[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-changelog] [xen stable-4.5] libxl: poll: Avoid fd deregistration race POLLNVAL crash



commit 7f7642f778b78e8e204fc082ce03072bb26887c7
Author:     Ian Jackson <ian.jackson@xxxxxxxxxxxxx>
AuthorDate: Thu Jul 9 17:28:48 2015 +0100
Commit:     Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>
CommitDate: Wed Aug 12 12:44:30 2015 +0100

    libxl: poll: Avoid fd deregistration race POLLNVAL crash
    
    It can happen that an fd is deregistered, and closed, and then a new
    fd opened, and reregistered, all while another thread is in poll().
    
    If this happens poll might report POLLNVAL, but the event loop would
    think that the fd was supposed to have been valid, and then fail an
    assertion:
      libxl_event.c:1183: afterpoll_check_fd: Assertion `poller->fds_changed || 
!(fds[slot].revents & 0x020)' failed.
    
    We can't simply ignore POLLNVAL because if we have bugs which cause
    messed-up fds, it is a serious problem which we really need to detect.
    
    Instead, add extra tracking to spot when this possibility arises, and
    abort on POLLNVAL if we are sure that it is unexpected.
    
    Reported-by: Jim Fehlig <jfehlig@xxxxxxxx>
    Signed-off-by: Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>
    CC: Jim Fehlig <jfehlig@xxxxxxxx>
    Acked-by: Wei Liu <wei.liu2@xxxxxxxxxx>
    Tested-by: Jim Fehlig <jfehlig@xxxxxxxx>
    (cherry picked from commit 681ce1681622a46d111cfdc4fc07e4cb565ae131)
---
 tools/libxl/libxl.c          |    2 ++
 tools/libxl/libxl_event.c    |   12 +++++++++++-
 tools/libxl/libxl_internal.h |   13 +++++++++++++
 3 files changed, 26 insertions(+), 1 deletions(-)

diff --git a/tools/libxl/libxl.c b/tools/libxl/libxl.c
index 939e077..3536b5d 100644
--- a/tools/libxl/libxl.c
+++ b/tools/libxl/libxl.c
@@ -62,6 +62,7 @@ int libxl_ctx_alloc(libxl_ctx **pctx, int version,
     ctx->poller_app = 0;
     LIBXL_LIST_INIT(&ctx->pollers_event);
     LIBXL_LIST_INIT(&ctx->pollers_idle);
+    LIBXL_LIST_INIT(&ctx->pollers_fds_changed);
 
     LIBXL_LIST_INIT(&ctx->efds);
     LIBXL_TAILQ_INIT(&ctx->etimes);
@@ -185,6 +186,7 @@ int libxl_ctx_free(libxl_ctx *ctx)
     libxl__poller_put(ctx, ctx->poller_app);
     ctx->poller_app = NULL;
     assert(LIBXL_LIST_EMPTY(&ctx->pollers_event));
+    assert(LIBXL_LIST_EMPTY(&ctx->pollers_fds_changed));
     libxl__poller *poller, *poller_tmp;
     LIBXL_LIST_FOREACH_SAFE(poller, &ctx->pollers_idle, entry, poller_tmp) {
         libxl__poller_dispose(poller);
diff --git a/tools/libxl/libxl_event.c b/tools/libxl/libxl_event.c
index 06ad284..597f65d 100644
--- a/tools/libxl/libxl_event.c
+++ b/tools/libxl/libxl_event.c
@@ -220,6 +220,7 @@ int libxl__ev_fd_modify(libxl__gc *gc, libxl__ev_fd *ev, 
short events)
 void libxl__ev_fd_deregister(libxl__gc *gc, libxl__ev_fd *ev)
 {
     CTX_LOCK;
+    libxl__poller *poller;
 
     if (!libxl__ev_fd_isregistered(ev)) {
         DBG("ev_fd=%p deregister unregistered",ev);
@@ -232,6 +233,9 @@ void libxl__ev_fd_deregister(libxl__gc *gc, libxl__ev_fd 
*ev)
     LIBXL_LIST_REMOVE(ev, entry);
     ev->fd = -1;
 
+    LIBXL_LIST_FOREACH(poller, &CTX->pollers_fds_changed, fds_changed_entry)
+        poller->fds_changed = 1;
+
  out:
     CTX_UNLOCK;
 }
@@ -1031,6 +1035,8 @@ static int beforepoll_internal(libxl__gc *gc, 
libxl__poller *poller,
 
     *nfds_io = used;
 
+    poller->fds_changed = 0;
+
     libxl__ev_time *etime = LIBXL_TAILQ_FIRST(&CTX->etimes);
     if (etime) {
         int our_timeout;
@@ -1095,7 +1101,7 @@ static int afterpoll_check_fd(libxl__poller *poller,
             /* again, stale slot entry */
             continue;
 
-        assert(!(fds[slot].revents & POLLNVAL));
+        assert(poller->fds_changed || !(fds[slot].revents & POLLNVAL));
 
         /* we mask in case requested events have changed */
         int slot_revents = fds[slot].revents & events;
@@ -1531,6 +1537,7 @@ int libxl__poller_init(libxl__gc *gc, libxl__poller *p)
     int rc;
     p->fd_polls = 0;
     p->fd_rindices = 0;
+    p->fds_changed = 0;
 
     rc = libxl__pipe_nonblock(CTX, p->wakeup_pipe);
     if (rc) goto out;
@@ -1567,12 +1574,15 @@ libxl__poller *libxl__poller_get(libxl__gc *gc)
         }
     }
 
+    LIBXL_LIST_INSERT_HEAD(&CTX->pollers_fds_changed, p,
+                           fds_changed_entry);
     return p;
 }
 
 void libxl__poller_put(libxl_ctx *ctx, libxl__poller *p)
 {
     if (!p) return;
+    LIBXL_LIST_REMOVE(p, fds_changed_entry);
     LIBXL_LIST_INSERT_HEAD(&ctx->pollers_idle, p, entry);
 }
 
diff --git a/tools/libxl/libxl_internal.h b/tools/libxl/libxl_internal.h
index 4120feb..bb2ebc0 100644
--- a/tools/libxl/libxl_internal.h
+++ b/tools/libxl/libxl_internal.h
@@ -304,6 +304,18 @@ struct libxl__poller {
     int (*fd_rindices)[3]; /* see libxl_event.c:beforepoll_internal */
 
     int wakeup_pipe[2]; /* 0 means no fd allocated */
+
+    /*
+     * We also use the poller to record whether any fds have been
+     * deregistered since we entered poll.  Each poller which is not
+     * idle is on the list pollers_fds_changed.  fds_changed is
+     * cleared by beforepoll, and tested by afterpoll.  Whenever an fd
+     * event is deregistered, we set the fds_changed of all non-idle
+     * pollers.  So afterpoll can tell whether any POLLNVAL is
+     * plausibly due to an fd being closed and reopened.
+     */
+    LIBXL_LIST_ENTRY(libxl__poller) fds_changed_entry;
+    bool fds_changed;
 };
 
 struct libxl__gc {
@@ -345,6 +357,7 @@ struct libxl__ctx {
 
     libxl__poller *poller_app; /* libxl_osevent_beforepoll and _afterpoll */
     LIBXL_LIST_HEAD(, libxl__poller) pollers_event, pollers_idle;
+    LIBXL_LIST_HEAD(, libxl__poller) pollers_fds_changed;
 
     LIBXL_SLIST_HEAD(libxl__osevent_hook_nexi, libxl__osevent_hook_nexus)
         hook_fd_nexi_idle, hook_timeout_nexi_idle;
--
generated by git-patchbot for /home/xen/git/xen.git#stable-4.5

_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.