Changes in kern

Changes in kern http://nxr.netbsd.org/rss/src/sys/kern/ en Copyright 2005 Java kernel: Track what lock each lwp is waiting for without LOCKDEBUG. This is reasonably cheap -- one extra pointer on the stack, couple extra stores when _waiting_ (spinning or sleeping) for a _contended_ lock -- and doesn't change the kernel ABI. Will enable us to get diagnostics from crash dumps when, e.g., there's a softint deadlock tripping a heartbeat panic. PR kern/60030: ddb/crash: show all locks without LOCKDEBUG Sun Aug 16 21:00:00 UTC 2026 riastradh proc_vmspace_getref: Take p->p_lock around p->p_sflag access. Prompted by reviewing the locking rules in; PR kern/60568: panic locking against myself (p->p_lock) in NFS from sysctl_vmproc Fri Aug 14 03:00:00 UTC 2026 riastradh Rename the "basic 4BSD" disklabel routines readdisklabel_4bsd() and writedisklabel_4bsd(), but add weak aliases to the standard names. This easily allows machine-specific code to look for an odd-ball machine-firmware-specific format that might be required only for booting but otherwise use garden-variety BSD disklabels. Sun Aug 09 16:00:00 UTC 2026 thorpej proc_regio: Clarify control flow and sprinkle comments. Use `if (error) goto out' instead of `if (error == 0) error = ...' to make the error case non-linear and indented, and the success case a straight line of logic. No functional change intended. PR port-amd64/60556: panic in process_read_fpregs_xmm Sat Aug 08 00:00:00 UTC 2026 riastradh vflushbuf: Print `vflushbuf: dirty' warning only once per call. And do it only under DEBUG, and rate-limit it globally. This can happen when vflushbuf is competing with an onslaught of concurrent I/O on a snapshot or block device (which is done without holding the vnode lock, despite what the comment says). That concurrent I/O might be happening indefinitely, so we are likely to have to print a _lot_ of warnings, potentially overwhelming the console, until the underlying problem is fixed by teaching vflushbuf to wait only for the writes that began before have completed, not also all writes that have begun since. This stop-gap measure intended to be low-risk for pullup to release branches. PR kern/60531: looping "vflushbuf: dirty" message Thu Aug 06 22:00:00 UTC 2026 riastradh <sys/disklabel.h> not needed here. Sat Aug 01 20:00:00 UTC 2026 thorpej Trailing whitespace Sat Aug 01 11:00:00 UTC 2026 skrll Move consulation of RAW_PART from an #if to an if (). Tue Jul 21 14:00:00 UTC 2026 thorpej kobj_load(): Consider ELF{32,64}_MACHDEP_FLAGS_OK(). Fri Jul 17 14:00:00 UTC 2026 thorpej Slim this down for non-MP configurations. Fri Jul 17 02:00:00 UTC 2026 thorpej devhandle_impl_subclass(): make sure a handle never has the type INVALID. Fri Jul 17 02:00:00 UTC 2026 thorpej Don't bother signalling remote CPUs on non-MP configurations. Fri Jul 17 02:00:00 UTC 2026 thorpej Elide subr_ipi.c for non-MP configurations. Fri Jul 17 02:00:00 UTC 2026 thorpej kqueue(2): Tighten branch in previous change. If sleepok is true, then kmem_zalloc should never fail, and this routine is inline with constant arguments at every call site, so let's let the compiler prune dead branches for the call sites that can sleep. PR kern/60436: kqueue(2): random kernel null pointer deref in EVFILT_PROC NOTE_TRACK Fri Jul 10 14:00:00 UTC 2026 riastradh Prevent a DoS attack by crafted binaries that trigger KASSERTS. Change the KASSERTS into errors instead. Reported by Andrew Griffiths. Thu Jun 18 20:00:00 UTC 2026 christos m_print(): Support TH_ACK flag Wed Jun 17 09:00:00 UTC 2026 msaitoh Rename the needlessly over-descriptive and at the same time somewhat non-descriptive "struct powerhook_desc" to just "struct powerhook". Tue May 26 15:00:00 UTC 2026 simonb Add "rootspec" hooks. These hooks add a generic mechanism for devices to advertise extra root device choices. Use these to replace wedge-specific code in sys/kern/kern_subr.c . XXX todo: rootspec configuraton can be unloaded after the root filesystem is mounted, and rootspechook_establish() calls after the root filesystem is mounted could then be a no-op. Thanks riastradh@, mlelstv@ and chs@ for reviews and comments. Tue May 26 14:00:00 UTC 2026 simonb PGOFSET -> PAGE_MASK Mon May 11 02:00:00 UTC 2026 thorpej Fix a longstanding bug on 64 bit ports with ABIs requiring sign-extension when returning 32-bit values in registers. Alpha, MIPS N64, RISCV. With these ABIs it is necessary for the syscall return code to manually sign-extend all 32-bit return values. If we don't, everything may look just fine, until a syscall return value is spilled to the stack and then reloaded. At that point, the zero-extended form produced by C assignment into the 64-bit register for return to userspace will be sign-extended upon reload, producing a different, corrupt value which will fail comparisons. This was observed with a getuid() != geteuid() deep inside sshd on N64, on an unusual system that had some very large UIDs with bit 31 set - but is quite difficult to reproduce with small tests, since register pressure and compiler behavior determine whether the conditions necessary to trigger the bug will exist. The easiest way to see whether a new target needs this treatment is to look at its GCC configuratrion; if its PROMOTE_MODE has UNSIGNEDP = 0 for SImode, the sign-extension in syscall() is necessary. Sun May 10 23:00:00 UTC 2026 tls