Reorganized the patches, fixed some bugs, and so on. Available from:
ftp://ftp.kernel.org/pub/linux/kernel/people/wli/kernels/2.5.51-bk1-wli-1/
Notable events:
(1) bugfix for unconditionally setting nr_ioapics = 2
reported by Zwane Mwaikambo
(2) bugfix for misaligned order > 0 frees in NUMA-Q highpage init.
(3) dropped manfred's slab changes; they're no longer needed
(4) no preallocation of pmd's in the pgd ctor; they exhibited
extremely poor fragmentation characteristics when
that was done. Just slab allocate pmd's in pgd_alloc().
I never really commented on the theme of the tree. It's basically a
bunch of things that are obvious to me how to do and a place to dump
patches that don't really fit into other trees (e.g. -mm), though I
think some of the NUMA-Q stuff might eventually get thrown over the
wall in the direction of -mjb. =) Sound bite: misc patches from wli.
The net effect is that some odd bootstrapping, wait table hashing,
memory initialization, and tasklist scan removal patches landed here.
Other stuff will crop up as I poke around the tree.
01_driverfs_oops
fix for still-broken driverfs memblk and node registration.
I suspect this might hit mainline soon.
02_numaq_io
Workaround for PCI bridges on different PCI segments
with clashing bus numbers until it's fixed (by me). Drop
IO-APIC's and PCI buses off node 0 onto the floor. Bugfix
to set nr_ioapics only if (clustered_apic_mode) and take
min(2, nr_ioapics) to avoid elevating nr_ioapics. Some
NUMA-Q PCI code may appear soon to replace this...
03_do_sak
Remove tasklist iteration from __do_SAK(); it's simply trying
to kill off tasks in a given session; use for_each_task_pid().
04_proc_super
proc_fill_super() is iterating over the tasklist because there
is no count of processes (only threads). Use a process counter.
05_cap_set_pg
cap_set_pg() is iterating over the tasklist in search of tasks
within a process group; use for_each_task_pid().
06_vm86
vm86 wants to clean up references to tasks. Do so by cleaning
up stale references in release_thread(), not by GC that walks
the tasklist comparing (possibly reused) task_struct pointers.
This was originally part of a bugfix in -dj only half of which
made it to mainline.
07_uml_get_task
UML's get_task() is looking for a task with a given pid; use
find_task_by_pid() instead of walking the tasklist.
08_numaq_mem_map
Free higher-order pages in NUMA-Q highmem mem_map initialization
instead of freeing order 0 pages at a time. New bugfix here that
fixes some improperly-aligned frees. There's an irritated nested
min() duplicate const warning.
09_numaq_pgdat
Allocate pgdats from node-local memory on NUMA-Q
10_has_stopped_jobs
Remove the unused has_stopped_jobs() from kernel/exit.c
and rename __has_stopped_jobs() to has_stopped_jobs().
11_inode_wait
Increase the too-small inode wait table's size.
12_pgd_ctor
Use slab ctor's for pgd's and pmd's on PAE i386 boxen.
One of the primary benefits is that pmd's are actually
accounted (via /proc/slabinfo) so their lowmem consumption
is directly visible.
-
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/
This archive was generated by hypermail 2b29 : Sun Dec 15 2002 - 22:00:25 EST