PHB hotplug re-uses PHB device tree generation code and passes
it to a guest via RTAS. Doing this requires knowledge of where
exactly in the device tree the node describing the PHB begins.
Provide this via a new optional pointer that can be used to
store the PHB node's start offset.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <155059671912.1466090.10891589403973703473.stgit@bahia.lab.toulouse-stg.fr.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
To support PHB hotplug we need to clean up lingering references,
memory, child properties, etc. prior to the PHB object being
finalized. Generally this will be called as a result of calling
object_unparent() on the PHB object, which in turn would normally
be called as the result of an unplug() operation.
When the PHB is finalized, child objects will be unparented in
turn, and finalized if the PHB was the only reference holder. so
we don't bother to explicitly unparent child objects of the PHB,
with the notable exception of DRCs. This is needed to avoid a QEMU
crash when unplugging a PHB and resetting the machine before the
guest could handle the event. The DRCs are removed from the QOM tree
by pci_unregister_root_bus() and we must make sure we're not leaving
stale aliases under the global /dr-connector path.
The formula that gives the number of DMA windows is moved to an
inline function in the hw/pci-host/spapr.h header because it
will have other users.
The unrealize function is able to cope with partially realized PHBs.
It is hence used to implement proper rollback on the realize error
path.
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Message-Id: <155059669881.1466090.13515030705986041517.stgit@bahia.lab.toulouse-stg.fr.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <155059667346.1466090.326696113231137772.stgit@bahia.lab.toulouse-stg.fr.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
All this code is used with both the XICS and XIVE interrupt controllers.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Some files claim that the code is licensed under the GPL, but then
suddenly suggest that the user should have a look at the LGPL.
That's of course non-sense, replace it with the correct GPL wording
instead.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <1548255083-8190-1-git-send-email-thuth@redhat.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
The macro is only used in one place, where the purpose of the
value is obvious. Eliminate the macro so we don't need to rely
on stringify().
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20190107193020.21744-2-ehabkost@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
PHB hotplug will bring more users for it. Let's define it along with
the PHB defines from which it is derived for simplicity.
While here fix a misleading comment about manual placement, which was
abandoned with 30b3bc5aa9.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This function is only used when creating the default PHB. Let's rename
it and move it to the core machine code for clarity.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
So that we don't have to call qdev_get_machine() to get the machine
class and the sPAPRIrq backend holding the number of MSIs.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
>From observation of various OS sources it can be seen that the token register
introduced in 4e46dcdbd3 "PPC: Newworld: Add uninorth token register" is not
required, since the only register currently implemented is the uninorth hardware
version which is read-only.
Remove the token register implementation and instead return the uninorth
version corresponding to the hardware.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Commit 4e46dcdbd3 "PPC: Newworld: Add uninorth token register" added a TODO
which was to convert the uninorth registers hack to a proper device. Move
these registers to a new uninorth device, removing the old hacks from
mac_newworld.c.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The existing UNINState actually represents the PCI/AGP host bridge stage so
rename it accordingly.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Do this for both the uninorth main and uninorth u3 AGP buses, using the main
PCI bus for each machine (this ensures the IO addresses still match those
used by OpenBIOS).
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Now that the OpenPIC is wired up via the board, we can now remove our temporary
PIC qdev pointer property and replace it with an object link instead.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This is in preparation for moving the PCI bus wiring inside the uninorth
host bridge devices. In the future it will be possible to remove this once the
PICs have been switched to use qdev GPIOs.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
[dwg: Added hw/hw.h #include as suggested by Philippe Mathieu-Daudé]
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Add code needed to get a functional PCI subsytem when using in
conjunction with upstream Linux guest (4.13+). Tested to work against
"e1000e" (network adapter, using MSI interrupts) as well as
"usb-ehci" (USB controller, using legacy PCI interrupts).
Based on "i.MX6 Applications Processor Reference Manual" (Document
Number: IMX6DQRM Rev. 4) as well as corresponding dirver in Linux
kernel (circa 4.13 - 4.16 found in drivers/pci/dwc/*)
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This is the final stage in correcting the naming convention with respect to
sabre, APB and PBM. It is effectively a file rename from apb.c to sabre.c
along with touching up a few constants to remove the remaining references
to APB.
Note that as part of the rename process the configuration variable
CONFIG_PCI_APB is changed to CONFIG_PCI_SABRE.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: Artyom Tarasenko <atar4qemu@gmail.com>
Similarly rename the corresponding APBState typedef to SabreState.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: Artyom Tarasenko <atar4qemu@gmail.com>
As hinted in the comment at the top of the file, the naming convention for the
APB types/QOM functions isn't correct. As a starting point we can at least
rename the APB type and related functions to improve the readability of apb.c.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: Artyom Tarasenko <atar4qemu@gmail.com>
Move the QOM type and macros into a new include/hw/pci-bridge/simba.h
file, and add a new CONFIG_SIMBA Makefile.objs variable which is enabled
for sparc64-softmmu builds only.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
CC: Michael S. Tsirkin <mst@redhat.com>
CC: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: Artyom Tarasenko <atar4qemu@gmail.com>
By separating the sun4u IOMMU device into new sun4u_iommu.c and sun4m_iommu.h
files we noticeably simplify apb.c whilst bringing sun4u in line with all the
other IOMMU-supporting architectures.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Acked-by: Artyom Tarasenko <atar4qemu@gmail.com>
This is in preparation to split the IOMMU device out of the APB. As part of
this commit we also enforce separation of the IOMMU and APB devices by using
a QOM object link to pass the IOMMU reference and accessing the IOMMU registers
via a separate memory region mapped into the APB config space rather than
directly.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Acked-by: Artyom Tarasenko <atar4qemu@gmail.com>
Following on from the previous commit, we can also do the same with
with legacy OBIO interrupts in pci_pbmA_map_irq().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
This enables us to remove the static array mapping in the ISA IRQ
handler (and the embedded reference to the APB device) by formalising
the interrupt wiring via the qdev GPIO API.
For more clarity we replace the APB OBIO interrupt numbers with constants
designating the interrupt source, and rename isa_irq_handler() to
ebus_isa_irq_handler().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Since the previous commit the only remaining use of the qdev busA property is
to configure the PCI bridge in front of the onboard ebus devices differently
to allow early OpenBIOS serial console access.
Instead we can now manually update the PCI configuration for bridge A in
pci_pbm_reset() and thus completely remove the busA property from the
PBMPCIBridge state.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
By making the special_base and mem_base values qdev properties, we can move
the remaining parts of pci_apb_init() into the pbm init() and realize()
functions.
This finally allows us to instantiate the APB directly using standard qdev
create/init functions in sun4u.c.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
This enables us to remove these parameters from pci_apb_init().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
This is a first step towards removing pci_apb_init() completely.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
This is initialisation that should really take place in the ebus realize
function. As part of this we also rework the ebus IRQ mapping so that
instead of having to pass in the array of pbm_irqs, we obtain a reference
to them by looking up the APB device during ebus realize.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
This also includes the related IOMMUState typedef and defines.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Artyom Tarasenko <atar4qemu@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
xics_get_qirq() is only used by the sPAPR machine. Let's move it there
and change its name to reflect its scope. It will be useful for XIVE
support which will use its own set of qirqs.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
include/hw/pci/pci_bus.h contains several data structures related to PCI
bridges that aren't needed by most users of pci_bus.h. We already have
a pci_bridge.h, so move them there.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Currently there is no MMIO range over 4G
reserved for PCI hotplug. Since the 32bit PCI hole
depends on the number of cold-plugged PCI devices
and other factors, it is very possible is too small
to hotplug PCI devices with large BARs.
Fix it by reserving 2G for I4400FX chipset
in order to comply with older Win32 Guest OSes
and 32G for Q35 chipset.
Even if the new defaults of pci-hole64-size will appear in
"info qtree" also for older machines, the property was
not implemented so no changes will be visible to guests.
Note this is a regression since prev QEMU versions had
some range reserved for 64bit PCI hotplug.
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
To implement INTx to gsi routing we need to pass the gpex host
bridge the gsi associated to each INTx index. Let's introduce
irq_num array and gpex_set_irq_num setter function.
Signed-off-by: Pranavkumar Sawargaonkar <pranavkumar@linaro.org>
Signed-off-by: Tushar Jagad <tushar.jagad@linaro.org>
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Feng Kan <fkan@apm.com>
Reviewed-by: Andrew Jones <drjones@redhat.com>
Message-id: 1505296004-6798-2-git-send-email-eric.auger@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The q35 machine type currently lets the guest firmware select a 1MB, 2MB
or 8MB TSEG (basically, SMRAM) size. In edk2/OVMF, we use 8MB, but even
that is not enough when a lot of VCPUs (more than approx. 224) are
configured -- SMRAM footprint scales largely proportionally with VCPU
count.
Introduce a new property for "mch" called "extended-tseg-mbytes", which
expresses (in megabytes) the user's choice of TSEG (SMRAM) size.
Invent a new, QEMU-specific register in the config space of the DRAM
Controller, at offset 0x50, in order to allow guest firmware to query the
TSEG (SMRAM) size.
According to Intel Document Number 316966-002, Table 5-1 "DRAM Controller
Register Address Map (D0:F0)":
Warning: Address locations that are not listed are considered Intel
Reserved registers locations. Reads to Reserved registers may
return non-zero values. Writes to reserved locations may
cause system failures.
All registers that are defined in the PCI 2.3 specification,
but are not necessary or implemented in this component are
simply not included in this document. The
reserved/unimplemented space in the PCI configuration header
space is not documented as such in this summary.
Offsets 0x50 and 0x51 are not listed in Table 5-1. They are also not part
of the standard PCI config space header. And they precede the capability
list as well, which starts at 0xe0 for this device.
When the guest writes value 0xffff to this register, the value that can be
read back is that of "mch.extended-tseg-mbytes" -- unless it remains
0xffff. The guest is required to write 0xffff first (as opposed to a
read-only register) because PCI config space is generally not cleared on
QEMU reset, and after S3 resume or reboot, new guest firmware running on
old QEMU could read a guest OS-injected value from this register.
After reading the available "extended" TSEG size, the guest firmware may
actually request that TSEG size by writing pattern 11b to the ESMRAMC
register's TSEG_SZ bit-field. (The Intel spec referenced above defines
only patterns 00b (1MB), 01b (2MB) and 10b (8MB); 11b is reserved.)
On the QEMU command line, the value can be set with
-global mch.extended-tseg-mbytes=N
The default value for 2.10+ q35 machine types is 16. The value is limited
to 0xfff (4095) at the moment, purely so that the product (4095 MB) can be
stored to the uint32_t variable "tseg_size" in mch_update_smram(). Users
are responsible for choosing sensible TSEG sizes.
On 2.9 and earlier q35 machine types, the default value is 0. This lets
the 11b bit pattern in ESMRAMC.TSEG_SZ, and the register at offset 0x50,
keep their original behavior.
When "extended-tseg-mbytes" is nonzero, the new register at offset 0x50 is
set to that value on reset, for completeness.
PCI config space is migrated automatically, so no VMSD changes are
necessary.
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Ref: https://bugzilla.redhat.com/show_bug.cgi?id=1447027
Ref: https://lists.01.org/pipermail/edk2-devel/2017-May/010456.html
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The pointer drc->detach_cb is being used as a way of informing
the detach() function inside spapr_drc.c which cb to execute. This
information can also be retrieved simply by checking drc->type and
choosing the right callback based on it. In this context, detach_cb
is redundant information that must be managed.
After the previous spapr_lmb_release change, no detach_cb_opaques
are being used by any of the three callbacks functions. This is
yet another information that is now unused and, on top of that, can't
be migrated either.
This patch makes the following changes:
- removal of detach_cb_opaque. the 'opaque' argument was removed from
the callbacks and from the detach() function of sPAPRConnectorClass. The
attribute detach_cb_opaque of sPAPRConnector was removed.
- removal of detach_cb from the detach() call. The function pointer
detach_cb of sPAPRConnector was removed. detach() now uses a
switch(drc->type) to execute the apropriate callback. To achieve this,
spapr_core_release, spapr_lmb_release and spapr_phb_remove_pci_device_cb
callbacks were made public to be visible inside detach().
Signed-off-by: Daniel Henrique Barboza <danielhb@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
bb9986452 "spapr_pci: Advertise access to PCIe extended config space"
allowed guests to access the extended config space of PCI Express devices
via the PAPR interfaces, even though the paravirtualized bus mostly acts
like plain PCI.
However, that patch enabled access unconditionally, including for existing
machine types, which is an unwise change in behaviour. This patch limits
the change to pseries-2.9 (and later) machine types.
Suggested-by: Andrea Bolognani <abologna@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Add support for emulating the Xilinx AXI Root Port Bridge for PCI
Express as described by Xilinx' PG055 document. This is a PCIe
controller that can be used with certain series of Xilinx FPGAs, and is
used on the MIPS Boston board which will make use of this code.
Signed-off-by: Paul Burton <paul.burton@imgtec.com>
[yongbok.kim@imgtec.com:
removed returning on !level,
updated IRQ connection with GPIO logic,
moved xilinx_pcie_init() to boston.c
replaced stw_le_p() with pci_set_word()
and other cosmetic changes]
Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Add also a missing parenthesis in a comment.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Acked-by: Alistair Francis <alistair.francis@xilinx.com>
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
daa2369 "spapr_pci: Add a 64-bit MMIO window" subtly broke migration
from qemu-2.7 to the current version. It split the device's MMIO
window into two pieces for 32-bit and 64-bit MMIO.
The patch included backwards compatibility code to convert the old
property into the new format. However, the property value was also
transferred in the migration stream and compared with a (probably
unwise) VMSTATE_EQUAL. So, the "raw" value from 2.7 is compared to
the new style converted value from (pre-)2.8 giving a mismatch and
migration failure.
Along with the actual field that caused the breakage, there are
several other ill-advised VMSTATE_EQUAL()s. To fix forwards
migration, we read the values in the stream into scratch variables and
ignore them, instead of comparing for equality. To fix backwards
migration, we populate those scratch variables in pre_save() with
adjusted values to match the old behaviour.
To permit the eventual possibility of removing this cruft from the
stream, we only include these compatibility fields if a new
'pre-2.8-migration' property is set. We clear it on the pseries-2.8
machine type, which obviously can't be migrated backwards, but set it
on earlier machine type versions.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Currently, the MMIO space for accessing PCI on pseries guests begins at
1 TiB in guest address space. Each PCI host bridge (PHB) has a 64 GiB
chunk of address space in which it places its outbound PIO and 32-bit and
64-bit MMIO windows.
This scheme as several problems:
- It limits guest RAM to 1 TiB (though we have a limited fix for this
now)
- It limits the total MMIO window to 64 GiB. This is not always enough
for some of the large nVidia GPGPU cards
- Putting all the windows into a single 64 GiB area means that naturally
aligning things within there will waste more address space.
In addition there was a miscalculation in some of the defaults, which meant
that the MMIO windows for each PHB actually slightly overran the 64 GiB
region for that PHB. We got away without nasty consequences because
the overrun fit within an unused area at the beginning of the next PHB's
region, but it's not pretty.
This patch implements a new scheme which addresses those problems, and is
also closer to what bare metal hardware and pHyp guests generally use.
Because some guest versions (including most current distro kernels) can't
access PCI MMIO above 64 TiB, we put all the PCI windows between 32 TiB and
64 TiB. This is broken into 1 TiB chunks. The first 1 TiB contains the
PIO (64 kiB) and 32-bit MMIO (2 GiB) windows for all of the PHBs. Each
subsequent TiB chunk contains a naturally aligned 64-bit MMIO window for
one PHB each.
This reduces the number of allowed PHBs (without full manual configuration
of all the windows) from 256 to 31, but this should still be plenty in
practice.
We also change some of the default window sizes for manually configured
PHBs to saner values.
Finally we adjust some tests and libqos so that it correctly uses the new
default locations. Ideally it would parse the device tree given to the
guest, but that's a more complex problem for another time.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
On real hardware, and under pHyp, the PCI host bridges on Power machines
typically advertise two outbound MMIO windows from the guest's physical
memory space to PCI memory space:
- A 32-bit window which maps onto 2GiB..4GiB in the PCI address space
- A 64-bit window which maps onto a large region somewhere high in PCI
address space (traditionally this used an identity mapping from guest
physical address to PCI address, but that's not always the case)
The qemu implementation in spapr-pci-host-bridge, however, only supports a
single outbound MMIO window, however. At least some Linux versions expect
the two windows however, so we arranged this window to map onto the PCI
memory space from 2 GiB..~64 GiB, then advertised it as two contiguous
windows, the "32-bit" window from 2G..4G and the "64-bit" window from
4G..~64G.
This approach means, however, that the 64G window is not naturally aligned.
In turn this limits the size of the largest BAR we can map (which does have
to be naturally aligned) to roughly half of the total window. With some
large nVidia GPGPU cards which have huge memory BARs, this is starting to
be a problem.
This patch adds true support for separate 32-bit and 64-bit outbound MMIO
windows to the spapr-pci-host-bridge implementation, each of which can
be independently configured. The 32-bit window always maps to 2G.. in PCI
space, but the PCI address of the 64-bit window can be configured (it
defaults to the same as the guest physical address).
So as not to break possible existing configurations, as long as a 64-bit
window is not specified, a large single window can be specified. This
will appear the same way to the guest as the old approach, although it's
now implemented by two contiguous memory regions rather than a single one.
For now, this only adds the possibility of 64-bit windows. The default
configuration still uses the legacy mode.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
The 'spapr-pci-host-bridge' represents the virtual PCI host bridge (PHB)
for a PAPR guest. Unlike on x86, it's routine on Power (both bare metal
and PAPR guests) to have numerous independent PHBs, each controlling a
separate PCI domain.
There are two ways of configuring the spapr-pci-host-bridge device: first
it can be done fully manually, specifying the locations and sizes of all
the IO windows. This gives the most control, but is very awkward with 6
mandatory parameters. Alternatively just an "index" can be specified
which essentially selects from an array of predefined PHB locations.
The PHB at index 0 is automatically created as the default PHB.
The current set of default locations causes some problems for guests with
large RAM (> 1 TiB) or PCI devices with very large BARs (e.g. big nVidia
GPGPU cards via VFIO). Obviously, for migration we can only change the
locations on a new machine type, however.
This is awkward, because the placement is currently decided within the
spapr-pci-host-bridge code, so it breaks abstraction to look inside the
machine type version.
So, this patch delegates the "default mode" PHB placement from the
spapr-pci-host-bridge device back to the machine type via a public method
in sPAPRMachineClass. It's still a bit ugly, but it's about the best we
can do.
For now, this just changes where the calculation is done. It doesn't
change the actual location of the host bridges, or any other behaviour.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
This adds a numa id property to a PHB to allow linking passed PCI device
to CPU/memory. It is up to the management stack to do CPU/memory pinning
to the node with the actual PCI device.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
[dwg: Renamed property from "node" to "numa_node" to match the similar
one in the pxb device]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>