Detect and report the supported MMU for each CPU. Export the
capabilities to the rest of the kernel and use it in pmap_bootstrap() to
check for Sv48 support.
Reviewed by: markj
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D39814
Modify when and how we perform parsing and reporting. Most notably,
everything now executes on CPU 0.
The de-facto standard way to enumerate CPU features (ISA extensions) on
RISC-V is by parsing each CPU's ISA string. We currently obtain this
information from the device tree, and in the future will be able to pull
it from ACPI tables.
Eliminate the SYSINIT from identcpu.c. We still need to walk the /cpus
list in the device tree, but now do this one CPU at a time, as a step in
the identify_cpu() procedure. This is slightly less error prone, and
allows us to parse ISA features for CPU 0 much earlier.
Make use of the SMP hooks cpu_mp_start() and cpu_mp_announce() to
identify and print secondary CPU info, respectively. This causes
secondary processor identification to be printed much earlier in boot;
everything is done by SI_SUB_CPU, SI_ORDER_THIRD. Adjust some other
printf() calls so that we get enough useful info to debug under
bootverbose.
Reviewed by: markj (slightly earlier version)
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D39811
It is advantageous to have knowledge of ISA features as early as
possible. For example, the presence of newer virtual memory extensions
may be useful to pmap_bootstrap().
To achieve this, split out the printf() parts of identify_cpu() into a
separate function, printcpuinfo(). This latter function will be called
later in boot after the console has been initialized.
Reviewed by: markj
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D39810
Make better use of the RISC-V identification CSRs: mvendorid, marchid,
and mimpid. This code was written before these registers were
well-specified, or even available to the kernel. It currently fails to
recognize any CPU or platform.
Per the privileged specification, mvendorid contains the JEDEC vendor ID,
or zero.
The marchid register denotes the CPU microarchitecture. This is either
one of the globally allocated open-source implementation IDs, or the
field has a custom encoding. Therefore, for known vendors (SiFive) we
can also maintain a list of known marchid values. If we can not give a
name to the CPU but marchid is non-zero, then just print its value in
the report.
The mimpid (implementation ID) could be used in the future to more
uniquely identify the micro-architecture, but it really remains to be
seen how it gets used. For now we just print its value.
Thank you to Danjel Qyteza <danq1222@gmail.com> who submitted an early
version of this change to me, although it has been almost entirely
rewritten.
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D39809
When a DMA request using bounce pages completes, a swi is triggered to
schedule pending DMA requests using the just-freed bounce pages. For
a long time this bus_dma swi has been tied to a "virtual memory" swi
(swi_vm). However, all of the swi_vm implementations are the same and
consist of checking a flag (busdma_swi_pending) which is always true
and if set calling busdma_swi. I suspect this dates back to the
pre-SMPng days and that the intention was for swi_vm to serve as a
mux. However, in the current scheme there's no need for the mux.
Instead, remove swi_vm and vm_ih. Each bus_dma implementation that
uses bounce pages is responsible for creating its own swi (busdma_ih)
which it now schedules directly. This swi invokes busdma_swi directly
removing the need for busdma_swi_pending.
One consequence is that the swi now works on RISC-V which had previously
failed to invoke busdma_swi from swi_vm.
Reviewed by: imp, kib
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D33447
Replace a call to DELAY(1) with a new cpu_lock_delay() KPI. Currently
cpu_lock_delay() is defined to DELAY(1) on all platforms. However,
platforms with a DELAY() implementation that uses spin locks should
implement a custom cpu_lock_delay() doesn't use locks.
Reviewed by: kib
MFC after: 3 days
register to determine if trap is from userspace.
Otherwise if we jump to kernel address from userspace, then
TRAPF_USERMODE failed to detect usermode and then do_ast
triggers a panic "ast in kernel mode".
Reviewed by: markj@
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D16469
