Done to avoid fixing ACPI route description of traditional PCI interrupts on q35
and because we should probably move with the times anyway.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-5-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
We are missing necessary config write handling for AER emulation in
the CXL root port. Add it based on pcie_root_port.c
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-4-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
PCIe r6.0 Figure 6-3 "Pseudo Logic Diagram for Selected Error Message Control
and Status Bits" includes a right hand branch under "All PCI Express devices"
that allows for messages to be generated or sent onwards without SERR#
being set as long as the appropriate per error class bit in the PCIe
Device Control Register is set.
Implement that branch thus enabling routing of ERR_COR, ERR_NONFATAL
and ERR_FATAL under OSes that set these bits appropriately (e.g. Linux)
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-3-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
This register in AER should be both writeable and should
have a default value with a couple of the errors masked
including the Uncorrectable Internal Error used by CXL for
it's error reporting.
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Message-Id: <20230302133709.30373-2-Jonathan.Cameron@huawei.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Fan Ni <fan.ni@samsung.com>
Provide pcihp specific callback to check if bus is hotpluggable
and consolidate its scattered hotplug criteria there.
While at it clean up no longer needed
qbus_set_hotplug_handler(BUS(bus), NULL)
workarounds since callback makes qbus_is_hotpluggable() return
correct answer even if hotplug_handler is set on bus.
PS:
see ("pci: fix 'hotplugglable' property behavior") for details
why callback was introduced.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-35-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
... instead of duplicating them in piix4 and lpc and then
trying to pass them to pcihp routines as arguments.
it simplifies call sites and places pcihp specific in
its own structure.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-34-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-33-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Generic PCI enumeration code doesn't really need access to
BSEL value, it is only used as means to decide if hotplug
enumerator should be called.
Use stateless object_property_find() to do that, and move
the rest of BSEL handling into build_append_pcihp_slots()
where it belongs.
This cleans up generic code a bit from hotplug stuff
and follow up patch will remove remaining call to
build_append_pcihp_slots() from generic code, making
it possible to use without ACPI PCI hotplug dependencies.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-32-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
previous commit ("pci: fix 'hotplugglable' property behavior") fixed
pcie root port's 'hotpluggable' property to behave consistently.
So we don't need a BSEL crutch anymore to see of device is not
hotpluggable, drop it from 'generic' PCI slots description handling.
BSEL is still used to decide if hotplug part should be called
but that will be moved out of generic code to hotplug one by
followup patches.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-31-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
acpi-index is now working with non-hotpluggable buses
(pci/q35 machine hostbridge), it can be used even if
ACPI PCI hotplug is disabled and as result acpi-index
uniqueness check will be omitted (since the check is
done by ACPI PCI hotplug handler, which isn't wired
when ACPI PCI hotplug is disabled).
Move check and related code to generic PCIDevice so it
would be independent of ACPI PCI hotplug.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-30-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-28-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-27-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
describing all present devices on functions other than
0 was complicated when non hotplug and hotplug code
was intermixed. So QEMU has been excluding non zero
functions since they are not supported by hotplug code,
then a condition to whitelist coldplugged bridges was
added and later whitelisting of devices that advertise
presence of their own AML description.
With non hotplug and hotplug code separated, it is
possible to relax rules and allow describing all
non-hotpluggble functions and hence simplify
conditions whether PCI device should be enumerated by
generic (non-hotplug) code.
Price of that simplification is an extra few Device()
descriptors in DSDT exposing built-in chipset functions,
which has no functional effect on guest side.
Apart from that, the enumeration of non zero functions,
allows to attach more NICs with acpi-index enabled
directly on hostbridge (if hotplug is not required).
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-25-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-24-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Inject static _DSM (EDSM) if non-hotpluggable device has
acpi-index configured on it.
It lets use acpi-index non-hotpluggable devices / devices
attached to non-hotpluggable bus.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-22-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-21-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-20-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
it's a helper method for acpi-index support on PCI buses
that do no support or have disabled ACPI PCI hotplug
or for non-hotpluggble endpoint devices.
(like non-hotpluggble NICs, integrated endpoints and
later for machines that do not support ACPI PCI hotplug)
no functional change, commit adds only EDSM method in DSDT
without any users. (the follow up patches will use it)
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-18-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-17-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
it will be reused by follow up patches that will implement
static _DSM for non-hotpluggable devices.
no functional AML change, only context one, where 'cap' (Local1)
initialization is moved after UUID/revision checks.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-15-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-14-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Currently the property may flip its state
during VM bring up or just doesn't work as
the name implies.
In particular with PCIE root port that has
'hotplug={on|off}' property, and when it's
turned off, one would expect
'hotpluggable' == false
for any devices attached to it.
Which is not the case since qbus_is_hotpluggable()
used by the property just checks for presence
of any hotplug_handler set on bus.
The problem is that name BusState::hotplug_handler
from its inception is misnomer, as it handles
not only hotplug but also in many cases coldplug
as well (i.e. generic wiring interface), and
it's fine to have hotplug_handler set on bus
while it doesn't support hotplug (ex. pcie-slot
with hotplug=off).
Another case of root port flipping 'hotpluggable'
state when ACPI PCI hotplug is enabled in this
case root port with 'hotplug=off' starts as
hotpluggable and then later on, pcihp
hotplug_handler clears hotplug_handler
explicitly after checking root port's 'hotplug'
property.
So root-port hotpluggablity check sort of works
if pcihp is enabled but is broken if pcihp is
disabled.
One way to deal with the issue is to ask
hotplug_handler if bus it controls is hotpluggable
or not. To do that add is_hotpluggable_bus()
hook to HotplugHandler interface and use it in
'hotpluggable' property + teach pcie-slot to
actually look into 'hotplug' property state
before deciding if bus is hotpluggable.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-13-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
commit [1] added ability to disable ACPI PCI hotplug
on hostbridge but forgot to take into account that it
should disable all ACPI hotplug machinery in case both
hostbridge and bridge hotplug are disabled.
Commit [2] tried to fix that, however it forgot to
remove hotplug_handler override which hands hotplug
control over to piix4 hotplug controller
(uninitialized after [2]).
As result at the time bridge is plugged in, its default
(SHPC) hotplug handler is replaced by piix4 one in
acpi_pcihp_device_plug_cb()
...
if (!s->legacy_piix &&
...
qbus_set_hotplug_handler(BUS(sec), OBJECT(hotplug_dev));
which is acting on uninitialized s->legacy_piix value
(0 by default) that was supposed to be initialized by
acpi_pcihp_init(), that is no longer called due to
following condition being false:
piix4_acpi_system_hot_add_init()
if (s->use_acpi_hotplug_bridge || s->use_acpi_root_pci_hotplug) {
and the bridge ends up with piix4 as hotplug handler
instead of shpc one.
Followup hotplug on that bridge as result yields
piix4 specific error:
Error: Unsupported bus. Bus doesn't have property 'acpi-pcihp-bsel' set
1) 3d7e78aa77 (Introduce a new flag for i440fx to disable PCI hotplug on the root bus)
2) df4008c9c5 (piix4: don't reserve hw resources when hotplug is off globally)
Fixes: df4008c9c5 (piix4: don't reserve hw resources when hotplug is off globally)
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-12-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(I practice [1] hasn't broke anything since on hardware side we unset
hotplug_handler on such intermediate port => hotplug behind it has
never worked)
When deciding if bridge should be described, the original
condition was
cold_plugged_bridge && pcihp_bridge_en
which was replaced [1] by
bridge has ACPI_PCIHP_PROP_BSEL
the later however is not the same thing as the original
and flips to false if intermediate bridge has hotplug
turned off (root-port with 'hotplug=off' option).
Since we already in build_pci_bridge_aml(), the question
if it's bridge is answered. Use DeviceState::hotplugged
to make decision if bridge should describe its slots.
What's left out is pcihp_bridge_en, which tells us if
ACPI bridge hotplug is enabled.
With hotplug and non hotplug part now being mostly
separated, omitting this check will only lead to
colplugged bridges describe occupied slots in case
when ACPI bridge hotplug is disabled.
Which makes behavior consistent with occupied slots
on hostbridge.
Ex (pc/DSDT.hpbrroot diff):
...
Device (S20)
{
Name (_ADR, 0x00040000) // _ADR: Address
+ Device (S08)
+ {
+ Name (_ADR, 0x00010000) // _ADR: Address
+ }
+
+ Device (S10)
+ {
+ Name (_ADR, 0x00020000) // _ADR: Address
+ }
}
...
PS:
testing shows that above doesn't affect adversely guest OS
behavior: i.e. if ACPI bridge hotplug is enabled it's
expected behaviour, and with ACPI bridge hotplug is disabled
(a.k. native hotplug), it doesn't break slot enumeration
nor native hotplug. (tested with RHEL9.0 and WS2022).
1)
Fixes: 6c36ec46b0 ("pcihp: make bridge describe itself using AcpiDevAmlIfClass:build_dev_aml")
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-10-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
follow up fix for missing root-port AML will affect these tests
by adding non-hotpluggable Device descriptors of colplugged
bridges when bridge hotplug is disabled.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-9-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Beside BSEL numbers change (due to 2 extra root-ports in q35/miltibridge test),
following change is expected:
Scope (\_SB.PCI0)
{
...
+ Scope (S50)
+ {
+ Scope (S00)
+ {
+ Method (PCNT, 0, NotSerialized)
+ {
+ BNUM = Zero
+ DVNT (PCIU, One)
+ DVNT (PCID, 0x03)
+ }
+ }
+
+ Method (PCNT, 0, NotSerialized)
+ {
+ ^S00.PCNT
+ }
+ }
...
Method (PCNT, 0, NotSerialized)
{
+ ^S50.PCNT ()
^S13.PCNT ()
^S12.PCNT ()
^S11.PCNT ()
I practice [1] hasn't broke anything since on hardware side we unset
hotplug_handler on such intermediate port => hotplug behind it has
not been properly wired and as result not worked.
1)
Fixes: ddab4d3fae ("pcihp: compose PCNT callchain right before its user _GPE._E01")
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-8-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Following corner case wasn't covered:
-device pcie-root-port,id=NO_HOTPLUG,hotplug=off
-device pcie-root-port,bus=NO_HOTPLUG
when intermediate root-port has explicitly disabled hotplug,
all hierarchy below it is not described anymore (used to be
described in 7.2)
So as result we see only NO_HOTPLUG root-port described
+ Device (S50)
+ {
+ Name (_ADR, 0x000A0000) // _ADR: Address
+ }
and no children nor notification chain for them are being composed.
Follow up patches will fix missing leaf root-port descriptor
and notification chain that should accompany it.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-7-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-6-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
expected changes:
Basically adds devices present on root bus in form:
Device (SXX)
{
Name (_ADR, 0xYYYYYYYY) // _ADR: Address
}
On top of that For q35.noacpihp, all ACPI PCI hotplug
AML is removed and _OSC get native hotplug enabled:
CreateDWordField (Arg3, 0x04, CDW2)
CreateDWordField (Arg3, 0x08, CDW3)
Local0 = CDW3 /* \_SB_.PCI0._OSC.CDW3 */
- Local0 &= 0x1E
+ Local0 &= 0x1F
If ((Arg1 != One))
{
CDW1 |= 0x08
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-5-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
test bridge AML generator with ACPI PCI hotplug disabled
(i.e. with native hotplug enabled/disabled per bridge/root port)
PS:
while at make sure that devices on pci-bridge are starting
from addr=1.0 as slot 0 is not available there and test
passes only because of a bug in ACPI hotplug that will be
fixed by follow up patch
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-4-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
for q35.noacpihp use plain default Q35 DSDT table as a starting point.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-3-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This reverts commit c471eb4f40.
which broke acpi tables test and rebuild due to skipping some tests
even thought none of devices tests depend on weren't disabled.
As result it leads to some expected tables not being updated,
merge conflicts and tests failure.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230302161543.286002-2-imammedo@redhat.com>
Acked-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
vhost-vdpa devices can return this feature now that blockers have been
set in case some features are not met.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-15-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Since some actions move to the start function instead of init, the
device features may not be the parent vdpa device's, but the one
returned by vhost backend. If transition to SVQ is supported, the vhost
backend will return _F_LOG_ALL to signal the device is migratable.
Add VHOST_F_LOG_ALL. HW dirty page tracking can be added on top of this
change if the device supports it in the future.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-14-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Next patches enable devices to be migrated even if vdpa netdev has not
been started with x-svq. However, not all devices are migratable, so we
need to block migration if we detect that.
Block migration if we detect the device expose a feature SVQ does not
know how to work with.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-13-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
A vdpa net device must initialize with SVQ in order to be migratable at
this moment, and initialization code verifies some conditions. If the
device is not initialized with the x-svq parameter, it will not expose
_F_LOG so the vhost subsystem will block VM migration from its
initialization.
Next patches change this, so we need to verify migration conditions
differently.
QEMU only supports a subset of net features in SVQ, and it cannot
migrate state that cannot track or restore in the destination. Add a
migration blocker if the device offers an unsupported feature.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-12-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Devices with CVQ need to migrate state beyond vq state. Leaving this to
future series.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-11-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Although it does not make a big difference, its more correct and
simplifies the cleanup path in subsequent patches.
Move ram_block_discard_disable(false) call to the top of
vhost_vdpa_cleanup because:
* We cannot use vhost_vdpa_first_dev after dev->opaque = NULL
assignment.
* Improve the stack order in cleanup: since it is the last action taken
in init, it should be the first at cleanup.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-10-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This allows net to restart the device backend to configure SVQ on it.
Ideally, these changes should not be net specific and they could be done
in:
* vhost_vdpa_set_features (with VHOST_F_LOG_ALL)
* vhost_vdpa_set_vring_addr (with .enable_log)
* vhost_vdpa_set_log_base.
However, the vdpa net backend is the one with enough knowledge to
configure everything because of some reasons:
* Queues might need to be shadowed or not depending on its kind (control
vs data).
* Queues need to share the same map translations (iova tree).
Also, there are other problems that may have solutions but complicates
the implementation at this stage:
* We're basically duplicating vhost_dev_start and vhost_dev_stop, and
they could go out of sync. If we want to reuse them, we need a way to
skip some function calls to avoid recursiveness (either vhost_ops ->
vhost_set_features, vhost_set_vring_addr, ...).
* We need to traverse all vhost_dev of a given net device twice: one to
stop and get the vq state and another one after the reset to
configure properties like address, fd, etc.
Because of that it is cleaner to restart the whole net backend and
configure again as expected, similar to how vhost-kernel moves between
userspace and passthrough.
If more kinds of devices need dynamic switching to SVQ we can:
* Create a callback struct like VhostOps and move most of the code
there. VhostOps cannot be reused since all vdpa backend share them,
and to personalize just for networking would be too heavy.
* Add a parent struct or link all the vhost_vdpa or vhost_dev structs so
we can traverse them.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-9-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The function vhost.c:vhost_dev_stop calls vhost operation
vhost_dev_start(false). In the case of vdpa it totally reset and wipes
the device, making the fetching of the vring base (virtqueue state) totally
useless.
The kernel backend does not use vhost_dev_start vhost op callback, but
vhost-user do. A patch to make vhost_user_dev_start more similar to vdpa
is desirable, but it can be added on top.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-8-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The function vhost.c:vhost_dev_stop fetches the vring base so the vq
state can be migrated to other devices. However, this is unreliable in
vdpa, since we didn't signal the device to suspend the queues, making
the value fetched useless.
Suspend the device if possible before fetching first and subsequent
vring bases.
Moreover, vdpa totally reset and wipes the device at the last device
before fetch its vrings base, making that operation useless in the last
device. This will be fixed in later patches of this series.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-7-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This allows vhost_vdpa to track if it is safe to get the vring base from
the device or not. If it is not, vhost can fall back to fetch idx from
the guest buffer again.
No functional change intended in this patch, later patches will use this
field.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Message-Id: <20230303172445.1089785-6-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
At this moment it is only possible to migrate to a vdpa device running
with x-svq=on. As a protective measure, the rewind of the inflight
descriptors was done at the destination. That way if the source sent a
virtqueue with inuse descriptors they are always discarded.
Since this series allows to migrate also to passthrough devices with no
SVQ, the right thing to do is to rewind at the source so the base of
vrings are correct.
Support for inflight descriptors may be added in the future.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-5-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This is needed for qemu to know it can suspend the device to retrieve
its status and enable SVQ with it, so all the process is transparent to
the guest.
Signed-off-by: Eugenio Pérez <eperezma@redhat.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Message-Id: <20230303172445.1089785-4-eperezma@redhat.com>
Tested-by: Lei Yang <leiyang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
