From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:42112) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UIdOQ-0004yI-Bv for qemu-devel@nongnu.org; Thu, 21 Mar 2013 07:10:52 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1UIdOH-000257-Ux for qemu-devel@nongnu.org; Thu, 21 Mar 2013 07:10:46 -0400 Received: from ssl.dlhnet.de ([91.198.192.8]:34117 helo=ssl.dlh.net) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1UIdOH-00023f-PZ for qemu-devel@nongnu.org; Thu, 21 Mar 2013 07:10:37 -0400 Message-ID: <514AEAB7.2070506@dlhnet.de> Date: Thu, 21 Mar 2013 12:10:47 +0100 From: Peter Lieven MIME-Version: 1.0 References: <514AE4CF.70009@dlhnet.de> <514AE8F8.6010301@dlhnet.de> In-Reply-To: <514AE8F8.6010301@dlhnet.de> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Subject: Re: [Qemu-devel] iscsi / block segfault regression List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: "qemu-devel@nongnu.org" Cc: Kevin Wolf , Paolo Bonzini , ronnie sahlberg Segfault happens if I specify format=iscsi in the -drive option group. In this case bdrv_swap() is called. Maybe there is a handling of bs->opaque missing?! Peter On 21.03.2013 12:03, Peter Lieven wrote: > git bisect results: > > lieven@lieven-pc:~/git/qemu$ git bisect bad > f500a6d3c2b9ef0bb06d0080d91d8ed3c1d68f58 is the first bad commit > commit f500a6d3c2b9ef0bb06d0080d91d8ed3c1d68f58 > Author: Kevin Wolf > Date: Mon Nov 12 17:35:27 2012 +0100 > > block: Avoid second open for format probing > > This fixes problems that are caused by the additional open/close cycle > of the existing format probing, for example related to qemu-nbd without > -t option or file descriptor passing. > > Signed-off-by: Kevin Wolf > > :100644 100644 c4f5566b5de6cfa371d7b60dd8b0e1a2da7fd55f 2ec3afebfe6a6c12b573cba9ad36de800623fe6a M block.c > > > On 21.03.2013 11:45, Peter Lieven wrote: >> Hi, >> >> I currently try to debug a regression in git master with iscsi that happens with some iscsi volumes. >> I added debugging output to all aio calls into block/iscsi.c and it seems that immediately after >> brdv_open something is overwriting bs->opaque with a new pointer. >> >> ~/git/qemu$ x86_64-softmmu/qemu-system-x86_64 -k de -enable-kvm -m 1024 -drive format=iscsi,file=iscsi://172.21.200.31/iqn.2001-05.com.equallogic:0-8a0906-3c0f4e007-dce91829d0450add-lieven-storage-migration-test-hd0/0,if=virtio,cache=writeback,aio=native >> -vnc :1 -boot dc -monitor stdio >> iscsi open start >> iscsi_full_connect_sync: iscsilun = 140063347224208, iscsilun->iscsi = 140063347225536, bs = 140063347221696, filename=iscsi://172.21.200.31/iqn.2001-05.com.equallogic:0-8a0906-3c0f4e007-dce91829d0450add-lieven-storage-migration-test-hd0/0 >> iscsi open finish ret=0 >> iscsi_aio_flush: iscsilun = 140063347230736, iscsilun->iscsi = 0, bs = 140063347221696, filename=iscsi://172.21.200.31/iqn.2001-05.com.equallogic:0-8a0906-3c0f4e007-dce91829d0450add-lieven-storage-migration-test-hd0/0 >> iscsi_aio_flush_acb: iscsilun = 140063347230736, iscsilun->iscsi = 0, bs = 140063347221696, filename=iscsi://172.21.200.31/iqn.2001-05.com.equallogic:0-8a0906-3c0f4e007-dce91829d0450add-lieven-storage-migration-test-hd0/0 >> Speicherzugriffsfehler (Speicherabzug geschrieben) >> >> bs address is unchanged, but iscsilun address is changed after connection it seems to an area with zeroed memory. the pointer to the iscsi context >> is thus NULL. >> >> iscsilun, iscsilun->iscsi and bs are addresses. >> >> Peter >> >