Message ID | 20230915102531.55894-1-hreitz@redhat.com (mailing list archive) |
---|---|
Headers | show |
Series | vhost-user: Back-end state migration | expand |
On Fri, Sep 15, 2023 at 12:25:25PM +0200, Hanna Czenczek wrote: > RFC: > https://lists.nongnu.org/archive/html/qemu-devel/2023-03/msg04263.html > > v1: > https://lists.nongnu.org/archive/html/qemu-devel/2023-04/msg01575.html > > v2: > https://lists.nongnu.org/archive/html/qemu-devel/2023-07/msg02604.html > > Hi, > > I’ve decided not to work on vhost-user SUSPEND/RESUME for now – it is > not technically required for virtio-fs migration, which is the actual > priority for me now. While we do want to have SUSPEND/RESUME at some > point, the only practically existing reason for it is to be able to > implement vhost-level resetting in virtiofsd, but that is not related to > migration. QEMU sends VHOST_USER_SET_STATUS 0 in vhost_dev_stop(). Are you assuming that virtiofs back-ends do not reset the device upon receiving this message? > So one of the changes in v3 is that it no longer depends on the > vhost-user SUSPEND/RESUME series, and describes the migration protocol > without the device being suspended at any point, but merely that the > vrings are stopped. > > Other changes include: > > - Patch 1: > - Rephrased a lot > - Added a description for the VHOST_USER_SET_DEVICE_STATE_FD > parameters > - Renamed VHOST_USER_PROTOCOL_F_MIGRATORY_STATE to > VHOST_USER_PROTOCOL_F_DEVICE_STATE > - enum variants changed in value due to dropping the SUSPEND/RESUME > dependency > > - Patch 2: > - Pulled in, was a stand-alone patch before > - Dropped a sentence about ring state before feature negotiations, as > the rings are not to be used during that period anyway > - Bit of rephrasing > > - Patch 3: > - Renamed “migratory state” to “device state” > - enum variants changed in value due to dropping the SUSPEND/RESUME > dependency > > - Patch 4: > - Changed `f` to @f (referencing parameter “f”) in comments > - Use g_autofree for the transfer buffer > - Note SUSPEND state as a future feature, not currently existing > - Wrap read() and write() in RETRY_ON_EINTR() > > - Patch 5: > - Renamed “migratory state” to “device state” > - (kept R-b still) > > > Hanna Czenczek (5): > vhost-user.rst: Migrating back-end-internal state > vhost-user.rst: Clarify enabling/disabling vrings > vhost-user: Interface for migration state transfer > vhost: Add high-level state save/load functions > vhost-user-fs: Implement internal migration > > docs/interop/vhost-user.rst | 188 ++++++++++++++++++++++- > include/hw/virtio/vhost-backend.h | 24 +++ > include/hw/virtio/vhost.h | 114 ++++++++++++++ > hw/virtio/vhost-user-fs.c | 101 ++++++++++++- > hw/virtio/vhost-user.c | 148 ++++++++++++++++++ > hw/virtio/vhost.c | 241 ++++++++++++++++++++++++++++++ > 6 files changed, 810 insertions(+), 6 deletions(-) > > -- > 2.41.0 >
On 25.09.23 22:48, Stefan Hajnoczi wrote: > On Fri, Sep 15, 2023 at 12:25:25PM +0200, Hanna Czenczek wrote: >> RFC: >> https://lists.nongnu.org/archive/html/qemu-devel/2023-03/msg04263.html >> >> v1: >> https://lists.nongnu.org/archive/html/qemu-devel/2023-04/msg01575.html >> >> v2: >> https://lists.nongnu.org/archive/html/qemu-devel/2023-07/msg02604.html >> >> Hi, >> >> I’ve decided not to work on vhost-user SUSPEND/RESUME for now – it is >> not technically required for virtio-fs migration, which is the actual >> priority for me now. While we do want to have SUSPEND/RESUME at some >> point, the only practically existing reason for it is to be able to >> implement vhost-level resetting in virtiofsd, but that is not related to >> migration. > QEMU sends VHOST_USER_SET_STATUS 0 in vhost_dev_stop(). Are you assuming > that virtiofs back-ends do not reset the device upon receiving this > message? Absolutely. vhost_dev_stop() is not in the migration-specific path, but is called whenever the vCPUs are stopped, which indeed happens to be part of migration, but is also used in other cases, like QMP stop. We have identified that it is wrong that we reset the back-end just because the vCPUs are stopped (e.g. on migration), but it is what we do right now when the VM is paused (e.g. through QMP stop). Therefore, stateful back-ends cannot implement reset lest stop/cont breaks the device. I don’t think anybody really cares whether a vhost-user back-end actually resets its internal state (if there is any) when the guest driver asks for a reset on the virtio level, as long as the guest driver is then able to initialize the device afterwards. I do think people care that stop/cont works, so it follows to me that no stateful back-end will reset its internal state when receiving a virtio/vhost reset. If they do, stop/cont breaks, which is a user-visible bug that needs to be addressed – either properly by implementing SUSPEND/RESUME in both qemu and the affected back-ends, or by using a similar work-around to virtiofsd, which is to ignore reset commands. It’s hard for me to imagine that people don’t care that stop/cont breaks some vhost-user back-end, but suddenly would start to care that migration doesn’t work – especially given that first of all someone will need to manually implement any migration support in that back-end even with this series, which means that really, the only back-end we are talking about here is our virtiofsd. To this day I’m not even aware of any other back-end that has internal state. Hanna
On Tue, Sep 26, 2023, 09:33 Hanna Czenczek <hreitz@redhat.com> wrote: > On 25.09.23 22:48, Stefan Hajnoczi wrote: > > On Fri, Sep 15, 2023 at 12:25:25PM +0200, Hanna Czenczek wrote: > >> RFC: > >> https://lists.nongnu.org/archive/html/qemu-devel/2023-03/msg04263.html > >> > >> v1: > >> https://lists.nongnu.org/archive/html/qemu-devel/2023-04/msg01575.html > >> > >> v2: > >> https://lists.nongnu.org/archive/html/qemu-devel/2023-07/msg02604.html > >> > >> Hi, > >> > >> I’ve decided not to work on vhost-user SUSPEND/RESUME for now – it is > >> not technically required for virtio-fs migration, which is the actual > >> priority for me now. While we do want to have SUSPEND/RESUME at some > >> point, the only practically existing reason for it is to be able to > >> implement vhost-level resetting in virtiofsd, but that is not related to > >> migration. > > QEMU sends VHOST_USER_SET_STATUS 0 in vhost_dev_stop(). Are you assuming > > that virtiofs back-ends do not reset the device upon receiving this > > message? > > Absolutely. vhost_dev_stop() is not in the migration-specific path, but > is called whenever the vCPUs are stopped, which indeed happens to be > part of migration, but is also used in other cases, like QMP stop. We > have identified that it is wrong that we reset the back-end just because > the vCPUs are stopped (e.g. on migration), but it is what we do right > now when the VM is paused (e.g. through QMP stop). > > Therefore, stateful back-ends cannot implement reset lest stop/cont > breaks the device. I don’t think anybody really cares whether a > vhost-user back-end actually resets its internal state (if there is any) > when the guest driver asks for a reset on the virtio level, as long as > the guest driver is then able to initialize the device afterwards. Some devices send configuration change notifications. For example, virtio-net speed and duplex changes. Imagine a network boot ROM runs and the firmware resets the virtio-net device when transferring control to the loaded kernel. Before the kernel driver initializes the device again, the vhost-user-net back-end reports a speed or duplex change and sends a Configuration Change Notification to the guest. The guest receives a spurious interrupt because the vhost-user-net device wasn't actually reset. I'm concerned that ignoring reset matters (admittedly in corner cases) and think that stateful device functionality shouldn't be added to the vhost-user protocol without a solution for reset. This patch series changes the vhost-user protocol, which is used by many different devices, not just virtiofs. The solution should work for vhost-user devices of any type and not be based on what we can get away with when running the current QEMU + virtiofsd. I do > think people care that stop/cont works, so it follows to me that no > stateful back-end will reset its internal state when receiving a > virtio/vhost reset. If they do, stop/cont breaks, which is a > user-visible bug that needs to be addressed – either properly by > implementing SUSPEND/RESUME in both qemu and the affected back-ends, or > by using a similar work-around to virtiofsd, which is to ignore reset > commands. > Properly, please. > It’s hard for me to imagine that people don’t care that stop/cont breaks > some vhost-user back-end, but suddenly would start to care that > migration doesn’t work – especially given that first of all someone will > need to manually implement any migration support in that back-end even > with this series, which means that really, the only back-end we are > talking about here is our virtiofsd. To this day I’m not even aware of > any other back-end that has internal state. > Another one I can think of is vhost-user-gpu. Why did you give up on implementing SUSPEND/RESUME? Stefan > Hanna > > >
Hi Hanna, I was thinking about how this could work without SUSPEND/RESUME. What do you think of the following? 1. The front-end sends VHOST_USER_RESET_DEVICE (or VHOST_USER_RESET_OWNER, when necessary) when the guest driver resets the device but not on vhost_dev_start()/vhost_dev_stop(). 2. Suspend the device when all virtqueues are stopped via VHOST_USER_GET_VRING_BASE. Resume the device after at least one virtqueue is started and enabled. 3. Ignore VHOST_USER_SET_STATUS. Reset would work. The device would suspend and resume without losing state. Existing vhost-user backends already behave like this in practice (they often don't implement RESET_DEVICE). It's close enough to what you're proposing that it doesn't require much additional work, but I think it covers the cases. Two concerns: 1. It's specific to vhost-user and diverges from vDPA. 2. VHOST_USER_SET_STATUS might be needed in the future even though it's not useful today. Stefan
On 26.09.23 21:20, Stefan Hajnoczi wrote: > > > On Tue, Sep 26, 2023, 09:33 Hanna Czenczek <hreitz@redhat.com> wrote: > > On 25.09.23 22:48, Stefan Hajnoczi wrote: > > On Fri, Sep 15, 2023 at 12:25:25PM +0200, Hanna Czenczek wrote: > >> RFC: > >> > https://lists.nongnu.org/archive/html/qemu-devel/2023-03/msg04263.html > >> > >> v1: > >> > https://lists.nongnu.org/archive/html/qemu-devel/2023-04/msg01575.html > >> > >> v2: > >> > https://lists.nongnu.org/archive/html/qemu-devel/2023-07/msg02604.html > >> > >> Hi, > >> > >> I’ve decided not to work on vhost-user SUSPEND/RESUME for now – > it is > >> not technically required for virtio-fs migration, which is the > actual > >> priority for me now. While we do want to have SUSPEND/RESUME > at some > >> point, the only practically existing reason for it is to be able to > >> implement vhost-level resetting in virtiofsd, but that is not > related to > >> migration. > > QEMU sends VHOST_USER_SET_STATUS 0 in vhost_dev_stop(). Are you > assuming > > that virtiofs back-ends do not reset the device upon receiving this > > message? > > > Absolutely. vhost_dev_stop() is not in the migration-specific > path, but > is called whenever the vCPUs are stopped, which indeed happens to be > part of migration, but is also used in other cases, like QMP > stop. We > have identified that it is wrong that we reset the back-end just > because > the vCPUs are stopped (e.g. on migration), but it is what we do right > now when the VM is paused (e.g. through QMP stop). > > Therefore, stateful back-ends cannot implement reset lest stop/cont > breaks the device. I don’t think anybody really cares whether a > vhost-user back-end actually resets its internal state (if there > is any) > when the guest driver asks for a reset on the virtio level, as > long as > the guest driver is then able to initialize the device afterwards. > > > Some devices send configuration change notifications. For example, > virtio-net speed and duplex changes. > > Imagine a network boot ROM runs and the firmware resets the virtio-net > device when transferring control to the loaded kernel. Before the > kernel driver initializes the device again, the vhost-user-net > back-end reports a speed or duplex change and sends a Configuration > Change Notification to the guest. The guest receives a spurious > interrupt because the vhost-user-net device wasn't actually reset. I don’t see how this relates to my argument that no stateful back-end can implement a full reset because doing so would break stop/cont. If vhost-user-net were stateful (which it isn’t, AFAIK), it could choose to implement a work-around such that it would stop sending notifications on reset, but not reset its internal state. Then, when qemu restores vring state in vhost_dev_start(), it would resume sending notifications. But again, I fail to see how this is not already an issue for stop/cont. > I'm concerned that ignoring reset matters (admittedly in corner cases) > and think that stateful device functionality shouldn't be added to the > vhost-user protocol without a solution for reset. I disagree. We have a stateful device already, whether we add functionality acknowledging this to the protocol or not. The problem exists. It is independent of migration. If there’s a problem because of this with migration, there’s a problem with stop/cont, too, that must already have been worked around. > This patch series changes the vhost-user protocol, which is used by > many different devices, not just virtiofs. The solution should work > for vhost-user devices of any type and not be based on what we can get > away with when running the current QEMU + virtiofsd. My argument was generic. Any existing stateful device implementation must implement reset in such a way that it won’t break stop/cont, i.e., it must not reset its internal state. > > I do > think people care that stop/cont works, so it follows to me that no > stateful back-end will reset its internal state when receiving a > virtio/vhost reset. If they do, stop/cont breaks, which is a > user-visible bug that needs to be addressed – either properly by > implementing SUSPEND/RESUME in both qemu and the affected > back-ends, or > by using a similar work-around to virtiofsd, which is to ignore reset > commands. > > > Properly, please. You misunderstand me. I’m not presenting the choice I have. I’m presenting the choices existing implementations *have had until this point*. *None* chose to do it properly. I don’t know of stateful implementations besides virtiofsd, but virtiofsd chose to be content with not implementing reset and thus having things “just work”. The work-arounds must exist already. > > > It’s hard for me to imagine that people don’t care that stop/cont > breaks > some vhost-user back-end, but suddenly would start to care that > migration doesn’t work – especially given that first of all > someone will > need to manually implement any migration support in that back-end > even > with this series, which means that really, the only back-end we are > talking about here is our virtiofsd. To this day I’m not even > aware of > any other back-end that has internal state. > > > Another one I can think of is vhost-user-gpu. I sure hope stop/cont works for them. > Why did you give up on implementing SUSPEND/RESUME? Because I think it’s unnecessary for implementing migration, and migration is what’s on my priority list. None of these issues are new, they have always existed with stop/cont, work-arounds must be in place to make stop/cont work, and because I don’t see the difference in how stop/cont is used outside of migration and how it is used during migration, I assume those work-arounds must work for migration as well. Implementing virtio-fs migration is a chain of dependencies. We need at least the specification to be in qemu before we can start sending merge requests to the vhost Rust crates to implement support there. We need that support there before we can make the changes to virtiofsd. Adding SUSPEND/RESUME adds another hard dependency to the whole discussion (this would have to go on before this series), which has proven absolutely clearly in the past months that it is a very complex finnicky isue that would take a ton of time still. And I can’t justify that for myself, given that I don’t see any practically existing problem. PS: As far as I remember, vhost-user doesn’t even have a working reset today. vhost_dev_stop() calls vhost_reset_status(), which is a no-op unless the back-end supports SET_STATUS. The only back-end implementation we found (while discussing SUSPEND/RESUME) to support SET_STATUS was dpdk, but while it logs SET_STATUS 0 as a reset, it doesn’t do a reset, i.e. doesn’t call reset_device(), which it would do on RESET_OWNER. Hanna
On 26.09.23 22:10, Stefan Hajnoczi wrote: > Hi Hanna, > I was thinking about how this could work without SUSPEND/RESUME. What > do you think of the following? > > 1. The front-end sends VHOST_USER_RESET_DEVICE (or > VHOST_USER_RESET_OWNER, when necessary) when the guest driver resets > the device but not on vhost_dev_start()/vhost_dev_stop(). This is half the work of SUSPEND/RESUME. It isn’t easy to do. > 2. Suspend the device when all virtqueues are stopped via > VHOST_USER_GET_VRING_BASE. Resume the device after at least one > virtqueue is started and enabled. > 3. Ignore VHOST_USER_SET_STATUS. > > Reset would work. The device would suspend and resume without losing > state. Existing vhost-user backends already behave like this in > practice (they often don't implement RESET_DEVICE). I don’t understand the point, though. Today, reset in practice is a no-op anyway, precisely because we only send SET_STATUS 0, don’t fall back to RESET_OWNER/RESET_DEVICE, and no back-end implements SET_STATUS 0 as a reset. By sending RESET_* in case of a guest-initiated reset and nothing in case of stop/cont, we effectively don’t change anything about the latter (which is what SUSPEND/RESUME would be for), but only fix the former case. While I agree that it’s wrong that we don’t really reset the back-end in case of a guest-initiated reset, this is the first time in this whole discussion that that part has been presented as a problem that needs fixing now. So the proposal effectively changes nothing for the vhost_dev_stop()/start() case where we’d want to make use of SUSPEND/RESUME, but only for the case where we would not use it. Hanna
On Wed, Sep 27, 2023 at 10:32:14AM +0200, Hanna Czenczek wrote: > On 26.09.23 22:10, Stefan Hajnoczi wrote: > > Hi Hanna, > > I was thinking about how this could work without SUSPEND/RESUME. What > > do you think of the following? > > > > 1. The front-end sends VHOST_USER_RESET_DEVICE (or > > VHOST_USER_RESET_OWNER, when necessary) when the guest driver resets > > the device but not on vhost_dev_start()/vhost_dev_stop(). > > This is half the work of SUSPEND/RESUME. It isn’t easy to do. I sent a patch series to bring VHOST_USER_RESET_DEVICE to all vhost-user devices: https://lore.kernel.org/qemu-devel/20230927192737.528280-1-stefanha@redhat.com/T/#t > > > 2. Suspend the device when all virtqueues are stopped via > > VHOST_USER_GET_VRING_BASE. Resume the device after at least one > > virtqueue is started and enabled. > > 3. Ignore VHOST_USER_SET_STATUS. > > > > Reset would work. The device would suspend and resume without losing > > state. Existing vhost-user backends already behave like this in > > practice (they often don't implement RESET_DEVICE). > > I don’t understand the point, though. Today, reset in practice is a no-op > anyway, precisely because we only send SET_STATUS 0, don’t fall back to > RESET_OWNER/RESET_DEVICE, and no back-end implements SET_STATUS 0 as a > reset. By sending RESET_* in case of a guest-initiated reset and nothing in > case of stop/cont, we effectively don’t change anything about the latter > (which is what SUSPEND/RESUME would be for), but only fix the former case. > While I agree that it’s wrong that we don’t really reset the back-end in > case of a guest-initiated reset, this is the first time in this whole > discussion that that part has been presented as a problem that needs fixing > now. > > So the proposal effectively changes nothing for the vhost_dev_stop()/start() > case where we’d want to make use of SUSPEND/RESUME, but only for the case > where we would not use it. We discussed this on a call today. 2 & 3 are additions to the spec that Hanna has agreed to work on. Stefan