Andreas Färber [Wed, 24 Jul 2013 23:09:03 +0000 (01:09 +0200)]
pl110: QOM'ify pl110, pl110_versatile and pl111
Let pl110_versatile and pl111 inherit from pl110 and use PL110() cast;
set their version index in an instance_init.
Reviewed-by: Hu Tao <[email protected]>
[AF: Set version PL110 explicitly, suggested by Hu Tao; fixed inheritence] Signed-off-by: Andreas Färber <[email protected]>
Anthony Liguori [Mon, 29 Jul 2013 16:33:48 +0000 (11:33 -0500)]
Merge remote-tracking branch 'stefanha/block' into staging
# By Stefan Hajnoczi (4) and others
# Via Stefan Hajnoczi
* stefanha/block:
dataplane: refuse to start if device is already in use
dataplane: enable virtio-blk x-data-plane=on live migration
migration: fix spice migration
migration: notify migration state before starting thread
block: Repair the throttling code.
gluster: Add image resize support
Anthony Liguori [Mon, 29 Jul 2013 16:33:34 +0000 (11:33 -0500)]
Merge remote-tracking branch 'afaerber/tags/qom-cpu-for-anthony' into staging
QOM CPUState refactorings
* Fix x86 cpu-add
* Change KVM PMU behavior for 1.6
# gpg: Signature made Mon 29 Jul 2013 10:28:18 AM CDT using RSA key ID 3E7E013F
# gpg: Can't check signature: public key not found
# By Eduardo Habkost (2) and Andreas Färber (1)
# Via Andreas Färber
* afaerber/tags/qom-cpu-for-anthony:
target-i386: Disable PMU CPUID leaf by default
target-i386: Pass X86CPU object to cpu_x86_find_by_name()
cpu: Partially revert "cpu: Change qemu_init_vcpu() argument to CPUState"
qemu-char: Register ring buffer driver with correct name "ringbuf"
The driver is new in 1.4, with the documented name "ringbuf".
However, it's actual name is the completely undocumented "memory".
Screwed up in commit 3949e59. Fix code to match documentation.
Keep the undocumented name working as an alias for compatibility.
Commit 51767e7 "qemu-char: Add new char backend CirMemCharDriver"
introduced a memory ring buffer character device driver named
"memory". Commit 3949e59 "qemu-char: Saner naming of memchar stuff &
doc fixes" changed the driver name to "ringbuf", along with a whole
bunch of other names, with the following rationale:
Naming is a mess. The code calls the device driver
CirMemCharDriver, the public API calls it "memory", "memchardev",
or "memchar", and the special commands are named like
"memchar-FOO". "memory" is a particularly unfortunate choice,
because there's another character device driver called
MemoryDriver. Moreover, the device's distinctive property is that
it's a ring buffer, not that's in memory.
This is what we released in 1.4.0.
Unfortunately, the rename missed a critical instance of "memory": the
actual driver name. Thus, the new device could be used only by an
entirely undocumented name. The documented name did not work.
Bummer.
Commit 6a85e60 fixes this by changing the documentation to match the
code. It also changes some, but not all related occurences of
"ringbuf" to "memory". Left alone are identifiers in C code, HMP and
QMP commands. The latter are external interface, so they can't be
changed.
The result is an inconsistent mess. Moreover, "memory" is a rotten
name. The device's distinctive property is that it's a ring buffer,
not that's in memory. User's don't care whether it's in RAM, flash,
or carved into chocolate tablets by Oompa Loompas.
Revert the commit. Next commit will fix just the bug.
Generator function get_expr() yields one section after the
other, as a string. An unindented, non-empty line that
isn't a comment starts a new section.
Lexing:
Split section into a list of tokens (strings), with help
of generator function tokenize().
Parsing:
Parse the first expression from the list of tokens, with
parse(), throw away any remaining tokens.
In parse_schema(): record value of an enum, union or
struct key (if any) in the appropriate global table,
append expression to the list of expressions.
Return list of expressions.
Known issues:
(1) Indentation is significant, unlike in real JSON.
(2) Neither lexer nor parser have any idea of source positions. Error
reporting is hard, let's go shopping.
(3) The one error we bother to detect, we "report" via raise.
(4) The lexer silently ignores invalid characters.
(5) If everything in a section gets ignored, the parser crashes.
(6) The lexer treats a string containing a structural character exactly
like the structural character.
(7) Tokens trailing the first expression in a section are silently
ignored.
(8) The parser accepts any token in place of a colon.
(9) The parser treats comma as optional.
(10) parse() crashes on unexpected EOF.
(11) parse_schema() crashes when a section's expression isn't a JSON
object.
Replace this piece of original art by a thoroughly unoriginal design.
Takes care of (1), (2), (5), (6) and (7), and lays the groundwork for
addressing the others. Generated source files remain unchanged.
Peter Maydell [Mon, 29 Jul 2013 11:22:11 +0000 (12:22 +0100)]
tpm.c: Don't try to put -1 in a variable of type TpmModel
The TpmModel type is an enum (valid values 0 and 1), which means
the compiler can legitimately decide that comparisons like
'tpm_models[i] == -1' are never true. (For example it could
pick 'unsigned char' as its type for representing the enum.)
Avoid this issue by using TPM_MODEL_MAX to mark entries in
the tpm_models[] array which aren't filled in, instead of -1.
This silences a clang warning:
tpm.c:43:27: error: comparison of constant -1 with expression of type
'enum TpmModel' is always false [-Werror,-Wtautological-constant-out-of-range-compare]
if (tpm_models[i] == -1) {
~~~~~~~~~~~~~ ^ ~~
Categorize devices that appear as output to "-device ?" command
by logical functionality. Sort the devices by logical categories
before showing them to user.
The sort is done by functionality rather than alphabetical.
The driver calculates SOLNT bit from UCSOLNT and SCSOLNT bits from
the request. The iu pointer has a type of srp_iu* which points to a union,
so cmd and rsp overlap. As the vscsi_send_rsp function calls
memset(iu, 0, sizeof(rsp)), it clears first 36 bytes of both cmd and rsp
so cmd.sol_not is always zero at the moment of calculating rsp.sol_not.
Anthony Liguori [Thu, 18 Jul 2013 19:33:04 +0000 (14:33 -0500)]
xics: rename types to be sane and follow coding style
Basically, in HW the layout of the interrupt network is:
- One ICP per processor thread (the "presenter"). This contains the
registers to fetch a pending interrupt (ack), EOI, and control the
processor priority.
- One ICS per logical source of interrupts (ie, one per PCI host
bridge, and a few others here or there). This contains the per-interrupt
source configuration (target processor(s), priority, mask) and the
per-interrupt internal state.
Under PAPR, there is a single "virtual" ICS ... somewhat (it's a bit
oddball what pHyp does here, arguably there are two but we can ignore
that distinction). There is no register level access. A pair of firmware
(RTAS) calls is used to configure each virtual interrupt.
So our model here is somewhat the same. We have one ICS in the emulated
XICS which arguably *is* the emulated XICS, there's no point making it a
separate "device", that would just be gross, and each VCPU has an
associated ICP.
Yet we call the "XICS" struct icp_state and then the ICPs
'struct icp_server_state'. It's particularly confusing when all of the
functions have xics_prefixes yet take *icp arguments.
At present, the savevm / migration support for the pseries machine will not
work when KVM is enabled. That's because KVM manages the guest's hash page
table in the host kernel, so qemu has no visibility of it. This patch
fixes this by using new kernel interfaces to extract and reinsert the
guest's hash table during the migration process.
David Gibson [Thu, 18 Jul 2013 19:33:01 +0000 (14:33 -0500)]
pseries: savevm support for pseries machine
This adds the necessary pieces to implement savevm / migration for the
pseries machine. The most complex part here is migrating the hash
table - for the paravirtualized pseries machine the guest's hash page
table is not stored within guest memory, but externally and the guest
accesses it via hypercalls.
This patch uses a hypervisor reserved bit of the HPTE as a dirty bit
(tracking changes to the HPTE itself, not the page it references).
This is used to implement a live migration style incremental save and
restore of the hash table contents.
Normally a hash table is 16MB but it can get bigger depending on how
much RAM the guest has. Due to its nature, updates to it are random so
the live migration style is used for it.
In addition it adds VMStateDescription information to save and restore
the (few) remaining pieces of state information needed by the pseries
machine.
The patch reimplements handling of indirect requests in order to
simplify upcoming live migration support.
- all pointers (except SCSIRequest*) were replaces with integer
indexes and offsets;
- DMA'ed srp_direct_buf kept untouched (ie. BE format);
- vscsi_fetch_desc() is added, now it is the only place where
descriptors are fetched and byteswapped;
- vscsi_req struct fields converted to migration-friendly types;
- many dprintf()'s fixed.
This also removed an unused field 'lun' from the spapr_vscsi device
which is assigned, but never used. So, remove it.
target-ppc: Convert ppc cpu savevm to VMStateDescription
The savevm code for the powerpc cpu emulation is currently based around
the old register_savevm() rather than register_vmstate() method. It's also
rather broken, missing some important state on some CPU models.
This patch completely rewrites the savevm for target-ppc, using the new
VMStateDescription approach. Exactly what needs to be saved in what
configurations has been more carefully examined, too. This introduces a
new version (5) of the cpu save format. The old load function is retained
to support version 4 images.
Stefan Hajnoczi [Mon, 29 Jul 2013 13:02:00 +0000 (15:02 +0200)]
dataplane: refuse to start if device is already in use
Dataplane must check whether a block device is in use before launching
the dataplane thread. This is necessary since the thread does not
synchronize with the main loop and I/O requests could cause corruption.
One example is when a drive is added and a block job is started before
hotplugging the virtio-blk-pci adapter. In this case we must not use
dataplane mode.
Stefan Hajnoczi [Mon, 29 Jul 2013 13:01:59 +0000 (15:01 +0200)]
dataplane: enable virtio-blk x-data-plane=on live migration
Although the dataplane thread does not cooperate with dirty memory
logging yet it's fairly easy to temporarily disable dataplane during
live migration. This way virtio-blk can live migrate when
x-data-plane=on.
The dataplane thread will restart after migration is cancelled or if the
guest resuming virtio-blk operation after migration completes.
Stefan Hajnoczi [Mon, 29 Jul 2013 13:01:58 +0000 (15:01 +0200)]
migration: fix spice migration
Commit 29ae8a4133082e16970c9d4be09f4b6a15034617 ("rdma: introduce
MIG_STATE_NONE and change MIG_STATE_SETUP state transition") changed the
state transitions during migration setup.
Spice used to be notified with MIG_STATE_ACTIVE and it detected this
using migration_is_active(). Spice is now notified with
MIG_STATE_SETUP and migration_is_active() no longer works.
Replace migration_is_active() with migration_in_setup() to fix spice
migration.
Stefan Hajnoczi [Mon, 29 Jul 2013 13:01:57 +0000 (15:01 +0200)]
migration: notify migration state before starting thread
The migration thread runs outside the QEMU global mutex when possible.
Therefore we must notify migration state change *before* starting the
migration thread.
This allows registered listeners to act before live migration iterations
begin. Therefore they can get into a state that allows for live
migration. When the migration thread starts everything will be ready.
Without this patch there is a race condition during migration setup,
depending on whether the migration thread has already transitioned from
SETUP to ACTIVE state.
The throttling code was segfaulting since commit 02ffb504485f0920cfc75a0982a602f824a9a4f4 because some qemu_co_queue_next caller
does not run in a coroutine.
qemu_co_queue_do_restart assume that the caller is a coroutinne.
As suggested by Stefan fix this by entering the coroutine directly.
Also make sure like suggested that qemu_co_queue_next() and
qemu_co_queue_restart_all() can be called only in coroutines.
Bug description: QEMU currently gets all bits from GET_SUPPORTED_CPUID
for CPUID leaf 0xA and passes them directly to the guest. This makes
the guest ABI depend on host kernel and host CPU capabilities, and
breaks live migration if we migrate between hosts with different
capabilities (e.g., different number of PMU counters).
Add a "pmu" property to X86CPU, and set it to true only on "-cpu host",
or on pc-*-1.5 and older machine-types.
For now, setting pmu=on will enable the current passthrough mode that
doesn't have any ABI stability guarantees, but in the future we may
implement a mode where the PMU CPUID bits are stable and configurable.
target-i386: Pass X86CPU object to cpu_x86_find_by_name()
This will help us change the initialization code to not require carrying
some intermediate values in a x86_def_t struct (and eventually kill the
x86_def_t struct entirely).
Andreas Färber [Sat, 27 Jul 2013 00:53:25 +0000 (02:53 +0200)]
cpu: Partially revert "cpu: Change qemu_init_vcpu() argument to CPUState"
Commit c643bed99 moved qemu_init_vcpu() calls to common CPUState code.
This causes x86 cpu-add to fail with "KVM: setting VAPIC address failed".
The reason for the failure is that CPUClass::kvm_fd is not yet
initialized in the following call graph:
->x86_cpu_realizefn
->x86_cpu_apic_realize
->qdev_init
->device_set_realized
->device_reset (hotplugged == 1)
->apic_reset_common
->vapic_base_update
->kvm_apic_vapic_base_update
This causes attempted KVM vCPU ioctls to fail.
By contrast, in the non-hotplug case the APIC is reset much later, when
the vCPU is already initialized.
As a quick and safe solution, move the qemu_init_vcpu() call back into
the targets' realize functions.
Anthony Liguori [Mon, 29 Jul 2013 12:30:31 +0000 (07:30 -0500)]
Merge remote-tracking branch 'sstabellini/xen-130729' into staging
# By Paul Durrant (1) and Stefano Stabellini (1)
# Via Stefano Stabellini
* sstabellini/xen-130729:
Xen PV Device
xen_disk: support "direct-io-safe" backend option
Anthony Liguori [Mon, 29 Jul 2013 12:30:21 +0000 (07:30 -0500)]
Merge remote-tracking branch 'agraf/s390-for-upstream' into staging
# By Alexander Graf (1) and others
# Via Alexander Graf
* agraf/s390-for-upstream:
s390: update s390-ccw.img
s390/ipl: Fix boot order
s390/IPL: Allow boot from other ssid than 0
Anthony Liguori [Mon, 29 Jul 2013 12:30:06 +0000 (07:30 -0500)]
Merge remote-tracking branch 'sweil/w32' into staging
# By Stefan Weil (10) and others
# Via Aurelien Jarno (1) and Stefan Weil (1)
* sweil/w32: (27 commits)
w32, w64: Add build rule for installer
target-mips: fix mipsdsp_mul_q31_q31
mips_malta: fix copy of the 0x1fc00000 region
linux-user: correct argument number for sys_mremap and sys_splice
target-mips: Remove assignment to a variable which is never used
target-mips: fix mipsdsp_trunc16_sat16_round
hw/mips: align initrd to 64KB to avoid kernel error
pflash_cfi01: duplicate status byte from bits 23:16 for 32bit reads
mips_malta: generate SMBUS EEPROM data
mips_malta: cap BIOS endian swap length at 0x3e0000 bytes
mips_malta: generate SPD EEPROM data at runtime
mips_malta: correct reading MIPS revision at 0x1fc00010
mips_malta: fix BIOS endianness swapping
mips_malta: QOM cast cleanup
target-mips: fix branch in likely delay slot tcg assert
target-mips: fix multiplication in mipsdsp_rndq15_mul_q15_q15
target-mips: Remove assignment to a variable which is never used
misc: Use g_assert_not_reached for code which is expected to be unreachable
qemu-options: mention C-a h in the -nographic doc
misc: Fix new typos in comments and strings
...
Paul Durrant [Mon, 29 Jul 2013 10:58:01 +0000 (10:58 +0000)]
Xen PV Device
Introduces a new Xen PV PCI device which will act as a binding point for
PV drivers for Xen.
The device has parameterized vendor-id, device-id and revision to allow to
be configured as a binding point for any vendor's PV drivers.
projects / qemu.git / log