On Wed, Aug 31, 2022 at 5:41 PM Richard W.M. Jones <rjones at redhat.com> wrote:> > On Wed, Aug 31, 2022 at 05:29:13PM +0800, Ming Lei wrote: > > I can reproduce it, please try the top patch in aio branch, which fixed > > hang in my reproducer with your test setting. > > > > https://github.com/ming1/ubdsrv/commits/aio > > (https://github.com/ming1/ubdsrv/commit/0a293b6eb7149dc5ee83e5d07d242accdb840c85) > > Yes, that seems to fix it. > > I have two loops, one git-cloning the kernel, and another copying the > source of nbdkit and recompiling it, and they are both working without > problems.BTW, aio branch has been merged to ublksrv/master, but the interface of ublksrv_aio_submit_req() changed, so please sync with ublksrv. thanks,
Richard W.M. Jones
2022-Aug-31 15:06 UTC
[Libguestfs] [PATCH libnbd] ublk: Add new nbdublk program
On Wed, Aug 31, 2022 at 10:49:42PM +0800, Ming Lei wrote:> On Wed, Aug 31, 2022 at 5:41 PM Richard W.M. Jones <rjones at redhat.com> wrote: > > > > On Wed, Aug 31, 2022 at 05:29:13PM +0800, Ming Lei wrote: > > > I can reproduce it, please try the top patch in aio branch, which fixed > > > hang in my reproducer with your test setting. > > > > > > https://github.com/ming1/ubdsrv/commits/aio > > > > (https://github.com/ming1/ubdsrv/commit/0a293b6eb7149dc5ee83e5d07d242accdb840c85) > > > > Yes, that seems to fix it. > > > > I have two loops, one git-cloning the kernel, and another copying the > > source of nbdkit and recompiling it, and they are both working without > > problems. > > BTW, aio branch has been merged to ublksrv/master, but the interface > of ublksrv_aio_submit_req() changed, so please sync with ublksrv.OK I have updated nbdublk: https://gitlab.com/rwmjones/libnbd/-/commits/nbdublk/ I have also added the changes to Fedora: https://koji.fedoraproject.org/koji/taskinfo?taskID=91468555 Please consider becoming a Fedora packager so you can comaintain this package. On the subject of the kernel issue, I have been running builds and git clones in a loop for many hours on a local (virtio-scsi) filesystem, and I have *not* seen the kernel memory corruption problem happening. I will go back to testing ublk again to see if the kernel memory corruption bug is reproducible. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones Read my programming and virtualization blog: http://rwmj.wordpress.com libguestfs lets you edit virtual machines. Supports shell scripting, bindings from many languages. http://libguestfs.org
On Wed, Aug 31, 2022 at 04:06:16PM +0100, Richard W.M. Jones wrote:> On Wed, Aug 31, 2022 at 10:49:42PM +0800, Ming Lei wrote: > > On Wed, Aug 31, 2022 at 5:41 PM Richard W.M. Jones <rjones at redhat.com> wrote: > > > > > > On Wed, Aug 31, 2022 at 05:29:13PM +0800, Ming Lei wrote: > > > > I can reproduce it, please try the top patch in aio branch, which fixed > > > > hang in my reproducer with your test setting. > > > > > > > > https://github.com/ming1/ubdsrv/commits/aio > > > > > > (https://github.com/ming1/ubdsrv/commit/0a293b6eb7149dc5ee83e5d07d242accdb840c85) > > > > > > Yes, that seems to fix it. > > > > > > I have two loops, one git-cloning the kernel, and another copying the > > > source of nbdkit and recompiling it, and they are both working without > > > problems. > > > > BTW, aio branch has been merged to ublksrv/master, but the interface > > of ublksrv_aio_submit_req() changed, so please sync with ublksrv. > > OK I have updated nbdublk: > > https://gitlab.com/rwmjones/libnbd/-/commits/nbdublk/ > > I have also added the changes to Fedora: > > https://koji.fedoraproject.org/koji/taskinfo?taskID=91468555 > > Please consider becoming a Fedora packager so you can comaintain > this package.OK, is there any guide about how to be a Fedora packager?> > On the subject of the kernel issue, I have been running builds and git > clones in a loop for many hours on a local (virtio-scsi) filesystem, > and I have *not* seen the kernel memory corruption problem happening. > > I will go back to testing ublk again to see if the kernel memory > corruption bug is reproducible.I am also running xfstests on both ublk and nbd, and compares results from the two. Thanks, Ming