A new shared header tcg-pool.inc.c adds new_pool_label,
for registering a tcg_target_ulong to be emitted after
the generated code, plus relocation data to install a
pointer to the data.
A new pointer is added to the TCGContext, so that we
dump the constant pool as data, not code.
Dispense with TCGBackendData, as it has never been used for more than
holding a single pointer. Use a define in the cpu/tcg-target.h to
signal requirement for TCGLabelQemuLdst, so that we can drop the no-op
tcg-be-null.h stubs. Rename tcg-be-ldst.h to tcg-ldst.inc.c.
tcg: Move USE_DIRECT_JUMP discriminator to tcg/cpu/tcg-target.h
Replace the USE_DIRECT_JUMP ifdef with a TCG_TARGET_HAS_direct_jump
boolean test. Replace the tb_set_jmp_target1 ifdef with an unconditional
function tb_target_set_jmp_target.
While we're touching all backends, add a parameter for tb->tc_ptr;
we're going to need it shortly for some backends.
Move tb_set_jmp_target and tb_add_jump from exec-all.h to cpu-exec.c.
This opens the possibility for TCG_TARGET_HAS_direct_jump to be
a runtime decision -- based on host cpu capabilities, the size of
code_gen_buffer, or a future debugging switch.
* remotes/ericb/tags/pull-nbd-2017-09-06:
nbd: Use new qio_channel_*_all() functions
io: Add new qio_channel_read{, v}_all_eof functions
io: Yield rather than wait when already in coroutine
iotests: blacklist 194 with the luks driver
iotests: rewrite 192 to use _launch_qemu to fix LUKS support
Peter Maydell [Thu, 7 Sep 2017 15:42:55 +0000 (16:42 +0100)]
Merge remote-tracking branch 'remotes/pmaydell/tags/pull-target-arm-20170907' into staging
target-arm:
* cleanups converting to DEFINE_PROP_LINK
* allwinner-a10: mark as not user-creatable
* initial patches working towards ARMv8M support
* implement generating aborts on memory transaction failures
* make BXJ behave correctly (ie not UNDEF) on ARMv6-and-later
* remotes/pmaydell/tags/pull-target-arm-20170907: (31 commits)
target/arm: Add Jazelle feature
target/arm: Implement new do_transaction_failed hook
hw/arm: Set ignore_memory_transaction_failures for most ARM boards
boards.h: Define new flag ignore_memory_transaction_failures
target/arm: Implement BXNS, and banked stack pointers
target/arm: Move regime_is_secure() to target/arm/internals.h
target/arm: Make CFSR register banked for v8M
target/arm: Make MMFAR banked for v8M
target/arm: Make CCR register banked for v8M
target/arm: Make MPU_CTRL register banked for v8M
target/arm: Make MPU_RNR register banked for v8M
target/arm: Make MPU_RBAR, MPU_RLAR banked for v8M
target/arm: Make MPU_MAIR0, MPU_MAIR1 registers banked for v8M
target/arm: Make VTOR register banked for v8M
nvic: Add NS alias SCS region
target/arm: Make CONTROL register banked for v8M
target/arm: Make FAULTMASK register banked for v8M
target/arm: Make PRIMASK register banked for v8M
target/arm: Make BASEPRI register banked for v8M
target/arm: Add MMU indexes for secure v8M
...
Peter Maydell [Thu, 7 Sep 2017 14:26:06 +0000 (15:26 +0100)]
Merge remote-tracking branch 'remotes/dgilbert/tags/pull-migration-20170906a' into staging
migration pull 2017-09-06
# gpg: Signature made Wed 06 Sep 2017 19:39:23 BST
# gpg: using RSA key 0x0516331EBC5BFDE7
# gpg: Good signature from "Dr. David Alan Gilbert (RH2) <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 45F5 C71B 4A0C B7FB 977A 9FA9 0516 331E BC5B FDE7
* remotes/dgilbert/tags/pull-migration-20170906a:
migration: dump str in migrate_set_state trace
snapshot/tests: Try loadvm twice
migration: Reset rather than destroy main_thread_load_event
runstate/migrate: Two more transitions
host-utils: Simplify pow2ceil()
host-utils: Proactively fix pow2floor(), switch to unsigned
xbzrle: Drop unused cache_resize()
migration: Report when bdrv_inactivate_all fails
Peter Maydell [Thu, 7 Sep 2017 13:34:25 +0000 (14:34 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/pull-tgt-20170906' into staging
tcg generic translate loop v15
# gpg: Signature made Wed 06 Sep 2017 17:02:31 BST
# gpg: using RSA key 0x64DF38E8AF7E215F
# gpg: Good signature from "Richard Henderson <[email protected]>"
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A 05C0 64DF 38E8 AF7E 215F
* remotes/rth/tags/pull-tgt-20170906: (32 commits)
target/arm: Perform per-insn cross-page check only for Thumb
target/arm: Split out thumb_tr_translate_insn
target/arm: Move ss check to init_disas_context
target/arm: [a64] Move page and ss checks to init_disas_context
target/arm: [tcg] Port to generic translation framework
target/arm: [tcg,a64] Port to disas_log
target/arm: [tcg] Port to disas_log
target/arm: [tcg,a64] Port to tb_stop
target/arm: [tcg] Port to tb_stop
target/arm: [tcg,a64] Port to translate_insn
target/arm: [tcg] Port to translate_insn
target/arm: [tcg,a64] Port to breakpoint_check
target/arm: [tcg,a64] Port to insn_start
target/arm: [tcg] Port to insn_start
target/arm: [tcg] Port to tb_start
target/arm: [tcg,a64] Port to init_disas_context
target/arm: [tcg] Port to init_disas_context
target/arm: [tcg] Port to DisasContextBase
target/i386: [tcg] Port to generic translation framework
target/i386: [tcg] Port to disas_log
...
This adds a feature bit indicating support of the (trivial) Jazelle
implementation if ARM_FEATURE_V6 is set or if the processor is arm926
or arm1026. This fixes the issue that any BXJ instruction will
result in an illegal_op. BXJ instructions will now check if the
architecture supports ARM_FEATURE_JAZELLE.
Peter Maydell [Thu, 7 Sep 2017 12:54:54 +0000 (13:54 +0100)]
hw/arm: Set ignore_memory_transaction_failures for most ARM boards
Set the MachineClass flag ignore_memory_transaction_failures
for almost all ARM boards. This means they retain the legacy
behaviour that accesses to unimplemented addresses will RAZ/WI
rather than aborting, when a subsequent commit adds support
for external aborts.
The exceptions are:
* virt -- we know that guests won't try to prod devices
that we don't describe in the device tree or ACPI tables
* mps2 -- this board was written to use unimplemented-device
for all the ranges with devices we don't yet handle
New boards should not set the flag, but instead be written
like the mps2.
Peter Maydell [Thu, 7 Sep 2017 12:54:54 +0000 (13:54 +0100)]
boards.h: Define new flag ignore_memory_transaction_failures
Define a new MachineClass field ignore_memory_transaction_failures.
If this is flag is true then the CPU will ignore memory transaction
failures which should cause the CPU to take an exception due to an
access to an unassigned physical address; the transaction will
instead return zero (for a read) or be ignored (for a write). This
should be set only by legacy board models which rely on the old
RAZ/WI behaviour for handling devices that QEMU does not yet model.
New board models should instead use "unimplemented-device" for all
memory ranges where the guest will attempt to probe for a device that
QEMU doesn't implement and a stub device is required.
We need this for ARM boards, where we're about to implement support for
generating external aborts on memory transaction failures. Too many
of our legacy board models rely on the RAZ/WI behaviour and we
would break currently working guests when their "probe for device"
code provoked an external abort rather than a RAZ.
Peter Maydell [Thu, 7 Sep 2017 12:54:54 +0000 (13:54 +0100)]
target/arm: Implement BXNS, and banked stack pointers
Implement the BXNS v8M instruction, which is like BX but will do a
jump-and-switch-to-NonSecure if the branch target address has bit 0
clear.
This is the first piece of code which implements "switch to the
other security state", so the commit also includes the code to
switch the stack pointers around, which is the only complicated
part of switching security state.
BLXNS is more complicated than just "BXNS but set the link register",
so we leave it for a separate commit.
Peter Maydell [Thu, 7 Sep 2017 12:54:54 +0000 (13:54 +0100)]
target/arm: Make CCR register banked for v8M
Make the CCR register banked if v8M security extensions are enabled.
This is slightly more complicated than the other "add banking"
patches because there is one bit in the register which is not
banked. We keep the live data in the NS copy of the register,
and adjust it on register reads and writes. (Since we don't
currently implement the behaviour that the bit controls, there
is nowhere else that needs to care.)
This patch includes the enforcement of the bits which are newly
RES1 in ARMv8M.
Peter Maydell [Thu, 7 Sep 2017 12:54:53 +0000 (13:54 +0100)]
target/arm: Make MPU_RBAR, MPU_RLAR banked for v8M
Make the MPU registers MPU_MAIR0 and MPU_MAIR1 banked if v8M security
extensions are enabled.
We can freely add more items to vmstate_m_security without
breaking migration compatibility, because no CPU currently
has the ARM_FEATURE_M_SECURITY bit enabled and so this
subsection is not yet used by anything.
Peter Maydell [Thu, 7 Sep 2017 12:54:53 +0000 (13:54 +0100)]
nvic: Add NS alias SCS region
For v8M the range 0xe002e000..0xe002efff is an alias region which
for secure accesses behaves like a NonSecure access to the main
SCS region. (For nonsecure accesses including when the security
extension is not implemented, it is RAZ/WI.)
Peter Maydell [Thu, 7 Sep 2017 12:54:52 +0000 (13:54 +0100)]
target/arm: Make FAULTMASK register banked for v8M
Make the FAULTMASK register banked if v8M security extensions are enabled.
Note that we do not yet implement the functionality of the new
AIRCR.PRIS bit (which allows the effect of the NS copy of FAULTMASK to
be restricted).
This patch includes the code to determine for v8M which copy
of FAULTMASK should be updated on exception exit; further
changes will be required to the exception exit code in general
to support v8M, so this is just a small piece of that.
The v8M ARM ARM introduces a notation where individual paragraphs
are labelled with R (for rule) or I (for information) followed
by a random group of subscript letters. In comments where we want
to refer to a particular part of the manual we use this convention,
which should be more stable across document revisions than using
section or page numbers.
Peter Maydell [Thu, 7 Sep 2017 12:54:52 +0000 (13:54 +0100)]
target/arm: Add MMU indexes for secure v8M
Now that MPU lookups can return different results for v8M
when the CPU is in secure vs non-secure state, we need to
have separate MMU indexes; add the secure counterparts
to the existing three M profile MMU indexes.
Peter Maydell [Thu, 7 Sep 2017 12:54:52 +0000 (13:54 +0100)]
target/arm: Add state field, feature bit and migration for v8M secure state
As the first step in implementing ARM v8M's security extension:
* add a new feature bit ARM_FEATURE_M_SECURITY
* add the CPU state field that indicates whether the CPU is
currently in the secure state
* add a migration subsection for this new state
(we will add the Secure copies of banked register state
to this subsection in later patches)
* add a #define for the one new-in-v8M exception type
* make the CPU debug log print S/NS status
Peter Maydell [Thu, 7 Sep 2017 12:54:51 +0000 (13:54 +0100)]
target/arm: Implement ARMv8M's PMSAv8 registers
As part of ARMv8M, we need to add support for the PMSAv8 MPU
architecture.
PMSAv8 differs from PMSAv7 both in register/data layout (for instance
using base and limit registers rather than base and size) and also in
behaviour (for example it does not have subregions); rather than
trying to wedge it into the existing PMSAv7 code and data structures,
we define separate ones.
This commit adds the data structures which hold the state for a
PMSAv8 MPU and the register interface to it. The implementation of
the MPU behaviour will be added in a subsequent commit.
Thomas Huth [Thu, 7 Sep 2017 12:54:51 +0000 (13:54 +0100)]
hw/arm/allwinner-a10: Mark the allwinner-a10 device with user_creatable = false
QEMU currently exits unexpectedly when the user accidentially
tries to do something like this:
$ aarch64-softmmu/qemu-system-aarch64 -S -M integratorcp -nographic
QEMU 2.9.93 monitor - type 'help' for more information
(qemu) device_add allwinner-a10
Unsupported NIC model: smc91c111
Exiting just due to a "device_add" should not happen. Looking closer
at the the realize and instance_init function of this device also
reveals that it is using serial_hds and nd_table directly there, so
this device is clearly not creatable by the user and should be marked
accordingly.
Eric Blake [Tue, 5 Sep 2017 19:11:14 +0000 (14:11 -0500)]
nbd: Use new qio_channel_*_all() functions
Rather than open-coding our own read/write-all functions, we
can make use of the recently-added qio code. It slightly
changes the error message in one of the iotests.
Eric Blake [Tue, 5 Sep 2017 19:11:13 +0000 (14:11 -0500)]
io: Add new qio_channel_read{, v}_all_eof functions
Some callers want to distinguish between clean EOF (no bytes read)
vs. a short read (at least one byte read, but EOF encountered
before reaching the desired length), as it allows clients the
ability to do a graceful shutdown when a server shuts down at
defined safe points in the protocol, rather than treating all
shutdown scenarios as an error due to EOF. However, we don't want
to require all callers to have to check for early EOF. So add
another wrapper function that can be used by the callers that care
about the distinction.
Eric Blake [Tue, 5 Sep 2017 19:11:12 +0000 (14:11 -0500)]
io: Yield rather than wait when already in coroutine
The new qio_channel_{read,write}{,v}_all functions are documented
as yielding until data is available. When used on a blocking
channel, this yield is done via qio_channel_wait() which spawns
a nested event loop under the hood (so it is that secondary loop
which yields as needed); but if we are already in a coroutine (at
which point QIO_CHANNEL_ERR_BLOCK is only possible if we are a
non-blocking channel), we want to yield the current coroutine
instead of spawning a nested event loop.
Reviewed-by: Emilio G. Cota <[email protected]> Signed-off-by: Lluís Vilanova <[email protected]>
Message-Id: <150002073981.22386.9870422422367410100[email protected]>
[rth: Moved max_insns adjustment from tb_start to init_disas_context.
Removed pc_next return from translate_insn.
Removed tcg_check_temp_count from generic loop.
Moved gen_io_end to exactly match gen_io_start.
Use qemu_log instead of error_report for temporary leaks.
Moved TB size/icount assignments before disas_log.] Signed-off-by: Richard Henderson <[email protected]>
There's nothing magic about the exception that we generate in order
to execute the magic kernel page. We can and should allow gdb to
set a breakpoint at this location.
Fold DISAS_EXC and DISAS_TB_JUMP into DISAS_NORETURN.
In both cases all following code is dead. In the first
case because we have exited the TB via exception; in the
second case because we have exited the TB via goto_tb
and its associated machinery.
This target is not sophisticated in its use of cleanups at the
end of the translation loop. For the most part, any condition
that exits the TB is dealt with by emitting the exiting opcode
right then and there. Therefore the only is_jmp indicator that
is needed is DISAS_NORETURN.
For two stack segment modifying cases, we have not yet exited
the TB (therefore DISAS_NORETURN feels wrong), but intend to exit.
The caller of gen_movl_seg_T0 currently checks for any non-zero
value, therefore DISAS_TOO_MANY seems acceptable for that usage.
migration: Reset rather than destroy main_thread_load_event
migration_incoming_state_destroy doesn't really destroy, it cleans up.
After a loadvm it's called, but the loadvm command can be run twice,
and so destroying an init-once mutex breaks on the second loadvm.
host-utils: Proactively fix pow2floor(), switch to unsigned
The function's stated contract is simple enough: "round down to the
nearest power of 2". Suggests the domain is the representable numbers
>= 1, because that's the smallest power of two.
The implementation doesn't check for domain errors, but returns
garbage instead:
* For negative arguments, pow2floor() returns -2^63, which is not even
a power of two, let alone the nearest one.
What sort of works is passing *unsigned* arguments >= 2^63. The
implicit conversion to signed is implementation defined, but
commonly yields the (negative) two's complement. pow2floor() then
returns -2^63. Callers that convert that back to unsigned get the
correct value 2^63.
* For a zero argument, pow2floor() shifts right by 64. Undefined
behavior. Common actual behavior is to shift by 0, yielding -2^63.
Fix by switching from int64_t to uint64_t and amending the contract to
map zero to zero.
Callers are fine with that:
* memory_access_size()
This function makes no sense unless the argument is positive and the
return value fits into int.
* raw_refresh_limits()
Passes an int between 1 and BDRV_REQUEST_MAX_BYTES.
* iscsi_refresh_limits()
Passes an integer between 0 and INT_MAX, converts the result to
uint32_t. Passing zero would be undefined behavior, but commonly
yield zero. The patch gives us the zero without the undefined
behavior.
* cache_init()
Passes a positive int64_t argument.
* xbzrle_cache_resize()
Passes a positive int64_t argument (>= TARGET_PAGE_SIZE, actually).
* spapr_node0_size()
Passes a positive uint64_t argument, and converts the result to
hwaddr, i.e. uint64_t.
* spapr_populate_memory()
Passes a positive hwaddr argument, and converts the result to
hwaddr.
If the bdrv_inactivate_all fails near the end of the migration,
the migration will fail and often the only diagnostics in the log
are an I/O error which you can't distinguish from an error on
the socket connection.
Add an error so we know when it's actually a block problem.
Pavel Butsykin [Mon, 4 Sep 2017 10:18:00 +0000 (13:18 +0300)]
qcow2: move qcow2_store_persistent_dirty_bitmaps() before cache flushing
After calling qcow2_inactivate(), all qcow2 caches must be flushed, but this
may not happen, because the last call qcow2_store_persistent_dirty_bitmaps()
can lead to marking l2/refcont cache as dirty.
Let's move qcow2_store_persistent_dirty_bitmaps() before the caсhe flushing
to fix it.
block/throttle.c uses existing I/O throttle infrastructure inside a
block filter driver. I/O operations are intercepted in the filter's
read/write coroutines, and referred to block/throttle-groups.c
The driver can be used with the syntax
-drive driver=throttle,file.filename=foo.qcow2,throttle-group=bar
which registers the throttle filter node with the ThrottleGroup 'bar'. The
given group must be created beforehand with object-add or -object.
projects / qemu.git / log