On Fri, Dec 03, 2004 at 10:51:29PM -0600, Marc Schwartz wrote:
> 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.
Is the fix for the b44 issue
incorporated into the kernels yet?
yes. I actually trimmed too much of the changelog.
There's another few days of changes that went in.
Also, does the change in the 4g/4g status influence the b44 fix?
In that the bug only affects -hugemem. So it's doubly fixed 8)
* Mon Nov 22 2004 Dave Jones <davej(a)redhat.com>
- Update -ac patch to 2.6.9-ac11
- make tulip_stop_rxtx() wait for DMA to fully stop. (#138240)
- ACPI: Make LEqual less strict about operand types matching.
- scsi: avoid extra 'put' on devices in __scsi_iterate_device() (#138135)
- Fix bugs with SOCK_SEQPACKET AF_UNIX sockets
- Reenable token ring drivers. (#119345)
- SELinux: Map Unix seqpacket sockets to appropriate security class
- SELinux: destroy avtab node cache in policy load error path.
- AF_UNIX: Serialize dgram read using semaphore just like stream.
- lockd: NLM blocks locks don't sleep
- NFS lock recovery fixes
- Add more MODULE_VERSION tags (#136403)
- Update qlogic driver to 2.6.10rc2 level.
- cciss: fixes for clustering
- ieee802.11 update.
- ipw2100: update to ver 1.0.0
- ipw2200: update to ver 1.0.0
- Enable promisc mode on ipw2100
- 3c59x: reload EEPROM values at rmmod for needy cards
- ppc64: Prevent sigsuspend stomping on r4 and r5
- ppc64: Alternative single-step fix.
- fix for recursive netdump oops on x86_64
- ia64: Fix IRQ routing fix when booted with maxcpus= (#138236)
- ia64: search the iommu for the correct size
- Deal with fraglists correctly on ipv4/ipv6 output
- Various statm accounting fixes (#139447)
- Reenable CMM /proc interface for s390 (#137397)
* Fri Nov 19 2004 Dave Jones <davej(a)redhat.com>
- e100: fix improper enabling of interrupts. (#139706)
- autofs4: allow map update recognition
- Various TCP fixes from 2.6.10rc
- Various netlink fixes from 2.6.10rc
- [IPV4]: Do not try to unhash null-netdev nexthops.
- ppc64: Make NUMA map CPU->node before bringing up the CPU (#128063)
- ppc64: sched domains / cpu hotplug cleanup. (#128063)
- ppc64: Add a CPU_DOWN_PREPARE hotplug CPU notifier (#128063)
- ppc64: Register a cpu hotplug notifier to reinitialize the
scheduler domains hierarchy (#128063)
- ppc64: Introduce CPU_DOWN_FAILED notifier (#128063)
- ppc64: Make arch_destroy_sched_domains() conditional (#128063)
- ppc64: Use CPU_DOWN_FAILED notifier in the sched-domains hotplug code (#128063)
- Various updates to the SCSI midlayer from 2.6.10rc.
- vlan_dev: return 0 on vlan_dev_change_mtu success. (#139760)
- Update Emulex lpfc driver to v8013
- Fix problem with b44 driver and 4g/4g patch. (#118165)
- Prevent oops when loading aic79xx on machine without hardware. (#125982)
- Use correct spinlock functions in token ring net code. (#135462)
- scsi: Add reset ioctl capability to ULDs
- scsi: update ips driver to 7.10.18
- Reenable ACPI hotplug driver. (#139976, #140130, #132691)