lib: sbi: Resolve the uninitialized complaint in sbi_pmu

A recent commit (b28f0700) turns to read from CSR_MCOUNTINHIBIT after
checking the SBI_HART_HAS_MCOUNTINHIBIT HART feature, which leaves the
mctr_inhbt uninitialized in pmu_ctr_find_hw().
As a result, compiler will complain:
error: 'mctr_inhbt' may be used uninitialized in this function.

This commit resolves the issue by assigning an initial value to mctr_inhbt.

Signed-off-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: Fix missing space

Fix missing space in sbi_hart_pmp_configure
error printing.

Signed-off-by: Nikita Shubin <n.shubin@yadro.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Xiang w <wxjstz@126.com>

lib: sbi: Fix spelling of "address" in sbi_domain.c

Fix a spelling typo in error print.

Signed-off-by: Vagrant Cascadian <vagrant@debian.org>
Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: utils: Add LiteX UART support

Add support for the UART provided by the LiteX SoC
framework (https://github.com/enjoy-digital/litex),
based on its FDT info (described in the Linux tree at
Documentation/devicetree/bindings/serial/litex,liteuart.yaml).

Signed-off-by: Gabriel Somlo <gsomlo@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: clear pmpcfg.A before setting in pmp_set()

We should clear A bits in prot variable before enabling A_NA4 or A_NAPOT.

Signed-off-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

platform: sifive_fu740: add platform reset driver

da9063 PMIC can be used to reset/shutdown the
Sifive Unmatched board.

shutdown is done simply by writing SHUTDOWN bit to
DA9063_REG_CONTROL_F register.

reset via setting WAKEUP bit in DA9063_REG_CONTROL_F
register followed by masking POWER and POWER1 domains
and setting STANDBY bit in DA9063_REG_CONTROL_A,
originally discovered by Alexandre Ghiti on linux-riscv
maillists.

Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Tested-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Signed-off-by: Nikita Shubin <n.shubin@yadro.com>

lib: utils/i2c: Add minimal SiFive I2C driver

Minimum SiFive I2C driver to read/send bytes over I2C bus.

This allows querying information and perform operation of onboard PMIC,
as well as power-off and reset.

Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Tested-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Signed-off-by: Nikita Shubin <n.shubin@yadro.com>

lib: utils/i2c: Add simple FDT based I2C framework

FDT based I2C framework on the top of I2C library.

The drivers are probed on demand by fdt_i2c_adapter_get
function.

Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Tested-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Signed-off-by: Nikita Shubin <n.shubin@yadro.com>

lib: utils/i2c: Add generic I2C configuration library

Helper library to keep track of registered I2C adapters,
identified by dts offset, basic send/read functions and
adapter configuration (enable, set dividers, etc...).

Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Tested-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Signed-off-by: Nikita Shubin <n.shubin@yadro.com>

lib: utils/reset: separate driver init func

Move driver init code to separate function, so it can be reused
elsewhere.

Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Tested-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Signed-off-by: Nikita Shubin <n.shubin@yadro.com>

lib: utils/reset: add priority to gpio reset

Make gpio_system_reset_check return priority instead of just true/false.

Make default 128 priority for reset/shutdown.

Tested-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Tested-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Signed-off-by: Nikita Shubin <n.shubin@yadro.com>

lib: sbi: Fix PMP address bits detection

We should ensure that pmpcfg0.pmp0cfg is set to zero before using
pmpaddr0 CSR for detecting implemented PMP address bits.

Fixes: bf21632860b4 ("lib: sbi: Detect PMP granularity and number
of address bits")
Signed-off-by: Vasan VS <vasan.vs@gmail.com>
Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>

lib: sbi: Fix initial value mask while updating the counters

The first 32 bits of the initial value for the counter should be
preserved while updating the mhpmcounter for 32bit.

Fixes: 13d40f21d588e ("lib: sbi: Add PMU support")
Signed-off-by: Atish Patra <atish.patra@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils: Rename the prefix in PMU DT properties

As per the DT schema rules, the prefix should be vendor. As the PMU
properties are generic for all vendors, change the prefix to riscv
instead of pmu.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Enable PMU extension for platforms without mcountinhibit

Some platforms such as hifive unmatched doesn't implement mcountinhibit
csr. However, it has hardware events that can be monitored using 2
hpmcounter it has (i.e. mhpmcounter3 & mhpmcounter4).

Currently, PMU extension disabled if mcountinhibit is absent. That's not
really necessary as long as the supervisor OS keeps track of the delta
value of the counters. Without mcountinhibit, the delta value won't be
entirely accurate because the counters are freely running. However, that
should be fine to produce an approximate counter value which can help
performance analysis. Perf sampling won't work though as sscof extension
is not present in hifive unmatched.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Counter info width should be zero indexed

The mhpm bits represent the number of bits available in mhpmcounter
while counter width describes a zero indexed value. Fix the counter width
calculation.

Fixes: 13d40f21d588 ("lib: sbi: Add PMU support")
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Reset the mhpmevent value upon counter reset

The hardware solely relies on the event selector value in mhpmevent
to figure out what event to monitor using that counter. It should be
reset when counter reset happens.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Allow programmable counters to monitor cycle/instret events

A platform may use programmable counters for cycle/instret events.
The priv spec allows that provided that cycle/instret also report those
events in addition to the programmable counters. We should allow that
functionality in OpenSBI.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Always enable access for all counters

OpenSBI doesn't use any counters for its own usage. Thus, all the counters
can be made accessible for lower privilege mode always. However, the
mcountinhibit must be set so that the counter doesn't increment.
As a result, we don't have to enable/disable mcounteren at every start/stop.

Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Support sscofpmf extension in OpenSBI

This patch adds sscofpmf extension in pmu module which includes
following things.

1. Enable overflow irq when starting a counter.
2. Setting the correct event filters passed from supervisor.
3. Delegating the overflow interrupt to the supervisor.
4. Add RV32 support for sscofpmf.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Delegate PMU counter overflow interrupt to S mode

OpenSBI doesn't handle PMU counters for now.

Delegate the overflow counter to S-mode always.

Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Detect Sscofpmf extension at run time

Sscofpmf ISA extension introduces PMU counter overflow and filtering support.
It introduces a read only `scountovf` csr that can be used to detect if
a hart supports this extension at runtime. However, this feature is only
useful if the hart already supports mcounteren and mcountinhibit.

Add a dynamic detection mechanism and boot time print message if sscofpmf
is present.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Update csr_read/write_num for PMU

The Sscofpmf extension introduces mhpmevent[h] csrs to handle filtering
/overflow bits in RV32. There is no way to read/write mcountinhibit
using mcountinhibit csr using a variable.

Updated the support to read/write mhpmevent[h] and mcountinhibit csr.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

riscv: Add new CSRs introduced by Sscofpmf[1] extension

[1] https://drive.google.com/file/d/1KcjgbLM5L1ZKY8934aJl8aQwGlMz6Cbo/view

Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: error handling in fdt_reset_init()

The initialization of a reset driver may fail for various reasons, like
a PMIC based reset driver not finding the required I2C driver. The return
code of the init routine may take other error values than -ENODEV.

If the initialization of a reset driver fails, this should not lead to the
board hanging. It is enough that the reset driver does not call
sbi_system_reset_add_device() to avoid invoking the driver for a device
that could not be initialized.

Change the return type of fdt_reset_init() to void.
Print a message if an error occurs.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

platform: generic: move fdt_reset_init to final_init

Move the fdt_reset_init() invocation from generic_early_init() to
generic_final_init(). This allows to print error messages.

Ignore the return value of fdt_reset_init() as we should not stop booting
due to failure to initialize reset drivers.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: move sbi_boot_print_general()

Moving the sbi_boot_print_general() call after the
sbi_platform_final_init() call allows to print devices initialized in the
latter.

To keep the overall print sequence the same also move
sbi_boot_print_domains().

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils/gpio: use list for drivers

Convert static array to sbi_list.

This removes size limitation, makes add/remove more efficient and
saves space.

Signed-off-by: Nikita Shubin <n.shubin@yadro.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: Fix GPA passed to __sbi_hfence_gvma_xyz() functions

The parameter passed to HFENCE.GVMA instruction in rs1 register
is guest physical address right shifted by 2 (i.e. divided by 4).

Unfortunately, we overlooked the semantics of rs1 registers for
HFENCE.GVMA instruction and never right shifted guest physical
address by 2. This issue did not manifest for hypervisors till
now because all H-extension implementations (such as QEMU, Spike,
Rocket Core FPGA, etc) we tried till now were conservatively
flushing everything upon any HFENCE.GVMA instruction.

This patch fixes GPA passed to __sbi_hfence_gvma_vmid_gpa()
and __sbi_hfence_gvma_gpa() functions.

Fixes: 331ff6a162c1 ("lib: Support stage1 and stage2 tlb flushing")
Reported-by: Ian Huang <ihuang@ventanamicro.com>
Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>

include: sbi_utils: Introduce an helper to get fdt base address

This simply adds an helper to get fdt address which is more explicit than
sbi_scratch_thishart_arg1_ptr.

Signed-off-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: Refine addr format in sbi_printf

Although we have PRILX to help us print unsigned long without
considering the 32bit/64bit differences, there are still some
places using 08lx and 016lx manually --- leading to redundant code.

This commit fixes the issue by using PRILX all the time.

Signed-off-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils/reset: Register separate GPIO system reset devices

Now that sbi_system support multiple system reset devices, we should
register separate devices for GPIO restart and GPIO poweroff because
DT describes these as separate devices with different compatible
strings.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>

lib: utils/irqchip: Automatically delegate T-HEAD PLIC access

The T-HEAD PLIC implementation requires setting a delegation bit
to allow access from S-mode. Now that the T-HEAD PLIC has its own
compatible string, set this bit automatically from the PLIC driver,
instead of reaching into the PLIC's MMIO space from another driver.

Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

Makefile: Add build time and compiler info string

When we are doing opensbi development, we want to know the build time
and compiler info for debug purpose.

To enable this message, please add "BUILD_INFO=y", like:

```
make BUILD_INFO=y
```

NOTE: Using `BUILD_INFO=y` without specifying SOURCE_DATE_EPOCH will
violate "reproducible builds". So it's ONLY for development and debug
purpose, and should NOT be used in a product which follows "reproducible
builds".

Signed-off-by: Wei Fu <wefu@redhat.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>

lib: sbi: system reset with invalid parameters

The SBI specification requires that sbi_system_reset() returns
SBI_ERR_INVALID_PARAM if reset_type or reset_reason are not valid.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: sbi: Save context for all non-retentive suspend types

Instead of saving context only for default non-retentive suspend,
we should save context for all non-retentive suspend types.

Fixes: 74756891cc35 ("lib: sbi: Implement SBI HSM suspend function")
Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: sbi: add priority for reset handler

Let's make system_reset_check returning priority instead of only
true/false. In that case 0 - means not supported, and anything above
means priority that makes existing reset handlers being used in first
place, unless it is decided to lower their priority.

The handler with the most priority wins.

Signed-off-by: Nikita Shubin <n.shubin@yadro.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: convert reset to list

To support different handlers for different types of resets, we are
adding a sbi_list of restart handlers.

Instead of sbi_system_reset_set_device we use
sbi_system_reset_add_device to reflect the actual meaning.

Signed-off-by: Nikita Shubin <n.shubin@yadro.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils: identify supported GPIO reset methods

The GPIO reset driver supports reset and poweroff. But not all boards
support both. gpio_system_reset_check() must detect this situation.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils/reset: use sbi_timer_mdelay() in gpio reset driver

We should use sbi_timer_mdelay() instead of custom gpio_mdelay() in
the gpio reset driver.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: sbi: Add generic timer delay loop function

We now have frequency of the timer device provided by the platform
support so we can emulate desired delay using a loop where the number
loop iterations are based on timer frequency.

This patch provides sbi_timer_delay_loop() for above purpose.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: sbi: Print timer frequency at boot time

We now have frequency in timer device instance provided by platform
so let's print timer frequency as part of the boot prints.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: sbi: Add timer frequency to struct sbi_timer_device

Generic mdelay() and udelay() functions can be provided by the
sbi_timer framework if timer frequency is available in the timer
instance provided by the platform support or timer driver.

This patch adds timer frequency (timer_freq) member in the
struct sbi_timer_device for above purpose.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: utils/fdt: Add fdt_parse_timebase_frequency() function

We add fdt_parse_timebase_frequency() function which can be used
by ACLINT mtimer driver and platform code to get timebase frequency.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>

lib: sbi: Add BUG() macro for csr_read/write_num() and misa_string()

We use BUG() macro in csr_read_num(), csr_write_num(), and
misa_string() functions for unhandled cases.

Signed-off-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: add some macros to detect BUG at runtime

Three macros are added. One is called BUG, which is used to put in an
unreachable branch. One is called BUG_ON, which is used to check bugs
and assert conditions are opposite. One is called SBI_ASSERT, used for
assertion checking.

Signed-off-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: fix ctz bug

The original version of ctz will cause an endless loop, if the parameter
passed in is 0. This commit fixes this bug.

Signed-off-by: Xiang W <wxjstz@126.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils/reset: Add a sunxi watchdog reset driver

One of the watchdogs in the D1 SoC provides a "soft reset" function,
which allows software to immediately reset the entire SoC. Add a driver
so it can implement the SBI system reset function.

Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils/reset: Sort fdt_reset driver list

In preparation for adding a new fdt_reset driver, ensure the existing
lists are sorted alphabetically.

Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: Refine the way to construct platform features

sbi_platform_get_features_str() uses sbi_snprintf() to construct the
features_str. However, it passes the wrong length value (i.e., the nfstr),
which should be (nfstr-offset) as the starting point of str (i.e.,
features_str + offset) changes.

This commit also checks the return value of snprintf, and handles the
corner case that the string buffer is full.

Signed-off-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Checking fifo validness in sbi_fifo_is_empty and is_full

As other exported fifo functions, we should check whether the fifo is
valid in sbi_fifo_is_empty and sbi_fifo_is_full. To this end, this patch
changes the retval from bool to int, and the two functions will return
SBI_EINVAL in the case the fifo is invalid.

Signed-off-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: protect dprintf output with spinlock

Avoid getting messages from multiple harts (using dprintf and printf)
concurrently with a spinlock serializaing calls to sbi_dprintf(),
sbi_printf() and sbi_puts()

Signed-off-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: Correct typo in faults delegation CSR name

Correcting the name of faults delegation CSR
%s/mfdeleg/medeleg

Signed-off-by: Rahul Pathak <rpathak@ventanamicro.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

payloads/test: Add support for SBI v0.2 ecalls

It can be useful to make SBI v0.2 or newer ecalls from this payload
for testing purposes. To support this, convert the macros to use the
extension/function parameter convention.

Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: sbi_trap: Restore redirect for access faults

commit 764a17d852a8 ("lib: sbi: Implement firmware counters") added
switch cases for CAUSE_LOAD_ACCESS and CAUSE_STORE_ACCESS. This caused
them to stop being redirected to U or S mode, as that is handled in the
default switch case. As a result, an error in userspace could cause the
system to hang. Fix this by allowing the acces fault case to fall
through to the default case.

Fixes: 764a17d852a8 ("lib: sbi: Implement firmware counters")
Signed-off-by: Samuel Holland <samuel@sholland.org>
Reviewed-by: Atish Patra <atish.patra@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>

lib: sbi: Fix bug in sbi_ecall_rfence that misses checking

In sbi_ecall_rfence_handler, it will compare the funcid with
REMOTE_HFENCE_GVMA and REMOTE_HFENCE_VVMA_ASID. Later it check
whether the misa includes H-extension. This checking is incomplete
which misses REMOTE_HFENCE_GVMA_VMID and REMOTE_HFENCE_VVMA.

Fix the issue by updating the checking range.

Signed-off-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/timer: Simplify MTIMER synchronization

We simplify MTIMER synchronization as follows:

1) Detect MTIMER devices with unique (or non-shared) MTIME
   register at boot-time
2) Select first MTIMER device with no associated HART as our
   reference MTIMER device
3) Only synchronize MTIMER devices with unique (or non-shared)
   MTIME register using reference MTIMER device
4) Directly update the MTIME register at time of synchronization
   because MTIME is a read/write register.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/timer: Allow ACLINT MTIMER supporting only 32-bit MMIO

We can have ACLINT MTIMER devices which only support 32-bit MMIO
accesses on RV64 system so this patch adds a boolean DT property
"mtimer,no-64bit-mmio" to detect this from MTIMER DT node.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/fdt: Extend fdt_parse_aclint_node() function

The fdt_parse_aclint_node() is used to parse DT node for SiFive
CLINT, ACLINT MTIMER, and ACLINT MSWI devices.

The ACLINT MTIMER has undergone following changes:
1) MTIMER DT node now requires separate addresses in for MTIME
   register and MTIMECMPx registers in the reg DT property.
2) MTIMER DT node might have no interrupts-extended DT property
   when the MTIMER device has no associated HARTs (i.e. the
   MTIMER device has no MTIMECMPx registers)

This patch extends fdt_parse_aclint_node() to handle above
mentioned changes in ACLINT MTIMER DT bindings.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils: Extend fdt_get_node_addr_size() for multiple register sets

We add "index" parameter to fdt_get_node_addr_size() API so that
calling function can specify index of desired register set. This
will allow fdt_get_node_addr_size() to handle DT nodes with
multiple register sets.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/timer: Allow separate base addresses for MTIME and MTIMECMP

We extend the ACLINT library to support separate base addresses
for MTIME and MTIMECMP registers.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/fdt: Change addr and size to uint64_t

The maximum address and size encoded in DT are 64-bit numbers, so we
should use uint64_t for 'addr' and 'size' in fdt_get_node_addr_size().

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: Fix bug in strncmp function when count is 0

No need to compare characters when the count turns to 0.
Fix the issue in sbi_strncmp.

Signed-off-by: Dong Du <Dd_nirvana@sjtu.edu.cn>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils: support both of gpio-poweroff, gpio-reset

The generic GPIO reset driver has two entries in the match table:
"gpio-poweroff", "gpio-reset". Only the first entry is considered by
fdt_reset_init().

Define "gpio-poweroff" and "gpio-reset" as compatibility strings of two
separate reset drivers. They still can share code.

Fixes: e3d6919d10d7 ("lib: utils/reset: Add generic GPIO reset driver")
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils: remove unused variable in fdt_reset_init

The value of variable current_driver is unused. Remove the variable.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

generic: fu740: add workaround for CIP-1200 errata

Instruction TLB can fail to respect a non-global SFENCE thus we need to
flush the TLB using SFENCE.VMA x0, x0

See full description of CIP-1200 in Errata_FU740-C000_20210205 from
https://www.sifive.com/boards/hifive-unmatched

Signed-off-by: David Abdurachmanov <david.abdurachmanov@sifive.com>
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>

firmware: use _fw_start for load address

The previous code uses _start as the load address, this default .entry is
the first segment, using _fw_start does not need to make this assumption.

Signed-off-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

Makefile: Manually forward RELAX_FLAG to the assembler when linking with LLD

When generating code with -mno-relax, GCC puts .option norelax in
the generated assembly, and so doesn’t bother passing on -mno-relax
to the assembler. This has the unfortunate effect that, when using
GCC to assemble hand-written assembly, -mno-relax does nothing,
and we have to pass -Wa,-mno-relax to manually forward it to the
assembler.

This is an old GCC bug that was fixed [1] recently. For the time
being, let's pass "-Wa,-mno-relax" to ASFLAGS for the GCC + LLD
combination to work, e.g.:

  $ make CC=riscv64-unknown-elf-gcc LLVM=1 PLATFORM=generic

[1] https://github.com/gcc-mirror/gcc/commit/3b0a7d624e64eeb81e4d5e8c62c46d86ef521857

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

README: Update toolchain information

Recent FW_PIC=y changes actually require toolchains with PIE support
and it is on by default. Existing doc uses a GNU bare-metal toolchain
as examples but it does not support PIE.

Replace references of bare-metal toolchain prefix with Linux toolchain
prefix everywhere in documentation.

Also indicate that "riscv64-unknown-freebsd-" as an alternative to
"riscv64-linux-" GNU toolchain.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

docs: Document FW_PIC compile time option

FW_PIC=y is on by default, but the doc is missing when this was
introduced. Add some description for it.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

docs: Document parameters passed to firmware and alignment requirement

This updates documentation to describe parameters passed to firmware
from previous booting stage, and corresponding address alignment
requirement.

This also fixes a typo in fw_dynamic.md (it's => its).

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

firmware: Remove the unhelpful alignment codes before fdt relocation

If the device tree is at an address that is not __SIZEOF_POINTER__
aligned, the fdt relocation code tries to align both source and
destination address to __SIZEOF_POINTER__ before the memory copy.
But such alignment can lead to unexpected results if either source
or destination address is not aligned.

In fact libfdt requires that the device tree must be at an 8-byte
aligned address. Hence remove the unhelpful alignment codes.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

include: types: Use __builtin_offsetof when supported

Clang provides a __builtin_offsetof which can be detected using
__has_builtin().

Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

firmware: use __SIZEOF_LONG__ for field offsets in fw_dynamic.h

The fields are of type unsigned long and are not pointers. While this
happens to be the same for RV32/RV64, it is not correct when compiling
for a CHERI-RISC-V system where pointers are twice the size of long.

Signed-off-by: Alex Richardson <Alexander.Richardson@cl.cam.ac.uk>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: utils/reset: Add generic GPIO reset driver

We add generic GPIO reset driver inspired from gpio-restart
and gpio-poweroff drivers of Linux kernel.

Signed-off-by: Green Wan <green.wan@sifive.com>
Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/gpio: Add minimal SiFive GPIO driver

We add a minimal SiFive GPIO driver so that we can do GPIO based
system power-off and reset.

Signed-off-by: Green Wan <green.wan@sifive.com>
Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/gpio: Add simple FDT based GPIO framework

We add a simple FDT based GPIO framework which is built on top
of generic GPIO library. The phandle of FDT GPIO chip DT node
is treated as unique GPIO chip ID required by the generic GPIO
library. The FDT based GPIO chip drivers will be probed on-demand
from fdt_gpio_pin_get() called by the GPIO client drivers.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/gpio: Add generic GPIO configuration library

We add generic GPIO configuration library which is independent of
hardware description format (FDT or ACPI). The OpenSBI platform
support or GPIO drivers can register GPIO chip instances which
can be discovered and used by different GPIO clients. Each GPIO
chip instance has a unique ID which can be used by GPIO clients
to lookup GPIO chip instance.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils/fdt: Add fdt_parse_phandle_with_args() API

The libfdt project does not have a generic API to parse phandle
with args from a DT node so we add fdt_parse_phandle_with_args()
for this purpose. This new API will be useful to FDT based drivers.

Signed-off-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

lib: utils: Drop dependency on libgcc by importing part of FreeBSD's libquad

We only need libgcc for 64-bit division on RV32. Whilst GCC toolchains
bundle libgcc, Clang toolchains tend not to ship libclang_rt.builtins
given every compiler is a cross-compiler for every target and so you
would need a silly number of builds of it, with only the native library
available; only vendor-provided Clang toolchains specifically for bare
metal cross-compiling are likely to provide it.

Thus, import part of FreeBSD's implementation of the division support
functions needed and stop linking against libgcc.

Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Tested-by: Anup Patel <anup.patel@wdc.com>

Makefile: Support building with Clang and LLVM binutils

This is intended to mirror the Linux kernel. Building with CC=clang will
use Clang as the compiler but default to using the existing binutils.
Building with LLVM=1 will default to using Clang and LLVM binutils.

Whilst GCC will accept the -N linker option and forward it on to the
linker, Clang will not, and so in order to support both compilers we
must use -Wl, to forward it to the linker as is required for most other
linker options.

Note that there is currently a bug when using Clang as the compiler and
riscv64-linux-gnu-ld as the linker for FW_PIC=y. At first glance this
appears to be a bug in GNU binutils, but this could also be Clang or
OpenSBI at fault in some subtle way. Thus, for now, advise that this
combination be avoided.

Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Tested-by: Anup Patel <anup.patel@wdc.com>

firmware: Only default FW_PIC to y if supported

Bare-metal GNU ld does not support PIE, so if using it this will result
in a failure to build. Instead, default to FW_PIC=n if not supported.
Note that an explicit FW_PIC=y is not overridden, to ensure the build
fails rather than silently producing a position-dependent binary.

Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Tested-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

firmware: Explicitly pass -pie to the linker, not just the driver

When using Clang with a bare-metal triple, -pie does not get passed to
the linker as it's not normally a thing that makes sense, unlike GCC
which will unconditionally forward it on and potentially result in a
linker error. However, LLD does support it, and manually forwarding it
on works as desired, so do so to fully support FW_PIC with Clang and
LLD.

Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Tested-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

fw_base: Put data in .data rather than .text

The -N linker option is supposed to make .text writable, but GNU ld and
LLD differ in interpreting what that means. GNU ld will happily let you
have relocations in it, but LLD will see that the input section is
read-only (even though the output section is writable) and give an
error. It's unclear if either of them intend to have that behaviour in
this edge case, but regardless there's no reason not to just put the
data in a writable .data section.

Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Tested-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

fw_base: Don't mark fw_platform_init as both global and weak

These are mutually exclusive. GNU as and LLVM both let later binding
directives override earlier ones so this works as intended, but LLVM 12
turned this into a warning as there's no good reason to do such a thing
and could be a potential bug. Thus, remove the redundant and incorrect
.globl directive for fw_platform_init.

Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Tested-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

lib: sbi: Fix sbi_pmu_exit() for systems not having MCOUNTINHIBIT csr

The sbi_pmu_exit() crashes on systems not having MCOUNTINHIBIT csr
so to fix this we check SBI_HART_HAS_MCOUNTINHIBIT feature in
sbi_pmu_exit() and do nothing if it is not available.

Fixes: 13d40f21d588 ("lib: sbi: Add PMU support")
Signed-off-by: Anup Patel <anup.patel@wdc.com>

firmware: Define a macro for version of struct fw_dynamic_info

Avoid using a magic number, instead use a macro for the version of
struct fw_dynamic_info.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

firmware: Remove the sanity checks in fw_save_info()

The sanity checks on the magic and version was already done in
fw_boot_hart(), which happens before fw_save_info() is called.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

docs: Make <xyz> visible in the rendered platform guide

At present in the rendered platform guide, all instances of <xyz>
are missing. Use &lt; and &gt; to replace <> to make them visible.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

docs: Correct a typo in platform_guide.md

It's riscv-pk, not riskv-pk.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Anup Patel <anup.patel@wdc.com>

firmware: Minor optimization in _scratch_init()

Before entering _scratch_init(), register t3 already holds a copy
of the firmware end address, hence there is no need to calculate
it again. This reduces 3 instructions in each _scratch_init() loop.

Signed-off-by: Bin Meng <bmeng.cn@gmail.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Reviewed-by: Atish Patra <atish.patra@wdc.com>

platform: generic: Add PMU support

Add PMU support for generic platform. Generic platform solely relies on
the device tree to parse all pmu related information. If any event is
not described in device tree, generic platform will not support it.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Improve TLB function naming

Follow the standard conventon for static function names:

All global functions should be start with sbi_<module name>_
All static functions should be start with <module name>_

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Implement firmware counters

RISC-V SBI v0.3 specification defines a set of firmware events that can
provide additional information about the current firmware context. All
of the firmware event monitoring are enabled now. The firmware
events must be defined as raw perf event with MSB set as specified in the
specification.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Implement SBI PMU extension

RISC-V SBI specfication 0.3 defines a PMU extension that allows supervisor
mode to start/stop/configure pmu related events. This patch implements
all of the functionality defined in the specification.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

utils: fdt: Add fdt helper functions to parse PMU DT nodes

The PMU DT node bindings are defined in docs/pmu_support.md
Add few fdt helper functions to parse the DT node and update the
event-counter mapping tables.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Add PMU support

RISC-V SBI v0.3 specification defined a PMU extension to configure/start/stop
the hardware/firmware pmu events.

Implement PMU support in OpenSBI library. The implementation is agnostic of
event to counter mapping & mhpmevent value configuration. That means, it
expects platform hooks will be used to set up the mapping and provide
the mhpmevent value at runtime.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Add PMU specific platform hooks

A platform hook to initialize PMU allows platform vendors to provide
their own mechanism to define pmu event-counter mappings in addition
to the DT based approach.

Another platform hook that allows platform vendors customize the
final mhpmevent value configuration.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Use csr_read/write_num to read/update PMU counters

Currently, csr_read/write_num functions are used to read/write PMP related
CSRs where CSR value is decided at runtime. Expand this function to include
PMU related CSRs as well.

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Reviewed-by: Xiang W <wxjstz@126.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>

lib: sbi: Remove redundant boot time print statement

This patch removes redundant print from sbi_boot_print_hart().

Reviewed-by: Anup Patel <anup.patel@wdc.com>
Signed-off-by: Atish Patra <atish.patra@wdc.com>