In a future patch the machine code will need to retrieve the MMU
information from KVM during machine initialization before the CPUs
are created.
Actually, KVM_PPC_GET_SMMU_INFO is a VM class ioctl, and thus, we
don't need to have a CPU object around. We just need for KVM to
be initialized and use the kvm_state global. This patch just does
that.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Now that we're checking our MMU configuration is supported by KVM,
rather than adjusting it to KVM, it doesn't really make sense to
have a fallback for kvm_get_smmu_info(). If KVM is too old or buggy
to provide the details, we should rather treat this as an error.
This patch thus adds error reporting to kvm_get_smmu_info() and get
rid of the fallback code. QEMU will now terminate if KVM fails to
provide MMU details. This may break some very old setups, but the
simplification is worth the sacrifice.
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The FPSCR[FI] bit indicates if the last floating point instruction had a result that was rounded. Each consecutive floating point instruction is suppose to set this bit to the correct value. What currently happens is this bit is not set as often as it should be. I have verified that this is the behavior of a real PowerPC 950. This patch fixes that problem by deciding to set this bit after each floating point instruction.
https://www.pdfdrive.net/powerpc-microprocessor-family-the-programming-environments-for-32-e3087633.html
Page 63 in table 2-4 is where the description of this bit can be found.
Signed-off-by: John Arbuckle <programmingkidx@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The store twin case was stubbed out. For now, implement it only within
a serial context, forcing parallel execution to synchronize. It would
be possible to implement with a cmpxchg loop, if we care, but the loose
alignment requirements (simply no crossing 32-byte boundary) might send
us back to the serial context anyway.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
These cases were stubbed out. For now, implement them only within
a serial context, forcing parallel execution to synchronize. It
would be possible to implement these with cmpxchg loops, if we care.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
These operations were previously unimplemented for ppc.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This avoids the need for gen_check_align entirely.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Move the guts of ST_ATOMIC to a function. Use foo_tl for the operations
instead of foo_i32 or foo_i64 specifically. Use MO_ALIGN instead of an
explicit call to gen_check_align.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Move the guts of LD_ATOMIC to a function. Use foo_tl for the operations
instead of foo_i32 or foo_i64 specifically. Use MO_ALIGN instead of an
explicit call to gen_check_align.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Leave only the minimal amount of code within the LDAR macro,
moving the rest of the code into gen_load_locked. Use MO_ALIGN
and remove the explicit call to gen_check_align.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Leave only the minimal amount of code within the STCX macro,
moving the rest of the code into gen_conditional_store.
Remove the explicit call to gen_check_align; the matching LDAX will
have already checked alignment, and we verify the same address.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Always use the gen_conditional_store implementation that uses
atomic_cmpxchg. Make sure and clear reserve_addr across most
interrupts crossing the cpu_loop.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
When running in a parallel context, we must use a helper in order
to perform the 128-bit atomic operation. When running in a serial
context, do the compare before the store.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Section 1.4 of the Power ISA v3.0B states that this insn is
single-copy atomic. As we cannot (yet) issue 128-bit stores
within TCG, use the generic helpers provided.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Section 1.4 of the Power ISA v3.0B states that both of these
instructions are single-copy atomic. As we cannot (yet) issue
128-bit loads within TCG, use the generic helpers provided.
Since TCG cannot (yet) return a 128-bit value, add a slot within
CPUPPCState for returning the high half of a 128-bit return value.
This solution is preferred to the helper assigning to architectural
registers directly, as it avoids clobbering all TCG live values.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This allows faults from MO_ALIGN to have the same effect
as from gen_check_align.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The architecture supports 128 TLB entries. There is no reason
not to provide all of them. In the process we need to fix a
bug that failed to parameterize the configuration register that
tells the operating system the number of entries.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
---
v2:
- Change VMState version.
This hook is only used by CONFIG_USER_ONLY.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
The sizes are already the same, however, we can improve things
if they are identical by design.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
The code in cpu_mmu_index does not properly honor SR_DME.
This bug has workarounds elsewhere in that we flush the
tlb more often than necessary, on the state changes that
should be reflected in a change of mmu_index.
Fixing this means that we can respect the mmu_index that
is given to tlb_flush.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
The previous code was confused, avoiding the flush of the old entry
if the new entry is invalid. We need to flush the old page if the
old entry is valid and the new page if the new entry is valid.
This bug was masked by over-flushing elsewhere.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
While we had defines for *_WAYS, we didn't define more than 1.
Reduce the complexity by eliminating this unused dimension.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
With tlb_fill in mmu.c, we can simplify things further.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
There is no reason to use an indirect branch instead
of simply testing the SR bits that control mmu state.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
There is no reason to allocate this separately. This was probably
copied from target/mips which makes the same mistake.
While doing so, move tlb into the clear-on-reset range. While not
all of the TLB bits are guaranteed zero on reset, all of the valid
bits are cleared, and the rest of the bits are unspecified.
Therefore clearing the whole of the TLB is correct.
Reviewed-by: Stafford Horne <shorne@gmail.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
Rather than pass base+offset to the helper, pass the full index.
In most cases the base is r0 and optimization yields a constant.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
A store to SR changes interrupt state, which should return
to the main loop to recognize that state.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
This allows us to limit the amount of ifdefs and isolate
the test for usermode.
Reviewed-by: Stafford Horne <shorne@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
Track direct jumps via dc->jmp_pc_imm. Use that in
preference to jmp_pc when possible. Emit goto_tb in
that case, and lookup_and_goto_tb otherwise.
Reviewed-by: Stafford Horne <shorne@gmail.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
We failed to store to cpu_pc before raising the exception,
which caused us to re-execute the same insn that we stepped.
Reviewed-by: Stafford Horne <shorne@gmail.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
No need to use the interrupt mechanisms when we can
simply exit the tb directly.
Reviewed-by: Stafford Horne <shorne@gmail.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
These values are unused.
Reviewed-by: Stafford Horne <shorne@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
Rather than emit disassembly while translating, reuse the
generated decoder to build a separate disassembler.
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
Correct the output of the "info mem" and "info tlb" monitor commands to
correctly show canonical addresses.
In 48-bit addressing mode, the upper 16 bits of linear addresses are
equal to bit 47. In 57-bit addressing mode (LA57), the upper 7 bits of
linear addresses are equal to bit 56.
Signed-off-by: Doug Gale <doug16k@gmail.com>
Message-Id: <20180617084025.29198-1-doug16k@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This implements NPT suport for SVM by hooking into
x86_cpu_handle_mmu_fault where it reads the stage-1 page table. Whether
we need to perform this 2nd stage translation, and how, is decided
during vmrun and stored in hflags2, along with nested_cr3 and
nested_pg_mode.
As get_hphys performs a direct cpu_vmexit in case of NPT faults, we need
retaddr in that function. To avoid changing the signature of
cpu_handle_mmu_fault, this passes the value from tlb_fill to get_hphys
via the CPU state.
This was tested successfully via the Jailhouse hypervisor.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-Id: <567473a0-6005-5843-4c73-951f476085ca@web.de>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
It eases code review, unit is explicit.
Patch generated using:
$ git grep -E '(1024|2048|4096|8192|(<<|>>).?(10|20|30))' hw/ include/hw/
and modified manually.
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: David Gibson <david@gibson.dropbear.id.au>
Message-Id: <20180625124238.25339-33-f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Missing break when this feature was added in 89e71e873d
("target/openrisc: implement shadow registers"). This was causing
strange issues as we get writes into the translation block jump cache
and other bits of state.
Fixes: 89e71e873d ("target/openrisc: implement shadow registers")
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Stafford Horne <shorne@gmail.com>
Add support for Hyper-V TLB flush which recently got added to KVM.
Just like regular Hyper-V we announce HV_EX_PROCESSOR_MASKS_RECOMMENDED
regardless of how many vCPUs we have. Windows is 'smart' and uses less
expensive non-EX Hypercall whenever possible (when it wants to flush TLB
for all vCPUs or the maximum vCPU index in the vCPU set requires flushing
is less than 64).
Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
Message-Id: <20180610184927.19309-1-vkuznets@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
tcg_s390_tod_updated() is always called with the iothread being locked
(e.g. from S390TODClass->set() e.g. via HELPER(sck) or on incoming
migration). The helper we call takes the lock itself - bad.
Let's change that by factoring out updating the ckc timer. This now looks
much nicer than having to call a helper from another function.
While touching it we also make sure that env->ckc is updated even if the
new value is -1ULL, for now it would not have been modified in that case.
Reported-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180629170520.13671-1-david@redhat.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Let's do this for completeness reason, although we don't support e.g.
PCDIMM/NVDIMM, which would use the alignment for placing the memory
region in guest physical memory. But maybe someday we would want to
support something like this - then we don't forget about this if
allowing multiple allocations in legacy_s390_alloc().
Use the same alignment as we would set in qemu_anon_ram_alloc(). Our
fixed address satisfies this alignment (1MB). This implicitly sets the
alignment of the underlying memory region.
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180628113817.30814-3-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
We always allocate at a fixed address, a second allocation can therefore
of course never work. We would simply overwrite mappings.
This can e.g. happen in s390_memory_init(), if trying to allocate more
than > 8TB. Let's just bail out, as there is no need for supporting it
(legacy handling for z/VM).
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180628113817.30814-2-david@redhat.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
run_on_cpu() doesn't seem to work reliably until the CPU has been fully
created if the single-threaded TCG main loop is already running.
Therefore, hotplugging a CPU under single-threaded TCG does currently
not work. We should use the direct call instead of going via
run_on_cpu().
So let's use run_on_cpu() for KVM only - KVM requires it due to the initial
CPU reset ioctl. As a nice side effect, we get rid of the ifdef.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180627134410.4901-10-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
If the CPU data is migrated after the TOD clock, the CKC timer of a CPU
is not rearmed. Let's rearm it when loading the CPU state.
Introduce tcg-stub.c just like kvm-stub.c for tcg specific stubs.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180627134410.4901-9-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
This allows a guest to change its TOD. We already take care of updating
all CKC timers from within S390TODClass.
Use MO_ALIGN to load the operand manually - this will properly trigger a
SPECIFICATION exception.
Acked-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180627134410.4901-8-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Let's stop the timer and delete any pending CKC IRQ before doing
anything else.
While at it, add a comment why the check for ckc == -1ULL is needed.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180627134410.4901-7-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Right now, each CPU has its own TOD. Especially, the TOD will differ
based on creation time of a CPU - e.g. when hotplugging a CPU the times
will differ quite a lot, resulting in stall warnings in the guest.
Let's use a single TOD by implementing our new TOD device. Prepare it
for TOD-clock epoch extension.
Most importantly, whenever we set the TOD, we have to update the CKC
timer.
Introduce "tcg_s390x.h" just like "kvm_s390x.h" for tcg specific
function declarations that should not go into cpu.h.
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Hildenbrand <david@redhat.com>
Message-Id: <20180627134410.4901-6-david@redhat.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
