BSD Commit Log Search

   s/kernel_GENERIC/kernel_GENERIC64, sync with arm64.

   Allow the user to decline roll forward by passing option "-o nolog" to
   mount_lfs.  Internally this is implemented by interpreting the MNT_LOG
   flag to mean "do roll forward".  Option "-o log" defaults on in mount_lfs(8).

   Don't panic if inodes that won't fit in the Ifile are encountered when
   rolling forward.  Instead, extend the Ifile to accommodate them.

   regen.

   Add various KX-7000/KX-6000G PCI devices.
   Shorten and improve some descriptions.
   Change some definitions for consistency.

   regen.

   Add VIA VX11 integrated graphics HDMI.
   Add various Zhaoxin SoC integrated graphics HDMI.

   Remove locking protocol error introduced in rev 1.32.

   aarch64: simplify rtypetbl initialisation in kobj_machdep.c

   libpcap 1.10.7 out

   whitespace. (blank line removal)

   aarch64: KNF elf_machdep.h

   aarch64: alignment consistency in elf_machdep.h

   aarch64: provide and use a AARCH64REG_READWRITE_INLINE

   aarch64: add ich_hcr_el2 to armreg.h with some bit definitions

   aarch64: Remove hypervisor.h and replace with armreg.h defines

   The previous value of the RES1 bits in CPTR_EL2 in non-VHE mode was even
   wrong.

   aarch64: fix relocation names to have R_AARCH64_ and not R_AARCH_

   compat defines are not included as very unlikely that they're needed

   aarch64: Teach kobj_reloc about R_AARCH64_LDST128_ABS_LO12_NC.

   PR port-arm/60660: missing R_AARCH_LDST128_ABS_LO12_NC

   From Taylor

   Protect seg_iocuont with lfs_lock instead of lfs_prelock.

   Decrement seg_iocount only while the segment lock is held.  May help
   with the deadlock reported in PR kern/60504.

   Instead of writing partial-segments synchronously when sync is requested,
   write them async, but wait for vp->v_numoutput to drop to zero if the
   calling semantics require it.  That way, even if we might have to wait,
   nobody else is required to wait with us.

   Be more discerning about when to write the inode of a file, and when
   to mark a file IN_MODIFIED.  Prevents what amounts to a vnode leak.

   x86/ALL: Enable nvmm(4).

   This way we can exercise the build of statically configured nvmm,
   when it is still normally used as as the module nvmm.kmod.

   doc/3RDPARTY: new postfix

   Postfix 3.11.7 out.

   +RDS  Reliable Datagram Sockets

   apply the same cascaded interrupt patch as on ofppc
   should fix PR60675

   a kernel config for the IBM 43P/150

   - use ofprint() for debug output so we have a chance to see it
   - do not BAT-map anything beyond 0xff000000 - my 43P gets quite grouchy if we do
   - copy EDID properties from OF like macppc does, for things like mach64 where
     X can program video modes but can't (relably) do DDC

   mips: throw a build error if the exception vector exceed 0x80 bytes.

   Use cpu_tlb_info instead of directly referencing ci_tlb_info. This way
   the PMAP_TLB_MAX == 1 optimisation can work.