epoll: support for disabling items, and a self-test app

Enhanced epoll_ctl to support EPOLL_CTL_DISABLE, which disables an epoll
item.  If epoll_ctl doesn't return -EBUSY in this case, it is then safe to
delete the epoll item in a multi-threaded environment.  Also added a new
test_epoll self- test app to both demonstrate the need for this feature
and test it.

Signed-off-by: Paton J. Lewis <palewis@adobe.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Jason Baron <jbaron@redhat.com>
Cc: Paul Holland <pholland@adobe.com>
Cc: Davide Libenzi <davidel@xmailserver.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

5 files changed, 392 insertions, 4 deletions

diff --git a/fs/eventpoll.c b/fs/eventpoll.c
index cd96649bfe62..da72250ddc1c 100644
--- a/fs/eventpoll.c
+++ b/fs/eventpoll.c
@@ -346,7 +346,7 @@ static inline struct epitem *ep_item_from_epqueue(poll_table *p)
 /* Tells if the epoll_ctl(2) operation needs an event copy from userspace */
 static inline int ep_op_has_event(int op)
 {
-        return op != EPOLL_CTL_DEL;
+        return op == EPOLL_CTL_ADD || op == EPOLL_CTL_MOD;
 }
 
 /* Initialize the poll safe wake up structure */
@@ -676,6 +676,34 @@ static int ep_remove(struct eventpoll *ep, struct epitem *epi)
         return 0;
 }
 
+/*
+ * Disables a "struct epitem" in the eventpoll set. Returns -EBUSY if the item
+ * had no event flags set, indicating that another thread may be currently
+ * handling that item's events (in the case that EPOLLONESHOT was being
+ * used). Otherwise a zero result indicates that the item has been disabled
+ * from receiving events. A disabled item may be re-enabled via
+ * EPOLL_CTL_MOD. Must be called with "mtx" held.
+ */
+static int ep_disable(struct eventpoll *ep, struct epitem *epi)
+{
+        int result = 0;
+        unsigned long flags;
+
+        spin_lock_irqsave(&ep->lock, flags);
+        if (epi->event.events & ~EP_PRIVATE_BITS) {
+                if (ep_is_linked(&epi->rdllink))
+                        list_del_init(&epi->rdllink);
+                /* Ensure ep_poll_callback will not add epi back onto ready
+                   list: */
+                epi->event.events &= EP_PRIVATE_BITS;
+                }
+        else
+                result = -EBUSY;
+        spin_unlock_irqrestore(&ep->lock, flags);
+
+        return result;
+}
+
 static void ep_free(struct eventpoll *ep)
 {
         struct rb_node *rbp;
@@ -1020,8 +1048,6 @@ static void ep_rbtree_insert(struct eventpoll *ep, struct epitem *epi)
         rb_insert_color(&epi->rbn, &ep->rbr);
 }
 
-
-
 #define PATH_ARR_SIZE 5
 /*
  * These are the number paths of length 1 to 5, that we are allowing to emanate
@@ -1787,6 +1813,12 @@ SYSCALL_DEFINE4(epoll_ctl, int, epfd, int, op, int, fd,
                 } else
                         error = -ENOENT;
                 break;
+        case EPOLL_CTL_DISABLE:
+                if (epi)
+                        error = ep_disable(ep, epi);
+                else
+                        error = -ENOENT;
+                break;
         }
         mutex_unlock(&ep->mtx);
 
diff --git a/include/linux/eventpoll.h b/include/linux/eventpoll.h
index f4bb378ccf6a..41085d0f3955 100644
--- a/include/linux/eventpoll.h
+++ b/include/linux/eventpoll.h
@@ -25,6 +25,7 @@
 #define EPOLL_CTL_ADD 1
 #define EPOLL_CTL_DEL 2
 #define EPOLL_CTL_MOD 3
+#define EPOLL_CTL_DISABLE 4
 
 /*
  * Request the handling of system wakeup events so as to prevent system suspends
diff --git a/tools/testing/selftests/Makefile b/tools/testing/selftests/Makefile
index 85baf11e2acd..43480149119e 100644
--- a/tools/testing/selftests/Makefile
+++ b/tools/testing/selftests/Makefile
@@ -1,4 +1,4 @@
-TARGETS = breakpoints kcmp mqueue vm cpu-hotplug memory-hotplug
+TARGETS = breakpoints kcmp mqueue vm cpu-hotplug memory-hotplug epoll
 
 all:
         for TARGET in $(TARGETS); do \
diff --git a/tools/testing/selftests/epoll/Makefile b/tools/testing/selftests/epoll/Makefile
new file mode 100644
index 000000000000..19806ed62f50
--- /dev/null
+++ b/tools/testing/selftests/epoll/Makefile
@@ -0,0 +1,11 @@
+# Makefile for epoll selftests
+
+all: test_epoll
+%: %.c
+        gcc -pthread -g -o $@ $^
+
+run_tests: all
+        ./test_epoll
+
+clean:
+        $(RM) test_epoll
diff --git a/tools/testing/selftests/epoll/test_epoll.c b/tools/testing/selftests/epoll/test_epoll.c
new file mode 100644
index 000000000000..e0fcff1e8331
--- /dev/null
+++ b/tools/testing/selftests/epoll/test_epoll.c
@@ -0,0 +1,344 @@
+/*
+ *  tools/testing/selftests/epoll/test_epoll.c
+ *
+ *  Copyright 2012 Adobe Systems Incorporated
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  Paton J. Lewis <palewis@adobe.com>
+ *
+ */
+
+#include <errno.h>
+#include <fcntl.h>
+#include <pthread.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <sys/epoll.h>
+#include <sys/socket.h>
+
+/*
+ * A pointer to an epoll_item_private structure will be stored in the epoll
+ * item's event structure so that we can get access to the epoll_item_private
+ * data after calling epoll_wait:
+ */
+struct epoll_item_private {
+        int index;  /* Position of this struct within the epoll_items array. */
+        int fd;
+        uint32_t events;
+        pthread_mutex_t mutex;  /* Guards the following variables... */
+        int stop;
+        int status;  /* Stores any error encountered while handling item. */
+        /* The following variable allows us to test whether we have encountered
+           a problem while attempting to cancel and delete the associated
+           event. When the test program exits, 'deleted' should be exactly
+           one. If it is greater than one, then the failed test reflects a real
+           world situation where we would have tried to access the epoll item's
+           private data after deleting it: */
+        int deleted;
+};
+
+struct epoll_item_private *epoll_items;
+
+/*
+ * Delete the specified item from the epoll set. In a real-world secneario this
+ * is where we would free the associated data structure, but in this testing
+ * environment we retain the structure so that we can test for double-deletion:
+ */
+void delete_item(int index)
+{
+        __sync_fetch_and_add(&epoll_items[index].deleted, 1);
+}
+
+/*
+ * A pointer to a read_thread_data structure will be passed as the argument to
+ * each read thread:
+ */
+struct read_thread_data {
+        int stop;
+        int status;  /* Indicates any error encountered by the read thread. */
+        int epoll_set;
+};
+
+/*
+ * The function executed by the read threads:
+ */
+void *read_thread_function(void *function_data)
+{
+        struct read_thread_data *thread_data =
+                (struct read_thread_data *)function_data;
+        struct epoll_event event_data;
+        struct epoll_item_private *item_data;
+        char socket_data;