From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:58855) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1fG7Xh-0004CC-M9 for qemu-devel@nongnu.org; Tue, 08 May 2018 14:40:56 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1fG7Xd-00074v-PZ for qemu-devel@nongnu.org; Tue, 08 May 2018 14:40:53 -0400 Received: from mx3-rdu2.redhat.com ([66.187.233.73]:53572 helo=mx1.redhat.com) by eggs.gnu.org with esmtps (TLS1.0:DHE_RSA_AES_256_CBC_SHA1:32) (Exim 4.71) (envelope-from ) id 1fG7Xd-00074A-Jo for qemu-devel@nongnu.org; Tue, 08 May 2018 14:40:49 -0400 Date: Tue, 8 May 2018 19:40:36 +0100 From: "Dr. David Alan Gilbert" Message-ID: <20180508184035.GT2500@work-vm> References: <1525618499-1560-1-git-send-email-lidongchen@tencent.com> <1525618499-1560-2-git-send-email-lidongchen@tencent.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1525618499-1560-2-git-send-email-lidongchen@tencent.com> Subject: Re: [Qemu-devel] [PATCH 2/2] migration: not wait RDMA_CM_EVENT_DISCONNECTED event after rdma_disconnect List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Lidong Chen Cc: quintela@redhat.com, qemu-devel@nongnu.org, galsha@mellanox.com, aviadye@mellanox.com, adido@mellanox.com, Lidong Chen * Lidong Chen (jemmy858585@gmail.com) wrote: > When cancel migration during RDMA precopy, the source qemu main thread hangs sometime. > > The backtrace is: > (gdb) bt > #0 0x00007f249eabd43d in write () from /lib64/libpthread.so.0 > #1 0x00007f24a1ce98e4 in rdma_get_cm_event (channel=0x4675d10, event=0x7ffe2f643dd0) at src/cma.c:2189 > #2 0x00000000007b6166 in qemu_rdma_cleanup (rdma=0x6784000) at migration/rdma.c:2296 > #3 0x00000000007b7cae in qio_channel_rdma_close (ioc=0x3bfcc30, errp=0x0) at migration/rdma.c:2999 > #4 0x00000000008db60e in qio_channel_close (ioc=0x3bfcc30, errp=0x0) at io/channel.c:273 > #5 0x00000000007a8765 in channel_close (opaque=0x3bfcc30) at migration/qemu-file-channel.c:98 > #6 0x00000000007a71f9 in qemu_fclose (f=0x527c000) at migration/qemu-file.c:334 > #7 0x0000000000795b96 in migrate_fd_cleanup (opaque=0x3b46280) at migration/migration.c:1162 > #8 0x000000000093a71b in aio_bh_call (bh=0x3db7a20) at util/async.c:90 > #9 0x000000000093a7b2 in aio_bh_poll (ctx=0x3b121c0) at util/async.c:118 > #10 0x000000000093f2ad in aio_dispatch (ctx=0x3b121c0) at util/aio-posix.c:436 > #11 0x000000000093ab41 in aio_ctx_dispatch (source=0x3b121c0, callback=0x0, user_data=0x0) > at util/async.c:261 > #12 0x00007f249f73c7aa in g_main_context_dispatch () from /lib64/libglib-2.0.so.0 > #13 0x000000000093dc5e in glib_pollfds_poll () at util/main-loop.c:215 > #14 0x000000000093dd4e in os_host_main_loop_wait (timeout=28000000) at util/main-loop.c:263 > #15 0x000000000093de05 in main_loop_wait (nonblocking=0) at util/main-loop.c:522 > #16 0x00000000005bc6a5 in main_loop () at vl.c:1944 > #17 0x00000000005c39b5 in main (argc=56, argv=0x7ffe2f6443f8, envp=0x3ad0030) at vl.c:4752 > > It does not get the RDMA_CM_EVENT_DISCONNECTED event after rdma_disconnect sometime. > I do not find out the root cause why not get RDMA_CM_EVENT_DISCONNECTED event, but > it can be reproduced if not invoke ibv_dereg_mr to release all ram blocks which fixed > in previous patch. Does this happen without your other changes? Can you give me instructions to repeat it and also say which cards you wereusing? > Anyway, it should not invoke rdma_get_cm_event in main thread, and the event channel > is also destroyed in qemu_rdma_cleanup. > > Signed-off-by: Lidong Chen > --- > migration/rdma.c | 12 ++---------- > migration/trace-events | 1 - > 2 files changed, 2 insertions(+), 11 deletions(-) > > diff --git a/migration/rdma.c b/migration/rdma.c > index 0dd4033..92e4d30 100644 > --- a/migration/rdma.c > +++ b/migration/rdma.c > @@ -2275,8 +2275,7 @@ static int qemu_rdma_write(QEMUFile *f, RDMAContext *rdma, > > static void qemu_rdma_cleanup(RDMAContext *rdma) > { > - struct rdma_cm_event *cm_event; > - int ret, idx; > + int idx; > > if (rdma->cm_id && rdma->connected) { > if ((rdma->error_state || > @@ -2290,14 +2289,7 @@ static void qemu_rdma_cleanup(RDMAContext *rdma) > qemu_rdma_post_send_control(rdma, NULL, &head); > } > > - ret = rdma_disconnect(rdma->cm_id); > - if (!ret) { > - trace_qemu_rdma_cleanup_waiting_for_disconnect(); > - ret = rdma_get_cm_event(rdma->channel, &cm_event); > - if (!ret) { > - rdma_ack_cm_event(cm_event); > - } > - } > + rdma_disconnect(rdma->cm_id); I'm worried whether this change could break stuff: The docs say for rdma_disconnect that it flushes any posted work requests to the completion queue; so unless we wait for the event do we know the stuff has been flushed? In the normal non-cancel case I'm worried that means we could lose something. (But I don't know the rdma/infiniband specs well enough to know if it's really a problem). Dave > trace_qemu_rdma_cleanup_disconnect(); > rdma->connected = false; > } > diff --git a/migration/trace-events b/migration/trace-events > index d6be74b..64573ff 100644 > --- a/migration/trace-events > +++ b/migration/trace-events > @@ -125,7 +125,6 @@ qemu_rdma_accept_pin_state(bool pin) "%d" > qemu_rdma_accept_pin_verbsc(void *verbs) "Verbs context after listen: %p" > qemu_rdma_block_for_wrid_miss(const char *wcompstr, int wcomp, const char *gcompstr, uint64_t req) "A Wanted wrid %s (%d) but got %s (%" PRIu64 ")" > qemu_rdma_cleanup_disconnect(void) "" > -qemu_rdma_cleanup_waiting_for_disconnect(void) "" > qemu_rdma_close(void) "" > qemu_rdma_connect_pin_all_requested(void) "" > qemu_rdma_connect_pin_all_outcome(bool pin) "%d" > -- > 1.8.3.1 > -- Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK