w/r/t: your [0/7] from initial series... As much as you don't want to keep living Groundhog Day - resolution of bugs like this are job security :-)... The subtleties of hostdev device Attach, Reattach, ReAttach, and Detach are such a tangle morass of shinola (ref: Steve Martin in "The Jerk")... w/r/t Suggestions for deamon restart issues... Seems like we need a way to save/restore the hostdev_mgr active/inactive lists using XML/JSON similar to how domains, storage, etc. handle it. Guess I just assumed that was already there ;-) since /var/run/libvirt/hostdevmgr exists. It seems that network stuff can be restored - virHostdevNetConfigRestore. Do you really think this series should be "held up" waiting to create some sort of status tracking? On 01/25/2016 11:20 AM, Andrea Bolognani wrote: Since I reviewed them all... I think the comment changes from 7/9 should just be inlined here and patch 4 instead of a separate patch Why not just reuse the function name you deleted? IOW: Is there a reason for "Only"? (not that I'm one that can complain about naming functions, but this just seems strange). Interesting ... In 2 instances, this will be a pointer to the "copy" element, while in the third instance this will be the "actual" on inactive list element. For a copy element, we'd *have* to search inactive; however, for an 'actual' we don't "need" to. This is tricky - the only time we care about the return status (now) is when skipUnmanaged == false, a/k/a the path where we pass the onlist element.. In my first pass through the changes I thought - oh no if we return -1, then this would be a path that could get that generic libvirt failed for some reason message; however, closer inspection noted that we guaranteed it was on the inactive list before calling here. <sigh>, unrelated and existing - virPCIDeviceGetManaged probably should return bool instead of unsigned int Existing, but if retries == 0, then cleanup never succeeded; however, perhaps one can assume that the subsequent call would fall over and play dead? Not that it really gets checked... Curious why the decision for a second loop - if activeDev matches, then it almost seems a shame to loop again. Why not (within if (activeDev): actual = virPCIDeviceListSteal(hostdev_mgr->activePCIHostdevs, activeDev); /* !actual should never happen, but better safe than sorry */ if (actual && virPCIDeviceListAdd(hostdev_mgr->inactivePCIHostdevs, actual) < 0) virPCIDeviceFree(actual); /* You could also... */ virPCIDeviceListDel(pcidevs, dev); } NOTE: Whether there is one or two loops, the second loop would need to call virPCIDeviceFree(actual) since we'd leak it otherwise. You'll also note I didn't keep the goto cleanup. Previously the code would completely go through the pcidevs Loop 4 regardless of whether virHostdevReattachPCIDevice code had failures. The new code has the subtle difference of jumping to cleanup if a failure is found. That could leave things in an awkward state especially since virHostdevReAttachPCIDevices is a void. Since failure for DeviceListAdd is because 1. device is already there (which I would *hope* isn't the case) or 2. memory allocation failure (in which case not much else successful will probably happen anyway), then perhaps continuing on rather than jumping to cleanup isn't a bad idea... We could be returning some memory that someone may find useful. My concerns about jumping to cleanup are that this API is called in error recovery paths as well as part of the ominous comment "For upgrade purpose:..." (comment before function header). So it seems the "existing logic" is try to clean up as many as possible. By potentially erroring out too soon could lead to more problems. So the question becomes what havoc is introduced if we cannot add to the inactive list but decide to continue as before... It seems we'll end up "failing" in virHostdevOnlyReattachPCIDevice since it's not in the inactiveList, but our Loop 4 logic doesn't care. Of course we could delete 'dev' from 'pcidevs' too before then... Hopefully this makes sense... It's been an 'edit in process'... This was a curious placement for *ListDel... If !activeDev, then calling *ListDel also won't find 'dev' on activelist... If the choice is to use two loops (and perhaps keep the cleanups)... 1. If this Steal fails, then something is seriously wrong, but we don't even give it a VIR_DEBUG message. 2. If this Steal fails, then going to cleanup is again a subtle difference with the prior logic that said, well I couldn't do anything with this, but I'm going to keep processing. 3. If we keep processing, then something on 'pcidevs' won't be in 'inactivePCIHostdevs', causing virHostdevOnlyReattachPCIDevice to fail. But that does not matter since we ignore the return value in Loop 4. 4. If we do Steal and if the subsequent Add fails, then we leak 'actual', so prior to the goto cleanup call virPCIDeviceFree(actual); (or instead if the goto cleanup;'s are removed). John

On 01/27/2016 12:26 PM, Andrea Bolognani wrote: I assume that restart is already broken... This series does fix some issues in the "normal" flow though. Perhaps a chicken and egg type problem. If you fix restart first, what of this series would be beneficial to ensure restart doesn't have issues... Right - I started looking at v2 virHostdevReattachPCIDeviceVeryCarefully :-) But since it's in the comment of the code: virHostdevReattachPCIDeviceToHost I kept losing focus on when something was on the inactive list or not. Then of course trying to reconcile 'pci' and 'dev' variable name usage. I think so - a lot of typing out loud which at least gives you my review context... I finally convinced myself there wasn't an issue even though it's strange to return something and in the end, no one really cares... Perhaps using a 4th parameter 'actual' (could be NULL) would make it more obvious that we knew on input that the device was already found on the inactive list. Determining whether we have a copy or an actual wasn't readily apparent. Consider some future "user" of this function - how easy is it to decide what to pass? I saw that today... That's fine - it was just another of those typing out loud moments. Your call - we go through the pcidevs list many times so it's not that big a deal... Yes, the error path - if you fail to add actual to inactive, then it was dropped on the floor Understood, but if !activeDev, then the virPCIDeviceListDel calls virPCIDeviceListSteal which calls virPCIDeviceListStealIndex using virPCIDeviceListFindIndex... You'll note activeList is sourced by calling virPCIDeviceListFind which calls virPCIDeviceListFindIndex So I was pointing out how pointless it would be to call virPCIDeviceListDel if activeDev == NULL (because it too would do nothing). Try to keep it under 100 patches please ;-) John

Need something here... Perhaps to the effect of instead of inlining most of virHostdevOnlyReattachPCIDevice, call virHostdevOnlyReattachPCIDevice. One extra "call" or check being made by doing this is the virPCIDeviceGetStubDriver check and virPCIDeviceWaitForCleanup. Are there any "timing" considerations from this path? Now for every PCI hostdev found, we'll call *WaitForCleanup. That should perhaps be noted. This looks reasonable otherwise. John On 01/25/2016 11:20 AM, Andrea Bolognani wrote:

This ensures the behavior for reattach is consistent, no matter how it was triggered (eg. 'virsh nodedev-reattach', 'virsh detach-device' or shutdown of a domain that is configured to use hostdevs). --- src/util/virhostdev.c | 33 ++++++++++++++++++++++++++++++--- 1 file changed, 30 insertions(+), 3 deletions(-) diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c index 586937e..f40d636 100644 --- a/src/util/virhostdev.c +++ b/src/util/virhostdev.c @@ -1637,10 +1637,24 @@ virHostdevPCINodeDeviceDetach(virHostdevManagerPtr hostdev_mgr, return ret; } +/** + * virHostdevPCINodeDeviceReAttach: + * @hostdev_mgr: hostdev manager + * @pci: PCI device + * + * Reattach a specific PCI device to the host. + * + * This function makes sure the device is not in use before reattaching it + * to the host; if the device has already been reattached to the host, the + * operation is considered successful. + * + * Returns: 0 on success, <0 on failure + */ int virHostdevPCINodeDeviceReAttach(virHostdevManagerPtr hostdev_mgr, virPCIDevicePtr pci) { + virPCIDevicePtr actual; struct virHostdevIsPCINodeDeviceUsedData data = {hostdev_mgr, NULL, false}; int ret = -1; @@ -1651,10 +1665,23 @@ virHostdevPCINodeDeviceReAttach(virHostdevManagerPtr hostdev_mgr, if (virHostdevIsPCINodeDeviceUsed(virPCIDeviceGetAddress(pci), &data)) goto out; - virPCIDeviceReattachInit(pci); + /* We want this function to be idempotent, so if the device has already + * been removed from the inactive list (and is not active either, as + * per the check above) just return right away. We also need to retrieve + * the actual device from the inactive list because that's the one that + * contains state information such as the stub driver */ + if (!(actual = virPCIDeviceListFind(hostdev_mgr->inactivePCIHostdevs, + pci))) { + VIR_DEBUG("PCI device %s is already attached to the host", + virPCIDeviceGetName(pci)); + ret = 0; + goto out; + } - if (virPCIDeviceReattach(pci, hostdev_mgr->activePCIHostdevs, - hostdev_mgr->inactivePCIHostdevs) < 0) + /* Reattach the device. We don't want to skip unmanaged devices in + * this case, because the user explicitly asked for the device to + * be reattached to the host driver */ + if (virHostdevOnlyReattachPCIDevice(hostdev_mgr, actual, false) < 0) goto out; ret = 0; -- 2.5.0

This function mirrors virHostdevOnlyDetachPCIDevice(). The handling of active and inactive devices is updated and made more explicit, which means virHostdevPreparePCIDevices() has to be updated as well. --- src/util/virhostdev.c | 87 ++++++++++++++++++++++++++++++++++++--------------- 1 file changed, 61 insertions(+), 26 deletions(-) diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c index f40d636..6f14574 100644 --- a/src/util/virhostdev.c +++ b/src/util/virhostdev.c @@ -594,6 +594,56 @@ virHostdevOnlyReattachPCIDevice(virHostdevManagerPtr mgr, return ret; } +/** + * virHostdevOnlyDetachPCIDevice: + * @mgr: hostdev manager + * @pci: PCI device to be detached + * @skipUnmanaged: whether to skip unmanaged devices + * + * Detach a PCI device from the host. + * + * This function only performs the base detach steps that are required + * regardless of whether the device is being attached to a domain or + * is simply being detached from the host for later use. + * + * The PCI related parts of @mgr (inactivePCIHostdevs, activePCIHostdevs) + * must have been locked beforehand using virObjectLock(). + * + * Returns: 0 on success, <0 on failure + */ +static int +virHostdevOnlyDetachPCIDevice(virHostdevManagerPtr mgr, + virPCIDevicePtr pci, + bool skipUnmanaged) +{ + int ret = -1; + + /* Skip unmanaged devices if asked to do so */ + if (!virPCIDeviceGetManaged(pci) && skipUnmanaged) { + VIR_DEBUG("Not detaching unmanaged PCI device %s", + virPCIDeviceGetName(pci)); + ret = 0; + goto out; + } + + VIR_DEBUG("Detaching PCI device %s", virPCIDeviceGetName(pci)); + if (virPCIDeviceDetach(pci, + mgr->activePCIHostdevs, + mgr->inactivePCIHostdevs) < 0) { + virErrorPtr err = virGetLastError(); + VIR_ERROR(_("Failed to detach PCI device %s: %s"), + virPCIDeviceGetName(pci), + err ? err->message : _("unknown error")); + virResetError(err); + goto out; + } + + ret = 0; + + out: + return ret; +} + int virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, const char *drv_name, @@ -664,17 +714,8 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); - if (virPCIDeviceGetManaged(dev)) { - VIR_DEBUG("Detaching managed PCI device %s", - virPCIDeviceGetName(dev)); - if (virPCIDeviceDetach(dev, - hostdev_mgr->activePCIHostdevs, - hostdev_mgr->inactivePCIHostdevs) < 0) - goto reattachdevs; - } else { - VIR_DEBUG("Not detaching unmanaged PCI device %s", - virPCIDeviceGetName(dev)); - } + if (virHostdevOnlyDetachPCIDevice(hostdev_mgr, dev, true) < 0) + goto reattachdevs; } /* At this point, all devices are attached to the stub driver and have @@ -704,23 +745,21 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, last_processed_hostdev_vf = i; } - /* Loop 5: Now mark all the devices as active */ + /* Step 5: move all devices from the inactive list to the active list */ for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); + virPCIDevicePtr actual; - VIR_DEBUG("Adding PCI device %s to active list", + VIR_DEBUG("Removing PCI device %s from inactive list", virPCIDeviceGetName(dev)); - if (virPCIDeviceListAdd(hostdev_mgr->activePCIHostdevs, dev) < 0) + if (!(actual = virPCIDeviceListSteal(hostdev_mgr->inactivePCIHostdevs, + dev))) goto inactivedevs; - } - - /* Loop 6: Now remove the devices from inactive list. */ - for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { - virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); - VIR_DEBUG("Removing PCI device %s from inactive list", - virPCIDeviceGetName(dev)); - virPCIDeviceListDel(hostdev_mgr->inactivePCIHostdevs, dev); + VIR_DEBUG("Adding PCI device %s to active list", + virPCIDeviceGetName(actual)); + if (virPCIDeviceListAdd(hostdev_mgr->activePCIHostdevs, actual) < 0) + goto inactivedevs; } /* Loop 7: Now set the used_by_domain of the device in @@ -771,10 +810,6 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, virPCIDeviceFree(dev); } - /* Loop 9: Now steal all the devices from pcidevs */ - while (virPCIDeviceListCount(pcidevs) > 0) - virPCIDeviceListStealIndex(pcidevs, 0); - ret = 0; goto cleanup; -- 2.5.0

On 01/25/2016 11:21 AM, Andrea Bolognani wrote: A new PCI device to be detach from the host IOW: Not currently on some active/inactive list, right? Seems reasonable otherwise though. John

Some of the comments are no longer accurate after the recent changes, others have been expanded and hopefully improved. The initial summary of the operations has been removed so that two separate edits are not needed when making changes. --- src/util/virhostdev.c | 68 +++++++++++++++++++++++---------------------------- 1 file changed, 30 insertions(+), 38 deletions(-) diff --git a/src/util/virhostdev.c b/src/util/virhostdev.c index ab17547..0892dbf 100644 --- a/src/util/virhostdev.c +++ b/src/util/virhostdev.c @@ -675,11 +675,10 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, * must be reset before being marked as active. */ - /* Loop 1: validate that non-managed device isn't in use, eg - * by checking that device is either un-bound, or bound - * to pci-stub.ko - */ + /* Detaching devices from the host involves several steps; each of them + * is described at length below */ + /* Step 1: perform safety checks, eg. ensure the devices are assignable */ for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); bool strict_acs_check = !!(flags & VIR_HOSTDEV_STRICT_ACS_CHECK); @@ -710,7 +709,8 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, } } - /* Loop 2: detach managed devices (i.e. bind to appropriate stub driver) */ + /* Step 2: detach managed devices (i.e. bind to appropriate stub driver). + * detachPCIDevices() will also mark devices as inactive */ for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); @@ -718,11 +718,10 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, goto reattachdevs; } - /* At this point, all devices are attached to the stub driver and have + /* At this point, devices are attached to the stub driver and have * been marked as inactive */ - /* Loop 3: Now that all the PCI hostdevs have been detached, we - * can safely reset them */ + /* Step 3: perform a PCI reset on all devices */ for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); @@ -732,8 +731,7 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, goto reattachdevs; } - /* Loop 4: For SRIOV network devices, Now that we have detached the - * the network device, set the netdev config */ + /* Step 4: set the netdev config for SRIOV network devices */ for (i = 0; i < nhostdevs; i++) { virDomainHostdevDefPtr hostdev = hostdevs[i]; if (!virHostdevIsPCINetDevice(hostdev)) @@ -762,9 +760,7 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, goto inactivedevs; } - /* Loop 7: Now set the used_by_domain of the device in - * activePCIHostdevs as domain name. - */ + /* Step 6: set driver and domain information */ for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev, activeDev; @@ -777,7 +773,7 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, virPCIDeviceSetUsedBy(activeDev, drv_name, dom_name); } - /* Loop 8: Now set the original states for hostdev def */ + /* Step 7: set the original states for hostdev def */ for (i = 0; i < nhostdevs; i++) { virPCIDevicePtr dev; virPCIDevicePtr pcidev; @@ -792,10 +788,9 @@ virHostdevPreparePCIDevices(virHostdevManagerPtr hostdev_mgr, dev = virPCIDeviceNew(pcisrc->addr.domain, pcisrc->addr.bus, pcisrc->addr.slot, pcisrc->addr.function); - /* original states "unbind_from_stub", "remove_slot", - * "reprobe" were already set by pciDettachDevice in - * loop 2. - */ + /* original states for "unbind_from_stub", "remove_slot" and + * "reprobe" (used when reattaching) were already set by + * detachPCIDevices() in a previous step */ VIR_DEBUG("Saving network configuration of PCI device %s", virPCIDeviceGetName(dev)); if ((pcidev = virPCIDeviceListFind(pcidevs, dev))) { @@ -875,16 +870,11 @@ virHostdevReAttachPCIDevices(virHostdevManagerPtr hostdev_mgr, goto cleanup; } - /* Loop through the assigned devices 4 times: 1) delete them all from - * activePCIHostdevs, 2) restore network config of SRIOV netdevs, 3) Do a - * PCI reset on each device, 4) reattach the devices to their host drivers - * (managed) or add them to inactivePCIHostdevs (!managed). - */ + /* Reattaching devices to the host involves several steps; each of them + * is described at length below */ - /* - * Loop 1: verify that each device in the hostdevs list really was in use - * by this domain, and remove them all from the activePCIHostdevs list. - */ + /* Step 1: verify that each device in the hostdevs list really was in use + * by this domain */ i = 0; while (i < virPCIDeviceListCount(pcidevs)) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); @@ -922,14 +912,11 @@ virHostdevReAttachPCIDevices(virHostdevManagerPtr hostdev_mgr, goto cleanup; } - /* At this point, any device that had been used by the guest is in - * pcidevs, but has been removed from activePCIHostdevs. - */ + /* At this point, devices are no longer marked as active and have been + * marked as inactive instead */ - /* - * Loop 2: restore original network config of hostdevs that used - * <interface type='hostdev'> - */ + /* Step 3: restore original network config of hostdevs that used + * <interface type='hostdev'> */ for (i = 0; i < nhostdevs; i++) { virDomainHostdevDefPtr hostdev = hostdevs[i]; @@ -950,7 +937,7 @@ virHostdevReAttachPCIDevices(virHostdevManagerPtr hostdev_mgr, } } - /* Loop 3: perform a PCI Reset on all devices */ + /* Step 4: perform a PCI Reset on all devices */ for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); @@ -964,15 +951,20 @@ virHostdevReAttachPCIDevices(virHostdevManagerPtr hostdev_mgr, } } - /* Loop 4: reattach devices to their host drivers (if managed) or place - * them on the inactive list (if not managed) - */ + /* Step 5: reattach managed devices to their host drivers; unmanaged + * devices need no further processing. reattachPCIDevices() will + * remove devices from the inactive list as they are reattached + * to the host */ for (i = 0; i < virPCIDeviceListCount(pcidevs); i++) { virPCIDevicePtr dev = virPCIDeviceListGet(pcidevs, i); ignore_value(virHostdevOnlyReattachPCIDevice(hostdev_mgr, dev, true)); } + /* At this point, all devices are either marked as inactive (if they + * were unmanaged), or have been reattached to the host driver (if they + * were managed) and are no longer tracked by any of our device lists */ + cleanup: virObjectUnref(pcidevs); virObjectUnlock(hostdev_mgr->activePCIHostdevs); -- 2.5.0

On Tue, 2016-01-26 at 18:59 -0500, John Ferlan wrote: That comment is still there by mistake: the comment was supposed to replace it, the idea being that if we list and describe briefly the steps in a comment, as we currently do, we have to keep both the actual comments and the summary in sync with no real gain. Missed when renaming the function between v1 and v2 :) Okay. Same as above :) Well, the comment was intended to refer to the situation after Step 5 has been performed, implying that no error that could have caused us to jump straight to cleanup had occurred. Maybe that can be made more obvious someway, I'll think about it. Cheers. --  Andrea Bolognani Software Engineer - Virtualization Team

Unbinding a PCI device from the stub driver can require several steps, and it can be useful for debugging to be able to trace which of these steps are performed and which are skipped for each device. --- src/util/virpci.c | 22 ++++++++++++++++++---- 1 file changed, 18 insertions(+), 4 deletions(-) diff --git a/src/util/virpci.c b/src/util/virpci.c index f56dbe6..51a3f88 100644 --- a/src/util/virpci.c +++ b/src/util/virpci.c @@ -1106,26 +1106,37 @@ virPCIDeviceUnbindFromStub(virPCIDevicePtr dev) if (!driver) { /* The device is not bound to any driver and we are almost done. */ + VIR_DEBUG("PCI device %s is not bound to any driver", dev->name); goto reprobe; } - if (!dev->unbind_from_stub) + if (!dev->unbind_from_stub) { + VIR_DEBUG("Unbind from stub skipped for PCI device %s", dev->name); goto remove_slot; + } /* If the device isn't bound to a known stub, skip the unbind. */ if (virPCIStubDriverTypeFromString(driver) < 0 || - virPCIStubDriverTypeFromString(driver) == VIR_PCI_STUB_DRIVER_NONE) + virPCIStubDriverTypeFromString(driver) == VIR_PCI_STUB_DRIVER_NONE) { + VIR_DEBUG("Unbind from stub skipped for PCI device %s because of " + "unknown stub driver", dev->name); goto remove_slot; + } - VIR_DEBUG("Found stub driver %s", driver); + VIR_DEBUG("Found stub driver %s for PCI device %s", driver, dev->name); + VIR_DEBUG("Unbinding PCI device %s", dev->name); if (virPCIDeviceUnbind(dev) < 0) goto cleanup; dev->unbind_from_stub = false; remove_slot: - if (!dev->remove_slot) + if (!dev->remove_slot) { + VIR_DEBUG("Slot removal skipped for PCI device %s", dev->name); goto reprobe; + } + + VIR_DEBUG("Removing slot for PCI device %s", dev->name); /* Xen's pciback.ko wants you to use remove_slot on the specific device */ if (!(path = virPCIDriverFile(driver, "remove_slot"))) @@ -1141,10 +1152,13 @@ virPCIDeviceUnbindFromStub(virPCIDevicePtr dev) reprobe: if (!dev->reprobe) { + VIR_DEBUG("Reprobe skipped for PCI device %s", dev->name); result = 0; goto cleanup; } + VIR_DEBUG("Reprobing for PCI device %s", dev->name); + /* Trigger a re-probe of the device is not in the stub's dynamic * ID table. If the stub is available, but 'remove_id' isn't * available, then re-probing would just cause the device to be -- 2.5.0