Graeme Gregory [Fri, 28 Aug 2020 09:02:43 +0000 (10:02 +0100)]
hw/arm/sbsa-ref: fix typo breaking PCIe IRQs
Fixing a typo in a previous patch that translated an "i" to a 1
and therefore breaking the allocation of PCIe interrupts. This was
discovered when virtio-net-pci devices ceased to function correctly.
Peter Maydell [Thu, 27 Aug 2020 15:59:02 +0000 (16:59 +0100)]
Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging
virtio,pc,acpi: features, fixes
better number of queues for vhost
smbios speed options
acpi fixes
Signed-off-by: Michael S. Tsirkin <[email protected]>
# gpg: Signature made Thu 27 Aug 2020 13:33:49 BST
# gpg: using RSA key 5D09FD0871C8F85B94CA8A0D281F0DB8D28D5469
# gpg: issuer "[email protected]"
# gpg: Good signature from "Michael S. Tsirkin <[email protected]>" [full]
# gpg: aka "Michael S. Tsirkin <[email protected]>" [full]
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17 0970 C350 3912 AFBE 8E67
# Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA 8A0D 281F 0DB8 D28D 5469
* remotes/mst/tags/for_upstream:
tests/bios-tables-test: add smbios cpu speed test
hw/smbios: add options for type 4 max-speed and current-speed
vhost-user-blk-pci: default num_queues to -smp N
virtio-blk-pci: default num_queues to -smp N
virtio-scsi-pci: default num_queues to -smp N
virtio-scsi: introduce a constant for fixed virtqueues
virtio-pci: add virtio_pci_optimal_num_queues() helper
Introduce a new flag for i440fx to disable PCI hotplug on the root bus
acpi: update expected DSDT files with _UID changes
disassemble-aml: -o actually works
arm/acpi: fix an out of spec _UID for PCI root
i386/acpi: fix inconsistent QEMU/OVMF device paths
acpi: allow DSDT changes
Ying Fang [Thu, 6 Aug 2020 03:56:34 +0000 (11:56 +0800)]
tests/bios-tables-test: add smbios cpu speed test
Add smbios type 4 CPU speed check for we added new options to set
smbios type 4 "max speed" and "current speed". The default value
should be 2000 when no option is specified, just as the old version
did.
We add the test case to one machine of each architecture, though it
doesn't really run on aarch64 platform for smbios test can't run on
uefi only platform yet.
Ying Fang [Thu, 6 Aug 2020 03:56:33 +0000 (11:56 +0800)]
hw/smbios: add options for type 4 max-speed and current-speed
Common VM users sometimes care about CPU speed, so we add two new
options to allow VM vendors to present CPU speed to their users.
Normally these information can be fetched from host smbios.
Strictly speaking, the "max speed" and "current speed" in type 4
are not really for the max speed and current speed of processor, for
"max speed" identifies a capability of the system, and "current speed"
identifies the processor's speed at boot (see smbios spec), but some
applications do not tell the differences.
Stefan Hajnoczi [Tue, 18 Aug 2020 14:33:48 +0000 (15:33 +0100)]
vhost-user-blk-pci: default num_queues to -smp N
Automatically size the number of request virtqueues to match the number
of vCPUs. This ensures that completion interrupts are handled on the
same vCPU that submitted the request. No IPI is necessary to complete
an I/O request and performance is improved. The maximum number of MSI-X
vectors and virtqueues limit are respected.
Stefan Hajnoczi [Tue, 18 Aug 2020 14:33:47 +0000 (15:33 +0100)]
virtio-blk-pci: default num_queues to -smp N
Automatically size the number of virtio-blk-pci request virtqueues to
match the number of vCPUs. Other transports continue to default to 1
request virtqueue.
A 1:1 virtqueue:vCPU mapping ensures that completion interrupts are
handled on the same vCPU that submitted the request. No IPI is
necessary to complete an I/O request and performance is improved. The
maximum number of MSI-X vectors and virtqueues limit are respected.
Performance improves from 78k to 104k IOPS on a 32 vCPU guest with 101
virtio-blk-pci devices (ioengine=libaio, iodepth=1, bs=4k, rw=randread
with NVMe storage).
Stefan Hajnoczi [Tue, 18 Aug 2020 14:33:46 +0000 (15:33 +0100)]
virtio-scsi-pci: default num_queues to -smp N
Automatically size the number of virtio-scsi-pci, vhost-scsi-pci, and
vhost-user-scsi-pci request virtqueues to match the number of vCPUs.
Other transports continue to default to 1 request virtqueue.
A 1:1 virtqueue:vCPU mapping ensures that completion interrupts are
handled on the same vCPU that submitted the request. No IPI is
necessary to complete an I/O request and performance is improved. The
maximum number of MSI-X vectors and virtqueues limit are respected.
Stefan Hajnoczi [Tue, 18 Aug 2020 14:33:45 +0000 (15:33 +0100)]
virtio-scsi: introduce a constant for fixed virtqueues
The event and control virtqueues are always present, regardless of the
multi-queue configuration. Define a constant so that virtqueue number
calculations are easier to read.
Multi-queue devices achieve the best performance when each vCPU has a
dedicated queue. This ensures that virtqueue used notifications are
handled on the same vCPU that submitted virtqueue buffers. When another
vCPU handles the the notification an IPI will be necessary to wake the
submission vCPU and this incurs a performance overhead.
Provide a helper function that virtio-pci devices will use in later
patches to automatically select the optimal number of queues.
The function handles guests with large numbers of CPUs by limiting the
number of queues to fit within the following constraints:
1. The maximum number of MSI-X vectors.
2. The maximum number of virtqueues.
Ani Sinha [Fri, 21 Aug 2020 16:54:03 +0000 (22:24 +0530)]
Introduce a new flag for i440fx to disable PCI hotplug on the root bus
We introduce a new global flag 'acpi-root-pci-hotplug' for i440fx with which
we can turn on or off PCI device hotplug on the root bus. This flag can be
used to prevent all PCI devices from getting hotplugged or unplugged from the
root PCI bus.
This feature is targetted mostly towards Windows VMs. It is useful in cases
where some hypervisor admins want to deploy guest VMs in a way so that the
users of the guest OSes are not able to hot-eject certain PCI devices from
the Windows system tray. Laine has explained the use case here in detail:
https://www.redhat.com/archives/libvir-list/2020-February/msg00110.html
Julia has resolved this issue for PCIE buses with the following commit: 530a0963184e57e71a5b538 ("pcie_root_port: Add hotplug disabling option")
This commit attempts to introduce similar behavior for PCI root buses used in
i440fx machine types (although in this case, we do not have a per-slot
capability to turn hotplug on or off).
Usage:
-global PIIX4_PM.acpi-root-pci-hotplug=off
By default, this option is enabled which means that hotplug is turned on for
the PCI root bus.
The previously existing flag 'acpi-pci-hotplug-with-bridge-support' for PCI-PCI
bridges remain as is and can be used along with this new flag to control PCI
hotplug on PCI bridges.
This change has been tested using a Windows 2012R2 server guest image and also
with a Windows 2019 server guest image on a Ubuntu 18.04 host using the latest
master qemu from upstream.
On ARM/virt machine type QEMU currently reports an incorrect _UID in
ACPI.
The particular node in question is the primary PciRoot (PCI0 in ACPI),
which gets assigned PCI0 in ACPI UID and 0 in the
DevicePath. This is due to the _UID assigned to it by build_dsdt in
hw/arm/virt-acpi-build.c Which does not correspond to the primary PCI
identifier given by pcibus_num in hw/pci/pci.c
In UEFI v2.8, section "10.4.2 Rules with ACPI _HID and _UID" ends with
the paragraph,
Root PCI bridges will use the plug and play ID of PNP0A03, This will
be stored in the ACPI Device Path _HID field, or in the Expanded
ACPI Device Path _CID field to match the ACPI name space. The _UID
in the ACPI Device Path structure must match the _UID in the ACPI
name space.
(See especially the last sentence.)
A similar bug has been reported on i386, on that architecture it has
been reported to confuse at least macOS which uses ACPI UIDs to build
the DevicePath for NVRAM boot options, while OVMF firmware gets them via
an internal channel through QEMU. When UEFI firmware and ACPI have
different values, this makes the underlying operating system unable to
report its boot option.
macOS uses ACPI UIDs to build the DevicePath for NVRAM boot options,
while OVMF firmware gets them via an internal channel through QEMU.
Due to a bug in QEMU ACPI currently UEFI firmware and ACPI have
different values, and this makes the underlying operating system
unable to report its boot option.
The particular node in question is the primary PciRoot (PCI0 in ACPI),
which for some reason gets assigned 1 in ACPI UID and 0 in the
DevicePath. This is due to the _UID assigned to it by build_dsdt in
hw/i386/acpi-build.c Which does not correspond to the primary PCI
identifier given by pcibus_num in hw/pci/pci.c
Reference with the device paths, OVMF startup logs, and ACPI table
dumps (SysReport):
https://github.com/acidanthera/bugtracker/issues/1050
In UEFI v2.8, section "10.4.2 Rules with ACPI _HID and _UID" ends with
the paragraph,
Root PCI bridges will use the plug and play ID of PNP0A03, This will
be stored in the ACPI Device Path _HID field, or in the Expanded
ACPI Device Path _CID field to match the ACPI name space. The _UID
in the ACPI Device Path structure must match the _UID in the ACPI
name space.
(See especially the last sentence.)
Considering *extra* root bridges / root buses (with bus number > 0),
QEMU's ACPI generator actually does the right thing; since QEMU commit c96d9286a6d7 ("i386/acpi-build: more traditional _UID and _HID for PXB
root buses", 2015-06-11).
However, the _UID values for root bridge zero (on both i440fx and q35)
have always been "wrong" (from UEFI perspective), going back in QEMU to
commit 74523b850189 ("i386: add ACPI table files from seabios",
2013-10-14).
Even in SeaBIOS, these _UID values have always been 1; see commit a4d357638c57 ("Port rombios32 code from bochs-bios.", 2008-03-08) for
i440fx, and commit ecbe3fd61511 ("seabios: q35: add dsdt", 2012-12-01)
for q35.
Peter Maydell [Wed, 26 Aug 2020 21:23:53 +0000 (22:23 +0100)]
Merge remote-tracking branch 'remotes/hdeller/tags/target-hppa-v3-pull-request' into staging
artist out of bounds fixes
# gpg: Signature made Wed 26 Aug 2020 22:09:55 BST
# gpg: using EDDSA key BCE9123E1AD29F07C049BBDEF712B510A23A0F5F
# gpg: Good signature from "Helge Deller <[email protected]>" [unknown]
# gpg: aka "Helge Deller <[email protected]>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 4544 8228 2CD9 10DB EF3D 25F8 3E5F 3D04 A7A2 4603
# Subkey fingerprint: BCE9 123E 1AD2 9F07 C049 BBDE F712 B510 A23A 0F5F
* remotes/hdeller/tags/target-hppa-v3-pull-request:
hw/display/artist: Fix invalidation of lines near screen border
hw/display/artist: Fix invalidation of lines in artist_draw_line()
hw/display/artist: Unbreak size mismatch memory accesses
hw/display/artist: Prevent out of VRAM buffer accesses
Revert "hw/display/artist: Avoid drawing line when nothing to display"
hw/display/artist: Refactor artist_rop8() to avoid buffer over-run
hw/display/artist: Check offset in draw_line to avoid buffer over-run
hw/hppa/lasi: Don't abort on invalid IMR value
hw/display/artist.c: fix out of bounds check
hw/hppa: Implement proper SeaBIOS version check
seabios-hppa: Update to SeaBIOS hppa version 1
hw/hppa: Sync hppa_hardware.h file with SeaBIOS sources
Sven Schnelle [Sat, 8 Aug 2020 18:51:57 +0000 (20:51 +0200)]
hw/display/artist: Fix invalidation of lines near screen border
If parts of the invalidated screen lines are outside of the VRAM buffer,
the code skips the whole invalidate. This is incorrect when only parts
of the buffer are invisble - which is the case when the mouse cursor is
located near the screen border.
Commit 5d971f9e6725 ("memory: Revert "memory: accept mismatching sizes
in memory_region_access_valid") broke the artist driver in a way that
the dtwm window manager on HP-UX rendered wrong.
Fixes: 5d971f9e6725 ("memory: Revert "memory: accept mismatching sizes in memory_region_access_valid") Signed-off-by: Sven Schnelle <[email protected]> Signed-off-by: Helge Deller <[email protected]>
Helge Deller [Sun, 9 Aug 2020 13:35:38 +0000 (15:35 +0200)]
hw/display/artist: Prevent out of VRAM buffer accesses
Simplify various bounds checks by changing parameters like row and column
numbers to become unsigned instead of signed.
With that we can check if the calculated offset is bigger than the size of the
VRAM region and bail out if not.
hw/display/artist: Refactor artist_rop8() to avoid buffer over-run
Invalid I/O writes can craft an offset out of the vram_buffer range.
Instead of passing an unsafe pointer to artist_rop8(), pass the vram_buffer and
the offset. We can now check if the offset is in range before accessing it.
We avoid:
Program terminated with signal SIGSEGV, Segmentation fault.
284 *dst &= ~plane_mask;
(gdb) bt
#0 0x000056367b2085c0 in artist_rop8 (s=0x56367d38b510, dst=0x7f9f972fffff <error: Cannot access memory at address 0x7f9f972fffff>, val=0 '\000') at hw/display/artist.c:284
#1 0x000056367b209325 in draw_line (s=0x56367d38b510, x1=-20480, y1=-1, x2=0, y2=17920, update_start=true, skip_pix=-1, max_pix=-1) at hw/display/artist.c:646
Helge Deller [Mon, 3 Aug 2020 05:41:46 +0000 (07:41 +0200)]
hw/hppa/lasi: Don't abort on invalid IMR value
NetBSD initializes the LASI IMR value with 0xffffffff to disable all LASI
interrupts. This triggered an assert() and stopped the emulation. By replacing
the check with a warning in the guest log we now allow NetBSD to boot again.
* remotes/maxreitz/tags/pull-block-2020-08-26: (34 commits)
iotests: Add tests for qcow2 images with extended L2 entries
qcow2: Assert that expand_zero_clusters_in_l1() does not support subclusters
qcow2: Allow preallocation and backing files if extended_l2 is set
qcow2: Add the 'extended_l2' option and the QCOW2_INCOMPAT_EXTL2 bit
qcow2: Add prealloc field to QCowL2Meta
qcow2: Add subcluster support to qcow2_measure()
qcow2: Add subcluster support to qcow2_co_pwrite_zeroes()
qcow2: Add subcluster support to handle_alloc_space()
qcow2: Clear the L2 bitmap when allocating a compressed cluster
qcow2: Update L2 bitmap in qcow2_alloc_cluster_link_l2()
qcow2: Add subcluster support to check_refcounts_l2()
qcow2: Add subcluster support to discard_in_l2_slice()
qcow2: Add subcluster support to zero_in_l2_slice()
qcow2: Add subcluster support to qcow2_get_host_offset()
qcow2: Add subcluster support to calculate_l2_meta()
qcow2: Handle QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC
qcow2: Replace QCOW2_CLUSTER_* with QCOW2_SUBCLUSTER_*
qcow2: Add cluster type parameter to qcow2_get_host_offset()
qcow2: Add qcow2_cluster_is_allocated()
qcow2: Add qcow2_get_subcluster_range_type()
...
Peter Maydell [Tue, 25 Aug 2020 21:50:42 +0000 (22:50 +0100)]
Merge remote-tracking branch 'remotes/alistair/tags/pull-riscv-to-apply-20200825' into staging
This pull request first adds support for multi-socket NUMA RISC-V
machines. The Spike and Virt machines both support NUMA sockets.
This PR also updates the current experimental Hypervisor support to the
v0.6.1 spec.
# gpg: Signature made Tue 25 Aug 2020 19:47:41 BST
# gpg: using RSA key F6C4AC46D4934868D3B8CE8F21E10D29DF977054
# gpg: Good signature from "Alistair Francis <[email protected]>" [full]
# Primary key fingerprint: F6C4 AC46 D493 4868 D3B8 CE8F 21E1 0D29 DF97 7054
* remotes/alistair/tags/pull-riscv-to-apply-20200825:
target/riscv: Support the Virtual Instruction fault
target/riscv: Return the exception from invalid CSR accesses
target/riscv: Support the v0.6 Hypervisor extension CRSs
target/riscv: Only support little endian guests
target/riscv: Only support a single VSXL length
target/riscv: Update the CSRs to the v0.6 Hyp extension
target/riscv: Update the Hypervisor trap return/entry
target/riscv: Fix the interrupt cause code
target/riscv: Convert MSTATUS MTL to GVA
target/riscv: Don't allow guest to write to htinst
target/riscv: Do two-stage lookups on hlv/hlvx/hsv instructions
target/riscv: Allow generating hlv/hlvx/hsv instructions
target/riscv: Allow setting a two-stage lookup in the virt status
hw/riscv: virt: Allow creating multiple NUMA sockets
hw/riscv: spike: Allow creating multiple NUMA sockets
hw/riscv: Add helpers for RISC-V multi-socket NUMA machines
hw/riscv: Allow creating multiple instances of PLIC
hw/riscv: Allow creating multiple instances of CLINT
Alistair Francis [Wed, 12 Aug 2020 19:13:46 +0000 (12:13 -0700)]
target/riscv: Return the exception from invalid CSR accesses
When performing a CSR access let's return a negative exception value on
an error instead of -1. This will allow us to specify the exception in
future patches.
Anup Patel [Fri, 15 May 2020 09:28:50 +0000 (14:58 +0530)]
hw/riscv: virt: Allow creating multiple NUMA sockets
We extend RISC-V virt machine to allow creating a multi-socket
machine. Each RISC-V virt machine socket is a NUMA node having
a set of HARTs, a memory instance, a CLINT instance, and a PLIC
instance. Other devices are shared between all sockets. We also
update the generated device tree accordingly.
By default, NUMA multi-socket support is disabled for RISC-V virt
machine. To enable it, users can use "-numa" command-line options
of QEMU.
Example1: For two NUMA nodes with 2 CPUs each, append following
to command-line options: "-smp 4 -numa node -numa node"
Example2: For two NUMA nodes with 1 and 3 CPUs, append following
to command-line options:
"-smp 4 -numa node -numa node -numa cpu,node-id=0,core-id=0 \
-numa cpu,node-id=1,core-id=1 -numa cpu,node-id=1,core-id=2 \
-numa cpu,node-id=1,core-id=3"
The maximum number of sockets in a RISC-V virt machine is 8
but this limit can be changed in future.
Anup Patel [Thu, 14 May 2020 11:09:43 +0000 (16:39 +0530)]
hw/riscv: spike: Allow creating multiple NUMA sockets
We extend RISC-V spike machine to allow creating a multi-socket
machine. Each RISC-V spike machine socket is a NUMA node having
a set of HARTs, a memory instance, and a CLINT instance. Other
devices are shared between all sockets. We also update the
generated device tree accordingly.
By default, NUMA multi-socket support is disabled for RISC-V spike
machine. To enable it, users can use "-numa" command-line options
of QEMU.
Example1: For two NUMA nodes with 2 CPUs each, append following
to command-line options: "-smp 4 -numa node -numa node"
Example2: For two NUMA nodes with 1 and 3 CPUs, append following
to command-line options:
"-smp 4 -numa node -numa node -numa cpu,node-id=0,core-id=0 \
-numa cpu,node-id=1,core-id=1 -numa cpu,node-id=1,core-id=2 \
-numa cpu,node-id=1,core-id=3"
The maximum number of sockets in a RISC-V spike machine is 8
but this limit can be changed in future.
Anup Patel [Fri, 29 May 2020 07:56:22 +0000 (13:26 +0530)]
hw/riscv: Add helpers for RISC-V multi-socket NUMA machines
We add common helper routines which can be shared by RISC-V
multi-socket NUMA machines.
We have two types of helpers:
1. riscv_socket_xyz() - These helper assist managing multiple
sockets irrespective whether QEMU NUMA is enabled/disabled
2. riscv_numa_xyz() - These helpers assist in providing
necessary QEMU machine callbacks for QEMU NUMA emulation
Anup Patel [Fri, 15 May 2020 04:55:33 +0000 (10:25 +0530)]
hw/riscv: Allow creating multiple instances of PLIC
We extend PLIC emulation to allow multiple instances of PLIC in
a QEMU RISC-V machine. To achieve this, we remove first HART id
zero assumption from PLIC emulation.
Anup Patel [Thu, 14 May 2020 10:21:31 +0000 (15:51 +0530)]
hw/riscv: Allow creating multiple instances of CLINT
We extend CLINT emulation to allow multiple instances of CLINT in
a QEMU RISC-V machine. To achieve this, we remove first HART id
zero assumption from CLINT emulation.
Direct leak of 2500 byte(s) in 1 object(s) allocated from:
#0 0x7f1b5a8b8d28 in __interceptor_calloc (/usr/lib/x86_64-linux-gnu/libasan.so.4+0xded28)
#1 0x7f1b5a514b10 in g_malloc0 (/usr/lib/x86_64-linux-gnu/libglib-2.0.so.0+0x51b10)
#2 0xd79ea4e4c0ad31c3 (<unknown module>)
SUMMARY: AddressSanitizer: 2500 byte(s) leaked in 1 allocation(s).
Call 'g_rand_free' in the end of function to avoid this.
Fixes: 4d3a329af59("tests/util-sockets: add abstract unix socket cases") Signed-off-by: Li Qiang <[email protected]> Reviewed-by: Philippe Mathieu-Daudé <[email protected]>
Reviewed-by: xiaoqiang zhao <[email protected]> Signed-off-by: Daniel P. Berrangé <[email protected]>
Alberto Garcia [Fri, 10 Jul 2020 16:13:15 +0000 (18:13 +0200)]
qcow2: Assert that expand_zero_clusters_in_l1() does not support subclusters
This function is only used by qcow2_expand_zero_clusters() to
downgrade a qcow2 image to a previous version. This would require
transforming all extended L2 entries into normal L2 entries but this
is not a simple task and there are no plans to implement this at the
moment.
Alberto Garcia [Fri, 10 Jul 2020 16:13:14 +0000 (18:13 +0200)]
qcow2: Allow preallocation and backing files if extended_l2 is set
Traditional qcow2 images don't allow preallocation if a backing file
is set. This is because once a cluster is allocated there is no way to
tell that its data should be read from the backing file.
Extended L2 entries have individual allocation bits for each
subcluster, and therefore it is perfectly possible to have an
allocated cluster with all its subclusters unallocated.
Alberto Garcia [Fri, 10 Jul 2020 16:13:13 +0000 (18:13 +0200)]
qcow2: Add the 'extended_l2' option and the QCOW2_INCOMPAT_EXTL2 bit
Now that the implementation of subclusters is complete we can finally
add the necessary options to create and read images with this feature,
which we call "extended L2 entries".
Alberto Garcia [Fri, 10 Jul 2020 16:13:12 +0000 (18:13 +0200)]
qcow2: Add prealloc field to QCowL2Meta
This field allows us to indicate that the L2 metadata update does not
come from a write request with actual data but from a preallocation
request.
For traditional images this does not make any difference, but for
images with extended L2 entries this means that the clusters are
allocated normally in the L2 table but individual subclusters are
marked as unallocated.
This will allow preallocating images that have a backing file.
There is one special case: when we resize an existing image we can
also request that the new clusters are preallocated. If the image
already had a backing file then we have to hide any possible stale
data and zero out the new clusters (see commit 955c7d6687 for more
details).
In this case the subclusters cannot be left as unallocated so the L2
bitmap must be updated.
Alberto Garcia [Fri, 10 Jul 2020 16:13:10 +0000 (18:13 +0200)]
qcow2: Add subcluster support to qcow2_co_pwrite_zeroes()
This works now at the subcluster level and pwrite_zeroes_alignment is
updated accordingly.
qcow2_cluster_zeroize() is turned into qcow2_subcluster_zeroize() with
the following changes:
- The request can now be subcluster-aligned.
- The cluster-aligned body of the request is still zeroized using
zero_in_l2_slice() as before.
- The subcluster-aligned head and tail of the request are zeroized
with the new zero_l2_subclusters() function.
There is just one thing to take into account for a possible future
improvement: compressed clusters cannot be partially zeroized so
zero_l2_subclusters() on the head or the tail can return -ENOTSUP.
This makes the caller repeat the *complete* request and write actual
zeroes to disk. This is sub-optimal because
1) if the head area was compressed we would still be able to use
the fast path for the body and possibly the tail.
2) if the tail area was compressed we are writing zeroes to the
head and the body areas, which are already zeroized.
Alberto Garcia [Fri, 10 Jul 2020 16:13:09 +0000 (18:13 +0200)]
qcow2: Add subcluster support to handle_alloc_space()
The bdrv_co_pwrite_zeroes() call here fills complete clusters with
zeroes, but it can happen that some subclusters are not part of the
write request or the copy-on-write. This patch makes sure that only
the affected subclusters are overwritten.
A potential improvement would be to also fill with zeroes the other
subclusters if we can guarantee that we are not overwriting existing
data. However this would waste more disk space, so we should first
evaluate if it's really worth doing.
Alberto Garcia [Fri, 10 Jul 2020 16:13:07 +0000 (18:13 +0200)]
qcow2: Update L2 bitmap in qcow2_alloc_cluster_link_l2()
The L2 bitmap needs to be updated after each write to indicate what
new subclusters are now allocated. This needs to happen even if the
cluster was already allocated and the L2 entry was otherwise valid.
In some cases however a write operation doesn't need change the L2
bitmap (because all affected subclusters were already allocated). This
is detected in calculate_l2_meta(), and qcow2_alloc_cluster_link_l2()
is never called in those cases.
Alberto Garcia [Fri, 10 Jul 2020 16:13:06 +0000 (18:13 +0200)]
qcow2: Add subcluster support to check_refcounts_l2()
The offset field of an uncompressed cluster's L2 entry must be aligned
to the cluster size, otherwise it is invalid. If the cluster has no
data then it means that the offset points to a preallocation, so we
can clear the offset field without affecting the guest-visible data.
This is what 'qemu-img check' does when run in repair mode.
On traditional qcow2 images this can only happen when QCOW_OFLAG_ZERO
is set, and repairing such entries turns the clusters from ZERO_ALLOC
into ZERO_PLAIN.
Extended L2 entries have no ZERO_ALLOC clusters and no QCOW_OFLAG_ZERO
but the idea is the same: if none of the subclusters are allocated
then we can clear the offset field and leave the bitmap untouched.
Alberto Garcia [Fri, 10 Jul 2020 16:13:05 +0000 (18:13 +0200)]
qcow2: Add subcluster support to discard_in_l2_slice()
Two things need to be taken into account here:
1) With full_discard == true the L2 entry must be cleared completely.
This also includes the L2 bitmap if the image has extended L2
entries.
2) With full_discard == false we have to make the discarded cluster
read back as zeroes. With normal L2 entries this is done with the
QCOW_OFLAG_ZERO bit, whereas with extended L2 entries this is done
with the individual 'all zeroes' bits for each subcluster.
Note however that QCOW_OFLAG_ZERO is not supported in v2 qcow2
images so, if there is a backing file, discard cannot guarantee
that the image will read back as zeroes. If this is important for
the caller it should forbid it as qcow2_co_pdiscard() does (see 80f5c01183 for more details).
Alberto Garcia [Fri, 10 Jul 2020 16:13:04 +0000 (18:13 +0200)]
qcow2: Add subcluster support to zero_in_l2_slice()
The QCOW_OFLAG_ZERO bit that indicates that a cluster reads as
zeroes is only used in standard L2 entries. Extended L2 entries use
individual 'all zeroes' bits for each subcluster.
This must be taken into account when updating the L2 entry and also
when deciding that an existing entry does not need to be updated.
Alberto Garcia [Fri, 10 Jul 2020 16:13:03 +0000 (18:13 +0200)]
qcow2: Add subcluster support to qcow2_get_host_offset()
The logic of this function remains pretty much the same, except that
it uses count_contiguous_subclusters(), which combines the logic of
count_contiguous_clusters() / count_contiguous_clusters_unallocated()
and checks individual subclusters.
qcow2_cluster_to_subcluster_type() is not necessary as a separate
function anymore so it's inlined into its caller.
Alberto Garcia [Fri, 10 Jul 2020 16:13:02 +0000 (18:13 +0200)]
qcow2: Add subcluster support to calculate_l2_meta()
If an image has subclusters then there are more copy-on-write
scenarios that we need to consider. Let's say we have a write request
from the middle of subcluster #3 until the end of the cluster:
1) If we are writing to a newly allocated cluster then we need
copy-on-write. The previous contents of subclusters #0 to #3 must
be copied to the new cluster. We can optimize this process by
skipping all leading unallocated or zero subclusters (the status of
those skipped subclusters will be reflected in the new L2 bitmap).
2) If we are overwriting an existing cluster:
2.1) If subcluster #3 is unallocated or has the all-zeroes bit set
then we need copy-on-write (on subcluster #3 only).
2.2) If subcluster #3 was already allocated then there is no need
for any copy-on-write. However we still need to update the L2
bitmap to reflect possible changes in the allocation status of
subclusters #4 to #31. Because of this, this function checks
if all the overwritten subclusters are already allocated and
in this case it returns without creating a new QCowL2Meta
structure.
After all these changes l2meta_cow_start() and l2meta_cow_end()
are not necessarily cluster-aligned anymore. We need to update the
calculation of old_start and old_end in handle_dependencies() to
guarantee that no two requests try to write on the same cluster.
Alberto Garcia [Fri, 10 Jul 2020 16:13:00 +0000 (18:13 +0200)]
qcow2: Replace QCOW2_CLUSTER_* with QCOW2_SUBCLUSTER_*
In order to support extended L2 entries some functions of the qcow2
driver need to start dealing with subclusters instead of clusters.
qcow2_get_host_offset() is modified to return the subcluster type
instead of the cluster type, and all callers are updated to replace
all values of QCow2ClusterType with their QCow2SubclusterType
equivalents.
This patch only changes the data types, there are no semantic changes.
Alberto Garcia [Fri, 10 Jul 2020 16:12:59 +0000 (18:12 +0200)]
qcow2: Add cluster type parameter to qcow2_get_host_offset()
This function returns an integer that can be either an error code or a
cluster type (a value from the QCow2ClusterType enum).
We are going to start using subcluster types instead of cluster types
in some functions so it's better to use the exact data types instead
of integers for clarity and in order to detect errors more easily.
This patch makes qcow2_get_host_offset() return 0 on success and
puts the returned cluster type in a separate parameter. There are no
semantic changes.
Alberto Garcia [Fri, 10 Jul 2020 16:12:57 +0000 (18:12 +0200)]
qcow2: Add qcow2_get_subcluster_range_type()
There are situations in which we want to know how many contiguous
subclusters of the same type there are in a given cluster. This can be
done by simply iterating over the subclusters and repeatedly calling
qcow2_get_subcluster_type() for each one of them.
However once we determined the type of a subcluster we can check the
rest efficiently by counting the number of adjacent ones (or zeroes)
in the bitmap. This is what this function does.
Alberto Garcia [Fri, 10 Jul 2020 16:12:56 +0000 (18:12 +0200)]
qcow2: Add QCow2SubclusterType and qcow2_get_subcluster_type()
This patch adds QCow2SubclusterType, which is the subcluster-level
version of QCow2ClusterType. All QCOW2_SUBCLUSTER_* values have the
the same meaning as their QCOW2_CLUSTER_* equivalents (when they
exist). See below for details and caveats.
In images without extended L2 entries clusters are treated as having
exactly one subcluster so it is possible to replace one data type with
the other while keeping the exact same semantics.
With extended L2 entries there are new possible values, and every
subcluster in the same cluster can obviously have a different
QCow2SubclusterType so functions need to be adapted to work on the
subcluster level.
There are several things that have to be taken into account:
a) QCOW2_SUBCLUSTER_COMPRESSED means that the whole cluster is
compressed. We do not support compression at the subcluster
level.
b) There are two different values for unallocated subclusters:
QCOW2_SUBCLUSTER_UNALLOCATED_PLAIN which means that the whole
cluster is unallocated, and QCOW2_SUBCLUSTER_UNALLOCATED_ALLOC
which means that the cluster is allocated but the subcluster is
not. The latter can only happen in images with extended L2
entries.
c) QCOW2_SUBCLUSTER_INVALID is used to detect the cases where an L2
entry has a value that violates the specification. The caller is
responsible for handling these situations.
To prevent compatibility problems with images that have invalid
values but are currently being read by QEMU without causing side
effects, QCOW2_SUBCLUSTER_INVALID is only returned for images
with extended L2 entries.
qcow2_cluster_to_subcluster_type() is added as a separate function
from qcow2_get_subcluster_type(), but this is only temporary and both
will be merged in a subsequent patch.
Alberto Garcia [Fri, 10 Jul 2020 16:12:55 +0000 (18:12 +0200)]
qcow2: Update get/set_l2_entry() and add get/set_l2_bitmap()
Extended L2 entries are 128-bit wide: 64 bits for the entry itself and
64 bits for the subcluster allocation bitmap.
In order to support them correctly get/set_l2_entry() need to be
updated so they take the entry width into account in order to
calculate the correct offset.
This patch also adds the get/set_l2_bitmap() functions that are
used to access the bitmaps. For convenience we allow calling
get_l2_bitmap() on images without subclusters. In this case the
returned value is always 0 and has no meaning.
Alberto Garcia [Fri, 10 Jul 2020 16:12:54 +0000 (18:12 +0200)]
qcow2: Add l2_entry_size()
qcow2 images with subclusters have 128-bit L2 entries. The first 64
bits contain the same information as traditional images and the last
64 bits form a bitmap with the status of each individual subcluster.
Because of that we cannot assume that L2 entries are sizeof(uint64_t)
anymore. This function returns the proper value for the image.
Alberto Garcia [Fri, 10 Jul 2020 16:12:51 +0000 (18:12 +0200)]
qcow2: Add subcluster-related fields to BDRVQcow2State
This patch adds the following new fields to BDRVQcow2State:
- subclusters_per_cluster: Number of subclusters in a cluster
- subcluster_size: The size of each subcluster, in bytes
- subcluster_bits: No. of bits so 1 << subcluster_bits = subcluster_size
Images without subclusters are treated as if they had exactly one
subcluster per cluster (i.e. subcluster_size = cluster_size).
Alberto Garcia [Fri, 10 Jul 2020 16:12:50 +0000 (18:12 +0200)]
qcow2: Add dummy has_subclusters() function
This function will be used by the qcow2 code to check if an image has
subclusters or not.
At the moment this simply returns false. Once all patches needed for
subcluster support are ready then QEMU will be able to create and
read images with subclusters and this function will return the actual
value.
Alberto Garcia [Fri, 10 Jul 2020 16:12:49 +0000 (18:12 +0200)]
qcow2: Document the Extended L2 Entries feature
Subcluster allocation in qcow2 is implemented by extending the
existing L2 table entries and adding additional information to
indicate the allocation status of each subcluster.
This patch documents the changes to the qcow2 format and how they
affect the calculation of the L2 cache size.
Alberto Garcia [Fri, 10 Jul 2020 16:12:48 +0000 (18:12 +0200)]
qcow2: Add get_l2_entry() and set_l2_entry()
The size of an L2 entry is 64 bits, but if we want to have subclusters
we need extended L2 entries. This means that we have to access L2
tables and slices differently depending on whether an image has
extended L2 entries or not.
This patch replaces all l2_slice[] accesses with calls to
get_l2_entry() and set_l2_entry().
Alberto Garcia [Fri, 10 Jul 2020 16:12:47 +0000 (18:12 +0200)]
qcow2: Process QCOW2_CLUSTER_ZERO_ALLOC clusters in handle_copied()
When writing to a qcow2 file there are two functions that take a
virtual offset and return a host offset, possibly allocating new
clusters if necessary:
- handle_copied() looks for normal data clusters that are already
allocated and have a reference count of 1. In those clusters we
can simply write the data and there is no need to perform any
copy-on-write.
- handle_alloc() looks for clusters that do need copy-on-write,
either because they haven't been allocated yet, because their
reference count is != 1 or because they are ZERO_ALLOC clusters.
The ZERO_ALLOC case is a bit special because those are clusters that
are already allocated and they could perfectly be dealt with in
handle_copied() (as long as copy-on-write is performed when required).
In fact, there is extra code specifically for them in handle_alloc()
that tries to reuse the existing allocation if possible and frees them
otherwise.
This patch changes the handling of ZERO_ALLOC clusters so the
semantics of these two functions are now like this:
- handle_copied() looks for clusters that are already allocated and
which we can overwrite (NORMAL and ZERO_ALLOC clusters with a
reference count of 1).
- handle_alloc() looks for clusters for which we need a new
allocation (all other cases).
One important difference after this change is that clusters found
in handle_copied() may now require copy-on-write, but this will be
necessary anyway once we add support for subclusters.
Alberto Garcia [Fri, 10 Jul 2020 16:12:44 +0000 (18:12 +0200)]
qcow2: Convert qcow2_get_cluster_offset() into qcow2_get_host_offset()
qcow2_get_cluster_offset() takes an (unaligned) guest offset and
returns the (aligned) offset of the corresponding cluster in the qcow2
image.
In practice none of the callers need to know where the cluster starts
so this patch makes the function calculate and return the final host
offset directly. The function is also renamed accordingly.
There is a pre-existing exception with compressed clusters: in this
case the function returns the complete cluster descriptor (containing
the offset and size of the compressed data). This does not change with
this patch but it is now documented.
Alberto Garcia [Fri, 10 Jul 2020 16:12:43 +0000 (18:12 +0200)]
qcow2: Make Qcow2AioTask store the full host offset
The file_cluster_offset field of Qcow2AioTask stores a cluster-aligned
host offset. In practice this is not very useful because all users(*)
of this structure need the final host offset into the cluster, which
they calculate using
There is no reason why Qcow2AioTask cannot store host_offset directly
and that is what this patch does.
(*) compressed clusters are the exception: in this case what
file_cluster_offset was storing was the full compressed cluster
descriptor (offset + size). This does not change with this patch
but it is documented now.
Gerd Hoffmann [Mon, 24 Aug 2020 07:40:57 +0000 (09:40 +0200)]
meson: drop keymaps symlink
We are building the keymaps by default now. Drop the keymaps symlink
so the generated files are actually written to the build tree not the
source tree.
Peter Maydell [Mon, 24 Aug 2020 15:39:52 +0000 (16:39 +0100)]
Merge remote-tracking branch 'remotes/cschoenebeck/tags/pull-9p-20200812' into staging
9pfs: Fix severe performance issue of Treaddir requests.
# gpg: Signature made Wed 12 Aug 2020 11:06:21 BST
# gpg: using RSA key 96D8D110CF7AF8084F88590134C2B58765A47395
# gpg: issuer "[email protected]"
# gpg: Good signature from "Christian Schoenebeck <[email protected]>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: ECAB 1A45 4014 1413 BA38 4926 30DB 47C3 A012 D5F4
# Subkey fingerprint: 96D8 D110 CF7A F808 4F88 5901 34C2 B587 65A4 7395
* remotes/cschoenebeck/tags/pull-9p-20200812:
9pfs: clarify latency of v9fs_co_run_in_worker()
9pfs: differentiate readdir lock between 9P2000.u vs. 9P2000.L
9pfs: T_readdir latency optimization
9pfs: add new function v9fs_co_readdir_many()
9pfs: split out fs driver core of v9fs_co_readdir()
9pfs: make v9fs_readdir_response_size() public
tests/virtio-9p: added split readdir tests
Peter Maydell [Mon, 24 Aug 2020 11:13:09 +0000 (12:13 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20200824' into staging
target-arm queue:
* hw/cpu/a9mpcore: Verify the machine use Cortex-A9 cores
* hw/arm/smmuv3: Implement SMMUv3.2 range-invalidation
* docs/system/arm: Document the Xilinx Versal Virt board
* target/arm: Make M-profile NOCP take precedence over UNDEF
* target/arm: Use correct FPST for VCMLA, VCADD on fp16
* target/arm: Various cleanups preparing for fp16 support
* remotes/pmaydell/tags/pull-target-arm-20200824: (27 commits)
target/arm: Use correct FPST for VCMLA, VCADD on fp16
target/arm: Implement FPST_STD_F16 fpstatus
target/arm: Make A32/T32 use new fpstatus_ptr() API
target/arm: Replace A64 get_fpstatus_ptr() with generic fpstatus_ptr()
target/arm: Delete unused ARM_FEATURE_CRC
target/arm/translate.c: Delete/amend incorrect comments
target/arm: Delete unused VFP_DREG macros
target/arm: Remove ARCH macro
target/arm: Convert T32 coprocessor insns to decodetree
target/arm: Do M-profile NOCP checks early and via decodetree
target/arm: Tidy up disas_arm_insn()
target/arm: Convert A32 coprocessor insns to decodetree
target/arm: Separate decode from handling of coproc insns
target/arm: Pull handling of XScale insns out of disas_coproc_insn()
docs/system/arm: Document the Xilinx Versal Virt board
hw/arm/smmuv3: Advertise SMMUv3.2 range invalidation
hw/arm/smmuv3: Support HAD and advertise SMMUv3.1 support
hw/arm/smmuv3: Let AIDR advertise SMMUv3.0 support
hw/arm/smmuv3: Fix IIDR offset
hw/arm/smmuv3: Get prepared for range invalidation
...
Peter Maydell [Thu, 6 Aug 2020 10:44:53 +0000 (11:44 +0100)]
target/arm: Use correct FPST for VCMLA, VCADD on fp16
When we implemented the VCMLA and VCADD insns we put in the
code to handle fp16, but left it using the standard fp status
flags. Correct them to use FPST_STD_F16 for fp16 operations.
Peter Maydell [Thu, 6 Aug 2020 10:44:52 +0000 (11:44 +0100)]
target/arm: Implement FPST_STD_F16 fpstatus
Architecturally, Neon FP16 operations use the "standard FPSCR" like
all other Neon operations. However, this is defined in the Arm ARM
pseudocode as "a fixed value, except that FZ16 (and AHP) follow the
FPSCR bits". In QEMU, the softfloat float_status doesn't include
separate flush-to-zero for FP16 operations, so we must keep separate
fp_status for "Neon non-FP16" and "Neon fp16" operations, in the
same way we do already for the non-Neon "fp_status" vs "fp_status_f16".
Add the extra float_status field to the CPU state structure,
ensure it is correctly initialized and updated on FPSCR writes,
and make fpstatus_ptr(FPST_STD_F16) return a pointer to it.
Peter Maydell [Thu, 6 Aug 2020 10:44:50 +0000 (11:44 +0100)]
target/arm: Replace A64 get_fpstatus_ptr() with generic fpstatus_ptr()
We currently have two versions of get_fpstatus_ptr(), which both take
an effectively boolean argument:
* the one for A64 takes "bool is_f16" to distinguish fp16 from other ops
* the one for A32/T32 takes "int neon" to distinguish Neon from other ops
This is confusing, and to implement ARMv8.2-FP16 the A32/T32 one will
need to make a four-way distinction between "non-Neon, FP16",
"non-Neon, single/double", "Neon, FP16" and "Neon, single/double".
The A64 version will then be a strict subset of the A32/T32 version.
To clean this all up, we want to go to a single implementation which
takes an enum argument with values FPST_FPCR, FPST_STD,
FPST_FPCR_F16, and FPST_STD_F16. We rename the function to
fpstatus_ptr() so that unconverted code gets a compilation error
rather than silently passing the wrong thing to the new function.
This commit implements that new API, and converts A64 to use it:
get_fpstatus_ptr(false) -> fpstatus_ptr(FPST_FPCR)
get_fpstatus_ptr(true) -> fpstatus_ptr(FPST_FPCR_F16)
Peter Maydell [Wed, 5 Aug 2020 21:08:48 +0000 (22:08 +0100)]
target/arm: Delete unused ARM_FEATURE_CRC
In commit 962fcbf2efe57231a9f5df we converted the uses of the
ARM_FEATURE_CRC bit to use the aa32_crc32 isar_feature test
instead. However we forgot to remove the now-unused definition
of the feature name in the enum. Delete it now.
In arm_tr_init_disas_context() we have a FIXME comment that suggests
"cpu_M0 can probably be the same as cpu_V0". This isn't in fact
possible: cpu_V0 is used as a temporary inside gen_iwmmxt_shift(),
and that function is called in various places where cpu_M0 contains a
live value (i.e. between gen_op_iwmmxt_movq_M0_wRn() and
gen_op_iwmmxt_movq_wRn_M0() calls). Remove the comment.
We also have a comment on the declarations of cpu_V0/V1/M0 which
claims they're "for efficiency". This isn't true with modern TCG, so
replace this comment with one which notes that they're only used with
the iwmmxt decode.
Peter Maydell [Mon, 3 Aug 2020 11:18:49 +0000 (12:18 +0100)]
target/arm: Remove ARCH macro
The ARCH() macro was used a lot in the legacy decoder, but
there are now just two uses of it left. Since a macro which
expands out to a goto is liable to be confusing when reading
code, replace the last two uses with a simple open-coded
qeuivalent.
projects / qemu.git / log