This allows the virtio-scsi-pci device to batch the setup of all its
host notifiers. This significantly improves boot time of VMs with a
high number of vCPUs, e.g. from 6m5.563s down to 1m2.884s for a
pseries machine with 384 vCPUs.
Note that memory_region_transaction_commit() must be called before
virtio_bus_cleanup_host_notifier() because the latter might close
ioeventfds that the transaction still assumes to be around when it
commits.
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20210407143501.244343-5-groug@kaod.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Host notifiers are guaranteed to be idle until the callbacks are
hooked up with virtio_queue_aio_set_host_notifier_handler(). They
thus don't need to be set or unset with the AioContext lock held.
Do this outside the critical section, like virtio-blk already
does : basically downgrading virtio_scsi_vring_init() to only
setup the host notifier and set the callback in the caller.
This will allow to batch addition/deletion of ioeventds in
a single memory transaction, which is expected to greatly
improve initialization time.
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20210407143501.244343-4-groug@kaod.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
This allows the virtio-blk-pci device to batch the setup of all its
host notifiers. This significantly improves boot time of VMs with a
high number of vCPUs, e.g. from 3m26.186s down to 0m58.023s for a
pseries machine with 384 vCPUs.
Note that memory_region_transaction_commit() must be called before
virtio_bus_cleanup_host_notifier() because the latter might close
ioeventfds that the transaction still assumes to be around when it
commits.
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20210407143501.244343-3-groug@kaod.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
When dataplane multiqueue support was added in QEMU 2.7, the path
that would rollback guest notifiers assignment in case of error
simply got dropped.
Later on, when Error was added to blk_set_aio_context() in QEMU 4.1,
another error path was introduced, but it ommits to rollback both
host and guest notifiers.
It seems cleaner to fix the rollback path in one go. The patch is
simple enough that it can be adjusted if backported to a pre-4.1
QEMU.
Fixes: 51b04ac5c6 ("virtio-blk: dataplane multiqueue support")
Cc: stefanha@redhat.com
Fixes: 97896a4887 ("block: Add Error to blk_set_aio_context()")
Cc: kwolf@redhat.com
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210407143501.244343-2-groug@kaod.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The get_vmstate_memory_region() method from PCDIMMDeviceClass is only
ever called from this class and is never overridden, so it can be converted
into an ordinary function.
This saves us from having to do an indirect call in order to reach it.
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Message-Id: <f42da25471dc4b967796642388294e61e6587047.1619303649.git.maciej.szmigiero@oracle.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Address was swapped with value when writing MMIO registers, so the user
saw garbage in lot of cases. The interrupt status was not correctly set.
Signed-off-by: Roman Kapl <rka@sysgo.com>
Message-Id: <20210427110504.10878-1-rka@sysgo.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The VirtIOFeature structure isn't modified, mark it const.
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210511104157.2880306-2-philmd@redhat.com>
Do the same as in commit
(4d027afeb3 Virt: ACPI: fix qemu assert due to re-assigned table data address)
for remaining tables that happen to use saved at
the beginning pointer to build header to avoid assert
when table_data is relocated due to implicit re-size.
In this case user is trying to start Windows 10 and getting assert at
hw/acpi/bios-linker-loader.c:239:
bios_linker_loader_add_checksum: Assertion `start_offset < file->blob->len' failed.
Fixes: https://bugs.launchpad.net/bugs/1923497
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <20210414084356.3792113-1-imammedo@redhat.com>
Cc: mst@redhat.com, qemu-stable@nongnu.org
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
AMD IOMMU PTEs have a special mode allowing to specify an arbitrary page
size. Quoting the AMD IOMMU specification: "When the Next Level bits [of
a pte] are 7h, the size of the page is determined by the first zero bit
in the page address, starting from bit 12."
So if the lowest bits of the page address is 0, the page is 8kB. If the
lowest bits are 011, the page is 32kB. Currently pte_override_page_mask()
doesn't compute the right value for this page size and amdvi_translate()
can return the wrong guest-physical address. With a Linux guest, DMA
from SATA devices accesses the wrong memory and causes probe failure:
qemu-system-x86_64 ... -device amd-iommu -drive id=hd1,file=foo.bin,if=none \
-device ahci,id=ahci -device ide-hd,drive=hd1,bus=ahci.0
[ 6.613093] ata1.00: qc timeout (cmd 0xec)
[ 6.615062] ata1.00: failed to IDENTIFY (I/O error, err_mask=0x4)
Fix the page mask.
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Message-Id: <20210421084007.1190546-1-jean-philippe@linaro.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Commit 1901b4967c changed the nvme device from using a bar exclusive
for MSI-x to sharing it on bar0.
Unfortunately, the msix_uninit_exclusive_bar() call remains in
nvme_exit() which causes havoc when the device is removed with, say,
device_del. Fix this.
Additionally, a subregion is added but it is not removed on exit which
causes a reference to linger and the drive to never be unlocked.
Fixes: 1901b4967c ("hw/block/nvme: move msix table and pba to BAR 0")
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
While processing SCSI i/o requests in mptsas_process_scsi_io_request(),
the Megaraid emulator appends new MPTSASRequest object 'req' to
the 's->pending' queue. In case of an error, this same object gets
dequeued in mptsas_free_request() only if SCSIRequest object
'req->sreq' is initialised. This may lead to a use-after-free issue.
Since s->pending is actually not used, simply remove it from
MPTSASState.
Cc: qemu-stable@nongnu.org
Signed-off-by: Michael Tokarev <mjt@tls.msk.ru>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reported-by: Cheolwoo Myung <cwmyung@snu.ac.kr>
Message-id: 20210419134247.1467982-1-f4bug@amsat.org
Message-Id: <20210416102243.1293871-1-mjt@msgid.tls.msk.ru>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Reported-by: Cheolwoo Myung <cwmyung@snu.ac.kr>
BugLink: https://bugs.launchpad.net/qemu/+bug/1914236 (CVE-2021-3392)
Fixes: e351b82611 ("hw: Add support for LSI SAS1068 (mptsas) device")
[PMD: Reworded description, added more tags]
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The SSE-300 has a Cortex-M55 (which was the whole reason for us
modelling it), but we forgot to actually update the code to let it
have a different CPU type from the IoTKit and SSE-200. Add CPU type
as a field for ARMSSEInfo instead of hardcoding it to always use a
Cortex-M33.
Buglink: https://bugs.launchpad.net/qemu/+bug/1923861
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210416104010.13228-1-peter.maydell@linaro.org
SSE-300 currently shares the SSE-200 Property array. This is
bad principally because the default values of the CPU0_FPU
and CPU0_DSP properties disable the FPU and DSP on the CPU.
That is correct for the SSE-200 but not the SSE-300.
Give the SSE-300 its own Property array with the correct
SSE-300 specific settings:
* SSE-300 has only one CPU, so no CPU1* properties
* SSE-300 CPU has FPU and DSP
Buglink: https://bugs.launchpad.net/qemu/+bug/1923861
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210415182353.8173-1-peter.maydell@linaro.org
Make virtio-fs take into account server capabilities.
Just returning requested features assumes they all of then are implemented
by server and results in setting unsupported configuration if some of them
are absent.
Signed-off-by: Anton Kuchin <antonkuchin@yandex-team.ru>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
With changes suggested by Stefan
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Here's what I hope is the last ppc related pull request for qemu-6.0.
The 2 patches here revert a behavioural change that after further
discussion we concluded was a bad idea (adding a timeout for
possibly-failed hot unplug requests). Instead it implements a
different approach to the original problem: we again let unplug
requests the guest doesn't respond to remain pending indefinitely, but
no longer allow those to block attempts to retry the same unplug
again.
The change is a bit more complex than I'd like for this late in the
freeze. Nonetheless, I think it's important to merge this for 6.0, so
we don't allow a release which has the probably-a-bad-idea timeout
behaviour.
-----BEGIN PGP SIGNATURE-----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=fIJT
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/dg-gitlab/tags/ppc-for-6.0-20210412' into staging
ppc patch queue for 2021-04-21
Here's what I hope is the last ppc related pull request for qemu-6.0.
The 2 patches here revert a behavioural change that after further
discussion we concluded was a bad idea (adding a timeout for
possibly-failed hot unplug requests). Instead it implements a
different approach to the original problem: we again let unplug
requests the guest doesn't respond to remain pending indefinitely, but
no longer allow those to block attempts to retry the same unplug
again.
The change is a bit more complex than I'd like for this late in the
freeze. Nonetheless, I think it's important to merge this for 6.0, so
we don't allow a release which has the probably-a-bad-idea timeout
behaviour.
# gpg: Signature made Mon 12 Apr 2021 06:25:58 BST
# gpg: using RSA key 75F46586AE61A66CC44E87DC6C38CACA20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>" [full]
# gpg: aka "David Gibson (Red Hat) <dgibson@redhat.com>" [full]
# gpg: aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>" [full]
# gpg: aka "David Gibson (kernel.org) <dwg@kernel.org>" [unknown]
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E 87DC 6C38 CACA 20D9 B392
* remotes/dg-gitlab/tags/ppc-for-6.0-20210412:
spapr.c: always pulse guest IRQ in spapr_core_unplug_request()
spapr: rollback 'unplug timeout' for CPU hotunplugs
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
When adding the Reset register in commit 5790b757cf we
forgot to migrate it.
While it is possible a VM using the PIIX4 is migrated just
after requesting a system shutdown, it is very unlikely.
However when restoring a migrated VM, we might have the
RCR bit #4 set on the stack and when the VM resume it
directly shutdowns.
Add a post_load() migration handler and set the default
RCR value to 0 for earlier versions, assuming the VM was
not going to shutdown before migration.
Fixes: 5790b757cf ("piix4: Add the Reset Control Register")
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Message-Id: <20210324200334.729899-1-f4bug@amsat.org>
TYPE_VIA_PM calls apm_init() in via_pm_realize(), so
requires APM to be selected.
Reported-by: BALATON Zoltan <balaton@eik.bme.hu>
Fixes: dd0ff8191a ("isa: express SuperIO dependencies with Kconfig")
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210302080531.913802-1-f4bug@amsat.org>
When a CDB has been received and is about to be submitted to the SCSI layer
via one of the ESP select commands, ensure that do_cmd is set to zero before
executing the command.
Otherwise a guest executing 2 valid CDBs in quick sequence can invoke the SCSI
.transfer_data callback again before do_cmd is set to zero by the callback
function triggering an assert at the start of esp_transfer_data().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210407195801.685-12-mark.cave-ayland@ilande.co.uk>
Instead let the SCSI layer invoke the .cancel callback itself to cancel and
reset the request state.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210407195801.685-11-mark.cave-ayland@ilande.co.uk>
If a guest transfers the message out/command phase data using DMA with a TC
that is larger than the cmdfifo size then the cmdfifo overflows triggering
an assert. Limit the size of the transfer to the free space available in
cmdfifo.
Buglink: https://bugs.launchpad.net/qemu/+bug/1919036
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-10-mark.cave-ayland@ilande.co.uk>
If the guest tries to read a CDB using DMA and cmdfifo is not empty then it is
possible to overflow cmdfifo.
Since this can only occur by issuing deliberately incorrect instruction
sequences, ensure that the maximum length of the CDB transferred to cmdfifo is
limited to the available free space within cmdfifo.
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-9-mark.cave-ayland@ilande.co.uk>
If the guest tries to execute a CDB when cmdfifo is not empty before the start
of the message out phase then clearing the message out phase data will cause
cmdfifo to underflow due to cmdfifo_cdb_offset being larger than the amount of
data within.
Since this can only occur by issuing deliberately incorrect instruction
sequences, ensure that the maximum length of esp_fifo_pop_buf() is limited to
the size of the data within cmdfifo.
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-8-mark.cave-ayland@ilande.co.uk>
When about to execute a SCSI command, ensure that cmdfifo is not empty and
current_dev is non-NULL. This can happen if the guest tries to execute a TI
(Transfer Information) command without issuing one of the select commands
first.
Buglink: https://bugs.launchpad.net/qemu/+bug/1910723
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-7-mark.cave-ayland@ilande.co.uk>
The const pointer returned by fifo8_pop_buf() lies directly within the array used
to model the FIFO. Building with address sanitizers enabled shows that if the
caller expects a minimum number of bytes present then if the FIFO is nearly full,
the caller may unexpectedly access past the end of the array.
Introduce esp_fifo_pop_buf() which takes a destination buffer and performs a
memcpy() in it to guarantee that the caller cannot overwrite the FIFO array and
update all callers to use it. Similarly add underflow protection similar to
esp_fifo_push() and esp_fifo_pop() so that instead of triggering an assert()
the operation becomes a no-op.
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20210407195801.685-6-mark.cave-ayland@ilande.co.uk>
Each FIFO currently has its own pop functions with the only difference being
the capacity check. The original reason for this was that the fifo8
implementation doesn't have a formal API for retrieving the FIFO capacity,
however there are multiple examples within QEMU where the capacity field is
accessed directly.
Change esp_fifo_pop() to access the FIFO capacity directly and then consolidate
esp_cmdfifo_pop() into esp_fifo_pop().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-5-mark.cave-ayland@ilande.co.uk>
Each FIFO currently has its own push functions with the only difference being
the capacity check. The original reason for this was that the fifo8
implementation doesn't have a formal API for retrieving the FIFO capacity,
however there are multiple examples within QEMU where the capacity field is
accessed directly.
Change esp_fifo_push() to access the FIFO capacity directly and then consolidate
esp_cmdfifo_push() into esp_fifo_push().
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-4-mark.cave-ayland@ilande.co.uk>
The code for write_response() has always used the FIFO to store the data for
the status/message in phases, even for DMA transactions. Switch to using a
separate buffer that can be used directly for DMA transactions and restrict
the FIFO use to the non-DMA case.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210407195801.685-3-mark.cave-ayland@ilande.co.uk>
After issuing a SCSI command the SCSI layer can call the SCSIBusInfo .cancel
callback which resets both current_req and current_dev to NULL. If any data
is left in the transfer buffer (async_len != 0) then the next TI (Transfer
Information) command will attempt to reference the NULL pointer causing a
segfault.
Buglink: https://bugs.launchpad.net/qemu/+bug/1910723
Buglink: https://bugs.launchpad.net/qemu/+bug/1909247
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Tested-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20210407195801.685-2-mark.cave-ayland@ilande.co.uk>
If QEMU is launched with the -S option then the ESPState mig_version_id property
is left unset due to the ordering of the VMState fields in the VMStateDescription
for sysbusespscsi and pciespscsi. If the VM is migrated and restored in this
stopped state, the version tests in the vmstate_esp VMStateDescription and
esp_post_load() become confused causing the migration to fail.
Fix the ordering problem by moving the setting of mig_version_id to a common
esp_pre_save() function which is invoked first by both sysbusespscsi and
pciespscsi rather than at the point where ESPState is itself serialised into the
migration stream.
Buglink: https://bugs.launchpad.net/qemu/+bug/1922611
Fixes: 0bd005be78 ("esp: add vmstate_esp version to embedded ESPState")
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20210407124842.32695-1-mark.cave-ayland@ilande.co.uk>
Each board in mps2-tz.c specifies a RAMInfo[] array providing
information about each RAM in the board. The .mpc field of the
RAMInfo struct specifies which MPC, if any, the RAM is attached to.
We already assert if the array doesn't have any entry for an MPC, but
we don't diagnose the error of using the same MPC number twice (which
is quite easy to do by accident if copy-and-pasting structure
entries).
Enhance find_raminfo_for_mpc() so that it detects multiple entries
for the MPC as well as missing entries.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20210409150527.15053-3-peter.maydell@linaro.org
The AN524 has three MPCs: one for the BRAM, one for the QSPI flash,
and one for the DDR. We incorrectly set the .mpc field in the
RAMInfo struct for the SRAM block to 1, giving it the same MPC we are
using for the QSPI. The effect of this was that the QSPI didn't get
mapped into the system address space at all, via an MPC or otherwise,
and guest programs which tried to read from the QSPI would get a bus
error. Correct the SRAM RAMInfo to indicate that it does not have an
associated MPC.
Fixes: 25ff112a8c ("hw/arm/mps2-tz: Add new mps3-an524 board")
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20210409150527.15053-2-peter.maydell@linaro.org
In emulation of the CFGI_STE_RANGE command, we now take StreamID as the
start of the invalidation range, regardless of whatever the Range is,
whilst the spec clearly states that
- "Invalidation is performed for an *aligned* range of 2^(Range+1)
StreamIDs."
- "The bottom Range+1 bits of the StreamID parameter are IGNORED,
aligning the range to its size."
Take CFGI_ALL (where Range == 31) as an example, if there are some random
bits in the StreamID field, we'll fail to perform the full invalidation but
get a strange range (e.g., SMMUSIDRange={.start=1, .end=0}) instead. Rework
the emulation a bit to get rid of the discrepancy with the spec.
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Acked-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20210402100449.528-1-yuzenghui@huawei.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The GSIV values in SMMUv3 IORT node are not correct as they don't match
the SMMUIrq enumeration, which describes the IRQ<->PIN mapping used by
our emulated vSMMU.
Fixes: a703b4f6c1 ("hw/arm/virt-acpi-build: Add smmuv3 node in IORT table")
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
Acked-by: Eric Auger <eric.auger@redhat.com>
Message-id: 20210402084731.93-1-yuzenghui@huawei.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
For most commands, when issuing an AIO, the BlockAIOCB is stored in the
NvmeRequest aiocb pointer when the AIO is issued. The main use of this
is cancelling AIOs when deleting submission queues (it is currently not
used for Abort).
However, some commands like Dataset Management Zone Management Send
(zone reset) may involve more than one AIO and here the AIOs are issued
without saving a reference to the BlockAIOCB. This is a problem since
nvme_del_sq() will attempt to cancel outstanding AIOs, potentially with
an invalid BlockAIOCB since the aiocb pointer is not NULL'ed when the
request structure is recycled.
Fix this by
1. making sure the aiocb pointer is NULL'ed when requests are recycled
2. only attempt to cancel the AIO if the aiocb is non-NULL
3. if any AIOs could not be cancelled, drain all aio as a last resort.
Fixes: dc04d25e2f ("hw/block/nvme: add support for the format nvm command")
Fixes: c94973288c ("hw/block/nvme: add broadcast nsid support flush command")
Fixes: e4e430b3d6 ("hw/block/nvme: add simple copy command")
Fixes: 5f5dc4c6a9 ("hw/block/nvme: zero out zones on reset")
Fixes: 2605257a26 ("hw/block/nvme: add the dataset management command")
Cc: Gollu Appalanaidu <anaidu.gollu@samsung.com>
Cc: Minwoo Im <minwoo.im@samsung.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Minwoo Im <minwoo.im.dev@gmail.com>
nvme_compare() fails to store the aiocb from the blk_aio_preadv() call.
Fix this.
Fixes: 0a384f923f ("hw/block/nvme: add compare command")
Cc: Gollu Appalanaidu <anaidu.gollu@samsung.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
Reviewed-by: Minwoo Im <minwoo.im.dev@gmail.com>
nvme_map_prp needs to calculate the number of list entries based on the
offset value. For the subsequent PRP2 list, need to ensure the number of
entries is within the MAX number of PRP entries for a page.
Signed-off-by: Padmakar Kalghatgi <p.kalghatgi@samsung.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Commit 47c8c915b1 fixed a problem where multiple spapr_drc_detach()
requests were breaking QEMU. The solution was to just spapr_drc_detach()
once, and use spapr_drc_unplug_requested() to filter whether we already
detached it or not. The commit also tied the hotplug request to the
guest in the same condition.
Turns out that there is a reliable way for a CPU hotunplug to fail. If a
guest with one CPU hotplugs a CPU1, then offline CPU0s via 'echo 0 >
/sys/devices/system/cpu/cpu0/online', then attempts to hotunplug CPU1,
the kernel will refuse it because it's the last online CPU of the
system. Given that we're pulsing the IRQ only in the first try, in a
failed attempt, all other CPU1 hotunplug attempts will fail, regardless
of the online state of CPU1 in the kernel, because we're simply not
letting the guest know that we want to hotunplug the device.
Let's move spapr_hotplug_req_remove_by_index() back out of the "if
(!spapr_drc_unplug_requested(drc))" conditional, allowing for multiple
'device_del' requests to the same CPU core to reach the guest, in case
the CPU core didn't fully hotunplugged previously.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20210401000437.131140-3-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The pseries machines introduced the concept of 'unplug timeout' for CPU
hotunplugs. The idea was to circunvent a deficiency in the pSeries
specification (PAPR), that currently does not define a proper way for
the hotunplug to fail. If the guest refuses to release the CPU (see [1]
for an example) there is no way for QEMU to detect the failure.
Further discussions about how to send a QAPI event to inform about the
hotunplug timeout [2] exposed problems that weren't predicted back when
the idea was developed. Other QEMU machines don't have any type of
hotunplug timeout mechanism for any device, e.g. ACPI based machines
have a way to make hotunplug errors visible to the hypervisor. This
would make this timeout mechanism exclusive to pSeries, which is not
ideal.
The real problem is that a QAPI event that reports hotunplug timeouts
puts the management layer (namely Libvirt) in a weird spot. We're not
telling that the hotunplug failed, because we can't be 100% sure of
that, and yet we're resetting the unplug state back, preventing any
DEVICE_DEL events to reach out in case the guest decides to release the
device. Libvirt would need to inspect the guest itself to see if the
device was released or not, otherwise the internal domain states will be
inconsistent. Moreover, Libvirt already has an 'unplug timeout'
concept, and a QEMU side timeout would need to be juggled together with
the existing Libvirt timeout.
All this considered, this solution ended up creating more trouble than
it solved. This patch reverts the 3 commits that introduced the timeout
mechanism for CPU hotplugs in pSeries machines.
This reverts commit 4515a5f786
"qemu_timer.c: add timer_deadline_ms() helper"
This reverts commit d1c2e3ce3d
"spapr_drc.c: add hotunplug timeout for CPUs"
This reverts commit 51254ffb32
"spapr_drc.c: introduce unplug_timeout_timer"
[1] https://bugzilla.redhat.com/show_bug.cgi?id=1911414
[2] https://lists.gnu.org/archive/html/qemu-devel/2021-03/msg04682.html
CC: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20210401000437.131140-2-danielhb413@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Calling qdev_get_machine() from a QOM instance_init function is
fragile because we can't be sure the machine object actually
exists. And this happens to break when passing ",help" on the
command line to get the list of properties for a CPU core
device types :
$ ./qemu-system-ppc64 -device power8_v2.0-spapr-cpu-core,help
qemu-system-ppc64: ../../hw/core/machine.c:1290:
qdev_get_machine: Assertion `machine != NULL' failed.
Aborted (core dumped)
This used to work before QEMU 5.0, but commit 3df261b667
unwillingly introduced a subtle regression : the above command
line needs to create an instance but the instance_init function
of the base class calls qdev_get_machine() before
qemu_create_machine() has been called, which is a programming bug.
Use current_machine instead. It is okay to skip the setting of
nr_thread in this case since only its type is displayed.
Fixes: 3df261b667 ("softmmu/vl.c: Handle '-cpu help' and '-device help' before 'no default machine'")
Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Greg Kurz <groug@kaod.org>
Cc: peter.maydell@linaro.org
Message-Id: <20210409160339.500167-3-groug@kaod.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Setting the 'fallback' property corrupts the QOM instance state
(FDCtrlSysBus) because it accesses an incorrect offset (it uses
the offset of the FDCtrlISABus state).
Cc: qemu-stable@nongnu.org
Fixes: a73275dd6f ("fdc: Add fallback option")
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210407133742.1680424-1-f4bug@amsat.org>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
ccw_dstream_read/write functions returned values are sometime
not taking into account and reported back to the upper level
of interpretation of CCW instructions.
It follows that accessing an invalid address does not trigger
a subchannel status program check to the guest as it should.
Let's test the return values of ccw_dstream_write[_buf] and
ccw_dstream_read[_buf] and report it to the caller.
Cc: qemu-stable@nongnu.org
Signed-off-by: Pierre Morel <pmorel@linux.ibm.com>
Acked-by: Halil Pasic <pasic@linux.ibm.com>
Message-Id: <1617899529-9329-2-git-send-email-pmorel@linux.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Greg Kurz
Maciej S. Szmigiero
Roman Kapl
Philippe Mathieu-Daudé
Igor Mammedov
Jean-Philippe Brucker
Klaus Jensen
Michael Tokarev
Peter Maydell
Anton Kuchin
Philippe Mathieu-Daudé
Mark Cave-Ayland
Zenghui Yu
Padmakar Kalghatgi
Daniel Henrique Barboza
Pierre Morel