From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:44910) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ad0nC-0004Fk-EX for qemu-devel@nongnu.org; Mon, 07 Mar 2016 14:26:11 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1ad0n7-0003yq-96 for qemu-devel@nongnu.org; Mon, 07 Mar 2016 14:26:10 -0500 Received: from mx1.redhat.com ([209.132.183.28]:60406) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1ad0n7-0003yP-2q for qemu-devel@nongnu.org; Mon, 07 Mar 2016 14:26:05 -0500 From: Markus Armbruster Date: Mon, 7 Mar 2016 20:25:38 +0100 Message-Id: <1457378754-21649-27-git-send-email-armbru@redhat.com> In-Reply-To: <1457378754-21649-1-git-send-email-armbru@redhat.com> References: <1457378754-21649-1-git-send-email-armbru@redhat.com> Subject: [Qemu-devel] [PATCH v2 26/42] ivshmem: Plug leaks on unplug, fix peer disconnect List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: claudio.fontana@huawei.com, cam@cs.ualberta.ca, mlureau@redhat.com, david.marchand@6wind.com, pbonzini@redhat.com close_peer_eventfds() cleans up three things: ioeventfd triggers if they exist, eventfds, and the array to store them. Commit 98609cd (v1.2.0) fixed it not to clean up ioeventfd triggers when they don't exist (property ioeventfd=off, which is the default). Unfortunately, the fix also made it skip cleanup of the eventfds and the array then. This is a memory and file descriptor leak on unplug. Additionally, the reset of nb_eventfds is skipped. Doesn't matter on unplug. On peer disconnect, however, this permanently wedges the interrupt vectors used for that peer's ID. The eventfds stay behind, but aren't connected to a peer anymore. When the ID gets recycled for a new peer, the new peer's eventfds get assigned to vectors after the old ones. Commonly, the device's number of vectors matches the server's, so the new ones get dropped with a "Too many eventfd received" message. Interrupts either don't work (common case) or go to the wrong vector. Fix by narrowing the conditional to just the ioeventfd trigger cleanup. While there, move the "invalid" peer check to the only caller where it can actually happen, and tighten it to reject own ID. Cc: Paolo Bonzini Signed-off-by: Markus Armbruster --- hw/misc/ivshmem.c | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) diff --git a/hw/misc/ivshmem.c b/hw/misc/ivshmem.c index e568263..2f2f43f 100644 --- a/hw/misc/ivshmem.c +++ b/hw/misc/ivshmem.c @@ -428,21 +428,17 @@ static void close_peer_eventfds(IVShmemState *s, int posn) { int i, n; - if (!ivshmem_has_feature(s, IVSHMEM_IOEVENTFD)) { - return; - } - if (posn < 0 || posn >= s->nb_peers) { - error_report("invalid peer %d", posn); - return; - } - + assert(posn >= 0 && posn < s->nb_peers); n = s->peers[posn].nb_eventfds; - memory_region_transaction_begin(); - for (i = 0; i < n; i++) { - ivshmem_del_eventfd(s, posn, i); + if (ivshmem_has_feature(s, IVSHMEM_IOEVENTFD)) { + memory_region_transaction_begin(); + for (i = 0; i < n; i++) { + ivshmem_del_eventfd(s, posn, i); + } + memory_region_transaction_commit(); } - memory_region_transaction_commit(); + for (i = 0; i < n; i++) { event_notifier_cleanup(&s->peers[posn].eventfds[i]); } @@ -598,6 +594,10 @@ static void process_msg_shmem(IVShmemState *s, int fd) static void process_msg_disconnect(IVShmemState *s, uint16_t posn) { IVSHMEM_DPRINTF("posn %d has gone away\n", posn); + if (posn >= s->nb_peers || posn == s->vm_id) { + error_report("invalid peer %d", posn); + return; + } close_peer_eventfds(s, posn); } -- 2.4.3