Messages in this thread | | | Date | Sun, 20 Jul 2003 00:02:47 +0200 (CEST) | From | Ingo Molnar <> | Subject | [patch] 4G/4G split patch, 2.6.0-test1-G7 |
| |
the latest 4G/4G split patch can be found at:
http://redhat.com/~mingo/4g-patches/4g-2.6.0-test1-mm1-G7
besides being a merge to 2.6.0-test-mm1, this version also includes many cleanups, bugfixes and speedups. All quirks are fixed and sysenter based syscalls work now too.
Changes:
- merge to the 2.6.0-test1-mm1 kernel
- the 4G/4G trampoline is now linked to the high address - this is faster and also enabled the removal of the ugly runtime linking code.
- removed all 4G/4G hardcoded limits, they are not generated by asm-offsets.c, compile-time.
- moved the TI_ hardcoded constants to asm-offsets.c as well.
- add remap_numa_kva() again, to fix NUMA booting.
- fixed hardware-debug-register handling.
- entry.S entry/exit assembly speedups. Got rid of %cr3 accesses. Reload %esp only if necessary.
- restore lcall7/lcall27 binary-compatibility gates
- increased the TSS mappings - it should now work up to 32 CPUs.
- increased the entry trampoline to 2 pages
- made the FPU code not use __get_user loops for FPU conversion, but a conversion loop into a kernel-internal scratch area, and a user-copy afterwards.
- made the signal code use less user-copy looping in the signal-frame construction code. This change and the FPU change brought down signal latency from 120 usecs to 28 usecs.
- got rid of __PAGE_OFFSET_USER, it's now done via TASK_SIZE.
- fixed SMP deadlock in the futex code.
- fixed the reboot code's 4G/4G-incompatible assumptions.
- remove assumptions about the # of pmds the trampoline window uses.
Thanks go to everyone who provided feedback: Kirill Korotaev, William Lee Irwin, Martin J. Bligh, Dave Hansen and others.
(there are no known bugs or design problems at the moment, so please resend any item that got lost in this cycle.)
Ingo
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |