meson: remove CONFIG_POSIX and CONFIG_WIN32 from config_targetos

For consistency with other OSes, use if...endif for rules that are
target-independent.

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

meson: remove OS definitions from config_targetos

CONFIG_DARWIN, CONFIG_LINUX and CONFIG_BSD are used in some rules, but
only CONFIG_LINUX has substantial use.  Convert them all to if...endif.

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

meson: always probe u2f and canokey if the option is enabled

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

meson: move subdirs to "Collect sources" section

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

meson: move config-host.h definitions together

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

meson: move CFI detection code with other compiler flags

Keep it together with the other compiler modes, and before dependencies.

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

meson: keep subprojects together

And move away dependencies that are not subprojects anymore.

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

meson: move accelerator dependency checks together

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

meson: move option validation together

Check options before compiler flags, because some compiler flags are
incompatible with modules.

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

meson: move program checks together

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

meson: add more sections to main meson.build

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

configure: unify again the case arms in probe_target_compiler

Remove assignments that match the default, and group the
targets for debian-legacy-test-cross and debian-all-test-cross
into a single arm.

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

configure: remove unnecessary subshell

Do not use a subshell to hide the shadowing of $config_host_mak.

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

Makefile: clean qemu-iotests output

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

meson: use version_compare() to compare version

While a simple lexicographic comparison usually works, it is less
robust than a more specific algorithm designed to compare versions.

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

meson: remove unused variable

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

meson: fix type of "relocatable" option

Since the option is of boolean type, the default value should be a boolean
rather than a string.

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

esp: check for NULL result from scsi_device_find()

Add a 'current_lun' check for a null value
to avoid null pointer dereferencing and
recover host if NULL return

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: 4eb8606560 (esp: store lun coming from the MESSAGE OUT phase)
Signed-off-by: Alexandra Diupina <adiupina@astralinux.ru>
Message-ID: <20231229152647.19699-1-adiupina@astralinux.ru>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: the sgx_epc_get_section stub is reachable

The sgx_epc_get_section stub is reachable from cpu_x86_cpuid.  It
should not assert, instead it should just return true just like
the "real" sgx_epc_get_section does when SGX is disabled.

Reported-by: Vladimír Beneš <vbenes@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-ID: <20220201190941.106001-1-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: implement CMPccXADD

The main difficulty here is that a page fault when writing to the destination
must not overwrite the flags.  Therefore, the flags computation must be
inlined instead of using gen_jcc1*.

For simplicity, I am using an unconditional cmpxchg operation, that becomes
a NOP if the comparison fails.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: introduce flags writeback mechanism

ALU instructions can write to both memory and flags.  If the CC_SRC*
and CC_DST locations have been written already when a memory access
causes a fault, the value in CC_SRC* and CC_DST might be interpreted
with the wrong CC_OP (the one that is in effect before the instruction.

Besides just using the wrong result for the flags, something like
subtracting -1 can have disastrous effects if the current CC_OP is
CC_OP_EFLAGS: this is because QEMU does not expect bits outside the ALU
flags to be set in CC_SRC, and env->eflags can end up set to all-ones.
In the case of the attached testcase, this sets IOPL to 3 and would
cause an assertion failure if SUB is moved to the new decoder.

This mechanism is not really needed for BMI instructions, which can
only write to a register, but put it to use anyway for cleanliness.
In the case of BZHI, the code has to be modified slightly to ensure
that decode->cc_src is written, otherwise the new assertions trigger.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: adjust decoding of J operand

gen_jcc() has been changed to accept a relative offset since the
new decoder was written.  Adjust the J operand, which is meant
to be used with jump instructions such as gen_jcc(), to not
include the program counter and to not truncate the result, as
both operations are now performed by common code.

The result is that J is now the same as the I operand.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: move operand load and writeback out of gen_cmovcc1

Similar to gen_setcc1, make gen_cmovcc1 receive TCGv.  This is more friendly
to simultaneous implementation in the old and the new decoder.

A small wart is that s->T0 of CMOV is currently the *second* argument (which
would ordinarily be in T1).  Therefore, the condition has to be inverted in
order to overwrite s->T0 with cpu_regs[reg] if the MOV is not performed.

This only applies to the old decoder, and this code will go away soon.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: prepare for implementation of STOS/SCAS in new decoder

Do not use gen_op, and pull the load from the accumulator into
disas_insn.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: do not use s->tmp0 for jumps on ECX ==/!= 0

Create a new temporary, to ease the register allocator's work.

Creation of the temporary is pushed into gen_ext_tl, which
also allows NULL as the first parameter now.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: do not use s->tmp4 for push

Just create a temporary for the occasion.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: split eflags computation out of gen_compute_eflags

The new x86 decoder wants the gen_* functions to compute EFLAGS before
writeback, which can be an issue for instructions with a memory
destination such as ARPL or shifts.

Extract code to compute the EFLAGS without clobbering CC_SRC, in case
the memory write causes a fault.  The flags writeback mechanism will
take care of copying the result to CC_SRC.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: do not clobber T0 on string operations

The new decoder would rather have the operand in T0 when expanding SCAS, rather
than use R_EAX directly as gen_scas currently does.  This makes SCAS more similar
to CMP and SUB, in that CC_DST = T0 - T1.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: do not clobber A0 in POP translation

The new decoder likes to compute the address in A0 very early, so the
gen_lea_v_seg in gen_pop_T0 would clobber the address of the memory
operand.  Instead use T0 since it is already available and will be
overwritten immediately after.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: do not decode string source/destination into decode->mem

decode->mem is only used if one operand has has_ea == true.  String
operations will not use decode->mem and will load A0 on their own, because
they are the only case of two memory operands in a single instruction.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: add X86_SPECIALs for MOVSX and MOVZX

Usually the registers are just moved into s->T0 without much care for
their operand size.  However, in some cases we can get more efficient
code if the operand fetching logic syncs with the emission function
on what is nicer.

All the current uses are mostly demonstrative and only reduce the code
in the emission functions, because the instructions do not support
memory operands.  However the logic is generic and applies to several
more instructions such as MOVSXD (aka movslq), one-byte shift
instructions, multiplications, XLAT, and indirect calls/jumps.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: rename zext0/zext2 and make them closer to the manual

X86_SPECIAL_ZExtOp0 and X86_SPECIAL_ZExtOp2 are poorly named; they are a hack
that is needed by scalar insertion and extraction instructions, and not really
related to zero extension: for PEXTR the zero extension is done by the generation
functions, for PINSR the high bits are not used at all and in fact are *not*
filled with zeroes when loaded into s->T1.

Rename the values to match the effect described in the manual, and explain
better in the comments.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: avoid trunc and ext for MULX and RORX

Use _tl operations for 32-bit operands on 32-bit targets, and only go
through trunc and extu ops for 64-bit targets.  While the trunc/ext
ops should be pretty much free after optimization, the optimizer also
does not like having the same temporary used in multiple EBBs.
Therefore it is nicer to not use tmpN* unless necessary.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: reimplement check for validity of LOCK prefix

The previous check erroneously allowed CMP to be modified with LOCK.
Instead, tag explicitly the instructions that do support LOCK.

Acked-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: document more deviations from the manual

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: clean up cpu_cc_compute_all

cpu_cc_compute_all() has an argument that is always equal to CC_OP for historical
reasons (dating back to commit a7812ae4123, "TCG variable type checking.", 2008-11-17,
which added the argument to helper_cc_compute_all).  It does not make sense for the
argument to have any other value, so remove it and clean up some lines that are not
too long anymore.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: remove unnecessary truncations

gen_lea_v_seg (called by gen_add_A0_ds_seg) already zeroes any
bits of s->A0 beyond s->aflag.  It does so before summing the
segment base and, if not in 64-bit mode, also after summing it.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: remove unnecessary arguments from raise_interrupt

is_int is always 1, and error_code is always zero.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: speedup JO/SETO after MUL or IMUL

OF is equal to the carry flag, so use the same CCPrepare.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

target/i386: optimize computation of JL and JLE from flags

Take advantage of the fact that there can be no 1 bits between SF and OF.
If they were adjacent, you could sum SF and get a carry only if SF was
already set.  Then the value of OF in the sum is the XOR of OF itself,
the carry (which is SF) and 0 (the value of the OF bit in the addend):
this is OF^SF exactly.

Because OF and SF are not adjacent, just place more 1 bits to the
left so that the carry propagates, which means summing CC_O - CC_S.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

configure: use a native non-cross compiler for linux-user

Commit c2118e9e1ab ("configure: don't try a "native" cross for linux-user",
2023-11-23) sought to avoid issues with using the native compiler with a
cross-endian or cross-bitness setup.  However, in doing so it ended up
requiring a cross compiler setup (and most likely a slow compiler setup)
even when building TCG tests that are native to the host architecture.
Always allow the host compiler in that case.

Cc: qemu-stable@nongnu.org
Fixes: c2118e9e1ab ("configure: don't try a "native" cross for linux-user", 2023-11-23)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Merge tag 'pull-request-2023-12-20' of https://gitlab.com/thuth/qemu into staging

* Add compat machines for QEMU 9.0
* Some header clean-ups by Philippe
* Restrict type names to alphanumerical range (and a few special characters)
* Fix analyze-migration.py script on s390x
* Clean up and improve some tests
* Document handling of commas in CLI options parameters

# -----BEGIN PGP SIGNATURE-----
#
# iQJFBAABCAAvFiEEJ7iIR+7gJQEY8+q5LtnXdP5wLbUFAmWCtYsRHHRodXRoQHJl
# ZGhhdC5jb20ACgkQLtnXdP5wLbWLnw//cNJrxG0V+j0iakX+C7HRumVrLBDI4KYY
# Cp2Hx92SyeQ0Kk8DJS6JueTV0SLjMsV77APu2YPH7ELmPlk+CB9gqmV7xVoYNvsm
# QbRPlIjFw8MHLekadc2A+C+pn48tWACoOdBEDIfazKrxybnf0B57RC/fIfMKHjbs
# 2ALCoFbbgphs7yWuzTHK8ayKaGMhUVkWfzHQwpnq899olHyZBhkl951uKJA6VmLx
# KvggePkpszLjmmXA8MH1hDCcizki31cB0ZKTbQFCyE42s2S3Hvg0GueU90O7Y1cj
# lS5tPVQxyEhUYMLL+/hudlf2OYqVn2BalB7ieUQIy6rG8yoc9zxfIKQi0ccl+2oA
# s8HRq5S0bSjtilQogU1LQL/Gk6W1/N9MmnhKvCGB+BTK5KX7s4EQk02y9gGZm/8s
# pMErMyaXTG4dLiTAK42VgMVDqCYvzBmE+Gj91OmoUR7fb+VMrsWxeBFxMPDn+VtL
# TMJegIFsjw2QCSitcU4v+nP0qtKgXGbuZtrGXKabrxH5PmeQFJDSM7TwpTK4qvjK
# QMIQKBbz8BfJnUzN8qAaaJEpp1T5tcMJClKtfcgxq/+VyaSaHLmD0cljqBC+g+y7
# FTo+fa7oYx44sAlqapdEXBSGn4T+J26iuCef13CCCiPfYBv/tk3b2E0AWHj4y58I
# +VpInjUaPBQ=
# =TA1/
# -----END PGP SIGNATURE-----
# gpg: Signature made Wed 20 Dec 2023 04:36:11 EST
# gpg:                using RSA key 27B88847EEE0250118F3EAB92ED9D774FE702DB5
# gpg:                issuer "thuth@redhat.com"
# gpg: Good signature from "Thomas Huth <th.huth@gmx.de>" [full]
# gpg:                 aka "Thomas Huth <thuth@redhat.com>" [full]
# gpg:                 aka "Thomas Huth <huth@tuxfamily.org>" [full]
# gpg:                 aka "Thomas Huth <th.huth@posteo.de>" [unknown]
# Primary key fingerprint: 27B8 8847 EEE0 2501 18F3  EAB9 2ED9 D774 FE70 2DB5

* tag 'pull-request-2023-12-20' of https://gitlab.com/thuth/qemu:
  tests/unit/test-qmp-event: Replace fixture by global variables
  tests/unit/test-qmp-event: Simplify event emission check
  tests/unit/test-qmp-event: Drop superfluous mutex
  tests/qtest/npcm7xx_pwm-test: Only do full testing in slow mode
  qemu-options: Clarify handling of commas in options parameters
  tests/qtest/migration-test: Fix analyze-migration.py for s390x
  qom/object: Limit type names to alphanumerical and some few special characters
  tests/unit/test-io-task: Rename "qemu:dummy" to avoid colon in the name
  memory: Remove "qemu:" prefix from the "qemu:ram-discard-manager" type name
  hw: Replace anti-social QOM type names (again)
  docs/system/arm: Fix for rename of type "xlnx.bbram-ctrl"
  target: Restrict 'sysemu/reset.h' to system emulation
  hw/s390x/ipl: Remove unused 'exec/exec-all.h' included header
  hw/misc/mips_itu: Remove unnecessary 'exec/exec-all.h' header
  hw/ppc/spapr_hcall: Remove unused 'exec/exec-all.h' included header
  system/qtest: Restrict QTest API to system emulation
  system/qtest: Include missing 'hw/core/cpu.h' header
  MAINTAINERS: Add some more vmware-related files to the corresponding section
  hw: Add compat machines for 9.0

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

Merge tag 'pull-target-arm-20231219' of https://git.linaro.org/people/pmaydell/qemu-arm into staging

target-arm queue:
 * arm/kvm: drop the split between "common KVM support" and
   "64-bit KVM support", since 32-bit Arm KVM no longer exists
 * arm/kvm: clean up APIs to be consistent about CPU arguments
 * Don't implement *32_EL2 registers when EL1 is AArch64 only
 * Restrict DC CVAP & DC CVADP instructions to TCG accel
 * Restrict TCG specific helpers
 * Propagate MDCR_EL2.HPMN into PMCR_EL0.N
 * Include missing 'exec/exec-all.h' header
 * fsl-imx: add simple RTC emulation for i.MX6 and i.MX7 boards

# -----BEGIN PGP SIGNATURE-----
#
# iQJNBAABCAA3FiEE4aXFk81BneKOgxXPPCUl7RQ2DN4FAmWB6o0ZHHBldGVyLm1h
# eWRlbGxAbGluYXJvLm9yZwAKCRA8JSXtFDYM3mxMEACRpRxJ81pLs8fFYC5BgRhU
# BCxr+ZqarBygzsH9YWUN2TFFKlEZi7mLu6lzFsfN/qEmYCg8VslPbulQHqcGkx51
# kVxXFp/KuGlKt4zGRagZUJxgYAwwU5mnK6dTZT5/ZF6yWX67dXn8V7MP9lqqEPw5
# 5gut7Mu4f7MiAQbwZY1CWP+iu5uZmdsBuKxA6zkxOWJh/A1SfaqQRO6xVQttLAxS
# DPMTpQGmwPS4I+3gGNnqlSu6etp2tdy2K0cW3fhMp6hx70uNMHmFNzRhT/6TaKka
# 9AqXQsFHQiFXDGAm6PmCvfQI6KpLljDyNL/TuUkQWi72bGEHjUsJAdG0aXVOa30W
# uC7vuJkdZrP/t5P1AkZhWQUrlawDRV2YHNDD+gY4fxJL/STkGyU6M8R1nm1J+InN
# n0SeK0VHRC6DRPXCMQhC5QwKUH6ZjFZRs/r2opTu9p+ThQAQRmZBiVfdISCDMYnN
# DCiSb78gIFaUkwtiP44qq8MJQjsHnXtTD1Akqyo2fXSKs66jDK9Gnc8gENYdpghe
# 7V36bOp6scROHOB2a/r8gT42RKzSN6uh6xByaaToza63/bPgvHnn8vvQQbB01AgX
# zJC1xs3dwY8JMyqDefda0K0NDPS8TzNsXYmgxxxcQJpUvB4VVjet9VIMF3T+d8HO
# Pas41Z1gsQY+rcaRk/9mPA==
# =GWIA
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 19 Dec 2023 14:10:05 EST
# gpg:                using RSA key E1A5C593CD419DE28E8315CF3C2525ED14360CDE
# gpg:                issuer "peter.maydell@linaro.org"
# gpg: Good signature from "Peter Maydell <peter.maydell@linaro.org>" [full]
# gpg:                 aka "Peter Maydell <pmaydell@gmail.com>" [full]
# gpg:                 aka "Peter Maydell <pmaydell@chiark.greenend.org.uk>" [full]
# gpg:                 aka "Peter Maydell <peter@archaic.org.uk>" [unknown]
# Primary key fingerprint: E1A5 C593 CD41 9DE2 8E83  15CF 3C25 25ED 1436 0CDE

* tag 'pull-target-arm-20231219' of https://git.linaro.org/people/pmaydell/qemu-arm: (43 commits)
  fsl-imx: add simple RTC emulation for i.MX6 and i.MX7 boards
  target/arm/helper: Propagate MDCR_EL2.HPMN into PMCR_EL0.N
  target/arm/tcg: Including missing 'exec/exec-all.h' header
  target/arm: Restrict DC CVAP & DC CVADP instructions to TCG accel
  target/arm: Restrict TCG specific helpers
  target/arm: Don't implement *32_EL2 registers when EL1 is AArch64 only
  target/arm/kvm: Have kvm_arm_hw_debug_active take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_handle_debug take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_handle_dabt_nisv take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_verify_ext_dabt_pending take a ARMCPU arg
  target/arm/kvm: Have kvm_arm_[get|put]_virtual_time take ARMCPU argument
  target/arm/kvm: Have kvm_arm_vcpu_finalize take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_vcpu_init take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_pmu_set_irq take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_pmu_init take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_pvtime_init take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_set_device_attr take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_sve_get_vls take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_sve_set_vls take a ARMCPU argument
  target/arm/kvm: Have kvm_arm_add_vcpu_properties take a ARMCPU argument
  ...

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

Merge tag 'pull-vfio-20231219' of https://github.com/legoater/qemu into staging

vfio queue:

* Introduce an IOMMU interface backend for VFIO devices
* Convert IOMMU type1 and sPAPR IOMMU to respective backends
* Introduce a new IOMMUFD backend for ARM, x86_64 and s390x platforms

# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEEoPZlSPBIlev+awtgUaNDx8/77KEFAmWB34AACgkQUaNDx8/7
# 7KGOMxAAqXegvAneHqIlu4c8TzTuUR2rkYgev9RdfIHRDuY2XtaX14xlWn/rpTXZ
# qSgeta+iT8Cv4YV1POJeHWFDNs9E29p1w+R7nLcH1qTIIaZHtxwbVVQ3s7kAo1Vb
# 1S1G0/zIznzGVI50a0lj1gO2yQJnu/79nXpnICgA5REW0CscMssnvboQODlwq17V
# ZLNVM8CSAvKl6ppkmzRdfNXCfq6x7bf4MsvnuXsqda4TBbvyyTjAqdo/8sjKiGly
# gSDQqhgy6cvEXIF0UUHPJzFApf0YdXUDlL8hzH90hvRVu4W/t24dPmT7UkVIX9Ek
# TA7RVxv7iJlHtFDqfSTAJFr7nKO9Tm2V9N7xbD1OJUKrMoPZRT6+0R1hMKqsZ5z+
# nG6khqHGzuo/aI9n70YxYIPXt+vs/EHI4WUtslGLUTL0xv8lUzk6cxyIJupFRmDS
# ix6GM9TXOV8RyOveL2knHVymlFnAR6dekkMB+6ljUTuzDwG0oco4vno8z9bi7Vct
# j36bM56U3lhY+w+Ljoy0gPwgrw/FROnGG3mp1mwp1KRHqtEDnUQu8CaLbJOBsBGE
# JJDP6AKAYMczdmYVkd4CvE0WaeSxtOUxW5H5NCPjtaFQt0qEcght2lA2K15g521q
# jeojoJ/QK5949jnNCqm1Z66/YQVL79lPyL0E+mxEohwu+yTORk4=
# =U0x5
# -----END PGP SIGNATURE-----
# gpg: Signature made Tue 19 Dec 2023 13:22:56 EST
# gpg:                using RSA key A0F66548F04895EBFE6B0B6051A343C7CFFBECA1
# gpg: Good signature from "Cédric Le Goater <clg@redhat.com>" [unknown]
# gpg:                 aka "Cédric Le Goater <clg@kaod.org>" [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: A0F6 6548 F048 95EB FE6B  0B60 51A3 43C7 CFFB ECA1

* tag 'pull-vfio-20231219' of https://github.com/legoater/qemu: (47 commits)
  hw/ppc/Kconfig: Imply VFIO_PCI
  docs/devel: Add VFIO iommufd backend documentation
  vfio: Introduce a helper function to initialize VFIODevice
  vfio/ccw: Move VFIODevice initializations in vfio_ccw_instance_init
  vfio/ap: Move VFIODevice initializations in vfio_ap_instance_init
  vfio/platform: Move VFIODevice initializations in vfio_platform_instance_init
  vfio/pci: Move VFIODevice initializations in vfio_instance_init
  hw/i386: Activate IOMMUFD for q35 machines
  kconfig: Activate IOMMUFD for s390x machines
  hw/arm: Activate IOMMUFD for virt machines
  vfio: Make VFIOContainerBase poiner parameter const in VFIOIOMMUOps callbacks
  vfio/ccw: Make vfio cdev pre-openable by passing a file handle
  vfio/ccw: Allow the selection of a given iommu backend
  vfio/ap: Make vfio cdev pre-openable by passing a file handle
  vfio/ap: Allow the selection of a given iommu backend
  vfio/platform: Make vfio cdev pre-openable by passing a file handle
  vfio/platform: Allow the selection of a given iommu backend
  vfio/pci: Make vfio cdev pre-openable by passing a file handle
  vfio/pci: Allow the selection of a given iommu backend
  vfio/iommufd: Enable pci hot reset through iommufd cdev interface
  ...

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

tests/unit/test-qmp-event: Replace fixture by global variables

The fixture buys us exactly nothing, as we need a global variable
anyway, for test_qapi_event_emit().  Drop it.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20231122072456.2518816-4-armbru@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/unit/test-qmp-event: Simplify event emission check

The generated qapi_event_send_FOO() call an event emitter function.
It's test_qapi_event_emit() in this test.  It compares the actual
event to the expected event, and sets a flag to record it was called.
The test functions set expected data and clear the flag before calling
qapi_event_send_FOO(), and check the flag afterwards.

Make test_qapi_event_emit() consume expected data, and the test
functions check it was consumed.  Delete the flag.  This is simpler.
It also catches extraneous calls of test_qapi_event_emit().  Catching
that is not worthwhile, but since the cost is negative...

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20231122072456.2518816-3-armbru@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/unit/test-qmp-event: Drop superfluous mutex

Mutex @test_event_lock is held from fixture setup to teardown,
protecting global variable @test_event_data.  But tests always run one
after the other, so this is superfluous.  It also confuses Coverity.
Drop the mutex.

Fixes: CID 1527425
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-ID: <20231122072456.2518816-2-armbru@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/qtest/npcm7xx_pwm-test: Only do full testing in slow mode

The npcm7xx_pwm-test can take quite a while when running with
--enable-debug on a loaded system. The tests here are quite
repetitive - by default it should be fine if we only execute
some of them and only execute all when running in slow testing mode.

Message-ID: <20231215143524.49241-1-thuth@redhat.com>
Reviewed-by: "Daniel P. Berrangé" <berrange@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

qemu-options: Clarify handling of commas in options parameters

Provide explicit guidance on dealing with option parameters as arbitrary
strings containing commas, such as in "file=my,file" and "string=a,b". The
updated documentation emphasizes the need to double commas when they
appear within such parameters.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1839
Signed-off-by: Yihuan Pan <xun794@gmail.com>
Message-ID: <20231213141706.629833-2-xun794@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/qtest/migration-test: Fix analyze-migration.py for s390x

The migration stream on s390x contains data for the storage_attributes
which the analyze-migration.py cannot handle yet. Add the basic code
for handling this, so we can re-enable the check in the migration-test.

Message-ID: <20231120113951.162090-1-thuth@redhat.com>
Reviewed-by: Fabiano Rosas <farosas@suse.de>
Signed-off-by: Thomas Huth <thuth@redhat.com>

qom/object: Limit type names to alphanumerical and some few special characters

QOM names currently don't have any enforced naming rules. This
can be problematic, e.g. when they are used on the command line
for the "-device" option (where the comma is used to separate
properties). To avoid that such problematic type names come in
again, let's restrict the set of acceptable characters during the
type registration.

Ideally, we'd apply here the same rules as for QAPI, i.e. all type
names should begin with a letter, and contain only ASCII letters,
digits, hyphen, and underscore. However, we already have so many
pre-existing types like:

    486-x86_64-cpu
    cfi.pflash01
    power5+_v2.1-spapr-cpu-core
    virt-2.6-machine
    pc-i440fx-3.0-machine

... so that we have to allow "." and "+" for now, too. While the
dot is used in a lot of places, the "+" can fortunately be limited
to two classes of legacy names ("power" and "Sun-UltraSparc" CPUs).

We also cannot enforce the rule that names must start with a letter
yet, since there are lot of types that start with a digit. Still,
at least limiting the first characters to the alphanumerical range
should be way better than nothing.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231117114457.177308-6-thuth@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/unit/test-io-task: Rename "qemu:dummy" to avoid colon in the name

Type names should not contain special characters like ":" (so that
they are easier to use with QAPI and other parts). We are going to
forbid such names in an upcoming patch. Thus let's replace the ":"
here with a "-".

Reviewed-by: "Daniel P. Berrangé" <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231117114457.177308-5-thuth@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

memory: Remove "qemu:" prefix from the "qemu:ram-discard-manager" type name

Type names should not contain special characters like ":". Let's
remove the whole prefix here since it does not really seem to be
helpful to have such a prefix here. The type name is only used
internally for an interface type, so the renaming should not affect
the user interface or migration.

Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231117114457.177308-4-thuth@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

hw: Replace anti-social QOM type names (again)

QOM type names containing ',' result in awful UI.  We got rid of them
in v6.0.0 (commit e178113ff64 hw: Replace anti-social QOM type names).
A few have crept back since:

    xlnx,cframe-reg
    xlnx,efuse
    xlnx,pmc-efuse-cache
    xlnx,versal-cfu-apb
    xlnx,versal-cfu-fdro
    xlnx,versal-cfu-sfr
    xlnx,versal-crl
    xlnx,versal-efuse
    xlnx,zynqmp-efuse

These are all device types.  They can't be plugged with -device /
device_add, except for "xlnx,efuse" (I'm not sure that one is
intentional).

They *can* be used with -device / device_add to request help.
Usability is poor, though: you have to double the comma, like this:

    $ qemu-system-aarch64 -device xlnx,,pmc-efuse-cache,help

They can also be used with -global, where you must *not* double the
comma:

    $ qemu-system-aarch64 -global xlnx,efuse.drive-index=2

Trap for the unwary.

"xlnx,efuse", "xlnx,versal-efuse", "xlnx,pmc-efuse-cache",
"xlnx-zynqmp-efuse" are from v6.2.0, "xlnx,versal-crl" is from v7.1.0,
and the remainder are new.

Rename them all to "xlnx-FOO", like commit e178113ff64 did.

Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Francisco Iglesias <francisco.iglesias@amd.com>
Message-ID: <20231117114457.177308-3-thuth@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

docs/system/arm: Fix for rename of type "xlnx.bbram-ctrl"

Fixes: b65b4b7ae3c8 (xlnx-bbram: hw/nvram: Use dot in device type name)
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[thuth: Use longhand syntax to avoid problems with the "." in the name]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-ID: <20231117114457.177308-2-thuth@redhat.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

target: Restrict 'sysemu/reset.h' to system emulation

vCPU "reset" is only possible with system emulation.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Warner Losh <imp@bsdimp.com>
Reviewed-by: Song Gao <gaosong@loongson.cn>
Message-ID: <20231212113640.30287-5-philmd@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Zhao Liu <zhao1.liu@intel.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

hw/s390x/ipl: Remove unused 'exec/exec-all.h' included header

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231212113640.30287-4-philmd@linaro.org>
Reviewed-by: Eric Farman <farman@linux.ibm.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

hw/misc/mips_itu: Remove unnecessary 'exec/exec-all.h' header

mips_itu.c only requires declarations from "hw/core/cpu.h"
and "cpu.h". Avoid including the huge "exec/exec-all.h" header.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231212113640.30287-3-philmd@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

hw/ppc/spapr_hcall: Remove unused 'exec/exec-all.h' included header

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231212113640.30287-2-philmd@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>

system/qtest: Restrict QTest API to system emulation

Outside of system emulation, only qtest_enabled() can be used.

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231212113016.29808-3-philmd@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

system/qtest: Include missing 'hw/core/cpu.h' header

"hw/core/cpu.h" declares 'first_cpu'. Include it to avoid
when unrelated headers are refactored:

  system/qtest.c:548:33: error: use of undeclared identifier 'first_cpu'
            address_space_write(first_cpu->as, addr, MEMTXATTRS_UNSPECIFIED,
                                ^

Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20231212113016.29808-2-philmd@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

MAINTAINERS: Add some more vmware-related files to the corresponding section

These files are obviously related to Vmware emulation, so let's list
them in the corresponding section in the MAINTAINERS file.

Message-ID: <20231107102738.14797-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Thomas Huth <thuth@redhat.com>

hw: Add compat machines for 9.0

Add 9.0 machine types for arm/i440fx/m68k/q35/s390x/spapr.

Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Message-ID: <20231120094259.1191804-1-cohuck@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Reviewed-by: Gavin Shan <gshan@redhat.com>
Acked-by: Eric Farman <farman@linux.ibm.com> # s390x
Signed-off-by: Thomas Huth <thuth@redhat.com>

hw/ppc/Kconfig: Imply VFIO_PCI

When the legacy and iommufd backends were introduced, a set of common
vfio-pci routines were exported in pci.c for both backends to use :

  vfio_pci_pre_reset
  vfio_pci_get_pci_hot_reset_info
  vfio_pci_host_match
  vfio_pci_post_reset

This introduced a build failure on PPC when --without-default-devices
is use because VFIO is always selected in ppc/Kconfig but VFIO_PCI is
not.

Use an 'imply VFIO_PCI' in ppc/Kconfig and bypass compilation of the
VFIO EEH hooks routines defined in hw/ppc/spapr_pci_vfio.c with
CONFIG_VFIO_PCI.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

docs/devel: Add VFIO iommufd backend documentation

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio: Introduce a helper function to initialize VFIODevice

Introduce a helper function to replace the common code to initialize
VFIODevice in pci, platform, ap and ccw VFIO device.

No functional change intended.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/ccw: Move VFIODevice initializations in vfio_ccw_instance_init

Some of the VFIODevice initializations is in vfio_ccw_realize,
move all of them in vfio_ccw_instance_init.

No functional change intended.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Eric Farman <farman@linux.ibm.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/ap: Move VFIODevice initializations in vfio_ap_instance_init

Some of the VFIODevice initializations is in vfio_ap_realize,
move all of them in vfio_ap_instance_init.

No functional change intended.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Eric Farman <farman@linux.ibm.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/platform: Move VFIODevice initializations in vfio_platform_instance_init

Some of the VFIODevice initializations is in vfio_platform_realize,
move all of them in vfio_platform_instance_init.

No functional change intended.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/pci: Move VFIODevice initializations in vfio_instance_init

Some of the VFIODevice initializations is in vfio_realize,
move all of them in vfio_instance_init.

No functional change intended.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

hw/i386: Activate IOMMUFD for q35 machines

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

kconfig: Activate IOMMUFD for s390x machines

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Eric Farman <farman@linux.ibm.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

hw/arm: Activate IOMMUFD for virt machines

Signed-off-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio: Make VFIOContainerBase poiner parameter const in VFIOIOMMUOps callbacks

Some of the callbacks in VFIOIOMMUOps pass VFIOContainerBase poiner,
those callbacks only need read access to the sub object of VFIOContainerBase.
So make VFIOContainerBase, VFIOContainer and VFIOIOMMUFDContainer as const
in these callbacks.

Local functions called by those callbacks also need same changes to avoid
build error.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/ccw: Make vfio cdev pre-openable by passing a file handle

This gives management tools like libvirt a chance to open the vfio
cdev with privilege and pass FD to qemu. This way qemu never needs
to have privilege to open a VFIO or iommu cdev node.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Farman <farman@linux.ibm.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/ccw: Allow the selection of a given iommu backend

Now we support two types of iommu backends, let's add the capability
to select one of them. This depends on whether an iommufd object has
been linked with the vfio-ccw device:

If the user wants to use the legacy backend, it shall not
link the vfio-ccw device with any iommufd object:

 -device vfio-ccw,sysfsdev=/sys/bus/mdev/devices/XXX

This is called the legacy mode/backend.

If the user wants to use the iommufd backend (/dev/iommu) it
shall pass an iommufd object id in the vfio-ccw device options:

 -object iommufd,id=iommufd0
 -device vfio-ccw,sysfsdev=/sys/bus/mdev/devices/XXX,iommufd=iommufd0

Suggested-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Farman <farman@linux.ibm.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/ap: Make vfio cdev pre-openable by passing a file handle

This gives management tools like libvirt a chance to open the vfio
cdev with privilege and pass FD to qemu. This way qemu never needs
to have privilege to open a VFIO or iommu cdev node.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/ap: Allow the selection of a given iommu backend

Now we support two types of iommu backends, let's add the capability
to select one of them. This depends on whether an iommufd object has
been linked with the vfio-ap device:

if the user wants to use the legacy backend, it shall not
link the vfio-ap device with any iommufd object:

 -device vfio-ap,sysfsdev=/sys/bus/mdev/devices/XXX

This is called the legacy mode/backend.

If the user wants to use the iommufd backend (/dev/iommu) it
shall pass an iommufd object id in the vfio-ap device options:

 -object iommufd,id=iommufd0
 -device vfio-ap,sysfsdev=/sys/bus/mdev/devices/XXX,iommufd=iommufd0

Suggested-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/platform: Make vfio cdev pre-openable by passing a file handle

This gives management tools like libvirt a chance to open the vfio
cdev with privilege and pass FD to qemu. This way qemu never needs
to have privilege to open a VFIO or iommu cdev node.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/platform: Allow the selection of a given iommu backend

Now we support two types of iommu backends, let's add the capability
to select one of them. This depends on whether an iommufd object has
been linked with the vfio-platform device:

If the user wants to use the legacy backend, it shall not
link the vfio-platform device with any iommufd object:

 -device vfio-platform,host=XXX

This is called the legacy mode/backend.

If the user wants to use the iommufd backend (/dev/iommu) it
shall pass an iommufd object id in the vfio-platform device options:

 -object iommufd,id=iommufd0
 -device vfio-platform,host=XXX,iommufd=iommufd0

Suggested-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/pci: Make vfio cdev pre-openable by passing a file handle

This gives management tools like libvirt a chance to open the vfio
cdev with privilege and pass FD to qemu. This way qemu never needs
to have privilege to open a VFIO or iommu cdev node.

Together with the earlier support of pre-opening /dev/iommu device,
now we have full support of passing a vfio device to unprivileged
qemu by management tool. This mode is no more considered for the
legacy backend. So let's remove the "TODO" comment.

Add helper functions vfio_device_set_fd() and vfio_device_get_name()
to set fd and get device name, they will also be used by other vfio
devices.

There is no easy way to check if a device is mdev with FD passing,
so fail the x-balloon-allowed check unconditionally in this case.

There is also no easy way to get BDF as name with FD passing, so
we fake a name by VFIO_FD[fd].

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/pci: Allow the selection of a given iommu backend

Now we support two types of iommu backends, let's add the capability
to select one of them. This depends on whether an iommufd object has
been linked with the vfio-pci device:

If the user wants to use the legacy backend, it shall not
link the vfio-pci device with any iommufd object:

 -device vfio-pci,host=0000:02:00.0

This is called the legacy mode/backend.

If the user wants to use the iommufd backend (/dev/iommu) it
shall pass an iommufd object id in the vfio-pci device options:

 -object iommufd,id=iommufd0
 -device vfio-pci,host=0000:02:00.0,iommufd=iommufd0

Suggested-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/iommufd: Enable pci hot reset through iommufd cdev interface

Implement the newly introduced pci_hot_reset callback named
iommufd_cdev_pci_hot_reset to do iommufd specific check and
reset operation.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/pci: Introduce a vfio pci hot reset interface

Legacy vfio pci and iommufd cdev have different process to hot reset
vfio device, expand current code to abstract out pci_hot_reset callback
for legacy vfio, this same interface will also be used by iommufd
cdev vfio device.

Rename vfio_pci_hot_reset to vfio_legacy_pci_hot_reset and move it
into container.c.

vfio_pci_[pre/post]_reset and vfio_pci_host_match are exported so
they could be called in legacy and iommufd pci_hot_reset callback.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/pci: Extract out a helper vfio_pci_get_pci_hot_reset_info

This helper will be used by both legacy and iommufd backends.

No functional changes intended.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/iommufd: Add support for iova_ranges and pgsizes

Some vIOMMU such as virtio-iommu use IOVA ranges from host side to
setup reserved ranges for passthrough device, so that guest will not
use an IOVA range beyond host support.

Use an uAPI of IOMMUFD to get IOVA ranges of host side and pass to
vIOMMU just like the legacy backend, if this fails, fallback to
64bit IOVA range.

Also use out_iova_alignment returned from uAPI as pgsizes instead of
qemu_real_host_page_size() as a fallback.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/iommufd: Relax assert check for iommufd backend

Currently iommufd doesn't support dirty page sync yet,
but it will not block us doing live migration if VFIO
migration is force enabled.

So in this case we allow set_dirty_page_tracking to be NULL.
Note we don't need same change for query_dirty_bitmap because
when dirty page sync isn't supported, query_dirty_bitmap will
never be called.

Suggested-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/iommufd: Implement the iommufd backend

The iommufd backend is implemented based on the new /dev/iommu user API.
This backend obviously depends on CONFIG_IOMMUFD.

So far, the iommufd backend doesn't support dirty page sync yet.

Co-authored-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/common: return early if space isn't empty

This is a trivial optimization. If there is active container in space,
vfio_reset_handler will never be unregistered. So revert the check of
space->containers and return early.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

util/char_dev: Add open_cdev()

/dev/vfio/devices/vfioX may not exist. In that case it is still possible
to open /dev/char/$major:$minor instead. Add helper function to abstract
the cdev open.

Suggested-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

backends/iommufd: Introduce the iommufd object

Introduce an iommufd object which allows the interaction
with the host /dev/iommu device.

The /dev/iommu can have been already pre-opened outside of qemu,
in which case the fd can be passed directly along with the
iommufd object:

This allows the iommufd object to be shared accross several
subsystems (VFIO, VDPA, ...). For example, libvirt would open
the /dev/iommu once.

If no fd is passed along with the iommufd object, the /dev/iommu
is opened by the qemu code.

Suggested-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Tested-by: Eric Auger <eric.auger@redhat.com>
Tested-by: Nicolin Chen <nicolinc@nvidia.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/spapr: Move hostwin_list into spapr container

No functional changes intended.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/spapr: Move prereg_listener into spapr container

No functional changes intended.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/spapr: switch to spapr IOMMU BE add/del_section_window

No functional change intended.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/spapr: Introduce spapr backend and target interface

Introduce an empty spapr backend which will hold spapr specific
content, currently only prereg_listener and hostwin_list.

Also introduce two spapr specific callbacks add/del_window into
VFIOIOMMUOps. Instantiate a spapr ops with a helper setup_spapr_ops
and assign it to bcontainer->ops.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/container: Implement attach/detach_device

No functional change intended.

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Yi Sun <yi.y.sun@linux.intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/container: Move iova_ranges to base container

Meanwhile remove the helper function vfio_free_container as it
only calls g_free now.

No functional change intended.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/container: Move dirty_pgsizes and max_dirty_bitmap_size to base container

No functional change intended.

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Yi Sun <yi.y.sun@linux.intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/container: Move listener to base container

Move listener to base container. Also error and initialized fields
are moved at the same time.

No functional change intended.

Signed-off-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Yi Liu <yi.l.liu@intel.com>
Signed-off-by: Yi Sun <yi.y.sun@linux.intel.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>

vfio/container: Move vrdl_list to base container

No functional change intended.

Signed-off-by: Zhenzhong Duan <zhenzhong.duan@intel.com>
Reviewed-by: Cédric Le Goater <clg@redhat.com>
Signed-off-by: Cédric Le Goater <clg@redhat.com>