Immediate counter re-load on setting (or on starting to run with)
counter = 0 is a wrong behaviour for some of the timers. Add "no
immediate reload" policy that provides correct behaviour for such timers.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Message-id: bf9385cd2550ca451d564fa46007688cee3f3d9d.1475421224.git.digetx@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Performing trigger on setting (or starting to run with) counter = 0 could
be a wrong behaviour for some of the timers, provide "no immediate trigger"
policy to maintain correct behaviour for such timers.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Message-id: 72c0319cf2ec599f22397b7da280c06c34dc40dd.1475421224.git.digetx@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Currently, periodic timer that has load = delta = 0 performs trigger
on timer reload and stops, printing a "period zero" error message.
Introduce new policy that makes periodic timer to continuously trigger
with a period interval in case of load = 0.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Message-id: 632b23dd11055d9bd5e338d66b38fac0bd51462e.1475421224.git.digetx@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Currently, periodic counter wraps around immediately once counter reaches
"0", this is wrong behaviour for some of the timers, resulting in one period
being lost. Add new ptimer policy that provides correct behaviour for such
timers, so that counter stays with "0" for a one period before wrapping
around.
Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Message-id: f22a670cf1f4be298b31640cb5f4be1df0f20ab6.1475421224.git.digetx@gmail.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Support target CPUs having a page size which isn't knownn
at compile time. To use this, the CPU implementation should:
* define TARGET_PAGE_BITS_VARY
* not define TARGET_PAGE_BITS
* define TARGET_PAGE_BITS_MIN to the smallest value it
might possibly want for TARGET_PAGE_BITS
* call set_preferred_target_page_bits() in its realize
function to indicate the actual preferred target page
size for the CPU (and report any error from it)
In CONFIG_USER_ONLY, the CPU implementation should continue
to define TARGET_PAGE_BITS appropriately for the guest
OS page size.
Machines which want to take advantage of having the page
size something larger than TARGET_PAGE_BITS_MIN must
set the MachineClass minimum_page_bits field to a value
which they guarantee will be no greater than the preferred
page size for any CPU they create.
Note that changing the target page size by setting
minimum_page_bits is a migration compatibility break
for that machine.
For debugging purposes, attempts to use TARGET_PAGE_SIZE
before it has been finally confirmed will assert.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
It's only used in qmp_chardev_add(), so use a create() argument instead.
Also switched to typedef functions for CharDriverParse/CharDriverCreate.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022100951.19562-7-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
No need to count the users of a CharDriverState, it can rely on the fact
of whether there is a CharBackend associated or if there is enough space
in the muxer.
Simplify and fold chr_mux_new_fe() in qemu_chr_fe_init() since there is
a single user now. Also switch from fprintf to raising error instead.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022100951.19562-5-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
I introduced this function in d61b0c9a2f, but it isn't
used. Furthermore, it was incomplete, as it would need to translate QEMU
chr events to Spice port events.
(presumably it was used in the follow-up NBD-spice series that was not
completed: http://lists.gnu.org/archive/html/qemu-devel/2013-11/msg02024.html)
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022100951.19562-4-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This may help to catch unhandled cases, and avoid having to maintain
numbering.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022100951.19562-3-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Usage has long been removed, since commit f220174de8.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022100951.19562-2-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The fe_open state belongs to front end.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022100951.19562-1-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
No need to keep explicit_fe_open around if it affects only a
qemu_chr_fe_set_handlers(). Use an additional argument instead.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-24-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The function is used to free the backend opaque pointer, let's name it
accordingly.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-23-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since the hanlders are associated with a CharBackend, rather than the
CharDriverState, it is more appropriate to store in CharBackend. This
avoids the handler copy dance in qemu_chr_fe_set_handlers() then
mux_chr_update_read_handler(), by storing the CharBackend pointer
directly.
Also a mux CharDriver should go through mux->backends[focused], since
chr->be will stay NULL. Before that, it was possible to call
chr->handler by mistake with surprising results, for ex through
qemu_chr_be_can_write(), which would result in calling the last set
handler front end, not the one with focus.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-22-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
In most cases, front ends do not care about the side effect of
CharBackend, so we can simply skip the checks and call the qemu_chr_fe
functions even without associated CharDriver.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-20-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that all front end use qemu_chr_fe_init(), we can move chardev
claiming in init(), and add a function deinit() to release the chardev
and cleanup handlers.
The qemu_chr_fe_claim_no_fail() for property are gone, since the
property will raise an error instead. In other cases, where there is
already an error path, an error is raised instead. Finally, other cases
are handled by &error_abort in qemu_chr_fe_init().
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-19-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
qemu_chr_add_handlers*() have been removed in previous change, so the
common qemu_chr_set_handlers() is no longer needed.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-17-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This also switches from qemu_chr_add_handlers() to
qemu_chr_fe_set_handlers(). Note that qemu_chr_fe_set_handlers() now
takes the focus when fe_open (qemu_chr_add_handlers() did take the
focus)
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-16-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
qemu_chr_accept_input() and qemu_chr_disconnect() are only used by
frontend, so use qemu_chr_fe prefix.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-14-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Store the property in a CharBackend instead of CharDriverState*. This
also replace systematically chr by chr.chr to access the
CharDriverState*. The following patches will replace it with calls to
qemu_chr_fe CharBackend functions.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-12-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This new structure is meant to keep the details associated with a char
driver usage. On initialization, it gets a tag from the mux backend.
It can change its handlers thanks to qemu_chr_fe_set_handlers().
This structure is introduced so that all frontend will be moved to hold
and use a CharBackend. This will allow to better track char usage and
allocation, and help prevent some memory leaks or corruption.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-10-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Make qemu_chr_add_handlers_full() aware of mux handling. This allows
introduction of a tag associated with the fe handlers and a
qemu_chr_set_handlers() function to set the handler for a particular
tag. That will allow to get rid of qemu_chr_add_handlers*() in later
changes, in favor of qemu_chr_fe_set_handler().
To this end, chr_update_read_handler callback is enhanced with a tag
argument, and mux_chr_update_read_handler() is splitted in new
functions: mux_chr_new_handler_tag(), mux_chr_set_handlers(),
mux_set_focus().
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-9-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
The CharDriverState.init() callback is no longer set since commit
a61ae7f88c and thus unused. The only user, the malta FGPA display has
been converted to use an event "opened" callback instead.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-7-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Since commit b6607a1a20, serial_hds_isa_init() was introduced to
factor out serial_isa_init() loops. However, sun4uv shouldn't start from
0 when there is a mm serial on 0 already. Add a "from" argument to
serial_hds_isa_init().
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20161022095318.17775-5-marcandre.lureau@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This speeds up MEMORY_LISTENER_CALL noticeably. Right now,
with many PCI devices you have N regions added to M AddressSpaces
(M = # PCI devices with bus-master enabled) and each call looks
up the whole listener list, with at least M listeners in it.
Because most of the regions in N are BARs, which are also roughly
proportional to M, the whole thing is O(M^3). This changes it
to O(M^2), which is the best we can do without rewriting the
whole thing.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This introduces load-acquire and store-release operations in QEMU.
For now, just use them as an implementation detail of atomic_mb_read
and atomic_mb_set.
Since docs/atomics.txt documents that atomic_mb_read only synchronizes
with an atomic_mb_set of the same variable, we can use the new implementation
everywhere instead of seq-cst loads and stores.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Introduce CTR mode support for the cipher APIs.
CTR mode uses a counter rather than a traditional IV.
The counter has additional properties, including a nonce
and initial counter block. We reuse the ctx->iv as
the counter for conveniences.
Both libgcrypt and nettle are support CTR mode, the
cipher-builtin doesn't support yet.
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
It can't guarantee all cipher modes are supported
if one cipher algorithm is supported by a backend.
Let's extend qcrypto_cipher_supports() to take both
the algorithm and mode as parameters.
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Daniel P. Berrange <berrange@redhat.com>
We should avoid exposing new hardware (through DT and ACPI) on older
machine types. This patch keeps 2.7 and older from changing, despite
the introduction of ITS support for 2.8.
Signed-off-by: Andrew Jones <drjones@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 1476117341-32690-3-git-send-email-drjones@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The SMC controllers on the Aspeed AST2500 SoC are very similar to the
ones found on the AST2400. The differences are on the number of
supported flash modules and their default mappings in the SoC address
space.
The Aspeed AST2500 has one SPI controller for the BMC firmware and two
for the host firmware. All controllers have now the same set of
registers compatible with the AST2400 FMC controller and the legacy
'SMC' controller is fully gone.
We keep the FMC object to act as the BMC SPI controller and add a new
SPI controller for the host. We also have to introduce new type names
to handle the differences in the flash modules memory mappping.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Message-id: 1474977462-28032-5-git-send-email-clg@kaod.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The AST2500 SoC has two. Let's prepare ground for the next changes
which will add the required definitions for the second host SPI
controller.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Message-id: 1474977462-28032-4-git-send-email-clg@kaod.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This will ease the definition of the new controllers for the AST2500
SoC and also ease the support of the segment registers, which provide
a way to reconfigure the mapping window of each slave.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Message-id: 1474977462-28032-3-git-send-email-clg@kaod.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The Aspeed SoC has three different types of SMC (Static Memory
Controller) controllers: the SMC (legacy), the FMC (the new one) and
the SPI for the host PNOR. The FMC and the SPI models are now
converging on the AST2500 SoC and the SMC, which was still available
on the AST2400 SoC, was removed.
The Aspeed SoC does not provide support for the legacy SMC
controller. So, let's rename the 'smc' object to 'fmc' to clarify its
nature.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Message-id: 1474977462-28032-2-git-send-email-clg@kaod.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Cluster x2APIC cannot work without KVM's x2apic API when the maximal
APIC ID is greater than 8 and only KVM's LAPIC can support x2APIC, so we
forbid other APICs and also the old KVM case with less than 9, to
simplify the code.
There is no point in enabling EIM in forbidden APICs, so we keep it
enabled only for the KVM APIC; unconditionally, because making the
option depend on KVM version would be a maintanance burden.
Old QEMUs would enable eim whenever intremap was on, which would trick
guests into thinking that they can enable cluster x2APIC even if any
interrupt destination would get clamped to 8 bits.
Depending on your configuration, QEMU could notice that the destination
LAPIC is not present and report it with a very non-obvious:
KVM: injection failed, MSI lost (Operation not permitted)
Or the guest could say something about unexpected interrupts, because
clamping leads to aliasing so interrupts were being delivered to
incorrect VCPUs.
KVM_X2APIC_API is the feature that allows us to enable EIM for KVM.
QEMU 2.7 allowed EIM whenever interrupt remapping was enabled. In order
to keep backward compatibility, we again allow guests to misbehave in
non-obvious ways, and make it the default for old machine types.
A user can enable the buggy mode it with "x-buggy-eim=on".
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
The default (auto) emulates the current behavior.
A user can now control EIM like
-device intel-iommu,intremap=on,eim=off
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
The MMIO based interface to APIC doesn't work well with MSIs that have
upper address bits set (remapped x2APIC MSIs). A specialized interface
is a quick and dirty way to avoid the shortcoming.
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Every configuration has only up to one APIC class and we'll be extending
the class with a function that can be called without an instanced
object, so a direct access to the class is convenient.
This patch will break compilation if some code uses apic_get_class()
with CONFIG_USER_ONLY.
Suggested-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Pass an error object to prepare for migration to VFIO-PCI realize.
In vfio platform vfio_base_device_init we currently just report the
error. Subsequent patches will propagate the error up to the realize
function.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Pass an error object to prepare for migration to VFIO-PCI realize.
For the time being let's just simply report the error in
vfio platform's vfio_base_device_init(). A subsequent patch will
duly propagate the error up to vfio_platform_realize.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
To prepare for migration to realize, let's use a local error
object in vfio_initfn. Also let's use the same error prefix for all
error messages.
On top of the 1-1 conversion, we start using a common error prefix for
all error messages. We also introduce a similar warning prefix which will
be used later on.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Highlights:
* Significant rework of how PCI IO windows are placed for the
pseries machine type
* A number of extra tests added for ppc
* Other tests clean up / fixed
* Some cleanups to the XICS interrupt controller in preparation
for the 'powernv' machine type
A number of the test changes aren't strictly in ppc related code, but
are included via my tree because they're primarily focused on
improving test coverage for ppc.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=tnCe
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.8-20161017' into staging
ppc patch queue 2016-10-17
Highlights:
* Significant rework of how PCI IO windows are placed for the
pseries machine type
* A number of extra tests added for ppc
* Other tests clean up / fixed
* Some cleanups to the XICS interrupt controller in preparation
for the 'powernv' machine type
A number of the test changes aren't strictly in ppc related code, but
are included via my tree because they're primarily focused on
improving test coverage for ppc.
# gpg: Signature made Mon 17 Oct 2016 03:42:41 BST
# gpg: using RSA key 0x6C38CACA20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>"
# gpg: aka "David Gibson (Red Hat) <dgibson@redhat.com>"
# gpg: aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>"
# gpg: aka "David Gibson (kernel.org) <dwg@kernel.org>"
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E 87DC 6C38 CACA 20D9 B392
* remotes/dgibson/tags/ppc-for-2.8-20161017:
spapr: Improved placement of PCI host bridges in guest memory map
spapr_pci: Add a 64-bit MMIO window
spapr: Adjust placement of PCI host bridge to allow > 1TiB RAM
spapr_pci: Delegate placement of PCI host bridges to machine type
libqos: Limit spapr-pci to 32-bit MMIO for now
libqos: Correct error in PCI hole sizing for spapr
libqos: Isolate knowledge of spapr memory map to qpci_init_spapr()
ppc/xics: Split ICS into ics-base and ics class
ppc/xics: Make the ICSState a list
spapr: fix inheritance chain for default machine options
target-ppc: implement vexts[bh]2w and vexts[bhw]2d
tests/boot-sector: Increase time-out to 90 seconds
tests/boot-sector: Use mkstemp() to create a unique file name
tests/boot-sector: Use minimum length for the Forth boot script
qtest: ask endianness of the target in qtest_init()
tests: minor cleanups in usb-hcd-uhci-test
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
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>
Dmitry Osipenko
Peter Maydell
Marc-André Lureau
Paolo Bonzini
Gonglei
Andrew Jones
Cédric Le Goater
Radim Krčmář
Eric Auger
David Gibson