arcbios: enable FPU around ARCS calls
This seems required for the SGI O2 for PROM graphics console IO to work
when it scrolls. See PR port-sgimips/60204 for more details.
The issue was introduced in v1.214 of sys/arch/mips/mips/locore.S .
Notably maya@ disabled the FPU early in boot, expecting the rest of
boot to run with the FPU disabled. The FPU is being explicitly
enabled/disabled here to keep to the spirit of maya@'s above commit.
A "better" solution would be to use the pcu API and mark the thread
as needing the FPU for the duration of the arcbios call, however:
* The current MIPS pcu FPU code in sys/arch/mips/mips/mips_fpu.c
doesn't support FPU use in kernel yet; and
* this stuff is called super early during boot and once the OS starts
the ARCBIOS API is no longer supposed to be used.
call this bozohttpd 20260503, and update the CHANGES for the last 2 years
o fix the default minimum TLS version to 1.1 from 1.3. the
manual already said 1.1 was the default. fixes PR#58878.
o log the correct port with TLS connections. fixes PR#59644.
o fix use-after-free, double-free, and bounds checking problems.
from shm.
o better lint support.
o several updates for the manual. from lukem.
add D Bohdan to the contributors list.
Hide platform vmparam differences away from user-space and modules.
For modules that need to know USRSTACK (exec / compat modules), provide
a way to get that via a variable initialized in cpu_startup().
Undo some of the previous, restore flexibility
This is related to PR bin/59957 (in that it is a continuation/
alteration of the previous fix). This was primarily designed
to (hopefully) fix the ~80 extra ATF test failures that the previous
solution caused, by allowing the utilities to work as they had
previously, rather than attempting to enforce one universal true
world order.
Change the openspecial() function to be findspecial() as it no longer
opens anything - but leave it in openspecial.c for several reasons:
First, it might make sense to recreate openspecial() for use in just
those utilities that want to do all that it did (just 2 of the four
that were modified to use it).
Also, this function (or functions) really should be moved to libutil,
rather than buried in sbin/fsck (which doesn't even use it/them at all
- though could perhaps use the findspecial() variant), and I didn't see
[27 lines not shown]
Despite pooka's dislike of the situation, it is sometimes unavoidable
that systems within a given $MACHINE may have different vmparams.
Furthermore, rump is fundamentally a user-space entity on NetBSD, and
it's absolutely true that different $MACHINEs within a $MACHINE_ARCH
may legitimately have different vmparams related to physical memory
layout, and so using the real <machine/vmparams.h> is an impediment
to a $MACHINE_ARCH-generic user-space build.
As such, rump once again has its own vmparams.h file with values that
should be perfectly adequate for the rump virtual environment and its
non-existent subordinate user-space.
PR 58762: disable MKCOMPAT for earm*.
If someone is interested in re-adding support for oabi compat library
builds, they can figure out the missing bits. But for now, stop producing
bogus compat32/debug32 sets on all evbarm builds.
import zlib-1.3.2 (previous was 1.3.1)
Version 1.3.2 has these key updates from 1.3.1:
Address findings of the 7ASecurity audit of zlib.
Check for negative lengths in crc32_combine functions.
Copy only the initialized window contents in inflateCopy.
Prevent the use of insecure functions without an explicit request.
Add compressBound_z and deflateBound_z functions for large values.
Use atomics to build inflate fixed tables once.
Add --undefined option to ./configure for UBSan checker.
Copy only the initialized deflate state in deflateCopy.
Zero inflate state on allocation.
Add compress_z and uncompress_z functions.
Complete rewrite of cmake support.
Remove untgz from contrib.
Vectorize the CRC-32 calculation on the s390x.
Remove vstudio projects in lieu of cmake-generated projects.
[6 lines not shown]
Introduce a new function, uvm_lwp_is_pagedaemon(), which returns true
if the specified LWP is the pagedaemon LWP. This test is inline for
the kernel proper, and directed to a function for modules. This, in turn,
allows us to hide "struct uvm_cpu" and "struct uvm" from modules, as
these have arrays in them that are not a consistent size certainly across
$MACHINEs within a given $MACHINE_ARCH, and may not necessarily even be
consistent within a given $MACHINE.
NetBSD -> 11.99.6 because new symbol referenced by modules.
Import acpica-20260408 (previous was 20251212)
8 April 2026. Summary of changes for version 20260804:
Major changes:
* ACPI Table Enhancements:
Updated IORT and MADT template compilations to add new IWB and
GICv5 translator/router/ITS entries. Introduced full support for
the new DTPR table, including compiler and disassembler handling,
template fixes, and register structure refinements.
- Jose Marinho and Michal Camacho Romero
* Expanded IDs, GUIDs, and Methods:
Added Framework Computer Hardware IDs, modern standby _DSM GUIDs,
and TPM2 start methods. Updated _DSM Arg3 formatting and _CPC
handling to align with newer ACPI specifications and avoid Linux
compatibility warnings.
- Daniel Schaefer and Saket Dumbre
[28 lines not shown]
