New testing kernel.

[Dave Jones]

I just made a 2.6.9-1.698_FC3 set of kernels which fix
a number of bugs (Full changelog below), including some
of the more popular ones that were introduced during the
last update (namely, smbfs should work again, and hopefully
 the palm/visor oopses should be gone too).

There's still a number of bugs that affect a lot of users
(like the acpi_power_off issue which I still don't
 have a handle on, despite being able to reproduce it
 and spending countless hours building kernels with
 extra debugging).  Hopefully I'll get some of the
other nastier issues knocked off soon.

So, in the meantime, try out -698, and file bugs if
something new broke. (And please close any old ones
that went away with this kernel).  The last 2.6.9
update closed quite a few long-standing bugs, however
a *lot* of bugs are still in NEEDINFO awaiting
confirmation with the new kernel(s).

The biggest change in this kernel over the older ones
that affects all x86 users, is that the 4g/4g
memory split is no longer the default.  If you
have a lot of memory (16GB or more), or have a workload
that benefits from being able to have more address
space, you can use the -hugemem kernel to run with
4g/4g again.  The regular non-hugemem kernels have
the traditional 3:1 split.
Due to this quite large change, this kernel
could really use folks jumping up and down on it
for a while before it gets pushed into updates-proper.


Thanks,

		Dave


* Fri Dec  3 2004 Dave Jones <davej(a)redhat.com&gt;
- Pull in bits of -ac12
  Should fix the smbfs & visor issues among others.

* Thu Dec  2 2004 Dave Jones <davej(a)redhat.com&gt;
- Drop the futex debug patch, it served its purpose.
- XFRM layer bug fixes
- ppc64: Convert to using ibm,read-slot-reset-state2 RTAS call
- ide: Make CSB6 driver support configurations.
- ide: Handle early EOF on CDs.
- Fix sx8 device naming in sysfs
- e100/e1000: return -EINVAL when setting rx-mini or rx-jumbo. (#140793)

* Wed Dec  1 2004 Dave Jones <davej(a)redhat.com&gt;
- Disable 4G/4G for i686.
- Workaround for the E1000 erratum 23 (#140047)
- Remove bogus futex warning. (#138179)
- x86_64: Fix lost edge triggered irqs on UP kernel.
- x86_64: Reenable DRI for MGA.
- Workaround E1000 post-maturely writing back to TX descriptors (#133261)
- 3c59x: add EEPROM_RESET for 3c900 Boomerang
- Fix buffer overrun in arch/x86_64/sys_ia32.c:sys32_ni_syscall()
- ext3: improves ext3's error logging when we encounter an on-disk corruption.
- ext3: improves ext3's ability to deal with corruption on-disk
- ext3: Handle double-delete of indirect blocks.
- Disable SCB2 flash driver for RHEL4. (#141142)

* Tue Nov 30 2004 Dave Jones <davej(a)redhat.com&gt;
- x86_64: add an option to configure oops stack dump
- x86[64]: display phys_proc_id only when it is initialized
- x86_64: no TIOCSBRK/TIOCCBRK in ia32 emulation
- via-rhine: references __init code during resume
- Add barriers to generic timer code to prevent race. (#128242)
- ppc64: Add PURR and version data to /proc/ppc64/lparcfg
- Prevent xtime value becoming incorrect.
- scsi: return full SCSI status byte in SG_IO
- Fix show_trace() in irq context with CONFIG_4KSTACKS
- Adjust alignment of pagevec structure.
- md: make sure md always uses rdev_dec_pending properly.
- Make proc_pid_status not dereference dead task structs.
- sg: Fix oops of sg_cmd_done and sg_release race (#140648)
- fix bad segment coalescing in blk_recalc_rq_segments()
- fix missing security_*() check in net/compat.c
- ia64/x86_64/s390 overlapping vma fix
- Update Emulex lpfc to 8.0.15

* Mon Nov 29 2004 Dave Jones <davej(a)redhat.com&gt;
- Add another card reader to whitelist. (#141022)
- Fix possible hang in do_wait() (#140042)
- Fix ps showing wrong ppid. (#132030)
- Print advice to use -hugemem if >=16GB of memory is detected.
- Enable ICOM serial driver. (#136150)
- Enable acpi hotplug driver for IA64.
- SCSI: fix USB forced remove oops.
- ia64: add missing sn2 timer mask in time_interpolator code. (#140580)
- ia64: Fix hang reading /proc/pal/cpu0/tr_info (#139571)
- ia64: bump number of UARTS. (#139100)
- Fix ACPI debug level (#141292)
- Make EDD runtime configurable, and reenable.
- ppc64: IBM VSCSI driver race fix. (#138725)
- ppc64: Ensure PPC64 interrupts don't end up hard-disabled. (#139020, #131590)
- ppc64: Yet more sigsuspend/singlestep fixing. (#140102, #137931)
- x86-64: Implement ACPI based reset mechanism. (#139104)
- Backport 2.6.10rc sysfs changes needed for IBM hotplug driver. (#140372)
- Update Emulex lpfc driver to v8.0.14
- Optimize away the unconditional write to debug registers on signal delivery path.
- Fix up scsi_test_unit_ready() to work correctly with CD-ROMs.
- md: fix two little bugs in raid10
- Remove incorrect ELF check from module loading. (#140954)
- Plug leaks in error paths of aic driver.
- Add refcounting to scsi command allocation.
- Taint oopses on machine checks, bad_page()'s calls and forced rmmod's.
- Share Intel cache descriptors between x86 & x86-64.
- rx checksum support for gige nForce ethernet
- vm: vm_dirty_ratio initialisation fix

* Mon Nov 29 2004 Soeren Sandmann <sandmann(a)redhat.com&gt;
- Build FC-3 kernel in RHEL build root

* Sun Nov 28 2004 Dave Jones <davej(a)redhat.com&gt;
- Move 4g/4g kernel into -hugemem.

* Sat Nov 27 2004 Dave Jones <davej(a)redhat.com&gt;
- Recognise Shuttle SN85G4 card reader. (#139163)

* Tue Nov 23 2004 Dave Jones <davej(a)redhat.com&gt;
- Add futex debug patch.