Re: [Qemu-devel] [Qemu-block] [PATCH RFC 00/16] block: Protect AIO context change with perm API

[Fam Zheng <famz@redhat.com>]

On Mon, 04/10 10:04, Stefan Hajnoczi wrote:
> On Tue, Mar 21, 2017 at 11:16:19AM +0800, Fam Zheng wrote:
> > Eject / change of scsi-cd on a virtio-scsi dataplane bus causes abort() because
> > the new BDS doesn't get proper bdrv_set_aio_context().
> > 
> > Store the AioContext in BB and do it in blk_insert_bs. That is done by
> > Vladimir's patch.
> > 
> > Other patches are to make sure such a bdrv_set_aio_context() doesn't interfere
> > with other BBs using other nodes from this graph.
> > 
> > RFC note:
> > 
> > Unfortunately, a use-after-free crash in iotests 030 appears since patch 7,
> > which I believe is a latent bug that bdrv_reopen is "reentered" in existing
> > code, rather than from this series:
> > 
> > > #4  0x0000561ab90425a7 in bdrv_reopen
> > > #5  0x0000561ab8e1d28e in stream_complete
> > > #6  0x0000561ab9048543 in block_job_defer_to_main_loop_bh
> > > #7  0x0000561ab91305bc in aio_bh_call
> > > #8  0x0000561ab9130659 in aio_bh_poll
> > > #9  0x0000561ab9135656 in aio_poll
> > > #10 0x0000561ab90a6cf5 in bdrv_flush
> > > #11 0x0000561ab904285a in bdrv_reopen_prepare
> > > #12 0x0000561ab90423f0 in bdrv_reopen_multiple
> > > #13 0x0000561ab90425ef in bdrv_reopen
> > > #14 0x0000561ab909fa49 in commit_active_start
> > > #15 0x0000561ab8dd6ffb in qmp_block_commit
> > > #16 0x0000561ab8ded485 in qmp_marshal_block_commit
> > > #17 0x0000561ab9123e6c in do_qmp_dispatch
> > > #18 0x0000561ab9123fa4 in qmp_dispatch
> > > #19 0x0000561ab8ca26b7 in handle_qmp_command
> > 
> > I have a fix that I'll post separately.
> > 
> > The last patches are an alternative to patch 7, to "workaround" this in a
> > really non-obvious way.
> 
> Are there qemu-iotests that cover both success and failure scenarios for
> the new permission?

No, I'll add them.

Fam