There is at least one guest (OS) such that although it does not rely on
the guarantees provided by ORB 1 word 9 bit (aka unlimited prefetch, aka
P bit) not being set, it fails to tell this to the machine.
Usually this ain't a big deal, as the original purpose of the P bit is to
allow for performance optimizations. vfio-ccw however can not provide the
guarantees required if the bit is not set.
It is not possible to implement support for the P bit not set without
transitioning to lower level protocols for vfio-ccw. So let's give the
user the opportunity to force setting the P bit, if the user knows this
is safe. For self modifying channel programs forcing the P bit is not
safe. If the P bit is forced for a self modifying channel program things
are expected to break in strange ways.
Let's also avoid warning multiple about P bit not set in the ORB in case
P bit is not told to be forced, and designate the affected vfio-ccw
device.
Signed-off-by: Halil Pasic <pasic@linux.ibm.com>
Suggested-by: Dong Jia Shi <bjsdjshi@linux.ibm.com>
Acked-by: Jason J. Herne <jjherne@linux.ibm.com>
Tested-by: Jason J. Herne <jjherne@linux.ibm.com>
Message-Id: <20180524175828.3143-2-pasic@linux.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Coverity recently started complaining about virtio_ccw_notify(). Turns
out, there is a couple of things that can be cleaned up. Let's clean!
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Fixes: CID 1390619
Signed-off-by: Halil Pasic <pasic@linux.ibm.com>
Message-Id: <20180516132757.68558-1-pasic@linux.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
This is wrong. object_finalize_child_property()'s unref balances the
ref in object_property_add_child(). qbus_realize's unref balances the
ref that was initially placed by object_new/object_initialize.
This reverts commit f3d58385a6.
Reported-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20180613172815.32738-4-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The USB device don't hold the bus. There is no ASAN related reports
anymore.
This reverts commit cd7bc87868.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20180613172815.32738-3-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
The bus is not owned by the device.
This reverts commit 410a096adf.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20180613172815.32738-2-marcandre.lureau@redhat.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Commit 3d85885a1b tried to fix error handling, but it actually
went into the wrong direction by dropping the local Error *.
In the default KVM case, the rationale is to try the in-kernel XICS first,
and if not possible, to fallback to userland XICS. Passing errp everywhere
makes this fallback impossible if errp is &error_fatal (which happens to
be the case). And anyway, if the caller would pass a regular &local_err,
things would be worse: we could possibly pass an already set *errp to
error_setg() and crash, or return an error even in case of success.
So we definitely need a local Error * and only propagate it when we're
done with the fallback logic. This is what this patch does.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This patch adds gen_io_start()/gen_io_end() to various instructions as required
in order to boot my OpenBIOS test images on qemu-system-sparc64 with icount
enabled.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
The instance_init function of devices should always succeed to be able
to introspect the device. However, the instance_init function of the
"openprom" device can currently fail, for example like this:
$ echo "{'execute':'qmp_capabilities'}"\
"{'execute':'device-list-properties',"\
" 'arguments':{'typename':'openprom'}}" \
| sparc64-softmmu/qemu-system-sparc64 -M sun4v,accel=qtest -qmp stdio
{"QMP": {"version": {"qemu": {"micro": 91, "minor": 11, "major": 2},
"package": "build-all"}, "capabilities": []}}
{"return": {}}
RAMBlock "sun4u.prom" already registered, abort!
Aborted (core dumped)
This should not happen. Fix this problem by moving the affected code from
instance_init into a realize function instead.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
On the Alpha DP264 machine, the Cirrus VGA is I/O mapped
in the 3C0H-3CFH range, thus I/O base used by the parallel
device clashes, and since a4cb773928 the VGA is not
working:
(qemu) info mtree
address-space: memory
0000000000000000-ffffffffffffffff (prio 0, i/o): system
00000801fc000000-00000801fdffffff (prio 0, i/o): pci0-io
...
00000801fc0003b4-00000801fc0003b5 (prio 0, i/o): vga
00000801fc0003ba-00000801fc0003ba (prio 0, i/o): vga
00000801fc0003bc-00000801fc0003c3 (prio 0, i/o): parallel
^^^ ^^^^^^^^
00000801fc0003c0-00000801fc0003cf (prio 0, i/o): vga
^^^
00000801fc0003d4-00000801fc0003d5 (prio 0, i/o): vga
00000801fc0003da-00000801fc0003da (prio 0, i/o): vga
...
As there is no particular reason to use this base address
(introduced in 7bea0dd434), change to 378H which is the
default on PC machines.
Reported-by: Emilio G. Cota <cota@braap.org>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Emilio G. Cota <cota@braap.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20180614233935.26585-1-f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
CPUPPCState currently contains a number of fields containing the state of
the VPA. The VPA is a PAPR specific concept covering several guest/host
shared memory areas used to communicate some information with the
hypervisor.
As a PAPR concept this is really machine specific information, although it
is per-cpu, so it doesn't really belong in the core CPU state structure.
There's also other information that's per-cpu, but platform/machine
specific. So create a (void *)machine_data in PowerPCCPU which can be
used by the machine to locate per-cpu data. Intialization, lifetime and
cleanup of machine_data is entirely up to the machine type.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Greg Kurz <groug@kaod.org>
Tested-by: Greg Kurz <groug@kaod.org>
This extracts from the PvChip realize routine the part creating the
cores. On Power9, we will need to create the cores after the Xive
interrupt controller is created.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This moves some code out from spapr_cpu_core_realize() for clarity. No
functional change.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The spapr_realize_vcpu() function doesn't rollback in case of error.
This isn't a problem with coldplugged CPUs because the machine won't
start and QEMU will exit. Hotplug is a different story though: the
CPU thread is started under object_property_set_bool() and it assumes
it can access the CPU object.
If icp_create() fails, we return an error without unregistering the
reset handler for this CPU, and we let the underlying QEMU thread for
this CPU alive. Since spapr_cpu_core_realize() doesn't care to unrealize
already realized CPUs either, but happily frees all of them anyway, the
CPU thread crashes instantly:
(qemu) device_add host-spapr-cpu-core,core-id=1,id=gku
GKU: failing icp_create (cpu 0x11497fd0)
^^^^^^^^^^
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffee3feaa0 (LWP 24725)]
0x00000000104c8374 in object_dynamic_cast_assert (obj=0x11497fd0,
^^^^^^^^^^^^^^
pointer to the CPU object
623 trace_object_dynamic_cast_assert(obj ? obj->class->type->name
(gdb) p obj->class->type
$1 = (Type) 0x0
(gdb) p * obj
$2 = {class = 0x10ea9c10, free = 0x11244620,
^^^^^^^^^^
should be g_free
(gdb) p g_free
$3 = {<text variable, no debug info>} 0x7ffff282bef0 <g_free>
obj is a dangling pointer to the CPU that was just destroyed in
spapr_cpu_core_realize().
This patch adds proper rollback to both spapr_realize_vcpu() and
spapr_cpu_core_realize().
Signed-off-by: Greg Kurz <groug@kaod.org>
[dwg: Fixed a conflict due to a change in my tree]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Commit 94ad93bd97 (QEMU 2.12) switched to instantiate CPUs separately
but it missed to adapt the error path accordingly. If something fails in
the CPU creation loop, then the CPU object that was just created is leaked.
The error paths in this function are a bit obfuscated, and adding
yet another label to free this CPU object makes it worse. We should
move the block of the loop to a separate function, with a proper
rollback path, but this is a bigger cleanup.
For now, let's just fix the bug by adding the missing calls to
object_unref(). This will allow easier backport to older QEMU
versions.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Currently we don't have any unrealize path for pnv cpu cores. We get away
with this because we don't yet support cpu hotplug for pnv.
However, we're going to want it eventually, and in the meantime, it makes
it non-obvious why there are a bunch of allocations on the realize() path
that don't have matching frees.
So, implement the missing unrealize path.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
pnv_cpu_init() is only called from the the pnv cpu core realize path, and
really only can be called from there. So fold it into its caller, which
we also rename for brevity.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Currently, we allocate space for all the cpu objects within a single core
in one big block. This was copied from an older version of the spapr code
and requires some ugly pointer manipulation to extract the individual
objects.
This design was due to a misunderstanding of qemu lifetime conventions and
has already been changed in spapr (in 94ad93bd "spapr_cpu_core: instantiate
CPUs separately".
Make an equivalent change in pnv_core to get rid of the nasty pointer
arithmetic.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
In pnv_core_realize() we call two functions with an Error * parameter in
succession, which will go badly if they both cause errors. In fact, a
failure in either of them indicates a qemu internal error, so we can just
use &error_abort in both cases.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
spapr_cpu_init() and spapr_cpu_destroy() are only called from the spapr
cpu core realize/unrealize paths, and really can only be called from there.
Those are all short functions, so fold the pairs together for simplicity.
While we're there rename some functions and change some parameter types
for brevity and clarity.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
When writing registers that have read only bits we have to avoid
changing these bits as they may have non zero values. Make sure we use
the correct masks to mask out read only and reserved bits when
changing registers.
Also remove extra spaces from dram_control and arbitration_control
assignments.
Signed-off-by: BALATON Zoltan <balaton@eik.bme.hu>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
In the case where we have an interrupt generated externally from inputs to
bits 1 and 2 of port A and/or port B, it is necessary to expose
mos6522_update_irq() so it can be called by the interrupt source.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The datasheet indicates that the interrupt is generated by ANDing the
interrupt flags register (IFR) with the interrupt enable register (IER)
but currently there is an extra filter for the SR and timer interrupts.
Remove this extra filter to allow interrupts to be generated by external
inputs on bits 1 and 2 of ports A and B.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
According to the 6522 datasheet the shift register (SR) interrupt flag is
cleared upon write with no mention of any other interrupt flags.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
On CentOS 7.5, gcc-4.8.5-28.el7_5.1.ppc64le fails to build QEMU due to :
hw/intc/xics_kvm.c: In function ‘ics_set_kvm_state’:
hw/intc/xics_kvm.c:281:13: error: ‘ret’ may be used uninitialized in this
function [-Werror=maybe-uninitialized]
return ret;
Fix the breakage and also remove the extra error reporting as
kvm_device_access() already provides a substantial error message.
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The PMU device supercedes the CUDA device found on older New World Macs and
is supported by a larger number of guest OSs from OS 9 to OS X 10.5.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
MacOS 9 has a bug in its PMU driver whereby after configuring the ADB bus
devices it sends another write to reg 3 on both devices resetting them
both back to the same address.
Add a new disable_direct_reg3_writes property to ADBDevice to disable these
direct writes which can enabled just for the upcoming pmu-adb support.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
According to the Apple ADB documentation, register 3 is a 2-byte register
with the device address in the first byte, and the handler ID in the second
byte.
This is currently the opposite away to which QEMU returns them so switch the
order around.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The programmer switch is wired up via an external GPIO pin and can be used
to aid debugging Mac guests.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
PMU-enabled New World Macs expose their GPIOs via a separate memory region
within the macio device.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This option allows the VIA configuration to be controlled between 3
different possible setups: cuda, pmu-adb and pmu with USB rather than ADB
keyboard/mouse.
For the moment we don't do anything with the configuration except to pass
it to the macio device (the via-cuda parent) and also to the firmware via
the fw_cfg interface so that it can present the correct device tree.
The default is cuda which is the current default and so will have no
change in behaviour.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This is in preparation for adding configuration controlled via machine
options.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
If the negotiated compat mode can't be set, but raw mode is supported,
we decide to ignore the error. An so, we should free it to prevent a
memory leak.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Commit 9d6f106552 moved the last line in this block to somewhere else,
but it forgot to remove the now useless #if/#endif.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
In default_caps_with_cpu() we set spapr_cap_cfpc to broken for POWER8
processors and before.
Since we no longer require private l1d cache on POWER8 for this cap to
be set to workaround change this to default to broken for POWER7
processors and before.
Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
For cap_ppc_safe_cache to be set to workaround, we require both a l1d
cache flush instruction and private l1d cache.
On POWER8 don't require private l1d cache. This means a guest on a
POWER8 machine can make use of the cache flush workarounds.
Signed-off-by: Suraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Also, assert that we don't overflow any of two different offsets into
the TB. Both unwind and goto_tb both record a uint16_t for later use.
This fixes an arm-softmmu test case utilizing NEON in which there is
a TB generated that runs to 7800 opcodes, and compiles to 96k on an
x86_64 host. This overflows the 16-bit offset in which we record the
goto_tb reset offset. Because of that overflow, we install a jump
destination that goes to neverland. Boom.
With this reduced op count, the same TB compiles to about 48k for
aarch64, ppc64le, and x86_64 hosts, and neither assertion fires.
Cc: qemu-stable@nongnu.org
Reported-by: "Jason A. Donenfeld" <Jason@zx2c4.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Currently to run Avocado acceptance tests in an out-of-tree
build directory, we need to use the full path to the test:
build_dir$ avocado run /full/path/to/sources/qemu/tests/acceptance/boot_linux_console.py
This patch adds a symlink in the build tree to simplify the
tests invocation, allowing the same command than in in-tree builds:
build_dir$ avocado run tests/acceptance/boot_linux_console.py
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20180612173437.14462-1-f4bug@amsat.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
This test boots a Linux kernel, and checks that the given command
line was effective in two ways:
* It makes the kernel use the set "console device" as a console
* The kernel records the command line as expected in the console
Given that way too many error conditions may occur, and detecting the
kernel boot progress status may not be trivial, this test relies on a
timeout to handle unexpected situations. Also, it's *not* tagged as a
quick test for obvious reasons.
It may be useful, while interactively running/debugging this test, or
tests similar to this one, to show some of the logging channels.
Example:
$ avocado --show=QMP,console run boot_linux_console.py
Signed-off-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20180530184156.15634-6-crosa@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
The set_console() method is intended to ease higher level use cases
that require a console device.
The amount of intelligence is limited on purpose, requiring either the
device type explicitly, or the existence of a machine (pattern)
definition.
Because of the console device type selection criteria (by machine
type), users should also be able to define that. It'll then be used
for both '-machine' and for the console device type selection.
Users of the set_console() method will certainly be interested in
accessing the console device, and for that a console_socket property
has been added.
Signed-off-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20180530184156.15634-5-crosa@redhat.com>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
This patch adds a few simple behavior tests for VNC.
Signed-off-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20180530184156.15634-4-crosa@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Tests will often need to add extra arguments to QEMU command
line arguments.
Signed-off-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20180530184156.15634-3-crosa@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
This patch adds the very minimum infrastructure necessary for writing
and running functional/acceptance tests, including:
* Documentation
* The avocado_qemu.Test base test class
* One example tests (version.py)
Additional functionality is expected to be added along the tests that
require them.
Signed-off-by: Cleber Rosa <crosa@redhat.com>
Message-Id: <20180530184156.15634-2-crosa@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
[ehabkost: fix typo on testing.rst]
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
The COPYING.PYTHON file was added when we added the compatibility
argparse.py module, which was licensed under the Python Software
Foundation License Version 2.
Now the compatibility argparse.py module was removed, and we are
not carrying any code under that license anymore. Remove
COPYING.PYTHON.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <20180611180152.2681-1-ehabkost@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Use mmap_lock in user-mode to protect TCG state and the page descriptors.
In !user-mode, each vCPU has its own TCG state, so no locks needed.
Per-page locks are used to protect the page descriptors.
Per-TB locks are used in both modes to protect TB jumps.
Some notes:
- tb_lock is removed from notdirty_mem_write by passing a
locked page_collection to tb_invalidate_phys_page_fast.
- tcg_tb_lookup/remove/insert/etc have their own internal lock(s),
so there is no need to further serialize access to them.
- do_tb_flush is run in a safe async context, meaning no other
vCPU threads are running. Therefore acquiring mmap_lock there
is just to please tools such as thread sanitizer.
- Not visible in the diff, but tb_invalidate_phys_page already
has an assert_memory_lock.
- cpu_io_recompile is !user-only, so no mmap_lock there.
- Added mmap_unlock()'s before all siglongjmp's that could
be called in user-mode while mmap_lock is held.
+ Added an assert for !have_mmap_lock() after returning from
the longjmp in cpu_exec, just like we do in cpu_exec_step_atomic.
Performance numbers before/after:
Host: AMD Opteron(tm) Processor 6376
ubuntu 17.04 ppc64 bootup+shutdown time
700 +-+--+----+------+------------+-----------+------------*--+-+
| + + + + + *B |
| before ***B*** ** * |
|tb lock removal ###D### *** |
600 +-+ *** +-+
| ** # |
| *B* #D |
| *** * ## |
500 +-+ *** ### +-+
| * *** ### |
| *B* # ## |
| ** * #D# |
400 +-+ ** ## +-+
| ** ### |
| ** ## |
| ** # ## |
300 +-+ * B* #D# +-+
| B *** ### |
| * ** #### |
| * *** ### |
200 +-+ B *B #D# +-+
| #B* * ## # |
| #* ## |
| + D##D# + + + + |
100 +-+--+----+------+------------+-----------+------------+--+-+
1 8 16 Guest CPUs 48 64
png: https://imgur.com/HwmBHXe
debian jessie aarch64 bootup+shutdown time
90 +-+--+-----+-----+------------+------------+------------+--+-+
| + + + + + + |
| before ***B*** B |
80 +tb lock removal ###D### **D +-+
| **### |
| **## |
70 +-+ ** # +-+
| ** ## |
| ** # |
60 +-+ *B ## +-+
| ** ## |
| *** #D |
50 +-+ *** ## +-+
| * ** ### |
| **B* ### |
40 +-+ **** # ## +-+
| **** #D# |
| ***B** ### |
30 +-+ B***B** #### +-+
| B * * # ### |
| B ###D# |
20 +-+ D ##D## +-+
| D# |
| + + + + + + |
10 +-+--+-----+-----+------------+------------+------------+--+-+
1 8 16 Guest CPUs 48 64
png: https://imgur.com/iGpGFtv
The gains are high for 4-8 CPUs. Beyond that point, however, unrelated
lock contention significantly hurts scalability.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
tb_lock was needed when the function did retranslation. However,
since fca8a500d5 ("tcg: Save insn data and use it in
cpu_restore_state_from_tb") we don't do retranslation.
Get rid of the comment.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The acquisition of tb_lock was added when the async tlb_flush
was introduced in e3b9ca810 ("cputlb: introduce tlb_flush_* async work.")
tb_lock was there to allow us to do memset() on the tb_jmp_cache's.
However, since f3ced3c592 ("tcg: consistently access cpu->tb_jmp_cache
atomically") all accesses to tb_jmp_cache are atomic, so tb_lock
is not needed here. Get rid of it.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
This applies to both user-mode and !user-mode emulation.
Instead of relying on a global lock, protect the list of incoming
jumps with tb->jmp_lock. This lock also protects tb->cflags,
so update all tb->cflags readers outside tb->jmp_lock to use
atomic reads via tb_cflags().
In order to find the destination TB (and therefore its jmp_lock)
from the origin TB, we introduce tb->jmp_dest[].
I considered not using a linked list of jumps, which simplifies
code and makes the struct smaller. However, it unnecessarily increases
memory usage, which results in a performance decrease. See for
instance these numbers booting+shutting down debian-arm:
Time (s) Rel. err (%) Abs. err (s) Rel. slowdown (%)
------------------------------------------------------------------------------
before 20.88 0.74 0.154512 0.
after 20.81 0.38 0.079078 -0.33524904
GTree 21.02 0.28 0.058856 0.67049808
GHashTable + xxhash 21.63 1.08 0.233604 3.5919540
Using a hash table or a binary tree to keep track of the jumps
doesn't really pay off, not only due to the increased memory usage,
but also because most TBs have only 0 or 1 jumps to them. The maximum
number of jumps when booting debian-arm that I measured is 35, but
as we can see in the histogram below a TB with that many incoming jumps
is extremely rare; the average TB has 0.80 incoming jumps.
n_jumps: 379208; avg jumps/tb: 0.801099
dist: [0.0,1.0)|▄█▁▁▁▁▁▁▁▁▁▁▁ ▁▁▁▁▁▁ ▁▁▁ ▁▁▁ ▁|[34.0,35.0]
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
