Rather than hand-rolling scaled pointer arithmetic with casts and
shifts, let the compiler do so by indexing an array of GPRs, taking
advantage of the language's type system to scale based on whatever type
the register happens to be. This makes it easier to support CHERI where
the registers are capabilities, not plain integers, and so this pointer
arithmetic would need to change (and currently REGBYTES is both the size
of a register and the size of an integer word upstream).
Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Link: https://lore.kernel.org/r/20250709232932.37622-1-jrtc27@jrtc27.com
Signed-off-by: Anup Patel <anup@brainfault.org>
Since this persists in the preprocessed output (so that it can affect
the subsequent compilation), it ends up in the input to dtc and is a
syntax error, breaking the k210 build. Ideally we wouldn't add the
-include flag to DTSCPPFLAGS in the first place as this header is wholly
pointless there, but that's a more invasive build system change compared
to just making this header safe to include there.
Fixes: 86c01a73ff ("lib: sbi: Avoid GOT indirection for global symbol references")
Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Samuel Holland <samuel.holland@sifive.com>
Tested-by: Xiang W <wxjstz@126.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Link: https://lore.kernel.org/r/20250709232840.37551-1-jrtc27@jrtc27.com
Signed-off-by: Anup Patel <anup@brainfault.org>
The BSWAPxx() macros are now throwing the following warnings with
newer gcc versions. This is due to throwing an argument in that may
be evaluated more than one (I think) and therefore things like the
example below should be avoided.
Fix by making a set of BSWAPxx() wrappers which specifically only
evaluate 'x' once.
In file included lib/sbi/sbi_mpxy.c:21:
lib/sbi/sbi_mpxy.c: In function ‘sbi_mpxy_write_attrs’:
ib/sbi/sbi_mpxy.c:632:63: error: operation on ‘mem_idx’ may be undefined [-Werror=sequence-point]
632 | attr_val = le32_to_cpu(mem_ptr[mem_idx++]);
| ~~~~~~~^~
Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
Reviewed-by: Rahul Pathak <rahul@summations.net>
Reviewed-by: Xiang W <wxjstz@126.com>
Link: https://lore.kernel.org/r/20250704122938.897832-1-ben.dooks@codethink.co.uk
Signed-off-by: Anup Patel <anup@brainfault.org>
Using hsm stop in hsm wait loop causes secondary harts to be stuck
forever in OpenSBI on RISC-V platforms where HSM hart hotplug is
available and all harts come-up at the same time during system
power-on.
For example, lets say we have two harts A and B on a RISC-V platform
with HSM hart hotplug which come-up at the same time during system
power-on. The hart A enters OpenSBI before hart B hence it becomes
the primary (or cold-boot) hart whereas hart B becomes the secondary
(or warm-boot) hart. The hart A follows the OpenSBI cold-boot path
and registers hsm device before hart B enters OpenSBI. The hart B
eventually enters OpenSBI and follows the OpenSBI warm-boot path
so it will increment it's own entry_count before entering hsm wait
loop where it sees hsm device and stops itself. Later as part of
the Linux boot-up sequence, hart A issues SBI HSM start call to
bring-up hart B but OpenSBI sees entry_count != init_count for
hart B in sbi_hsm_hart_start() hence hsm_device_hart_start() is
not called for hart B resulting in hart B stuck forever in OpenSBI.
To fix the above issue, revert entry_count before doing hsm stop
in hsm wait loop.
Fixes: d844deadec ("lib: sbi: Use hsm stop for hsm wait")
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Nick Hu <nick.hu@sifive.com>
Link: https://lore.kernel.org/r/20250527124821.2113467-1-apatel@ventanamicro.com
Signed-off-by: Anup Patel <anup@brainfault.org>
Zicntr extension specifies three read-only CSRs, time, cycle and
instret. It isn't sufficient to report Zicntr is fully supported with
only time CSR detected.
This patch introduces a bitmap to sbi_hart_features to record
availability of these CSRs, which are detected using traps. Zicntr is
reported as present if and only if three CSRs are all available on the
HARTs.
Sites originally depending on SBI_HART_EXT_ZICNTR for detecting
existence of time CSR are switched to detect SBI_HART_CSR_TIME instead.
Suggested-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Yao Zi <ziyao@disroot.org>
Reviewed-by: Anup Patel <anup@brainfault.org>
Link: https://lore.kernel.org/r/20250516133352.36617-3-ziyao@disroot.org
Signed-off-by: Anup Patel <anup@brainfault.org>
It seems that current implementation doesn't fail on fdt_mpxy_init(),
because platforms might not have any MPXY devices. In fact, if there are
no MPXY devices, fdt_driver_init_all() will return SBI_OK.
More importantly, if there is any MPXY device which fails the
initialization, OpenSBI must check the error code and stop the booting.
Thus, this commit adds the return value for fdt_mpxy_init().
Signed-off-by: Alvin Chang <alvinga@andestech.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Link: https://lore.kernel.org/r/20250430091007.3768180-1-alvinga@andestech.com
Signed-off-by: Anup Patel <anup@brainfault.org>
The P2A doorbell system MSI index is expected to be discovered from
device tree instead of RPMI system MSI service group attribute. This
is based on ARC feedback before RPMI spec was frozen.
Let's parse P2A doorbell system MSI index from device tree and also
expose it as rpmi channel attribute to RPMI client drivers.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Link: https://lore.kernel.org/r/20250512083827.804151-3-apatel@ventanamicro.com
Signed-off-by: Anup Patel <anup@brainfault.org>
Most CSRs are XLEN bits wide, but some are 64 bit, so rv32 needs two
accesses, plaguing the code with ifdefs.
Add new helpers that split 64 bit operation into two operations on rv32.
The helpers don't use "csr + 0x10", but append "H" at the end of the csr
name to get a compile-time error when accessing a non 64 bit register.
This has the downside that you have to use the name when accessing them.
e.g. csr_read64(0x1234) or csr_read64(CSR_SATP) won't compile and the
error messages you get for these bugs are not straightforward.
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Radim Krčmář <rkrcmar@ventanamicro.com>
Link: https://lore.kernel.org/r/20250429142549.3673976-3-rkrcmar@ventanamicro.com
Signed-off-by: Anup Patel <anup@brainfault.org>
Similarly to what is done for SPELP, handle SSTATUS.SDT upon event
injection. In order to mimick an interrupt, set SDT to 1 for injection and
save its previous value in interrupted_flags[5:5]. Restore it upon
completion.
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
As raised during the ARC review, SPELP was not handled during the event
injection process. Save it as part of the interrupted flags, clear it
before injecting the event and restore it after completion.
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
As raised by Andrew on the kvm-unit-test review, this flags are meant to
hold SSTATUS bits in the specification. Rename them to match that.
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
The latest specification added new high priority RAS events and renamed
the PMU to PMU_OVERFLOW.
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Simplify the code and improve consistency by using the new macros where
possible. sbi_hart_count() obsoletes sbi_scratch_last_hartindex().
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
There is currently no helper for iterating through the harts in a
system, and code must choose between sbi_scratch_last_hartindex() and
sbi_platform_hart_count() for the loop condition.
sbi_scratch_last_hartindex() has unusual semantics, leading to the
likelihood of off-by-one errors, and sbi_platform_hart_count() is
provided by the platform and so may not be properly bounded.
Add a new helper which definitively reports the number of harts managed
by this OpenSBI instance, i.e. the number of valid hart indexes, and a
convenient iterator macro.
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
The compiler generates much better code for sbi_hartindex_to_hartid()
and sbi_hartindex_to_scratch() when using a constant for the bounds
check. This works out nicely because the underlying arrays are already
a constant size, so the only change needed is to fill the remainder of
each array with the appropriate default/out-of-bounds value. The
ellipsis in the designated initializer is a GCC extension (also
supported by Clang), but avoids runtime initialization of the array.
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
The internal limit on the number of harts is SBI_HARTMASK_MAX_BITS, as
this value determines the size of various bitmaps and arrays (including
hartindex_to_hartid_table and hartindex_to_scratch_table). Clamp the
value provided by the platform, and drop the extra array element.
Update the documentation to indicate that hart_index2id must be sized
based on hart_count, and that hart indexes must be contiguous. As of
commit 5e90e54a1a ("lib: utils:Check that hartid is valid"), there is
no restriction on the valid hart ID values.
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
OpenSBI is compiled with -fPIE, which generally implies dynamic linking.
This causes the compiler to generate GOT references for global symbols
in order to support runtime symbol interposition. However, OpenSBI does
not actually perform dynamic linking, so the GOT indirection just adds
unnecessary overhead.
The GOT references can be avoided by declaring global symbols with
hidden visibility, thus making them local to this dynamic object and
non-interposable. GCC/Clang's -fvisibility parameter is insufficient for
this purpose when referencing objects from other translation units;
either __attribute__((visibility(...)) or the pragma is required. Use
the pragma since it is easier to apply to every symbol. Additionally
clean up the one GOT reference from inline assembly.
With this change, a firmware linked with LLD does not contain either a
GOT or a PLT, and a firmware linked with BFD ld contains only a GOT with
a single (unreferenced, legacy) _GLOBAL_OFFSET_TABLE_ entry.
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
This patch adds unit tests for verifying the sbi_ecall version,
impid handling, and extension registration functions. The tests
ensure that the extension registration and unregistration work
as expected.
Signed-off-by: Akshay Behl <akshaybehl231@gmail.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Add SBI_ERR_DENIED_LOCKED and set it as the SBI_LAST_ERR which was
wrongly set to SBI_ERR_BAD_RANGE.
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
This reduces .bss size by 8 KiB, and should reduce overall memory usage
since most platforms will have significantly fewer than 512 entries in
this table. At the same time, it removes the fixed table size limit.
Since the table is only used within fdt_pmu.c, instead of updating the
extern declaration, make the table local to this file.
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
Currently, the sbi_sse_init() in cold boot path is called after
sbi_domain_finalize() so boot HART of non-root domains will start
before SSE cold boot init which can cause warm boot of such HARTs
to crash in sbi_sse_init().
To address the above issue, factor-out the non-root domain startup
from sbi_domain_finalize() function as a separate sbi_domain_startup()
function which can be called after sbi_sse_init() in cold boot path.
Fixes: 93f7d819fd ("lib: sbi: sse: allow adding new events")
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
The latest SBI 3.0 spec defines a new sbi_mpxy_get_shmem_size()
function and simplifies sbi_mpxy_set_shmem() function so update
the MPXY framework and SBI extension accordingly.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
The supervisor software can directly receive most of the system MSIs
except P2A doorbell and MSIs preferred to be handled in M-mode.
Add MPXY RPMI mailbox client driver for the System MSI service group.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
The service group ID assignment and some of the BASE services have
changes in the latest RPMI specification so let's update the RPMI
implementation accordingly.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Samuel Holland <samuel.holland@sifive.com>
Use the RPMI mailbox channel attributes to populate MPXY channel
attributes instead of hard coding them.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
To allow clients query service group version of a RPMI mailbox channel,
implement get_attribute() callback for the RPMI shared memory mailbox
controller.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
The rpmi_mailbox.h uses structures defined in mailbox.h so let's
include mailbox.h in rpmi_mailbox.h header.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Samuel Holland <samuel.holland@sifive.com>
Currently, the common MPXY RPMI mailbox client does not support
notifications so no need for dummy notifications support.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Reviewed-by: Samuel Holland <samuel.holland@sifive.com>
Some of the RPMI service groups may need additional context and
special handling when transferring messages via underlying mailbox
channel so introduce optional MPXY RPMI service group operations
for this purpose.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
Instead of having one common FDT MPXY RPMI mailbox client drivers
for various RPMI service groups, split this driver into two parts:
1) Common MPXY RPMI mailbox client library
2) MPXY driver for RPMI clock service group
The above split enables having a separate MPXY driver for each
RPMI clock service group and #1 (above) will allow code sharing
between various MPXY RPMI drivers.
Signed-off-by: Anup Patel <apatel@ventanamicro.com>
For driver subsystems that are not tightly integrated into the OpenSBI
init sequence, it is not important that the drivers are initialized in
any particular order. By putting all of these drivers in one array, they
can all be initialized with a single pass through the devicetree. This
saves about 10 ms of boot time on HiFive Unmatched.
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Anup Patel <anup@brainfault.org>
