When QEMU is started with hotplugged bridges (think migration):
QEMU -S -monitor stdio \
-device pci-bridge,chassis_nr=1 \
-device pci-bridge,bus=pci.1,addr=1.0,chassis_nr=2
(qemu) device_add pci-bridge,id=hpbr,bus=pci.1,addr=2.0,chassis_nr=3
(qemu) cont
it will generate AML calls to hpbr's PCNT, which doesn't exists
since it's hotplugged bridge. As result DSDT becomes malformed,
with consequences that hotplug might stop working at best or
crash guest OS at worst, when it attempts to call non existing
PCNT method or during OS guest reboot when parsing DSDT again.
IASL de-compiles malformed AML of above config DSDT as:
+ External (_SB_.PCI0.S18_.S10_.PCNT, MethodObj) // Warning: Unknown method, guessing 1 arguments
+ External (_SB_.PCI0.S18_.S19_.PCNT, MethodObj) // Warning: Unknown method, guessing 2 arguments
...
BNUM = One
DVNT (PCIU, One)
DVNT (PCID, 0x03)
- ^S08.PCNT ()
+ ^S19.PCNT (^S10.PCNT (^S08.PCNT ()))
}
}
With BSEL assignment limited only to coldplugged bridges [1],
it should be possible to add PCNT call to a child bridge only
if the child has BSEL property, otherwise ignore it since it's
hotplugged. Which should fix the issue.
1) ("pci: acpihp: assign BSEL only to coldplugged bridges")
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230112140312.3096331-13-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
ACPI PCI hotplug would broken after bridge hotplug and then migration
if hotplugged bridge were specified on target at command line.
Currently it's not possible since, 'hotplugged' property was made
read-only for some time now.
The issue would happen due to BSEL being assigned to all bridges
during 1st 'reset':
source seq:
1. start 'pc' machine => sets BSEL to 0 on pci.0 (host-bridge)
2. hotplug bridge, no bsel is assigned (so far is ok)
target seq:
1. start 'pc' machine with
-S -device pci-bridge,id=hp_br,hotplugged=on
BSEL gets assigned to as follows
hp_br: 0
pci.0: 1
as result hotplug requests with migrated AML generated on source
would be misdirected to 'hp_br' instead of intended pci.0
While it's not issue at the moment, it's based on implicit assumptions
* 'hotplugged' property is read-only
* 1st reset happens before QEMU drops into monitor mode
which lets add hotplugged on source bridges as hotplugged ones
(anything added at that stage counts as hotplugged
(yet another assumption))
All of it looks too fragile to me, so lets restrict BSEL only
to cold-plugged bridges explicitly.
Migration wise it shouldn't break anything since assignment order
stays the same:
* user can't specify 'hotplugged=on' on CLI
* user can't specify 'hotplugged=off' at monitor stage or later
on older QEMU versions where 'hotplugged' is RW, hotplug is broken
after migration anyways and we cannot do anything to fix that.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230112140312.3096331-12-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
piix4_pm_reset() is calling acpi_pcihp_reset() when ACPI PCI hotplug
is disabled, which leads to assigning BSEL properties to bridges on path
acpi_set_bsel()
...
if (qbus_is_hotpluggable(BUS(bus))) {
// above happens to be true by default (though it's SHPC hotplug handler)
// set BSEL
}
At the moment the issue is masked by the fact that we use not only BSEL,
to decide if we should generated hoplug AML but also pcihp_bridge_en knob.
However the later patches will drop dependency on pcihp_bridge_en,
and use only BSEL exclusively to decide if hotplug AML for slots should be built,
which exposes issue.
We should not ever call acpi_pcihp_reset() if ACPI PCI hotplug is disabled,
make it so.
PS:
* Q35 does the right thing (i.e. it calls acpi_pcihp_reset only when pcihp is enabled)
* the issue also makes acpi_pcihp_update() logic run on SHPC enabled bridges,
which seems to be harmless
Fixes: 3d7e78aa77 ("Introduce a new flag for i440fx to disable PCI hotplug on the root bus")
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230112140312.3096331-11-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When ACPI PCI hotplug for Q35 was introduced (6.1), it was implemented
by hiding HPC capability on PCIE slot. That however led to a number of
regressions and to fix it, it was decided to keep HPC cap exposed
in ACPI PCI hotplug case and force guest in ACPI PCI hotplug mode
by other means [1].
That reduced meaning of x-native-hotplug to a compat knob [2] for
broken 6.1 machine type.
Rename property to match its current purpose.
1) 211afe5c69 (hw/i386/acpi-build: Deny control on PCIe Native Hot-plug in _OSC)
2) c318bef762 (hw/acpi/ich9: Add compat prop to keep HPC bit set for 6.1 machine type)
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230112140312.3096331-10-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
'use_uefi' is used for the flag is a part of 'test_data *data'
argument that is passed to the same functions, which
makes use_uefi argument redundant.
Drop it and use 'data::uefi_*' directly, instead.
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20230112140312.3096331-7-imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The setup_data links are appended to the compressed kernel image. Since
the kernel image is typically loaded at 0x100000, setup_data lives at
`0x100000 + compressed_size`, which does not get relocated during the
kernel's boot process.
The kernel typically decompresses the image starting at address
0x1000000 (note: there's one more zero there than the compressed image
above). This usually is fine for most kernels.
However, if the compressed image is actually quite large, then
setup_data will live at a `0x100000 + compressed_size` that extends into
the decompressed zone at 0x1000000. In other words, if compressed_size
is larger than `0x1000000 - 0x100000`, then the decompression step will
clobber setup_data, resulting in crashes.
Visually, what happens now is that QEMU appends setup_data to the kernel
image:
kernel image setup_data
|--------------------------||----------------|
0x100000 0x100000+l1 0x100000+l1+l2
The problem is that this decompresses to 0x1000000 (one more zero). So
if l1 is > (0x1000000-0x100000), then this winds up looking like:
kernel image setup_data
|--------------------------||----------------|
0x100000 0x100000+l1 0x100000+l1+l2
d e c o m p r e s s e d k e r n e l
|-------------------------------------------------------------|
0x1000000 0x1000000+l3
The decompressed kernel seemingly overwriting the compressed kernel
image isn't a problem, because that gets relocated to a higher address
early on in the boot process, at the end of startup_64. setup_data,
however, stays in the same place, since those links are self referential
and nothing fixes them up. So the decompressed kernel clobbers it.
Fix this by appending setup_data to the cmdline blob rather than the
kernel image blob, which remains at a lower address that won't get
clobbered.
This could have been done by overwriting the initrd blob instead, but
that poses big difficulties, such as no longer being able to use memory
mapped files for initrd, hurting performance, and, more importantly, the
initrd address calculation is hard coded in qboot, and it always grows
down rather than up, which means lots of brittle semantics would have to
be changed around, incurring more complexity. In contrast, using cmdline
is simple and doesn't interfere with anything.
The microvm machine has a gross hack where it fiddles with fw_cfg data
after the fact. So this hack is updated to account for this appending,
by reserving some bytes.
Fixup-by: Michael S. Tsirkin <mst@redhat.com>
Cc: x86@kernel.org
Cc: Philippe Mathieu-Daudé <philmd@linaro.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Message-Id: <20221230220725.618763-1-Jason@zx2c4.com>
Message-ID: <20230128061015-mutt-send-email-mst@kernel.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Eric Biggers <ebiggers@google.com>
Tested-by: Mathias Krause <minipli@grsecurity.net>
It seems not super clear on when iova_tree is used, and why. Add a rich
comment above iova_tree to track why we needed the iova_tree, and when we
need it.
Also comment for the map/unmap messages, on how they're used and
implications (e.g. unmap can be larger than the mapped ranges).
Suggested-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <20230109193727.1360190-1-peterx@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Fixup the migration compatibility for existing machine types
so that they do not enable msi-x.
Symptom:
(qemu) qemu: get_pci_config_device: Bad config data: i=0x34 read: 84 device: 98 cmask: ff wmask: 0 w1cmask:0
qemu: Failed to load PCIDevice:config
qemu: Failed to load virtio-rng:virtio
qemu: error while loading state for instance 0x0 of device '0000:00:03.0/virtio-rng'
qemu: load of migration failed: Invalid argument
Note: This fix will break migration from 7.2->7.2-fixed with this patch
bz: https://bugzilla.redhat.com/show_bug.cgi?id=2155749
Fixes: 9ea02e8f1 ("virtio-rng-pci: Allow setting nvectors, so we can use MSI-X")
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <20230109105809.163975-1-dgilbert@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: David Daney <david.daney@fungible.com>
Fixes: 9ea02e8f1 ("virtio-rng-pci: Allow setting nvectors, so we can use MSI-X")<br>
Signed-off-by: Dr. David Alan Gilbert <<a href="mailto:dgilbert@redhat.com" target="_blank">dgilbert@redhat.com</a>><br>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Pressing attention button has special meaning when power indicator is
blinking. Better just not do it.
For example, trying to remove device immediately after hotplug leads to
both commands succeded but device not actually unrealized.
Same thing for PCIE hotplug was done in
81124b3c7a "pcie: add power indicator blink check"
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
Message-Id: <20221116214458.82090-1-vsementsov@yandex-team.ru>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This test is failing in gtk-vnc on Darwin:
$ make check-qtest-aarch64
...
19/20 qemu:qtest+qtest-aarch64 / qtest-aarch64/vnc-display-test
ERROR **: 10:42:35.488: vnc-error: Unsupported auth type 17973672
While QEMU picks the sigaltstack coroutine backend, gtk-vnc uses
the ucontext coroutine backend, which might be broken on Darwin.
Disable this test (current problem being investigated in this thread:
https://lore.kernel.org/qemu-devel/Y8kw6X6keB5l53nl@redhat.com/).
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230119120514.28778-4-philmd@linaro.org>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
If we don't specify any machine, an architecture default
might be picked. But some architectures don't provide any
default, such ARM:
$ make check-qtest-aarch64
...
19/20 qemu:qtest+qtest-aarch64 / qtest-aarch64/vnc-display-test
qemu-system-aarch64: No machine specified, and there is no default
Since we don't need any particular machine to run this VNC
test, use the 'none' machine.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230119120514.28778-3-philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
While this test is skipped on Windows, we still get when building:
tests/qtest/vnc-display-test.c:22:20: warning: unused function 'on_vnc_error' [-Wunused-function]
static inline void on_vnc_error(VncConnection* self,
^
tests/qtest/vnc-display-test.c:28:20: warning: unused function 'on_vnc_auth_failure' [-Wunused-function]
static inline void on_vnc_auth_failure(VncConnection *self,
^
2 warnings generated.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-Id: <20230119120514.28778-2-philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Since commit efc6c070ac ("configure: Add a test for the minimum
compiler version") the minimum compiler version required for GCC
is 4.8, which supports __builtin_bswap().
Remove the NetBSD specific ifdef'ry.
This reverts commit 1360677cfe
("makes NetBSD use the native bswap functions").
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230111163147.71761-7-philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Since commit efc6c070ac ("configure: Add a test for the minimum
compiler version") the minimum compiler version required for GCC
is 4.8, which supports __builtin_bswap().
Remove the FreeBSD specific ifdef'ry.
This reverts commit de03c3164a
("bswap: Fix build on FreeBSD 10.0").
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230111163147.71761-6-philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Since commit efc6c070ac ("configure: Add a test for the minimum
compiler version") the minimum compiler version required for GCC
is 4.8, which supports __builtin_bswap().
Remove the Haiku specific ifdef'ry.
This reverts commit 652a46ebba
("bswap.h: Include <endian.h> on Haiku for bswap operations").
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230111163147.71761-5-philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
No need to recompile the dtc submodule here again and again, we can
use the pre-built binary from the distribution instead.
(And this will also help in case we finally get rid of the dtc submodule
in QEMU one day)
Message-Id: <20230124143824.844040-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Each job uses its own addons section nowadays, so the generic section
is completely unused and outdated, thus we can remove it now.
Message-Id: <20230119135914.2040853-1-thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
No need to compile-test third party submodules over and over again if
we can simply use the pre-build library from the distribution instead.
By also adding --enable-fdt=system to the configure options, we can
also avoid to check out the "dtc" submodule here.
Message-Id: <20230120075330.2076773-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
qemu-system-nios2 uses the functions from libfdt in hw/nios2/boot.c,
so this target has to be marked with TARGET_NEED_FDT=y in its config
file.
Message-Id: <20230119125745.2028814-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
'-drive if=none' is meant for configuring back-end devices only, so this
got marked as deprecated in QEMU 6.2. Users should now only use the new
way with '-drive if=pflash' instead.
Message-Id: <20230112083921.887828-1-thuth@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
We are facing the issues that our test logs in the gitlab CI are
too big (and thus cut off). The bios-tables-test is one of the few
qtests that prints many lines of output by default when running with
V=1, so it contributes to this problem. Almost all other qtests are
silent with V=1 and only print debug messages with V=2 and higher.
Thus let's change the bios-tables-test to behave more like the
other tests and only print the debug messages with V=2 (or higher).
Message-Id: <20230118125132.1694469-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
We are still facing the issues that our test logs in the gitlab CI
are too big (and thus cut off). A huge part is still caused by the
qom-test that prints the path and name of each object it looks at
by default. That's too much. Let's be silent by default, and only
print the object path+name when running with V=2 (and the properties
only with V=3 and higher).
Message-Id: <20230118122557.1668860-1-thuth@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Igor Mammedov