- Tentatively scheduled as base version for RHEL4-U5.

- The current 2.6.18 x86_64 kernel has changed the IRQ-stack-to-
  process-stack linkage, where until now the link value was a pointer
  to the exception frame on the process stack, but has been changed
  to point to a location on the process stack above the exception
  frame.  Because of that, after displaying the trace data from the
  IRQ stack, "bt" would then display an invalid exception frame,
  which was reported as a "possibly bogus exception frame".
  (anderson@redhat.com)

- Also in x86_64 kernels, fix for the "bt" command.  When the backtrace
  started on the NMI exception stack, it was displaying the correct
  exception frame data, but was erroneously reporting that it was a
  "possibly bogus exception frame".  (anderson@redhat.com)

- And again in x86_64 kernels, fix for the "bt" command.  When making
  the transition from the IRQ stack back to the process stack, when
  the IRQ stack entry was made via the relatively new "call_softirq"
  entry point.  In that case, there is no exception frame on the
  process stack, because it's essentially just a cross-stack call
  from do_softirq().  However, a bogus exception frame was being
  displayed, along with a "possibly bogus exception frame" message;
  and if the RIP value in the truly bogus exception frame happened
  to fall in the user virtual address range, the remainder of the
  process stack trace was not displayed at all.  (anderson@redhat.com)

- Fix for 2.6.18-era ia64 DISCONTIGMEM kernels, which would fail
  during initialization with the error message: "crash: invalid
  (optional) structure member offsets: pglist_data_node_next or
  pglist_data_pgdat_next".  (anderson@redhat.com)

- Adapted Olivier Daudel's nifty enhancement to the "struct" command,
  which allows the single "struct.member" argument to optionally be
  expressed in a "struct.member[,member,member] format, in order to
  display multiple members of a given structure.  This also applies to
  the "union" and "*" commands, as all three functions have now been
  combined into one behind the scenes.  Fixed the display for applying
  a minus count, and given that it opened up a the door to a number of
  entry errors, I also added additional error-catching/handling to avoid
  the display of incorrect structure data.
  (olivier.daudel@u-paris10.fr, anderson@redhat.com)

- Fixed three sources of potential segmentation violations when using
  the "bt" command when the experimental dwarf CFI unwind backtrace
  facility was turned on.  (anderson@redhat.com)

- Added a new machdep_init(POST_VM) call, which is currently only being
  used by the x86_64 architecture; it calls init_unwind_table(), which
  has to be done after vm_init() in order to access the unwind tables
  of kernel modules.  (anderson@redhat.com)

Download from: http://people.redhat.com/anderson