When FEAT_RME is implemented, these bits override the value of
CNT[VP]_CTL_EL0.IMASK in Realm and Root state. Move the IRQ state update
into a new gt_update_irq() function and test those bits every time we
recompute the IRQ state.
Since we're removing the IRQ state from some trace events, add a new
trace event for gt_update_irq().
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Message-id: 20230809123706.1842548-7-jean-philippe@linaro.org
[PMM: only register change hook if not USER_ONLY and if TCG]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
At the moment we only handle Secure and Nonsecure security spaces for
the AT instructions. Add support for Realm and Root.
For AArch64, arm_security_space() gives the desired space. ARM DDI0487J
says (R_NYXTL):
If EL3 is implemented, then when an address translation instruction
that applies to an Exception level lower than EL3 is executed, the
Effective value of SCR_EL3.{NSE, NS} determines the target Security
state that the instruction applies to.
For AArch32, some instructions can access NonSecure space from Secure,
so we still need to pass the state explicitly to do_ats_write().
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20230809123706.1842548-5-jean-philippe@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
GPC checks are not performed on the output address for AT instructions,
as stated by ARM DDI 0487J in D8.12.2:
When populating PAR_EL1 with the result of an address translation
instruction, granule protection checks are not performed on the final
output address of a successful translation.
Rename get_phys_addr_with_secure(), since it's only used to handle AT
instructions.
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20230809123706.1842548-4-jean-philippe@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
When HCR_EL2.E2H is enabled, TLB entries are formed using the EL2&0
translation regime, instead of the EL2 translation regime. The TLB VAE2*
instructions invalidate the regime that corresponds to the current value
of HCR_EL2.E2H.
At the moment we only invalidate the EL2 translation regime. This causes
problems with RMM, which issues TLBI VAE2IS instructions with
HCR_EL2.E2H enabled. Update vae2_tlbmask() to take HCR_EL2.E2H into
account.
Add vae2_tlbbits() as well, since the top-byte-ignore configuration is
different between the EL2&0 and EL2 regime.
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20230809123706.1842548-3-jean-philippe@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The PAR_EL1.SH field documents that for the cases of:
* Device memory
* Normal memory with both Inner and Outer Non-Cacheable
the field should be 0b10 rather than whatever was in the
translation table descriptor field. (In the pseudocode this
is handled by PAREncodeShareability().) Perform this
adjustment when assembling a PAR value.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20230807141514.19075-16-peter.maydell@linaro.org
arm_hcr_el2_eff_secstate() takes a bool secure, which it uses to
determine whether EL2 is enabled in the current security state.
With the advent of FEAT_RME this is no longer sufficient, because
EL2 can be enabled for Secure state but not for Root, and both
of those will pass 'secure == true' in the callsites in ptw.c.
As it happens in all of our callsites in ptw.c we either avoid making
the call or else avoid using the returned value if we're doing a
translation for Root, so this is not a behaviour change even if the
experimental FEAT_RME is enabled. But it is less confusing in the
ptw.c code if we avoid the use of a bool secure that duplicates some
of the information in the ArmSecuritySpace argument.
Make arm_hcr_el2_eff_secstate() take an ARMSecuritySpace argument
instead. Because we always want to know the HCR_EL2 for the
security state defined by the current effective value of
SCR_EL3.{NSE,NS}, it makes no sense to pass ARMSS_Root here,
and we assert that callers don't do that.
To avoid the assert(), we thus push the call to
arm_hcr_el2_eff_secstate() down into the cases in
regime_translation_disabled() that need it, rather than calling the
function and ignoring the result for the Root space translations.
All other calls to this function in ptw.c are already in places
where we have confirmed that the mmu_idx is a stage 2 translation
or that the regime EL is not 3.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20230807141514.19075-7-peter.maydell@linaro.org
Unlike architectures with precise self-modifying code semantics
(e.g. x86) ARM processors do not maintain coherency for instruction
execution and memory, requiring an instruction synchronization
barrier on every core that will execute the new code, and on many
models also the explicit use of cache management instructions.
While this is required to make JITs work on actual hardware, QEMU
has gotten away with not handling this since it does not emulate
caches, and unconditionally invalidates code whenever the softmmu
or the user-mode page protection logic detects that code has been
modified.
Unfortunately the latter does not work in the face of dual-mapped
code (a common W^X workaround), where one page is executable and
the other is writable: user-mode has no way to connect one with the
other as that is only known to the kernel and the emulated
application.
This commit works around the issue by telling software that
instruction cache invalidation is required by clearing the
CPR_EL0.DIC flag (regardless of whether the emulated processor
needs it), and then invalidating code in IC IVAU instructions.
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1034
Co-authored-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: John Högberg <john.hogberg@ericsson.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 168778890374.24232.3402138851538068785-1@git.sr.ht
[PMM: removed unnecessary AArch64 feature check; moved
"clear CTR_EL1.DIC" code up a bit so it's not in the middle
of the vfp/neon related tests]
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Some registers whose 'cooked' writefns induce TLB maintenance do
not have raw_writefn ops defined. If only the writefn ops is set
(ie. no raw_writefn is provided), it is assumed the cooked also
work as the raw one. For those registers it is not obvious the
tlb_flush works on KVM mode so better/safer setting the raw write.
Signed-off-by: Eric Auger <eric.auger@redhat.com>
Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
DC CVAP and DC CVADP instructions can be executed in EL0 on Linux,
either directly when SCTLR_EL1.UCI == 1 or emulated by the kernel (see
user_cache_maint_handler() in arch/arm64/kernel/traps.c).
This patch enables execution of the two instructions in user mode
emulation.
Signed-off-by: Zhuojia Shen <chaosdefinition@hotmail.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
In check_s2_mmu_setup() we have a check that is attempting to
implement the part of AArch64.S2MinTxSZ that is specific to when EL1
is AArch32:
if !s1aarch64 then
// EL1 is AArch32
min_txsz = Min(min_txsz, 24);
Unfortunately we got this wrong in two ways:
(1) The minimum txsz corresponds to a maximum inputsize, but we got
the sense of the comparison wrong and were faulting for all
inputsizes less than 40 bits
(2) We try to implement this as an extra check that happens after
we've done the same txsz checks we would do for an AArch64 EL1, but
in fact the pseudocode is *loosening* the requirements, so that txsz
values that would fault for an AArch64 EL1 do not fault for AArch32
EL1, because it does Min(old_min, 24), not Max(old_min, 24).
You can see this also in the text of the Arm ARM in table D8-8, which
shows that where the implemented PA size is less than 40 bits an
AArch32 EL1 is still OK with a configured stage2 T0SZ for a 40 bit
IPA, whereas if EL1 is AArch64 then the T0SZ must be big enough to
constrain the IPA to the implemented PA size.
Because of part (2), we can't do this as a separate check, but
have to integrate it into aa64_va_parameters(). Add a new argument
to that function to indicate that EL1 is 32-bit. All the existing
callsites except the one in get_phys_addr_lpae() can pass 'false',
because they are either doing a lookup for a stage 1 regime or
else they don't care about the tsz/tsz_oob fields.
Cc: qemu-stable@nongnu.org
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/1627
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20230509092059.3176487-1-peter.maydell@linaro.org
Since commit a0e61807a3 ("qapi: Remove QMP events and commands from
user-mode builds") we don't generate the "qapi-commands-machine.h"
header in a user-emulation-only build.
Move the QMP functions from helper.c (which is always compiled)
to monitor.c (which is only compiled when system-emulation
is selected). Rename monitor.c to arm-qmp-cmds.c.
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20230223155540.30370-2-philmd@linaro.org>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[Straightforward conflict with commit 9def656e7a resolved]
Mark up the sysreg definitions for the registers trapped
by HFGRTR/HFGWTR bits 36..63.
Of these, some correspond to RAS registers which we implement as
always-UNDEF: these don't need any extra handling for FGT because the
UNDEF-to-EL1 always takes priority over any theoretical
FGT-trap-to-EL2.
Bit 50 (NACCDATA_EL1) is for the ACCDATA_EL1 register which is part
of the FEAT_LS64_ACCDATA feature which we don't yet implement.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Fuad Tabba <tabba@google.com>
Message-id: 20230130182459.3309057-14-peter.maydell@linaro.org
Message-id: 20230127175507.2895013-14-peter.maydell@linaro.org
Implement the machinery for fine-grained traps on normal sysregs.
Any sysreg with a fine-grained trap will set the new field to
indicate which FGT register bit it should trap on.
FGT traps only happen when an AArch64 EL2 enables them for
an AArch64 EL1. They therefore are only relevant for AArch32
cpregs when the cpreg can be accessed from EL0. The logic
in access_check_cp_reg() will check this, so it is safe to
add a .fgt marking to an ARM_CP_STATE_BOTH ARMCPRegInfo.
The DO_BIT and DO_REV_BIT macros define enum constants FGT_##bitname
which can be used to specify the FGT bit, eg
.fgt = FGT_AFSR0_EL1
(We assume that there is no bit name duplication across the FGT
registers, for brevity's sake.)
Subsequent commits will add the .fgt fields to the relevant register
definitions and define the FGT_nnn values for them.
Note that some of the FGT traps are for instructions that we don't
handle via the cpregs mechanisms (mostly these are instruction traps).
Those we will have to handle separately.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Fuad Tabba <tabba@google.com>
Message-id: 20230130182459.3309057-10-peter.maydell@linaro.org
Message-id: 20230127175507.2895013-10-peter.maydell@linaro.org
Define the system registers which are provided by the
FEAT_FGT fine-grained trap architectural feature:
HFGRTR_EL2, HFGWTR_EL2, HDFGRTR_EL2, HDFGWTR_EL2, HFGITR_EL2
All these registers are a set of bit fields, where each bit is set
for a trap and clear to not trap on a particular system register
access. The R and W register pairs are for system registers,
allowing trapping to be done separately for reads and writes; the I
register is for system instructions where trapping is on instruction
execution.
The data storage in the CPU state struct is arranged as a set of
arrays rather than separate fields so that when we're looking up the
bits for a system register access we can just index into the array
rather than having to use a switch to select a named struct member.
The later FEAT_FGT2 will add extra elements to these arrays.
The field definitions for the new registers are in cpregs.h because
in practice the code that needs them is code that also needs
the cpregs information; cpu.h is included in a lot more files.
We're also going to add some FGT-specific definitions to cpregs.h
in the next commit.
We do not implement HAFGRTR_EL2, because we don't implement
FEAT_AMUv1.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Fuad Tabba <tabba@google.com>
Message-id: 20230130182459.3309057-9-peter.maydell@linaro.org
Message-id: 20230127175507.2895013-9-peter.maydell@linaro.org
The HSTR_EL2 register is not supposed to have an effect unless EL2 is
enabled in the current security state. We weren't checking for this,
which meant that if the guest set up the HSTR_EL2 register we would
incorrectly trap even for accesses from Secure EL0 and EL1.
Add the missing checks. (Other places where we look at HSTR_EL2
for the not-in-v8A bits TTEE and TJDBX are already checking that
we are in NS EL0 or EL1, so there we alredy know EL2 is enabled.)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Fuad Tabba <tabba@google.com>
Message-id: 20230130182459.3309057-8-peter.maydell@linaro.org
Message-id: 20230127175507.2895013-8-peter.maydell@linaro.org
The AArch32 ATS12NSO* address translation operations are supposed to
trap to either EL2 or EL3 if they're executed at Secure EL1 (which
can only happen if EL3 is AArch64). We implement this, but we got
the syndrome value wrong: like other traps to EL2 or EL3 on an
AArch32 cpreg access, they should report the 0x3 syndrome, not the
0x0 'uncategorized' syndrome. This is clear in the access pseudocode
for these instructions.
Fix the syndrome value for these operations by correcting the
returned value from the ats_access() function.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Fuad Tabba <tabba@google.com>
Message-id: 20230130182459.3309057-3-peter.maydell@linaro.org
Message-id: 20230127175507.2895013-3-peter.maydell@linaro.org
