From: Alexey Kardashevskiy <aik@ozlabs.ru>
To: Alex Williamson <alex.williamson@redhat.com>,
Jike Song <jike.song@intel.com>
Cc: linuxppc-dev@lists.ozlabs.org,
David Gibson <david@gibson.dropbear.id.au>,
Paul Mackerras <paulus@samba.org>,
kvm-ppc@vger.kernel.org, kvm@vger.kernel.org
Subject: Re: [PATCH kernel v2 10/11] vfio: Check for unregistered notifiers when group is actually released
Date: Tue, 20 Dec 2016 09:41:29 +1100 [thread overview]
Message-ID: <533b7e4c-8c3a-5c37-4c25-66bdbd265ec5@ozlabs.ru> (raw)
In-Reply-To: <20161219092805.1887ea5d@t450s.home>
On 20/12/16 03:28, Alex Williamson wrote:
> On Mon, 19 Dec 2016 18:41:05 +0800
> Jike Song <jike.song@intel.com> wrote:
>
>> On 12/18/2016 09:28 AM, Alexey Kardashevskiy wrote:
>>> This moves a check for unregistered notifiers from fops release
>>> callback to the place where the group will actually be released.
>>>
>>> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
>>> ---
>>>
>>> This is going to be used in the following patch in cleanup
>>> path. Since the next patch is RFC, this one might not be needed.
>
> Alexey, this is intended to be a bug fix, it should be sent separately,
> not buried in an unrelated patch series.
Well, I was pretty unsure this is a correct fix and I was sure that it
would make sense without the context which is quite weird :)
>
>> I didn't find any use in the following patch 11/11, did you mean
>> something else?
>>
>> BTW the warning in vfio_group_release seems too late, the user
>> should actually unregister everything by close()?
>
> The thing is, it's not the user that registered the notifiers, it's the
> vendor driver. The vendor driver should know via the device release to
> unregister the notifier, which we're counting on to happen before the
> group release. Can we rely on that ordering even in the case where a
> user is SIGKILL'd?
>
>>> ---
>>> drivers/vfio/vfio.c | 5 ++---
>>> 1 file changed, 2 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
>>> index 6b9a98508939..083b581e87c0 100644
>>> --- a/drivers/vfio/vfio.c
>>> +++ b/drivers/vfio/vfio.c
>>> @@ -403,6 +403,8 @@ static void vfio_group_release(struct kref *kref)
>>> struct iommu_group *iommu_group = group->iommu_group;
>>>
>>> WARN_ON(!list_empty(&group->device_list));
>>> + /* Any user didn't unregister? */
>>> + WARN_ON(group->notifier.head);
>
> Even if this is a bug, this is the wrong fix. This is when the group
> is being destroyed. Yes, it would be a bug to still have any notifiers
> here, but the intention is to make sure there are no notifiers when the
> group is idle and unused.
Out of curiosity - vendor drivers are supposed to hold a group file open,
not just reference vfio_grop/iommu_group objects?
>
>>>
>>> list_for_each_entry_safe(unbound, tmp,
>>> &group->unbound_list, unbound_next) {
>>> @@ -1584,9 +1586,6 @@ static int vfio_group_fops_release(struct inode *inode, struct file *filep)
>>>
>>> filep->private_data = NULL;
>>>
>>> - /* Any user didn't unregister? */
>>> - WARN_ON(group->notifier.head);
>>> -
>>> vfio_group_try_dissolve_container(group);
>>>
>>> atomic_dec(&group->opened);
>>>
>
--
Alexey
next prev parent reply other threads:[~2016-12-19 22:41 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-12-18 1:28 [PATCH kernel v2 00/11] powerpc/kvm/vfio: Enable in-kernel acceleration Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 01/11] KVM: PPC: Reserve KVM_CAP_SPAPR_TCE_VFIO capability number Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 02/11] powerpc/iommu: Cleanup iommu_table disposal Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 03/11] powerpc/vfio_spapr_tce: Add reference counting to iommu_table Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 04/11] powerpc/mmu: Add real mode support for IOMMU preregistered memory Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 05/11] KVM: PPC: Use preregistered memory API to access TCE list Alexey Kardashevskiy
2016-12-21 4:08 ` David Gibson
2016-12-21 8:57 ` Alexey Kardashevskiy
2017-01-11 6:35 ` Alexey Kardashevskiy
2017-01-12 5:49 ` David Gibson
2016-12-18 1:28 ` [PATCH kernel v2 06/11] powerpc/powernv/iommu: Add real mode version of iommu_table_ops::exchange() Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 07/11] KVM: PPC: Enable IOMMU_API for KVM_BOOK3S_64 permanently Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 08/11] KVM: PPC: Pass kvm* to kvmppc_find_table() Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 09/11] vfio iommu: Add helpers to (un)register blocking notifiers per group Alexey Kardashevskiy
2016-12-21 6:04 ` David Gibson
2016-12-22 1:25 ` Alexey Kardashevskiy
2016-12-18 1:28 ` [PATCH kernel v2 10/11] vfio: Check for unregistered notifiers when group is actually released Alexey Kardashevskiy
2016-12-19 10:41 ` Jike Song
2016-12-19 16:28 ` Alex Williamson
2016-12-19 22:41 ` Alexey Kardashevskiy [this message]
2016-12-18 1:29 ` [PATCH kernel v2 11/11] KVM: PPC: Add in-kernel acceleration for VFIO Alexey Kardashevskiy
2016-12-20 6:52 ` [PATCH kernel v3] " Alexey Kardashevskiy
2017-01-12 5:04 ` David Gibson
2017-01-12 8:09 ` Alexey Kardashevskiy
2017-01-12 23:53 ` David Gibson
2017-01-13 2:23 ` Alexey Kardashevskiy
2017-01-13 2:38 ` David Gibson
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=533b7e4c-8c3a-5c37-4c25-66bdbd265ec5@ozlabs.ru \
--to=aik@ozlabs.ru \
--cc=alex.williamson@redhat.com \
--cc=david@gibson.dropbear.id.au \
--cc=jike.song@intel.com \
--cc=kvm-ppc@vger.kernel.org \
--cc=kvm@vger.kernel.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=paulus@samba.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).