt the stress library for thread safety, this seems like a
good opportunity to remove it, so this commit does so.

Signed-off-by: Ben Pfaff <[email protected]>

lib/automake.mk

@@ -182,8 +182,6 @@ lib_libopenvswitch_a_SOURCES = \
 	lib/stream-unix.c \
 	lib/stream.c \
 	lib/stream.h \
-	lib/stress.c \
-	lib/stress.h \
 	lib/string.c \
 	lib/string.h \
 	lib/svec.c \
@@ -306,7 +304,6 @@ MAN_FRAGMENTS += \
 	lib/ssl-peer-ca-cert.man \
 	lib/ssl.man \
 	lib/ssl-syn.man \
-	lib/stress-unixctl.man \
 	lib/table.man \
 	lib/unixctl.man \
 	lib/unixctl-syn.man \
@@ -383,11 +380,6 @@ lib/coverage.def: $(DIST_SOURCES)
 	sed -n 's|^COVERAGE_DEFINE(\([_a-zA-Z0-9]\{1,\}\)).*$$|COVERAGE_COUNTER(\1)|p' $(all_sources) | LC_ALL=C sort -u > $@
 CLEANFILES += lib/coverage.def
 
-lib/stress.$(OBJEXT): lib/stress.def
-lib/stress.def: $(DIST_SOURCES)
-	sed -n '/^STRESS_OPTION(/,/);$$/{s/);$$/)/;p}' $(all_sources) > $@
-CLEANFILES += lib/stress.def
-
 lib/vlog.$(OBJEXT): lib/vlog-modules.def
 lib/vlog-modules.def: $(DIST_SOURCES)
 	sed -n 's|^VLOG_DEFINE_\(THIS_\)\{0,1\}MODULE(\([_a-zA-Z0-9]\{1,\}\)).*$$|VLOG_MODULE(\2)|p' $(all_sources) | LC_ALL=C sort -u > $@

lib/netlink-socket.c

@@ -31,7 +31,6 @@
 #include "ofpbuf.h"
 #include "poll-loop.h"
 #include "socket-util.h"
-#include "stress.h"
 #include "util.h"
 #include "vlog.h"
 
@@ -309,15 +308,6 @@ nl_sock_send_seq(struct nl_sock *sock, const struct ofpbuf *msg,
     return nl_sock_send__(sock, msg, nlmsg_seq, wait);
 }
 
-/* This stress option is useful for testing that OVS properly tolerates
- * -ENOBUFS on NetLink sockets.  Such errors are unavoidable because they can
- * occur if the kernel cannot temporarily allocate enough GFP_ATOMIC memory to
- * reply to a request.  They can also occur if messages arrive on a multicast
- * channel faster than OVS can process them. */
-STRESS_OPTION(
-    netlink_overflow, "simulate netlink socket receive buffer overflow",
-    5, 1, -1, 100);
-
 static int
 nl_sock_recv__(struct nl_sock *sock, struct ofpbuf *buf, bool wait)
 {
@@ -373,10 +363,6 @@ nl_sock_recv__(struct nl_sock *sock, struct ofpbuf *buf, bool wait)
         return EPROTO;
     }
 
-    if (STRESS(netlink_overflow)) {
-        return ENOBUFS;
-    }
-
     buf->size = MIN(retval, buf->allocated);
     if (retval > buf->allocated) {
         COVERAGE_INC(netlink_recv_jumbo);

lib/stream-fd.c

@@ -26,7 +26,6 @@
 #include "fatal-signal.h"
 #include "poll-loop.h"
 #include "socket-util.h"
-#include "stress.h"
 #include "util.h"
 #include "stream-provider.h"
 #include "stream.h"
@@ -89,38 +88,22 @@ fd_connect(struct stream *stream)
     return check_connection_completion(s->fd);
 }
 
-STRESS_OPTION(
-    stream_flaky_recv, "simulate failure of fd stream recvs",
-    100, 0, -1, 0);
-
 static ssize_t
 fd_recv(struct stream *stream, void *buffer, size_t n)
 {
     struct stream_fd *s = stream_fd_cast(stream);
     ssize_t retval;
 
-    if (STRESS(stream_flaky_recv)) {
-        return -EIO;
-    }
-
     retval = read(s->fd, buffer, n);
     return retval >= 0 ? retval : -errno;
 }
 
-STRESS_OPTION(
-    stream_flaky_send, "simulate failure of fd stream sends",
-    100, 0, -1, 0);
-
 static ssize_t
 fd_send(struct stream *stream, const void *buffer, size_t n)
 {
     struct stream_fd *s = stream_fd_cast(stream);
     ssize_t retval;
 
-    if (STRESS(stream_flaky_send)) {
-        return -EIO;
-    }
-
     retval = write(s->fd, buffer, n);
     return (retval > 0 ? retval
             : retval == 0 ? -EAGAIN