Merge tag 'pull-riscv-to-apply-20241220' of https://github.com/alistair23/qemu into staging

RISC-V PR for 10.0

* Correct the validness check of iova
* Fix APLIC in_clrip and clripnum write emulation
* Support riscv-iommu-sys device
* Add Tenstorrent Ascalon CPU
* Add AIA userspace irqchip_split support
* Add Microblaze V generic board
* Upgrade ACPI SPCR table to support SPCR table revision 4 format
* Remove tswap64() calls from HTIF
* Support 64-bit address of initrd
* Introduce svukte ISA extension
* Support ssstateen extension
* Support for RV64 Xiangshan Nanhu CPU

 # -----BEGIN PGP SIGNATURE-----
 #
 # iQIzBAABCAAdFiEEaukCtqfKh31tZZKWr3yVEwxTgBMFAmdkzjgACgkQr3yVEwxT
 # gBOcyA//e0XhAQciQglCZZCfINdOyI8qSh+P2K0qtrXZ4VERHEMp7UoD5CQr2cZv
 # h8ij1EkatXCwukVELx0rNckxG33bEFgG1oESnQSrwGE0Iu4csNW24nK5WlUS0/r+
 # A5oD2wtzEF+cbhTKrVSDBN/PvlnWTKGEoJRkuXWfz5d4uR9eyQhfED0S2j36lNEC
 # X1x/OZoKM89XuXtOFe9g55Z5UNzAatcdTISozL0FydiPh7QeVjTLHh28/tt559MX
 # 7v5aJFlQuZ78z1mIHkZmPSorSrJ0zqhkP6NWe1ae06oMgzwRQQhYLppDILV4ZgUF
 # 3mSDRoXmBycQXiYNPcHep3LdXfvxr+PpWHSevx8gH1jwm93On7Y/H7Uol6TDXzfC
 # mrFjalfV5tzrD90ZvB+s5bCMF1q5Z8Dlj0pYF9aN9P1ILoWy3dndFAPJB6uKKDP7
 # Qd4qOQ3dVyHAX9jLmVkB6QvAV/vTDrYTsAxaF/EaoLOy0IoKhjTvgda3XzE1MFKA
 # gVafLluADIfSEdqa2QR2ExL8d1SZVoiObCp5TMLRer0HIpg/vQZwjfdbo4BgQKL3
 # 7Q6wBxcZUNqrFgspXjm5WFIrdk2rfS/79OmvpNM6SZaK6BnklntdJHJHtAWujGsm
 # EM310AUFpHMp2h6Nqnemb3qr5l4d20KSt8DhoPAUq1IE59Kb8XY=
 # =0iQW
 # -----END PGP SIGNATURE-----
 # gpg: Signature made Thu 19 Dec 2024 20:54:00 EST
 # gpg:                using RSA key 6AE902B6A7CA877D6D659296AF7C95130C538013
 # gpg: Good signature from "Alistair Francis <alistair@alistair23.me>" [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: 6AE9 02B6 A7CA 877D 6D65  9296 AF7C 9513 0C53 8013

* tag 'pull-riscv-to-apply-20241220' of https://github.com/alistair23/qemu: (39 commits)
  target/riscv: add support for RV64 Xiangshan Nanhu CPU
  target/riscv: add ssstateen
  target/riscv/tcg: hide warn for named feats when disabling via priv_ver
  target/riscv: Include missing headers in 'internals.h'
  target/riscv: Include missing headers in 'vector_internals.h'
  target/riscv: Check svukte is not enabled in RV32
  target/riscv: Expose svukte ISA extension
  target/riscv: Check memory access to meet svukte rule
  target/riscv: Support hstatus[HUKTE] bit when svukte extension is enabled
  target/riscv: Support senvcfg[UKTE] bit when svukte extension is enabled
  target/riscv: Add svukte extension capability variable
  hw/riscv: Add the checking if DTB overlaps to kernel or initrd
  hw/riscv: Add a new struct RISCVBootInfo
  hw/riscv: Support to load DTB after 3GB memory on 64-bit system.
  hw/char/riscv_htif: Clarify MemoryRegionOps expect 32-bit accesses
  hw/char/riscv_htif: Explicit little-endian implementation
  MAINTAINERS: Cover RISC-V HTIF interface
  tests/qtest/bios-tables-test: Update virt SPCR golden reference for RISC-V
  hw/acpi: Upgrade ACPI SPCR table to support SPCR table revision 4 format
  qtest: allow SPCR acpi table changes
  ...

Conflicts:
  target/riscv/cpu.c

  Merge conflict with DEFINE_PROP_END_OF_LIST() removal. No Property
  array terminator is needed anymore.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

Merge tag 'for-upstream' of https://gitlab.com/bonzini/qemu into staging

* qdev: second part of Property cleanups
* rust: second part of QOM rework
* rust: callbacks wrapper
* rust: pl011 bugfixes
* kvm: cleanup errors in kvm_convert_memory()

 # -----BEGIN PGP SIGNATURE-----
 #
 # iQFIBAABCAAyFiEE8TM4V0tmI4mGbHaCv/vSX3jHroMFAmdkaEkUHHBib256aW5p
 # QHJlZGhhdC5jb20ACgkQv/vSX3jHroN0/wgAgIJg8BrlRKfmiz14NZfph8/jarSj
 # TOWYVxL2v4q98KBuL5pta2ucObgzwqyqSyc02S2DGSOIMQCIiBB5MaCk1iMjx+BO
 # pmVU8gNlD8faO8SSmnnr+jDQt+G+bQ/nRgQJOAReF8oVw3O2aC/FaVKpitMzWtvv
 # PLnJWdrqqpGq14OzX8iNCzSujxppAuyjrhT4lNlekzDoDfdTez72r+rXkvg4GzZL
 # QC3xLYg/LrT8Rs+zgOhm/AaIyS4bOyMlkU9Du1rQ6Tyne45ey2FCwKVzBKrJdGcw
 # sVbzEclxseLenoTbZqYK6JTzLdDoThVUbY2JwoCGUaIm+74P4NjEsUsTVg==
 # =TuQM
 # -----END PGP SIGNATURE-----
 # gpg: Signature made Thu 19 Dec 2024 13:39:05 EST
 # gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
 # gpg:                issuer "pbonzini@redhat.com"
 # gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [full]
 # gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>" [full]
 # Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
 #      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* tag 'for-upstream' of https://gitlab.com/bonzini/qemu: (42 commits)
  rust: pl011: simplify handling of the FIFO enabled bit in LCR
  rust: pl011: fix migration stream
  rust: pl011: extend registers to 32 bits
  rust: pl011: fix break errors and definition of Data struct
  rust: pl011: always use reset() method on registers
  rust: pl011: match break logic of C version
  rust: pl011: fix declaration of LineControl bits
  target/i386: Reset TSCs of parked vCPUs too on VM reset
  kvm: consistently return 0/-errno from kvm_convert_memory
  rust: qemu-api: add a module to wrap functions and zero-sized closures
  rust: qom: add initial subset of methods on Object
  rust: qom: add casting functionality
  rust: tests: allow writing more than one test
  bql: add a "mock" BQL for Rust unit tests
  rust: re-export C types from qemu-api submodules
  rust: rename qemu-api modules to follow C code a bit more
  rust: qom: add possibility of overriding unparent
  rust: qom: put class_init together from multiple ClassInitImpl<>
  Constify all opaque Property pointers
  hw/core/qdev-properties: Constify Property argument to PropertyInfo.print
  ...

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

target/riscv: add support for RV64 Xiangshan Nanhu CPU

Add a CPU entry for the RV64 XiangShan NANHU CPU which
supports single-core and dual-core configurations. More
details can be found at
https://docs.xiangshan.cc/zh-cn/latest/integration/overview

Signed-off-by: MollyChen <xiaoou@iscas.ac.cn>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241205073622.46052-1-xiaoou@iscas.ac.cn>
[ Changes by AF
 - Fixup code formatting
]
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: add ssstateen

ssstateen is defined in RVA22 as:

"Supervisor-mode view of the state-enable extension. The supervisor-mode
(sstateen0-3) and hypervisor-mode (hstateen0-3) state-enable registers
must be provided."

Add ssstateen as a named feature that is available if we also have
smstateen.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Message-ID: <20241113171755.978109-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/tcg: hide warn for named feats when disabling via priv_ver

Commit 68c9e54bea handled a situation where a warning was being shown
when using the 'sifive_e' cpu when disabling the named extension zic64b.
It makes little sense to show user warnings for named extensions that
users can't control, and the solution taken was to disable zic64b
manually in riscv_cpu_update_named_features().

This solution won't scale well when adding more named features, and can
eventually end up repeating riscv_cpu_disable_priv_spec_isa_exts().

Change riscv_cpu_disable_priv_spec_isa_exts() to not show warnings when
disabling a named feature. This will accomplish the same thing we're
doing today while avoiding having two points where we're disabling
exts via priv_ver mismatch.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241113171755.978109-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Include missing headers in 'internals.h'

Rather than relying on implicit includes, explicit them,
in order to avoid when refactoring unrelated headers:

  target/riscv/internals.h:49:15: error: use of undeclared identifier 'PRV_S'
     49 |         ret = PRV_S;
        |               ^
  target/riscv/internals.h:93:9: error: call to undeclared function 'env_archcpu'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
     93 |     if (env_archcpu(env)->cfg.ext_zfinx) {
        |         ^
  target/riscv/internals.h:101:15: error: unknown type name 'float32'; did you mean 'float'?
    101 | static inline float32 check_nanbox_s(CPURISCVState *env, uint64_t f)
        |               ^~~~~~~
        |               float

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241203200828.47311-3-philmd@linaro.org>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Include missing headers in 'vector_internals.h'

Rather than relying on implicit includes, explicit them,
in order to avoid when refactoring unrelated headers:

  target/riscv/vector_internals.h:36:12: error: call to undeclared function 'FIELD_EX32'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
     36 |     return FIELD_EX32(simd_data(desc), VDATA, NF);
        |            ^

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241203200828.47311-2-philmd@linaro.org>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Check svukte is not enabled in RV32

The spec explicitly says svukte doesn't support RV32. So check that it
is not enabled in RV32.

Signed-off-by: Fea.Wang <fea.wang@sifive.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241203034932.25185-7-fea.wang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Expose svukte ISA extension

Add "svukte" in the ISA string when svukte extension is enabled.

Signed-off-by: Fea.Wang <fea.wang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241203034932.25185-6-fea.wang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Check memory access to meet svukte rule

Follow the Svukte spec, do the memory access address checking

1. Include instruction fetches or explicit memory accesses
2. System run in effective privilege U or VU
3. Check senvcfg[UKTE] being set, or hstatus[HUKTE] being set if
instruction is HLV, HLVX, HSV and execute from U mode to VU mode
4. Depend on Sv39 and check virtual addresses bit[SXLEN-1]
5. Raises a page-fault exception corresponding to the original access
type.

Ref: https://github.com/riscv/riscv-isa-manual/pull/1564/files

Signed-off-by: Frank Chang <frank.chang@sifive.com>
Signed-off-by: Fea.Wang <fea.wang@sifive.com>
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241203034932.25185-5-fea.wang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Support hstatus[HUKTE] bit when svukte extension is enabled

Svukte extension add HUKTE bit, bit[24] in hstatus CSR. The written
value will be masked when the svukte extension is not enabled.

When hstatus[HUKTE] bit is set, HLV/HLVX/HSV work in the U-mode should
do svukte check.

Signed-off-by: Fea.Wang <fea.wang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241203034932.25185-4-fea.wang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Support senvcfg[UKTE] bit when svukte extension is enabled

Svukte extension add UKTE bit, bit[8] in senvcfg CSR. The bit will be
supported when the svukte extension is enabled.

When senvcfg[UKTE] bit is set, the memory access from U-mode should do
the svukte check only except HLV/HLVX/HSV H-mode instructions which
depend on hstatus[HUKTE].

Signed-off-by: Fea.Wang <fea.wang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241203034932.25185-3-fea.wang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Add svukte extension capability variable

Refer to the draft of svukte extension from:
https://github.com/riscv/riscv-isa-manual/pull/1564

Svukte provides a means to make user-mode accesses to supervisor memory
raise page faults in constant time, mitigating attacks that attempt to
discover the supervisor software's address-space layout.

Signed-off-by: Fea.Wang <fea.wang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241203034932.25185-2-fea.wang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: Add the checking if DTB overlaps to kernel or initrd

DTB is placed to the end of memory, so we will check if the start
address of DTB overlaps to the address of kernel/initrd.

Signed-off-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241120153935.24706-4-jim.shu@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: Add a new struct RISCVBootInfo

Add a new struct RISCVBootInfo to sync boot information between multiple
boot functions.

Signed-off-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241120153935.24706-3-jim.shu@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: Support to load DTB after 3GB memory on 64-bit system.

Larger initrd image will overlap the DTB at 3GB address. Since 64-bit
system doesn't have 32-bit addressable issue, we just load DTB to the end
of dram in 64-bit system.

Signed-off-by: Jim Shu <jim.shu@sifive.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241120153935.24706-2-jim.shu@sifive.com>
[ Changes by AF
 -  Store fdt_load_addr_hi32 in the reset vector
]
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/char/riscv_htif: Clarify MemoryRegionOps expect 32-bit accesses

Looking at htif_mm_ops[] read/write handlers, we notice they
expect 32-bit values to accumulate into to the 'fromhost' and
'tohost' 64-bit variables. Explicit by setting the .impl
min/max fields.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241129154304.34946-4-philmd@linaro.org>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/char/riscv_htif: Explicit little-endian implementation

Since our RISC-V system emulation is only built for little
endian, the HTIF device aims to interface with little endian
memory accesses, thus we can explicit htif_mm_ops:endianness
being DEVICE_LITTLE_ENDIAN.

In that case tswap64() is equivalent to le64_to_cpu(), as in
"convert this 64-bit little-endian value into host cpu order".
Replace to simplify.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241129154304.34946-3-philmd@linaro.org>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

MAINTAINERS: Cover RISC-V HTIF interface

The HTIF interface is RISC-V specific, add
it within the MAINTAINERS section covering
hw/riscv/.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241129154304.34946-2-philmd@linaro.org>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

tests/qtest/bios-tables-test: Update virt SPCR golden reference for RISC-V

Update the virt SPCR golden reference file for RISC-V to accommodate the
SPCR Table revision 4 [1], utilizing the iasl binary compiled from the
latest ACPICA repository. The SPCR table has been modified to
adhere to the revision 4 format [2].

[1]: https://learn.microsoft.com/en-us/windows-hardware/drivers/serports/serial-port-console-redirection-table
[2]: https://github.com/acpica/acpica/pull/931

Diffs from iasl:
/*
  * Intel ACPI Component Architecture
  * AML/ASL+ Disassembler version 20200925 (64-bit version)
  * Copyright (c) 2000 - 2020 Intel Corporation
  *
- * Disassembly of tests/data/acpi/riscv64/virt/SPCR, Wed Aug 28 18:28:19 2024
+ * Disassembly of /tmp/aml-MN0NS2, Wed Aug 28 18:28:19 2024
  *
  * ACPI Data Table [SPCR]
  *
  * Format: [HexOffset DecimalOffset ByteLength]  FieldName : FieldValue
  */

 [000h 0000   4]                    Signature : "SPCR"    [Serial Port Console Redirection table]
-[004h 0004   4]                 Table Length : 00000050
-[008h 0008   1]                     Revision : 02
-[009h 0009   1]                     Checksum : B9
+[004h 0004   4]                 Table Length : 0000005A
+[008h 0008   1]                     Revision : 04
+[009h 0009   1]                     Checksum : 13
 [00Ah 0010   6]                       Oem ID : "BOCHS "
 [010h 0016   8]                 Oem Table ID : "BXPC    "
 [018h 0024   4]                 Oem Revision : 00000001
 [01Ch 0028   4]              Asl Compiler ID : "BXPC"
 [020h 0032   4]        Asl Compiler Revision : 00000001

-[024h 0036   1]               Interface Type : 00
+[024h 0036   1]               Interface Type : 12
 [025h 0037   3]                     Reserved : 000000

 [028h 0040  12]         Serial Port Register : [Generic Address Structure]
 [028h 0040   1]                     Space ID : 00 [SystemMemory]
 [029h 0041   1]                    Bit Width : 20
 [02Ah 0042   1]                   Bit Offset : 00
 [02Bh 0043   1]         Encoded Access Width : 01 [Byte Access:8]
 [02Ch 0044   8]                      Address : 0000000010000000

 [034h 0052   1]               Interrupt Type : 10
 [035h 0053   1]          PCAT-compatible IRQ : 00
 [036h 0054   4]                    Interrupt : 0000000A
 [03Ah 0058   1]                    Baud Rate : 07
 [03Bh 0059   1]                       Parity : 00
 [03Ch 0060   1]                    Stop Bits : 01
 [03Dh 0061   1]                 Flow Control : 00
 [03Eh 0062   1]                Terminal Type : 00
 [04Ch 0076   1]                     Reserved : 00
 [040h 0064   2]                PCI Device ID : FFFF
 [042h 0066   2]                PCI Vendor ID : FFFF
 [044h 0068   1]                      PCI Bus : 00
 [045h 0069   1]                   PCI Device : 00
 [046h 0070   1]                 PCI Function : 00
 [047h 0071   4]                    PCI Flags : 00000000
 [04Bh 0075   1]                  PCI Segment : 00
-[04Ch 0076   4]                     Reserved : 00000000
+[04Ch 0076 004h]             Uart Clock Freq : 00000000
+[050h 0080 004h]           Precise Baud rate : 00000000
+[054h 0084 002h]       NameSpaceStringLength : 0002
+[056h 0086 002h]       NameSpaceStringOffset : 0058
+[058h 0088 002h]             NamespaceString : "."

-Raw Table Data: Length 80 (0x50)
+Raw Table Data: Length 90 (0x5A)

-    0000: 53 50 43 52 50 00 00 00 02 B9 42 4F 43 48 53 20  // SPCRP.....BOCHS
+    0000: 53 50 43 52 5A 00 00 00 04 13 42 4F 43 48 53 20  // SPCRZ.....BOCHS
     0010: 42 58 50 43 20 20 20 20 01 00 00 00 42 58 50 43  // BXPC    ....BXPC
-    0020: 01 00 00 00 00 00 00 00 00 20 00 01 00 00 00 10  // ......... ......
+    0020: 01 00 00 00 12 00 00 00 00 20 00 01 00 00 00 10  // ......... ......
     0030: 00 00 00 00 10 00 0A 00 00 00 07 00 01 00 00 03  // ................
     0040: FF FF FF FF 00 00 00 00 00 00 00 00 00 00 00 00  // ................
+    0050: 00 00 00 00 02 00 58 00 2E 00                    // ......X...

Signed-off-by: Sia Jee Heng <jeeheng.sia@starfivetech.com>
Reviewed-by: Sunil V L <sunilvl@ventanamicro.com>
Message-ID: <20241028015744.624943-4-jeeheng.sia@starfivetech.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/acpi: Upgrade ACPI SPCR table to support SPCR table revision 4 format

Update the SPCR table to accommodate the SPCR Table revision 4 [1].
The SPCR table has been modified to adhere to the revision 4 format [2].

[1]: https://learn.microsoft.com/en-us/windows-hardware/drivers/serports/serial-port-console-redirection-table
[2]: https://github.com/acpica/acpica/pull/931

Signed-off-by: Sia Jee Heng <jeeheng.sia@starfivetech.com>
Reviewed-by: Sunil V L <sunilvl@ventanamicro.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bibo Mao <maobibo@loongson.cn>
Message-ID: <20241028015744.624943-3-jeeheng.sia@starfivetech.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

qtest: allow SPCR acpi table changes

Signed-off-by: Sia Jee Heng <jeeheng.sia@starfivetech.com>
Reviewed-by: Sunil V L <sunilvl@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241028015744.624943-2-jeeheng.sia@starfivetech.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: Add Microblaze V generic board

Add a basic board with interrupt controller (intc), timer, serial
(uartlite), small memory called LMB@0 (128kB) and DDR@0x80000000
(configured via command line eg. -m 2g).
This is basic configuration which matches HW generated out of AMD Vivado
(design tools). But initial configuration is going beyond what it is
configured by default because validation should be done on other
configurations too. That's why wire also additional uart16500, axi
ethernet(with axi dma).
GPIOs, i2c and qspi is also listed for completeness.

IRQ map is: (addr)
0 - timer (0x41c00000)
1 - uartlite (0x40600000)
2 - i2c (0x40800000)
3 - qspi (0x44a00000)
4 - uart16550 (0x44a10000)
5 - emaclite (0x40e00000)
6 - timer2 (0x41c10000)
7 - axi emac (0x40c00000)
8 - axi dma (0x41e00000)
9 - axi dma
10 - gpio (0x40000000)
11 - gpio2 (0x40010000)
12 - gpio3 (0x40020000)

Signed-off-by: Sai Pavan Boddu <sai.pavan.boddu@amd.com>
Signed-off-by: Michal Simek <michal.simek@amd.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241125134739.18189-1-sai.pavan.boddu@amd.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

docs: update riscv/virt.rst with kernel-irqchip=split support

Also add a new page, docs/specs/riscv-aia.rst, where we're documenting
the state of AIA support in QEMU w.r.t the controllers being emulated or
not depending on the AIA and accelerator settings.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-9-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/kvm: remove irqchip_split() restriction

Remove the 'irqchip_split()' restriction in kvm_arch_init() now that
we have support for "-accel kvm,kernel-irqchip=split".

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-8-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/intc/riscv_aplic: add kvm_msicfgaddr for split mode aplic-imsic

The last step to enable KVM AIA aplic-imsic with irqchip in split mode
is to deal with how MSIs are going to be sent. In our current design we
don't allow an APLIC controller to send MSIs unless it's on m-mode. And
we also do not allow Supervisor MSI address configuration via the
'smsiaddrcfg' and 'smsiaddrcfgh' registers unless it's also a m-mode
APLIC controller.

Add a new RISCVACPLICState attribute called 'kvm_msicfgaddr'. This
attribute represents the base configuration address for MSIs, in our
case the base addr of the IMSIC controller. This attribute is being set
only when running irqchip_split() mode with aia=aplic-imsic.

During riscv_aplic_msi_send() we'll check if the attribute was set to
skip the check for a m-mode APLIC controller and to change the resulting
MSI addr by adding kvm_msicfgaddr right before address_space_stl_le().

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c, riscv_aplic.c: add 'emulated_aplic' helpers

The current logic to determine if we don't need an emulated APLIC
controller, i.e. KVM will provide for us, is to determine if we're
running KVM, with in-kernel irqchip support, and running
aia=aplic-imsic. This is modelled by riscv_is_kvm_aia_aplic_imsic() and
virt_use_kvm_aia_aplic_imsic().

This won't suffice to support irqchip_split() mode: it will match
exactly the same conditions as the one above, but setting the irqchip to
'split' mode will now require us to emulate an APLIC s-mode controller,
like we're doing with 'aia=aplic'.

Create a new riscv_use_emulated_aplic() helper that will encapsulate
this logic. Replace the uses of "riscv_is_kvm_aia_aplic_imsic()" with
this helper every time we're taking a decision on emulate an APLIC
controller or not. Do the same in virt.c with virt_use_emulated_aplic().

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-6-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv/kvm: consider irqchip_split() in aia_create()

Before adding support to kernel-irqchip=split when using KVM AIA we need
to change how we create the in-kernel AIA device.

In the use case we have so far, i.e. in-kernel irqchip without split
mode, both the s-mode APLIC and IMSIC controllers are provided by the
irqchip. In irqchip_split() mode we'll emulate the s-mode APLIC
controller, which will send MSIs to the in-kernel IMSIC controller. To
do that we need to change kvm_riscv_aia_create() to not create the
in-kernel s-mode APLIC controller.

In the kernel source arch/riscv/kvm/aia_aplic.c, function
kvm_riscv_aia_aplic_init(), we verify that the APLIC controller won't be
instantiated by KVM if we do not set 'nr_sources', which is set via
KVM_DEV_RISCV_AIA_CONFIG_SRCS. For QEMU this means that we should not
set 'aia_irq_num' during kvm_riscv_aia_create() in irqchip_split() mode.

In this same condition, skip KVM_DEV_RISCV_AIA_ADDR_APLIC as well since
it is used to set the base address for the in-kernel APLIC controller.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c: rename helper to virt_use_kvm_aia_aplic_imsic()

Similar to the riscv_is_kvm_aia_aplic_imsic() helper from riscv_aplic.c,
the existing virt_use_kvm_aia() is testing for KVM aia=aplic-imsic with
in-kernel irqchip enabled. It is not checking for a generic AIA support.

Rename the helper to virt_use_kvm_aia_aplic_imsic() to reflect what the
helper is doing, and use the existing riscv_is_kvm_aia_aplic_imsic() to
obscure details such as the presence of the in-kernel irqchip.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c: reduce virt_use_kvm_aia() usage

In create_fdt_sockets() we have the following pattern:

    if (kvm_enabled() && virt_use_kvm_aia(s)) {
        (... do stuff ...)
    } else {
        (... do other stuff ...)
    }
    if (kvm_enabled() && virt_use_kvm_aia(s)) {
        (... do more stuff ...)
    } else {
        (... do more other stuff)
    }

Do everything in a single if/else clause to reduce the usage of
virt_use_kvm_aia() helper and to make the code a bit less repetitive.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/intc/riscv_aplic: rename is_kvm_aia()

The helper is_kvm_aia() is checking not only for AIA, but for
aplic-imsic (i.e. "aia=aplic-imsic" in 'virt' RISC-V machine) with an
in-kernel chip present.

Rename it to be a bit clear what the helper is doing since we'll add
more AIA helpers in the next patches.

Make the helper public because the 'virt' machine will use it as well.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241119191706.718860-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

target/riscv: Add Tenstorrent Ascalon CPU

Add a CPU entry for the Tenstorrent Ascalon CPU, a series of 2 wide to
8 wide RV64 cores. More details can be found at
https://tenstorrent.com/ip/tt-ascalon

Signed-off-by: Anton Blanchard <antonb@tenstorrent.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241113110459.1607299-1-antonb@tenstorrent.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

docs/specs: add riscv-iommu-sys information

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241106133407.604587-8-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/riscv-iommu: implement reset protocol

Add a riscv_iommu_reset() helper in the base emulation code that
implements the expected reset behavior as defined by the riscv-iommu
spec.

Devices can then use this helper in their own reset callbacks.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241106133407.604587-7-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt.c, riscv-iommu-sys.c: add MSIx support

MSIx support is added in the RISC-V IOMMU platform device by including
the required MSIx facilities to alow software to properly setup the MSIx
subsystem.

We took inspiration of what is being done in the riscv-iommu-pci device,
mainly msix_init() and msix_notify(), while keeping in mind that
riscv-iommu-sys isn't a true PCI device and we don't need to copy/paste
all the contents of these MSIx functions.

Two extra MSI MemoryRegions were added: 'msix-table' and 'msix-pba'.
They are used to manage r/w of the MSI table and Pending Bit Array (PBA)
respectively. Both are subregions of the main IOMMU memory region,
iommu->regs_mr, initialized during riscv_iommu_realize(), and each one
has their own handlers for MSIx reads and writes.

This is the expected memory map when using this device in the 'virt'
machine:

    0000000003010000-0000000003010fff (prio 0, i/o): riscv-iommu-regs
      0000000003010300-000000000301034f (prio 0, i/o): msix-table
      0000000003010400-0000000003010407 (prio 0, i/o): msix-pba

We're now able to set IGS to RISCV_IOMMU_CAP_IGS_BOTH, and userspace is
free to decide which interrupt model to use.

Enabling MSIx support for this device in the 'virt' machine requires
adding 'msi-parent' in the iommu-sys DT.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241106133407.604587-6-dbarboza@ventanamicro.com>
[ Changes by AF:
 - Used PRIx64 in trace
]
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/virt: Add IOMMU as platform device if the option is set

Add a new machine option called 'iommu-sys' that enables a
riscv-iommu-sys platform device for the 'virt' machine. The option is
default 'off'.

The device will use IRQs 36 to 39.

We will not support both riscv-iommu-sys and riscv-iommu-pci devices in
the same board in this first implementation. If a riscv-iommu-pci device
is added in the command line we will disable the riscv-iommu-sys device.

Signed-off-by: Sunil V L <sunilvl@ventanamicro.com>
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241106133407.604587-5-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv: add riscv-iommu-sys platform device

This device models the RISC-V IOMMU as a sysbus device. The same design
decisions taken in the riscv-iommu-pci device were kept, namely the
existence of 4 vectors are available for each interrupt cause.

The WSIs are emitted using the input of the s->notify() callback as a
index to an IRQ list. The IRQ list starts at 'base_irq' and goes until
base_irq + 3. This means that boards must have 4 contiguous IRQ lines
available, starting from 'base_irq'.

Signed-off-by: Tomasz Jeznach <tjeznach@rivosinc.com>
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241106133407.604587-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/riscv-iommu: parametrize CAP.IGS

Interrupt Generation Support (IGS) is a capability that is tied to the
interrupt deliver mechanism, not with the core IOMMU emulation. We
should allow device implementations to set IGS as they wish.

A new helper is added to make it easier for device impls to set IGS. Use
it in our existing IOMMU device (riscv-iommu-pci) to set
RISCV_IOMMU_CAPS_IGS_MSI.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241106133407.604587-3-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/riscv-iommu.c: add riscv_iommu_instance_init()

Move all the static initializion of the device to an init() function,
leaving only the dynamic initialization to be done during realize.

With this change s->cap is initialized with RISCV_IOMMU_CAP_DBG during
init(), and realize() will increment s->cap with the extra caps.

This will allow callers to add IOMMU capabilities before the
realization.

Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241106133407.604587-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/intc/riscv_aplic: Fix APLIC in_clrip and clripnum write emulation

In the section "4.7 Precise effects on interrupt-pending bits"
of the RISC-V AIA specification defines that:

"If the source mode is Level1 or Level0 and the interrupt domain
is configured in MSI delivery mode (domaincfg.DM = 1):
The pending bit is cleared whenever the rectified input value is
low, when the interrupt is forwarded by MSI, or by a relevant
write to an in_clrip register or to clripnum."

Update the riscv_aplic_set_pending() to match the spec.

Fixes: bf31cf06eb ("hw/intc/riscv_aplic: Fix setipnum_le write emulation for APLIC MSI-mode")
Signed-off-by: Yong-Xuan Wang <yongxuan.wang@sifive.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241029085349.30412-1-yongxuan.wang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

hw/riscv/riscv-iommu.c: Correct the validness check of iova

From RISCV IOMMU spec section 2.1.3:
When SXL is 1, the following rules apply:
- If the first-stage is not Bare, then a page fault corresponding to the
original access type occurs if the IOVA has bits beyond bit 31 set to 1.
- If the second-stage is not Bare, then a guest page fault corresponding
to the original access type occurs if the incoming GPA has bits beyond bit
33 set to 1.

From RISCV IOMMU spec section 2.3 step 17:
Use the process specified in Section "Two-Stage Address Translation" of
the RISC-V Privileged specification to determine the GPA accessed by the
transaction.

From RISCV IOMMU spec section 2.3 step 19:
Use the second-stage address translation process specified in Section
"Two-Stage Address Translation" of the RISC-V Privileged specification
to translate the GPA A to determine the SPA accessed by the transaction.

This commit adds the iova check with the following rules:
- For Sv32, Sv32x4, Sv39x4, Sv48x4 and Sv57x4, the iova must be zero
extended.
- For Sv39, Sv48 and Sv57, the iova must be signed extended with most
significant bit.

Signed-off-by: Jason Chien <jason.chien@sifive.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Message-ID: <20241114065617.25133-1-jason.chien@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>

Merge tag 'pull-loongarch-20241219' of https://gitlab.com/bibo-mao/qemu into staging

loongarch queue

# -----BEGIN PGP SIGNATURE-----
#
# iHUEABYKAB0WIQQNhkKjomWfgLCz0aQfewwSUazn0QUCZ2PKBQAKCRAfewwSUazn
# 0QAZAQCxbLnvzOb9TPORlg5w0n/xFaKCL7dJbJE4WjlM7dhLkAEA5G8JVoP5Ju2B
# mcK7wbymyXNX1ocsukL/JM2JavHS+AI=
# =JoSk
# -----END PGP SIGNATURE-----
# gpg: Signature made Thu 19 Dec 2024 02:23:49 EST
# gpg:                using EDDSA key 0D8642A3A2659F80B0B3D1A41F7B0C1251ACE7D1
# gpg: Good signature from "bibo mao <maobibo@loongson.cn>" [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: 7044 3A00 19C0 E97A 31C7  13C4 8E86 8FB7 A176 9D4C
#      Subkey fingerprint: 0D86 42A3 A265 9F80 B0B3  D1A4 1F7B 0C12 51AC E7D1

* tag 'pull-loongarch-20241219' of https://gitlab.com/bibo-mao/qemu:
  hw/intc/loongarch_extioi: Code cleanup about loongarch_extioi
  hw/intc/loongarch_extioi: Add pre_save interface
  hw/intc/loongarch_extioi: Inherit from loongarch_extioi_common
  hw/intc/loongarch_extioi: Add common file loongarch_extioi_common
  hw/intc/loongarch_extioi: Add unrealize interface
  hw/intc/loongarch_extioi: Add common realize interface
  hw/intc/loongarch_extioi: Rename LoongArchExtIOI with LoongArchExtIOICommonState
  include: Rename LoongArchExtIOI with LoongArchExtIOICommonState
  include: Move struct LoongArchExtIOI to header file loongarch_extioi_common
  include: Add loongarch_extioi_common header file
  hw/intc/loongarch_pch: Code cleanup about loongarch_pch_pic
  hw/intc/loongarch_pch: Add pre_save and post_load interfaces
  hw/intc/loongarch_pch: Inherit from loongarch_pic_common
  hw/intc/loongarch_pch: Move some functions to file loongarch_pic_common
  hw/intc/loongarch_pch: Rename LoongArchPCHPIC with LoongArchPICCommonState
  hw/intc/loongarch_pch: Merge instance_init() into realize()
  include: Move struct LoongArchPCHPIC to loongarch_pic_common header file
  include: Add loongarch_pic_common header file

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

rust: pl011: simplify handling of the FIFO enabled bit in LCR

Use ==/!= instead of going through bool and xor.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: pl011: fix migration stream

The Rust vmstate macros lack the type-safety of their C equivalents (so
safe, much abstraction), and therefore they were predictably wrong.

The registers have already been changed to 32-bits in the previous patch,
but read_pos/read_count/read_trigger also have to be u32 instead of usize.
The easiest way to do so is to let the FIFO use u32 indices instead
of usize.

My plan for making VMStateField typesafe is to have a trait to retrieve
a basic VMStateField; for example something like vmstate_uint32 would
become an implementation of the VMState trait on u32.  Then you'd write
something like "vmstate_of!(Type, field).with_version_id(2)".  That is,
vmstate_of retrieves the basic VMStateField and fills in the offset,
and then more changes can be applied on top.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: pl011: extend registers to 32 bits

The PL011 Technical Reference Manual lists the "real" size of the
registers in table 3-1, and only rounds up to the next byte when
describing the registers; for example, UARTDR is listed as having
width 12/8 (12 bits read, 8 written) and only bits 15:0 are listed
in "Table 3-2 UARTDR Register".

However, in practice these are 32-bit registers, accessible only
through 32-bit MMIO accesses; preserving the fiction that they're
smaller introduces multiple casts (to go from the bilge bitfield
type to e.g u16 to u64) and more importantly it breaks the
migration stream because the Rust vmstate macros are not yet
type safe.

So, just make everything 32-bits wide.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: pl011: fix break errors and definition of Data struct

The Data struct is wrong, and does not show how bits 8-15 of DR
are the receive status.  Fix it, and use it to fix break
errors ("c >> 8" in the C code does not translate to
"c.to_be_bytes()[3]").

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: pl011: always use reset() method on registers

For CR, the ugly-ish "0.into()" idiom is already hidden within the
reset method.  Do not repeat it.

For FR, standardize on reset() being equivalent to "*self = Self::default()"
and let reset_fifo toggle only the bits that are related to FIFOs.  This
commit also reproduces C commit 02b1f7f6192 ("hw/char/pl011: Split RX/TX
path of pl011_reset_fifo()", 2024-09-13).

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: pl011: match break logic of C version

Check loopback_enabled(), not fifo_enabled(), like the C code.

Also, set_break_error() must not happen until the break is read from
the FIFO.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: pl011: fix declaration of LineControl bits

The bits in the LineControl struct were backwards. :(

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: Reset TSCs of parked vCPUs too on VM reset

Since commit 5286c3662294 ("target/i386: properly reset TSC on reset")
QEMU writes the special value of "1" to each online vCPU TSC on VM reset
to reset it.

However parked vCPUs don't get that handling and due to that their TSCs
get desynchronized when the VM gets reset.
This in turn causes KVM to turn off PVCLOCK_TSC_STABLE_BIT in its exported
PV clock.
Note that KVM has no understanding of vCPU being currently parked.

Without PVCLOCK_TSC_STABLE_BIT the sched clock is marked unstable in
the guest's kvm_sched_clock_init().
This causes a performance regressions to show in some tests.

Fix this issue by writing the special value of "1" also to TSCs of parked
vCPUs on VM reset.

Reproducing the issue:
1) Boot a VM with "-smp 2,maxcpus=3" or similar

2) device_add host-x86_64-cpu,id=vcpu,node-id=0,socket-id=0,core-id=2,thread-id=0

3) Wait a few seconds

4) device_del vcpu

5) Inside the VM run:
# echo "t" >/proc/sysrq-trigger; dmesg | grep sched_clock_stable
Observe the sched_clock_stable() value is 1.

6) Reboot the VM

7) Once the VM boots once again run inside it:
# echo "t" >/proc/sysrq-trigger; dmesg | grep sched_clock_stable
Observe the sched_clock_stable() value is now 0.

Fixes: 5286c3662294 ("target/i386: properly reset TSC on reset")
Signed-off-by: Maciej S. Szmigiero <maciej.szmigiero@oracle.com>
Link: https://lore.kernel.org/r/5a605a88e9a231386dc803c60f5fed9b48108139.1734014926.git.maciej.szmigiero@oracle.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

kvm: consistently return 0/-errno from kvm_convert_memory

In case of incorrect parameters, kvm_convert_memory() was returning
-1 instead of -EINVAL.  The guest won't notice because it will move
anyway to RUN_STATE_INTERNAL_ERROR, but fix this for consistency and
clarity.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: qemu-api: add a module to wrap functions and zero-sized closures

One recurring issue when writing Rust bindings is how to convert a Rust
function ("fn" or "impl Fn") to a C function, and how to pass around
"self" to a C function that only takes a void*.

An easy solution would be to store on the heap a pair consisting of
a pointer to the Rust function and the pointer to "self", but it is
possible to do better.  If an "Fn" has zero size (that is, if it is a
zero-capture closures or a function pointer---which in turn includes all
methods), it is possible to build a generic Rust function that calls it
even if you only have the type; you don't need either the pointer to the
function itself (because the address of the code is part of the type)
or any closure data (because it has size zero).

Introduce a wrapper that provides the functionality of calling the
function given only its type.

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: qom: add initial subset of methods on Object

Add an example of implementing instance methods and converting the
result back to a Rust type.  In this case the returned types are a
string (actually a Cow<str>; but that's transparent as long as it derefs
to &str) and a QOM class.

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: qom: add casting functionality

Add traits that let client cast typecast safely between object types.
In particular, an upcast is compile-time guaranteed to succeed, and a
YOLO C-style downcast must be marked as unsafe.

The traits are based on an IsA<> trait that declares what
is a subclass of what, which is an idea taken from glib-rs
(https://docs.rs/glib/latest/glib/object/trait.IsA.html).
The four primitives are also taken from there
(https://docs.rs/glib/latest/glib/object/trait.Cast.html).  However,
the implementation of casting itself is a bit different and uses the
Deref trait.

This removes some pointer arithmetic from the pl011 device; it is also a
prerequisite for the definition of methods, so that they can be invoked
on all subclass structs.  This will use the IsA<> trait to detect the
structs that support the methods.

glib also has a "monadic" casting trait which could be implemented on
Option (as in https://docs.rs/glib/latest/glib/object/trait.CastNone.html)
and perhaps even Result.  For now I'm leaving it out, as the patch is
already big enough and the benefit seems debatable.

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: tests: allow writing more than one test

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

bql: add a "mock" BQL for Rust unit tests

Right now, the stub BQL in stubs/iothread-lock.c always reports itself as
unlocked.  However, Rust would like to run its tests in an environment where
the BQL *is* locked.  Provide an extremely dirty function that flips the
return value of bql_is_locked() to true.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: re-export C types from qemu-api submodules

Long term we do not want device code to use "bindings" at all, so make it
possible to get the relevant types from the other modules of qemu-api.

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: rename qemu-api modules to follow C code a bit more

A full match would mean calling them qom::object and hw::core::qdev.  For now,
keep the names shorter but still a bit easier to find.

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: qom: add possibility of overriding unparent

Add a blanket definition of ClassInitImpl<ObjectClass> that thunks
ObjectImpl::UNPARENT and overrides it in ObjectClass if it is not
None.

ClassInitImpl<DeviceClass> can now call its superclass's ClassInitImpl,
so that the C and Rust hierarchies match more closely.

This is mostly done as an example of implementing the metaclass
hierarchy under ClassInitImpl.

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust: qom: put class_init together from multiple ClassInitImpl<>

Parameterize the implementation of ClassInitImpl so that it is
possible to call up the chain of implementations, one superclass at
a time starting at ClassInitImpl<Self::Class>.

In order to avoid having to implement (for example)
ClassInitImpl<PL011Class>, also remove the dummy PL011Class and
PL011LuminaryClass structs and specify the same ObjectType::Class as
the superclass.  In the future this default behavior can be handled by
a procedural macro, by looking at the first field in the struct.

Note that the new trait is safe: the calls are started by
rust_class_init<>(), which is not public and can convert the class
pointer to a Rust reference.

Since CLASS_BASE_INIT applies to the type that is being defined,
and only to it, move it to ObjectImpl.

Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Constify all opaque Property pointers

Via sed "s/  Property [*]/  const Property */".

The opaque pointers passed to ObjectProperty callbacks are
the last instances of non-const Property pointers in the tree.
For the most part, these callbacks only use object_field_prop_ptr,
which now takes a const pointer itself.

This logically should have accompanied d36f165d952 which
allowed const Property to be registered.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-25-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/core/qdev-properties: Constify Property argument to PropertyInfo.print

This logically should have accompanied d36f165d952 which
allowed const Property to be registered.

There is exactly one instance of this method: print_pci_devfn.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-24-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/core/qdev-properties: Constify Property argument to object_field_prop_ptr

This logically should have accompanied d36f165d952 which
allowed const Property to be registered.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-23-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

include/hw/qdev-properties: Shrink struct Property

Before, via pahole:

arm32, a 32-bit host which aligns uint64_t:

struct Property {
        const char  *              name;                 /*     0     4 */
        const PropertyInfo  *      info;                 /*     4     4 */
        ptrdiff_t                  offset;               /*     8     4 */
        uint8_t                    bitnr;                /*    12     1 */
        /* XXX 3 bytes hole, try to pack */
        uint64_t                   bitmask;              /*    16     8 */
        _Bool                      set_default;          /*    24     1 */
        /* XXX 7 bytes hole, try to pack */
        union {
                int64_t            i;                    /*    32     8 */
                uint64_t           u;                    /*    32     8 */
        } defval;                                        /*    32     8 */
        int                        arrayoffset;          /*    40     4 */
        const PropertyInfo  *      arrayinfo;            /*    44     4 */
        int                        arrayfieldsize;       /*    48     4 */
        const char  *              link_type;            /*    52     4 */
        /* size: 56, cachelines: 1, members: 11 */
        /* sum members: 46, holes: 2, sum holes: 10 */
};

arm64, an arbitrary 64-bit host:

struct Property {
        const char  *              name;                 /*     0     8 */
        const PropertyInfo  *      info;                 /*     8     8 */
        ptrdiff_t                  offset;               /*    16     8 */
        uint8_t                    bitnr;                /*    24     1 */
        /* XXX 7 bytes hole, try to pack */
        uint64_t                   bitmask;              /*    32     8 */
        _Bool                      set_default;          /*    40     1 */
        /* XXX 7 bytes hole, try to pack */
        union {
                int64_t            i;                    /*    48     8 */
                uint64_t           u;                    /*    48     8 */
        } defval;                                        /*    48     8 */
        int                        arrayoffset;          /*    56     4 */
        /* XXX 4 bytes hole, try to pack */
        const PropertyInfo  *      arrayinfo;            /*    64     8 */
        int                        arrayfieldsize;       /*    72     4 */
        /* XXX 4 bytes hole, try to pack */
        const char  *              link_type;            /*    80     8 */
        /* size: 88, cachelines: 2, members: 11 */
        /* sum members: 66, holes: 4, sum holes: 22 */
};

Afterward there are no holes in either structure.
For arm32, size 48, padding 2, saved 8 bytes.
For arm64, size 72, padding 6, saved 16 bytes.

Saves 20k from qemu-system-aarch64 on a 64-bit host.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-22-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

include/hw/qdev-properties: Remove DEFINE_PROP_END_OF_LIST

Now that all of the Property arrays are counted, we can remove
the terminator object from each array.  Update the assertions
in device_class_set_props to match.

With struct Property being 88 bytes, this was a rather large
form of terminator.  Saves 30k from qemu-system-aarch64.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-21-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/riscv: Do not abuse DEFINE_PROP_END_OF_LIST

These are not arrays of Property and had no business
using DEFINE_PROP_END_OF_LIST.  Use plain { } instead.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-20-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/core: Remove device_class_set_props function

All uses of device_class_set_props() are now using arrays.
Validate this compile-time in the device_class_set_props macro and
call device_class_set_props_n using the known size of the array.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-19-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

rust/qemu-api: Use device_class_set_props_n

This means we can update declare_properties to drop the
zero terminator at the end of the array as well.

Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-18-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/arm/armsse: Use device_class_set_props_n

We must remove DEFINE_PROP_END_OF_LIST so the count is correct.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-17-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/scsi/megasas: Use device_class_set_props_n

We must remove DEFINE_PROP_END_OF_LIST so the count is correct.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-16-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

migration: Use device_class_set_props_n

Export the migration_properties array size from options.c;
use that to feed device_class_set_props_n.  We must remove
DEFINE_PROP_END_OF_LIST so the count is correct.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-15-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/core: Introduce device_class_set_props_n

Record the size of the array in DeviceClass.props_count_.
Iterate with known count in qdev_prop_walk.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-14-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

include/hw/qdev-core: Detect most empty Property lists at compile time

Add a macro expansion of device_class_set_props which can check
on the type and size of PROPS before calling the function.

Avoid the macro in migration.c because migration_properties
is defined externally with indeterminate size.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-13-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/virtio: Remove empty Property lists

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-12-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/sparc: Remove empty Property lists

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-11-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/xen: Remove empty Property lists

There is no point in registering no properties.
Remove xen_sysdev_class_init entirely, as it did nothing else.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-10-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/s390x: Remove empty Property lists

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-9-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/tricore: Remove empty Property lists

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-8-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/ppc: Only register spapr_nvdimm_properties if CONFIG_LIBPMEM

Do not register an empty set of properties.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-7-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/pci-host/astro: Remove empty Property list

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-6-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/s390x: Use s390x_cpu_properties for system mode only

Avoid the empty property list for user-only mode.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-5-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/ppc: Remove empty property list

Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-4-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/ide: Constify sysbus_ahci_properties

Reviewed-by: Bernhard Beschow <shentey@gmail.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-3-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

migration: Constify migration_properties

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Lei Yang <leiyang@redhat.com>
Link: https://lore.kernel.org/r/20241218134251.4724-2-richard.henderson@linaro.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/intc/loongarch_extioi: Code cleanup about loongarch_extioi

Remove definition about LoongArchExtIOI and LOONGARCH_EXTIOI, and
replace them with LoongArchExtIOICommonState and macro
LOONGARCH_EXTIOI_COMMON separately. Also remove unnecessary header
files.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_extioi: Add pre_save interface

Add vmstate pre_save interface, which can be used extioi kvm driver
in future.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_extioi: Inherit from loongarch_extioi_common

Set TYPE_LOONGARCH_EXTIOI inherit from TYPE_LOONGARCH_EXTIOI_COMMON
object, it shares vmsate and property of TYPE_LOONGARCH_EXTIOI_COMMON,
and has its own realize() function.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_extioi: Add common file loongarch_extioi_common

Add new common file loongarch_extioi_common.c, and move vmstate
and property structure to common file.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_extioi: Add unrealize interface

For loongarch extioi emulation driver, add unrealize interface and
remove instance_finalize interface and move the code to unrealize
interface.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_extioi: Add common realize interface

Add common realize function, it is only to check validity of property.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_extioi: Rename LoongArchExtIOI with LoongArchExtIOICommonState

With some structure such as vmstate and property, rename LoongArchExtIOI
with LoongArchExtIOICommonState, these common structure will be moved
to common file.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

include: Rename LoongArchExtIOI with LoongArchExtIOICommonState

Rename structure LoongArchExtIOI with LoongArchExtIOICommonState,
since it is defined in file loongarch_extioi_common.h

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

include: Move struct LoongArchExtIOI to header file loongarch_extioi_common

Move definiton of structure LoongArchExtIOI from header file loongarch_extioi.h
to file loongarch_extioi_common.h.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

include: Add loongarch_extioi_common header file

Add common header file include/hw/intc/loongarch_extioi_common.h, and
move some macro definition from include/hw/intc/loongarch_extioi.h to
the common header file.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_pch: Code cleanup about loongarch_pch_pic

Remove definition about LoongArchPCHPIC and LOONGARCH_PCH_PIC, and
replace them with LoongArchPICCommonState and LOONGARCH_PIC_COMMON
separately. Also remove unnecessary header files.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_pch: Add pre_save and post_load interfaces

Add vmstate pre_save and post_load interfaces, which can be used
by pic kvm driver in future.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_pch: Inherit from loongarch_pic_common

Set TYPE_LOONGARCH_PIC inherit from TYPE_LOONGARCH_PIC_COMMON object,
it shares vmsate and property of TYPE_LOONGARCH_PIC_COMMON, and has
its own realize() function.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_pch: Move some functions to file loongarch_pic_common

Move some common functions to file loongarch_pic_common.c, the common
functions include loongarch_pic_common_realize(), property structure
loongarch_pic_common_properties and vmstate structure
vmstate_loongarch_pic_common.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_pch: Rename LoongArchPCHPIC with LoongArchPICCommonState

With pic vmstate, rename structure name vmstate_loongarch_pch_pic with
vmstate_loongarch_pic_common, and with pic property rename
loongarch_pch_pic_properties with loongarch_pic_common_properties.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>

hw/intc/loongarch_pch: Merge instance_init() into realize()

Memory region is created in instance_init(), merge it into function
realize(). There is no special class_init() for loongarch_pch object.

Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Reviewed-by: Song Gao <gaosong@loongson.cn>