Previously, the number of PMP regions was hardcoded to 16 in QEMU.
This patch replaces the fixed value with a new `pmp_regions` field,
allowing platforms to configure the number of PMP regions.
If no specific value is provided, the default number of PMP regions
remains 16 to preserve the existing behavior.
A new CPU parameter num-pmp-regions has been introduced to the QEMU
command line. For example:
-cpu rv64, g=true, c=true, pmp=true, num-pmp-regions=8
Signed-off-by: Jay Chang <jay.chang@sifive.com>
Reviewed-by: Frank Chang <frank.chang@sifive.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20250606072525.17313-3-jay.chang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Björn reported in [1] a case where a rv64 CPU is going through the
profile code path to enable satp mode. In this case,the amount of
extensions on top of the rv64 CPU made it compliant with the RVA22S64
profile during the validation of CPU 0. When the subsequent CPUs were
initialized the static profile object has the 'enable' flag set,
enabling the profile code path for those CPUs.
This happens because we are initializing and realizing each CPU before
going to the next, i.e. init and realize CPU0, then init and realize
CPU1 and so on. If we change any persistent state during the validation
of CPU N it will interfere with the init/realization of CPU N+1.
We're using the 'enabled' profile flag to do two distinct things: inform
cpu_init() that we want profile extensions to be enabled, and telling
QMP that a profile is currently enabled in the CPU. We want to be
flexible enough to recognize profile support for all CPUs that has the
extension prerequisites, but we do not want to force the profile code
path if a profile wasn't set too.
Add a new 'present' flag for profiles that will coexist with the 'enabled'
flag. Enabling a profile means "we want to switch on all its mandatory
extensions". A profile is 'present' if we asserted during validation
that the CPU has the needed prerequisites.
This means that the case reported by Björn now results in
RVA22S64.enabled=false and RVA22S64.present=true. QMP will recognize it
as a RVA22 compliant CPU and we won't force the CPU into the profile
path.
[1] https://lore.kernel.org/qemu-riscv/87y0usiz22.fsf@all.your.base.are.belong.to.us/
Reported-by: Björn Töpel <bjorn@kernel.org>
Fixes: 2af005d610 ("target/riscv/tcg: validate profiles during finalize")
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Tested-by: Björn Töpel <bjorn@rivosinc.com>
Message-ID: <20250528184407.1451983-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
While at it, constify it so that the RISCVCSR array in RISCVCPUDef
can also be const.
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Profile CPUs reuse the instance_init function for bare CPUs; make them
proper subclasses instead. Enabling a profile is now done based on the
RISCVCPUDef struct: even though there is room for only one in RISCVCPUDef,
subclasses check that the parent class's profile is enabled through the
parent profile mechanism.
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Start from the top of the hierarchy: dynamic and vendor CPUs are just
markers, whereas bare CPUs can have their instance_init function
replaced by RISCVCPUDef.
The only difference is that the maximum supported SATP mode has to
be specified separately for 32-bit and 64-bit modes.
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Allow using RISCVCPUDef to replicate all the logic of custom .instance_init
functions. To simulate inheritance, merge the child's RISCVCPUDef with
the parent and then finally move it to the CPUState at the end of
TYPE_RISCV_CPU's own instance_init function.
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Prepare for adding more fields to RISCVCPUDef and reading them in
riscv_cpu_init: instead of storing the misa_mxl_max field in
RISCVCPUClass, ensure that there's always a valid RISCVCPUDef struct
and go through it.
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Start putting all the CPU definitions in a struct. Later this will replace
instance_init functions with declarative code, for now just remove the
ugly cast of class_data.
Reviewed-by: Alistair Francis <alistair23@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
They are used to provide the nice QOM properties for svNN,
but the canonical source of the CPU configuration is now
cpu->cfg.max_satp_mode. Store them in the ArchCPU struct.
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Almost all users of cpu->cfg.satp_mode care about the "max" value
satp_mode_max_from_map(cpu->cfg.satp_mode.map). Convert the QOM
properties back into it. For TCG, deduce the bitmap of supported modes
from valid_vm[].
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
At this moment we're not checking if the host has support for any
specific CSR before doing get/put regs. This will cause problems if the
host KVM doesn't support it (see [1] as an example).
We'll use the same approach done with the CPU extensions: read all known
KVM CSRs during init() to check for availability, then read/write them
if they are present. This will be made by either using get-reglist or by
directly reading the CSRs.
For now we'll just convert the CSRs to use a kvm_csr_cfg[] array,
reusing the same KVMCPUConfig abstraction we use for extensions, and use
the array in (get|put)_csr_regs() instead of manually listing them. A
lot of boilerplate will be added but at least we'll automate the get/put
procedure for CSRs, i.e. adding a new CSR in the future will be a matter
of adding it in kvm_csr_regs[] and everything else will be taken care
of.
Despite all the code changes no behavioral change is made.
[1] https://lore.kernel.org/qemu-riscv/CABJz62OfUDHYkQ0T3rGHStQprf1c7_E0qBLbLKhfv=+jb0SYAw@mail.gmail.com/
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20250429124421.223883-6-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Cc: qemu-stable@nongnu.org
For some targets, simply remove the local definition.
For other targets, move the inline definition out of line.
Reviewed-by: Pierrick Bouvier <pierrick.bouvier@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
To avoid including the huge "cpu.h" for a simple definition,
move TARGET_INSN_START_EXTRA_WORDS to "cpu-param.h".
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
The current 'parent' mechanic for profiles allows for one profile to be
a child of a previous/older profile, enabling all its extensions (and
the parent profile itself) and sparing us from tediously listing all
extensions for every profile.
This works fine for u-mode profiles. For s-mode profiles this is not
enough: a s-mode profile extends not only his equivalent u-mode profile
but also the previous s-mode profile. This means, for example, that
RVA23S64 extends both RVA23U64 and RVA22S64.
To fit this usage, rename the existing 'parent' to 'u_parent' and add a
new 's_parent' attribute for profiles. Handle both like we were doing
with the previous 'parent' attribute, i.e. if set, enable it. This
change does nothing for the existing profiles but will make RVA23S64
simpler.
Suggested-by: Andrew Jones <ajones@ventanamicro.com>
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20250115184316.2344583-4-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Because the RNMI interrupt trap handler address is implementation defined.
We add the 'rnmi-interrupt-vector' and 'rnmi-exception-vector' as the property
of the harts. It’s very easy for users to set the address based on their
expectation. This patch also adds the functionality to handle the RNMI signals.
Signed-off-by: Frank Chang <frank.chang@sifive.com>
Signed-off-by: Tommy Wu <tommy.wu@sifive.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20250106054336.1878291-4-frank.chang@sifive.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Do a cosmetic change in riscv_raise_exception() to change 'exception'
type from uint32_t to RISCVException, making it a bit clear that the
arg is directly correlated to the RISCVException enum.
As a side effect, change 'excp' type from int to RISCVException in
generate_exception() to guarantee that all callers of
riscv_raise_exception() will use the enum.
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Message-ID: <20250106173734.412353-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Convert all targets simultaneously, as the gen_intermediate_code
function disappears from the target. While there are possible
workarounds, they're larger than simply performing the conversion.
Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
zicfiss introduces a new state ssp ("shadow stack register") in cpu.
ssp is expressed as a new unprivileged csr (CSR_SSP=0x11) and holds
virtual address for shadow stack as programmed by software.
Shadow stack (for each mode) is enabled via bit3 in *envcfg CSRs.
Shadow stack can be enabled for a mode only if it's higher privileged
mode had it enabled for itself. M mode doesn't need enabling control,
it's always available if extension is available on cpu.
This patch also implements helper bcfi function which determines if bcfi
is enabled at current privilege or not.
Adds ssp to migration state as well.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241008225010.1861630-12-debug@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
zicfilp protects forward control flow (if enabled) by enforcing all
indirect call and jmp must land on a landing pad instruction `lpad`. If
target of an indirect call or jmp is not `lpad` then cpu/hart must raise
a sw check exception with tval = 2.
This patch implements the mechanism using TCG. Target architecture branch
instruction must define the end of a TB. Using this property, during
translation of branch instruction, TB flag = FCFI_LP_EXPECTED can be set.
Translation of target TB can check if FCFI_LP_EXPECTED flag is set and a
flag (fcfi_lp_expected) can be set in DisasContext. If `lpad` gets
translated, fcfi_lp_expected flag in DisasContext can be cleared. Else
it'll fault.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241008225010.1861630-7-debug@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
sw check exception support was recently added. This patch further augments
sw check exception by providing support for additional code which is
provided in *tval. Adds `sw_check_code` field in cpuarchstate. Whenever
sw check exception is raised *tval gets the value deposited in
`sw_check_code`.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241008225010.1861630-6-debug@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
elp state is recorded in *status on trap entry (less privilege to higher
privilege) and restored in elp from *status on trap exit (higher to less
privilege).
Additionally this patch introduces a forward cfi helper function to
determine if current privilege has forward cfi is enabled or not based on
*envcfg (for U, VU, S, VU, HS) or mseccfg csr (for M).
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241008225010.1861630-5-debug@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
zicfilp introduces a new state elp ("expected landing pad") in cpu.
During normal execution, elp is idle (NO_LP_EXPECTED) i.e not expecting
landing pad. On an indirect call, elp moves LP_EXPECTED. When elp is
LP_EXPECTED, only a subsquent landing pad instruction can set state back
to NO_LP_EXPECTED. On reset, elp is set to NO_LP_EXPECTED.
zicfilp is enabled via bit2 in *envcfg CSRs. Enabling control for M-mode
is in mseccfg CSR at bit position 10.
On trap, elp state is saved away in *status.
Adds elp to the migration state as well.
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Jim Shu <jim.shu@sifive.com>
Co-developed-by: Andy Chiu <andy.chiu@sifive.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241008225010.1861630-4-debug@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Execution environment config CSR controlling user env and current
privilege state shouldn't be limited to qemu-system only. *envcfg
CSRs control enabling of features in next lesser mode. In some cases
bits *envcfg CSR can be lit up by kernel as part of kernel policy or
software (user app) can choose to opt-in by issuing a system call
(e.g. prctl). In case of qemu-user, it should be no different because
qemu is providing underlying execution environment facility and thus
either should provide some default value in *envcfg CSRs or react to
system calls (prctls) initiated from application. priv is set to PRV_U
and menvcfg/senvcfg set to 0 for qemu-user on reest.
`henvcfg` has been left for qemu-system only because it is not expected
that someone will use qemu-user where application is expected to have
hypervisor underneath which is controlling its execution environment. If
such a need arises then `henvcfg` could be exposed as well.
Relevant discussion:
https://lore.kernel.org/all/CAKmqyKOTVWPFep2msTQVdUmJErkH+bqCcKEQ4hAnyDFPdWKe0Q@mail.gmail.com/
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <20241008225010.1861630-2-debug@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Both CSRRS and CSRRC always read the addressed CSR and cause any read side
effects regardless of rs1 and rd fields. Note that if rs1 specifies a register
holding a zero value other than x0, the instruction will still attempt to write
the unmodified value back to the CSR and will cause any attendant side effects.
So if CSRRS or CSRRC tries to write a read-only CSR with rs1 which specifies
a register holding a zero value, an illegal instruction exception should be
raised.
Signed-off-by: Yu-Ming Chang <yumin686@andestech.com>
Signed-off-by: Alvin Chang <alvinga@andestech.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-ID: <172100444279.18077.6893072378718059541-0@git.sr.ht>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Currently, if a counter monitoring cycle/instret is stopped via
mcountinhibit we just update the state while the value is saved
during the next read. This is not accurate as the read may happen
many cycles after the counter is stopped. Ideally, the read should
return the value saved when the counter is stopped.
Thus, save the value of the counter during the inhibit update
operation and return that value during the read if corresponding bit
in mcountihibit is set.
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Signed-off-by: Atish Patra <atishp@rivosinc.com>
Message-ID: <20240711-smcntrpmf_v7-v8-8-b7c38ae7b263@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Privilege mode filtering can also be emulated for cycle/instret by
tracking host_ticks/icount during each privilege mode switch. This
patch implements that for both cycle/instret and mhpmcounters. The
first one requires Smcntrpmf while the other one requires Sscofpmf
to be enabled.
The cycle/instret are still computed using host ticks when icount
is not enabled. Otherwise, they are computed using raw icount which
is more accurate in icount mode.
Co-Developed-by: Rajnesh Kanwal <rkanwal@rivosinc.com>
Signed-off-by: Rajnesh Kanwal <rkanwal@rivosinc.com>
Reviewed-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Acked-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Atish Patra <atishp@rivosinc.com>
Message-ID: <20240711-smcntrpmf_v7-v8-7-b7c38ae7b263@rivosinc.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
