This patch fixes decrement of the pointers for subx mem, mem instructions.
Without the patch pointers are decremented by OS_* constant value instead of
retrieving the corresponding data size and using it as a decrement.
Backports commit 355d4d1c00e708907ff391c24ca708f1c9c06bf0 from qemu
This is a bug fix to ensure 64-bit reads of these registers don't read
adjacent data.
Backports commit e4e91a217c17fff4045dd4b423cdcb471b3d6a0e from qemu
Because the design of the PMU requires that the counter values be
converted between their delta and guest-visible forms for mode
filtering, an additional hook which occurs before the EL is changed is
necessary.
Backports commit b5c53d1b3886387874f8c8582b205aeb3e4c3df6 from qemu
This eliminates the need for fetching it from el_change_hook_opaque, and
allows for supporting multiple el_change_hooks without having to hack
something together to find the registered opaque belonging to GICv3.
Backports commit d5a5e4c93dae0dc3feb402cf7ee78d846da1a7e1 from qemu
In commit 95695effe8caa552b8f2 we changed the v7M/v8M stack
pop code to use a new v7m_stack_read() function that checks
whether the read should fail due to an MPU or bus abort.
We missed one call though, the one which reads the signature
word for the callee-saved register part of the frame.
Correct the omission.
Backports commit 4818bad98c8212fbbb0525d10761b6b65279ab92 from qemu
Remove a stale TODO comment -- we have now made the arm_ldl_ptw()
and arm_ldq_ptw() functions propagate physical memory read errors
out to their callers.
Backports commit 145772707fe80395b87c244ccf5699a756f1946b from qemu
Without bounding the increment, we can overflow exp either here
in scalbn_decomposed or when adding the bias in round_canonical.
This can result in e.g. underflowing to 0 instead of overflowing
to infinity.
The old softfloat code did bound the increment.
Backports commit ce8d4082054519f2eaac39958edde502860a7fc6 from qemu
The re-factoring of div_floats changed the order of checking meaning
an operation like -inf/0 erroneously raises the divbyzero flag.
IEEE-754 (2008) specifies this should only occur for operations on
finite operands.
We fix this by moving the check on the dividend being Inf/0 to before
the divisor is zero check.
Backports commit 9cb4e398c2f95c1e837fe9c570e124a55259f725 from qemu
68000 CPUs do not save format in the exception stack frame.
This patch adds feature checking to prevent format saving for 68000.
m68k_ret() already includes this modification, this patch fixes
the exception processing function too.
Backports commit 000761dc0c97d70e7314db3e8f52783880325a22 from qemu
The re-factor broke the raising of INVALID when NaN/Inf is passed to
the float_to_int conversion functions. round_to_uint_and_pack got this
right for NaN but also missed out the Inf handling.
Fixes https://bugs.launchpad.net/qemu/+bug/1759264
Backports commit 801bc56336a127d9b351b3a2cc0336e4d0cb2686 from qemu
Before 8936006 ("fpu/softfloat: re-factor minmax", 2018-02-21),
we used to return +Zero for maxnummag(-Zero,+Zero); after that
commit, we return -Zero.
Fix it by making {min,max}nummag consistent with {min,max}num,
deferring to the latter when the absolute value of the operands
is the same.
With this fix we now pass fp-test.
Backports commit 6245327a367292b354489c54e965646823023919 from qemu
The MIPS TCG target makes the assumption that the offset from the
target env pointer to the tlb_table is less than about 64K. This
used to be true, but gradual addition of features to the Arm
target means that it's no longer true there. This results in
the build-time assertion failing:
In file included from /home/pm215/qemu/include/qemu/osdep.h:36:0,
from /home/pm215/qemu/tcg/tcg.c:28:
/home/pm215/qemu/tcg/mips/tcg-target.inc.c: In function ‘tcg_out_tlb_load’:
/home/pm215/qemu/include/qemu/compiler.h:90:36: error: static assertion failed: "not expecting: offsetof(CPUArchState, tlb_table[NB_MMU_MODES - 1][1]) > 0x7ff0 + 0x7fff"
^
/home/pm215/qemu/include/qemu/compiler.h:98:30: note: in expansion of macro ‘QEMU_BUILD_BUG_MSG’
^
/home/pm215/qemu/tcg/mips/tcg-target.inc.c:1236:9: note: in expansion of macro ‘QEMU_BUILD_BUG_ON’
QEMU_BUILD_BUG_ON(offsetof(CPUArchState,
^
/home/pm215/qemu/rules.mak:66: recipe for target 'tcg/tcg.o' failed
An ideal long term approach would be to rearrange the CPU state
so that the tlb_table was not so far along it, but this is tricky
because it would move it from the "not cleared on CPU reset" part
of the struct to the "cleared on CPU reset" part. As a simple fix
for the 2.12 release, make the MIPS TCG target handle an arbitrary
offset by emitting more add instructions. This will mean an extra
instruction in the fastpath for TCG loads and stores for the
affected guests (currently just aarch64-softmmu)
Backports commit 161dfd1e7fad1203840c0390f235030eba3fd23c from qemu
In icount mode, instructions that access io memory spaces in the middle
of the translation block invoke TB recompilation. After recompilation,
such instructions become last in the TB and are allowed to access io
memory spaces.
When the code includes instruction like i386 'xchg eax, 0xffffd080'
which accesses APIC, QEMU goes into an infinite loop of the recompilation.
This instruction includes two memory accesses - one read and one write.
After the first access, APIC calls cpu_report_tpr_access, which restores
the CPU state to get the current eip. But cpu_restore_state_from_tb
resets the cpu->can_do_io flag which makes the second memory access invalid.
Therefore the second memory access causes a recompilation of the block.
Then these operations repeat again and again.
This patch moves resetting cpu->can_do_io flag from
cpu_restore_state_from_tb to cpu_loop_exit* functions.
It also adds a parameter for cpu_restore_state which controls restoring
icount. There is no need to restore icount when we only query CPU state
without breaking the TB. Restoring it in such cases leads to the
incorrect flow of the virtual time.
In most cases new parameter is true (icount should be recalculated).
But there are two cases in i386 and openrisc when the CPU state is only
queried without the need to break the TB. This patch fixes both of
these cases.
Backports commit afd46fcad2dceffda35c0586f5723c127b6e09d8 from qemu
We are still seeing signals during translation time when we walk over
a page protection boundary. This expands the check to ensure the host
PC is inside the code generation buffer. The original suggestion was
to check versus tcg_ctx.code_gen_ptr but as we now segment the
translation buffer we have to settle for just a general check for
being inside.
I've also fixed up the declaration to make it clear it can deal with
invalid addresses. A later patch will fix up the call sites.
Backports commit d25f2a72272b9ffe0d06710d6217d1169bc2cc7d from qemu
The parameters for tcg_gen_insn_start are target_ulong, which may be split
into two TCGArg parameters for storage in the opcode on 32-bit hosts.
Fixes the ARM target and its direct use of tcg_set_insn_param, which would
set the wrong argument in the 64-on-32 case.
Backports commit 9743cd5736263e90d312b2c33bd739ffe1eae70d from qemu
Currently our PMSAv7 and ARMv7M MPU implementation cannot handle
MPU region sizes smaller than our TARGET_PAGE_SIZE. However we
report that in a slightly confusing way:
DRSR[3]: No support for MPU (sub)region alignment of 9 bits. Minimum is 10
The problem is not the alignment of the region, but its size;
tweak the error message to say so:
DRSR[3]: No support for MPU (sub)region size of 512 bytes. Minimum is 1024.
Backports commit 8aec759b45fa6986c0b159cb27353d6abb0d5d73 from qemu
Make sure we are not treating architecturally Undefined instructions
as a SWP, by verifying the opcodes as per section A8.8.229 of ARMv7-A
specification. Bits [21:20] must be zero for this to be a SWP or SWPB.
We also choose to UNDEF for the architecturally UNPREDICTABLE case of
bits [11:8] not being zero.
Backports commit c4869ca630a57f4269bb932ec7f719cef5bc79b8 from qemu
In commit 7073fbada733c8d10992f00772c9b9299d740e9b, the `andn` instruction
was implemented via `tcg_gen_andc` but passes the operands in the wrong
order:
- X86 defines `andn dest,src1,src2` as: dest = ~src1 & src2
- TCG defines `andc dest,src1,src2` as: dest = src1 & ~src2
The following simple test shows the issue:
int main(void) {
uint32_t ret = 0;
__asm (
"mov $0xFF00, %%ecx\n"
"mov $0x0F0F, %%eax\n"
"andn %%ecx, %%eax, %%ecx\n"
"mov %%ecx, %0\n"
: "=r" (ret));
printf("%08X\n", ret);
return 0;
}
This patch fixes the problem by simply swapping the order of the two last
arguments in `tcg_gen_andc_tl`.
Backports commit 5cd10051c2e02b7a86eae49919d6c65a87dbea46 from qemu
A mistake in the type passed to sizeof, that happens to work
when the out-of-line fallback itself is using host vectors,
but fails when using only the base types.
Backports commit 6cb1d3b8517572031a22675280ec642972cdb395 from qemu
configure tries to detect if the compiler
supports 16-byte vector operations.
As stated in the comment of the detection
program, there is a problem with the system
compiler on GCC on Centos 7.
This program doesn't actually detect the problem
with GCC on RHEL7 on PPC64LE (Red Hat 4.8.5-28).
This patch updates the test to look more like
it is in QEMU helpers, and now detects the problem.
The error reported is:
CC ppc64-softmmu/accel/tcg/tcg-runtime-gvec.o
..//accel/tcg/tcg-runtime-gvec.c: In function ‘helper_gvec_shl8i’:
../accel/tcg/tcg-runtime-gvec.c:558:26: internal compiler error: in emit_move_insn, at expr.c:3495
*(vec8 *)(d + i) = *(vec8 *)(a + i) << shift;
^
Fixes: db43267 "tcg: Add generic vector expanders"
Backports commit 74912f6dad207bfdd5bfa8c6012c364ea9405fc7 from qemu
Failure to do so results in the tcg optimizer sign-extending
any constant fold from 32-bits. This turns out to be visible
in the RISC-V testsuite using a host that emits these opcodes
(e.g. any non-x86_64).
Backports commit f2f1dde75160cac6ede330f3db50dc817d01a2d6 from qemu
For debug exceptions due to breakpoints or the BKPT instruction which
are taken to AArch32, the Fault Address Register is architecturally
UNKNOWN. We were using that as license to simply not set
env->exception.vaddress, but this isn't correct, because it will
expose to the guest whatever old value was in that field when
arm_cpu_do_interrupt_aarch32() writes it to the guest IFSR. That old
value might be a FAR for a previous guest EL2 or secure exception, in
which case we shouldn't show it to an EL1 or non-secure exception
handler. It might also be a non-deterministic value, which is bad
for record-and-replay.
Clear env->exception.vaddress before taking breakpoint debug
exceptions, to avoid this minor information leak.
Backports commit 548f514cf89dd9ab39c0cb4c063097bccf141fdd from qemu
Now that we have a helper function specifically for the BRK and
BKPT instructions, we can set the exception.fsr there rather
than in arm_cpu_do_interrupt_aarch32(). This allows us to
use our new arm_debug_exception_fsr() helper.
In particular this fixes a bug where we were hardcoding the
short-form IFSR value, which is wrong if the target exception
level has LPAE enabled.
Fixes: https://bugs.launchpad.net/qemu/+bug/1756927
Backports commit 62b94f31d0df75187bb00684fc29e8639eacc0c5 from qemu
When a debug exception is taken to AArch32, it appears as a Prefetch
Abort, and the Instruction Fault Status Register (IFSR) must be set.
The IFSR has two possible formats, depending on whether LPAE is in
use. Factor out the code in arm_debug_excp_handler() which picks
an FSR value into its own utility function, update it to use
arm_fi_to_lfsc() and arm_fi_to_sfsc() rather than hard-coded constants,
and use the correct condition to select long or short format.
In particular this fixes a bug where we could select the short
format because we're at EL0 and the EL1 translation regime is
not using LPAE, but then route the debug exception to EL2 because
of MDCR_EL2.TDE and hand EL2 the wrong format FSR.
Backports commit 81621d9ab8a0f07956e67850b15eebf6d6992eec from qemu
The MDCR_EL2.TDE bit allows the exception level targeted by debug
exceptions to be set to EL2 for code executing at EL0. We handle
this in the arm_debug_target_el() function, but this is only used for
hardware breakpoint and watchpoint exceptions, not for the exception
generated when the guest executes an AArch32 BKPT or AArch64 BRK
instruction. We don't have enough information for a translate-time
equivalent of arm_debug_target_el(), so instead make BKPT and BRK
call a special purpose helper which can do the routing, rather than
the generic exception_with_syndrome helper.
Backports commit c900a2e62dd6dde11c8f5249b638caad05bb15be from qemu
In OE project 4.15 linux kernel boot hang was observed under
single cpu aarch64 qemu. Kernel code was in a loop waiting for
vtimer arrival, spinning in TC generated blocks, while interrupt
was pending unprocessed. This happened because when qemu tried to
handle vtimer interrupt target had interrupts disabled, as
result flag indicating TCG exit, cpu->icount_decr.u16.high,
was cleared but arm_cpu_exec_interrupt function did not call
arm_cpu_do_interrupt to process interrupt. Later when target
reenabled interrupts, it happened without exit into main loop, so
following code that waited for result of interrupt execution
run in infinite loop.
To solve the problem instructions that operate on CPU sys state
(i.e enable/disable interrupt), and marked as DISAS_UPDATE,
should be considered as DISAS_EXIT variant, and should be
forced to exit back to main loop so qemu will have a chance
processing pending CPU state updates, including pending
interrupts.
This change brings consistency with how DISAS_UPDATE is treated
in aarch32 case.
Backports commit a75a52d62418dafe462be4fe30485501d1010bb9 from qemu
Currently CPUState::cpu_index is monotonically increasing and a newly
created CPU always gets the next higher index. The next available
index is calculated by counting the existing number of CPUs. This is
fine as long as we only add CPUs, but there are architectures which
are starting to support CPU removal, too. For an architecture like PowerPC
which derives its CPU identifier (device tree ID) from cpu_index, the
existing logic of generating cpu_index values causes problems.
With the currently proposed method of handling vCPU removal by parking
the vCPU fd in QEMU
(Ref: http://lists.gnu.org/archive/html/qemu-devel/2015-02/msg02604.html),
generating cpu_index this way will not work for PowerPC.
This patch changes the way cpu_index is handed out by maintaining
a bit map of the CPUs that tracks both addition and removal of CPUs.
The CPU bitmap allocation logic is part of cpu_exec_init(), which is
called by instance_init routines of various CPU targets. Newly added
cpu_exec_exit() API handles the deallocation part and this routine is
called from generic CPU instance_finalize.
Note: This new CPU enumeration is for !CONFIG_USER_ONLY only.
CONFIG_USER_ONLY continues to have the old enumeration logic.
Backports commit b7bca7333411bd19c449147e8202ae6b0e4a8e09 from qemu
Add an Error argument to cpu_exec_init() to let users collect the
error. This is in preparation to change the CPU enumeration logic
in cpu_exec_init(). With the new enumeration logic, cpu_exec_init()
can fail if cpu_index values corresponding to max_cpus have already
been handed out.
Since all current callers of cpu_exec_init() are from instance_init,
use error_abort Error argument to abort in case of an error.
Backports commit 5a790cc4b942e651fec7edc597c19b637fad5a76 from qemu
SRC_EA() and gen_extend() can return either a temporary
TCGv or a memory allocated one. Mark them when they are
allocated, and free them automatically at end of the
instruction translation.
We want to free locally allocated TCGv to avoid
overflow in sequence like:
0xc00ae406: movel %fp@(-132),%fp@(-268)
0xc00ae40c: movel %fp@(-128),%fp@(-264)
0xc00ae412: movel %fp@(-20),%fp@(-212)
0xc00ae418: movel %fp@(-16),%fp@(-208)
0xc00ae41e: movel %fp@(-60),%fp@(-220)
0xc00ae424: movel %fp@(-56),%fp@(-216)
0xc00ae42a: movel %fp@(-124),%fp@(-252)
0xc00ae430: movel %fp@(-120),%fp@(-248)
0xc00ae436: movel %fp@(-12),%fp@(-260)
0xc00ae43c: movel %fp@(-8),%fp@(-256)
0xc00ae442: movel %fp@(-52),%fp@(-276)
0xc00ae448: movel %fp@(-48),%fp@(-272)
...
That can fill a lot of TCGv entries in a sequence,
especially since 15fa08f845 ("tcg: Dynamically allocate TCGOps")
we have no limit to fill the TCGOps cache and we can fill
the entire TCG variables array and overflow it.
Backports commit ecc207d2fc1d45fabb16c38742a6675a7ba56cbc from qemu
Intel processor trace should be disabled when
CPUID.(EAX=14H,ECX=0H).ECX.[bit31] is set.
Generated packets which contain IP payloads will have LIP
values when this bit is set, or IP payloads will have RIP
values.
Currently, The information of CPUID 14H is constant to make
live migration safty and this bit is always 0 in guest even
if host support LIP values.
Guest sees the bit is 0 will expect IP payloads with RIP
values, but the host CPU will generate IP payloads with
LIP values if this bit is set in HW.
To make sure the value of IP payloads correctly, Intel PT
should be disabled when bit[31] is set.
Backports relevant parts of commit c078ca968c6c7cb62781c1843d840cb0f5c72781 from qemu
both do nothing as for the first all callers
parse_cpu_model() and qmp_query_cpu_model_()
should provide non NULL value, so just abort if it's not so.
While at it drop cpu_common_class_by_name() which is not need
any more as every target has CPUClass::class_by_name callback
by now, though abort in case a new arch will forget to define one.
Backports commit 99193d8f2ef594648ad67cc3d007b0e4fb2f8cf8 from qemu
cpu_init(cpu_model) were replaced by cpu_create(cpu_type) so
no users are left, remove it.
Backports commit 3f71e724e283233753f1b5b3d6a30948d3084636 from qemu
With all targets defining CPU_RESOLVING_TYPE, refactor
cpu_parse_cpu_model(type, cpu_model) to parse_cpu_model(cpu_model)
so that callers won't have to know internal resolving cpu
type. Place it in exec.c so it could be called from both
target independed vl.c and *-user/main.c.
That allows us to stop abusing cpu type from
MachineClass::default_cpu_type
as resolver class in vl.c which were confusing part of
cpu_parse_cpu_model().
Also with new parse_cpu_model(), the last users of cpu_init()
in null-machine.c and bsd/linux-user targets could be switched
to cpu_create() API and cpu_init() API will be removed by
follow up patch.
With no longer users left remove MachineState::cpu_model field,
new code should use MachineState::cpu_type instead and
leave cpu_model parsing to generic code in vl.c.
Backports commit 2278b93941d42c30e2950d4b8dff4943d064e7de from qemu
there are 2 use cases to deal with:
1: fixed CPU models per board/soc
2: boards with user configurable cpu_model and fallback to
default cpu_model if user hasn't specified one explicitly
For the 1st
drop intermediate cpu_model parsing and use const cpu type
directly, which replaces:
typename = object_class_get_name(
cpu_class_by_name(TYPE_ARM_CPU, cpu_model))
object_new(typename)
with
object_new(FOO_CPU_TYPE_NAME)
or
cpu_generic_init(BASE_CPU_TYPE, "my cpu model")
with
cpu_create(FOO_CPU_TYPE_NAME)
as result 1st use case doesn't have to invoke not necessary
translation and not needed code is removed.
For the 2nd
1: set default cpu type with MachineClass::default_cpu_type and
2: use generic cpu_model parsing that done before machine_init()
is run and:
2.1: drop custom cpu_model parsing where pattern is:
typename = object_class_get_name(
cpu_class_by_name(TYPE_ARM_CPU, cpu_model))
[parse_features(typename, cpu_model, &err) ]
2.2: or replace cpu_generic_init() which does what
2.1 does + create_cpu(typename) with just
create_cpu(machine->cpu_type)
as result cpu_name -> cpu_type translation is done using
generic machine code one including parsing optional features
if supported/present (removes a bunch of duplicated cpu_model
parsing code) and default cpu type is defined in an uniform way
within machine_class_init callbacks instead of adhoc places
in boadr's machine_init code.
Backports commit ba1ba5cca3962a9cc400c713c736b4fb8db1f38e from qemu
define default CPU type in generic way in pc_machine_class_init()
and let common machine code to handle cpu_model parsing
Patch also introduces TARGET_DEFAULT_CPU_TYPE define for 2 purposes:
* make foo_machine_class_init() look uniform on every target
* use define in [bsd|linux]-user targets to pick default
cpu type
Backports commit 311ca98d16bbb6a2a38b38ba898baa4a4d4ab9a7 from qemu
All machines that support user specified cpu_model either call
cpu_generic_init() or cpu_class_by_name()/CPUClass::parse_features
to parse feature string and to get CPU type to create.
Which leads to code duplication and hard-codding default CPU model
within machine_foo_init() code. Which makes it impossible to
get CPU type before machine_init() is run.
So instead of setting default CPUs models and doing parsing in
target specific machine_foo_init() in various ways, provide
a generic data driven cpu_model parsing before machine_init()
is called.
in follow up per target patches, it will allow to:
* define default CPU type in consistent/generic manner
per machine type and drop custom code that fallbacks
to default if cpu_model is NULL
* drop custom features parsing in targets and do it
in centralized way.
* for cases of
cpu_generic_init(TYPE_BASE/DEFAULT_CPU, "some_cpu")
replace it with
cpu_create(machine->cpu_type) || cpu_create(TYPE_FOO)
depending if CPU type is user settable or not.
not doing useless parsing and clearly documenting where
CPU model is user settable or fixed one.
Patch allows machine subclasses to define default CPU type
per machine class at class_init() time and if that is set
generic code will parse cpu_model into a MachineState::cpu_type
which will be used to create CPUs for that machine instance
and allows gradual per board conversion.
Backports commit 6063d4c0f98b35a27ca018393d328a1825412a7e from qemu
