Alexander Graf [Tue, 29 Aug 2017 11:38:18 +0000 (13:38 +0200)]
console: add question-mark escape operator
Some termcaps (found using SLES11SP1) use [? sequences. According to man
console_codes (http://linux.die.net/man/4/console_codes) the question mark
is a nop and should simply be ignored.
This patch does exactly that, rendering screen output readable when
outputting guest serial consoles to the graphical console emulator.
virtio-gpu can trigger the assert added by commit "6905b93447 console:
add same surface replace pre-condition" in multihead setups (where
surface can be NULL for secondary displays). Allow surface being NULL.
Drop pixman submodule and support for the "internal" pixman build.
pixman should be reasonably well established meanwhile so we don't
need the fallback submodule any more. While being at it also drop
some #ifdefs for pixman versions older than what we require in
configure anyway.
target/m68k: Switch fpu_rom from make_floatx80() to make_floatx80_init()
GCC 4.7.2 on SunOS reports that the values assigned to array members are not
real constants:
target/m68k/fpu_helper.c:32:5: error: initializer element is not constant
target/m68k/fpu_helper.c:32:5: error: (near initialization for 'fpu_rom[0]')
rules.mak:66: recipe for target 'target/m68k/fpu_helper.o' failed
Convert the array to make_floatx80_init() to fix it.
Replace floatx80_pi-like constants with make_floatx80_init() as they are
defined as make_floatx80().
Peter Maydell [Fri, 8 Sep 2017 13:44:44 +0000 (14:44 +0100)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.11-20170908' into staging
ppc patch queue 2017-09-08
This is the first batch of ppc related patches for qemu-2.11, and it's
accumulated quite a few things. Includes:
* A cleanup to handling of ppc cpu models from Igor
* First parts of fixes to handling of guest vs. host SMT modes from
Sam Bobroff
* Preliminary patches towards supporting the Sam460 board from
Balaton Zoltan
* Several fixes for hotplug logic
* Assorted other fixes and cleanups
* remotes/dgibson/tags/ppc-for-2.11-20170908: (40 commits)
ppc: spapr: Move VCPU ID calculation into sPAPR
ppc: remove non implemented cpu models
ppc: drop caching ObjectClass from PowerPCCPUAlias
ppc: simplify cpu model lookup by PVR
ppc: replace inter-function cyclic dependency/recurssion with 2 simple lookups
ppc: make cpu alias point only to real cpu models
ppc: make cpu_model translation to type consistent
ppc: use macros to make cpu type name from string literal
target/ppc: Remove old STATUS file
PPC: KVM: Support machine option to set VSMT mode
spapr: fallback to raw mode if best compat mode cannot be set during CAS
hw/nvram/spapr_nvram: Device can not be created by the users
hw/ppc/spapr_cpu_core: Add a proper check for spapr machine
ppc4xx: Export ECB and PLB emulation
ppc4xx_i2c: Move to hw/i2c
ppc4xx_i2c: QOMify
ppc4xx: Split off 4xx I2C emulation from ppc405_uc to its own file
ppc4xx: Make MAL emulation more generic
ppc4xx: Move MAL from ppc405_uc to ppc4xx_devs
spapr_iommu: Realloc guest visible TCE table when hot(un)plugging vfio-pci
...
Mark Cave-Ayland [Sun, 16 Jul 2017 20:27:34 +0000 (21:27 +0100)]
pci: add reserved slot check to do_pci_register_device()
Add a new slot_reserved_mask bitmask to PCIBus indicating whether or not each
PCI slot on the bus is reserved. Ensure that it is initialised to zero to
maintain the existing behaviour that all slots are available by default, and
add the additional check with appropriate error reporting to
do_pci_register_device().
This compat property sole function is to prevent the device from being
instantiated. Instead of requiring an extra compat property, check if
fw_cfg has DMA enabled.
fw_cfg is a built-in device that is initialized very early by the
machine init code. We have at least one other device that also
assumes fw_cfg_find() can be safely used on realize: pvpanic.
This has the additional benefit of handling other cases properly, like:
$ qemu-system-x86_64 -device vmgenid -machine none
qemu-system-x86_64: -device vmgenid: vmgenid requires DMA write support in fw_cfg, which this machine type does not provide
$ qemu-system-x86_64 -device vmgenid -machine pc-i440fx-2.9 -global fw_cfg.dma_enabled=off
qemu-system-x86_64: -device vmgenid: vmgenid requires DMA write support in fw_cfg, which this machine type does not provide
$ qemu-system-x86_64 -device vmgenid -machine pc-i440fx-2.6 -global fw_cfg.dma_enabled=on
[boots normally]
Peter Xu [Thu, 17 Aug 2017 05:56:14 +0000 (13:56 +0800)]
intel_iommu: fix missing BQL in pt fast path
In vtd_switch_address_space() we did the memory region switch, however
it's possible that the caller of it has not taken the BQL at all. Make
sure we have it.
hw/pci: add QEMU-specific PCI capability to the Generic PCI Express Root Port
To enable hotplugging of a newly created pcie-pci-bridge,
we need to tell firmware (e.g. SeaBIOS) to reserve
additional buses or IO/MEM/PREF space for pcie-root-port.
Additional bus reservation allows us to hotplug pcie-pci-bridge into this root port.
The number of buses and IO/MEM/PREF space to reserve are provided to the device via
a corresponding property, and to the firmware via new PCI capability.
The properties' default values are -1 to keep default behavior unchanged.
hw/pci: introduce bridge-only vendor-specific capability to provide some hints to firmware
On PCI init PCI bridges may need some extra info about bus number,
IO, memory and prefetchable memory to reserve. QEMU can provide this
with a special vendor-specific PCI capability.
This patch prevents PCI passthrough hotplug on Xen. Even if the Xen tool
stack prepares its own ACPI tables, we still rely on QEMU for hotplug
ACPI notifications.
The original issue is fixed by the two previous patch:
hw/acpi: Limit hotplug to root bus on legacy mode
hw/acpi: Move acpi_set_pci_info to pcihp
Anthony PERARD [Wed, 6 Sep 2017 13:40:32 +0000 (14:40 +0100)]
hw/acpi: Move acpi_set_pci_info to pcihp
HW part of ACPI PCI hotplug in QEMU depends on ACPI_PCIHP_PROP_BSEL
being set on a PCI bus that supports ACPI hotplug. It should work
regardless of the source of ACPI tables (QEMU generator/legacy SeaBIOS/Xen).
So move ACPI_PCIHP_PROP_BSEL initialization into HW ACPI implementation
part from QEMU's ACPI table generator.
To do PCI passthrough with Xen, the property ACPI_PCIHP_PROP_BSEL needs
to be set, but this was done only when ACPI tables are built which is
not needed for a Xen guest. The need for the property starts with commit
"pc: pcihp: avoid adding ACPI_PCIHP_PROP_BSEL twice"
(f0c9d64a68b776374ec4732424a3e27753ce37b6).
Adding find_i440fx into stubs so that mips-softmmu target can be built.
Alex Williamson [Thu, 7 Sep 2017 20:27:09 +0000 (14:27 -0600)]
vhost: Release memory references on cleanup
vhost registers a MemoryListener where it adds and removes references
to MemoryRegions as the MemoryRegionSections pass through. The
region_add callback is invoked for each existing section when the
MemoryListener is registered, but unregistering the MemoryListener
performs no reciprocal region_del callback. It's therefore the
owner of the MemoryListener's responsibility to cleanup any persistent
changes, such as these memory references, after unregistering.
The consequence of this bug is that if we have both a vhost device
and a vfio device, the vhost device will reference any mmap'd MMIO of
the vfio device via this MemoryListener. If the vhost device is then
removed, those references remain outstanding. If we then attempt to
remove the vfio device, it never gets finalized and the only way to
release the kernel file descriptors is to terminate the QEMU process.
Peter Maydell [Fri, 8 Sep 2017 11:57:28 +0000 (12:57 +0100)]
Merge remote-tracking branch 'remotes/jasowang/tags/net-pull-request' into staging
# gpg: Signature made Fri 08 Sep 2017 03:00:34 BST
# gpg: using RSA key 0xEF04965B398D6211
# gpg: Good signature from "Jason Wang (Jason Wang on RedHat) <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 215D 46F4 8246 689E C77F 3562 EF04 965B 398D 6211
* remotes/jasowang/tags/net-pull-request:
colo-compare: Update the COLO document to add the IOThread configuration
colo-compare: Use IOThread to Check old packet regularly and Process pactkets of the primary
qemu-iothread: IOThread supports the GMainContext event loop
net/colo-compare.c: Fix comments and scheme
net/colo-compare.c: Adjust net queue pop order for performance
net/colo-compare.c: Optimize unpredictable tcp options comparison
e1000: Rename the SEC symbol to SEQEC
net/socket: Improve -net socket error reporting
net/net: Convert parse_host_port() to Error
net/socket: Convert several helper functions to Error
net/socket: Don't treat odd socket type as SOCK_STREAM
MAINTAINERS: Update mail address for COLO Proxy
net: rtl8139: do not use old_mmio accesses
net/rocker: Fix the unusual macro name
net/rocker: Convert to realize()
net/rocker: Plug memory leak in pci_rocker_init()
net/rocker: Remove the dead error handling
net/filter-rewriter.c: Fix rewirter checksum bug when use virtio-net
Peter Maydell [Fri, 8 Sep 2017 10:38:55 +0000 (11:38 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tcg-20170907' into staging
TCG constant pools
# gpg: Signature made Thu 07 Sep 2017 23:35:45 BST
# gpg: using RSA key 0x64DF38E8AF7E215F
# gpg: Good signature from "Richard Henderson <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* remotes/rth/tags/pull-tcg-20170907: (23 commits)
tcg/ppc: Use constant pool for movi
tcg/ppc: Look for shifted constants
tcg/ppc: Change TCG_REG_RA to TCG_REG_TB
tcg/arm: Use constant pool for call
tcg/arm: Use constant pool for movi
tcg/arm: Extract INSN_NOP
tcg/arm: Code rearrangement
tcg/arm: Tighten tlb indexing offset test
tcg/arm: Improve tlb load for armv7
tcg/sparc: Use constant pool for movi
tcg/sparc: Introduce TCG_REG_TB
tcg/aarch64: Use constant pool for movi
tcg/s390: Use constant pool for cmpi
tcg/s390: Use constant pool for xori
tcg/s390: Use constant pool for ori
tcg/s390: Use constant pool for andi
tcg/s390: Use constant pool for movi
tcg/s390: Fix sign of patch_reloc addend
tcg/s390: Introduce TCG_REG_TB
tcg/i386: Store out-of-range call targets in constant pool
...
Peter Maydell [Fri, 8 Sep 2017 10:02:54 +0000 (11:02 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/pull-pa-20170907' into staging
Conversion to TranslatorOps
# gpg: Signature made Thu 07 Sep 2017 19:42:48 BST
# gpg: using RSA key 0x64DF38E8AF7E215F
# gpg: Good signature from "Richard Henderson <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* remotes/rth/tags/pull-pa-20170907:
target/hppa: Convert to TranslatorOps
target/hppa: Convert to DisasContextBase
target/hppa: Convert to DisasJumpType
Peter Maydell [Fri, 8 Sep 2017 09:21:54 +0000 (10:21 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/pull-axp-20170907' into staging
Queued target/alpha patches
# gpg: Signature made Thu 07 Sep 2017 19:17:22 BST
# gpg: using RSA key 0x64DF38E8AF7E215F
# gpg: Good signature from "Richard Henderson <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* remotes/rth/tags/pull-axp-20170907:
target/alpha: Switch to do_transaction_failed() hook
target/alpha: Convert to TranslatorOps
target/alpha: Convert to DisasContextBase
target/alpha: Convert to DisasJumpType
Wang Yong [Tue, 29 Aug 2017 07:22:38 +0000 (15:22 +0800)]
colo-compare: Use IOThread to Check old packet regularly and Process pactkets of the primary
Remove the task which check old packet in the comparing thread,
then use IOthread context timer to handle it.
Process pactkets in the IOThread which arrived over the socket.
we use iothread_get_g_main_context to create a new g_main_loop in
the IOThread.then the packets from the primary and the secondary
are processed in the IOThread.
Finally remove the colo-compare thread using the IOThread instead.
Wang Yong [Tue, 29 Aug 2017 07:22:37 +0000 (15:22 +0800)]
qemu-iothread: IOThread supports the GMainContext event loop
IOThread uses AioContext event loop and does not run a GMainContext.
Therefore,chardev cannot work in IOThread,such as the chardev is
used for colo-compare packets reception.
This patch makes the IOThread run the GMainContext event loop,
chardev and IOThread can work together.
net/colo-compare.c: Adjust net queue pop order for performance
The packet_enqueue() use g_queue_push_tail() to
enqueue net packet, so it is more efficent way use
g_queue_pop_head() to get packet for compare.
That will improve the success rate of comparison.
In my test the performance of ftp put 1000M file
will increase 10%
When network is busy, some tcp options(like sack) will unpredictable
occur in primary side or secondary side. it will make packet size
not same, but the two packet's payload is identical. colo just
care about packet payload, so we skip the option field.
Mao Zhongyi [Mon, 4 Sep 2017 14:35:40 +0000 (22:35 +0800)]
net/socket: Improve -net socket error reporting
When -net socket fails, it first reports a specific error, then
a generic one, like this:
$ ./x86_64-softmmu/qemu-system-x86_64 -net socket,mcast=230.0.0.1:1234,listen
qemu-system-x86_64: -net socket,mcast=230.0.0.1:1234,listen: exactly one of listen=, connect=, mcast= or udp= is required
qemu-system-x86_64: -net socket,mcast=230.0.0.1:1234,listen: Device 'socket' could not be initialized
Convert net_socket_*_init() to Error to get rid of the superfluous second
error message. After the patch, the effect like this:
$ ./x86_64-softmmu/qemu-system-x86_64 -net socket,mcast=230.0.0.1:1234,listen
qemu-system-x86_64: -net socket,mcast=230.0.0.1:1234,listen: exactly one of listen=, connect=, mcast= or udp= is requireda
This also fixes a few silent failures to report an error.
Mao Zhongyi [Mon, 4 Sep 2017 14:35:38 +0000 (22:35 +0800)]
net/socket: Convert several helper functions to Error
Currently, net_socket_mcast_create(), net_socket_fd_init_dgram() and
net_socket_fd_init() use the function such as fprintf(), perror() to
report an error message.
Mao Zhongyi [Mon, 4 Sep 2017 14:35:37 +0000 (22:35 +0800)]
net/socket: Don't treat odd socket type as SOCK_STREAM
In net_socket_fd_init(), the 'default' case is odd: it warns,
then continues as if the socket type was SOCK_STREAM. The
comment explains "this could be a eg. a pty", but that makes
no sense. If @fd really was a pty, getsockopt() would fail
with ENOTSOCK. If @fd was a socket, but neither SOCK_DGRAM nor
SOCK_STREAM. It should not be treated as if it was SOCK_STREAM.
Turn this case into an Error. If there is a genuine reason to
support something like SOCK_RAW, it should be explicitly
handled.
Matt Parker [Tue, 15 Aug 2017 20:06:07 +0000 (21:06 +0100)]
net: rtl8139: do not use old_mmio accesses
Both io and memory use the same mmio functions in the rtl8139 device.
This patch removes the separate MemoryRegionOps and old_mmio accessors
for memory, and replaces it with an alias to the io memory region.
Mao Zhongyi [Mon, 14 Aug 2017 03:33:09 +0000 (11:33 +0800)]
net/rocker: Convert to realize()
The rocker device still implements the old PCIDeviceClass .init()
instead of the new .realize(). All devices need to be converted to
.realize().
.init() reports errors with fprintf() and return 0 on success, negative
number on failure. Meanwhile, when -device rocker fails, it first report
a specific error, then a generic one, like this:
$ x86_64-softmmu/qemu-system-x86_64 -device rocker,name=qemu-rocker
rocker: name too long; please shorten to at most 9 chars
qemu-system-x86_64: -device rocker,name=qemu-rocker: Device initialization failed
Now, convert it to .realize() that passes errors to its callers via its
errp argument. Also avoid the superfluous second error message. After
the patch, effect like this:
$ x86_64-softmmu/qemu-system-x86_64 -device rocker,name=qemu-rocker
qemu-system-x86_64: -device rocker,name=qemu-rocker: name too long; please shorten to at most 9 chars
Mao Zhongyi [Mon, 14 Aug 2017 03:33:08 +0000 (11:33 +0800)]
net/rocker: Plug memory leak in pci_rocker_init()
pci_rocker_init() leaks a World when the name more than 9 chars,
then return a negative value directly, doesn't make a correct
cleanup. So add a new goto label to fix it.
Mao Zhongyi [Mon, 14 Aug 2017 03:33:07 +0000 (11:33 +0800)]
net/rocker: Remove the dead error handling
Memory allocation functions like world_alloc, desc_ring_alloc etc,
they are all wrappers around g_malloc, g_new etc. But g_malloc and
similar functions doesn't return null. Because they ignore the fact
that g_malloc() of 0 bytes returns null. So error checks for these
allocation failure are superfluous. Now, remove them entirely.
Sam Bobroff [Thu, 31 Aug 2017 06:38:46 +0000 (16:38 +1000)]
ppc: spapr: Move VCPU ID calculation into sPAPR
Move the calculation of a CPU's VCPU ID out of the generic PPC code
(ppc_cpu_realizefn()) and into sPAPR specific code
(spapr_cpu_core_realize()) where it belongs.
Unfortunately, due to the way things are ordered, we still need to
default the VCPU ID in ppc_cpu_realizfn() but at least doing that
doesn't require any interaction with sPAPR.
Igor Mammedov [Wed, 30 Aug 2017 13:24:34 +0000 (15:24 +0200)]
ppc: drop caching ObjectClass from PowerPCCPUAlias
Caching there practically doesn't give any benefits
and that at slow path druring querying supported CPU list.
But it introduces non conventional path of where from
comes used CPU type name (kvm_ppc_register_host_cpu_type).
Taking in account that kvm_ppc_register_host_cpu_type()
fixes up models the aliases point to, it's sufficient to
make ppc_cpu_class_by_name() translate cpu alias to
correct cpu type name.
So drop PowerPCCPUAlias::oc field + ppc_cpu_class_by_alias()
and let ppc_cpu_class_by_name() do conversion to cpu type name,
which simplifies code a little bit saving ~20LOC and trouble
wondering why ppc_cpu_class_by_alias() is necessary.
Igor Mammedov [Wed, 30 Aug 2017 13:24:32 +0000 (15:24 +0200)]
ppc: replace inter-function cyclic dependency/recurssion with 2 simple lookups
previous patches cleaned up cpu model/alias naming which
allows to simplify cpu model/alias to cpu type lookup a bit
byt removing recurssion and dependency of ppc_cpu_class_by_name() /
ppc_cpu_class_by_alias() on each other.
Besides of simplifying code it reduces it by ~15LOC.
Igor Mammedov [Wed, 30 Aug 2017 13:24:31 +0000 (15:24 +0200)]
ppc: make cpu alias point only to real cpu models
alias pointing to another alias forces lookup code to
do recurrsive translation till real cpu model is reached.
Drop this nonsence and make each alias point to cpu model
that has corresponding CPU type. It will allow to drop
recurrsion in cpu model translation code and actually
make ppc_cpu_aliases[] content use PowerPCCPUAlias
fields properly
(i.e. alias goes into .alias and model goes into .model)
While at it add TODO defines around aliases that point to
cpu models excluded by the same TODO defines.
Igor Mammedov [Wed, 30 Aug 2017 13:24:30 +0000 (15:24 +0200)]
ppc: make cpu_model translation to type consistent
PPC handles -cpu FOO rather incosistently,
i.e. it does case-insensitive matching of FOO to
a CPU type (see: ppc_cpu_compare_class_name) but
handles alias names as case-sensitive, as result:
# qemu-system-ppc64 -M mac99 -cpu g3
qemu-system-ppc64: unable to find CPU model ' kN�U'
# qemu-system-ppc64 -cpu 970MP_V1.1
qemu-system-ppc64: Unable to find sPAPR CPU Core definition
Igor Mammedov [Wed, 30 Aug 2017 13:24:29 +0000 (15:24 +0200)]
ppc: use macros to make cpu type name from string literal
Replace
"-" TYPE_POWERPC_CPU
when composing cpu type name from cpu model string literal
and the same pattern in format strings with
POWERPC_CPU_TYPE_SUFFIX and POWERPC_CPU_TYPE_NAME(model)
macroses like we do in x86.
Later POWERPC_CPU_TYPE_NAME() will be used to define default
cpu type per machine type and as bonus it will be consistent
and easy grep-able pattern across all other targets that I'm
plannig to treat the same way.
Thomas Huth [Fri, 1 Sep 2017 07:38:12 +0000 (09:38 +0200)]
target/ppc: Remove old STATUS file
The target/ppc/STATUS file has seen its last real update 10 years
ago - so the information in there is not up to date anymore. Since
nobody seems to care about this file, let's simply remove it.
Sam Bobroff [Fri, 18 Aug 2017 05:50:22 +0000 (15:50 +1000)]
PPC: KVM: Support machine option to set VSMT mode
KVM now allows writing to KVM_CAP_PPC_SMT which has previously been
read only. Doing so causes KVM to act, for that VM, as if the host's
SMT mode was the given value. This is particularly important on Power
9 systems because their default value is 1, but they are able to
support values up to 8.
This patch introduces a way to control this capability via a new
machine property called VSMT ("Virtual SMT"). If the value is not set
on the command line a default is chosen that is, when possible,
compatible with legacy systems.
Note that the intialization of KVM_CAP_PPC_SMT has changed slightly
because it has changed (in KVM) from a global capability to a
VM-specific one. This won't cause a problem on older KVMs because VM
capabilities fall back to global ones.
Greg Kurz [Thu, 17 Aug 2017 11:23:50 +0000 (13:23 +0200)]
spapr: fallback to raw mode if best compat mode cannot be set during CAS
KVM PR doesn't allow to set a compat mode. This causes ppc_set_compat_all()
to fail and we return H_HARDWARE to the guest right away.
This is excessive: even if we favor compat mode since commit 152ef803ceb19,
we should at least fallback to raw mode if the guest supports it.
This patch modifies cas_check_pvr() so that it also reports that the real
PVR was found in the table supplied by the guest. Note that this is only
makes sense if raw mode isn't explicitely disabled (ie, the user didn't
set the machine "max-cpu-compat" property). If this is the case, we can
simply ignore ppc_set_compat_all() failures, and let the guest run in raw
mode.
This NVRAM device registers RTAS calls during its realize function
and thus can only be used once - and that's internally from spapr.c.
So let's mark the device with user_creatable = false to avoid that
the users can crash their QEMU this way.
Thomas Huth [Thu, 24 Aug 2017 03:52:32 +0000 (05:52 +0200)]
hw/ppc/spapr_cpu_core: Add a proper check for spapr machine
QEMU currently crashes when the user tries to add a spapr-cpu-core
on a non-pseries machine:
$ qemu-system-ppc64 -S -machine ppce500,accel=tcg \
-device POWER5+_v2.1-spapr-cpu-core
hw/ppc/spapr_cpu_core.c:178:spapr_cpu_core_realize_child:
Object 0x55cee1f55160 is not an instance of type spapr-machine
Aborted (core dumped)
So let's add a proper check for the correct machine time with
a more friendly error message here.
BALATON Zoltan [Sun, 20 Aug 2017 17:23:05 +0000 (19:23 +0200)]
ppc4xx: Export ECB and PLB emulation
Make these device models available outside ppc405_uc.c for reuse in
460EX emulation. They are left in their current place for now because
they are used mostly unchanged and I'm not sure these correctly model
the components in 440 SoCs (but they seem to be good enough). These
functions could be moved in a subsequent clean up series when this is
confirmed.
spapr_iommu: Realloc guest visible TCE table when hot(un)plugging vfio-pci
This replaces g_malloc() with spapr_tce_alloc_table() as this is
the standard way of allocating tables and this allows moving the table
back to KVM when unplugging a VFIO PCI device and VFIO TCE acceleration
support is not present in the KVM.
Although spapr_tce_alloc_table() is expected to fail with EBUSY
if called when previous fd is not closed yet, in practice we will not
see it because cap_spapr_vfio is false at the moment.
KONRAD Frederic [Mon, 7 Aug 2017 15:50:46 +0000 (17:50 +0200)]
booke206: fix tlbnps for fixed size TLB
Some OS don't populate the TSIZE field when using a fixed size TLB which result
in a 1KB TLB. When the TLB is a fixed size TLB the TSIZE field should be
ignored.
KONRAD Frederic [Mon, 7 Aug 2017 15:50:45 +0000 (17:50 +0200)]
booke206: fix booke206_tlbnps for mav 2.0
This fixes booke206_tlbnps for MAV 2.0 by checking the MMUCFG register and
return directly the right tlbnps instead of computing it from non existing
field.
Sam Bobroff [Wed, 9 Aug 2017 05:38:56 +0000 (15:38 +1000)]
ppc: spapr: Make VCPU ID handling private to SPAPR
The concept of a VCPU ID that differs from the CPU's index
(cpu->cpu_index) exists only within SPAPR machines so, move the
functions ppc_get_vcpu_id() and ppc_get_cpu_by_vcpu_id() into spapr.c
and rename them appropriately.
Sam Bobroff [Thu, 3 Aug 2017 06:28:44 +0000 (16:28 +1000)]
ppc: spapr: Rename cpu_dt_id to vcpu_id
This field actually records the VCPU ID used by KVM and, although the
value is also used in the device tree it is primarily the VCPU ID so
rename it as such.
Sam Bobroff [Thu, 3 Aug 2017 06:28:36 +0000 (16:28 +1000)]
e500: Use cpu_index instead of vcpu_dt_id
The e500 platform code uses the function ppc_get_vcpu_dt_id() to get
an id to put in its device tree. Which seems like it makes sense, but
ppc_get_vcpu_dt_id() is actually badly named - it only differs from
cpu_index in cases where you're running on KVM HV and the host's
number of threads differs from the guests. Since KVM HV only supports
PAPR, not e500, it doesn't make sense to use it here.
Simply use the cpu_index instead (which is 'i' in this context
because qemu_get_cpu(i) returns the cpu with cpu_index == i).
Greg Kurz [Tue, 25 Jul 2017 17:59:44 +0000 (19:59 +0200)]
spapr_drc: add unrealize method to physical DRC class
When hot-unplugging a PHB, all its PCI DRC connectors get unrealized. This
patch adds an unrealize method to the physical DRC class, in order to undo
registrations performed in realize_physical().
Greg Kurz [Tue, 25 Jul 2017 17:59:18 +0000 (19:59 +0200)]
spapr_pci: parent the MSI memory region to the PHB
This memory region should be owned by the PHB. This ensures the PHB
cannot be finalized as long as the the region is guest visible, or
used by a CPU or a device.
Greg Kurz [Tue, 25 Jul 2017 17:58:53 +0000 (19:58 +0200)]
spapr_drc: use g_strdup_printf() instead of snprintf()
Passing a stack allocated buffer of arbitrary length to snprintf()
without checking the return value can cause the resultant strings
to be silently truncated.
Greg Kurz [Tue, 25 Jul 2017 17:58:40 +0000 (19:58 +0200)]
spapr_iommu: use g_strdup_printf() instead of snprintf()
Passing a stack allocated buffer of arbitrary length to snprintf()
without checking the return value can cause the resultant strings
to be silently truncated.
This patch is a follow up on the discussions made in patch
"hw/ppc: disable hotplug before CAS is completed" that can be
found at [1].
At this moment, we do not support CPU/memory hotplug in early
boot stages, before CAS. When a hotplug occurs, the event is logged
in an internal RTAS event log queue and an IRQ pulse is fired. In
regular conditions, the guest handles the interrupt by executing
check_exception, fetching the generated hotplug event and enabling
the device for use.
In early boot, this IRQ isn't caught (SLOF does not handle hotplug
events), leaving the event in the rtas event log queue. If the guest
executes check_exception due to another hotplug event, the re-assertion
of the IRQ ends up de-queuing the first hotplug event as well. In short,
a device hotplugged before CAS is considered coldplugged by SLOF.
This leads to device misbehavior and, in some cases, guest kernel
Ooops when trying to unplug the device.
A proper fix would be to turn every device hotplugged before CAS
as a colplugged device. This is not trivial to do with the current
code base though - the FDT is written in the guest memory at
ppc_spapr_reset and can't be retrieved without adding extra state
(fdt_size for example) that will need to managed and migrated. Adding
the hotplugged DT in the middle of CAS negotiation via the updated DT
tree works with CPU devs, but panics the guest kernel at boot. Additional
analysis would be necessary for LMBs and PCI devices. There are
questions to be made in QEMU/SLOF/kernel level about how we can make
this change in a sustainable way.
With Linux guests, a fix would be the kernel executing check_exception
at boot time, de-queueing the events that happened in early boot and
processing them. However, even if/when the newer kernels start
fetching these events at boot time, we need to take care of older
kernels that won't be doing that.
This patch works around the situation by issuing a CAS reset if a hotplugged
device is detected during CAS:
- the DRC conditions that warrant a CAS reset is the same as those that
triggers a DRC migration - the DRC must have a device attached and
the DRC state is not equal to its ready_state. With that in mind, this
patch makes use of 'spapr_drc_needed' to determine if a CAS reset
is needed.
- In the middle of CAS negotiations, the function
'spapr_hotplugged_dev_before_cas' goes through all the DRCs to see
if there are any DRC that requires a reset, using spapr_drc_needed. If
that happens, returns '1' in 'spapr_h_cas_compose_response' which will set
spapr->cas_reboot to true, causing the machine to reboot.
The sPAPR machine isn't clearing up the pending events QTAILQ on
machine reboot. This allows for unprocessed hotplug/epow events
to persist in the queue after reset and, when reasserting the IRQs in
check_exception later on, these will be being processed by the OS.
This patch implements a new function called 'spapr_clear_pending_events'
that clears up the pending_events QTAILQ. This helper is then called
inside ppc_spapr_reset to clear up the events queue, preventing
old/deprecated events from persisting after a reset.
hw/ppc/spapr_drc.c: change spapr_drc_needed to use drc->dev
This patch makes a small fix in 'spapr_drc_needed' to change how we detect
if a DRC has a device attached. Previously it used dr_entity_sense for this,
which works for physical DRCs.
However, for logical DRCs, it didn't cover the case where a logical DRC has
a drc->dev but the state is LOGICAL_UNUSABLE (e.g. a hotplugged CPU before
CAS). In this case, the dr_entity_sense of this DRC returns UNUSABLE and the
code was considering that there were no dev attached, making spapr_drc_needed
return 'false' when in fact we would like to migrate the DRC.
Changing it to check for drc->dev instead works for all DRC types.
At this point the conversion is a wash. Loading of TB+ofs is
smaller, but the actual return address from exit_tb is larger.
There are a few more insns required to transition between TBs.
But the expectation is that accesses to the constant pool will
on the whole be smaller.
We are not going to use ldrd for loading the comparator
for 32-bit guests, so don't limit cmp_off to 8 bits then.
This eliminates one insn in the tlb load for some guests.
Use UBFX to avoid limitation on CPU_TLB_BITS. Since we're dropping
the initial shift, we need to replace the page masking. We can use
MOVW+BIC to do this without shifting. The result is the same size
as the armv6 path with one less conditional instruction.
projects / qemu.git / log