XIVE offers a 'cache watch facility', which allows software to read/update
a potentially cached table entry with no software lock. There's one such
facility in the Virtualization Controller (VC) to update the ESB and END
entries and one in the Presentation Controller (PC) to update the
NVP/NVG/NVC entries.
Each facility has 4 cache watch engines to control the updates and
firmware can request an available engine by querying the hardware
'watch_assign' register of the VC or PC. The engine is then reserved and
is released after the data is updated by reading the 'watch_spec' register
(which also allows to check for a conflict during the update).
If no engine is available, the special value 0xFF is returned and
firmware is expected to repeat the request until an engine becomes
available.
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Signed-off-by: Michael Kowal <kowal@linux.vnet.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
to log an error in case of bad configuration of the XIVE tables by the FW.
Reviewed-by: Frederic Barrat <fbarrat@linux.ibm.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The low-level functions to access the TIMA take a presenter object as
a first argument. When accessing the TIMA from the IC BAR,
i.e. indirect calls, we currently pass a NULL pointer for the
presenter argument. While it appears ok with the current usage, it's
dangerous. And it's pretty easy to figure out the presenter in that
context, so this patch fixes it.
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-ID: <20230705081400.218408-1-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Direct TIMA operations can be done through 4 pages, each with a
different privilege level dictating what fields can be accessed. On
the other hand, indirect TIMA accesses on P10 are done through a
single page, which is the equivalent of the most privileged page of
direct TIMA accesses.
The offset in the IC bar of an indirect access specifies what hw
thread is targeted (page shift bits) and the offset in the
TIMA being accessed (the page offset bits). When the indirect
access is calling the underlying direct access functions, it is
therefore important to clearly separate the 2, as the direct functions
assume any page shift bits define the privilege ring level. For
indirect accesses, those bits must be 0. This patch fixes the offset
passed to direct TIMA functions.
It didn't matter for SMT1, as the 2 least significant bits of the page
shift are part of the hw thread ID and always 0, so the direct TIMA
functions were accessing the privilege ring 0 page. With SMT4/8, it is
no longer true.
The fix is specific to P10, as indirect TIMA access on P9 was handled
differently.
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-ID: <20230703080858.54060-1-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Booting linux on the powernv10 machine logs a few errors like:
Invalid read at addr 0x38, size 1, region 'xive-ic-tm-indirect', reason: invalid size (min:8 max:8)
Invalid write at addr 0x38, size 1, region 'xive-ic-tm-indirect', reason: invalid size (min:8 max:8)
Invalid read at addr 0x38, size 1, region 'xive-ic-tm-indirect', reason: invalid size (min:8 max:8)
Those errors happen when linux is resetting XIVE. We're trying to
read/write the enablement bit for the hardware context and qemu
doesn't allow indirect TIMA accesses of less than 8 bytes. Direct TIMA
access can go through though, as well as indirect TIMA accesses on P9.
So even though there are some restrictions regarding the address/size
combinations for TIMA access, the example above is perfectly valid.
This patch lets indirect TIMA accesses of all sizes go through. The
special operations will be intercepted and the default "raw" handlers
will pick up all other requests and complain about invalid sizes as
appropriate.
Tested-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-ID: <20230626094057.1192473-1-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Accessing the TIMA from some specific ring/offset combination can
trigger a special operation, with or without side effects. It is
implemented in qemu with an array of special operations to compare
accesses against. Since the presenter on P10 is pretty similar to P9,
we had the full array defined for P9 and we just had a special case
for P10 to treat one access differently. With a recent change,
6f2cbd133d ("pnv/xive2: Handle TIMA access through all ports"), we
now ignore some of the bits of the TIMA address, but that patch
managed to botch the detection of the special case for P10.
To clean that up, this patch introduces a full array of special ops to
be used for P10. The code to detect a special access is common with
P9, only the array of operations differs. The presenter can pick the
correct array of special ops based on its configuration introduced in
a previous patch.
Fixes: Coverity CID 1512997, 1512998
Fixes: 6f2cbd133d ("pnv/xive2: Handle TIMA access through all ports")
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The presenters for xive on P9 and P10 are mostly similar but the
behavior can be tuned through a few CQ registers. This patch adds a
"get_config" method, which will allow to access that config from the
presenter in a later patch.
For now, just define the config for the TIMA version.
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
When dumping the END and NVP tables ("info pic" from the HMP) on the
P10 model, we're likely to be flooded with error messages such as:
XIVE[0] - VST: invalid NVPT entry f33800 !?
The error is printed when finding an empty VSD in an indirect
table (thus END and NVP tables with skiboot), which is going to happen
when dumping the xive state. So let's tune down those messages. They
can be re-enabled easily with a macro if needed.
Those errors were already hidden on xive/P9, for the same reason.
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20230531150537.369350-1-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
The Thread Interrupt Management Area (TIMA) can be accessed through 4
ports, targeted by the address. The base address of a TIMA
is using port 0 and the other ports are 0x80 apart. Using one port or
another can be useful to balance the load on the snoop buses. With
skiboot and linux, we currently use port 0, but as it tends to be
busy, another hypervisor is using port 1 for TIMA access.
The port address bits fall in between the special op indication
bits (the 2 MSBs) and the register offset bits (the 6 LSBs). They are
"don't care" for the hardware when processing a TIMA operation. This
patch filters out those port address bits so that a TIMA operation can
be triggered using any port.
It is also true for indirect access (through the IC BAR) and it's
actually nothing new, it was already the case on P9. Which helps here,
as the TIMA handling code is common between P9 (xive) and P10 (xive2).
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20230601121331.487207-6-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Fix what was probably a silly mistake and allow to write the Physical
Thread enable registers 0 and 1. Skiboot prefers to use the ENx_SET
variant so it went unnoticed, but there's no reason to discard a write
to the full register, it is Read-Write.
Fixes: da71b7e3ed ("ppc/pnv: Add a XIVE2 controller to the POWER10 chip")
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20230601121331.487207-4-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Add basic read/write support for the ESB cache configuration register
on P10. We don't model the ESB cache in qemu so reading/writing the
register won't do anything, but it avoids logging a guest error when
skiboot configures it:
qemu-system-ppc64 -machine powernv10 ... -d guest_errors
...
XIVE[0] - VC: invalid read @240
XIVE[0] - VC: invalid write @240
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20230601121331.487207-3-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Add basic read/write support for the TCTXT Config register on P10. qemu
doesn't do anything with it yet, but it avoids logging a guest error
when skiboot configures the fused-core state:
qemu-system-ppc64 -machine powernv10 ... -d guest_errors
...
[ 0.131670000,5] XIVE: [ IC 00 ] Initializing XIVE block ID 0...
XIVE[0] - TCTXT: invalid read @140
XIVE[0] - TCTXT: invalid write @140
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20230601121331.487207-2-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
PnvChipClass, PnvChip, Pnv8Chip, Pnv9Chip, and Pnv10Chip are defined
in pnv.h. Many users of the header don't actually need them. One
instance is this inclusion loop: hw/ppc/pnv_homer.h includes
hw/ppc/pnv.h for typedef PnvChip, and vice versa for struct PnvHomer.
Similar structs live in their own headers: PnvHomerClass and PnvHomer
in pnv_homer.h, PnvLpcClass and PnvLpcController in pci_lpc.h,
PnvPsiClass, PnvPsi, Pnv8Psi, Pnv9Psi, Pnv10Psi in pnv_psi.h, ...
Move PnvChipClass, PnvChip, Pnv8Chip, Pnv9Chip, and Pnv10Chip to new
pnv_chip.h, and adjust include directives. This breaks the inclusion
loop mentioned above.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20221222104628.659681-2-armbru@redhat.com>
It keeps repeating, move it to the header. This uses __builtin_ffsll() to
allow using the macros in #define.
This is not using the QEMU's FIELD macros as this would require changing
all such macros found in skiboot (the PPC PowerNV firmware).
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Message-Id: <20220628080544.1509428-1-aik@ozlabs.ru>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
When accessing a thread context through the IC BAR, the offset of the
page in the BAR identifies the CPU. From that offset, we can compute
the PIR (processor ID register) of the CPU to do the data structure
lookup. On P10, the current code assumes an access for node 0 when
computing the PIR. Everything is almost in place to allow access for
other nodes though. So this patch reworks how the PIR value is
computed so that we can access all thread contexts through the IC BAR.
The PIR is already correct on P9, so no need to modify anything there.
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20220602165310.558810-1-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
When writing a register from the TCTXT memory region (4th page within
the IC BAR), we were overwriting the Presentation Controller (PC)
register at the same offset. It looks like a silly cut and paste
error.
We were somehow lucky: the TCTXT registers being touched are
TCTXT_ENx/_SET/_RESET to enable physical threads and the PC registers
at the same offset are either not used by our model or the update was
harmless.
Found through code inspection.
Signed-off-by: Frederic Barrat <fbarrat@linux.ibm.com>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20220523151859.72283-1-fbarrat@linux.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
The XIVE interrupt controller on P10 can automatically save and
restore the state of the interrupt registers under the internal NVP
structure representing the VCPU. This saves a costly store/load in
guest entries and exits.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Add GEN1 config even if we don't use it yet in the core framework.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The thread interrupt management area (TIMA) is a set of pages mapped
in the Hypervisor and in the guest OS address space giving access to
the interrupt thread context registers for interrupt management, ACK,
EOI, CPPR, etc.
XIVE2 changes slightly the TIMA layout with extra bits for the new
features, larger CAM lines and the controller provides configuration
switches for backward compatibility. This is called the XIVE2
P9-compat mode, of Gen1 TIMA. It impacts the layout of the TIMA and
the availability of the internal features associated with it,
Automatic Save & Restore for instance. Using a P9 layout also means
setting the controller in such a mode at init time.
As the OPAL driver initializes the XIVE2 controller with a XIVE2/P10
TIMA directly, the XIVE2 model only has a simple support for the
compat mode in the OS TIMA.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Only the CAM line updates done by the hypervisor are specific to
POWER10. Instead of duplicating the TM ops table, we handle these
commands locally under the PowerNV XIVE2 model.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
These bits control the availability of interrupt features : StoreEOI,
PHB PQ_disable, PHB Address-Based Trigger and the overall XIVE
exploitation mode. These bits can be set at early boot time of the
system to activate/deactivate a feature for testing purposes. The
default value should be '1'.
The 'XIVE exploitation mode' bit is a software bit that skiboot could
use to disable the XIVE OS interface and propose a P8 style XICS
interface instead. There are no plans for that for the moment.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The trigger message coming from a HW source contains a special bit
informing the XIVE interrupt controller that the PQ bits have been
checked at the source or not. Depending on the value, the IC can
perform the check and the state transition locally using its own PQ
state bits.
The following changes add new accessors to the XiveRouter required to
query and update the PQ state bits. This only applies to the PowerNV
machine. sPAPR accessors are provided but the pSeries machine should
not be concerned by such complex configuration for the moment.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
The XIVE2 interrupt controller of the POWER10 processor follows the
same logic than on POWER9 but the HW interface has been largely
reviewed. It has a new register interface, different BARs, extra
VSDs, new layout for the XIVE2 structures, and a set of new features
which are described below.
This is a model of the POWER10 XIVE2 interrupt controller for the
PowerNV machine. It focuses primarily on the needs of the skiboot
firmware but some initial hypervisor support is implemented for KVM
use (escalation).
Support for new features will be implemented in time and will require
new support from the OS.
* XIVE2 BARS
The interrupt controller BARs have a different layout outlined below.
Each sub-engine has now own its range and the indirect TIMA access was
replaced with a set of pages, one per CPU, under the IC BAR:
- IC BAR (Interrupt Controller)
. 4 pages, one per sub-engine
. 128 indirect TIMA pages
- TM BAR (Thread Interrupt Management Area)
. 4 pages
- ESB BAR (ESB pages for IPIs)
. up to 1TB
- END BAR (ESB pages for ENDs)
. up to 2TB
- NVC BAR (Notification Virtual Crowd)
. up to 128
- NVPG BAR (Notification Virtual Process and Group)
. up to 1TB
- Direct mapped Thread Context Area (reads & writes)
OPAL does not use the grouping and crowd capability.
* Virtual Structure Tables
XIVE2 adds new tables types and also changes the field layout of the END
and NVP Virtualization Structure Descriptors.
- EAS
- END new layout
- NVT was splitted in :
. NVP (Processor), 32B
. NVG (Group), 32B
. NVC (Crowd == P9 block group) 32B
- IC for remote configuration
- SYNC for cache injection
- ERQ for event input queue
The setup is slighly different on XIVE2 because the indexing has changed
for some of the tables, block ID or the chip topology ID can be used.
* XIVE2 features
SCOM and MMIO registers have a new layout and XIVE2 adds a new global
capability and configuration registers.
The lowlevel hardware offers a set of new features among which :
- a configurable number of priorities : 1 - 8
- StoreEOI with load-after-store ordering is activated by default
- Gen2 TIMA layout
- A P9-compat mode, or Gen1, TIMA toggle bit for SW compatibility
- increase to 24bit for VP number
Other features will have some impact on the Hypervisor and guest OS
when activated, but this is not required for initial support of the
controller.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
