OpenSBI vendor extension calls must not cause scheduling, as they're
part of M-mode trap handling. Thus, comment out nxsig_usleep() as
well, which is occasionally taken and crashes the system in that
case. Fix this by commenting out lines that have the potential to
cause scheduling.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Summary:
- I noticed that QEMU shows high CPU usage if the number of
CPUs does not match the kernel configuration. (e.g. -smp 8
and CONFIG_SMP_NCPUS=2)
- This commit fixes this issue.
Impact:
- qemu-rv only
Testing:
- Tested with the following configs
- rv-virt:smp64 (CONFIG_NCPUS=1/2/8)
- rv-virt:nsh64
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
It is not really needed; g_hart_stacks is only used during SBI init as
a temporary stack area. We can use the scratch area buffers for this, as
the scratch areas define almost 4K of extra space, which is used for
exception stacks anyway.
The L3 table address was calculated incorrectly. For every 2MiB of
mapped memory, an offset of 4KiB is needed from the base of the L3
table. The old calculation failed if paddr was not aligned to a 2MiB
boundary.
RX_FIFO_ADDRs and TX_FIFO_ADDR were misconfigured. These addresses
overlapped causing data corruption during high USB loads. For
example, data corruption was present during the following conditions:
1. Composite USB driver was used (CDC/ACM + Mass storage)
2. /dev/ttyACM0 was accessed instantly from Linux side when
starting up.
3. Training data was sent to /dev/ttyACM0 from NuttX from the
very beginning periodically.
It was observed that while Mass storage was negotiating, sometimes
data sent from NuttX to Linux via CDC/ACM was corrupt, although it
was sent properly on the TX fifo.
Also, don't access TXCSRL_REG_EPN_TX_FIFO_NE_MASK for EP0 as it's
not applicable.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Initialize .domains_init function entry point in sbi_platform_operations when
CONFIG_OPENSBI_DOMAINS=y. In this case, the board specific code must provide
the "board_domains_init" function.
Signed-off-by: Jouni Ukkonen <jouni.ukkonen@unikie.com>
With faster data transfer rates, it was seen that the read
requests occasionally were issued while the USB RX operation
was actually in progress. This patch makes sure the system
doesn't accidentally read the RX fifo while it's being filled
up, but rather, checks for the RXCSRL_REG_EPN_RX_PKT_RDY_MASK
flag. This flag indicates the packet is ready to be read.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
cache_dbus_mmu_set and cache_ibus_mmu_set return positive values in case
of errors, so DEBUVERIFY could never detect them since this macro checks
for negative values.
Besides, the successful execution of those functions is mandatory for
the reliable operation under Protected Mode, so the verification is
always performed, even when DEBUG is not enabled.
Signed-off-by: Gustavo Henrique Nihei <gustavo.nihei@espressif.com>
forget to update in this patch:
commit b02db04e00
Author: Xiang Xiao <xiaoxiang@xiaomi.com>
Date: Sun Jun 5 17:10:19 2022 +0800
arch/assert: Keep the thread dump column order same as ps
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
1.Move __XSTR from include/arch.h to include/irq.h
2.Move FLOAD/FSTORE and REGLOAD/REGSTORE from include/arch.h to src/common/riscv_internal.h
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
The "p" format specifier already prepends the pointer address with "0x"
when printing.
Signed-off-by: Gustavo Henrique Nihei <gustavo.nihei@espressif.com>
Starting the rptun with the autostart flag set will cause significant
delays at the boot, as it will wait for the master to be up. U-boot/linux
combination may take more than 10 seconds to boot to the point where the
rpmsg bus is initialized.
For now, the user needs to initialize the rptun separately, for example,
by issuing the following command:
rptun start /dev/rptun/mpfs-ihc
This command will also block if started before the rpmsg bus master is up.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Move the linker defined symbols to a separate file, so they can be
accessed without pulling in everything from riscv_internal.h and
whatever it includes (e.g. syscall.h drags in a lot).
Add a function to easily enable event handling on fabric and mss gpios. This
is similar to what exists e.g. for stm32 arm chips.
Also fix some small bugs in mpfs_configgpio related to IRQ bits configuration
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
This patch fixes the following issues:
1. MPFS_EMMCSD_HRS06_EMM bitmask had to be 0x7, not 0x03
2. putreg32() caused outright memory corruption as the
arguments were in wrong order
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
The old implementation needed a contiguous memory block for user
ROM/RAM. This is because there was only 1 L3 page table which can only
map a contiguous memory area.
Also, remove the PMP configuration which just complicates things,
rely on the MMU mappings instead.
Update PLL configuration parameters to match the values provided
by the vendor.
Also remove extra call to mpfs_pll_config() as it's already called
at mpfs_clockconfig().
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Summary:
- I noticed that sometimes uart shows nothing on the maix-bit board.
- This commit adds a workaround to avoid such the issue
Impact:
- k210 only
Testing:
- Tested with maix-bit
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
This provides USB composite (CDC/ACM and Mass Storage) support
for mpfs board. In addition, a number of USB fixes are included:
- Support for Setup Out packets
- Proper support for larger than packet size writes
- Finishing setup packets properly
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Output "X" with showprogress and make a system reset.
Silently ignoring failed training is dangerous and will cause random behaviour if DDR is used
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
since the code could map the unsupported work to the
supported one and remove select SCHED_WORKQUEUE from
Kconfig since SCHED_[L|H]PWORK already do the selection
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Linux kernel uses M-mode trap for handling Inter-Hart Communication (IHC).
This patch provides all the required functionalities for this purpose.
Previously, HSS bootloader was required. Now, NuttX is run as the
bootloader providing OpenSBI vendor extensions instead. This setup has
been tested on the following configuration:
- Hart 0 has NuttX in bootloader mode with OpenSBI
- Hart 1 unused
- Hart 2 has NuttX configured at 0xa2000000
- Hart 3 has U-boot / Linux kernel (at 0x80200000)
- Hart 4 has U-boot / Linux kernel (at 0x80200000)
Upon startup, NuttX on hart 0 will initialize SD-card driver, loads
the hart 2 NuttX from the SD-card and loads the U-boot to 0x80200000.
Also the nuttx.sbi -binary is loaded from SD-card into address 0x80000000,
which is also marked as reserved area in the Linux kernel device tree (for
the chuck 0x80000000 - 0x80200000).
Hart 2 NuttX waits until Linux kernel (IHC master) is started. After the
initial handshake, RPMsg / virtIO bus along with the IHC may be used for
proper AMP mode.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Summary:
- I noticed that rv-virt:knsh64 crashes when it executes the init.
- This commit fixes this issue.
Impact:
- None
Testing:
- Tested with rv-virt:knsh64
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
The FPU register saving upon vfork entry was missing.
Also added macro that tells the actual size of an FPU reg, instead
of just having a coefficient for qfpu/no-qfpu.
- Remove most of the ifeq-conditions and replace them with variables.
- Move the -I flag for 3rd party headers to opensbi/Make.defs
This clean-up / generalization makes it much simpler to add a new SBI
implementation, without the need to add a bunch of ifeq / elif conditions
to the makefile.
Let the user pick what runs on the harts. For example, the
default configuration now supports NuttX on hart2 and Linux
kernel on harts 3 and 4. Also fix a few issues in the code.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
g_mpfs_ack_sig and g_mpfs_rx_sig are better used with
SEM_INITIALIZER(0) (signalling) rather than with
SEM_INITIALIZER(1) (mutual exclusion).
Co-authored-by: Petro Karashchenko <petro.karashchenko@gmail.com>
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
PR#6249 was already merged without the review fixes. Provide the
fixes here on a separate patch.
Co-authored-by: Petro Karashchenko <petro.karashchenko@gmail.com>
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Summary:
- This commit adds MMU support for qemu-rv
- Please note that mtimer is disabled for S-mode because
the mtimer needs to be accessed in M-mode
Impact:
- qemu-rv only
Testing:
- Tested with rv-virt:knsh64 (will be pushed later)
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
This provides an example of Asymmetric Multiprocessing (AMP). The
master from Linux sends pings that this NuttX echoes back. The system
uses RPMsg from OpenAMP.
The Inter-Hart Communication module is present in the vendor's software
stack with the tag "2021.11". The software is present on github at the
polarfire-soc project. The following conditions must be met:
1. FPGA programmed with 2021.11 software
2. HSS (Vendor bootloader) with 2021.11 software
3. U-boot and Linux kernel from 2011.11 software
Currently the IHC works as a slave only on the hart number 4.
On the NuttX side, this patch uses rptun that incorporates rpmsg and
virtio. If it used only rpmsg and virtio, the future maintenance would
likely be much heavier. Using rptun also simplifies many things.
Upon success, the master side from Linux may issue an example test:
root@icicle-kit-es-amp:/opt/microchip/amp/rpmsg-pingpong# ./rpmsg-pingpong
However, the rpmsg-pingpong.c (compiled on target with gcc), may need to
be modified as seen below to match the device id:
- char *rpmsg_dev="virtio0.rpmsg-amp-demo-channel.-1.0";
+ char *rpmsg_dev="virtio0.rpmsg-amp-demo-channel.-1.1024";
This work uses a separate linker script. Due to a bug yet unknown to date,
a small NuttX, when loaded by the vendor HSS bootloader, will cause the
Linux kernel to hang at boot. Thus, the binary size is increased with
a section 'filler_area' whose only purpose is to increase the image size
so that the Linux kernel will boot up.
Signed-off-by: Eero Nurkkala <eero.nurkkala@offcode.fi>
Implement a generic access rights modification procedure instead
of the procedures that only do one thing (enable/disable write)
to one section (text).
Summary:
- Because a context switch issue still exists with FPU,
it should be enabled only if EXPERIMENTAL=y
Impact:
- None
Testing:
- Tested with ostest
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
When the .elf file is loaded from disk, the kernel must be given write
access to the allocated .text section in the task's address environment.
The access is removed after the elf is loaded and relocations are done.
NOTE:
The reason this works for the ARM implementation, is that the ARM MMU
can be configured to give write access for the privileged mode, but
revoke write access for the user mode.
Regardless, it would be smart to revoke write access even for the
kernel, when the kernel does not need it. This framework allows doing
that, if someone wishes to take up the task.
Summary:
- This commit adds support for semihosting and hostfs
Impact:
- None
Testing:
- Tested with nsh and nsh64 (defconfig will be updated later)
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Mappings are done with vaddr=paddr.
- I/O space mapped with two gigapages
- Kernel space mapped to statically allocated page tables. 2MB of kernel
memory is supported.
- Page pool is mapped to the kernel space, to allow virtual memory access
for the kernel e.g. to initialize the page memory when it is allocated.
by renaming riscv_dispatch_syscall to sys_callx, so the caller
don't need the immediate step(syscallx->riscv_dispatch_syscall)
Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com>
Using the Kconfig macro does not work for RISC-V target, as there the
user heap follows .data/.bss and does not obey any Kconfig provided
boundary.
Added stubs for ARM and Z80 also.
Copy the kernel mappings to the new (user) address environment. The
copyuing is done exactly once. This relies on the fact that the kernel
L1/L2 mappings will never change, as all of the kernel memory is mapped
upon boot.
This implements initial support for kernel build (address environments,
page allocator) for RISC-V.
This is done a bit differently compared to the ARMV7 implementation:
- Support implemented for Sv39 MMU, however the implementation should be
extensible for other MMU types also.
- Instead of preserving and moving the L1 references around, a canonical
approach is used instead, where the page table base address register
is switched upon context switch.
- To preserve a bit of memory, only a single L1/L2 table is supported,
this gives access to 1GiB of virtual memory for each process, which
should be more than enough.
Some things worth noting:
- Assumes page pool is mapped with vaddr=paddr mappings
- The CONFIG_ARCH_XXXX_VBASE and CONFIG_ARCH_XXXX_NPAGES values are
ignored, with the exception of CONFIG_ARCH_DATA_VBASE which is used
for ARCH_DATA_RESERVE
- ARCH_DATA_RESERVE is placed at the beginning of the userspace task's
address environment
Summary:
- I noticed that maix-bit:smp does not work with QEMU.
- Actually, QEMU supports sifive_u (not K210) but it works
if FPU is disabled.
- This commit fixes this issue.
Impact:
- K210 with QEMU only
Testing:
- Tested with qemu-5.2
Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com>
Selecting this option will pass "-Map=$(TOPDIR)$(DELIM)nuttx.map" to ld
when linking NuttX ELF. That file can be useful for verifying
and debugging magic section games, and for seeing which
pieces of code get eliminated with DEBUG_OPT_UNUSED_SECTIONS.
Signed-off-by: chao.an <anchao@xiaomi.com>
Enable this option to optimization the unused input sections with the
linker by compiling with " -ffunction-sections -fdata-sections ", and
linking with " --gc-sections ".
Signed-off-by: chao.an <anchao@xiaomi.com>
