Thomas Huth [Tue, 15 Sep 2015 19:34:20 +0000 (21:34 +0200)]
ppc/spapr: Fix buffer overflow in spapr_populate_drconf_memory()
The buffer that is allocated in spapr_populate_drconf_memory()
is used for setting both, the "ibm,dynamic-memory" and the
"ibm,associativity-lookup-arrays" property. However, only the
size of the first one is taken into account when allocating the
memory. So if the length of the second property is larger than
the length of the first one, we run into a buffer overflow here!
Fix it by taking the length of the second property into account,
too.
Fixes: "spapr: Support ibm,dynamic-reconfiguration-memory" patch Signed-off-by: Thomas Huth <[email protected]> Reviewed-by: David Gibson <[email protected]> Signed-off-by: David Gibson <[email protected]>
David Gibson [Tue, 8 Sep 2015 01:21:52 +0000 (11:21 +1000)]
spapr: Fix default NUMA node allocation for threads
At present, if guest numa nodes are requested, but the cpus in each node
are not specified, spapr just uses the default behaviour or assigning each
vcpu round-robin to nodes.
If smp_threads != 1, that will assign adjacent threads in a core to
different NUMA nodes. As well as being just weird, that's a configuration
that can't be represented in the device tree we give to the guest, which
means the guest and qemu end up with different ideas of the NUMA topology.
This patch implements mc->cpu_index_to_socket_id in the spapr code to
make sure vcpus get assigned to nodes only at the socket granularity.
Bharata B Rao [Mon, 3 Aug 2015 05:35:43 +0000 (11:05 +0530)]
spapr: Move memory hotplug to RTAS_LOG_V6_HP_ID_DRC_COUNT type
Till now memory hotplug used RTAS_LOG_V6_HP_ID_DRC_INDEX hotplug type
which meant that we generated one hotplug type of EPOW event for every
256MB (SPAPR_MEMORY_BLOCK_SIZE). This quickly overruns the kernel
rtas log buffer thus resulting in loss of memory hotplug events. Switch
to RTAS_LOG_V6_HP_ID_DRC_COUNT hotplug type for memory so that we
generate only one event per hotplug request.
Bharata B Rao [Mon, 3 Aug 2015 05:35:41 +0000 (11:05 +0530)]
spapr: Revert to memory@XXXX representation for non-hotplugged memory
Don't represent non-hotluggable memory under drconf node. With this
we don't have to create DRC objects for them.
The effect of this patch is that we revert back to memory@XXXX representation
for all the memory specified with -m option and represent the cold
plugged memory and hot-pluggable memory under
ibm,dynamic-reconfiguration-memory.
Bharata B Rao [Mon, 3 Aug 2015 05:35:40 +0000 (11:05 +0530)]
spapr: Populate ibm,associativity-lookup-arrays correctly for non-NUMA
When NUMA isn't configured explicitly, assume node 0 is present for
the purpose of creating ibm,associativity-lookup-arrays property
under ibm,dynamic-reconfiguration-memory DT node. This ensures that
the associativity index property is correctly updated in ibm,dynamic-memory
for the LMB that is hotplugged.
Bharata B Rao [Mon, 29 Jun 2015 08:44:32 +0000 (14:14 +0530)]
spapr: Don't allow memory hotplug to memory less nodes
Currently PowerPC kernel doesn't allow hot-adding memory to memory-less
node, but instead will silently add the memory to the first node that has
some memory. This causes two unexpected behaviours for the user.
- Memory gets hotplugged to a different node than what the user specified.
- Since pc-dimm subsystem in QEMU still thinks that memory belongs to
memory-less node, a reboot will set things accordingly and the previously
hotplugged memory now ends in the right node. This appears as if some
memory moved from one node to another.
So until kernel starts supporting memory hotplug to memory-less
nodes, just prevent such attempts upfront in QEMU.
Bharata B Rao [Mon, 29 Jun 2015 08:44:30 +0000 (14:14 +0530)]
spapr: Make hash table size a factor of maxram_size
The hash table size is dependent on ram_size, but since with hotplug
the memory can grow till maxram_size. Hence make hash table size dependent
on maxram_size.
This allows to hotplug huge amounts of memory to the guest.
Parse ibm,architecture.vec table obtained from the guest and enable
memory node configuration via ibm,dynamic-reconfiguration-memory if guest
supports it. This is in preparation to support memory hotplug for
sPAPR guests.
This changes the way memory node configuration is done. Currently all
memory nodes are built upfront. But after this patch, only memory@0 node
for RMA is built upfront. Guest kernel boots with just that and rest of
the memory nodes (via memory@XXX or ibm,dynamic-reconfiguration-memory)
are built when guest does ibm,client-architecture-support call.
Note: This patch needs a SLOF enhancement which is already part of
SLOF binary in QEMU.
David Gibson [Wed, 12 Aug 2015 03:16:48 +0000 (13:16 +1000)]
spapr: Add LMB DR connectors
Enable memory hotplug for pseries 2.4 and add LMB DR connectors.
With memory hotplug, enforce RAM size, NUMA node memory size and maxmem
to be a multiple of SPAPR_MEMORY_BLOCK_SIZE (256M) since that's the
granularity in which LMBs are represented and hot-added.
LMB DR connectors will be used by the memory hotplug code.
Signed-off-by: Bharata B Rao <[email protected]> Signed-off-by: Michael Roth <[email protected]>
[spapr_drc_reset implementation]
[since this missed the 2.4 cutoff, changing to only enable for 2.5] Signed-off-by: David Gibson <[email protected]>
sPAPR uses hard coded limit of maximum 255 supported CPUs which is
exactly the same as QEMU-wide limit which is MAX_CPUMASK_BITS and also
defined as 255.
This makes use of a global CPU number limit for the "pseries" machine.
In order to anticipate future increase of the MAX_CPUMASK_BITS
(or to help debugging large systems), this also bumps the FDT_MAX_SIZE
limit from 256K to 1M assuming that 1 CPU core needs roughly 512 bytes
in the device tree so the new limit can cover up to 2048 CPU cores.
David Gibson [Wed, 16 Sep 2015 06:57:51 +0000 (16:57 +1000)]
spapr: Don't use QOM [*] syntax for DR connectors.
The dynamic reconfiguration (hotplug) code for the pseries machine type
uses a "DR connector" QOM object for each resource it will be possible
to hotplug. Each of these is added to its owner using
object_property_add_child(owner, "dr-connector[*], ...);
That works ok, mostly, but it means that the property indices are
arbitrary, depending on the order in which the connectors are constructed.
That might line up to something useful, but it doesn't have to.
It will get worse once we add hotplug RAM support. That will add a DR
connector object for every 256MB of potential memory. So if maxmem=2T,
for example, there are 8192 objects under the same parent.
The QOM interfaces aren't really designed for this. In particular
object_property_add() with [*] has O(n^2) time complexity (in the number of
existing children): first it has a linear search through array indices to
find a free slot, each of which is attempted to a recursive call to
object_property_add() with a specific [N]. Those calls are O(n) because
there's a linear search through all properties to check for duplicates.
By using a meaningful index value, which we already know is unique we can
avoid the [*] special behaviour. That lets us reduce the total time for
creating the DR objects from O(n^3) to O(n^2).
O(n^2) is still kind of crappy, but it's enough to reduce the startup time
of qemu (with in-progress memory hotplug support) with maxmem=2T from ~20
minutes to ~4 seconds.
Michael Roth [Thu, 10 Sep 2015 21:11:02 +0000 (16:11 -0500)]
spapr_drc: use RTAS return codes for methods called by RTAS
Certain methods in sPAPRDRConnector objects are only ever called by
RTAS and in many cases are responsible for the logic that determines
the RTAS return codes.
Rather than having a level of indirection requiring RTAS code to
re-interpret return values from such methods to determine the
appropriate return code, just pass them through directly.
This requires changing method return types to uint32_t to match the
type of values currently passed to RTAS helpers.
In the case of read accesses like drc->entity_sense() where we weren't
previously reporting any errors, just the read value, we modify the
function to return RTAS return code, and pass the read value back via
reference.
Michael Roth [Thu, 10 Sep 2015 21:11:03 +0000 (16:11 -0500)]
spapr_drc: don't allow 'empty' DRCs to be unisolated or allocated
Logical resources start with allocation-state:UNUSABLE /
isolation-state:ISOLATED. During hotplug, guests will transition
them to allocation-state:USABLE, and then to
isolation-state:UNISOLATED.
For cases where we cannot transition to allocation-state:USABLE,
in this case due to no device/resource being association with
the logical DRC, we should return an error -3.
For physical DRCs, we default to allocation-state:USABLE and stay
there, so in this case we should report an error -3 when the guest
attempts to make the isolation-state:ISOLATED transition for a DRC
with no device associated.
These are as documented in PAPR 2.7, 13.5.3.4.
We also ensure allocation-state:USABLE when the guest attempts
transition to isolation-state:UNISOLATED to deal with misbehaving
guests attempting to bring online an unallocated logical resource.
This is as documented in PAPR 2.7, 13.7.
Currently we implement no such error logic. Fix this by handling
these error cases as PAPR defines.
Michael Roth [Tue, 15 Sep 2015 21:34:59 +0000 (16:34 -0500)]
spapr_pci: fix device tree props for MSI/MSI-X
PAPR requires ibm,req#msi and ibm,req#msi-x to be present in the
device node to define the number of msi/msi-x interrupts the device
supports, respectively.
Currently we have ibm,req#msi-x hardcoded to a non-sensical constant
that happens to be 2, and are missing ibm,req#msi entirely. The result
of that is that msi-x capable devices get limited to 2 msi-x
interrupts (which can impact performance), and msi-only devices likely
wouldn't work at all. Additionally, if devices expect a minimum that
exceeds 2, the guest driver may fail to load entirely.
SLOF still owns the generation of these properties at boot-time
(although other device properties have since been offloaded to QEMU),
but for hotplugged devices we rely on the values generated by QEMU
and thus hit the limitations above.
Fix this by generating these properties in QEMU as expected by guests.
In the future it may make sense to modify SLOF to pass through these
values directly as we do with other props since we're duplicating SLOF
code.
For setting debug watchpoints, sPAPR guests use H_SET_MODE hypercall.
The existing QEMU H_SET_MODE handler does not support this but
the KVM handler in HV KVM does. However it is not enabled.
This enables the in-kernel H_SET_MODE handler which handles:
- Completed Instruction Address Breakpoint Register
- Watch point 0 registers.
David Gibson [Tue, 8 Sep 2015 01:21:31 +0000 (11:21 +1000)]
pseries: Fix incorrect calculation of threads per socket for chip-id
The device tree presented to pseries machine type guests includes an
ibm,chip-id property which gives essentially the socket number of each
vcpu core (individual vcpu threads don't get a node in the device
tree).
To calculate this, it uses a vcpus_per_socket variable computed as
(smp_cpus / #sockets). This is correct for the usual case where
smp_cpus == smp_threads * smp_cores * #sockets.
However, you can start QEMU with the number of cores and threads
mismatching the total number of vcpus (whether that _should_ be
permitted is a topic for another day). It's a bit hard to say what
the "real" number of vcpus per socket here is, but for most purposes
(smp_threads * smp_cores) will more meaningfully match how QEMU
behaves with respect to socket boundaries.
pseries: Update SLOF firmware image to qemu-slof-20150813
The changes are:
1. GPT support;
2. Much faster VGA support.
The full changelog is:
> Add missing half word access case to _FASTRMOVE and _FASTMOVE
> Remove unused RMOVE64 stub
> fbuffer: Implement RFILL as an accelerated primitive
> fbuffer: Implement MRMOVE as an accelerated primitive
> fbuffer: Precalculate line length in bytes
> terminal: Disable the terminal-write trace by default
> boot: remove trailing ":" in the bootpath
> ci: implement boot client interface
> boot: bootpath should be complete device path
> fbuffer: Use a smaller cursor
> fbuffer: Improve invert-region helper
> usb-hid: Caps is not always shift
> cas: Increase FDT buffer size to accomodate larger ibm, cas node properties
> README: Update with patch submittion note
> disk-label: add support for booting from GPT FAT partition
> disk-label: introduce helper to check fat filesystem
> introduce 8-byte LE helpers
> disk-label: simplify gpt-prep-partition? routine
> fbuffer: introduce the invert-region-x helper
> fbuffer: introduce the invert-region helper
> fbuffer: simplify address computations in fb8-toggle-cursor
Laurent Vivier [Thu, 13 Aug 2015 12:53:02 +0000 (14:53 +0200)]
pseries: define coldplugged devices as "configured"
When a device is hotplugged, attach() sets "configured" to
false, waiting an action from the OS to configure it and then
to call ibm,configure-connector. On ibm,configure-connector,
the hypervisor sets "configured" to true.
In case of coldplugged device, attach() sets "configured" to
false, but firmware and OS never call the ibm,configure-connector
in this case, so it remains set to false.
It could be harmless, but when we unplug a device, hypervisor
waits the device becomes configured because for it, a not configured
device is a device being configured, so it waits the end of configuration
to unplug it... and it never happens, so it is never unplugged.
This patch set by default coldplugged device to "configured=true",
hotplugged device to "configured=false".
Bharata B Rao [Mon, 31 Aug 2015 23:53:52 +0000 (09:53 +1000)]
spapr_rtas: Prevent QEMU crash during hotplug without a prior device_add
If drmgr is used in the guest to hotplug a device before a device_add
has been issued via the QEMU monitor, QEMU segfaults in configure_connector
call. This occurs due to accessing of NULL FDT which otherwise would have
been created and associated with the DRC during device_add command.
Check for NULL FDT and return failure from configure_connector call.
As per PAPR+, an error value of -9003 seems appropriate for this failure.
Thomas Huth [Tue, 1 Sep 2015 01:29:02 +0000 (11:29 +1000)]
ppc/spapr: Use qemu_log_mask() for hcall_dprintf()
To see the output of the hcall_dprintf statements, you currently have
to enable the DEBUG_SPAPR_HCALLS macro in include/hw/ppc/spapr.h.
This is ugly because a) not every user who wants to debug guest
problems can or wants to recompile QEMU to be able to see such issues,
and b) since this macro is disabled by default, the code in the
hcall_dprintf() brackets tends to bitrot until somebody temporarily
enables that macro again.
Since the hcall_dprintf statements except one indicate guest
problems, let's always use qemu_log_mask(LOG_GUEST_ERROR, ...) for
this macro instead. One spot indicated an unimplemented host feature,
so this is changed into qemu_log_mask(LOG_UNIMP, ...) instead. Now
it's possible to see all those messages by simply adding the CLI
parameter "-d guest_errors,unimp", without the need to re-compile
the binary.
David Gibson [Thu, 3 Sep 2015 00:08:23 +0000 (10:08 +1000)]
spapr_drc: Fix potential undefined behaviour
The DRC_INDEX_ID_MASK macro does a left shift on ~0, which is a signed
quantity, and therefore undefined behaviour according to the C spec. In
particular this causes warnings from the clang sanitizer.
This fixes it by calculating the same mask without using ~0 (I think the
new method is a more common idiom for generating masks anyway). For good
measure I also use 1ULL to force the expression's type to unsigned long
long, which should be good for assigning to anything we're going to want
to.
Andrew Jones [Tue, 1 Sep 2015 01:25:35 +0000 (11:25 +1000)]
spapr: add dumpdtb support
dumpdtb (-machine dumpdtb=<file>) allows one to inspect the generated
device tree of machine types that generate device trees. This is
useful for a) seeing what's there b) debugging/testing device tree
generator patches. It can be used as follows
Sam Bobroff [Tue, 1 Sep 2015 01:24:37 +0000 (11:24 +1000)]
spapr: SPLPAR Characteristics
Improve the SPLPAR Characteristics information:
Add MaxPlatProcs: set to max_cpus, the maximum CPUs that could be
addded to the system.
Add DesMem: set to the initial memory of the system.
Add DesProcs: set to smp_cpus, the inital number of CPUs in the
system.
Sam Bobroff [Tue, 1 Sep 2015 01:23:34 +0000 (11:23 +1000)]
spapr: Add /rtas/ibm,change-msix-capable
QEMU is MSI-X capable and makes it available via ibm,change-msi, so
we should indicate this by adding /rtas/ibm,change-msix-capable to the
device tree.
David Gibson [Wed, 12 Aug 2015 03:15:56 +0000 (13:15 +1000)]
spapr: Create pseries-2.5 machine
Add pseries-2.5 machine version.
Signed-off-by: Bharata B Rao <[email protected]>
[Altered to merge before memory hotplug -- dwg]
[Altered to work with b9f072d01 -- dwg] Signed-off-by: David Gibson <[email protected]>
spapr: Provide an error message when migration fails due to htab_shift mismatch
Include an error message when migration fails due to mismatch in
htab_shift values at source and target. This should provide a bit more
verbose message in addition to the current migration failure message
that reads like:
qemu-system-ppc64: error while loading state for instance 0x0 of device 'spapr/htab'
After this patch, the failure message will look like this:
qemu-system-ppc64: htab_shift mismatch: source 29 target 24
qemu-system-ppc64: error while loading state for instance 0x0 of device 'spapr/htab'
Currently device_del requires that the client provide the
device short ID. device_add allows devices to be created
without giving an ID, at which point there is no way to
delete them with device_del. The QOM object path, however,
provides an alternative way to identify the devices.
Allowing device_del to accept an object path ensures all
devices are deletable regardless of whether they have an
ID.
Peter Maydell [Mon, 21 Sep 2015 21:33:51 +0000 (22:33 +0100)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2015-09-21' into staging
qapi: QMP introspection
# gpg: Signature made Mon 21 Sep 2015 08:59:17 BST using RSA key ID EB918653
# gpg: Good signature from "Markus Armbruster <[email protected]>"
# gpg: aka "Markus Armbruster <[email protected]>"
* remotes/armbru/tags/pull-qapi-2015-09-21: (26 commits)
qapi-introspect: Hide type names
qapi: New QMP command query-qmp-schema for QMP introspection
qapi: Pseudo-type '**' is now unused, drop it
qapi-schema: Fix up misleading specification of netdev_add
qom: Don't use 'gen': false for qom-get, qom-set, object-add
qapi: Introduce a first class 'any' type
qapi: Make output visitor return qnull() instead of NULL
qapi: Improve built-in type documentation
qapi-commands: De-duplicate output marshaling functions
qapi: De-duplicate parameter list generation
qapi: Rename qmp_marshal_input_FOO() to qmp_marshal_FOO()
qapi-commands: Rearrange code
qapi-visit: Rearrange code a bit
qapi: Clean up after recent conversions to QAPISchemaVisitor
qapi: Replace dirty is_c_ptr() by method c_null()
qapi-event: Convert to QAPISchemaVisitor, fixing data with base
qapi-event: Eliminate global variable event_enum_value
qapi: De-duplicate enum code generation
qapi-commands: Convert to QAPISchemaVisitor
qapi-visit: Convert to QAPISchemaVisitor, fixing bugs
...
Peter Maydell [Mon, 21 Sep 2015 18:42:33 +0000 (19:42 +0100)]
Merge remote-tracking branch 'remotes/aurel/tags/pull-tcg-mips-20150921' into staging
TCG MIPS queue
- Fixes for 64-bit guests
- Small cleanups
# gpg: Signature made Sun 20 Sep 2015 23:33:15 BST using RSA key ID 1DDD8C9B
# gpg: Good signature from "Aurelien Jarno <[email protected]>"
# gpg: aka "Aurelien Jarno <[email protected]>"
# gpg: aka "Aurelien Jarno <[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: 7746 2642 A9EF 94FD 0F77 196D BA9C 7806 1DDD 8C9B
* remotes/aurel/tags/pull-tcg-mips-20150921:
tcg/mips: pass oi to tcg_out_tlb_load
tcg/mips: move tcg_out_addsub2
tcg/mips: Fix clobbering of qemu_ld inputs
Peter Maydell [Mon, 21 Sep 2015 16:01:46 +0000 (17:01 +0100)]
Merge remote-tracking branch 'remotes/agraf/tags/signed-ppc-for-upstream' into staging
Patch queue for ppc - 2015-09-20
Highlights this time around:
- e500: Fix u-boot boot with -M virt by updating to new version
- e500: fix ATMU reads
- book3s: Fixes (unaligned exceptions, vector instructions)
- yet another dbdma ide fix
I'm out taking care of my son for the next 2 months. During that time
please consider David Gibson the interim ppc queue maintainer. I'm sure
Aurelien will be more than happy to help him review patches as well ;-).
# gpg: Signature made Sun 20 Sep 2015 21:51:16 BST using RSA key ID 03FEDC60
# gpg: Good signature from "Alexander Graf <[email protected]>"
# gpg: aka "Alexander Graf <[email protected]>"
* remotes/agraf/tags/signed-ppc-for-upstream:
target-ppc: fix xscmpodp and xscmpudp decoding
target-ppc: fix vcipher, vcipherlast, vncipherlast and vpermxor
PPC: E500: Update u-boot to commit 79c884d7e4
target-ppc: Fix SRR0 when taking unaligned exceptions
PPC: e500 pci host: Fix ATMUs register reads
mac_dbdma: always clear FLUSH bit once DBDMA channel flush is complete
kvm_ppc: remove kvmppc_timer_hack
qapi: New QMP command query-qmp-schema for QMP introspection
qapi/introspect.json defines the introspection schema. It's designed
for QMP introspection, but should do for similar uses, such as QGA.
The introspection schema does not reflect all the rules and
restrictions that apply to QAPI schemata. A valid QAPI schema has an
introspection value conforming to the introspection schema, but the
converse is not true.
Introspection lowers away a number of schema details, and makes
implicit things explicit:
* The built-in types are declared with their JSON type.
All integer types are mapped to 'int', because how many bits we use
internally is an implementation detail. It could be pressed into
external interface service as very approximate range information,
but that's a bad idea. If we need range information, we better do
it properly.
* Implicit type definitions are made explicit, and given
auto-generated names:
- Array types, named by appending "List" to the name of their
element type, like in generated C.
- The enumeration types implicitly defined by simple union types,
named by appending "Kind" to the name of their simple union type,
like in generated C.
- Types that don't occur in generated C. Their names start with ':'
so they don't clash with the user's names.
* All type references are by name.
* The struct and union types are generalized into an object type.
* Base types are flattened.
* Commands take a single argument and return a single result.
Dictionary argument or list result is an implicit type definition.
The empty object type is used when a command takes no arguments or
produces no results.
The argument is always of object type, but the introspection schema
doesn't reflect that.
The 'gen': false directive is omitted as implementation detail.
The 'success-response' directive is omitted as well for now, even
though it's not an implementation detail, because it's not used by
QMP.
* Events carry a single data value.
Implicit type definition and empty object type use, just like for
commands.
The value is of object type, but the introspection schema doesn't
reflect that.
* Types not used by commands or events are omitted.
Indirect use counts as use.
* Optional members have a default, which can only be null right now
Instead of a mandatory "optional" flag, we have an optional default.
No default means mandatory, default null means optional without
default value. Non-null is available for optional with default
(possible future extension).
* Clients should *not* look up types by name, because type names are
not ABI. Look up the command or event you're interested in, then
follow the references.
TODO Should we hide the type names to eliminate the temptation?
New generator scripts/qapi-introspect.py computes an introspection
value for its input, and generates a C variable holding it.
It can generate awfully long lines. Marked TODO.
A new test-qmp-input-visitor test case feeds its result for both
tests/qapi-schema/qapi-schema-test.json and qapi-schema.json to a
QmpInputVisitor to verify it actually conforms to the schema.
New QMP command query-qmp-schema takes its return value from that
variable. Its reply is some 85KiBytes for me right now.
If this turns out to be too much, we have a couple of options:
* We can use shorter names in the JSON. Not the QMP style.
* Optionally return the sub-schema for commands and events given as
arguments.
Right now qmp_query_schema() sends the string literal computed by
qmp-introspect.py. To compute sub-schema at run time, we'd have to
duplicate parts of qapi-introspect.py in C. Unattractive.
* Let clients cache the output of query-qmp-schema.
It changes only on QEMU upgrades, i.e. rarely. Provide a command
query-qmp-schema-hash. Clients can have a cache indexed by hash,
and re-query the schema only when they don't have it cached. Even
simpler: put the hash in the QMP greeting.
qapi: Make output visitor return qnull() instead of NULL
Before commit 1d10b44, it crashed. Since then, it returns NULL, with
a FIXME comment. The FIXME is valid: code that assumes QObject *
can't be null exists. I'm not aware of a way to feed this problematic
return value to code that actually chokes on null in the current code,
but the next few commits will create one, failing "make check".
Commit 481b002 solved a very similar problem by introducing a special
null QObject. Using this special null QObject is clearly the right
way to resolve this FIXME, so do that, and update the test
accordingly.
However, the patch isn't quite right: it messes up the reference
counting. After about SIZE_MAX visits, the reference counter
overflows, failing the assertion in qnull_destroy_obj(). Because
that's many orders of magnitude more visits of nulls than we expect,
we take this patch despite its flaws, to get the QMP introspection
stuff in without further delay. We'll want to fix it for real before
the release.
gen_marshal_output() uses its parameter name only for name of the
generated function. Name it after the type being marshaled instead of
its caller, and drop duplicates.
Saves 7 copies of qmp_marshal_output_int() in qemu-ga, and one copy of
qmp_marshal_output_str() in qemu-system-*.
is_c_ptr() looks whether the end of the C text for the type looks like
a pointer. Works, but is fragile.
We now have a better tool: use QAPISchemaType method c_null(). The
initializers for non-pointers become prettier: 0, false or the
enumeration constant with the value 0 instead of {0}.
Duplicated in commit 21cd70d. Yes, we can't import qapi-types, but
that's no excuse. Move the helpers from qapi-types.py to qapi.py, and
replace the duplicates in qapi-event.py.
The generated event enumeration type's lookup table becomes
const-correct (see commit 2e4450f), and uses explicit indexes instead
of relying on order (see commit 912ae9c).
Fixes alternates to generate a visitor for their implicit enumeration
type. None of them are currently used, obviously. Example:
block-core.json's BlockdevRef now generates
visit_type_BlockdevRefKind().
Code is generated in a different order now, and therefore has got a
few new forward declarations. Doesn't matter.
The guard QAPI_VISIT_BUILTIN_VISITOR_DECL is renamed to
QAPI_VISIT_BUILTIN.
The previous commit's two ugly special cases exist here, too. Mark
both TODO.
{ 'struct': 'UserDefZero',
'data': { 'integer': 'int' } }
Patch's effect on UserDefFlatUnion:
struct UserDefFlatUnion {
/* Members inherited from UserDefUnionBase: */
+ int64_t integer;
char *string;
EnumOne enum1;
/* Own members: */
union { /* union tag is @enum1 */
void *data;
UserDefA *value1;
UserDefB *value2;
UserDefB *value3;
};
};
Flat union visitors remain broken. They'll be fixed next.
Code is generated in a different order now, but that doesn't matter.
The two guards QAPI_TYPES_BUILTIN_STRUCT_DECL and
QAPI_TYPES_BUILTIN_CLEANUP_DECL are replaced by just
QAPI_TYPES_BUILTIN.
Two ugly special cases for simple unions now stand out like sore
thumbs:
1. The type tag is named 'type' everywhere, except in generated C,
where it's 'kind'.
2. QAPISchema lowers simple unions to semantically equivalent flat
unions. However, the C generated for a simple unions differs from
the C generated for its equivalent flat union, and we therefore
need special code to preserve that pointless difference for now.
spice: surface switch fast path requires same format too.
Commit "555e72f spice: rework mirror allocation, add no-resize fast path"
adds a fast path for surface switches which does't go through the full
primary surface destroy and re-recreation in case the new surface is
identical to the old one (page-flip). It checks the size only though,
but the format must be identical too. This patch adds the format check.
Commit "0002a51 ui/spice: Support shared surface for most pixman
formats" increases the chance to actually trigger this.
tests/qapi-schema: Convert test harness to QAPISchemaVisitor
The old code prints the result of parsing (list of expression
dictionaries), and partial results of semantic analysis (list of enum
dictionaries, list of struct dictionaries).
The new code prints a trace of a schema visit, i.e. what the back-ends
are going to use. Built-in and array types are omitted, because
they're boring.
The QAPI code generators work with a syntax tree (nested dictionaries)
plus a few symbol tables (also dictionaries) on the side.
They have clearly outgrown these simple data structures. There's lots
of rummaging around in dictionaries, and information is recomputed on
the fly. For the work I'm going to do, I want more clearly defined
and more convenient interfaces.
Going forward, I also want less coupling between the back-ends and the
syntax tree, to make messing with the syntax easier.
Create a bunch of classes to represent QAPI schemata.
Have the QAPISchema initializer call the parser, then walk the syntax
tree to create the new internal representation, and finally perform
semantic analysis.
Shortcut: the semantic analysis still relies on existing check_exprs()
to do the actual semantic checking. All this code needs to move into
the classes. Mark as TODO.
Simple unions are lowered to flat unions. Flat unions and structs are
represented as a more general object type.
Catching name collisions in generated code would be nice. Mark as
TODO.
We generate array types eagerly, even though most of them aren't used.
Mark as TODO.
Nothing uses the new intermediate representation just yet, thus no
change to generated files.
The xscmpodp and xscmpudp instructions only have the AX, BX bits in
there encoding, the lowest bit (usually TX) is marked as an invalid
bit. We therefore can't decode them with GEN_XX2FORM, which decodes
the two lowest bit.
Introduce a new form GEN_XX2FORM, which decodes AX and BX and mark
the lowest bit as invalid.
target-ppc: fix vcipher, vcipherlast, vncipherlast and vpermxor
For vector instructions, the helpers get pointers to the vector register
in arguments. Some operands might point to the same register, including
the operand holding the result.
When emulating instructions which access the vector elements in a
non-linear way, we need to store the result in an temporary variable.
The current U-Boot binary in QEMU has a bug where it fails to support
dynamic CCSR addressing. Without this support, u-boot can not boot the
ppce500 machine anymore. This has been fixed upstream in u-boot commit e834975b.
Update the u-boot blob we carry in QEMU to the latest u-boot upstream,
so that we can successfully run u-boot with the ppce500 machine again.
Rudolf Marek [Fri, 14 Aug 2015 11:38:55 +0000 (13:38 +0200)]
PPC: e500 pci host: Fix ATMUs register reads
There is a bug in the register mask when reading
the ATMUs registers. As the result some registers
cannot be read, and read is aliased to the other
registers. Fix it.
Mark Cave-Ayland [Sun, 23 Aug 2015 10:50:55 +0000 (11:50 +0100)]
mac_dbdma: always clear FLUSH bit once DBDMA channel flush is complete
The code to flush the DBDMA channel was effectively duplicated in
dbdma_control_write(), except for the fact that the copy executed outside of a
RUN bit transition was broken by not clearing the FLUSH bit once the flush was
complete.
Newer PPC Linux kernels would timeout waiting for the FLUSH bit to clear again
after submitting a FLUSH command. Fix this by always clearing the FLUSH bit
once the channel flush is complete and removing the repeated code.
Peter Maydell [Sat, 19 Sep 2015 14:59:52 +0000 (15:59 +0100)]
Merge remote-tracking branch 'remotes/afaerber/tags/qom-devices-for-peter' into staging
QOM infrastructure fixes and device conversions
* QOM API error handling fixes
* Performance improvements for device GPIO property creation
* Remaining conversion of QEMUMachine to QOM
# gpg: Signature made Sat 19 Sep 2015 15:40:44 BST using RSA key ID 3E7E013F
# gpg: Good signature from "Andreas Färber <[email protected]>"
# gpg: aka "Andreas Färber <[email protected]>"
* remotes/afaerber/tags/qom-devices-for-peter: (21 commits)
machine: Eliminate QEMUMachine and qemu_register_machine()
Revert use of DEFINE_MACHINE() for registrations of multiple machines
Use DEFINE_MACHINE() to register all machines
mac_world: Break long line
machine: DEFINE_MACHINE() macro
exynos4: Declare each QEMUMachine as a separate variable
exynos4: Use MachineClass instead of exynos4_machines array
exynos4: Use EXYNOS4210_NCPUS instead of max_cpus on error message
machine: Set MachineClass::name automatically
machine: Ensure all TYPE_MACHINE subclasses have the right suffix
mac99: Use MACHINE_TYPE_NAME to encode class name
s390: Rename s390-ccw-virtio-2.4 class name to use MACHINE_TYPE_NAME
s390-virtio: Rename machine class name to use MACHINE_TYPE_NAME
pseries: Rename machine class names to use MACHINE_TYPE_NAME
arm: Rename virt machine class to use MACHINE_TYPE_NAME
vexpress: Rename machine classes to use MACHINE_TYPE_NAME
vexpress: Don't set name on abstract class
machine: MACHINE_TYPE_NAME macro
qdev: Do not use slow [*] expansion for GPIO creation
qom: Fix invalid error check in property_get_str()
...
Andreas Färber [Sat, 19 Sep 2015 08:49:44 +0000 (10:49 +0200)]
Revert use of DEFINE_MACHINE() for registrations of multiple machines
The script used for converting from QEMUMachine had used one
DEFINE_MACHINE() per machine registered. In cases where multiple
machines are registered from one source file, avoid the excessive
generation of module init functions by reverting this unrolling.
The macro will allow easy registration of a TYPE_MACHINE subclass, using
only the machine name and a MachineClass initialization function as
parameter.
Eduardo Habkost [Thu, 20 Aug 2015 21:54:36 +0000 (14:54 -0700)]
machine: Set MachineClass::name automatically
Now all TYPE_MACHINE subclasses use MACHINE_TYPE_NAME to generate the
class name. So instead of requiring each subclass to set
MachineClass::name manually, we can now set it automatically at the
TYPE_MACHINE class_base_init() function.
Signed-off-by: Eduardo Habkost <[email protected]> Reviewed-by: Marcel Apfelbaum <[email protected]>
[AF/ehabkost: Updated for s390-ccw machines]
[AF: Cleanup of intermediate virt and vexpress name handling] Signed-off-by: Andreas Färber <[email protected]>
Eduardo Habkost [Thu, 20 Aug 2015 21:54:33 +0000 (14:54 -0700)]
s390: Rename s390-ccw-virtio-2.4 class name to use MACHINE_TYPE_NAME
Machine class names should use the "-machine" suffix to allow
class-name-based machine class lookup to work. Rename the
s390-ccw-virtio-2.4 machine class using the MACHINE_TYPE_NAME macro.
Eduardo Habkost [Thu, 20 Aug 2015 21:54:32 +0000 (14:54 -0700)]
s390-virtio: Rename machine class name to use MACHINE_TYPE_NAME
Machine class names should use the "-machine" suffix to allow
class-name-based machine class lookup to work. Rename the s390-virtio
machine class using the MACHINE_TYPE_NAME macro.
Eduardo Habkost [Thu, 20 Aug 2015 21:54:31 +0000 (14:54 -0700)]
pseries: Rename machine class names to use MACHINE_TYPE_NAME
Machine class names should use the "-machine" suffix to allow
class-name-based machine class lookup to work. Rename the the pseries
machine classes using the MACHINE_TYPE_NAME macro.
Eduardo Habkost [Thu, 20 Aug 2015 21:54:30 +0000 (14:54 -0700)]
arm: Rename virt machine class to use MACHINE_TYPE_NAME
Machine class names should use the "-machine" suffix to allow
class-name-based machine class lookup to work. Rename the arm virt
machine class using the MACHINE_TYPE_NAME macro.
Eduardo Habkost [Thu, 20 Aug 2015 21:54:29 +0000 (14:54 -0700)]
vexpress: Rename machine classes to use MACHINE_TYPE_NAME
Machine class names should use the "-machine" suffix to allow
class-name-based machine class lookup to work. Rename the vexpress
machine classes using the MACHINE_TYPE_NAME macro.
Eduardo Habkost [Thu, 20 Aug 2015 21:54:28 +0000 (14:54 -0700)]
vexpress: Don't set name on abstract class
The MachineClass::name field won't be ever be used on TYPE_VEXPRESS, as
it is an abstract class and the machine class lookup code explicitly
skips abstract classes. We can remove it to make the code simpler.
Instead of computing mem_index and s_bits in both tcg_out_qemu_ld and
tcg_out_qemu_st function and passing them to tcg_out_tlb_load, directly
pass oi to the tcg_out_tlb_load function and compute mem_index and
s_bits there.
James Hogan [Mon, 14 Sep 2015 10:34:54 +0000 (11:34 +0100)]
tcg/mips: Fix clobbering of qemu_ld inputs
The MIPS TCG backend implements qemu_ld with 64-bit targets using the v0
register (base) as a temporary to load the upper half of the QEMU TLB
comparator (see line 5 below), however this happens before the input
address is used (line 8 to mask off the low bits for the TLB
comparison, and line 12 to add the host-guest offset). If the input
address (addrl) also happens to have been placed in v0 (as in the second
column below), it gets clobbered before it is used.
addrl in t2 addrl in v0
1 srl a0,t2,0x7 srl a0,v0,0x7
2 andi a0,a0,0x1fe0 andi a0,a0,0x1fe0
3 addu a0,a0,s0 addu a0,a0,s0
4 lw at,9136(a0) lw at,9136(a0) set TCG_TMP0 (at)
5 lw v0,9140(a0) lw v0,9140(a0) set base (v0)
6 li t9,-4093 li t9,-4093
7 lw a0,9160(a0) lw a0,9160(a0) set addend (a0)
8 and t9,t9,t2 and t9,t9,v0 use addrl
9 bne at,t9,0x836d8c8 bne at,t9,0x836d838 use TCG_TMP0
10 nop nop
11 bne v0,t8,0x836d8c8 bne v0,a1,0x836d838 use base
12 addu v0,a0,t2 addu v0,a0,v0 use addrl, addend
13 lw t0,0(v0) lw t0,0(v0)
Fix by using TCG_TMP0 (at) as the temporary instead of v0 (base),
pushing the load on line 5 forward into the delay slot of the low
comparison (line 10). The early load of the addend on line 7 also needs
pushing even further for 64-bit targets, or it will clobber a0 before
we're done with it. The output for 32-bit targets is unaffected.
Pavel Fedin [Fri, 31 Jul 2015 12:23:22 +0000 (15:23 +0300)]
qdev: Do not use slow [*] expansion for GPIO creation
Expansion of [*] suffix is very slow because index expansion is done using
trial and error strategy, starting every time from zero and retrying with
the next index until insertion succeeds. With large number of already added
properties this process takes huge amount of time (O(n^2) complexity).
Some architectures (like ARM) use very large amount of IRQ pins in interrupt
controller models. This flaw makes machine startup extremely slow
(~20 seconds for ARM64 with 32 CPUs). This patch decreases this time down to
~10 seconds.
Also in qdev_init_gpio_out_named() memset() is now called only once for the
whole array instead of per-cell cleaning
projects / qemu.git / log