improvements in 7.1

Marius Strobl marius at alchemy.franken.de
Tue Jan 13 11:51:32 PST 2009


FYI, in comparison to FreeBSD 7.0, 7.1 contains the following
improvements interesting for sparc64:
o ata(4) was tought about the mapping used by (Open)Solaris
  required to fit the geometries of larger, i.e. > 32GB,
  disks into the the 16-bit fields of the Sun and VTOC8 disk
  labels. This means it's now possible to use the full
  capacity of such disks without the need to label them with
  (Open)Solaris.
o dc(4) grew a workaround for a minor hardware bug of certain
  chips that it supports, including those used on-board in
  V100 and X1, which prevented the link state to be reported
  correctly, in turn causing failover of lagg(4) to not work.
  Solving this revealed bugs which could lead to panics in
  dc(4) itself though. If you're using dc(4) and updated
  some time between 7.0 and 7.1 you'll definitely should
  switch to 7.1.
o esp(4) received an overhaul adapting it to MPSAFE cam(4)
  and make it more robust regarding target failures that
  previously caused panics, amongst other improvements.
  Unfortunately, the esp(4) in 7.0 was incorrectly marked
  as being MPSAFE with unclear outcome. If you use esp(4)
  on a MP machine you thus should update to 7.1
o kbdmux(4) was enabled by default in the sparc64 GENERIC.
  Since sunkbd(4) emulates PS/2 keyboards by default for
  some time just like ukbd(4) does it's possible to mix
  RS232 and USB keyboards now (this actually was requested
  by someone some time ago...).
o gem(4) received some minor fixes. Support for the SBus-
  version of GEM will hit CURRENT soon and should be
  available in 7-STABLE some time before 7.2.
o hme(4) got some fixes and improvements, though nothing
  particular standing out.
o sf(4) was enabled by default in the sparc64 GENERIC
  after it received a major overhaul making it work on all
  architectures supported by FreeBSD.
o sym(4) was adapted to MPSAFE cam(4) and now correctly
  negotiates Ultra-3 with 53C1010 without NVRAM as
  found on sparc64.
o vr(4) was enabled by default in the sparc64 GENERIC
  after it received a major overhaul making it work on all
  architectures supported by FreeBSD.
o The sparc64 interrupt code was better adapted to the
  machine independent one, getting rid of unnecessary
  disabling and re-enabling of the interrupt source with
  ever acknowledge and allowing to load-balance interrupts
  across the CPUs in MP system as well as binding single
  interrupts to specific CPUs.
o The sparc64 OFW PCI code was switched over to use the
  PCI domain support, which is mandatory for bringing in
  support for USIII and beyond sun4u. This should be
  transparent for the average user, just pciconf(8)
  selectors might have changed on some machines as a
  result.
o A workaround in the sparc64 stack_save(9) for an old GCC
  bug was removed, which fixes backtracing of curthread as
  done by witness(4) for displaying information when using
  GCC 4.2 or later.
o A race concerning more or less (well, at least not limited
  to sparc64) machine dependent code of kmem(4), which
  allowed a panic to by triggered using for example fstat(1),
  was fixed.

Chances are also that we once again shipped with working
Firefox 2 and friends, though unfortunately version 3
was reported to again contain a bunch of LP64 bugs.

Marius



More information about the freebsd-sparc64 mailing list