hvf: Introduce hvf vcpu struct

We will need more than a single field for hvf going forward. To keep
the global vcpu struct uncluttered, let's allocate a special hvf vcpu
struct, similar to how hax does it.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
Tested-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-12-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Remove hvf-accel-ops.h

We can move the definition of hvf_vcpu_exec() into our internal
hvf header, obsoleting the need for hvf-accel-ops.h.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-11-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Make synchronize functions static

The hvf accel synchronize functions are only used as input for local
callback functions, so we can make them static.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-10-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Use cpu_synchronize_state()

There is no reason to call the hvf specific hvf_cpu_synchronize_state()
when we can just use the generic cpu_synchronize_state() instead. This
allows us to have less dependency on internal function definitions and
allows us to make hvf_cpu_synchronize_state() static.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-9-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Split out common code on vcpu init and destroy

Until now, Hypervisor.framework has only been available on x86_64 systems.
With Apple Silicon shipping now, it extends its reach to aarch64. To
prepare for support for multiple architectures, let's start moving common
code out into its own accel directory.

This patch splits the vcpu init and destroy functions into a generic and
an architecture specific portion. This also allows us to move the generic
functions into the generic hvf code, removing exported functions.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-8-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Remove use of hv_uvaddr_t and hv_gpaddr_t

The ARM version of Hypervisor.framework no longer defines these two
types, so let's just revert to standard ones.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-7-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Make hvf_set_phys_mem() static

The hvf_set_phys_mem() function is only called within the same file.
Make it static.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-6-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Move hvf internal definitions into common header

Until now, Hypervisor.framework has only been available on x86_64 systems.
With Apple Silicon shipping now, it extends its reach to aarch64. To
prepare for support for multiple architectures, let's start moving common
code out into its own accel directory.

This patch moves a few internal struct and constant defines over.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-5-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Move cpu functions into common directory

Until now, Hypervisor.framework has only been available on x86_64 systems.
With Apple Silicon shipping now, it extends its reach to aarch64. To
prepare for support for multiple architectures, let's start moving common
code out into its own accel directory.

This patch moves CPU and memory operations over. While at it, make sure
the code is consumable on non-i386 systems.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-4-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Move vcpu thread functions into common directory

Until now, Hypervisor.framework has only been available on x86_64 systems.
With Apple Silicon shipping now, it extends its reach to aarch64. To
prepare for support for multiple architectures, let's start moving common
code out into its own accel directory.

This patch moves the vCPU thread loop over.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-3-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hvf: Move assert_hvf_ok() into common directory

Until now, Hypervisor.framework has only been available on x86_64 systems.
With Apple Silicon shipping now, it extends its reach to aarch64. To
prepare for support for multiple architectures, let's start moving common
code out into its own accel directory.

This patch moves assert_hvf_ok() and introduces generic build infrastructure.

Signed-off-by: Alexander Graf <agraf@csgraf.de>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20210519202253.76782-2-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Enable BFloat16 extensions

Disable BF16 again for !have_neon and !have_vfp during realize.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-13-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

linux-user/aarch64: Enable hwcap bits for bfloat16

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-12-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement bfloat widening fma (indexed)

This is BFMLAL{B,T} for both AArch64 AdvSIMD and SVE,
and VFMA{B,T}.BF16 for AArch32 NEON.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-11-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement bfloat widening fma (vector)

This is BFMLAL{B,T} for both AArch64 AdvSIMD and SVE,
and VFMA{B,T}.BF16 for AArch32 NEON.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-10-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement bfloat16 matrix multiply accumulate

This is BFMMLA for both AArch64 AdvSIMD and SVE,
and VMMLA.BF16 for AArch32 NEON.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-9-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement bfloat16 dot product (indexed)

This is BFDOT for both AArch64 AdvSIMD and SVE,
and VDOT.BF16 for AArch32 NEON.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-8-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement bfloat16 dot product (vector)

This is BFDOT for both AArch64 AdvSIMD and SVE,
and VDOT.BF16 for AArch32 NEON.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-7-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

softfpu: Add float_round_to_odd_inf

For Arm BFDOT and BFMMLA, we need a version of round-to-odd
that overflows to infinity, instead of the max normal number.

Cc: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-6-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement vector float32 to bfloat16 conversion

This is BFCVT{N,T} for both AArch64 AdvSIMD and SVE,
and VCVT.BF16.F32 for AArch32 NEON.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-5-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Implement scalar float32 to bfloat16 conversion

This is the 64-bit BFCVT and the 32-bit VCVT{B,T}.BF16.F32.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-4-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Unify unallocated path in disas_fp_1src

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-3-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Add isar_feature_{aa32, aa64, aa64_sve}_bf16

Note that the SVE BFLOAT16 support does not require SVE2,
it is an independent extension.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210525225817.400336-2-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: use raise_exception_ra for stack limit exception

The sequence cpu_restore_state() + raise_exception() is equivalent to
raise_exception_ra(), so use that instead.  (In this case we never
cared about the syndrome value, because M-profile doesn't use the
syndrome; the old code was just written unnecessarily awkwardly.)

Cc: Richard Henderson <richard.henderson@linaro.org>
Cc: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Jamie Iles <jamie@nuviainc.com>
[PMM: Retain edited version of comment; rewrite commit message]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: use raise_exception_ra for MTE check failure

Now that raise_exception_ra restores the state before raising the
exception we can use restore_exception_ra to perform the state restore +
exception raising without clobbering the syndrome.

Cc: Richard Henderson <richard.henderson@linaro.org>
Cc: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Jamie Iles <jamie@nuviainc.com>
[PMM: Keep the one line of the comment that is still relevant]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: fold do_raise_exception into raise_exception

Now that there are no other users of do_raise_exception, fold it into
raise_exception.

Cc: Richard Henderson <richard.henderson@linaro.org>
Cc: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Jamie Iles <jamie@nuviainc.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: fix missing exception class

The DAIF and PAC checks used raise_exception_ra to raise an exception
and unwind CPU state but raise_exception_ra is currently designed for
handling data aborts as the syndrome is partially precomputed and
encoded in the TB and then merged in merge_syn_data_abort when handling
the data abort.  Using raise_exception_ra for DAIF and PAC checks
results in an empty syndrome being retrieved from data[2] in
restore_state_to_opc and setting ESR to 0.  This manifested as:

  kvm [571]: Unknown exception class: esr: 0x000000 –
  Unknown/Uncategorized

when launching a KVM guest when the host qemu used a CPU supporting
EL2+pointer authentication and enabling pointer authentication in the
guest.

Rework raise_exception_ra such that the state is restored before raising
the exception so that the exception is not clobbered by
restore_state_to_opc.

Fixes: 0d43e1a2d29a ("target/arm: Add PAuth helpers")
Cc: Richard Henderson <richard.henderson@linaro.org>
Cc: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Jamie Iles <jamie@nuviainc.com>
[PMM: added comment]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Mark LDS{MIN,MAX} as signed operations

The operands to tcg_gen_atomic_fetch_s{min,max}_i64 must
be signed, so that the inputs are properly extended.
Zero extend the result afterward, as needed.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/364
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20210602020720.47679-1-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

docs: Fix installation of man pages with Sphinx 4.x

The 4.x branch of Sphinx introduces a breaking change, as generated man
pages are now written to subdirectories corresponding to the manual
section they belong to. This results in `make install` erroring out when
attempting to install the man pages, because they are not where it
expects to find them.

This patch restores the behavior of Sphinx 3.x regarding man pages.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/256
Signed-off-by: Damien Goutte-Gattat <dgouttegattat@incenp.org>
Message-id: 20210503161422.15028-1-dgouttegattat@incenp.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

arm: Consistently use "Cortex-Axx", not "Cortex Axx"

The official punctuation for Arm CPU names uses a hyphen, like
"Cortex-A9". We mostly follow this, but in a few places usage
without the hyphen has crept in. Fix those so we consistently
use the same way of writing the CPU name.

This commit was created with:
  git grep -z -l 'Cortex ' | xargs -0 sed -i 's/Cortex /Cortex-/'

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20210527095152.10968-1-peter.maydell@linaro.org

target/arm: Allow board models to specify initial NS VTOR

Currently we allow board models to specify the initial value of the
Secure VTOR register, using an init-svtor property on the TYPE_ARMV7M
object which is plumbed through to the CPU.  Allow board models to
also specify the initial value of the Non-secure VTOR via a similar
init-nsvtor property.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-10-peter.maydell@linaro.org

target/arm: Make FPSCR.LTPSIZE writable for MVE

The M-profile FPSCR has an LTPSIZE field, but if MVE is not
implemented it is read-only and always reads as 4; this is how QEMU
currently handles it.

Make the field writable when MVE is implemented.

We can safely add the field to the MVE migration struct because
currently no CPUs enable MVE and so the migration struct is never
used.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-8-peter.maydell@linaro.org

target/arm: Implement M-profile VPR register

If MVE is implemented for an M-profile CPU then it has a VPR
register, which tracks predication information.

Implement the read and write handling of this register, and
the migration of its state.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-7-peter.maydell@linaro.org

target/arm: Fix return values in fp_sysreg_checks()

The fp_sysreg_checks() function is supposed to be returning an
FPSysRegCheckResult, which is an enum with three possible values.
However, three places in the function "return false" (a hangover from
a previous iteration of the design where the function just returned a
bool).  Make these return FPSysRegCheckFailed instead (for no
functional change, since both false and FPSysRegCheckFailed are
zero).

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-6-peter.maydell@linaro.org

target/arm: Add MVE check to VMOV_reg_sp and VMOV_reg_dp

Split out the handling of VMOV_reg_sp and VMOV_reg_dp so that we can
permit the insns if either FP or MVE are present.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-5-peter.maydell@linaro.org

target/arm: Move fpsp/fpdp isar check into callers of do_vfp_2op_sp/dp

The do_vfp_2op_sp() and do_vfp_2op_dp() functions currently check
whether floating point is supported via the aa32_fpdp_v2 and
aa32_fpsp_v2 isar checks.  For v8.1M MVE support, the VMOV_reg trans
functions (but not any of the others) need to update this to also
allow the insn if MVE is implemented.  Move the check out of the do_
function and into its callsites (which are all implemented via the
DO_VFP_2OP macro), so we have a place to change the check for the
VMOV insns.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-4-peter.maydell@linaro.org

target/arm: Update feature checks for insns which are "MVE or FP"

Some v8M instructions are present if either the floating point
extension or MVE is implemented.  Update our implementation of them
to check for MVE as well as for FP.

This is all the insns which use CheckDecodeFaults(ExtType_MveOrFp) or
CheckDecodeFaults(ExtType_MveOrDpFp) in their pseudocode, which are
essentially the loads and stores, moves and sysreg accesses, except
for VMOV_reg_sp and VMOV_reg_dp, which we handle in subsequent
patches because they need a refactor to provide a place to put the
new MVE check.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-3-peter.maydell@linaro.org

target/arm: Add isar feature check functions for MVE

Add the isar feature check functions we will need for v8.1M MVE:
 * a check for MVE present: this corresponds to the pseudocode's
   CheckDecodeFaults(ExtType_Mve)
 * a check for the optional floating-point part of MVE: this
   corresponds to CheckDecodeFaults(ExtType_MveFp)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210520152840.24453-2-peter.maydell@linaro.org

Merge remote-tracking branch 'remotes/dg-gitlab/tags/ppc-for-6.1-20210603' into staging

ppc patch queue 2021-06-03

Next batch of ppc target patches.  Highlights are:
 * A fix for a regression with single-step mode
 * Start of moving ppc to use decodetree
 * Implementation of some POWER10 64-bit prefixed instructions
 * Several cleanups to softmmu code
 * Continued progress towards allowing --disable-tcg
 * Fix for the POWER PEF implementation
 * Fix for LPCR handling of hotplugged CPUs
 * Assorted other bugfixes and cleanups

This patchset does contain a couple of changes to code outside my
normal scope of maintainership, related to the removal of cpu_dump and
cpu_statistics hooks.  ppc was the last target arch implementing these
at all, and they didn't really do anything there either.  The patches
should have relevant acks.

# gpg: Signature made Thu 03 Jun 2021 09:20:59 BST
# gpg:                using RSA key 75F46586AE61A66CC44E87DC6C38CACA20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>" [full]
# gpg:                 aka "David Gibson (Red Hat) <dgibson@redhat.com>" [full]
# gpg:                 aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>" [full]
# gpg:                 aka "David Gibson (kernel.org) <dwg@kernel.org>" [unknown]
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E  87DC 6C38 CACA 20D9 B392

* remotes/dg-gitlab/tags/ppc-for-6.1-20210603: (42 commits)
  target/ppc: fix single-step exception regression
  target/ppc: Move cmp/cmpi/cmpl/cmpli to decodetree
  target/ppc: Move addpcis to decodetree
  target/ppc: Implement vcfuged instruction
  target/ppc: Implement cfuged instruction
  target/ppc: Implement setbc/setbcr/stnbc/setnbcr instructions
  target/ppc: Implement prefixed integer store instructions
  target/ppc: Move D/DS/X-form integer stores to decodetree
  target/ppc: Implement prefixed integer load instructions
  target/ppc: Move D/DS/X-form integer loads to decodetree
  target/ppc: Implement PNOP
  target/ppc: Move ADDI, ADDIS to decodetree, implement PADDI
  target/ppc: Add infrastructure for prefixed insns
  target/ppc: Move page crossing check to ppc_tr_translate_insn
  target/ppc: Introduce macros to check isa extensions
  target/ppc: powerpc_excp: Consolidade TLB miss code
  target/ppc: powerpc_excp: Remove dump_syscall_vectored
  target/ppc: powerpc_excp: Move lpes code to where it is used
  target/ppc: overhauled and moved logic of storing fpscr
  target/ppc: removed all mentions to PPC_DUMP_CPU
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/ppc: fix single-step exception regression

Commit 6086c75 (target/ppc: Replace POWERPC_EXCP_BRANCH with
DISAS_NORETURN) broke the generation of exceptions when
CPU_SINGLE_STEP or CPU_BRANCH_STEP were set, due to nip always being
reset to the address of the current instruction.
This fix leaves nip untouched when generating the exception.

Signed-off-by: Luis Pires <luis.pires@eldorado.org.br>
Reported-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210602125103.332793-1-luis.pires@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Move cmp/cmpi/cmpl/cmpli to decodetree

Additionally, REQUIRE_64BIT when L=1 to match what is specified in The
Programming Environments Manual:

"For 32-bit implementations, the L field must be cleared, otherwise the
instruction form is invalid."

Some CPUs are known to deviate from this specification by ignoring the
L bit [1]. The stricter behavior, however, can help users that test
software with qemu, making it more likely to detect bugs that would
otherwise be silent.

If deemed necessary, a future patch can adapt this behavior based on
the specific CPU model.

[1] The 601 manual is the only one I've found that explicitly states
that the L bit is ignored, but we also observe this behavior in a 7447A
v1.2.

Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-15-matheus.ferst@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
[dwg: Corrected whitespace error]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Move addpcis to decodetree

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-14-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Implement vcfuged instruction

Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-13-matheus.ferst@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Implement cfuged instruction

Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-12-matheus.ferst@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Implement setbc/setbcr/stnbc/setnbcr instructions

Implements the following PowerISA v3.1 instructions:
setbc: Set Boolean Condition
setbcr: Set Boolean Condition Reverse
setnbc: Set Negative Boolean Condition
setnbcr: Set Negative Boolean Condition Reverse

Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-11-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Implement prefixed integer store instructions

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-10-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Move D/DS/X-form integer stores to decodetree

These are all connected by macros in the legacy decoding.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-9-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Implement prefixed integer load instructions

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-8-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Move D/DS/X-form integer loads to decodetree

These are all connected by macros in the legacy decoding.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-7-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Implement PNOP

The illegal suffix behavior matches what was observed in a
POWER10 DD2.0 machine.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-6-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Move ADDI, ADDIS to decodetree, implement PADDI

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-5-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Add infrastructure for prefixed insns

Signed-off-by: Luis Pires <luis.pires@eldorado.org.br>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-4-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Move page crossing check to ppc_tr_translate_insn

With prefixed instructions, the number of instructions
remaining until the page crossing is no longer constant.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-3-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: Introduce macros to check isa extensions

These will be used by the decodetree trans_* functions
to early-exit when the instruction set is not enabled.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Matheus Ferst <matheus.ferst@eldorado.org.br>
Message-Id: <20210601193528.2533031-2-matheus.ferst@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: powerpc_excp: Consolidade TLB miss code

The only difference in the code for Instruction fetch, Data load and
Data store TLB miss errors is that when called from an unsupported
processor (i.e. not one of 602, 603, 603e, G2, 7x5 or 74xx), they
abort with a message specific to the operation type (insn fetch, data
load/store).

If a processor does not support those interrupts we should not be
registering them in init_excp_<proc> to begin with, so that error
message would never be used.

I'm leaving the message in for completeness, but making it generic and
consolidating the three interrupts into the same case statement body.

Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Message-Id: <20210601214649.785647-4-farosas@linux.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: powerpc_excp: Remove dump_syscall_vectored

This function is identical to dump_syscall, so use the latter for
system call vectored as well.

Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Message-Id: <20210601214649.785647-3-farosas@linux.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: powerpc_excp: Move lpes code to where it is used

Signed-off-by: Fabiano Rosas <farosas@linux.ibm.com>
Message-Id: <20210601214649.785647-2-farosas@linux.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: overhauled and moved logic of storing fpscr

Followed the suggested overhaul to store_fpscr logic, and moved it to
cpu.c where it can be accessed in !TCG builds.

The overhaul was suggested because storing a value to fpscr should
never raise an exception, so we could remove all the mess that happened
with POWERPC_EXCP_FP.

We also moved fpscr_set_rounding_mode into cpu.c as it could now be moved
there, and it is needed when a value for the fpscr is being stored
directly.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210527163522.23019-1-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: removed all mentions to PPC_DUMP_CPU

This feature will no longer be useful as ppc moves to using decodetree
for TCG. And building with it enabled is no longer possible, due to
changes in opc_handler_t. Since the last commit that mentions it
happened in 2014, I think it is safe to remove it.

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210531145629.21300-5-bruno.larsen@eldorado.org.br>
Reviewed-by: Luis Pires <luis.pires@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: removed GEN_OPCODE decision tree

since both, PPC_DO_STATISTICS and PPC_DUMP_CPU, are obsoleted as
target/ppc moves to decodetree, we can remove this ifdef based decision
tree, and only have what is now the standard option for the macro.

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210531145629.21300-4-bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

HMP: added info cpustats to removed_features.rst

Documented the removal of the HMP command info cpustats

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210531145629.21300-3-bruno.larsen@eldorado.org.br>
Reviewed-by: Luis Pires <luis.pires@eldorado.org.br>
Reviewed-by: Lucas Mateus <lucas.araujo@eldorado.org.br>
Reviewed-by: Greg Kurz <groug@kaod.org>
[dwg: Fix a Sphinx formatting error]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

hw/core/cpu: removed cpu_dump_statistics function

No more architectures set the pointer to dump_statistics, so there's no
point in keeping it, or the related cpu_dump_statistics function.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210526202104.127910-6-bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Luis Pires <luis.pires@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Message-Id: <20210531145629.21300-2-bruno.larsen@eldorado.org.br>
Acked-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

ppc/pef.c: initialize cgs->ready in kvmppc_svm_init()

QEMU is failing to launch a CGS pSeries guest in a host that has PEF
support:

qemu-system-ppc64: ../softmmu/vl.c:2585: qemu_machine_creation_done: Assertion `machine->cgs->ready' failed.
Aborted

This is happening because we're not setting the cgs->ready flag that is
asserted in qemu_machine_creation_done() during machine start.

cgs->ready is set in s390_pv_kvm_init() and sev_kvm_init(). Let's set it
in kvmppc_svm_init() as well.

Reported-by: Ram Pai <linuxram@us.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20210528201619.52363-1-danielhb413@gmail.com>
Acked-by: Ram Pai <linuxram@us.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

monitor: removed cpustats command

Since ppc was the last architecture to collect these statistics and
it is currently phasing this collection out, the command that would query
this information is being removed.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210526202104.127910-5-bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Luis Pires <luis.pires@eldorado.org.br>
Acked-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: removed mentions to DO_PPC_STATISTICS

Removed the commented out definition and all ifdefs relating to
PPC_DUMP_STATISTICS, as it's hardly ever used.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210526202104.127910-4-bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Luis Pires <luis.pires@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: remove ppc_cpu_dump_statistics

This function requires surce code modification to be useful, which means
it probably is not used often, and the move to using decodetree means
the statistics won't even be collected anymore.

Also removed setting dump_statistics in ppc_cpu_realize, since it was
only useful when in conjunction with ppc_cpu_dump_statistics.

Suggested-by: Richard Henderson<richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210526202104.127910-3-bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Luis Pires <luis.pires@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: updated meson.build to support disable-tcg

updated build file to not compile some sources that are unnecessary if
TCG is disabled on the system.

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210525115355.8254-5-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: created tcg-stub.c file

Created a file with stubs needed to compile disabling TCG. *_ppc_opcodes
were created to make cpu_init.c have a few less ifdefs, since they are
not needed. softmmu_resize_hpt_* have to be created because the compiler
can't automatically know they aren't used, but they should never be
reached.

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210525115355.8254-4-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: added ifdefs around TCG-only code

excp_helper.c, mmu-hash64.c and mmu_helper.c have some function
declarations that are TCG-only, and couldn't be easily moved to a
TCG only file, so ifdefs were added around them.

We also needed ifdefs around some header files because helper-proto.h
includes trace/generated-helpers.h, which is never created when building
without TCG, and cpu_ldst.h includes tcg/tcg.h, whose containing folder
is not included as a -iquote. As future cleanup, we could change the
part of the configuration script to add those.

cpu_init.c also had a callback definition that is TCG only and could be
removed as part of a future cleanup (all the dump_statistics part is
almost never used and will become obsolete as we transition to using
decodetree).

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210525115355.8254-3-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: used ternary operator when registering MAS

The write calback decision when registering the MAS SPR has been turned
into a ternary operation, rather than an if-then-else block.

This was done because when building without TCG, even though the
compiler will optimize away the pointers to spr_write_generic*, it
doesn't optimize away the decision and assignment to the local pointer,
creating compiler errors. This cleanup looked better than using ifdefs,
so  we decided to with it.

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210525115355.8254-2-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Set LPCR to current AIL mode when starting a new CPU

TCG does not keep track of AIL mode in a central place, it's based on
the current LPCR[AIL] bits. Synchronize the new CPU's LPCR to the
current LPCR in rtas_start_cpu(), similarly to the way the ILE bit is
synchronized.

Open-code the ILE setting as well now that the caller's LPCR is
available directly, there is no need for the indirection.

Without this, under both TCG and KVM, adding a POWER8/9/10 class CPU
with a new core ID after a modern Linux has booted results in the new
CPU's LPCR missing the LPCR[AIL]=0b11 setting that the other CPUs have.
This can cause crashes and unexpected behaviour.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20210526091626.3388262-3-npiggin@gmail.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Remove stale comment about power-saving LPCR bits

Commit 47a9b551547 ("spapr: Clean up handling of LPCR power-saving exit
bits") moved this logic but did not remove the comment from the
previous location.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20210526091626.3388262-2-npiggin@gmail.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: fold ppc_store_ptcr into it's only caller

ppc_store_ptcr, defined in mmu_helper.c, was only used by
helper_store_ptcr, in misc_helper.c. To avoid possible confusion,
the function was folded into the helper.

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Message-Id: <20210526143516.125582-1-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: nvdimm: Fix the persistent-memory root node name in device tree

The FDT code is adding the pmem root node by name "persistent-memory"
which should have been "ibm,persistent-memory".

The linux fetches the device tree nodes by type and it has been working
correctly as the type is correct. If someone searches by its intended
name it would fail, so fix that.

Reported-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Message-Id: <162204278956.219.9061511386011411578.stgit@cc493db1e665>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: nvdimm: Forward declare and move the definitions

The subsequent patches add definitions which tend to get
the compilation to cyclic dependency. So, prepare with
forward declarations, move the definitions and clean up.

Signed-off-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Message-Id: <162133925415.610.11584121797866216417.stgit@4f1e6f2bd33e>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Fix EEH capability issue on KVM guest for PCI passthru

With upstream kernel, especially after commit 98ba956f6a389
("powerpc/pseries/eeh: Rework device EEH PE determination") we see that KVM
guest isn't able to enable EEH option for PCI pass-through devices anymore.

[root@atest-guest ~]# dmesg | grep EEH
[    0.032337] EEH: pSeries platform initialized
[    0.298207] EEH: No capable adapters found: recovery disabled.
[root@atest-guest ~]#

So far the linux kernel was assuming pe_config_addr equal to device's
config_addr and using it to enable EEH on the PE through ibm,set-eeh-option
RTAS call. Which wasn't the correct way as per PAPR. The linux kernel
commit 98ba956f6a389 fixed this flow. With that fixed, linux now uses PE
config address returned by ibm,get-config-addr-info2 RTAS call to enable
EEH option per-PE basis instead of per-device basis. However this has
uncovered a bug in qemu where ibm,set-eeh-option is treating PE config
address as per-device config address.

Hence in qemu guest with recent kernel the ibm,set-eeh-option RTAS call
fails with -3 return value indicating that there is no PCI device exist for
the specified PE config address. The rtas_ibm_set_eeh_option call uses
pci_find_device() to get the PC device that matches specific bus and devfn
extracted from PE config address passed as argument. Thus it tries to map
the PE config address to a single specific PCI device 'bus->devices[devfn]'
which always results into checking device on slot 0 'bus->devices[0]'.
This succeeds when there is a pass-through device (vfio-pci) present on
slot 0. But in cases where there is no pass-through device present in slot
0, but present in non-zero slots, ibm,set-eeh-option call fails to enable
the EEH capability.

hw/ppc/spapr_pci_vfio.c: spapr_phb_vfio_eeh_set_option()
   case RTAS_EEH_ENABLE: {
        PCIHostState *phb;
        PCIDevice *pdev;

        /*
         * The EEH functionality is enabled on basis of PCI device,
         * instead of PE. We need check the validity of the PCI
         * device address.
         */
        phb = PCI_HOST_BRIDGE(sphb);
        pdev = pci_find_device(phb->bus,
                               (addr >> 16) & 0xFF, (addr >> 8) & 0xFF);
        if (!pdev || !object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
            return RTAS_OUT_PARAM_ERROR;
        }

hw/pci/pci.c:pci_find_device()

PCIDevice *pci_find_device(PCIBus *bus, int bus_num, uint8_t devfn)
{
    bus = pci_find_bus_nr(bus, bus_num);

    if (!bus)
        return NULL;

    return bus->devices[devfn];
}

This patch fixes ibm,set-eeh-option to check for presence of any PCI device
(vfio-pci) under specified bus and enable the EEH if found. The current
code already makes sure that all the devices on that bus are from same
iommu group (within same PE) and fail very early if it does not.

After this fix guest is able to find EEH capable devices and enable EEH
recovery on it.

[root@atest-guest ~]# dmesg | grep EEH
[    0.048139] EEH: pSeries platform initialized
[    0.405115] EEH: Capable adapter found: recovery enabled.
[root@atest-guest ~]#

Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Mahesh Salgaonkar <mahesh@linux.ibm.com>
Message-Id: <162158429107.145117.5843504911924013125.stgit@jupiter>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

spapr: Don't hijack current_machine->boot_order

QEMU 6.0 moved all the -boot variables to the machine. Especially, the
removal of the boot_order static changed the handling of '-boot once'
from:

    if (boot_once) {
        qemu_boot_set(boot_once, &error_fatal);
        qemu_register_reset(restore_boot_order, g_strdup(boot_order));
    }

to

    if (current_machine->boot_once) {
        qemu_boot_set(current_machine->boot_once, &error_fatal);
        qemu_register_reset(restore_boot_order,
                            g_strdup(current_machine->boot_order));
    }

This means that we now register as subsequent boot order a copy
of current_machine->boot_once that was just set with the previous
call to qemu_boot_set(), i.e. we never transition away from the
once boot order.

It is certainly fragile^Wwrong for the spapr code to hijack a
field of the base machine type object like that. The boot order
rework simply turned this software boundary violation into an
actual bug.

Have the spapr code to handle that with its own field in
SpaprMachineState. Also kfree() the initial boot device
string when "once" was used.

Fixes: 4b7acd2ac821 ("vl: clean up -boot variables")
Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1960119
Cc: pbonzini@redhat.com
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <20210521160735.1901914-1-groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: removed unnecessary inclusion of helper-proto.h

These files included helper-proto.h, but didn't use or declare any
helpers, so the #include has been removed

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210521201759.85475-6-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: reduce usage of fpscr_set_rounding_mode

It is preferable to store the current rounding mode and retore from that
than recalculating from fpscr, so we changed the behavior of do_fri and
VSX_ROUND to do it like that.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210521201759.85475-4-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: moved ppc_store_lpcr and ppc_store_msr to cpu.c

These functions are used in hw/ppc logic, during machine startup, which
means it must be compiled when --disable-tcg is selected, and so it has
been moved into a common code file

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Message-Id: <20210521201759.85475-3-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

target/ppc: cleaned error_report from ppc_store_sdr1

Changed how the function ppc_store_sdr1, from error_report(...) to
qemu_log_mask(LOG_GUEST_ERROR, ...).

Signed-off-by: Bruno Larsen (billionai) <bruno.larsen@eldorado.org.br>
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20210521201759.85475-2-bruno.larsen@eldorado.org.br>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>

Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging

Block layer patches

- NBD server: Fix crashes related to switching between AioContexts
- file-posix: Workaround for discard/write_zeroes on buggy filesystems
- Follow-up fixes for the reopen vs. permission changes
- quorum: Fix error handling for flush
- block-copy: Refactor copy_range handling
- docs: Describe how to use 'null-co' block driver

# gpg: Signature made Wed 02 Jun 2021 14:44:15 BST
# gpg:                using RSA key DC3DEB159A9AF95D3D7456FE7F09B272C88F2FD6
# gpg:                issuer "kwolf@redhat.com"
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>" [full]
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* remotes/kevin/tags/for-upstream:
  docs/secure-coding-practices: Describe how to use 'null-co' block driver
  block-copy: refactor copy_range handling
  block-copy: fix block_copy_task_entry() progress update
  nbd/server: Use drained block ops to quiesce the server
  block-backend: add drained_poll
  block: improve permission conflict error message
  block: simplify bdrv_child_user_desc()
  block/vvfat: inherit child_vvfat_qcow from child_of_bds
  block: improve bdrv_child_get_parent_desc()
  block-backend: improve blk_root_get_parent_desc()
  block: document child argument of bdrv_attach_child_common()
  block/file-posix: Try other fallbacks after invalid FALLOC_FL_ZERO_RANGE
  block/file-posix: Fix problem with fallocate(PUNCH_HOLE) on GPFS
  block: drop BlockBackendRootState::read_only
  block: drop BlockDriverState::read_only
  block: consistently use bdrv_is_read_only()
  block/vvfat: fix vvfat_child_perm crash
  block/vvfat: child_vvfat_qcow: add .get_parent_aio_context, fix crash
  qemu-io-cmds: assert that we don't have .perm requested in no-blk case
  block/quorum: Provide .bdrv_co_flush instead of .bdrv_co_flush_to_disk

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/thuth-gitlab/tags/pull-request-2021-06-02' into staging

* Update the references to some doc files (use *.rst instead of *.txt)
* Bump minimum versions of some requirements after removing CentOS 7 support

# gpg: Signature made Wed 02 Jun 2021 08:12:18 BST
# 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

* remotes/thuth-gitlab/tags/pull-request-2021-06-02:
  configure: bump min required CLang to 6.0 / XCode 10.0
  configure: bump min required GCC to 7.5.0
  configure: bump min required glib version to 2.56
  tests/docker: drop CentOS 7 container
  tests/vm: convert centos VM recipe to CentOS 8
  crypto: drop used conditional check
  crypto: bump min gnutls to 3.5.18, dropping RHEL-7 support
  crypto: bump min gcrypt to 1.8.0, dropping RHEL-7 support
  crypto: drop back compatibility typedefs for nettle
  crypto: bump min nettle to 3.4, dropping RHEL-7 support
  patchew: move quick build job from CentOS 7 to CentOS 8 container
  block/ssh: Bump minimum libssh version to 0.8.7
  docs: fix references to docs/devel/s390-dasd-ipl.rst
  docs: fix references to docs/specs/tpm.rst
  docs: fix references to docs/devel/build-system.rst
  docs: fix references to docs/devel/atomics.rst
  docs: fix references to docs/devel/tracing.rst

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

docs/secure-coding-practices: Describe how to use 'null-co' block driver

Document that security reports must use 'null-co,read-zeroes=on'
because otherwise the memory is left uninitialized (which is an
on-purpose performance feature).

Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20210601162548.2076631-1-philmd@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block-copy: refactor copy_range handling

Currently we update s->use_copy_range and s->copy_size in
block_copy_do_copy().

It's not very good:

1. block_copy_do_copy() is intended to be a simple function, that wraps
bdrv_co_<io> functions for need of block copy. That's why we don't pass
BlockCopyTask into it. So, block_copy_do_copy() is bad place for
manipulation with generic state of block-copy process

2. We are going to make block-copy thread-safe. So, it's good to move
manipulation with state of block-copy to the places where we'll need
critical sections anyway, to not introduce extra synchronization
primitives in block_copy_do_copy().

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210528141628.44287-3-vsementsov@virtuozzo.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block-copy: fix block_copy_task_entry() progress update

Don't report successful progress on failure, when call_state->ret is
set.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210528141628.44287-2-vsementsov@virtuozzo.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

nbd/server: Use drained block ops to quiesce the server

Before switching between AioContexts we need to make sure that we're
fully quiesced ("nb_requests == 0" for every client) when entering the
drained section.

To do this, we set "quiescing = true" for every client on
".drained_begin" to prevent new coroutines from being created, and
check if "nb_requests == 0" on ".drained_poll". Finally, once we're
exiting the drained section, on ".drained_end" we set "quiescing =
false" and call "nbd_client_receive_next_request()" to resume the
processing of new requests.

With these changes, "blk_aio_attach()" and "blk_aio_detach()" can be
reverted to be as simple as they were before f148ae7d36.

RHBZ: https://bugzilla.redhat.com/show_bug.cgi?id=1960137
Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Sergio Lopez <slp@redhat.com>
Message-Id: <20210602060552.17433-3-slp@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block-backend: add drained_poll

Allow block backends to poll their devices/users to check if they have
been quiesced when entering a drained section.

This will be used in the next patch to wait for the NBD server to be
completely quiesced.

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Sergio Lopez <slp@redhat.com>
Message-Id: <20210602060552.17433-2-slp@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: improve permission conflict error message

Now permissions are updated as follows:
 1. do graph modifications ignoring permissions
 2. do permission update

 (of course, we rollback [1] if [2] fails)

So, on stage [2] we can't say which users are "old" and which are
"new" and exist only since [1]. And current error message is a bit
outdated. Let's improve it, to make everything clean.

While being here, add also a comment and some good assertions.

iotests 283, 307, qsd-jobs outputs are updated.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210601075218.79249-7-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: simplify bdrv_child_user_desc()

All child classes have this callback. So, drop unreachable code.

Still add an assertion to bdrv_attach_child_common(), to early detect
bad classes.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210601075218.79249-6-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/vvfat: inherit child_vvfat_qcow from child_of_bds

Recently we've fixed a crash by adding .get_parent_aio_context handler
to child_vvfat_qcow. Now we want it to support .get_parent_desc as
well. child_vvfat_qcow wants to implement own .inherit_options, it's
not bad. But omitting all other handlers is a bad idea. Let's inherit
the class from child_of_bds instead, similar to chain_child_class and
detach_by_driver_cb_class in test-bdrv-drain.c.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210601075218.79249-5-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: improve bdrv_child_get_parent_desc()

We have different types of parents: block nodes, block backends and
jobs. So, it makes sense to specify type together with name.

Next, this handler us used to compose an error message about permission
conflict. And permission conflict occurs in a specific place of block
graph. We shouldn't report name of parent device (as it refers another
place in block graph), but exactly and only the name of the node. So,
use bdrv_get_node_name() directly.

iotest 283 output is updated.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-Id: <20210601075218.79249-4-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block-backend: improve blk_root_get_parent_desc()

We have different types of parents: block nodes, block backends and
jobs. So, it makes sense to specify type together with name.

While being here also use g_autofree.

iotest 307 output is updated.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-Id: <20210601075218.79249-3-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: document child argument of bdrv_attach_child_common()

The logic around **child is not obvious: this reference is used not
only to return resulting child, but also to rollback NULL value on
transaction abort.

So, let's add documentation and some assertions.

While being here, drop extra declaration of bdrv_attach_child_noperm().

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210601075218.79249-2-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/file-posix: Try other fallbacks after invalid FALLOC_FL_ZERO_RANGE

If fallocate(... FALLOC_FL_ZERO_RANGE ...) returns EINVAL, it's likely
an indication that the file system is buggy and does not implement
unaligned accesses right. We still might be lucky with the other
fallback fallocate() calls later in this function, though, so we should
not return immediately and try the others first.
Since FALLOC_FL_ZERO_RANGE could also return EINVAL if the file descriptor
is not a regular file, we ignore this filesystem bug silently, without
printing an error message for the user.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20210527172020.847617-3-thuth@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/file-posix: Fix problem with fallocate(PUNCH_HOLE) on GPFS

A customer reported that running

 qemu-img convert -t none -O qcow2 -f qcow2 input.qcow2 output.qcow2

fails for them with the following error message when the images are
stored on a GPFS file system :

 qemu-img: error while writing sector 0: Invalid argument

After analyzing the strace output, it seems like the problem is in
handle_aiocb_write_zeroes(): The call to fallocate(FALLOC_FL_PUNCH_HOLE)
returns EINVAL, which can apparently happen if the file system has
a different idea of the granularity of the operation. It's arguably
a bug in GPFS, since the PUNCH_HOLE mode should not result in EINVAL
according to the man-page of fallocate(), but the file system is out
there in production and so we have to deal with it. In commit 294682cc3a
("block: workaround for unaligned byte range in fallocate()") we also
already applied the a work-around for the same problem to the earlier
fallocate(FALLOC_FL_ZERO_RANGE) call, so do it now similar with the
PUNCH_HOLE call. But instead of silently catching and returning
-ENOTSUP (which causes the caller to fall back to writing zeroes),
let's rather inform the user once about the buggy file system and
try the other fallback instead.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20210527172020.847617-2-thuth@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: drop BlockBackendRootState::read_only

Instead of keeping additional boolean field, let's store the
information in BDRV_O_RDWR bit of BlockBackendRootState::open_flags.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210527154056.70294-4-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: drop BlockDriverState::read_only

This variable is just a cache for !(bs->open_flags & BDRV_O_RDWR),
which we have to synchronize everywhere. Let's just drop it and
consistently use bdrv_is_read_only().

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210527154056.70294-3-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: consistently use bdrv_is_read_only()

It's better to use accessor function instead of bs->read_only directly.
In some places use bdrv_is_writable() instead of
checking both BDRV_O_RDWR set and BDRV_O_INACTIVE not set.

In bdrv_open_common() it's a bit strange to add one more variable, but
we are going to drop bs->read_only in the next patch, so new ro local
variable substitutes it here.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210527154056.70294-2-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block/vvfat: fix vvfat_child_perm crash

It's wrong to rely on s->qcow in vvfat_child_perm, as on permission
update during bdrv_open_child() call this field is not set yet.

Still prior to aa5a04c7db27eea6b36de32f241b155f0d9ce34d, it didn't
crash, as bdrv_open_child passed NULL as child to bdrv_child_perm(),
and NULL was equal to NULL in assertion (still, it was bad guarantee
for child being s->qcow, not backing :).

Since aa5a04c7db27eea6b36de32f241b155f0d9ce34d
"add bdrv_attach_child_noperm" bdrv_refresh_perms called on parent node
when attaching child, and new correct child pointer is passed to
.bdrv_child_perm. Still, s->qcow is NULL at the moment. Let's rely only
on role instead.

Without that fix,
./build/qemu-system-x86_64 -usb -device usb-storage,drive=fat16 \
    -drive \
    file=fat:rw:fat-type=16:"<path of a host folder>",id=fat16,format=raw,if=none

crashes:
(gdb) bt
0  raise () at /lib64/libc.so.6
1  abort () at /lib64/libc.so.6
2  _nl_load_domain.cold () at /lib64/libc.so.6
3  annobin_assert.c_end () at /lib64/libc.so.6
4  vvfat_child_perm (bs=0x559186f3d690, c=0x559186f1ed20, role=3,
                     reopen_queue=0x0, perm=0, shared=31,
                     nperm=0x7ffe56f28298, nshared=0x7ffe56f282a0) at
    ../block/vvfat.c:3214
5  bdrv_child_perm (bs=0x559186f3d690, child_bs=0x559186f60190,
                    c=0x559186f1ed20, role=3, reopen_queue=0x0,
                    parent_perm=0, parent_shared=31,
                    nperm=0x7ffe56f28298, nshared=0x7ffe56f282a0)
    at ../block.c:2094
6  bdrv_node_refresh_perm (bs=0x559186f3d690, q=0x0,
                           tran=0x559186f65850, errp=0x7ffe56f28530) at
    ../block.c:2336
7  bdrv_list_refresh_perms (list=0x559186db5b90 = {...}, q=0x0,
                            tran=0x559186f65850, errp=0x7ffe56f28530)
    at ../block.c:2358
8  bdrv_refresh_perms (bs=0x559186f3d690, errp=0x7ffe56f28530) at
    ../block.c:2419
9  bdrv_attach_child
    (parent_bs=0x559186f3d690, child_bs=0x559186f60190,
     child_name=0x559184d83e3d "write-target",
     child_class=0x5591852f3b00 <child_vvfat_qcow>, child_role=3,
     errp=0x7ffe56f28530) at ../block.c:2959
10 bdrv_open_child
    (filename=0x559186f5cb80 "/var/tmp/vl.7WYmFU",
     options=0x559186f66c20, bdref_key=0x559184d83e3d "write-target",
     parent=0x559186f3d690, child_class=0x5591852f3b00
     <child_vvfat_qcow>, child_role=3, allow_none=false,
     errp=0x7ffe56f28530) at ../block.c:3351
11 enable_write_target (bs=0x559186f3d690, errp=0x7ffe56f28530) at
    ../block/vvfat.c:3177
12 vvfat_open (bs=0x559186f3d690, options=0x559186f42db0, flags=155650,
               errp=0x7ffe56f28530) at ../block/vvfat.c:1236
13 bdrv_open_driver (bs=0x559186f3d690, drv=0x5591853d97e0
                     <bdrv_vvfat>, node_name=0x0,
                     options=0x559186f42db0, open_flags=155650,
                     errp=0x7ffe56f28640) at ../block.c:1557
14 bdrv_open_common (bs=0x559186f3d690, file=0x0,
                     options=0x559186f42db0, errp=0x7ffe56f28640) at
    ../block.c:1833
...

(gdb) fr 4
 #4  vvfat_child_perm (bs=0x559186f3d690, c=0x559186f1ed20, role=3,
                      reopen_queue=0x0, perm=0, shared=31,
                      nperm=0x7ffe56f28298, nshared=0x7ffe56f282a0) at
    ../block/vvfat.c:3214
3214        assert(c == s->qcow || (role & BDRV_CHILD_COW));
(gdb) p role
 $1 = 3   # BDRV_CHILD_DATA | BDRV_CHILD_METADATA
(gdb) p *c
 $2 = {bs = 0x559186f60190, name = 0x559186f669d0 "write-target", klass
     = 0x5591852f3b00 <child_vvfat_qcow>, role = 3, opaque =
         0x559186f3d690, perm = 3, shared_perm = 4, frozen = false,
         parent_quiesce_counter = 0, next = {le_next = 0x0, le_prev =
             0x559186f41818}, next_parent = {le_next = 0x0, le_prev =
                 0x559186f64320}}
(gdb) p s->qcow
 $3 = (BdrvChild *) 0x0

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210524101257.119377-3-vsementsov@virtuozzo.com>
Tested-by: John Arbuckle <programmingkidx@gmail.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>