Bus lock debug exception is a feature that can notify the kernel by
generate an #DB trap after the instruction acquires a bus lock when
CPL>0. This allows the kernel to enforce user application throttling or
mitigations.
This feature is enumerated via CPUID.(EAX=7,ECX=0).ECX[bit 24].
Backports 06e878b413766778a53be3d25c0373a23679d039
We were fudging TBI1 enabled to speed up the generated code.
Now that we've improved the code generation, remove this.
Also, tidy the comment to reflect the current code.
The pauth test was testing a kernel address (-1) and making
incorrect assumptions about TBI1; stick to userland addresses.
Backports 16c849784873d10d0da257d698e391fddea1f0e4
When working with performance monitoring counters, we look at
MDCR_EL2.HPMN as part of the check whether a counter is enabled. This
check fails, because MDCR_EL2.HPMN is reset to 0, meaning that no
counters are "enabled" for < EL2.
That's in violation of the Arm specification, which states that
> On a Warm reset, this field [MDCR_EL2.HPMN] resets to the value in
> PMCR_EL0.N
That's also what a comment in the code acknowledges, but the necessary
adjustment seems to have been forgotten when support for more counters
was added.
This change fixes the issue by setting the reset value to PMCR.N, which
is four.
Backports d3c1183ffeb71ca3a783eae3d7e1c51e71e8a621
cpsr has been treated as being the same as spsr, but it isn't.
Since PSTATE_SS isn't in cpsr, remove it and move it into env->pstate.
This allows us to add support for CPSR_DIT, adding helper functions
to merge SPSR_ELx to and from CPSR.
Backports f944a854ce4007000accf7c191b5b52916947198
Add support for FEAT_DIT. DIT (Data Independent Timing) is a required
feature for ARMv8.4. Since virtual machine execution is largely
nondeterministic and TCG is outside of the security domain, it's
implemented as a NOP.
Backports dc8b18534ea1dcc90d80ad9a61a3b0aa7eb312fb
The FW and AW bits of SCR_EL3 are RES1 only in some contexts. Force them
to 1 only when there is no support for AArch32 at EL1 or above.
The reset value will be 0x30 only if the CPU is AArch64-only; if there
is support for AArch32 at EL1 or above, it will be reset to 0.
Also adds helper function isar_feature_aa64_aa32_el1 to check if AArch32
is supported at EL1 or above.
Backports 10d0ef3e6cfe228df4b2d3e27325f1b0e2b71fd5
Expose the VMX exit/entry load pkrs control bits in
VMX_TRUE_EXIT_CTLS/VMX_TRUE_ENTRY_CTLS MSRs to guest, which supports the
PKS in nested VM.
Backports 52a44ad2b92ba4cd81c2b271cd5e4a2d820e91fc
Protection Keys for Supervisor-mode pages is a simple extension of
the PKU feature that QEMU already implements. For supervisor-mode
pages, protection key restrictions come from a new MSR. The MSR
has no XSAVE state associated to it.
PKS is only respected in long mode. However, in principle it is
possible to set the MSR even outside long mode, and in fact
even the XSAVE state for PKRU could be set outside long mode
using XRSTOR. So do not limit the migration subsections for
PKRU and PKRS to long mode.
Backports e7e7bdababeefff10736c6adf410c66d2f0d46fe
This patch fixes a translation bug for a subset of x86 BMI instructions
such as the following:
c4 e2 f9 f7 c0 shlxq %rax, %rax, %rax
Currently, these incorrectly generate an undefined instruction exception
when SSE is disabled via CR4, while instructions like "shrxq" work fine.
The problem appears to be related to BMI instructions encoded using VEX
and with a mandatory prefix of "0x66" (data). Instructions with this
data prefix (such as shlxq) are currently rejected. Instructions with
other mandatory prefixes (such as shrxq) translate as expected.
This patch removes the incorrect check in "gen_sse" that causes the
exception to be generated. For the non-BMI cases, the check is
redundant: prefixes are already checked at line 3696.
Buglink: https://bugs.launchpad.net/qemu/+bug/1748296
Backports 51909241d26fe6fe18a08def93ccc8273f61a8b3
32-bit targets by definition do not support long mode; therefore, the
bit must be masked in the features supported by the accelerator.
As a side effect, this avoids setting up the 0x80000008 CPUID leaf
for
qemu-system-i386 -cpu host
which since commit 5a140b255d ("x86/cpu: Use max host physical address
if -cpu max option is applied") would have printed this error:
qemu-system-i386: phys-bits should be between 32 and 36 (but is 48)
Backports 5ea9e9e239db83391a39c09f1de63c4099c20df5
commit 568496c0c0f1 ("cpu: Add callback to check architectural") and
commit 3826121d9298 ("target-arm: Implement checking of fired")
introduced an ARM-specific hack for cpu_check_watchpoint.
Make debug_check_watchpoint optional, and move it to tcg_ops.
Backports c73bdb35a91fb6b17c2c93b1ba381fc88a406f8d
commit 40612000599e ("arm: Correctly handle watchpoints for BE32 CPUs")
introduced this ARM-specific, TCG-specific hack to adjust the address,
before checking it with cpu_check_watchpoint.
Make adjust_watchpoint_address optional and move it to tcg_ops.
Backports 9ea9087bb4a86893e4ac6ff643837937dc9e5849
The TCG-specific CPU methods will be moved to a separate struct,
to make it easier to move accel-specific code outside generic CPU
code in the future. Start by moving tcg_initialize().
The new CPUClass.tcg_opts field may eventually become a pointer,
but keep it an embedded struct for now, to make code conversion
easier.
Backports e9e51b7154404efc9af8735ab87c658a9c434cfd
cc->do_interrupt is in theory a TCG callback used in accel/tcg only,
to prepare the emulated architecture to take an interrupt as defined
in the hardware specifications,
but in reality the _do_interrupt style of functions in targets are
also occasionally reused by KVM to prepare the architecture state in a
similar way where userspace code has identified that it needs to
deliver an exception to the guest.
In the case of ARM, that includes:
1) the vcpu thread got a SIGBUS indicating a memory error,
and we need to deliver a Synchronous External Abort to the guest to
let it know about the error.
2) the kernel told us about a debug exception (breakpoint, watchpoint)
but it is not for one of QEMU's own gdbstub breakpoints/watchpoints
so it must be a breakpoint the guest itself has set up, therefore
we need to deliver it to the guest.
So in order to reuse code, the same arm_do_interrupt function is used.
This is all fine, but we need to avoid calling it using the callback
registered in CPUClass, since that one is now TCG-only.
Fortunately this is easily solved by replacing calls to
CPUClass::do_interrupt() with explicit calls to arm_do_interrupt().
Backports 853bfef4e6d60244fd131ec55bbf1e7caa52599b. We don't support
KVM, so we just bring the comment addition over.
This was defined at some point before ARMv8.4, and will
shortly be used by new processor descriptions.
Backports 1d51bc96cc4a9b2d31a3f4cb8442ce47753088e2
When building with GCC 10.2 configured with --extra-cflags=-Os, we get:
target/arm/m_helper.c: In function ‘arm_v7m_cpu_do_interrupt’:
target/arm/m_helper.c:1811:16: error: ‘restore_s16_s31’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
1811 | if (restore_s16_s31) {
| ^
target/arm/m_helper.c:1350:10: note: ‘restore_s16_s31’ was declared here
1350 | bool restore_s16_s31;
| ^~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
Initialize the 'restore_s16_s31' variable to silence the warning.
Backports 0ae4f11ee57350dac0e705ba79516310400ff43c
These two were odd, in that do_pfirst_pnext passed the
count of 64-bit words rather than bytes. Change to pass
the standard pred_full_reg_size to avoid confusion.
Backports 86300b5d044064046395ae8ed605cc19e63f2a7c
SVE predicate operations cannot use the "usual" simd_desc
encoding, because the lengths are not a multiple of 8.
But we were abusing the SIMD_* fields to store values anyway.
This abuse broke when SIMD_OPRSZ_BITS was modified in e2e7168a214.
Introduce a new set of field definitions for exclusive use
of predicates, so that it is obvious what kind of predicate
we are manipulating. To be used in future patches
Backports b64ee454a4a086ed459bcda4c0bbb54e197841e4
On ARMv8-A, accesses by 32-bit secure EL1 to monitor registers trap to
the upper (64-bit) EL. With Secure EL2 support, we can no longer assume
that that is always EL3, so make room for the value to be computed at
run-time.
Backports 6b340aeb48e4f7f983e1c38790de65ae93079840
The stage_1_mmu_idx() already effectively keeps track of which
translation regimes have two stages. Don't hard-code another test.
Backports 7879460a6149ed5e80c29cac85449191d9c5754a
In the secure stage 2 translation regime, the VSTCR.SW and VTCR.NSW
bits can invert the secure flag for pagetable walks. This patchset
allows S1_ptw_translate() to change the non-secure bit.
Backports 3d4bd397433b12b148d150c8bc5655a696389bd1
The VTTBR write callback so far assumes that the underlying VM lies in
non-secure state. This handles the secure state scenario.
backports c4f060e89effd70ebdb23d3315495d33af377a09
This adds the MMU indices for EL2 stage 1 in secure state.
To keep code contained, which is largelly identical between secure and
non-secure modes, the MMU indices are reassigned. The new assignments
provide a systematic pattern with a non-secure bit.
Backports b6ad6062f1e55bd5b9407ce89e55e3a08b83827c
With the ARMv8.4-SEL2 extension, EL2 is a legal exception level in
secure mode, though it can only be AArch64.
This patch adds the target EL for exceptions from 64-bit S-EL2.
It also fixes the target EL to EL2 when HCR.{A,F,I}MO are set in secure
mode. Those values were never used in practice as the effective value of
HCR was always 0 in secure mode.
Backports 6c85f906261226e87211506bd9f787fd48a09f17
This adds a common helper to compute the effective value of MDCR_EL2.
That is the actual value if EL2 is enabled in the current security
context, or 0 elsewise.
Backports 59dd089cf9e4a9cddee596c8a1378620df51b9bb
Do not assume that EL2 is available in and only in non-secure context.
That equivalence is broken by ARMv8.4-SEL2.
Backports e6ef0169264b00cce552404f689ce137018ff290
In this context, the HCR value is the effective value, and thus is
zero in secure mode. The tests for HCR.{F,I}MO are sufficient.
Backports cc974d5cd84ea60a3dad59752aea712f3d47f8ce
The crypto overhead of emulating pauth can be significant for
some workloads. Add two boolean properties that allows the
feature to be turned off, on with the architected algorithm,
or on with an implementation defined algorithm.
We need two intermediate booleans to control the state while
parsing properties lest we clobber ID_AA64ISAR1 into an invalid
intermediate state.
Backports relevent members from eb94284d0812b4e7c11c5d075b584100ac1c1b9a
Without hardware acceleration, a cryptographically strong
algorithm is too expensive for pauth_computepac.
Even with hardware accel, we are not currently expecting
to link the linux-user binaries to any crypto libraries,
and doing so would generally make the --static build fail.
So choose XXH64 as a reasonably quick and decent hash.
Backports 283fc52ade85eb50141f3b8b85f82b07d016cb17
In commit cd8be50e58f63413c0 we converted the A32 coprocessor
insns to decodetree. This accidentally broke XScale/iWMMXt insns,
because it moved the handling of "cp insns which are handled
by looking up the cp register in the hashtable" from after the
call to the legacy disas_xscale_insn() decode to before it,
with the result that all XScale/iWMMXt insns now UNDEF.
Update valid_cp() so that it knows that on XScale cp 0 and 1
are not standard coprocessor instructions; this will cause
the decodetree trans_ functions to ignore them, so that
execution will correctly get through to the legacy decode again.
Backports e4d51ac6921dc861bfb3d20e4c7dcf345840a9da
When FEAT_MTE is implemented, the AArch64 view of CTR_EL0 adds the
TminLine field in bits [37:32].
Extend the ctr field to be able to hold this context.
Backports a5fd319ae7f6d496ff5448ec1dedcae8e2f59e9f
The AArch64 view of CLIDR_EL1 extends the ICB field to include also bit
32, as well as adding a Ttype<n> field when FEAT_MTE is implemented.
Extend the clidr field to be able to hold this context.
Backports f6450bcb6b2d3e4beae77141edce9e99cb8c277e
Now that we have implemented all the features needed by the v8.1M
architecture, we can add the model of the Cortex-M55. This is the
configuration without MVE support; we'll add MVE later
Backports 590e05d6b48937f6d3c631354fd706f8e005b8f6
Implement the v8.1M FPCXT_NS floating-point system register. This is
a little more complicated than FPCXT_S, because it has specific
handling for "current FP state is inactive", and it only wants to do
PreserveFPState(), not the full set of actions done by
ExecuteFPCheck() which vfp_access_check() implements.
Backports eb20dafdbff92063a88624176fdc396e01961bf3
In commit 64f863baeedc8659 we implemented the v8.1M FPCXT_S register,
but we got the write behaviour wrong. On read, this register reads
bits [27:0] of FPSCR plus the CONTROL.SFPA bit. On write, it doesn't
just write back those bits -- it writes a value to the whole FPSCR,
whose upper 4 bits are zeroes.
We also incorrectly implemented the write-to-FPSCR as a simple store
to vfp.xregs; this skips the "update the softfloat flags" part of
the vfp_set_fpscr helper so the value would read back correctly but
not actually take effect.
Fix both of these things by doing a complete write to the FPSCR
using the helper function.
Backports 7fbf95a037d79c5e923ffb51ac902dbe9599c87f
In 50244cc76abc we updated mte_check_fail to match the ARM
pseudocode, using the correct EL to select the TCF field.
But we failed to update MTE0_ACTIVE the same way, which led
to g_assert_not_reached().
Backports cc97b0019bb590b9b3c2a623e9ebee48831e0ce3
When the 'int N' instruction is executed in protected mode, the
pseudocode in the architecture manual specifies that we need to check:
* vector number within IDT limits
* selected IDT descriptor is a valid type (interrupt, trap or task gate)
* if this was a software interrupt then gate DPL < CPL
The way we had structured the code meant that the privilege check for
software interrupts ended up not in the code path taken for task gate
handling, because all of the task gate handling code was in the 'case 5'
of the switch which was checking "is this descriptor a valid type".
Move the task gate handling code out of that switch (so that it is now
purely doing the "valid type?" check) and below the software interrupt
privilege check.
The effect of this missing check was that in a guest userspace binary
executing 'int 8' would cause a guest kernel panic rather than the
userspace binary being handed a SEGV.
This is essentially the same bug fixed in VirtualBox in 2012:
https://www.halfdog.net/Security/2012/VirtualBoxSoftwareInterrupt0x8GuestCrash/
Note that for QEMU this is not a security issue because it is only
present when using TCG.
Backports 3df1a3d070575419859cbbab1083fafa7ec2669a
I found that there are many spelling errors in the comments of qemu/target/m68k.
I used spellcheck to check the spelling errors and found some errors in the folder.
Backports ce00ff729ee8461dc94a1593d25ceda65d973d3c
Per the SDM, when returning to outer privilege level, for segment
registers (ES, FS, GS, and DS) if the check fails, the segment
selector becomes null, but QEMU clears the base/limit/flags as well
as nullifying the segment selector, which should be a spec violation.
Real hardware seems to be compliant with the spec, at least on one
Coffee Lake board I tested.
Backports c2ba0515f2df58a661fcb5d6485139877d92ab1b
For PDEP and PEXT, the mask is provided in the memory (mod+r/m)
operand, and therefore is loaded in s->T0 by gen_ldst_modrm.
The source is provided in the second source operand (VEX.vvvv)
and therefore is loaded in s->T1. Fix the order in which
they are passed to the helpers.
Backports 75b208c28316095c4685e8596ceb9e3f656592e2
For v8.1M the architecture mandates that CPUs must provide at
least the "minimal RAS implementation" from the Reliability,
Availability and Serviceability extension. This consists of:
* an ESB instruction which is a NOP
-- since it is in the HINT space we need only add a comment
* an RFSR register which will RAZ/WI
* a RAZ/WI AIRCR.IESB bit
-- the code which handles writes to AIRCR does not allow setting
of RES0 bits, so we already treat this as RAZ/WI; add a comment
noting that this is deliberate
* minimal implementation of the RAS register block at 0xe0005000
-- this will be in a subsequent commit
* setting the ID_PFR0.RAS field to 0b0010
-- we will do this when we add the Cortex-M55 CPU model
Backports 46f4976f22a4549322307b34272e053d38653243
v8.1M introduces a new TRD flag in the CCR register, which enables
checking for stack frame integrity signatures on SG instructions.
Add the code in the SG insn implementation for the new behaviour.
Backports 7f484147369080d36c411c4ba969f90d025aed55
v8.1M adds new encodings of VLLDM and VLSTM (where bit 7 is set).
The only difference is that:
* the old T1 encodings UNDEF if the implementation implements 32
Dregs (this is currently architecturally impossible for M-profile)
* the new T2 encodings have the implementation-defined option to
read from memory (discarding the data) or write UNKNOWN values to
memory for the stack slots that would be D16-D31
We choose not to make those accesses, so for us the two
instructions behave identically assuming they don't UNDEF.
Backports fe6fa228a71f0eb8b8ee315452e6a7736c537b1f
In v8.1M a new exception return check is added which may cause a NOCP
UsageFault (see rule R_XLTP): before we clear s0..s15 and the FPSCR
we must check whether access to CP10 from the Security state of the
returning exception is disabled; if it is then we must take a fault.
(Note that for our implementation CPPWR is always RAZ/WI and so can
never cause CP10 accesses to fail.)
The other v8.1M change to this register-clearing code is that if MVE
is implemented VPR must also be cleared, so add a TODO comment to
that effect.
Backports 3423fbf10427db7680d3237d4f62d8370052fca0
In v8.0M, on exception entry the registers R0-R3, R12, APSR and EPSR
are zeroed for an exception taken to Non-secure state; for an
exception taken to Secure state they become UNKNOWN, and we chose to
leave them at their previous values.
In v8.1M the behaviour is specified more tightly and these registers
are always zeroed regardless of the security state that the exception
targets (see rule R_KPZV). Implement this.
Backports a59b1ed618415212c5f0f05abc1192e14ad5fdbb
Implement the new-in-v8.1M FPCXT_S floating point system register.
This is for saving and restoring the secure floating point context,
and it reads and writes bits [27:0] from the FPSCR and the
CONTROL.SFPA bit in bit [31].
Backports 64f863baeedc86590a608e2f1722dd8640aa9431
Factor out the code which handles M-profile lazy FP state preservation
from full_vfp_access_check(); accesses to the FPCXT_NS register are
a special case which need to do just this part (corresponding in the
pseudocode to the PreserveFPState() function), and not the full
set of actions matching the pseudocode ExecuteFPCheck() which
normal FP instructions need to do.
Backports 96dfae686628fc14ba4f993824322b93395e221b
We defined a constant name for the mask of NZCV bits in the FPCR/FPSCR
in the previous commit; use it in a couple of places in existing code,
where we're masking out everything except NZCV for the "load to Rt=15
sets CPSR.NZCV" special case.
Backports 6a017acdf83e3bb6bd5e85289ca90b2ea3282b7e
v8.1M defines a new FP system register FPSCR_nzcvqc; this behaves
like the existing FPSCR, except that it reads and writes only bits
[31:27] of the FPSCR (the N, Z, C, V and QC flag bits). (Unlike the
FPSCR, the special case for Rt=15 of writing the CPSR.NZCV is not
permitted.)
Implement the register. Since we don't yet implement MVE, we handle
the QC bit as RES0, with todo comments for where we will need to add
support later.
Backports 9542c30bcf13c495400d63616dd8dfa825b04685
Implement the new-in-v8.1M VLDR/VSTR variants which directly
read or write FP system registers to memory.
Backports 0bf0dd4dcbd9fab324700ac6e0cd061cd043de0d
The constant-expander functions like negate, plus_2, etc, are
generally useful; move them up in translate.c so we can use them in
the VFP/Neon decoders as well as in the A32/T32/T16 decoders.
Backports f7ed0c9433e7c5c157d2e6235eb5c8b93234a71a
Currently M-profile borrows the A-profile code for VMSR and VMRS
(access to the FP system registers), because all it needs to support
is the FPSCR. In v8.1M things become significantly more complicated
in two ways:
* there are several new FP system registers; some have side effects
on read, and one (FPCXT_NS) needs to avoid the usual
vfp_access_check() and the "only if FPU implemented" check
* all sysregs are now accessible both by VMRS/VMSR (which
reads/writes a general purpose register) and also by VLDR/VSTR
(which reads/writes them directly to memory)
Refactor the structure of how we handle VMSR/VMRS to cope with this:
* keep the M-profile code entirely separate from the A-profile code
* abstract out the "read or write the general purpose register" part
of the code into a loadfn or storefn function pointer, so we can
reuse it for VLDR/VSTR.
Backports 32a290b8c3c2dc85cd88bd8983baf900d575cab
For M-profile before v8.1M, the only valid register for VMSR/VMRS is
the FPSCR. We have a comment that states this, but the actual logic
to forbid accesses for any other register value is missing, so we
would end up with A-profile style behaviour. Add the missing check.
Backports ede97c9d71110821738a48f88ff9f10d6bec017f
In v8.1M the new CLRM instruction allows zeroing an arbitrary set of
the general-purpose registers and APSR. Implement this.
The encoding is a subset of the LDMIA T2 encoding, using what would
be Rn=0b1111 (which UNDEFs for LDMIA).
Backports 6e21a013fbdf54960a079dccc90772bb622e28e8
Implement the v8.1M VSCCLRM insn, which zeros floating point
registers if there is an active floating point context.
This requires support in write_neon_element32() for the MO_32
element size, so add it.
Because we want to use arm_gen_condlabel(), we need to move
the definition of that function up in translate.c so it is
before the #include of translate-vfp.c.inc.
Backports 83ff3d6add965c9752324de11eac5687121ea826
In arm_cpu_realizefn() we check whether the board code disabled EL3
via the has_el3 CPU object property, which we create if the CPU
starts with the ARM_FEATURE_EL3 feature bit. If it is disabled, then
we turn off ARM_FEATURE_EL3 and also zero out the relevant fields in
the ID_PFR1 and ID_AA64PFR0 registers.
This codepath was incorrectly being taken for M-profile CPUs, which
do not have an EL3 and don't set ARM_FEATURE_EL3, but which may have
the M-profile Security extension and so should have non-zero values
in the ID_PFR1.Security field.
Restrict the handling of the feature flag to A/R-profile cores.
Backports 4018818840f499d0a478508aedbb6802c8eae928
In v8.1M the PXN architecture extension adds a new PXN bit to the
MPU_RLAR registers, which forbids execution of code in the region
from a privileged mode.
This is another feature which is just in the generic "in v8.1M" set
and has no ID register field indicating its presence.
Backports cad8e2e3160dd10371552fce6cd8c6e171503e13
Using a target unsigned long would limit the Input Address to a LPAE
page-walk to 32 bits on AArch32 and 64 bits on AArch64. This is okay
for stage 1 or on AArch64, but it is insufficient for stage 2 on
AArch32. In that later case, the Input Address can have up to 40 bits.
Backports commit 98e8779770c40901ed585745aacc9a8e2b934a28
There is no "version 2" of the "Lesser" General Public License.
It is either "GPL version 2.0" or "Lesser GPL version 2.1".
This patch replaces all occurrences of "Lesser GPL version 2" with
"Lesser GPL version 2.1" in comment section.
Backport d9ff33ada7f32ca59f99b270a2d0eb223b3c9c8f
There is no "version 2" of the "Lesser" General Public License.
It is either "GPL version 2.0" or "Lesser GPL version 2.1".
This patch replaces all occurrences of "Lesser GPL version 2" with
"Lesser GPL version 2.1" in comment section.
Backports 50f57e09fda4b7ffbc5ba62aad6cebf660824023
Checks for UNDEF cases should go before the "is VFP enabled?" access
check, except in special cases. Move a stray UNDEF check in the VTBL
trans function up above the access check.
Backports b6c56c8a9a4064ea783f352f43c5df6231a110fa
The helper function did not get updated when we reorganized
the vector register file for SVE. Since then, the neon dregs
are non-sequential and cannot be simply indexed.
At the same time, make the helper function operate on 64-bit
quantities so that we do not have to call it twice.
Backports 604cef3e57eaeeef77074d78f6cf2eca1be11c62
Fix code style. Don't use '#' flag of printf format ('%#') in
format strings, use '0x' prefix instead
Backports 6eb55edbabb9eed1e4c7dfb233e7d738e8b5fa89
In arm_v7m_mmu_idx_for_secstate() we get the 'priv' level to pass to
armv7m_mmu_idx_for_secstate_and_priv() by calling arm_current_el().
This is incorrect when the security state being queried is not the
current one, because arm_current_el() uses the current security state
to determine which of the banked CONTROL.nPRIV bits to look at.
The effect was that if (for instance) Secure state was in privileged
mode but Non-Secure was not then we would return the wrong MMU index.
The only places where we are using this function in a way that could
trigger this bug are for the stack loads during a v8M function-return
and for the instruction fetch of a v8M SG insn.
Fix the bug by expanding out the M-profile version of the
arm_current_el() logic inline so it can use the passed in secstate
rather than env->v7m.secure.
Backports 7142eb9e24b4aa5118cd67038057f15694d782aa
Secure mode is not exempted from checking SCR_EL3.TLOR, and in the
future HCR_EL2.TLOR when S-EL2 is enabled.
Backports 9bd268bae5c4760870522292fb1d46e7da7e372a
The helper functions for performing the udot/sdot operations against
a scalar were not using an address-swizzling macro when converting
the index of the scalar element into a pointer into the vm array.
This had no effect on little-endian hosts but meant we generated
incorrect results on big-endian hosts.
For these insns, the index is indexing over group of 4 8-bit values,
so 32 bits per indexed entity, and H4() is therefore what we want.
(For Neon the only possible input indexes are 0 and 1.)
Backports d1a9254be5cc93afb15be19f7543da6ff4806256
In the neon_padd/pmax/pmin helpers for float16, a cut-and-paste error
meant we were using the H4() address swizzler macro rather than the
H2() which is required for 2-byte data. This had no effect on
little-endian hosts but meant we put the result data into the
destination Dreg in the wrong order on big-endian hosts.
Backports 552714c0812a10e5cff239bd29928e5fcb8d8b3b
In both cases, we can sink the write-back and perform
the accumulate into the normal destination temps
Backports 9f1a5f93c2dd345dc6c8fe86ed14bf1485056f6e
The only uses of this function are for loading VFP
double-precision values, and nothing to do with NEON.
Backports b38b96ca90827012ab8eb045c1337cea83a54c4b
The only uses of this function are for loading VFP
single-precision values, and nothing to do with NEON.
Backports 21c1c0e50b73c580c6bfc8f2314d1b6a14793561
We can then use this to improve VMOV (scalar to gp) and
VMOV (gp to scalar) so that we simply perform the memory
operation that we wanted, rather than inserting or
extracting from a 32-bit quantity.
These were the last uses of neon_load/store_reg, so remove them.
Backports 4d5fa5a80ac28f34b8497be1e85371272413a12e
Model these off the aa64 read/write_vec_element functions.
Use it within translate-neon.c.inc. The new functions do
not allocate or free temps, so this rearranges the calling
code a bit.
Backports a712266f5d5a36d04b22fe69fa15592d62bed019
This function makes it clear that we're talking about the whole
register, and not the 32-bit piece at index 0. This fixes a bug
when running on a big-endian host.
Backports 015ee81a4c06b644969f621fd9965cc6372b879e
If the M-profile low-overhead-branch extension is implemented, FPSCR
bits [18:16] are a new field LTPSIZE. If MVE is not implemented
(currently always true for us) then this field always reads as 4 and
ignores writes.
These bits used to be the vector-length field for the old
short-vector extension, so we need to take care that they are not
misinterpreted as setting vec_len. We do this with a rearrangement
of the vfp_set_fpscr() code that deals with vec_len, vec_stride
and also the QC bit; this obviates the need for the M-profile
only masking step that we used to have at the start of the function.
We provide a new field in CPUState for LTPSIZE, even though this
will always be 4, in preparation for MVE, so we don't have to
come back later and split it out of the vfp.xregs[FPSCR] value.
(This state struct field will be saved and restored as part of
the FPSCR value via the vmstate_fpscr in machine.c.)
Backports 8128c8e8cc9489a8387c74075974f86dc0222e7f
M-profile CPUs with half-precision floating point support should
be able to write to FPSCR.FZ16, but an M-profile specific masking
of the value at the top of vfp_set_fpscr() currently prevents that.
This is not yet an active bug because we have no M-profile
FP16 CPUs, but needs to be fixed before we can add any.
The bits that the masking is effectively preventing from being
set are the A-profile only short-vector Len and Stride fields,
plus the Neon QC bit. Rearrange the order of the function so
that those fields are handled earlier and only under a suitable
guard; this allows us to drop the M-profile specific masking,
making FZ16 writeable.
This change also makes the QC bit correctly RAZ/WI for older
no-Neon A-profile cores.
This refactoring also paves the way for the low-overhead-branch
LTPSIZE field, which uses some of the bits that are used for
A-profile Stride and Len.
Backports commit d31e2ce68d56f5bcc83831497e5fe4b8a7e18e85
v8.1M's "low-overhead-loop" extension has three instructions
for looping:
* DLS (start of a do-loop)
* WLS (start of a while-loop)
* LE (end of a loop)
The loop-start instructions are both simple operations to start a
loop whose iteration count (if any) is in LR. The loop-end
instruction handles "decrement iteration count and jump back to loop
start"; it also caches the information about the branch back to the
start of the loop to improve performance of the branch on subsequent
iterations.
As with the branch-future instructions, the architecture permits an
implementation to discard the LO_BRANCH_INFO cache at any time, and
QEMU takes the IMPDEF option to never set it in the first place
(equivalent to discarding it immediately), because for us a "real"
implementation would be unnecessary complexity.
(This implementation only provides the simple looping constructs; the
vector extension MVE (Helium) adds some extra variants to handle
looping across vectors. We'll add those later when we implement
MVE.)
Backports commit b7226369721896ab9ef71544e4fe95b40710e05a
v8.1M implements a new 'branch future' feature, which is a
set of instructions that request the CPU to perform a branch
"in the future", when it reaches a particular execution address.
In hardware, the expected implementation is that the information
about the branch location and destination is cached and then
acted upon when execution reaches the specified address.
However the architecture permits an implementation to discard
this cached information at any point, and so guest code must
always include a normal branch insn at the branch point as
a fallback. In particular, an implementation is specifically
permitted to treat all BF insns as NOPs (which is equivalent
to discarding the cached information immediately).
For QEMU, implementing this caching of branch information
would be complicated and would not improve the speed of
execution at all, so we make the IMPDEF choice to implement
all BF insns as NOPs.
Backports commit 05903f036edba8e3ed940cc215b8e27fb49265b9
The BLX immediate insn in the Thumb encoding always performs
a switch from Thumb to Arm state. This would be totally useless
in M-profile which has no Arm decoder, and so the instruction
does not exist at all there. Make the encoding UNDEF for M-profile.
(This part of the encoding space is used for the branch-future
and low-overhead-loop insns in v8.1M.)
Backports 920f04fa3ea789f8f85a52cee5395b8887b56cf7
