From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from lists.xenproject.org (lists.xenproject.org [192.237.175.120]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id 94D91C77B73 for ; Tue, 2 May 2023 19:41:34 +0000 (UTC) Received: from list by lists.xenproject.org with outflank-mailman.528794.822385 (Exim 4.92) (envelope-from ) id 1ptvsB-0007bx-Rc; Tue, 02 May 2023 19:41:15 +0000 X-Outflank-Mailman: Message body and most headers restored to incoming version Received: by outflank-mailman (output) from mailman id 528794.822385; Tue, 02 May 2023 19:41:15 +0000 Received: from localhost ([127.0.0.1] helo=lists.xenproject.org) by lists.xenproject.org with esmtp (Exim 4.92) (envelope-from ) id 1ptvsB-0007bq-O9; Tue, 02 May 2023 19:41:15 +0000 Received: by outflank-mailman (input) for mailman id 528794; Tue, 02 May 2023 19:41:14 +0000 Received: from se1-gles-flk1-in.inumbo.com ([94.247.172.50] helo=se1-gles-flk1.inumbo.com) by lists.xenproject.org with esmtp (Exim 4.92) (envelope-from ) id 1ptvsA-0007bk-Re for xen-devel@lists.xenproject.org; Tue, 02 May 2023 19:41:14 +0000 Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.129.124]) by se1-gles-flk1.inumbo.com (Halon) with ESMTPS id 4a9a5fbc-e921-11ed-8611-37d641c3527e; Tue, 02 May 2023 21:41:12 +0200 (CEST) Received: from mimecast-mx02.redhat.com (66.187.233.88 [66.187.233.88]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-335-YMLPGPKMO4SxOjhl3dEpxA-1; Tue, 02 May 2023 15:41:01 -0400 Received: from smtp.corp.redhat.com (int-mx02.intmail.prod.int.rdu2.redhat.com [10.11.54.2]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id E962D85A588; Tue, 2 May 2023 19:40:54 +0000 (UTC) Received: from localhost (unknown [10.39.192.230]) by smtp.corp.redhat.com (Postfix) with ESMTP id E85EC4020960; Tue, 2 May 2023 19:40:47 +0000 (UTC) X-BeenThere: xen-devel@lists.xenproject.org List-Id: Xen developer discussion List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Errors-To: xen-devel-bounces@lists.xenproject.org Precedence: list Sender: "Xen-devel" X-Inumbo-ID: 4a9a5fbc-e921-11ed-8611-37d641c3527e DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1683056471; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version:content-type:content-type: in-reply-to:in-reply-to:references:references; bh=6xENbjVvp+Pts3X84Cy2DsWSBfSYloGA+NzmbgbplRU=; b=fOt5e4RU5TCQFUGbDzQYsMURmBV4TTppPLwFTDXuiOX/1HAeWzGUfAZBXzXiWnCsNNQ+6K ZzSkLnn2K9rjM23rxRBj5B/SGsNXfh3w0WRwxeaPssy8iGjfAXZmMM26l+zMdFxLzhUNlf R+HFjIHQCOB0Hw5mFMP//N1FNQGQAhg= X-MC-Unique: YMLPGPKMO4SxOjhl3dEpxA-1 Date: Tue, 2 May 2023 15:40:45 -0400 From: Stefan Hajnoczi To: Kevin Wolf Cc: qemu-devel@nongnu.org, Daniel =?iso-8859-1?Q?P=2E_Berrang=E9?= , Juan Quintela , Julia Suvorova , xen-devel@lists.xenproject.org, eesposit@redhat.com, Richard Henderson , Fam Zheng , "Michael S. Tsirkin" , Coiby Xu , David Woodhouse , Marcel Apfelbaum , Peter Lieven , Paul Durrant , "Richard W.M. Jones" , qemu-block@nongnu.org, Stefano Garzarella , Anthony Perard , Stefan Weil , Xie Yongji , Paolo Bonzini , Aarushi Mehta , Philippe =?iso-8859-1?Q?Mathieu-Daud=E9?= , Eduardo Habkost , Stefano Stabellini , Hanna Reitz , Ronnie Sahlberg Subject: Re: [PATCH v4 06/20] block/export: wait for vhost-user-blk requests when draining Message-ID: <20230502194045.GC535070@fedora> References: <20230425172716.1033562-1-stefanha@redhat.com> <20230425172716.1033562-7-stefanha@redhat.com> MIME-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha256; protocol="application/pgp-signature"; boundary="XRuGd9g1z5oS2ell" Content-Disposition: inline In-Reply-To: X-Scanned-By: MIMEDefang 3.1 on 10.11.54.2 --XRuGd9g1z5oS2ell Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Tue, May 02, 2023 at 05:42:51PM +0200, Kevin Wolf wrote: > Am 25.04.2023 um 19:27 hat Stefan Hajnoczi geschrieben: > > Each vhost-user-blk request runs in a coroutine. When the BlockBackend > > enters a drained section we need to enter a quiescent state. Currently > > any in-flight requests race with bdrv_drained_begin() because it is > > unaware of vhost-user-blk requests. > >=20 > > When blk_co_preadv/pwritev()/etc returns it wakes the > > bdrv_drained_begin() thread but vhost-user-blk request processing has > > not yet finished. The request coroutine continues executing while the > > main loop thread thinks it is in a drained section. > >=20 > > One example where this is unsafe is for blk_set_aio_context() where > > bdrv_drained_begin() is called before .aio_context_detached() and > > .aio_context_attach(). If request coroutines are still running after > > bdrv_drained_begin(), then the AioContext could change underneath them > > and they race with new requests processed in the new AioContext. This > > could lead to virtqueue corruption, for example. > >=20 > > (This example is theoretical, I came across this while reading the > > code and have not tried to reproduce it.) > >=20 > > It's easy to make bdrv_drained_begin() wait for in-flight requests: add > > a .drained_poll() callback that checks the VuServer's in-flight counter. > > VuServer just needs an API that returns true when there are requests in > > flight. The in-flight counter needs to be atomic. > >=20 > > Signed-off-by: Stefan Hajnoczi > > --- > > include/qemu/vhost-user-server.h | 4 +++- > > block/export/vhost-user-blk-server.c | 16 ++++++++++++++++ > > util/vhost-user-server.c | 14 ++++++++++---- > > 3 files changed, 29 insertions(+), 5 deletions(-) > >=20 > > diff --git a/include/qemu/vhost-user-server.h b/include/qemu/vhost-user= -server.h > > index bc0ac9ddb6..b1c1cda886 100644 > > --- a/include/qemu/vhost-user-server.h > > +++ b/include/qemu/vhost-user-server.h > > @@ -40,8 +40,9 @@ typedef struct { > > int max_queues; > > const VuDevIface *vu_iface; > > =20 > > + unsigned int in_flight; /* atomic */ > > + > > /* Protected by ctx lock */ > > - unsigned int in_flight; > > bool wait_idle; > > VuDev vu_dev; > > QIOChannel *ioc; /* The I/O channel with the client */ > > @@ -62,6 +63,7 @@ void vhost_user_server_stop(VuServer *server); > > =20 > > void vhost_user_server_inc_in_flight(VuServer *server); > > void vhost_user_server_dec_in_flight(VuServer *server); > > +bool vhost_user_server_has_in_flight(VuServer *server); > > =20 > > void vhost_user_server_attach_aio_context(VuServer *server, AioContext= *ctx); > > void vhost_user_server_detach_aio_context(VuServer *server); > > diff --git a/block/export/vhost-user-blk-server.c b/block/export/vhost-= user-blk-server.c > > index 841acb36e3..092b86aae4 100644 > > --- a/block/export/vhost-user-blk-server.c > > +++ b/block/export/vhost-user-blk-server.c > > @@ -272,7 +272,20 @@ static void vu_blk_exp_resize(void *opaque) > > vu_config_change_msg(&vexp->vu_server.vu_dev); > > } > > =20 > > +/* > > + * Ensures that bdrv_drained_begin() waits until in-flight requests co= mplete. > > + * > > + * Called with vexp->export.ctx acquired. > > + */ > > +static bool vu_blk_drained_poll(void *opaque) > > +{ > > + VuBlkExport *vexp =3D opaque; > > + > > + return vhost_user_server_has_in_flight(&vexp->vu_server); > > +} > > + > > static const BlockDevOps vu_blk_dev_ops =3D { > > + .drained_poll =3D vu_blk_drained_poll, > > .resize_cb =3D vu_blk_exp_resize, > > }; >=20 > You're adding a new function pointer to an existing BlockDevOps... >=20 > > @@ -314,6 +327,7 @@ static int vu_blk_exp_create(BlockExport *exp, Bloc= kExportOptions *opts, > > vu_blk_initialize_config(blk_bs(exp->blk), &vexp->blkcfg, > > logical_block_size, num_queues); > > =20 > > + blk_set_dev_ops(exp->blk, &vu_blk_dev_ops, vexp); > > blk_add_aio_context_notifier(exp->blk, blk_aio_attached, blk_aio_d= etach, > > vexp); > > =20 > > blk_set_dev_ops(exp->blk, &vu_blk_dev_ops, vexp); >=20 > ..but still add a second blk_set_dev_ops(). Maybe a bad merge conflict > resolution with commit ca858a5fe94? Thanks, I probably didn't have ca858a5fe94 in my tree when writing this code. > > @@ -323,6 +337,7 @@ static int vu_blk_exp_create(BlockExport *exp, Bloc= kExportOptions *opts, > > num_queues, &vu_blk_iface, errp)) { > > blk_remove_aio_context_notifier(exp->blk, blk_aio_attached, > > blk_aio_detach, vexp); > > + blk_set_dev_ops(exp->blk, NULL, NULL); > > g_free(vexp->handler.serial); > > return -EADDRNOTAVAIL; > > } > > @@ -336,6 +351,7 @@ static void vu_blk_exp_delete(BlockExport *exp) > > =20 > > blk_remove_aio_context_notifier(exp->blk, blk_aio_attached, blk_ai= o_detach, > > vexp); > > + blk_set_dev_ops(exp->blk, NULL, NULL); > > g_free(vexp->handler.serial); > > } >=20 > These two hunks are then probably already fixes for ca858a5fe94 and > should be a separate patch if so. Sure, I can split them out. hw/ doesn't need to call blk_set_dev_ops(blk, NULL, NULL) because hw/core/qdev-properties-system.c:release_drive() -> blk_detach_dev() does it automatically, but block/export does. It's easy to overlook and that's probably why ca858a5fe94 didn't include it. > > diff --git a/util/vhost-user-server.c b/util/vhost-user-server.c > > index 1622f8cfb3..2e6b640050 100644 > > --- a/util/vhost-user-server.c > > +++ b/util/vhost-user-server.c > > @@ -78,17 +78,23 @@ static void panic_cb(VuDev *vu_dev, const char *buf) > > void vhost_user_server_inc_in_flight(VuServer *server) > > { > > assert(!server->wait_idle); > > - server->in_flight++; > > + qatomic_inc(&server->in_flight); > > } > > =20 > > void vhost_user_server_dec_in_flight(VuServer *server) > > { > > - server->in_flight--; > > - if (server->wait_idle && !server->in_flight) { > > - aio_co_wake(server->co_trip); > > + if (qatomic_fetch_dec(&server->in_flight) =3D=3D 1) { > > + if (server->wait_idle) { > > + aio_co_wake(server->co_trip); > > + } > > } > > } > > =20 > > +bool vhost_user_server_has_in_flight(VuServer *server) > > +{ > > + return qatomic_load_acquire(&server->in_flight) > 0; > > +} > > + >=20 > Any reason why you left the server->in_flight accesses in > vu_client_trip() non-atomic? I don't remember if it was a mistake or if there is a reason why it's safe. I'll replace those accesses with calls to vhost_user_server_has_in_flight(). Stefan --XRuGd9g1z5oS2ell Content-Type: application/pgp-signature; name="signature.asc" -----BEGIN PGP SIGNATURE----- iQEzBAEBCAAdFiEEhpWov9P5fNqsNXdanKSrs4Grc8gFAmRRZz0ACgkQnKSrs4Gr c8iEjwgAov+Ozs58HYaiYo0b4VnPoOnAM2QqHXYzmVRN4+5mqvFWNVKufgQMDbSG 1o1dgDNOCRU1tpOCUbNRYmvJxvYY6QA9Ho7AdLGPZc6jq2CR4LHarr5MP1Py5ktT dGAN6GFH3qzsf93j4wEa0HnWax5RvOdFEPxkK2JKgXRA+AesbOLRizK1q2P5p3TH 6I0SfPnLhlTeosVaQ4mRLkZuXNt5/bTeh54lW/NSLP6IpbBoB082Wqr1JqCwjdVO XxBNAMrMB/0oImJNHh9HSqpB2oHlL2FOa/yu++wNeZ1uzHdt7oRDpgwx2mQpMmiT /HhN2vFe1XsM621h8uV5aa4HZH7Hiw== =UcN5 -----END PGP SIGNATURE----- --XRuGd9g1z5oS2ell--