A Single RISC-V Patch for 3.1-rc1

This tag contains a single patch that I'd like to target for rc1: a fix
for a memory leak that was detected by static code analysis.

There are still three patch sets that I'd like to try to get up for 3.1:

* The patch set Basian just published that contains fixes for a pair of
  issues he found when converting our port to decodetree.
* An as-of-yet-unwritten fix to the third issue that Basian pointed out.
* A fix to our fflags bug, which is currently coupled to some CSR
  refactoring that I don't think is OK for 3.1.

I'm at Plumbers next week (and I think Alistair is there too?), but I'll
try to find a way to squeeze in as much as possible.

# gpg: Signature made Thu 08 Nov 2018 16:50:27 GMT
# gpg:                using RSA key EF4CA1502CCBAB41
# gpg: Good signature from "Palmer Dabbelt <palmer@dabbelt.com>"
# gpg:                 aka "Palmer Dabbelt <palmer@sifive.com>"
# 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: 00CE 76D1 8349 60DF CE88  6DF8 EF4C A150 2CCB AB41

* remotes/riscv/tags/riscv-for-master-3.1-rc1:
  riscv: spike: Fix memory leak in the board init
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>

Coverity caught a malloc() call that was never freed. This patch ensures
that we free the memory but also updates the allocation to use
g_strdup_printf() instead of malloc().
Signed-off-by: NAlistair Francis <alistair.francis@wdc.com>
Suggested-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NPalmer Dabbelt <palmer@sifive.com>
Signed-off-by: NPalmer Dabbelt <palmer@sifive.com>

The tcg-op.h header was missing the usual guard against multiple
inclusion; add it.

(Spotted by lgtm.com's static analyzer.)
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NMarc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: NPhilippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: NRichard Henderson <richard.henderson@linaro.org>
Tested-by: NPhilippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20181108125256.30986-1-peter.maydell@linaro.org

ppc patch queue 2018-11-08

Here's another patch of accumulated ppc patches for qemu-3.1.
Highlights are:
  * Support for nested HV KVM on POWER9 hosts
  * Remove Alex Graf as ppc maintainer
  * Emulation of external PID instructions

# gpg: Signature made Thu 08 Nov 2018 12:14:27 GMT
# gpg:                using RSA key 6C38CACA20D9B392
# gpg: Good signature from "David Gibson <david@gibson.dropbear.id.au>"
# gpg:                 aka "David Gibson (Red Hat) <dgibson@redhat.com>"
# gpg:                 aka "David Gibson (ozlabs.org) <dgibson@ozlabs.org>"
# gpg:                 aka "David Gibson (kernel.org) <dwg@kernel.org>"
# Primary key fingerprint: 75F4 6586 AE61 A66C C44E  87DC 6C38 CACA 20D9 B392

* remotes/dgibson/tags/ppc-for-3.1-20181108: (22 commits)
  ppc/spapr_caps: Add SPAPR_CAP_NESTED_KVM_HV
  target/ppc: Add one reg id for ptcr
  This patch fixes processing of rfi instructions in icount mode.
  hw/ppc/ppc440_uc: Remove dead code in sdram_size()
  MAINTAINERS: PPC: Remove myself
  ppc/pnv: check size before data buffer access
  target/ppc: fix mtmsr instruction for icount
  hw/ppc/mac_newworld: Free openpic_irqs array after use
  macio/pmu: Fix missing vmsd terminator
  spapr_pci: convert g_malloc() to g_new()
  target/ppc: Split out float_invalid_cvt
  target/ppc: Split out float_invalid_op_div
  target/ppc: Split out float_invalid_op_mul
  target/ppc: Split out float_invalid_op_addsub
  target/ppc: Introduce fp number classification
  target/ppc: Remove float_check_status
  target/ppc: Split up float_invalid_op_excp
  hw/ppc/spapr_rng: Introduce CONFIG_SPAPR_RNG switch for spapr_rng.c
  PPC: e500: convert SysBus init method to a realize method
  ppc4xx_pci: convert SysBus init method to a realize method
  ...
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>

Currently replay_get_byte() does not check for an error
from getc(). Coverity points out (CID 1390622) that this
could result in unexpected behaviour (such as looping
forever, if we use the replay_get_dword() return value
for a loop count). We don't expect reads from the replay
log to fail, and if they do there is no way we can
continue. So make them fatal errors.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NPavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Message-id: 20181106153330.5139-1-peter.maydell@linaro.org

* icount fix (Clement)
* dumping fixes for non-volatile memory (Marc-André, myself)
* x86 emulation fix (Rudolf)
* recent Hyper-V CPUID flag (Vitaly)
* Q35 doc fix (Daniel)
* lsi fix (Prasad)
* SCSI block limits emulation fixes (myself)
* qemu_thread_atexit rework (Peter)
* ivshmem memory leak fix (Igor)

# gpg: Signature made Tue 06 Nov 2018 21:34:30 GMT
# gpg:                using RSA key BFFBD25F78C7AE83
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>"
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>"
# Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* remotes/bonzini/tags/for-upstream:
  util/qemu-thread-posix: Fix qemu_thread_atexit* for OSX
  include/qemu/thread.h: Document qemu_thread_atexit* API
  scsi-generic: do not do VPD emulation for sense other than ILLEGAL_REQUEST
  scsi-generic: avoid invalid access to struct when emulating block limits
  scsi-generic: avoid out-of-bounds access to VPD page list
  scsi-generic: keep VPD page list sorted
  lsi53c895a: check message length value is valid
  scripts/dump-guest-memory: Synchronize with guest_phys_blocks_region_add
  memory-mapping: skip non-volatile memory regions in GuestPhysBlockList
  nvdimm: set non-volatile on the memory region
  memory: learn about non-volatile memory region
  target/i386: Clear RF on SYSCALL instruction
  MAINTAINERS: remove or downgrade myself to reviewer from some subsystems
  ivshmem: fix memory backend leak
  i386: clarify that the Q35 machine type implements a P35 chipset
  x86: hv_evmcs CPU flag support
  icount: fix deadlock when all cpus are sleeping
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>

Add the spapr cap SPAPR_CAP_NESTED_KVM_HV to be used to control the
availability of nested kvm-hv to the level 1 (L1) guest.

Assuming a hypervisor with support enabled an L1 guest can be allowed to
use the kvm-hv module (and thus run it's own kvm-hv guests) by setting:
-machine pseries,cap-nested-hv=true
or disabled with:
-machine pseries,cap-nested-hv=false
Signed-off-by: NSuraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

The ptcr (partition table control register) is used to store the address
and size of the partition table. For nested kvm-hv we have a level 1
guest register the location of it's partition table with the hypervisor.
Thus to support migration we need to be able to read this out of kvm
and restore it post migration.

Add the one reg id for the ptcr.
Signed-off-by: NSuraj Jitindar Singh <sjitindarsingh@gmail.com>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

In this mode writing to interrupt/peripheral state is controlled
by can_do_io flag. This flag must be set explicitly before helper
function invocation.
Signed-off-by: NMaria Klimushenkova <maria.klimushenkova@ispras.ru>
Signed-off-by: NPavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Tested-by: NMark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Coverity points out in CID 1390588 that the test for sh == 0
in sdram_size() can never fire, because we calculate sh with
    sh = 1024 - ((bcr >> 6) & 0x3ff);
which must result in a value between 1 and 1024 inclusive.

Without the relevant manual for the SoC, we're not completely
sure of the correct behaviour here, but we can remove the
dead code without changing how QEMU currently behaves.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NLaurent Vivier <lvivier@redhat.com>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

I haven't really been maintaining any PowerPC code for quite a while now,
so let's reflect reality: David does all the work and embedded PPC is in
"Odd Fixes" state rather than supported now.
Signed-off-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

While performing PowerNV memory r/w operations, the access length
'sz' could exceed the data[4] buffer size. Add check to avoid OOB
access.
Reported-by: NMoguofang <moguofang@huawei.com>
Signed-off-by: NPrasad J Pandit <pjp@fedoraproject.org>
Reviewed-by: NCédric Le Goater <clg@kaod.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

This patch fixes processing of mtmsr instructions in icount mode.
In this mode writing to interrupt/peripheral state is controlled
by can_do_io flag. This flag must be set explicitly before helper
function invocation.
Signed-off-by: NMaria Klimushenkova <maria.klimushenkova@ispras.ru>
Signed-off-by: NPavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Reviewed-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

In ppc_core99_init(), we allocate an openpic_irqs array, which
we then use to collect up the various qemu_irqs which we're
going to connect to the interrupt controller. Once we've
called sysbus_connect_irq() to connect them all up, the
array is no longer required, but we forgot to free it.

Since board init is only run once at startup, the memory
leak is not a significant one.

Spotted by Coverity: CID 1192916.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NPhilippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: NMark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Fix missing terminator in VMStateDescription

Fixes: d811d61fSigned-off-by: NDr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: NMark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: NLaurent Vivier <lvivier@redhat.com>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

When allocating an array, it is a recommended coding practice to call
g_new(FooType, n) instead of g_malloc(n * sizeof(FooType)) because
it takes care to avoid overflow when calculating the size of the
allocated block and it returns FooType *, which allows the compiler
to perform type checking.
Signed-off-by: NGreg Kurz <groug@kaod.org>
Reviewed-by: NPhilippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Signed-off-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Signed-off-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Signed-off-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Signed-off-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Having a separate, logical classifiation of numbers will
unify more error paths for different formats.
Signed-off-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Use do_float_check_status directly, so that we don't get confused
about which return address we're using.  And definitely don't use
helper_float_check_status.
Signed-off-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

The always_inline trick only works if the function is always
called from the outer-most helper.  But it isn't, so pass in
the outer-most return address.  There's no need for a switch
statement whose argument is always a constant.  Unravel the
switch and goto via more helpers.
Signed-off-by: NRichard Henderson <richard.henderson@linaro.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

The spapr-rng device is suboptimal when compared to virtio-rng, so
users might want to disable it in their builds. Thus let's introduce
a proper CONFIG switch to allow us to compile QEMU without this device.
The function spapr_rng_populate_dt is required for linking, so move it
to a different location.
Signed-off-by: NThomas Huth <thuth@redhat.com>
Reviewed-by: NGreg Kurz <groug@kaod.org>
Reviewed-by: NPhilippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Signed-off-by: NCédric Le Goater <clg@kaod.org>
Reviewed-by: NPhilippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: NGreg Kurz <groug@kaod.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Signed-off-by: NCédric Le Goater <clg@kaod.org>
Reviewed-by: NGreg Kurz <groug@kaod.org>
Reviewed-by: NPhilippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Signed-off-by: NCédric Le Goater <clg@kaod.org>
Reviewed-by: NGreg Kurz <groug@kaod.org>
Reviewed-by: NPhilippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

External PID is a mechanism present on BookE 2.06 that enables application to
store/load data from different address spaces. There are special version of some
instructions, which operate on alternate address space, which is specified in
the EPLC/EPSC regiser.

This implementation uses two additional MMU modes (mmu_idx) to provide the
address space for the load and store instructions. The QEMU TLB fill code was
modified to recognize these MMU modes and use the values in EPLC/EPSC to find
the proper entry in he PPC TLB. These two QEMU TLBs are also flushed on each
write to EPLC/EPSC.

Following instructions are implemented: dcbfep dcbstep dcbtep dcbtstep dcbzep
dcbzlep icbiep lbepx ldepx lfdepx lhepx lwepx stbepx stdepx stfdepx sthepx
stwepx.

Following vector instructions are not: evlddepx evstddepx lvepx lvepxl stvepx
stvepxl.
Signed-off-by: NRoman Kapl <rka@sysgo.com>
Signed-off-by: NDavid Gibson <david@gibson.dropbear.id.au>

Our current implementation of qemu_thread_atexit* is broken on OSX.
This is because it works by cerating a piece of thread-specific
data with pthread_key_create() and using the destructor function
for that data to run the notifier function passed to it by
the caller of qemu_thread_atexit_add(). The expected use case
is that the caller uses a __thread variable as the notifier,
and uses the callback to clean up information that it is
keeping per-thread in __thread variables.

Unfortunately, on OSX this does not work, because on OSX
a __thread variable may be destroyed (freed) before the
pthread_key_create() destructor runs. (POSIX imposes no
ordering constraint here; the OSX implementation happens
to implement __thread variables in terms of pthread_key_create((),
whereas Linux uses different mechanisms that mean the __thread
variables will still be present when the pthread_key_create()
destructor is run.)

Fix this by switching to a scheme similar to the one qemu-thread-win32
uses for qemu_thread_atexit: keep the thread's notifiers on a
__thread variable, and run the notifiers on calls to
qemu_thread_exit() and on return from the start routine passed
to qemu_thread_start(). We do this with the pthread_cleanup_push()
API.

We take advantage of the qemu_thread_atexit_add() API
permission not to run thread notifiers on process exit to
avoid having to special case the main thread.
Suggested-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NEric Blake <eblake@redhat.com>
Message-Id: <20181105135538.28025-3-peter.maydell@linaro.org>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Add documentation for the qemu_thread_atexit_add() and
qemu_thread_atexit_remove() functions.

We include a (previously undocumented) constraint that notifiers
may not be called if a thread is exiting because the entire
process is exiting. This is fine for our current use because
the callers use it only for cleaning up resources which go away
on process exit (memory, Win32 fibers), and we will need the
flexibility for the new posix implementation.
Signed-off-by: NPeter Maydell <peter.maydell@linaro.org>
Reviewed-by: NEric Blake <eblake@redhat.com>
Message-Id: <20181105135538.28025-2-peter.maydell@linaro.org>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Pass other sense, such as UNIT_ATTENTION or BUSY, directly to the
guest.
Reported-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Emulation of the block limits VPD page called back into scsi-disk.c,
which however expected the request to be for a SCSIDiskState and
accessed a scsi-generic device outside the bounds of its struct
(namely to retrieve s->max_unmap_size and s->max_io_size).

To avoid this, move the emulation code to a separate function that
takes a new SCSIBlockLimits struct and marshals it into the VPD
response format.
Reported-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

A device can report an excessive number of VPD pages when asked for a
list; this can cause an out-of-bounds access to buf in
scsi_generic_set_vpd_bl_emulation.  It should not happen, but
it is technically not incorrect so handle it: do not check any byte
past the allocation length that was sent to the INQUIRY command.
Reported-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Block limits emulation is just placing 0xb0 as the final byte of the
VPD pages list.  However, VPD page numbers must be sorted, so change
that to an in-place insert.  Since I couldn't find any disk that triggered
the loop more than once, this was tested by adding manually 0xb1
at the end of the list and checking that 0xb0 was added before.
Reported-by: NMax Reitz <mreitz@redhat.com>
Reviewed-by: NMax Reitz <mreitz@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

While writing a message in 'lsi_do_msgin', message length value
in 'msg_len' could be invalid due to an invalid migration stream.
Add an assertion to avoid an out of bounds access, and reject
the incoming migration data if it contains an invalid message
length.

Discovered by Deja vu Security. Reported by Oracle.
Signed-off-by: NPrasad J Pandit <pjp@fedoraproject.org>
Message-Id: <20181026194314.18663-1-ppandit@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Recent patches have removed ram_device and nonvolatile RAM
from dump-guest-memory's output.  Do the same for dumps
that are extracted from a QEMU core file.
Reviewed-by: NMarc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

GuestPhysBlockList is currently used to produce dumps. Given the size
and the typical usage of NVDIMM for storage, they are not a good idea
to have in the dumps. We may want to have an extra dump option to
include them. For now, skip non-volatile regions.

The TCG memory clear function is going to use the GuestPhysBlockList
as well, and will thus skip NVDIMM for similar reasons.
Signed-off-by: NMarc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20181003114454.5662-4-marcandre.lureau@redhat.com>
Reviewed-by: NLaszlo Ersek <lersek@redhat.com>
Reviewed-by: NDavid Hildenbrand <david@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

qemu-system-x86_64 -machine pc,nvdimm -m 2G,slots=4,maxmem=16G -enable-kvm -monitor stdio -object memory-backend-file,id=mem1,share=on,mem-path=/tmp/foo,size=1G -device nvdimm,id=nvdimm1,memdev=mem1

HMP info mtree command reflects the flag with "nv-" prefix on memory type:

(qemu) info mtree
0000000100000000-000000013fffffff (prio 0, nv-i/o): alias nvdimm-memory @/objects/mem1 0000000000000000-000000003fffffff

(qemu) info mtree -f
0000000100000000-000000013fffffff (prio 0, nv-ram): /objects/mem1
Signed-off-by: NMarc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20181003114454.5662-3-marcandre.lureau@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Add a new flag to mark memory region that are used as non-volatile, by
NVDIMM for example. That bit is propagated down to the flat view, and
reflected in HMP info mtree with a "nv-" prefix on the memory type.

This way, guest_phys_blocks_region_add() can skip the NV memory
regions for dumps and TCG memory clear in a following patch.

Cc: dgilbert@redhat.com
Cc: imammedo@redhat.com
Cc: pbonzini@redhat.com
Cc: guangrong.xiao@linux.intel.com
Cc: mst@redhat.com
Cc: xiaoguangrong.eric@gmail.com
Signed-off-by: NMarc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20181003114454.5662-2-marcandre.lureau@redhat.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>

Fix the SYSCALL instruction in 64-bit (long mode). The RF flag
should be cleared in R11 as well as in the RFLAGS. Intel
and AMD CPUs behave same. AMD has this documented in the
APM vol 3.
Signed-off-by: NRoman Kapl <rka@sysgo.com>
Signed-off-by: NRudolf Marek <rudolf.marek@sysgo.com>
Message-Id: <20181019122449.26387-1-rka@sysgo.com>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>