John Snow [Fri, 5 Feb 2016 18:12:33 +0000 (13:12 -0500)]
qemu-img: initialize MapEntry object
Commit 16b0d555 introduced an issue where we are not initializing
has_filename for the 'next' MapEntry object, which leads to interesting
errors in both Valgrind and Clang -fsanitize=undefined.
Zero the stack object at allocation AND make sure the utility to
populate the fields properly marks has_filename as false if applicable.
* remotes/awilliam/tags/vfio-update-20160219.1:
vfio/pci: use PCI_MSIX_FLAGS on retrieving the MSIX entries
hw/arm/sysbus-fdt: remove qemu_fdt_setprop returned value check
hw/arm/sysbus-fdt: enable amd-xgbe dynamic instantiation
hw/arm/sysbus-fdt: helpers for clock node generation
device_tree: qemu_fdt_getprop_cell converted to use the error API
device_tree: qemu_fdt_getprop converted to use the error API
device_tree: introduce qemu_fdt_node_path
device_tree: introduce load_device_tree_from_sysfs
hw/vfio/platform: amd-xgbe device
vfio/pci: replace 1 with PCI_CAP_LIST_NEXT to make code self-explain
pcie_aer: expose pcie_aer_msg() interface
aer: impove pcie_aer_init to support vfio device
vfio: make the 4 bytes aligned for capability size
pcie: modify the capability size assert
* remotes/pmaydell/tags/pull-softfloat-20160219:
MAINTAINERS: Add section for FPU emulation
osdep.h: Remove int_fast*_t Solaris compatibility code
fpu: Use plain 'int' rather than 'int_fast16_t' for exponents
fpu: Use plain 'int' rather than 'int_fast16_t' for shift counts
fpu: Remove use of int_fast16_t in conversions to int16
target-mips: Stop using uint_fast*_t types in r4k_tlb_t struct
This patch allows the instantiation of the vfio-amd-xgbe device
from the QEMU command line (-device vfio-amd-xgbe,host="<device>").
The guest is exposed with a device tree node that combines the description
of both XGBE and PHY (representation supported from 4.2 onwards kernel):
Documentation/devicetree/bindings/net/amd-xgbe.txt.
There are 5 register regions, 6 interrupts including 4 optional
edge-sensitive per-channel interrupts.
Some property values are inherited from host device tree. Host device tree
must feature a combined XGBE/PHY representation (>= 4.2 host kernel).
2 clock nodes (dma and ptp) also are created. It is checked those clocks
are fixed on host side.
AMD XGBE node creation function has a dependency on vfio Linux header and
more generally node creation function for VFIO platform devices only make
sense with CONFIG_LINUX so let's protect this code with #ifdef CONFIG_LINUX.
Eric Auger [Fri, 19 Feb 2016 16:42:31 +0000 (09:42 -0700)]
hw/arm/sysbus-fdt: helpers for clock node generation
Some passthrough'ed devices depend on clock nodes. Those need to be
generated in the guest device tree. This patch introduces some helpers
to build a clock node from information retrieved in the host device tree.
- copy_properties_from_host copies properties from a host device tree
node to a guest device tree node
- fdt_build_clock_node builds a guest clock node and checks the host
fellow clock is a fixed one.
fdt_build_clock_node will become static as soon as it gets used. A
dummy pre-declaration is needed for compilation of this patch.
Eric Auger [Fri, 19 Feb 2016 16:42:30 +0000 (09:42 -0700)]
device_tree: qemu_fdt_getprop_cell converted to use the error API
This patch aligns the prototype with qemu_fdt_getprop. The caller
can choose whether the function self-asserts on error (passing
&error_fatal as Error ** argument, corresponding to the legacy behavior),
or behaves differently such as simply output a message.
In this later case the caller can use the new lenp parameter to interpret
the error if any.
Eric Auger [Fri, 19 Feb 2016 16:42:30 +0000 (09:42 -0700)]
device_tree: qemu_fdt_getprop converted to use the error API
Current qemu_fdt_getprop exits if the property is not found. It is
sometimes needed to read an optional property, in which case we do
not wish to exit but simply returns a null value.
This patch converts qemu_fdt_getprop to accept an Error **, and existing
users are converted to pass &error_fatal. This preserves the existing
behaviour. Then to use the API with your optional semantic a null
parameter can be conveyed.
Eric Auger [Fri, 19 Feb 2016 16:42:29 +0000 (09:42 -0700)]
hw/vfio/platform: amd-xgbe device
This patch introduces the amd-xgbe VFIO platform device. It
allows the guest to do passthrough on a device exposing an
"amd,xgbe-seattle-v1a" compat string.
Chen Fan [Fri, 19 Feb 2016 16:42:28 +0000 (09:42 -0700)]
aer: impove pcie_aer_init to support vfio device
pcie_aer_init was used to emulate an aer capability for pcie device,
but for vfio device, the aer config space size is mutable and is not
always equal to PCI_ERR_SIZEOF(0x48). it depends on where the TLP Prefix
register required, so here we add a size argument.
Peter Maydell [Fri, 19 Feb 2016 16:25:01 +0000 (16:25 +0000)]
MAINTAINERS: Add section for FPU emulation
Add an entry to the MAINTAINERS file for our softfloat FPU
emulation code. This code is only 'odd fixes' but it's useful to
record who to cc on patches to it.
Peter Maydell [Fri, 19 Feb 2016 16:25:01 +0000 (16:25 +0000)]
osdep.h: Remove int_fast*_t Solaris compatibility code
We now do not use the int_fast*_t types anywhere in QEMU, so we can
remove the compatibility definitions we were providing for the
benefit of ancient Solaris versions.
Peter Maydell [Fri, 19 Feb 2016 16:25:01 +0000 (16:25 +0000)]
fpu: Use plain 'int' rather than 'int_fast16_t' for exponents
Use the plain 'int' type rather than 'int_fast16_t' for handling
exponents. Exponents don't need to be exactly 16 bits, so using int16_t
for them would confuse more than it clarified.
This should be a safe change because int_fast16_t semantics
permit use of 'int' (and on 32-bit glibc that is what you get).
Peter Maydell [Fri, 19 Feb 2016 16:25:00 +0000 (16:25 +0000)]
fpu: Use plain 'int' rather than 'int_fast16_t' for shift counts
Use the plain 'int' type rather than 'int_fast16_t' for shift counts
in the various shift related functions, since we don't actually care
about the size of the integer at all here, and using int16_t would
be confusing.
This should be a safe change because int_fast16_t semantics
permit use of 'int' (and on 32-bit glibc that is what you get).
Peter Maydell [Fri, 19 Feb 2016 16:25:00 +0000 (16:25 +0000)]
fpu: Remove use of int_fast16_t in conversions to int16
Make the functions which convert floating point to 16 bit integer
return int16_t rather than int_fast16_t, and correspondingly use
int_fast16_t in their internal implementations where appropriate.
(These functions are used only by the ARM target.)
Peter Maydell [Mon, 25 Jan 2016 17:40:49 +0000 (17:40 +0000)]
target-mips: Stop using uint_fast*_t types in r4k_tlb_t struct
The r4k_tlb_t structure uses the uint_fast*_t types. Most of these
uses are in bitfields and are thus pointless, because the bitfield
itself specifies the width of the type; just use 'unsigned int'
instead. (On glibc uint_fast16_t is defined as either 32 or 64 bits,
so we know the code is not reliant on it being exactly 16 bits.)
There is also one use of uint_fast8_t, which we replace with uint8_t,
because both are exactly 8 bits on glibc and this is the only
place outside the softfloat code which uses an int_fast*_t type.
Peter Maydell [Fri, 19 Feb 2016 15:19:13 +0000 (15:19 +0000)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-error-2016-02-19' into staging
Error reporting patches for 2016-02-19
# gpg: Signature made Fri 19 Feb 2016 12:47:50 GMT using RSA key ID EB918653
# gpg: Good signature from "Markus Armbruster <[email protected]>"
# gpg: aka "Markus Armbruster <[email protected]>"
* remotes/armbru/tags/pull-error-2016-02-19:
vl: Clean up machine selection in main().
vl: Set error location when parsing memory options
replay: Set error location properly when parsing options
vl: Reset location after handling command-line arguments
vl.c: Fix regression in machine error message
Peter Maydell [Fri, 19 Feb 2016 14:18:21 +0000 (14:18 +0000)]
Merge remote-tracking branch 'remotes/armbru/tags/pull-qapi-2016-02-19' into staging
QAPI patches for 2016-02-19
# gpg: Signature made Fri 19 Feb 2016 10:10:18 GMT using RSA key ID EB918653
# gpg: Good signature from "Markus Armbruster <[email protected]>"
# gpg: aka "Markus Armbruster <[email protected]>"
* remotes/armbru/tags/pull-qapi-2016-02-19:
qapi: Change visit_start_implicit_struct to visit_start_alternate
qapi: Don't box branches of flat unions
qapi: Don't box struct branch of alternate
qapi-visit: Use common idiom in gen_visit_fields_decl()
qapi: Emit structs used as variants in topological order
qapi: Adjust layout of FooList types
qapi-visit: Less indirection in visit_type_Foo_fields()
qapi-visit: Unify struct and union visit
qapi: Visit variants in visit_type_FOO_fields()
qapi-visit: Simplify how we visit common union members
qapi: Add tests of complex objects within alternate
qapi: Forbid 'any' inside an alternate
qapi: Forbid empty unions and useless alternates
qapi: Simplify excess input reporting in input visitors
qapi-visit: Honor prefix of discriminator enum
We set machine_class to the default first, and update it to the real
one later. Any use of machine_class in between is almost certainly
wrong (there are no such uses right now). Set it once and for all
instead.
Eduardo Habkost [Fri, 12 Feb 2016 19:02:25 +0000 (17:02 -0200)]
vl: Reset location after handling command-line arguments
After looping through all command-line arguments, error location
info becomes obsolete, and any function calling error_report()
will print misleading information. This breaks error reporting
for some option handling, like:
Marcel Apfelbaum [Fri, 12 Feb 2016 19:02:24 +0000 (17:02 -0200)]
vl.c: Fix regression in machine error message
Commit e1ce0c3cb (vl.c: fix regression when reading machine type
from config file) fixed the error message when the machine type
was supplied inside the config file. However now the option name
is not displayed correctly if the error happens when the machine
is specified at command line.
Running
./x86_64-softmmu/qemu-system-x86_64 -M q35-1.5 -redir tcp:8022::22
will result in the error message:
qemu-system-x86_64: -redir tcp:8022::22: unsupported machine type
Use -machine help to list supported machines
Fixed it by restoring the error location and also extracted the code
dealing with machine options into a separate function.
Peter Maydell [Fri, 19 Feb 2016 10:50:37 +0000 (10:50 +0000)]
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
vhost, virtio, pci, pxe
Fixes all over the place.
New tests for pxe.
Signed-off-by: Michael S. Tsirkin <[email protected]>
# gpg: Signature made Thu 18 Feb 2016 15:46:39 GMT using RSA key ID D28D5469
# gpg: Good signature from "Michael S. Tsirkin <[email protected]>"
# gpg: aka "Michael S. Tsirkin <[email protected]>"
* remotes/mst/tags/for_upstream:
tests/vhost-user-bridge: add scattering of incoming packets
vhost-user interrupt management fixes
rules: filter out irrelevant files
change type of pci_bridge_initfn() to void
dec: convert to realize()
tests: add pxe e1000 and virtio-pci tests
msix: fix msix_vector_masked
virtio: optimize virtio_access_is_big_endian() for little-endian targets
vhost: simplify vhost_needs_vring_endian()
vhost: move virtio 1.0 check to cross-endian helper
virtio: move cross-endian helper to vhost
vhost-net: revert support of cross-endian vnet headers
virtio-net: use the backend cross-endian capabilities
Eric Blake [Thu, 18 Feb 2016 06:48:29 +0000 (23:48 -0700)]
qapi: Change visit_start_implicit_struct to visit_start_alternate
After recent changes, the only remaining use of
visit_start_implicit_struct() is for allocating the space needed
when visiting an alternate. Since the term 'implicit struct' is
hard to explain, rename the function to its current usage. While
at it, we can merge the functionality of visit_get_next_type()
into the same function, making it more like visit_start_struct().
Eric Blake [Thu, 18 Feb 2016 06:48:27 +0000 (23:48 -0700)]
qapi: Don't box branches of flat unions
There's no reason to do two malloc's for a flat union; let's just
inline the branch struct directly into the C union branch of the
flat union.
Surprisingly, fewer clients were actually using explicit references
to the branch types in comparison to the number of flat unions
thus modified.
This lets us reduce the hack in qapi-types:gen_variants() added in
the previous patch; we no longer need to distinguish between
alternates and flat unions.
The change to unboxed structs means that u.data (added in commit cee2dedb) is now coincident with random fields of each branch of
the flat union, whereas beforehand it was only coincident with
pointers (since all branches of a flat union have to be objects).
Note that this was already the case for simple unions - but there
we got lucky. Remember, visit_start_union() blindly returns true
for all visitors except for the dealloc visitor, where it returns
the value !!obj->u.data, and that this result then controls
whether to proceed with the visit to the variant. Pre-patch,
this meant that flat unions were testing whether the boxed pointer
was still NULL, and thereby skipping visit_end_implicit_struct()
and avoiding a NULL dereference if the pointer had not been
allocated. The same was true for simple unions where the current
branch had pointer type, except there we bypassed visit_type_FOO().
But for simple unions where the current branch had scalar type, the
contents of that scalar meant that the decision to call
visit_type_FOO() was data-dependent - the reason we got lucky there
is that visit_type_FOO() for all scalar types in the dealloc visitor
is a no-op (only the pointer variants had anything to free), so it
did not matter whether the dealloc visit was skipped. But with this
patch, we would risk leaking memory if we could skip a call to
visit_type_FOO_fields() based solely on a data-dependent decision.
But notice: in the dealloc visitor, visit_type_FOO() already handles
a NULL obj - it was only the visit_type_implicit_FOO() that was
failing to check for NULL. And now that we have refactored things to
have the branch be part of the parent struct, we no longer have a
separate pointer that can be NULL in the first place. So we can just
delete the call to visit_start_union() altogether, and blindly visit
the branch type; there is no change in behavior except to the dealloc
visitor, where we now unconditionally visit the branch, but where that
visit is now always safe (for a flat union, we can no longer
dereference NULL, and for a simple union, visit_type_FOO() was already
safely handling NULL on pointer types).
Unfortunately, simple unions are not as easy to switch to unboxed
layout; because we are special-casing the hidden implicit type with
a single 'data' member, we really DO need to keep calling another
layer of visit_start_struct(), with a second malloc; although there
are some cleanups planned for simple unions in later patches.
visit_start_union() and gen_visit_implicit_struct() are now unused.
Drop them.
Note that after this patch, the only remaining use of
visit_start_implicit_struct() is for alternate types; the next patch
will do further cleanup based on that fact.
Eric Blake [Thu, 18 Feb 2016 06:48:26 +0000 (23:48 -0700)]
qapi: Don't box struct branch of alternate
There's no reason to do two malloc's for an alternate type visiting
a QAPI struct; let's just inline the struct directly as the C union
branch of the struct.
Surprisingly, no clients were actually using the struct member prior
to this patch outside of the testsuite; an earlier patch in the series
added some testsuite coverage to make the effect of this patch more
obvious.
In qapi.py, c_type() gains a new is_unboxed flag to control when we
are emitting a C struct unboxed within the context of an outer
struct (different from our other two modes of usage with no flags
for normal local variable declarations, and with is_param for adding
'const' in a parameter list). I don't know if there is any more
pythonic way of collapsing the two flags into a single parameter,
as we never have a caller setting both flags at once.
Ultimately, we want to also unbox branches for QAPI unions, but as
that touches a lot more client code, it is better as separate
patches. But since unions and alternates share gen_variants(), I
had to hack in a way to test if we are visiting an alternate type
for setting the is_unboxed flag: look for a non-object branch.
This works because alternates have at least two branches, with at
most one object branch, while unions have only object branches.
The hack will go away in a later patch.
The generated code difference to qapi-types.h is relatively small:
| struct BlockdevRef {
| QType type;
| union { /* union tag is @type */
| void *data;
|- BlockdevOptions *definition;
|+ BlockdevOptions definition;
| char *reference;
| } u;
| };
The corresponding spot in qapi-visit.c calls visit_type_FOO(), which
first calls visit_start_struct() to allocate or deallocate the member
and handle a layer of {} from the JSON stream, then visits the
members. To peel off the indirection and the memory management that
comes with it, we inline this call, then suppress allocation /
deallocation by passing NULL to visit_start_struct(), and adjust the
member visit:
Eric Blake [Thu, 18 Feb 2016 06:48:25 +0000 (23:48 -0700)]
qapi-visit: Use common idiom in gen_visit_fields_decl()
We have several instances of methods that do an early exit if
output is not needed, then log that output is being generated,
and finally produce the output; see qapi-types.py:gen_object()
and qapi-visit.py:gen_visit_implicit_struct(). The odd man
out was gen_visit_fields_decl(); rearrange it to be more like
the others. No semantic change or difference to generated code.
Eric Blake [Thu, 18 Feb 2016 06:48:24 +0000 (23:48 -0700)]
qapi: Emit structs used as variants in topological order
Right now, we emit the branches of union types as a boxed pointer,
and it suffices to have a forward declaration of the type. However,
a future patch will swap things to directly use the branch type,
instead of hiding it behind a pointer. For this to work, the
compiler needs the full definition of the type, not just a forward
declaration, prior to the union that is including the branch type.
This patch just adds topological sorting to hoist all types
mentioned in a branch of a union to be fully declared before the
union itself. The sort is always possible, because we do not
allow circular union types that include themselves as a direct
branch (it is, however, still possible to include a branch type
that itself has a pointer to the union, for a type that can
indirectly recursively nest itself - that remains safe, because
that the member of the branch type will remain a pointer, and the
QMP representation of such a type adds another {} for each recurring
layer of the union type).
Eric Blake [Thu, 18 Feb 2016 06:48:23 +0000 (23:48 -0700)]
qapi: Adjust layout of FooList types
By sticking the next pointer first, we don't need a union with
64-bit padding for smaller types. On 32-bit platforms, this
can reduce the size of uint8List from 16 bytes (or 12, depending
on whether 64-bit ints can tolerate 4-byte alignment) down to 8.
It has no effect on 64-bit platforms (where alignment still
dictates a 16-byte struct); but fewer anonymous unions is still
a win in my book.
It requires visit_next_list() to gain a size parameter, to know
what size element to allocate; comparable to the size parameter
of visit_start_struct().
I debated about going one step further, to allow for fewer casts,
by doing:
typedef GenericList GenericList;
struct GenericList {
GenericList *next;
};
struct FooList {
GenericList base;
Foo *value;
};
so that you convert to 'GenericList *' by '&foolist->base', and
back by 'container_of(generic, GenericList, base)' (as opposed to
the existing '(GenericList *)foolist' and '(FooList *)generic').
But doing that would require hoisting the declaration of
GenericList prior to inclusion of qapi-types.h, rather than its
current spot in visitor.h; it also makes iteration a bit more
verbose through 'foolist->base.next' instead of 'foolist->next'.
Note that for lists of objects, the 'value' payload is still
hidden behind a boxed pointer. Someday, it would be nice to do:
struct FooList {
FooList *next;
Foo value;
};
for one less level of malloc for each list element. This patch
is a step in that direction (now that 'next' is no longer at a
fixed non-zero offset within the struct, we can store more than
just a pointer's-worth of data as the value payload), but the
actual conversion would be a task for another series, as it will
touch a lot of code.
Eric Blake [Thu, 18 Feb 2016 06:48:22 +0000 (23:48 -0700)]
qapi-visit: Less indirection in visit_type_Foo_fields()
We were passing 'Foo **obj' to the internal helper function, but
all uses within the helper were via reads of '*obj'. Refactor
things to pass one less level of indirection, by having the
callers dereference before calling.
The refactoring will also make it easier to reuse the helpers in
a future patch when implicit structs are stored directly in the
parent struct rather than boxed through a pointer.
gen_visit_union() is now just like gen_visit_struct(). Rename
it to gen_visit_object(), use it for structs, and drop
gen_visit_struct(). Output is unchanged.
Eric Blake [Thu, 18 Feb 2016 06:48:20 +0000 (23:48 -0700)]
qapi: Visit variants in visit_type_FOO_fields()
We initially created the static visit_type_FOO_fields() helper
function for reuse of code - we have cases where the initial
setup for a visit has different allocation (depending on whether
the fields represent a stand-alone type or are embedded as part
of a larger type), but where the actual field visits are
identical once a pointer is available.
Up until the previous patch, visit_type_FOO_fields() was only
used for structs (no variants), so it was covering every field
for each type where it was emitted.
Meanwhile, the code for visiting unions looks like:
which splits the fields of the union visit across two functions.
Move the code to visit variants to live inside visit_type_U_fields(),
while making it conditional on having variants so that all other
instances of the helper function remain unchanged. This is also
a step closer towards unifying struct and union visits, and towards
allowing one union type to be the branch of another flat union.
The resulting diff to the generated code is a bit hard to read,
but it can be verified that it touches only union types, and that
the end result is the following general structure:
For a flat union FU with base B, it generates a visit of its base
fields:
visit_type_B_fields(v, (B **)obj, &err);
Instead, we can simply visit the common members using the same fields
visit function we use for structs, generated with
gen_visit_struct_fields(). This function visits the base if any, then
the local members.
For a simple union SU, visit_type_SU_fields() contains exactly the old
tag member visit, because there is no base, and the tag member is the
only member. For instance, the code generated for qapi-schema.json's
KeyValue changes like this:
For a flat union FU, visit_type_FU_fields() contains exactly the old
base fields visit, because there is a base, but no members. For
instance, the code generated for qapi-schema.json's CpuInfo changes
like this:
As you see, the generated code grows a bit, but in practice, it's lost
in the noise: qapi-schema.json's qapi-visit.c gains roughly 1%.
This simplification became possible with commit 441cbac "qapi-visit:
Convert to QAPISchemaVisitor, fixing bugs". It's a step towards
unifying gen_struct() and gen_union().
Eric Blake [Thu, 18 Feb 2016 06:48:18 +0000 (23:48 -0700)]
qapi: Add tests of complex objects within alternate
Upcoming patches will adjust how we visit an object branch of an
alternate; but we were completely lacking testsuite coverage.
Rectify this, so that the future patches will be able to highlight
the changes and still prove that we avoided regressions.
In particular, the use of a flat union UserDefFlatUnion rather
than a simple struct UserDefA as the branch will give us coverage
of an object with variants. And visiting an alternate as both
the top level and as a nested member gives confidence in correct
memory allocation handling, especially if the test is run under
valgrind.
Eric Blake [Thu, 18 Feb 2016 06:48:17 +0000 (23:48 -0700)]
qapi: Forbid 'any' inside an alternate
The whole point of an alternate is to allow some type-safety while
still accepting more than one JSON type. Meanwhile, the 'any'
type exists to bypass type-safety altogether. The two are
incompatible: you can't accept every type, and still tell which
branch of the alternate to use for the parse; fix this to give a
sane error instead of a Python stack trace.
Note that other types that can't be alternate members are caught
earlier, by check_type().
Eric Blake [Thu, 18 Feb 2016 06:48:16 +0000 (23:48 -0700)]
qapi: Forbid empty unions and useless alternates
Empty unions serve no purpose, and while we compile with gcc
which permits them, strict C99 forbids them. We happen to inject
a dummy 'void *data' member into the C unions that represent QAPI
unions and alternates, but we want to get rid of that member (it
pollutes the namespace for no good reason), which would leave us
with an empty union if the user didn't provide any branches. While
empty structs make sense in QAPI, empty unions don't add any
expressiveness to the QMP language. So prohibit them at parse
time. Update the documentation and testsuite to match.
Note that the documentation already mentioned that alternates
should have "two or more JSON data types"; so this also fixes
the code to enforce that. However, we have existing uses of a
union type with only one branch, so the 2-or-more strictness
is intentionally limited to alternates.
Eric Blake [Thu, 18 Feb 2016 06:48:15 +0000 (23:48 -0700)]
qapi: Simplify excess input reporting in input visitors
When reporting that an unvisited member remains at the end of an
input visit for a struct, we were using g_hash_table_find()
coupled with a callback function that always returns true, to
locate an arbitrary member of the hash table. But if all we
need is an arbitrary entry, we can get that from a single-use
iterator, without needing a tautological callback function.
Technically, our cast of &(GQueue *) to (void **) is not strict
C (while void * must be able to hold all other pointers, nothing
says a void ** has to be the same width or representation as a
GQueue **). The kosher way to write it would be the verbose:
void *tmp;
GQueue *any;
if (g_hash_table_iter_next(&iter, NULL, &tmp)) {
any = tmp;
But our code base (not to mention glib itself) already has other
cases of assuming that ALL pointers have the same width and
representation, where a compiler would have to go out of its way
to mis-compile our borderline behavior.
Eric Blake [Tue, 16 Feb 2016 23:39:25 +0000 (16:39 -0700)]
qapi-visit: Honor prefix of discriminator enum
When we added support for a user-specified prefix for an enum
type (commit 351d36e), we forgot to teach the qapi-visit code
to honor that prefix in the case of using a prefixed enum as
the discriminator for a flat union. While there is still some
on-list debate on whether we want to keep prefixes, we should
at least make it work as long as it is still part of the code
base.
Victor Kaplansky [Thu, 18 Feb 2016 14:45:05 +0000 (16:45 +0200)]
tests/vhost-user-bridge: add scattering of incoming packets
This patch adds to the vubr test the scattering of incoming
packets to the chain of RX buffer. Also, this patch corrects the
size of the header preceding the packet in RX buffers.
Note that this patch doesn't add the support for mergeable
buffers.
Peter Maydell [Thu, 18 Feb 2016 15:20:35 +0000 (15:20 +0000)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20160218-1' into staging
target-arm queue:
* implement or fix various EL3 trap behaviour for system registers
* clean up the trap/undef handling of the SRS instruction
* add some missing AArch64 performance monitor system registers
* implement reset for the PL061 GPIO device
* QOMify sd.c and the pxa2xx_mmci device
* SD card emulation fixes for booting Tianocore UEFI on RPi2
* QOMify various ARM timer devices
* remotes/pmaydell/tags/pull-target-arm-20160218-1: (36 commits)
hw/timer: QOM'ify pxa2xx_timer
hw/timer: QOM'ify pl031
hw/timer: QOM'ify exynos4210_rtc
hw/timer: QOM'ify exynos4210_pwm
hw/timer: QOM'ify exynos4210_mct
hw/timer: QOM'ify arm_timer (pass 2)
hw/timer: QOM'ify arm_timer (pass 1)
hw/sd: use guest error logging rather than fprintf to stderr
hw/sd: model a power-up delay, as a workaround for an EDK2 bug
hw/sd: implement CMD23 (SET_BLOCK_COUNT) for MMC compatibility
hw/sd/pxa2xx_mmci: Add reset function
hw/sd/pxa2xx_mmci: Convert to VMStateDescription
hw/sd/pxa2xx_mmci: Update to use new SDBus APIs
hw/sd/pxa2xx_mmci: convert to SysBusDevice object
sdhci_sysbus: Create SD card device in users, not the device itself
hw/sd/sdhci.c: Update to use SDBus APIs
hw/sd: Add QOM bus which SD cards plug in to
hw/sd/sd.c: Convert sd_reset() function into Device reset method
hw/sd/sd.c: QOMify
hw/sd/sdhci.c: Remove x-drive property
...
xiaoqiang.zhao [Thu, 18 Feb 2016 14:16:20 +0000 (14:16 +0000)]
hw/timer: QOM'ify arm_timer (pass 1)
* assign icp_pit_init to icp_pit_info.instance_init
* split the old SysBus init function into an instance_init
and a Device realize function
* use DeviceClass::realize instead of SysBusDeviceClass::init
Andrew Baumann [Thu, 18 Feb 2016 14:16:20 +0000 (14:16 +0000)]
hw/sd: use guest error logging rather than fprintf to stderr
Some of these errors may be harmless (e.g. probing unimplemented
commands, or issuing CMD12 in the wrong state), and may also be quite
frequent. Spamming the standard error output isn't desirable in such
cases.
Andrew Baumann [Thu, 18 Feb 2016 14:16:20 +0000 (14:16 +0000)]
hw/sd: model a power-up delay, as a workaround for an EDK2 bug
The SD spec for ACMD41 says that a zero argument is an "inquiry"
ACMD41, which does not start initialisation and is used only for
retrieving the OCR. However, Tianocore EDK2 (UEFI) has a bug [1]: it
first sends an inquiry (zero) ACMD41. If that first request returns an
OCR value with the power up bit (0x80000000) set, it assumes the card
is ready and continues, leaving the card in the wrong state. (My
assumption is that this works on hardware, because no real card is
immediately powered up upon reset.)
This change models a delay of 0.5ms from the first ACMD41 to the power
being up. However, it also immediately sets the power on upon seeing a
non-zero (non-enquiry) ACMD41. This speeds up UEFI boot, it should
also account for guests that simply delay after card reset and then
issue an ACMD41 that they expect will succeed.
[1] https://github.com/tianocore/edk2/blob/master/EmbeddedPkg/Universal/MmcDxe/MmcIdentification.c#L279
(This is the loop starting with "We need to wait for the MMC or SD
card is ready")
Peter Maydell [Thu, 18 Feb 2016 14:16:18 +0000 (14:16 +0000)]
sdhci_sysbus: Create SD card device in users, not the device itself
Move the creation of the SD card device from the sdhci_sysbus
device itself into the boards that create these devices.
This allows us to remove the cannot_instantiate_with_device_add
notation because we no longer call drive_get_next in the device
model.
Peter Maydell [Thu, 18 Feb 2016 14:16:18 +0000 (14:16 +0000)]
hw/sd: Add QOM bus which SD cards plug in to
Add a QOM bus for SD cards to plug in to.
Note that since sd_enable() is used only by one board and there
only as part of a broken implementation, we do not provide it in
the SDBus API (but instead add a warning comment about the old
function). Whoever converts OMAP and the nseries boards to QOM
will need to either implement the card switch properly or move
the enable hack into the OMAP MMC controller model.
In the SDBus API, the old-style use of sd_set_cb to register some
qemu_irqs for notification of card insertion and write-protect
toggling is replaced with methods in the SDBusClass which the
card calls on status changes and methods in the SDClass which
the controller can call to find out the current status. The
query methods will allow us to remove the abuse of the 'register
irqs' API by controllers in their reset methods to trigger
the card to tell them about the current status again.
Peter Maydell [Thu, 18 Feb 2016 14:16:18 +0000 (14:16 +0000)]
hw/sd/sd.c: QOMify
Turn the SD card into a QOM device.
This conversion only changes the device itself; the various
functions which are effectively methods on the device are not
touched at this point.
Peter Maydell [Thu, 18 Feb 2016 14:16:18 +0000 (14:16 +0000)]
hw/sd/sdhci.c: Remove x-drive property
The following commits will remove support for the old sdhci-pci
command line syntax using the x-drive property:
-device sdhci-pci,x-drive=mydrive -drive id=mydrive,[...]
and replace it with an explicit sd device:
-device sdhci-pci -drive id=mydrive,[...] -device sd,drive=mydrive
(This is OK because x-drive is experimental.)
This commit removes the x-drive property so that old style
command lines will fail with a reasonable error message:
-device sdhci-pci,x-drive=mydrive: Property '.x-drive' not found
Wei Huang [Thu, 18 Feb 2016 14:16:17 +0000 (14:16 +0000)]
ARM: PL061: Cleaning field of PL061 device state
This patch removes the float_high field of PL061State, which doesn't
seem to be used anywhere. Because this changes the device state, the
version ID is also bumped up for the reason of compatiblity.
Wei Huang [Thu, 18 Feb 2016 14:16:17 +0000 (14:16 +0000)]
ARM: PL061: Clear PL061 device state after reset
Current QEMU doesn't clear PL061 state after reset. This causes a
weird issue with guest reboot via GPIO. Here is the device state
with two reboot requests:
(PL061State fields) data old_in_data istate
VM boot 0 0 0
After 1st ACPI reboot request 8 8 8
After VM PL061 driver ACK 8 8 0
After VM reboot 8 8 0
------------------------------------------------------------
2nd ACPI reboot request 8
In the second reboot request above, because the old_in_data field is 8,
QEMU decides that there is a pending edge IRQ already (see
pl061_update()) in input; so it doesn't raise up IRQ again. As a result
the second reboot request is lost. The correct way is to clear PL061
device state after reset.
The default reset state is found from the documents listed below. Per
Peter's suggestion that QEMU automatically calls reset function after
device initialization, this patch removes calling pl061_reset() from
pl061_initfn().
Reference:
[1] PL061 Technical Reference Manual
[2] Stellaris LM3S8962 Microcontroller Data Sheet
[3] Stellaris LM3S5P31 Microcontroller Data Sheet
Peter Maydell [Thu, 18 Feb 2016 14:16:17 +0000 (14:16 +0000)]
target-arm: UNDEF in the UNPREDICTABLE SRS-from-System case
Make get_r13_banked() raise an exception at runtime for the
corner case of SRS from System mode, so that we can UNDEF it;
this brings us in to line with the ARM ARM's set of permitted
CONSTRAINED UNPREDICTABLE choices.
Peter Maydell [Thu, 18 Feb 2016 14:16:16 +0000 (14:16 +0000)]
target-arm: Combine user-only and softmmu get/set_r13_banked()
The user-mode versions of get/set_r13_banked() exist just to assert
if they're ever called -- the translate time code should never
emit calls to them because SRS from user mode always UNDEF.
There's no code in the softmmu versions that can't compile in
CONFIG_USER_ONLY, and the assertion is not particularly useful,
so combine the two functions rather than having completely split
versions under ifdefs.
Peter Maydell [Thu, 18 Feb 2016 14:16:16 +0000 (14:16 +0000)]
target-arm: Move get/set_r13_banked() to op_helper.c
Move get/set_r13_banked() from helper.c to op_helper.c. This will
let us add exception-raising code to them, and also puts them
in the same file as get/set_user_reg(), which makes some conceptual
sense.
(The original reason for the helper.c/op_helper.c split was that
only op_helper.c had access to the CPU env pointer; this distinction
has not been true for a long time, though, and so the split is
now rather arbitrary.)
Peter Maydell [Thu, 18 Feb 2016 14:16:16 +0000 (14:16 +0000)]
target-arm: Clean up trap/undef handling of SRS
The SRS instruction is:
* UNDEFINED in Hyp mode
* UNPREDICTABLE in User or System mode
* UNPREDICTABLE if the specified mode isn't accessible
* trapped to EL3 if EL3 is AArch64 and we are at Secure EL1
Clean up the code to handle all these cases cleanly, including
picking UNDEF as our choice of UNPREDICTABLE behaviour rather
blindly trusting the mode field passed in the instruction.
As part of this, move the check for IS_USER into gen_srs()
itself rather than having it done by the caller.
The exception is that we don't UNDEF for calls from System
mode, which need a runtime check. This will be dealt with in
the following commits.
Peter Maydell [Thu, 18 Feb 2016 14:16:16 +0000 (14:16 +0000)]
target-arm: Report correct syndrome for FPEXC32_EL2 traps
If access to FPEXC32_EL2 is trapped by CPTR_EL2.TFP or CPTR_EL3.TFP,
this should be reported with a syndrome register indicating an
FP access trap, not one indicating a system register access trap.
Peter Maydell [Thu, 18 Feb 2016 14:16:15 +0000 (14:16 +0000)]
target-arm: Implement MDCR_EL3.TDOSA and MDCR_EL2.TDOSA traps
Implement the traps to EL2 and EL3 controlled by the bits
MDCR_EL2.TDOSA MDCR_EL3.TDOSA. These can configurably trap
accesses to the "powerdown debug" registers.
Peter Maydell [Thu, 18 Feb 2016 14:16:15 +0000 (14:16 +0000)]
target-arm: Fix handling of SCR.SMD
We weren't quite implementing the handling of SCR.SMD correctly.
The condition governing whether the SMD bit should apply only
for NS state is "is EL3 is AArch32", not "is the current EL AArch32".
Fix the condition, and clarify the comment both to reflect this and
to expand slightly on what's going on for the v7-no-Virtualization case.
Peter Maydell [Thu, 18 Feb 2016 14:16:15 +0000 (14:16 +0000)]
target-arm: correct CNTFRQ access rights
Correct some corner cases we were getting wrong for
CNTFRQ access rights:
* should UNDEF from 32-bit Secure EL1
* only writable from the highest implemented exception level,
which might not be EL1 now
To clarify the code, provide a new utility function
arm_highest_el() which returns the highest implemented
exception level.
Victor Kaplansky [Thu, 18 Feb 2016 14:12:23 +0000 (16:12 +0200)]
vhost-user interrupt management fixes
Since guest_mask_notifier can not be used in vhost-user mode due
to buffering implied by unix control socket, force
use_mask_notifier on virtio devices of vhost-user interfaces, and
send correct callfd to the guest at vhost start.
Using guest_notifier_mask function in vhost-user case may
break interrupt mask paradigm, because mask/unmask is not
really done when returning from guest_notifier_mask call, instead
message is posted in a unix socket, and processed later.
Add an option boolean flag 'use_mask_notifier' to disable the use
of guest_notifier_mask in virtio pci.
Peter Maydell [Thu, 18 Feb 2016 10:29:47 +0000 (10:29 +0000)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-2.6-20160218' into staging
ppc patch queue for 2016-02-18
Currently accumulated patches for target-ppc, pseries machine type and
related devices.
* Some cleanups to management of SDR1 and the hashed page table
* Implementations of a number of simple PAPR hypercalls
* Significant improvements to the Macintosh CUDA device
* Several bugfixes
# gpg: Signature made Thu 18 Feb 2016 04:16:51 GMT using RSA key ID 20D9B392
# gpg: Good signature from "David Gibson <[email protected]>"
# gpg: aka "David Gibson (Red Hat) <[email protected]>"
# gpg: aka "David Gibson (ozlabs.org) <[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: 75F4 6586 AE61 A66C C44E 87DC 6C38 CACA 20D9 B392
* remotes/dgibson/tags/ppc-for-2.6-20160218: (26 commits)
hw/ppc/spapr: Halt CPU when powering off via RTAS call
pseries: Include missing pseries-2.5 compat properties in pseries-2.4
cuda: remove CUDA_GET_SET_IIC/CUDA_COMBINED_FORMAT_IIC commands
cuda: remove GET_6805_ADDR command
cuda: port SET_TIME command to new framework
cuda: port GET_TIME command to new framework
cuda: port SET_POWER_MESSAGES command to new framework
cuda: port FILE_SERVER_FLAG command to new framework
cuda: port RESET_SYSTEM command to new framework
cuda: port POWERDOWN command to new framework
cuda: port SET_DEVICE_LIST command to new framework
cuda: port SET_AUTO_RATE command to new framework
cuda: port AUTOPOLL command to new framework
cuda: move unknown commands reject out of switch
cuda: add a framework to handle commands
hw/ppc/spapr: Implement the h_set_xdabr hypercall
hw/ppc/spapr: Implement h_set_dabr
hw/ppc/spapr: Add h_set_sprg0 hypercall
migration: ensure htab_save_first completes after timeout
target-ppc: Remove hack for ppc_hash64_load_hpte*() with HV KVM
...
Thomas Huth [Wed, 17 Feb 2016 18:23:19 +0000 (19:23 +0100)]
hw/ppc/spapr: Halt CPU when powering off via RTAS call
The LoPAPR specification defines the following for the RTAS
power-off call: "On successful operation, does not return".
However, the implementation in QEMU currently returns and runs
the guest CPU again for some more cycles. This caused some
trouble with the new ppc implementation of the kvm-unit-tests
recently. So let's make sure that the QEMU implementation
follows the spec, thus stop the CPU to make sure that the
RTAS call does not return to the guest anymore.
It's often handy to make executables depend on each other, e.g. make a
test depend on a helper. This doesn't work now, as linker
will attempt to use the helper as an object.
To fix, filter only relevant file types before linking an executable.
David Gibson [Tue, 16 Feb 2016 23:25:37 +0000 (10:25 +1100)]
pseries: Include missing pseries-2.5 compat properties in pseries-2.4
Commit 4b23699 "pseries: Add pseries-2.6 machine type" added a new
SPAPR_COMPAT_2_5 macro in the usual way. However, it didn't add this
macro to the existing SPAPR_COMPAT_2_4 macro so that pseries-2.4
inherits newer compatibility properties which are needed for 2.5 and
earlier.
projects / qemu.git / log