hw/timer: Compile devices not target-dependent as common object
All these devices do not contain any target-specific. While most
of them are arch-specific, they are shared between different
targets of the same arch family (ARM and AArch64, MIPS32/MIPS64,
endianess, ...).
Put them into common-obj-y to compile them once for all targets.
Frediano Ziglio [Wed, 23 Oct 2019 12:26:51 +0000 (13:26 +0100)]
event_notifier: avoid dandling file descriptor in event_notifier_cleanup
If rfd is equal to wfd the file descriptor is closed but
rfd will still have the closed value.
The EventNotifier structure should not be used again after calling
event_notifier_cleanup or should be initialized again but make
sure to not have dandling file descriptors around.
Peter Maydell [Thu, 24 Oct 2019 15:22:58 +0000 (16:22 +0100)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-4.2-20191024' into staging
ppc patch queue 2019-10-24
Last pull request before soft freeze.
* Lots of fixes and cleanups for spapr interrupt controllers
* More SLOF updates to fix problems with full FDT rendering at CAS
time (alas, more yet are to come)
* A few other assorted changes
This isn't quite as well tested as I usually try to do before a pull
request. But I've been sick and running into some other difficulties,
and wanted to get this sent out before heading towards KVM forum.
* remotes/dgibson/tags/ppc-for-4.2-20191024: (28 commits)
spapr/xive: Set the OS CAM line at reset
ppc/pnv: Fix naming of routines realizing the CPUs
ppc: Reset the interrupt presenter from the CPU reset handler
ppc/pnv: Add a PnvChip pointer to PnvCore
ppc/pnv: Introduce a PnvCore reset handler
spapr_cpu_core: Implement DeviceClass::reset
spapr: move CPU reset after presenter creation
spapr: Don't request to unplug the same core twice
pseries: Update SLOF firmware image
spapr: Move SpaprIrq::nr_xirqs to SpaprMachineClass
spapr: Remove SpaprIrq::nr_msis
spapr, xics, xive: Move SpaprIrq::post_load hook to backends
spapr, xics, xive: Move SpaprIrq::reset hook logic into activate/deactivate
spapr: Remove SpaprIrq::init_kvm hook
spapr, xics, xive: Match signatures for XICS and XIVE KVM connect routines
spapr, xics, xive: Move dt_populate from SpaprIrq to SpaprInterruptController
spapr, xics, xive: Move print_info from SpaprIrq to SpaprInterruptController
spapr, xics, xive: Move set_irq from SpaprIrq to SpaprInterruptController
spapr: Formalize notion of active interrupt controller
spapr, xics, xive: Move irq claim and free from SpaprIrq to SpaprInterruptController
...
* remotes/ehabkost/tags/machine-next-pull-request:
hppa: drop usage of memory_region_allocate_system_memory() for ROM
ppc: rs6000_mc: drop usage of memory_region_allocate_system_memory()
sparc64: use memory_region_allocate_system_memory() only for '-m' specified RAM
target/i386: Introduce Denverton CPU model
Igor Mammedov [Tue, 8 Oct 2019 11:33:18 +0000 (07:33 -0400)]
hppa: drop usage of memory_region_allocate_system_memory() for ROM
machine_hppa_init() violates memory_region_allocate_system_memory() contract
by calling it multiple times which could break with -mem-path. Replace
the second usage (for 'rom') with memory_region_init_ram() instead.
Igor Mammedov [Tue, 8 Oct 2019 11:33:17 +0000 (07:33 -0400)]
ppc: rs6000_mc: drop usage of memory_region_allocate_system_memory()
rs6000mc_realize() violates memory_region_allocate_system_memory() contract
by calling it multiple times which could break -mem-path. Replace it with
plain memory_region_init_ram() instead.
Igor Mammedov [Tue, 8 Oct 2019 11:33:16 +0000 (07:33 -0400)]
sparc64: use memory_region_allocate_system_memory() only for '-m' specified RAM
memory_region_allocate_system_memory() was designed to be called for
allocating inital RAM. Using it mutiple times within one board is not
supported and could fail if -mem-path with non hugepage path is used.
Keep using memory_region_allocate_system_memory() only for initial
RAM and use memory_region_init_ram() for the rest fixed size regions.
Cédric Le Goater [Tue, 22 Oct 2019 16:38:12 +0000 (18:38 +0200)]
spapr/xive: Set the OS CAM line at reset
When a Virtual Processor is scheduled to run on a HW thread, the
hypervisor pushes its identifier in the OS CAM line. When running with
kernel_irqchip=off, QEMU needs to emulate the same behavior.
Set the OS CAM line when the interrupt presenter of the sPAPR core is
reset. This will also cover the case of hot-plugged CPUs.
This change also has the benefit to remove the use of CPU_FOREACH()
which can be unsafe.
Cédric Le Goater [Tue, 22 Oct 2019 16:38:10 +0000 (18:38 +0200)]
ppc: Reset the interrupt presenter from the CPU reset handler
On the sPAPR machine and PowerNV machine, the interrupt presenters are
created by a machine handler at the core level and are reset
independently. This is not consistent and it raises issues when it
comes to handle hot-plugged CPUs. In that case, the presenters are not
reset. This is less of an issue in XICS, although a zero MFFR could
be a concern, but in XIVE, the OS CAM line is not set and this breaks
the presenting algorithm. The current code has workarounds which need
a global cleanup.
Extend the sPAPR IRQ backend and the PowerNV Chip class with a new
cpu_intc_reset() handler called by the CPU reset handler and remove
the XiveTCTX reset handler which is now redundant.
Greg Kurz [Tue, 22 Oct 2019 16:38:07 +0000 (18:38 +0200)]
spapr_cpu_core: Implement DeviceClass::reset
Since vCPUs aren't plugged into a bus, we manually register a reset
handler for each vCPU. We also call this handler at realize time
to ensure hot plugged vCPUs are reset before being exposed to the
guest. This results in vCPUs being reset twice at machine reset.
It doesn't break anything but it is slightly suboptimal and above
all confusing.
The hotplug path in device_set_realized() already knows how to reset
a hotplugged device if the device reset handler is present. Implement
one for sPAPR CPU cores that resets all vCPUs under a core.
While here rename spapr_cpu_reset() to spapr_reset_vcpu() for
consistency with spapr_realize_vcpu() and spapr_unrealize_vcpu().
Greg Kurz [Wed, 23 Oct 2019 19:17:40 +0000 (21:17 +0200)]
spapr: Don't request to unplug the same core twice
We must not call spapr_drc_detach() on a detached DRC otherwise bad things
can happen, ie. QEMU hangs or crashes. This is easily demonstrated with
a CPU hotplug/unplug loop using QMP.
David Gibson [Fri, 27 Sep 2019 03:54:23 +0000 (13:54 +1000)]
spapr: Move SpaprIrq::nr_xirqs to SpaprMachineClass
For the benefit of peripheral device allocation, the number of available
irqs really wants to be the same on a given machine type version,
regardless of what irq backends we are using. That's the case now, but
only because we make sure the different SpaprIrq instances have the same
value except for the special legacy one.
Since this really only depends on machine type version, move the value to
SpaprMachineClass instead of SpaprIrq. This also puts the code to set it
to the lower value on old machine types right next to setting
legacy_irq_allocation, which needs to go hand in hand.
David Gibson [Fri, 27 Sep 2019 03:44:58 +0000 (13:44 +1000)]
spapr: Remove SpaprIrq::nr_msis
The nr_msis value we use here has to line up with whether we're using
legacy or modern irq allocation. Therefore it's safer to derive it based
on legacy_irq_allocation rather than having SpaprIrq contain a canned
value.
David Gibson [Fri, 27 Sep 2019 00:53:53 +0000 (10:53 +1000)]
spapr, xics, xive: Move SpaprIrq::post_load hook to backends
The remaining logic in the post_load hook really belongs to the interrupt
controller backends, and just needs to be called on the active controller
(after the active controller is set to the right thing based on the
incoming migration in the generic spapr_irq_post_load() logic).
David Gibson [Thu, 26 Sep 2019 13:58:36 +0000 (23:58 +1000)]
spapr, xics, xive: Move SpaprIrq::reset hook logic into activate/deactivate
It turns out that all the logic in the SpaprIrq::reset hooks (and some in
the SpaprIrq::post_load hooks) isn't really related to resetting the irq
backend (that's handled by the backends' own reset routines). Rather its
about getting the backend ready to be the active interrupt controller or
stopping being the active interrupt controller - reset (and post_load) is
just the only time that changes at present.
To make this flow clearer, move the logic into the explicit backend
activate and deactivate hooks.
David Gibson [Thu, 26 Sep 2019 13:02:41 +0000 (23:02 +1000)]
spapr: Remove SpaprIrq::init_kvm hook
This hook is a bit odd. The only caller is spapr_irq_init_kvm(), but
it explicitly takes an SpaprIrq *, so it's never really called through the
current SpaprIrq. Essentially this is just a way of passing through a
function pointer so that spapr_irq_init_kvm() can handle some
configuration and error handling logic without duplicating it between the
xics and xive reset paths.
So, make it just take that function pointer. Because of earlier reworks
to the KVM connect/disconnect code in the xics and xive backends we can
also eliminate some wrapper functions and streamline error handling a bit.
David Gibson [Thu, 26 Sep 2019 13:23:51 +0000 (23:23 +1000)]
spapr, xics, xive: Match signatures for XICS and XIVE KVM connect routines
Both XICS and XIVE have routines to connect and disconnect KVM with
similar but not identical signatures. This adjusts them to match
exactly, which will be useful for further cleanups later.
While we're there, we add an explicit return value to the connect path
to streamline error reporting in the callers. We remove error
reporting the disconnect path. In the XICS case this wasn't used at
all. In the XIVE case the only error case was if the KVM device was
set up, but KVM didn't have the capability to do so which is pretty
obviously impossible.
David Gibson [Mon, 30 Sep 2019 02:35:06 +0000 (12:35 +1000)]
spapr, xics, xive: Move dt_populate from SpaprIrq to SpaprInterruptController
This method depends only on the active irq controller. Now that we've
formalized the notion of active controller we can dispatch directly
through that, rather than dispatching via SpaprIrq with the dual
version having to do a second conditional dispatch.
David Gibson [Thu, 26 Sep 2019 06:12:05 +0000 (16:12 +1000)]
spapr, xics, xive: Move print_info from SpaprIrq to SpaprInterruptController
This method depends only on the active irq controller. Now that we've
formalized the notion of active controller we can dispatch directly
through that, rather than dispatching via SpaprIrq with the dual
version having to do a second conditional dispatch.
David Gibson [Thu, 26 Sep 2019 06:09:46 +0000 (16:09 +1000)]
spapr, xics, xive: Move set_irq from SpaprIrq to SpaprInterruptController
This method depends only on the active irq controller. Now that we've
formalized the notion of active controller we can dispatch directly through
that, rather than dispatching via SpaprIrq with the dual version having
to do a second conditional dispatch.
David Gibson [Thu, 26 Sep 2019 05:41:39 +0000 (15:41 +1000)]
spapr: Formalize notion of active interrupt controller
spapr now has the mechanism of constructing both XICS and XIVE instances of
the SpaprInterruptController interface. However, only one of the interrupt
controllers will actually be active at any given time, depending on feature
negotiation with the guest. This is handled in the current code via
spapr_irq_current() which checks the OV5 vector from feature negotiation to
determine the current backend.
Determining the active controller at the point we need it like this
can be pretty confusing, because it makes it very non obvious at what
points the active controller can change. This can make it difficult
to reason about the code and where a change of active controller could
appear in sequence with other events.
Make this mechanism more explicit by adding an 'active_intc' pointer
and an explicit spapr_irq_update_active_intc() function to update it
from the CAS state. We also add hooks on the intc backend which will
get called when it is activated or deactivated.
For now we just introduce the switch and hooks, later patches will
actually start using them.
David Gibson [Thu, 26 Sep 2019 04:11:23 +0000 (14:11 +1000)]
spapr, xics, xive: Move cpu_intc_create from SpaprIrq to SpaprInterruptController
This method essentially represents code which belongs to the interrupt
controller, but needs to be called on all possible intcs, rather than
just the currently active one. The "dual" version therefore calls
into the xics and xive versions confusingly.
Handle this more directly, by making it instead a method on the intc
backend, and always calling it on every backend that exists.
While we're there, streamline the error reporting a bit.
The SpaprIrq structure is used to represent ths spapr machine's irq
backend. Except that it kind of conflates two concepts: one is the
backend proper - a specific interrupt controller that we might or
might not be using, the other is the irq configuration which covers
the layout of irq space and which interrupt controllers are allowed.
This leads to some pretty confusing code paths for the "dual"
configuration where its hooks redirect to other SpaprIrq structures
depending on the currently active irq controller.
To clean this up, we start by introducing a new
SpaprInterruptController QOM interface to represent strictly an
interrupt controller backend, not counting anything configuration
related. We implement this interface in the XICs and XIVE interrupt
controllers, and in future we'll move relevant methods from SpaprIrq
into it.
Greg Kurz [Thu, 3 Oct 2019 12:02:00 +0000 (14:02 +0200)]
spapr: Set VSMT to smp_threads by default
Support for setting VSMT is available in KVM since linux-4.13. Most distros
that support KVM on POWER already have it. It thus seem reasonable enough
to have the default machine to set VSMT to smp_threads.
This brings contiguous VCPU ids and thus brings their upper bound down to
the machine's max_cpus. This is especially useful for XIVE KVM devices,
which may thus allocate only one VP descriptor per VCPU.
ppc/pnv: Use address_space_stq_be() when triggering an interrupt from PSI
Include the XIVE_TRIGGER_PQ bit in the trigger data which is how
hardware signals to the IC that the PQ bits of the interrupt source
have been checked.
The trigger data is used for both triggers of a HW source interrupts,
PHB, PSI, and triggers for rerouting interrupts between interrupt
controllers.
When an interrupt is rerouted, the trigger data follows an "END
trigger" format. In that case, the remote IC needs EAS containing an
END index to perform a lookup of an END.
An END trigger, bit0 of word0 set to '1', is defined as :
|0123|4567|0123|4567|0123|4567|0123|4567|
W0 E=1 |1P--|BLOC| END IDX |
W1 E=1 |M | END DATA |
An EAS is defined as :
|0123|4567|0123|4567|0123|4567|0123|4567|
W0 |V---|BLOC| END IDX |
W1 |M | END DATA |
The END trigger adds an extra 'PQ' bit, bit1 of word0 set to '1',
signaling that the PQ bits have been checked. That bit is unused in
the initial EAS definition.
When a HW device performs the trigger, the trigger data follows an
"EAS trigger" format because the trigger data in that case contains an
EAS index which the IC needs to look for.
An EAS trigger, bit0 of word0 set to '0', is defined as :
|0123|4567|0123|4567|0123|4567|0123|4567|
W0 E=0 |0P--|---- ---- ---- ---- ---- ---- ----|
W1 E=0 |BLOC| EAS INDEX |
There is also a 'PQ' bit, bit1 of word0 to '1', signaling that the
PQ bits have been checked.
Introduce these new trigger bits and rename the XIVE_SRCNO macros in
XIVE_EAS to reflect better the nature of the data.
target/ppc: Fix for optimized vsl/vsr instructions
In previous implementation, invocation of TCG shift function could request
shift of TCG variable by 64 bits when variable 'sh' is 0, which is not
supported in TCG (values can be shifted by 0 to 63 bits). This patch fixes
this by using two separate invocation of TCG shift functions, with maximum
shift amount of 32.
Name of variable 'shifted' is changed to 'carry' so variable naming
is similar to old helper implementation.
Variables 'avrA' and 'avrB' are replaced with variable 'avr'.
Greg Kurz [Fri, 4 Oct 2019 08:37:47 +0000 (10:37 +0200)]
xics: Make some device types not user creatable
Some device types of the XICS model are exposed to the QEMU command
line:
$ ppc64-softmmu/qemu-system-ppc64 -device help | grep ic[sp]
name "icp"
name "ics"
name "ics-spapr"
name "pnv-icp", desc "PowerNV ICP"
These are internal devices that shouldn't be instantiable by the
user. By the way, they can't be because their respective realize
functions expect link properties that can't be set from the command
line:
qemu-system-ppc64: -device icp: required link 'xics' not found:
Property '.xics' not found
qemu-system-ppc64: -device ics: required link 'xics' not found:
Property '.xics' not found
qemu-system-ppc64: -device ics-spapr: required link 'xics' not found:
Property '.xics' not found
qemu-system-ppc64: -device pnv-icp: required link 'xics' not found:
Property '.xics' not found
Hide them by setting dc->user_creatable to false in the base class
"icp" and "ics" init functions.
Greg Kurz [Fri, 4 Oct 2019 07:38:50 +0000 (09:38 +0200)]
xive: Make some device types not user creatable
Some device types of the XIVE model are exposed to the QEMU command
line:
$ ppc64-softmmu/qemu-system-ppc64 -device help | grep xive
name "xive-end-source", desc "XIVE END Source"
name "xive-source", desc "XIVE Interrupt Source"
name "xive-tctx", desc "XIVE Interrupt Thread Context"
These are internal devices that shouldn't be instantiable by the
user. By the way, they can't be because their respective realize
functions expect link properties that can't be set from the command
line:
qemu-system-ppc64: -device xive-source: required link 'xive' not found:
Property '.xive' not found
qemu-system-ppc64: -device xive-end-source: required link 'xive' not found:
Property '.xive' not found
qemu-system-ppc64: -device xive-tctx: required link 'cpu' not found:
Property '.cpu' not found
Hide them by setting dc->user_creatable to false in their respective
class init functions.
* remotes/armbru/tags/pull-qapi-2019-10-22-v3:
qapi: Allow introspecting fix for savevm's cooperation with blockdev
tests/qapi-schema: Cover feature documentation comments
tests: qapi: Test 'features' of commands
qapi: Add feature flags to commands
tests/qapi-schema: Tidy up test output indentation
qapi: Clear scripts/qapi/doc.py executable bits again
qapi: Split up scripts/qapi/common.py
qapi: Move gen_enum(), gen_enum_lookup() back to qapi/types.py
qapi: Speed up frontend tests
qapi: Eliminate accidental global frontend state
qapi: Store pragma state in QAPISourceInfo, not global state
qapi: Don't suppress doc generation without pragma doc-required
Peter Maydell [Tue, 22 Oct 2019 16:50:39 +0000 (17:50 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20191022-2' into staging
* Fix sign-extension for SMLAL* instructions
* Various ptimer device conversions to new transaction API
* Add a dummy Samsung SDHCI controller model to exynos4 boards
* Minor refactorings of RAM creation for some arm boards
* remotes/pmaydell/tags/pull-target-arm-20191022-2:
hw/arm/digic4: Inline digic4_board_setup_ram() function
hw/arm/omap1: Create the RAM in the board
hw/arm/omap2: Create the RAM in the board
hw/arm/collie: Create the RAM in the board
hw/arm/mps2: Use the IEC binary prefix definitions
hw/arm/xilinx_zynq: Use the IEC binary prefix definitions
hw/arm/exynos4210: Use the Samsung s3c SDHCI controller
hw/sd/sdhci: Add dummy Samsung SDHCI controller
hw/sd/sdhci: Add a comment to distinct the i.MX eSDHC functions
hw/m68k/mcf5208.c: Switch to transaction-based ptimer API
hw/watchdog/etraxfs_timer.c: Switch to transaction-based ptimer API
hw/timer/altera_timer.c: Switch to transaction-based ptimer API
hw/timer/lm32_timer: Switch to transaction-based ptimer API
hw/timer/sh_timer: Switch to transaction-based ptimer API
hw/timer/puv3_ost.c: Switch to transaction-based ptimer API
hw/timer/arm_mptimer.c: Undo accidental rename of arm_mptimer_init()
hw/timer/exynos4210_mct: Initialize ptimer before starting it
target/arm: Fix sign-extension for SMLAL*
hw/arm/digic4: Inline digic4_board_setup_ram() function
Having the RAM creation code in a separate function is not
very helpful. Move this code directly inside the board_init()
function, this will later allow the board to have the QOM
ownership of the RAM.
The SDRAM is incorrectly created in the OMAP310 SoC.
Move its creation in the board code, this will later allow the
board to have the QOM ownership of the RAM.
The SDRAM is incorrectly created in the OMAP2420 SoC.
Move its creation in the board code, this will later allow the
board to have the QOM ownership of the RAM.
The SDRAM is incorrectly created in the SA1110 SoC.
Move its creation in the board code, this will later allow the
board to have the QOM ownership of the RAM.
The Linux kernel access few S3C-specific registers [1] to set some
clock. We don't care about this part for device emulation [2]. Add
a dummy device to properly ignore these accesses, so we can focus
on the important registers missing.
Peter Maydell [Tue, 22 Oct 2019 15:50:36 +0000 (16:50 +0100)]
hw/m68k/mcf5208.c: Switch to transaction-based ptimer API
Switch the mcf5208 code away from bottom-half based ptimers to
the new transaction-based ptimer API. This just requires adding
begin/commit calls around the various places that modify the ptimer
state, and using the new ptimer_init() function to create the timer.
Peter Maydell [Tue, 22 Oct 2019 15:50:36 +0000 (16:50 +0100)]
hw/watchdog/etraxfs_timer.c: Switch to transaction-based ptimer API
Switch the etraxfs_timer code away from bottom-half based ptimers to
the new transaction-based ptimer API. This just requires adding
begin/commit calls around the various places that modify the ptimer
state, and using the new ptimer_init() function to create the timer.
Peter Maydell [Tue, 22 Oct 2019 15:50:36 +0000 (16:50 +0100)]
hw/timer/altera_timer.c: Switch to transaction-based ptimer API
Switch the altera_timer code away from bottom-half based ptimers to
the new transaction-based ptimer API. This just requires adding
begin/commit calls around the various places that modify the ptimer
state, and using the new ptimer_init() function to create the timer.
Peter Maydell [Tue, 22 Oct 2019 15:50:36 +0000 (16:50 +0100)]
hw/timer/lm32_timer: Switch to transaction-based ptimer API
Switch the lm32_timer code away from bottom-half based ptimers to the
new transaction-based ptimer API. This just requires adding
begin/commit calls around the various places that modify the ptimer
state, and using the new ptimer_init() function to create the ytimer.
Peter Maydell [Tue, 22 Oct 2019 15:50:36 +0000 (16:50 +0100)]
hw/timer/sh_timer: Switch to transaction-based ptimer API
Switch the sh_timer code away from bottom-half based ptimers to the
new transaction-based ptimer API. This just requires adding
begin/commit calls around the various places that modify the ptimer
state, and using the new ptimer_init() function to create the timer.
Peter Maydell [Tue, 22 Oct 2019 15:50:35 +0000 (16:50 +0100)]
hw/timer/puv3_ost.c: Switch to transaction-based ptimer API
Switch the puv3_ost code away from bottom-half based ptimers to the
new transaction-based ptimer API. This just requires adding
begin/commit calls around the various places that modify the ptimer
state, and using the new ptimer_init() function to create the timer.
Peter Maydell [Tue, 22 Oct 2019 15:50:35 +0000 (16:50 +0100)]
hw/timer/arm_mptimer.c: Undo accidental rename of arm_mptimer_init()
In commit b01422622b we did an automated rename of the ptimer_init()
function to ptimer_init_with_bh(). Unfortunately this caught the
unrelated arm_mptimer_init() function. Undo that accidental
renaming.
Guenter Roeck [Tue, 22 Oct 2019 15:50:35 +0000 (16:50 +0100)]
hw/timer/exynos4210_mct: Initialize ptimer before starting it
When booting a recent Linux kernel, the qemu message "Timer with delta
zero, disabling" is seen, apparently because a ptimer is started before
being initialized. Fix the problem by initializing the offending ptimer
before starting it.
The bug is effectively harmless in the old QEMUBH setup
because the sequence of events is:
* the delta zero means the timer expires immediately
* ptimer_reload() arranges for exynos4210_gfrc_event() to be called
* ptimer_reload() notices the zero delta and disables the timer
* later, the QEMUBH runs, and exynos4210_gfrc_event() correctly
configures the timer and restarts it
In the new transaction based API the bug is still harmless,
but differences of when the callback function runs mean the
message is not printed any more:
* ptimer_run() does nothing as it's inside a transaction block
* ptimer_transaction_commit() sees it has work to do and
calls ptimer_reload()
* the zero delta means the timer expires immediately
* ptimer_reload() calls exynos4210_gfrc_event() directly
* exynos4210_gfrc_event() configures the timer
* the delta is no longer zero so ptimer_reload() doesn't complain
(the zero-delta test is after the trigger-callback in
the ptimer_reload() function)
Regardless, the behaviour here was not intentional, and we should
just program the ptimer correctly to start with.
Signed-off-by: Guenter Roeck <[email protected]> Reviewed-by: Philippe Mathieu-Daudé <[email protected]> Tested-by: Philippe Mathieu-Daudé <[email protected]> Signed-off-by: Peter Maydell <[email protected]>
Message-id: 20191018143149[email protected]
[PMM: Expansion/clarification of the commit message:
the message is about a zero delta, not a zero period;
added detail to the commit message of the analysis of what
is happening and why the kernel boots even with the message;
added note that the message goes away with the new ptimer API] Signed-off-by: Peter Maydell <[email protected]>
1. add new modes:
connecting-wait: means, that reconnecting is in progress, and there
were small number of reconnect attempts, so all requests are
waiting for the connection.
connecting-nowait: reconnecting is in progress, there were a lot of
attempts of reconnect, all requests will return errors.
two old modes are used too:
connected: normal state
quit: exiting after fatal error or on close
Possible transitions are:
* -> quit
connecting-* -> connected
connecting-wait -> connecting-nowait (transition is done after
reconnect-delay seconds in connecting-wait mode)
connected -> connecting-wait
2. Implement reconnect in connection_co. So, in connecting-* mode,
connection_co, tries to reconnect unlimited times.
3. Retry nbd queries on channel error, if we are in connecting-wait
state.
Peter Krempa [Fri, 18 Oct 2019 08:14:54 +0000 (10:14 +0200)]
qapi: Allow introspecting fix for savevm's cooperation with blockdev
'savevm' was buggy as it considered all monitor-owned block device
nodes for snapshot. With the introduction of -blockdev, the common
usage made all nodes including protocol and backing file nodes be
monitor-owned and thus considered for snapshot.
This is a problem since the 'file' protocol nodes can't have internal
snapshots and it does not make sense to take snapshot of nodes
representing backing files.
This was fixed by commit 05f4aced658a02b02. Clients need to be able to
detect whether this fix is present.
Since savevm does not have an QMP alternative, add the feature for the
'human-monitor-command' backdoor which is used to call this command in
modern use.
Commit 8aa3a33e44 "tests/qapi-schema: Test for good feature lists in
structs" neglected to cover documentation comments, and the previous
commit followed its example. Make up for them.
Peter Krempa [Fri, 18 Oct 2019 08:14:51 +0000 (10:14 +0200)]
qapi: Add feature flags to commands
Similarly to features for struct types introduce the feature flags also
for commands. This will allow notifying management layers of fixes and
compatible changes in the behaviour of a command which may not be
detectable any other way.
The changes were heavily inspired by commit 6a8c0b51025.
The QAPI code generator clocks in at some 3100 SLOC in 8 source files.
Almost 60% of the code is in qapi/common.py. Split it into more
focused modules:
* Move QAPISchemaPragma and QAPISourceInfo to qapi/source.py.
* Move QAPIError and its sub-classes to qapi/error.py.
* Move QAPISchemaParser and QAPIDoc to parser.py. Use the opportunity
to put QAPISchemaParser first.
* Move check_expr() & friends to qapi/expr.py. Use the opportunity to
put the code into a more sensible order.
* Move QAPISchema & friends to qapi/schema.py
* Move QAPIGen and its sub-classes, ifcontext,
QAPISchemaModularCVisitor, and QAPISchemaModularCVisitor to qapi/gen.py
* Delete camel_case(), it's unused since commit e98859a9b9 "qapi:
Clean up after recent conversions to QAPISchemaVisitor"
A number of helper functions remain in qapi/common.py. I considered
moving the code generator helpers to qapi/gen.py, but decided not to.
Perhaps we should rewrite them as methods of QAPIGen some day.
* remotes/vivier2/tags/trivial-branch-pull-request:
tests/migration: fix a typo in comment
qemu-doc: Remove paragraph about requiring a HD image with -kernel
qapi: Move gen_enum(), gen_enum_lookup() back to qapi/types.py
The next commit will split up qapi/common.py. gen_enum() needs
QAPISchemaEnumMember, and that's in the way. Move it to qapi/types.py
along with its buddy gen_enum_lookup().
Permit me a short a digression on history: how did gen_enum() end up
in qapi/common.py? Commit 21cd70dfc1 "qapi script: add event support"
duplicated qapi-types.py's gen_enum() and gen_enum_lookup() in
qapi-event.py. Simply importing them would have been cleaner, but
wasn't possible as qapi-types.py was a program, not a module. Commit efd2eaa6c2 "qapi: De-duplicate enum code generation" de-duplicated by
moving them to qapi.py, which was a module.
Since then, program qapi-types.py has morphed into module types.py.
It's where gen_enum() and gen_enum_lookup() started, and where they
belong.
"make check-qapi-schema" takes around 10s user + system time for me.
With -j, it takes a bit over 3s real time. We have worse tests. It's
still annoying when you work on the QAPI generator.
Some 1.4s user + system time is consumed by make figuring out what to
do, measured by making a target that does nothing. There's nothing I
can do about that right now. But let's see what we can do about the
other 8s.
Almost 7s are spent running test-qapi.py for every test case, the rest
normalizing and diffing test-qapi.py output. We have 190 test cases.
If I downgrade to python2, it's 4.5s, but python2 is a goner.
Hacking up test-qapi.py to exit(0) without doing anything makes it
only marginally faster. The problem is Python startup overhead.
Our configure puts -B into $(PYTHON). Running without -B is faster:
4.4s.
We could improve the Makefile to run test cases only when the test
case or the generator changed. But I'm after improvement in the case
where the generator changed.
test-qapi.py is designed to be the simplest possible building block
for a shell script to do the complete job (it's actually a Makefile,
not a shell script; no real difference). Python is just not meant for
that. It's for bigger blocks.
Move the post-processing and diffing into test-qapi.py, and make it
capable of testing multiple schema files. Set executable bits while
there.
Running it once per test case now takes slightly longer than 8s. But
running it once for all of them takes under 0.2s.
Messing with the Makefile to run it only on the tests that need
retesting is clearly not worth the bother.
Expected error output changes because the new normalization strips off
$(SRCDIR)/tests/qapi-schema/ instead of just $(SRCDIR)/.
The .exit files go away, because there is no exit status to test
anymore.
The frontend can't be run more than once due to its global state.
A future commit will want to do that.
The only global frontend state remaining is accidental:
QAPISchemaParser.__init__()'s parameter previously_included=[].
Python evaluates the default once, at definition time. Any
modifications to it are visible in subsequent calls. Well-known
Python trap. Change the default to None and replace it by the real
default in the function body. Use the opportunity to convert
previously_included to a set.
qapi: Store pragma state in QAPISourceInfo, not global state
The frontend can't be run more than once due to its global state.
A future commit will want to do that.
Recent commit "qapi: Move context-sensitive checking to the proper
place" got rid of many global variables already, but pragma state is
still stored in global variables (that's why a pragma directive's
scope is the complete schema).
qapi: Don't suppress doc generation without pragma doc-required
Commit bc52d03ff5 "qapi: Make doc comments optional where we don't
need them" made scripts/qapi2texi.py fail[*] unless the schema had
pragma 'doc-required': true. The stated reason was inability to cope
with incomplete documentation.
When commit fb0bc835e5 "qapi-gen: New common driver for code and doc
generators" folded scripts/qapi2texi.py into scripts/qapi-gen.py, it
turned the failure into silent suppression.
The doc generator can cope with incomplete documentation now. I don't
know since when, or what the problem was, or even whether it ever
existed.
Drop the silent suppression.
[*] The fail part was broken, fixed in commit e8ba07ea9a.
Thomas Huth [Tue, 1 Oct 2019 11:01:11 +0000 (13:01 +0200)]
qemu-doc: Remove paragraph about requiring a HD image with -kernel
The need for specifying "-hda" together with "-kernel" has been removed in
commit 57a46d057995 ("Convert linux bootrom to external rom and fw_cfg"),
almost 10 years ago, so let's remove this description from our documentation
now, too.
s390x/kvm: Set default cpu model for all machine classes
We have to set the default model of all machine classes, not just for
the active one. Otherwise, "query-machines" will indicate the wrong
CPU model ("qemu-s390x-cpu" instead of "host-s390x-cpu") as
"default-cpu-type".
Doing a
{"execute":"query-machines"}
under KVM now results in
{"return": [
{
"hotpluggable-cpus": true,
"name": "s390-ccw-virtio-4.0",
"numa-mem-supported": false,
"default-cpu-type": "host-s390x-cpu",
"cpu-max": 248,
"deprecated": false},
{
"hotpluggable-cpus": true,
"name": "s390-ccw-virtio-2.7",
"numa-mem-supported": false,
"default-cpu-type": "host-s390x-cpu",
"cpu-max": 248,
"deprecated": false
} ...
Libvirt probes all machines via "-machine none,accel=kvm:tcg" and will
currently see the wrong CPU model under KVM.
s390x/tcg: Fix VECTOR SUBTRACT WITH BORROW COMPUTE BORROW INDICATION
The numbers are unsigned, the computation is wrong. "Each operand is
treated as an unsigned binary integer".
Let's implement as given in the PoP:
"A subtraction is performed by adding the contents of the second operand
with the bitwise complement of the third operand along with a borrow
indication from the rightmost bit of the fourth operand."
s390x/tcg: Fix VECTOR SUBTRACT WITH BORROW INDICATION
Testing this, there seems to be something messed up. We are dealing with
unsigned numbers. "Each operand is treated as an unsigned binary integer."
Let's just implement as written in the PoP:
"A subtraction is performed by adding the contents of
the second operand with the bitwise complement of
the third operand along with a borrow indication from
the rightmost bit position of the fourth operand and
the result is placed in the first operand."
Looks like my idea of what a "borrow" is was wrong. The PoP says:
"If the resulting subtraction results in a carry out of bit zero, a value
of one is placed in the corresponding element of the first operand;
otherwise, a value of zero is placed in the corresponding element"
As clarified by Richard, all we have to do is invert the result.
Andrew Jones [Wed, 16 Oct 2019 14:54:34 +0000 (16:54 +0200)]
s390x/cpumodel: Add missing visit_free
Beata Michalska noticed this missing visit_free() while reviewing
arm's implementation of qmp_query_cpu_model_expansion(), which is
modeled off this s390x implementation.
projects / qemu.git / log