From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([2001:4830:134:3::10]:36188) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1WrpqB-0007zJ-I0 for qemu-devel@nongnu.org; Tue, 03 Jun 2014 10:37:34 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1Wrpq6-000206-Lu for qemu-devel@nongnu.org; Tue, 03 Jun 2014 10:37:27 -0400 Received: from mx1.redhat.com ([209.132.183.28]:60576) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1Wrpq6-0001zw-E8 for qemu-devel@nongnu.org; Tue, 03 Jun 2014 10:37:22 -0400 Received: from int-mx02.intmail.prod.int.phx2.redhat.com (int-mx02.intmail.prod.int.phx2.redhat.com [10.5.11.12]) by mx1.redhat.com (8.14.4/8.14.4) with ESMTP id s53EbLiF009116 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=OK) for ; Tue, 3 Jun 2014 10:37:21 -0400 Date: Tue, 3 Jun 2014 16:37:19 +0200 From: Kevin Wolf Message-ID: <20140603143719.GD3264@noname.str.redhat.com> References: <1401804987-31085-1-git-send-email-pbonzini@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1401804987-31085-1-git-send-email-pbonzini@redhat.com> Subject: Re: [Qemu-devel] [PATCH] block: asynchronously stop the VM on I/O errors List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: Paolo Bonzini Cc: famz@redhat.com, qemu-devel@nongnu.org, stefanha@redhat.com Am 03.06.2014 um 16:16 hat Paolo Bonzini geschrieben: > With virtio-blk dataplane, I/O errors might occur while QEMU is > not in the main I/O thread. However, it's invalid to call vm_stop > when we're neither in a VCPU thread nor in the main I/O thread, > even if we were to take the iothread mutex around it. > > To avoid this problem, simply raise a request to the main I/O thread, > similar to what QEMU does when vm_stop is called from a CPU thread. > We know that bdrv_error_action is called from an AIO callback, and > the moment at which the callback will fire is not well-defined; it > depends on the moment at which the disk or OS finishes the operation, > which can happen at any time. > > Note that QEMU is certainly not in a CPU thread and we do not need to > call cpu_stop_current() like vm_stop() does. Do I understand correctly that this is not a fundamental truth of qemu's operation, but holds true only because the drivers that do support rerror/werror all use bdrv_aio_readv/writev(), which guarantees that a BH is used in error cases? Otherwise I think an I/O handler in a vcpu thread could directly call into the block layer and fail immediately (might happen for example if we added rerror/werror support to ATAPI). > This makes bdrv_error_action() thread safe. > > Signed-off-by: Paolo Bonzini > --- > block.c | 2 +- > stubs/vm-stop.c | 2 +- > 2 files changed, 2 insertions(+), 2 deletions(-) > > diff --git a/block.c b/block.c > index fc2edd3..fa41598 100644 > --- a/block.c > +++ b/block.c > @@ -3515,7 +3515,7 @@ void bdrv_error_action(BlockDriverState *bs, BlockErrorAction action, > assert(error >= 0); > bdrv_emit_qmp_error_event(bs, QEVENT_BLOCK_IO_ERROR, action, is_read); > if (action == BDRV_ACTION_STOP) { > - vm_stop(RUN_STATE_IO_ERROR); > + qemu_system_vmstop_request(RUN_STATE_IO_ERROR); > bdrv_iostatus_set_err(bs, error); By delaying the actual state change, does this break the invariant that bs->iostatus is BLOCK_DEVICE_IO_STATUS_OK while the VM is running? I know this invariant was mentioned occasionally. Not sure if anything actually breaks when it's violated, though. Kevin