|| ||Andrew Morton <firstname.lastname@example.org>|
|| ||lkml <email@example.com>|
|| ||Thu, 29 Aug 2002 01:36:46 -0700|
- Linus has merged per-zone-lru and per-zone-lock, so the NUMA guys should
be less than totally grumpy for several milliseconds.
- Linus has merged tlb-speedup. There are new stats in /proc/meminfo
which boast about how many global tlb invalidations have been saved.
These will disappear soon.
- The hashed rmap spinlocking patch has been shown to slow down the big
NUMA machines, and was of marginal benefit for SMP, and it added complexity.
It has been shuffled to apply after rmap-speedup and the highpte patch.
So it has basically been dropped. Daniel is cooking up something else.
+ A BUG_ON(page_count(page)) has been added to put_page_testzero()
(per Daniel's suggestion). Should catch the possible double-free
of LRU pages.
+ pagevec_lru_del() is removed
+ a change to the refill_inactive logic suggested by Ed Tomlinson
+ Janet Morgan's patch which speeds up readv&writev for O_DIRECT is
included. It needs some interface changes yet - generic_file_read
and generic_file_write should be changed to take an iovec, which will
allow lots of code to be removed.
+ Ed Tomlinson's slablru patch is now included. This puts slab pages
onto the LRU in an attempt to balance the various caches against
other demands on the page allocator. Seems to work nicely from a
quick test. It is early days for this code.
I added a field to /proc/meminfo which displays the total amount of
+ pte_highmem is still included, but it has mysteriously broken itself.
Don't enable this in config.
pte_highmem touches all architectures in possibly injurious ways. It
would be helpful if it received some non-x86 testing, please. Just the
"does it compile and boot" test.
2.5.32-mm2 hasn't had a ton of stability testing....
Incremental BK patch from Linus' tree
Fix block-highmem for scsi
Indexed directories for ext3
Check for zero-ref pages in put_page_testzero()
Clean up put_page() and page_cache_release().
Batched freeing and de-LRUing of anonymous pages
Writeback fixes and tuneups
Fix an ext3 deadlock
O_DIRECT support for ext3.
Convert page pointers into pfns for i386 NUMA
Rework setup_arch() for i386 NUMA
Restructure mem_init for i386 NUMA
discontigmem support for i386 NUMA
Clean up lots of open-coded uese of mem_map. For ia32 NUMA
Fix the boot-time reporting of each zone's available pages
Fix the __block_write_full_page() error path.
Back out the initial work for atomic copy_*_user()
spinlock/rwlock checking infrastructure
refill the inactive list more quickly
Use kmap_atomic() for generic_file_read()
Use kmap_atomic() for generic_file_write()
Fix throttling of heavy write()rs.
Make the global dirty memory accounting more accurate
Cleanup and fix the ramdisk driver (doesn't work right yet)
i386 discontigmem coding cleanups
i386 discontigmem cleanups
Downward adjustments to the default dirtymemory thresholds
Limit the consumption of ZONE_NORMAL by buffer_heads
rmap pte_chain space and CPU reductions
Resurrect CONFIG_HIGHPTE - ia32 pagetables in highmem
O_DIRECT support for readv/writev
age slab pages on the LRU
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to firstname.lastname@example.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/