Mark up the cpreginfo structs for the GIC CPU registers to indicate
the offsets from VNCR_EL2, as defined in table D8-66 in rule R_CSRPQ
in the Arm ARM.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Miguel Luis <miguel.luis@oracle.com>
The hypervisor can deliver (virtual) LPIs to a guest by setting up a
list register to have an intid which is an LPI. The GIC has to treat
these a little differently to standard interrupt IDs, because LPIs
have no Active state, and so the guest will only EOI them, it will
not also deactivate them. So icv_eoir_write() must do two things:
* if the LPI ID is not in any list register, we drop the
priority but do not increment the EOI count
* if the LPI ID is in a list register, we immediately deactivate
it, regardless of the split-drop-and-deactivate control
This can be seen in the VirtualWriteEOIR0() and VirtualWriteEOIR1()
pseudocode in the GICv3 architecture specification.
Without this fix, potentially a hypervisor guest might stall because
LPIs get stuck in a bogus Active+Pending state.
Cc: qemu-stable@nongnu.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Tested-by: Miguel Luis <miguel.luis@oracle.com>
The Big QEMU Lock (BQL) has many names and they are confusing. The
actual QemuMutex variable is called qemu_global_mutex but it's commonly
referred to as the BQL in discussions and some code comments. The
locking APIs, however, are called qemu_mutex_lock_iothread() and
qemu_mutex_unlock_iothread().
The "iothread" name is historic and comes from when the main thread was
split into into KVM vcpu threads and the "iothread" (now called the main
loop thread). I have contributed to the confusion myself by introducing
a separate --object iothread, a separate concept unrelated to the BQL.
The "iothread" name is no longer appropriate for the BQL. Rename the
locking APIs to:
- void bql_lock(void)
- void bql_unlock(void)
- bool bql_locked(void)
There are more APIs with "iothread" in their names. Subsequent patches
will rename them. There are also comments and documentation that will be
updated in later patches.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paul Durrant <paul@xen.org>
Acked-by: Fabiano Rosas <farosas@suse.de>
Acked-by: David Woodhouse <dwmw@amazon.co.uk>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Acked-by: Peter Xu <peterx@redhat.com>
Acked-by: Eric Farman <farman@linux.ibm.com>
Reviewed-by: Harsh Prateek Bora <harshpb@linux.ibm.com>
Acked-by: Hyman Huang <yong.huang@smartx.com>
Reviewed-by: Akihiko Odaki <akihiko.odaki@daynix.com>
Message-id: 20240102153529.486531-2-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The ICC_PMR_ELx and ICV_PMR_ELx bit masks returned from
ic{c,v}_fullprio_mask should technically also remove any
bit above 7 as these are marked reserved (read 0) and should
therefore should not be written as anything other than 0.
This was noted during a run of a proprietary test system and
discused on the mailing list [1] and initially thought not to
be an issue due to RES0 being technically allowed to be
written to and read back as long as the implementation does
not use the RES0 bits. It is very possible that the values
are used in comparison without masking, as pointed out by
Peter in [2], if (cs->hppi.prio >= cs->icc_pmr_el1) may well
do the wrong thing.
Masking these values in ic{c,v}_fullprio_mask() should fix
this and prevent any future problems with playing with the
values.
[1]: https://lists.nongnu.org/archive/html/qemu-arm/2023-11/msg00607.html
[2]: https://lists.nongnu.org/archive/html/qemu-arm/2023-11/msg00737.html
Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Message-id: 20231116172818.792364-1-ben.dooks@codethink.co.uk
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>
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
We currently only support GICv2 emulation. To also support GICv3, we will
need to pass a few system registers into their respective handler functions.
This patch adds support for HVF to call into the TCG callbacks for GICv3
system register handlers. This is safe because the GICv3 TCG code is generic
as long as we limit ourselves to EL0 and EL1 - which are the only modes
supported by HVF.
To make sure nobody trips over that, we also annotate callbacks that don't
work in HVF mode, such as EL state change hooks.
With GICv3 support in place, we can run with more than 8 vCPUs.
Signed-off-by: Alexander Graf <agraf@csgraf.de>
Message-id: 20230128224459.70676-1-agraf@csgraf.de
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
With commit 39f29e5993 ("hw/intc/arm_gicv3: Use correct number of
priority bits for the CPU") the number of priority bits was changed from
the maximum value 8 to typically 5. As a consequence a few of the lowest
bits in ICC_PMR_EL1 becomes RAZ/WI. However prior to this patch one of
these bits was still used since the supplied priority value is masked
before it's eventually right shifted with one bit. So the bit is not
lost as one might expect when the register is read again.
The Linux kernel depends on lowest valid bit to be reset to zero, see
commit 33625282adaa ("irqchip/gic-v3: Probe for SCR_EL3 being clear
before resetting AP0Rn") for details.
So fix this by masking the priority value after it may have been right
shifted by one bit.
Cc: qemu-stable@nongnu.org
Fixes: 39f29e5993 ("hw/intc/arm_gicv3: Use correct number of priority bits for the CPU")
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
We have about 30 instances of the typo/variant spelling 'writeable',
and over 500 of the more common 'writable'. Standardize on the
latter.
Change produced with:
sed -i -e 's/\([Ww][Rr][Ii][Tt]\)[Ee]\([Aa][Bb][Ll][Ee]\)/\1\2/g' $(git grep -il writeable)
and then hand-undoing the instance in linux-headers/linux/kvm.h.
Most of these changes are in comments or documentation; the
exceptions are:
* a local variable in accel/hvf/hvf-accel-ops.c
* a local variable in accel/kvm/kvm-all.c
* the PMCR_WRITABLE_MASK macro in target/arm/internals.h
* the EPT_VIOLATION_GPA_WRITABLE macro in target/i386/hvf/vmcs.h
(which is never used anywhere)
* the AR_TYPE_WRITABLE_MASK macro in target/i386/hvf/vmx.h
(which is never used anywhere)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Stefan Weil <sw@weilnetz.de>
Message-id: 20220505095015.2714666-1-peter.maydell@linaro.org
We previously open-coded the expression for the number of virtual APR
registers and the assertion that it was not going to cause us to
overflow the cs->ich_apr[] array. Factor this out into a new
ich_num_aprs() function, for consistency with the icc_num_aprs()
function we just added for the physical APR handling.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220512151457.3899052-7-peter.maydell@linaro.org
Message-id: 20220506162129.2896966-6-peter.maydell@linaro.org
Make the GICv3 set its number of bits of physical priority from the
implementation-specific value provided in the CPU state struct, in
the same way we already do for virtual priority bits. Because this
would be a migration compatibility break, we provide a property
force-8-bit-prio which is enabled for 7.0 and earlier versioned board
models to retain the legacy "always use 8 bits" behaviour.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220512151457.3899052-6-peter.maydell@linaro.org
Message-id: 20220506162129.2896966-5-peter.maydell@linaro.org
The GICv3 code has always supported a configurable number of virtual
priority and preemption bits, but our implementation currently
hardcodes the number of physical priority bits at 8. This is not
what most hardware implementations provide; for instance the
Cortex-A53 provides only 5 bits of physical priority.
Make the number of physical priority/preemption bits driven by fields
in the GICv3CPUState, the way that we already do for virtual
priority/preemption bits. We set cs->pribits to 8, so there is no
behavioural change in this commit. A following commit will add the
machinery for CPUs to set this to the correct value for their
implementation.
Note that changing the number of priority bits would be a migration
compatibility break, because the semantics of the icc_apr[][] array
changes.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220512151457.3899052-5-peter.maydell@linaro.org
Message-id: 20220506162129.2896966-4-peter.maydell@linaro.org
As noted in the comment, the PRIbits field in ICV_CTLR_EL1 is
supposed to match the ICH_VTR_EL2 PRIbits setting; that is, it is the
virtual priority bit setting, not the physical priority bit setting.
(For QEMU currently we always implement 8 bits of physical priority,
so the PRIbits field was previously 7, since it is defined to be
"priority bits - 1".)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220512151457.3899052-3-peter.maydell@linaro.org
Message-id: 20220506162129.2896966-2-peter.maydell@linaro.org
We allow a GICv3 to be connected to any CPU, but we don't do anything
to handle the case where the CPU type doesn't in hardware have a
GICv3 CPU interface and so the various GIC configuration fields
(gic_num_lrs, vprebits, vpribits) are not specified.
The current behaviour is that we will add the EL1 CPU interface
registers, but will not put in the EL2 CPU interface registers, even
if the CPU has EL2, which will leave the GIC in a broken state and
probably result in the guest crashing as it tries to set it up. This
only affects the virt board when using the cortex-a15 or cortex-a7
CPU types (both 32-bit) with -machine gic-version=3 (or 'max')
and -machine virtualization=on.
Instead of failing to set up the EL2 registers, if the CPU doesn't
define the GIC configuration set it to a reasonable default, matching
the standard configuration for most Arm CPUs.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220512151457.3899052-2-peter.maydell@linaro.org
Remove a possible source of error by removing REGINFO_SENTINEL
and using ARRAY_SIZE (convinently hidden inside a macro) to
find the end of the set of regs being registered or modified.
The space saved by not having the extra array element reduces
the executable's .data.rel.ro section by about 9k.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220501055028.646596-4-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Move ARMCPRegInfo and all related declarations to a new
internal header, out of the public cpu.h.
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220501055028.646596-2-richard.henderson@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Update the various GIC ID and feature registers for GICv4:
* PIDR2 [7:4] is the GIC architecture revision
* GICD_TYPER.DVIS is 1 to indicate direct vLPI injection support
* GICR_TYPER.VLPIS is 1 to indicate redistributor support for vLPIs
* GITS_TYPER.VIRTUAL is 1 to indicate vLPI support
* GITS_TYPER.VMOVP is 1 to indicate that our VMOVP implementation
handles cross-ITS synchronization for the guest
* ICH_VTR_EL2.nV4 is 0 to indicate direct vLPI injection support
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220408141550.1271295-38-peter.maydell@linaro.org
The maintenance interrupt state depends only on:
* ICH_HCR_EL2
* ICH_LR<n>_EL2
* ICH_VMCR_EL2 fields VENG0 and VENG1
Now we have a separate function that updates only the vIRQ and vFIQ
lines, use that in places that only change state that affects vIRQ
and vFIQ but not the maintenance interrupt.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220408141550.1271295-27-peter.maydell@linaro.org
The CPU interface changes to support vLPIs are fairly minor:
in the parts of the code that currently look at the list registers
to determine the highest priority pending virtual interrupt, we
must also look at the highest priority pending vLPI. To do this
we change hppvi_index() to check the vLPI and return a special-case
value if that is the right virtual interrupt to take. The callsites
(which handle HPPIR and IAR registers and the "raise vIRQ and vFIQ
lines" code) then have to handle this special-case value.
This commit includes two interfaces with the as-yet-unwritten
redistributor code:
* the new GICv3CPUState::hppvlpi will be set by the redistributor
(in the same way as the existing hpplpi does for physical LPIs)
* when the CPU interface acknowledges a vLPI it needs to set it
to non-pending; the new gicv3_redist_vlpi_pending() function
(which matches the existing gicv3_redist_lpi_pending() used
for physical LPIs) is a stub that will be filled in later
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220408141550.1271295-26-peter.maydell@linaro.org
The function gicv3_cpuif_virt_update() currently sets all of vIRQ,
vFIQ and the maintenance interrupt. This implies that it has to be
used quite carefully -- as the comment notes, setting the maintenance
interrupt will typically cause the GIC code to be re-entered
recursively. For handling vLPIs, we need the redistributor to be
able to tell the cpuif to update the vIRQ and vFIQ lines when the
highest priority pending vLPI changes. Since that change can't cause
the maintenance interrupt state to change, we can pull the "update
vIRQ/vFIQ" parts of gicv3_cpuif_virt_update() out into a separate
function, which the redistributor can then call without having to
worry about the reentrancy issue.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220408141550.1271295-25-peter.maydell@linaro.org
The trace_gicv3_icv_hppir_read trace event takes an integer value
which it uses to form the register name, which should be either
ICV_HPPIR0 or ICV_HPPIR1. We were passing in the 'grp' variable for
this, but that is either GICV3_G0 or GICV3_G1NS, which happen to be 0
and 2, which meant that tracing for the ICV_HPPIR1 register was
incorrectly printed as ICV_HPPIR2.
Use the same approach we do for all the other similar trace events,
and pass in 'ri->crm == 8 ? 0 : 1', deriving the index value
directly from the ARMCPRegInfo struct.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20220303202341.2232284-6-peter.maydell@linaro.org
gicv3_set_gicv3state() is used by arm_gicv3_common.c in
arm_gicv3_common_realize(). Since we want to restrict
arm_gicv3_cpuif.c to TCG, extract gicv3_set_gicv3state()
to a new file. Add this file to the meson 'specific'
source set, since it needs access to "cpu.h".
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20211115223619.2599282-2-philmd@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
According to the "Arm Generic Interrupt Controller Architecture
Specification GIC architecture version 3 and 4" (version G: page 345
for aarch64 or 509 for aarch32):
LRENP bit of ICH_MISR is set when ICH_HCR.LRENPIE==1 and
ICH_HCR.EOIcount is non-zero.
When only LRENPIE was set (and EOI count was zero), the LRENP bit was
wrongly set and MISR value was wrong.
As an additional consequence, if an hypervisor set ICH_HCR.LRENPIE,
the maintenance interrupt was constantly fired. It happens since patch
9cee1efe92 ("hw/intc: Set GIC maintenance interrupt level to only 0 or 1")
which fixed another bug about maintenance interrupt (most significant
bits of misr, including this one, were ignored in the interrupt trigger).
Fixes: 83f036fe3d ("hw/intc/arm_gicv3: Add accessors for ICH_ system registers")
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20211207094427.3473-1-damien.hedde@greensocs.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
It is valid for an OS to put virtual interrupt ID values into the
list registers ICH_LR<n> which are greater than 1023. This
corresponds to (for example) KVM using the in-kernel emulated ITS to
give a (nested) guest an ITS. LPIs are delivered by the L1 kernel to
the L2 guest via the list registers in the same way as non-LPI
interrupts.
QEMU's code for handling writes to ICV_IARn (which happen when the L2
guest acknowledges an interrupt) and to ICV_EOIRn (which happen at
the end of the interrupt) did not consider LPIs, so it would
incorrectly treat interrupt IDs above 1023 as invalid. Fix this by
using the correct condition, which is gicv3_intid_is_special().
Note that the condition in icv_dir_write() is correct -- LPIs
are not valid there and so we want to ignore both "special" ID
values and LPIs.
(In the pseudocode this logic is in:
- VirtualReadIAR0(), VirtualReadIAR1(), which call IsSpecial()
- VirtualWriteEOIR0(), VirtualWriteEOIR1(), which call
VirtualIdentifierValid(data, TRUE) meaning "LPIs OK"
- VirtualWriteDIR(), which calls VirtualIdentifierValid(data, FALSE)
meaning "LPIs not OK")
This bug doesn't seem to have any visible effect on Linux L2 guests
most of the time, because the two bugs cancel each other out: we
neither mark the interrupt active nor deactivate it. However it does
mean that the L2 vCPU priority while the LPI handler is running will
not be correct, so the interrupt handler could be unexpectedly
interrupted by a different interrupt.
(NB: this has nothing to do with using QEMU's emulated ITS.)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
The GICv3/v4 pseudocode has a function IsSpecial() which returns true
if passed a "special" interrupt ID number (anything between 1020 and
1023 inclusive). We open-code this condition in a couple of places,
so abstract it out into a new function gicv3_intid_is_special().
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
During sbsa acs level 3 testing, it is seen that the GIC maintenance
interrupts are not triggered and the related test cases fail. This
is because we were incorrectly passing the value of the MISR register
(from maintenance_interrupt_state()) to qemu_set_irq() as the level
argument, whereas the device on the other end of this irq line
expects a 0/1 value.
Fix the logic to pass a 0/1 level indication, rather than a
0/not-0 value.
Fixes: c5fc89b36c ("hw/intc/arm_gicv3: Implement gicv3_cpuif_virt_update()")
Signed-off-by: Shashi Mallela <shashi.mallela@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20210915205809.59068-1-shashi.mallela@linaro.org
[PMM: tweaked commit message; collapsed nested if()s into one]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Implemented lpi processing at redistributor to get lpi config info
from lpi configuration table,determine priority,set pending state in
lpi pending table and forward the lpi to cpuif.Added logic to invoke
redistributor lpi processing with translated LPI which set/clear LPI
from ITS device as part of ITS INT,CLEAR,DISCARD command and
GITS_TRANSLATER processing.
Signed-off-by: Shashi Mallela <shashi.mallela@linaro.org>
Tested-by: Neil Armstrong <narmstrong@baylibre.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20210910143951.92242-7-shashi.mallela@linaro.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
icv_eoir_write() and icv_dir_write() ignore invalid virtual IRQ numbers
(like LPIs). The issue is that these functions check against the number
of implemented IRQs (QEMU's default is num_irq=288) which can be lower
than the maximum virtual IRQ number (1020 - 1). The consequence is that
if a hypervisor creates an LR for an IRQ between 288 and 1020, then the
guest is unable to deactivate the resulting IRQ. Note that other
functions that deal with large IRQ numbers, like icv_iar_read, check
against 1020 and not against num_irq.
Fix the checks by using GICV3_MAXIRQ (1020) instead of the number of
implemented IRQs.
Signed-off-by: Ricardo Koller <ricarkol@google.com>
Message-id: 20210702233701.3369-1-ricarkol@google.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Commit 382c7160d1 ("hw/intc/arm_gicv3_cpuif: Fix EOIR write access
check logic") added an assert_not_reached() if the guest writes the EOIR
register while no interrupt is active.
It turns out some software does this: EDK2, in
GicV3ExitBootServicesEvent(), unconditionally write EOIR for all
interrupts that it manages. This now causes QEMU to abort when running
UEFI on a VM with GICv3. Although it is UNPREDICTABLE behavior and EDK2
does need fixing, the punishment seems a little harsh, especially since
icc_eoir_write() already tolerates writes of nonexistent interrupt
numbers. Display a guest error and tolerate spurious EOIR writes.
Fixes: 382c7160d1 ("hw/intc/arm_gicv3_cpuif: Fix EOIR write access check logic")
Signed-off-by: Jean-Philippe Brucker <jean-philippe@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 20210604130352.1887560-1-jean-philippe@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
In icc_eoir_write() we assume that we can identify the group of the
IRQ being completed based purely on which register is being written
to and the current CPU state, and that "CPU state matches group
indicated by register" is the only necessary access check.
This isn't correct: if the CPU is not in Secure state then EOIR1 will
only complete Group 1 NS IRQs, but if the CPU is in EL3 it can
complete both Group 1 S and Group 1 NS IRQs. (The pseudocode
ICC_EOIR1_EL1 makes this clear.) We were also missing the logic to
prevent EOIR0 writes completing G0 IRQs when they should not.
Rearrange the logic to first identify the group of the current
highest priority interrupt and then look at whether we should
complete it or ignore the access based on which register was accessed
and the state of the CPU. The resulting behavioural change is:
* EL3 can now complete G1NS interrupts
* G0 interrupt completion is now ignored if the GIC
and the CPU have the security extension enabled and
the CPU is not secure
Reported-by: Chan Kim <ckim@etri.re.kr>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20210510150016.24910-1-peter.maydell@linaro.org
In gicv3_init_cpuif() we copy the ARMCPU gicv3_maintenance_interrupt
into the GICv3CPUState struct's maintenance_irq field. This will
only work if the board happens to have already wired up the CPU
maintenance IRQ before the GIC was realized. Unfortunately this is
not the case for the 'virt' board, and so the value that gets copied
is NULL (since a qemu_irq is really a pointer to an IRQState struct
under the hood). The effect is that the CPU interface code never
actually raises the maintenance interrupt line.
Instead, since the GICv3CPUState has a pointer to the CPUState, make
the dereference at the point where we want to raise the interrupt, to
avoid an implicit requirement on board code to wire things up in a
particular order.
Reported-by: Jose Martins <josemartins90@gmail.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20201009153904.28529-1-peter.maydell@linaro.org
Reviewed-by: Luc Michel <luc@lmichel.fr>
The IAR0/IAR1 register is used to acknowledge an interrupt - a read of the
register activates the highest priority pending interrupt and provides its
interrupt ID. Activating an interrupt can change the CPU's virtual interrupt
state - this change makes sure the virtual irq state is updated.
Signed-off-by: Jeff Kubascik <jeff.kubascik@dornerworks.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20200113154607.97032-1-jeff.kubascik@dornerworks.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
In my "build everything" tree, changing hw/irq.h triggers a recompile
of some 5400 out of 6600 objects (not counting tests and objects that
don't depend on qemu/osdep.h).
hw/hw.h supposedly includes it for convenience. Several other headers
include it just to get qemu_irq and.or qemu_irq_handler.
Move the qemu_irq and qemu_irq_handler typedefs from hw/irq.h to
qemu/typedefs.h, and then include hw/irq.h only where it's still
needed. Touching it now recompiles only some 500 objects.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20190812052359.30071-13-armbru@redhat.com>
The ICC_CTLR_EL3 register includes some bits which are aliases
of bits in the ICC_CTLR_EL1(S) and (NS) registers. QEMU chooses
to keep those bits in the cs->icc_ctlr_el1[] struct fields.
Unfortunately a missing '~' in the code to update the bits
in those fields meant that writing to ICC_CTLR_EL3 would corrupt
the ICC_CLTR_EL1 register values.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20190520162809.2677-5-peter.maydell@linaro.org
In ich_vmcr_write() we enforce "writes of BPR fields to less than
their minimum sets them to the minimum" by doing a "read vbpr and
write it back" operation. A typo here meant that we weren't handling
writes to these fields correctly, because we were reading from VBPR0
but writing to VBPR1.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20190520162809.2677-4-peter.maydell@linaro.org
Replace arm_hcr_el2_{fmo,imo,amo} with a more general routine
that also takes SCR_EL3.NS (aka arm_is_secure_below_el3) into
account, as documented for the plethora of bits in HCR_EL2.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20181210150501.7990-2-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
The IMO, FMO and AMO bits in HCR_EL2 are defined to "behave as
1 for all purposes other than direct reads" if HCR_EL2.TGE
is set and HCR_EL2.E2H is 0, and to "behave as 0 for all
purposes other than direct reads" if HCR_EL2.TGE is set
and HRC_EL2.E2H is 1.
To avoid having to check E2H and TGE everywhere where we test IMO and
FMO, provide accessors arm_hcr_el2_imo(), arm_hcr_el2_fmo()and
arm_hcr_el2_amo(). We don't implement ARMv8.1-VHE yet, so the E2H
case will never be true, but we include the logic to save effort when
we eventually do get to that.
(Note that in several of these callsites the change doesn't
actually make a difference as either the callsite is handling
TGE specially anyway, or the CPU can't get into that situation
with TGE set; we change everywhere for consistency.)
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180724115950.17316-5-peter.maydell@linaro.org
In icc_dir_write() we were incorrectly checking HCR_EL2.FMO
when determining whether IRQ should be routed to EL2; this should
be HCR_EL2.IMO (compare the GICv3 pseudocode ICC_DIR_EL1[]).
Use the correct mask.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180723180337.17378-1-peter.maydell@linaro.org
There was a nasty flip in identifying which register group an access is
targeting. The issue caused spuriously raised priorities of the guest
when handing CPUs over in the Jailhouse hypervisor.
Cc: qemu-stable@nongnu.org
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-id: 28b927d3-da58-bce4-cc13-bfec7f9b1cb9@siemens.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
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.
Signed-off-by: Aaron Lindsay <alindsay@codeaurora.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1523997485-1905-6-git-send-email-alindsay@codeaurora.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
If the GIC has the security extension support enabled, then a
non-secure access to ICC_PMR must take account of the non-secure
view of interrupt priorities, where real priorities 0x00..0x7f
are secure-only and not visible to the non-secure guest, and
priorities 0x80..0xff are shown to the guest as if they were
0x00..0xff. We had the logic here wrong:
* on reads, the priority is in the secure range if bit 7
is clear, not if it is set
* on writes, we want to set bit 7, not mask everything else
Our ICC_RPR read code had the same error as ICC_PMR.
(Compare the GICv3 spec pseudocode functions ICC_RPR_EL1
and ICC_PMR_EL1.)
Fixes: https://bugs.launchpad.net/qemu/+bug/1748434
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Andrew Jones <drjones@redhat.com>
Message-id: 20180315133441.24149-1-peter.maydell@linaro.org
If EL3 is not implemented (ie only one security state) then the
one and only ICC_BPR1 register behaves like the Non-secure
ICC_BPR1 in an EL3-present configuration. In particular, its
reset value is GIC_MIN_BPR_NS, not GIC_MIN_BPR.
Correct the erroneous reset value; this fixes a problem where
we might hit the assert added in commit a89ff39ee9.
Reported-by: Alex Bennée <alex.bennee@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Tested-by: Alex Bennée <alex.bennee@linaro.org>
Message-id: 1496849369-30282-1-git-send-email-peter.maydell@linaro.org
When we calculate the mask to use to get the group priority from
an interrupt priority, the way that NS BPR1 is handled differs
from how BPR0 and S BPR1 work -- a BPR1 value of 1 means
the group priority is in bits [7:1], whereas for BPR0 and S BPR1
this is indicated by a 0 BPR value.
Subtract 1 from the BPR value before creating the mask if
we're using the NS BPR value, for both hardware and virtual
interrupts, as the GICv3 pseudocode does, and fix the comments
accordingly.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 1493226792-3237-4-git-send-email-peter.maydell@linaro.org
icc_bpr_write() was not enforcing that writing a value below the
minimum for the BPR should behave as if the BPR was set to the
minimum value. This doesn't make a difference for the secure
BPRs (since we define the minimum for the QEMU implementation
as zero) but did mean we were allowing the NS BPR1 to be set to
0 when 1 should be the lowest value.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 1493226792-3237-3-git-send-email-peter.maydell@linaro.org
We were setting the VBPR1 field of VMCR_EL2 to icv_min_vbpr()
on reset, but this is not correct. The field should reset to
the minimum value of ICV_BPR0_EL1 plus one.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 1493226792-3237-2-git-send-email-peter.maydell@linaro.org
Add gicv3state void pointer to CPUARMState struct
to store GICv3CPUState.
In case of usecase like CPU reset, we need to reset
GICv3CPUState of the CPU. In such scenario, this pointer
becomes handy.
Signed-off-by: Vijaya Kumar K <Vijaya.Kumar@cavium.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Message-id: 1487850673-26455-5-git-send-email-vijay.kilari@gmail.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This finally allows TCG to benefit from the iothread introduction: Drop
the global mutex while running pure TCG CPU code. Reacquire the lock
when entering MMIO or PIO emulation, or when leaving the TCG loop.
We have to revert a few optimization for the current TCG threading
model, namely kicking the TCG thread in qemu_mutex_lock_iothread and not
kicking it in qemu_cpu_kick. We also need to disable RAM block
reordering until we have a more efficient locking mechanism at hand.
Still, a Linux x86 UP guest and my Musicpal ARM model boot fine here.
These numbers demonstrate where we gain something:
20338 jan 20 0 331m 75m 6904 R 99 0.9 0:50.95 qemu-system-arm
20337 jan 20 0 331m 75m 6904 S 20 0.9 0:26.50 qemu-system-arm
The guest CPU was fully loaded, but the iothread could still run mostly
independent on a second core. Without the patch we don't get beyond
32206 jan 20 0 330m 73m 7036 R 82 0.9 1:06.00 qemu-system-arm
32204 jan 20 0 330m 73m 7036 S 21 0.9 0:17.03 qemu-system-arm
We don't benefit significantly, though, when the guest is not fully
loading a host CPU.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Message-Id: <1439220437-23957-10-git-send-email-fred.konrad@greensocs.com>
[FK: Rebase, fix qemu_devices_reset deadlock, rm address_space_* mutex]
Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
[EGC: fixed iothread lock for cpu-exec IRQ handling]
Signed-off-by: Emilio G. Cota <cota@braap.org>
[AJB: -smp single-threaded fix, clean commit msg, BQL fixes]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <rth@twiddle.net>
Reviewed-by: Pranith Kumar <bobby.prani@gmail.com>
[PM: target-arm changes]
Acked-by: Peter Maydell <peter.maydell@linaro.org>
Fix a broken expression in the calculation of ELRSR
register bits: instead of "(lr & ICH_LR_EL2_HW) == 1"
we want to check for != 0, because the HW bit is not
bit 0 so a test for == 1 is always false.
Fixes: https://bugs.launchpad.net/bugs/1658506
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-id: 1485255993-6322-1-git-send-email-peter.maydell@linaro.org
Implement the architecturally required traps from NS EL1
to EL2 for the CPU interface registers. These fall into
several different groups:
* group-0-only registers all trap if ICH_HRC_EL2.TALL0 is set
(exactly the registers covered by gicv3_fiq_access())
* group-1-only registers all trap if ICH_HRC_EL2.TALL1 is set
(exactly the registers covered by gicv3_irq_access())
* DIR traps if ICH_HCR_EL2.TC or ICH_HCR_EL2.TDIR are set
* PMR, RPR, CTLR trap if ICH_HCR_EL2.TC is set
* SGI0R, SGI1R, ASGI1R trap if ICH_HCR_EL2.TC is set or
if HCR_EL2.IMO or HCR_EL2.FMO are set
We split DIR and the SGI registers out into their own access
functions, leaving the existing gicv3_irqfiq_access() just
handling PMR, RPR and CTLR.
This commit doesn't implement support for trapping on
HSTR_EL2.T12 for the 32-bit registers, as we don't implement
any of those per-coprocessor trap bits currently and
probably will want to do those in some more centralized way.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1483977924-14522-14-git-send-email-peter.maydell@linaro.org
Implement the function which signals virtual interrupts to the
CPU as appropriate following CPU interface state changes.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 1483977924-14522-13-git-send-email-peter.maydell@linaro.org
