Reminder: it breaks top(1) and vmstat(1). Updates to these tools
are at http://surriel.com/procps/
Fix block-highmem for IDE. From Jens.
Fix some boot-time hang which I had. From Jens.
Debug patch to catch people calling things like kmalloc(__GFP_WAIT),
down(), etc from inside spinlocks. This is a much stronger test than
the one in schedule(). It's really only effective with CONFIG_PREEMPT=y.
Causes a storm of complaints and stack backtraces at boot due to
drivers/ide/ide-probe.c:init_irq() calling lots of things which it
shouldn't under ide_lock. It will find other bugs.
We don't need to hold mapping->private_lock while moving the page
onto its mapping's dirty_pages list.
Feature for software suspend to force page reclaim. Doesn't work right.
Fix the lockup which happens each time the aic7xxx driver encounters
an IO error.
Remove page->virtual, hash for it.
Kill off the secret benchmarking feature which allowed the dirty memory
threshold to be exceeded when running dbench. Makes the kernel really
strict about limiting the amount of dirty memory and should make the
buffer_head stripping code more effective.
Fix a task lockup which can happen in mempool_alloc() when the number
of sleeping tasks exceeds the size of the mempool.
Remove the write_mapping_buffers() function. It didn't work out.
Use the buffer_boundary() block-mapping hint for scheduling of IO
against indirect blocks. This _does_ work out. Large-file writeback
is now smooth and seekless.
Clean up the ll_rw_block() function.
Remove the lock_kernel() in ext2_readdir(). I'm fairly sure that it's
superfluous, and it was being irritating.
Make sure that contig_page_data is undefined on discontigmem builds.
Make the normal zone per-node rather than global on the NUMAQ's.
Clean up alloc_pages_node()
The dcache RCU code.
I do not intend to send this on to Linus. It is not my area and I
do not understand the code.
But a 15% hit in specweb99 on an 8-way is sigificant and no other
solution has presented itself. So this way, this code gets some extra
testing. Plus clearing away this problem helps us to identify other
I shall keep these patches ticking over as a general service to the
people who are interested in, and working on 2.5 performance.
No real progress has been made on the writer-starves-reader problem
which is impacting the `contest' io_fullload test. This is a severe
problem - a streaming write is decreasing read performance from the
same disk by a factor of 4000. Working on it.
Fix block-highmem for scsi
Indexed directories for ext3
spinlock/rwlock checking infrastructure
Cleanup and fix the ramdisk driver (doesn't work right yet)
debug code to detect might-sleep-inside-spinlock bugs
Infrastructure for communicating request queue congestion to the VM
avoid ext2 inode prereads if the queue is congested
non-blocking writeback infrastructure, use it for pdflush
Non-blocking page reclaim
don't call __mark_inode_dirty under spinlock
prepare_to_wait/finish_wait: new sleep/wakeup API
Use the faster wakeups in the VM and block layers
Speed up sys_sync() against multiple spindles
Early and smarter shrinking of slabs
Fix the generic_file_write-from-same-mmapped-page deadlock
Add /proc/buddyinfo - stats on the free pages pool
Remove struct free_area_struct and free_area_t, use `struct free_area'
Per-node kswapd instance
Simple topology API
radix tree gang lookup
Move the vm accounting out of /proc/stat
Add kswapd_steal to /proc/vmstat
I/O wait statistics
SARD disk accounting
Use fast wakeups in TCP/IPV4
Fix a tmpfs swapoff deadlock
page state cleanup
shmem_rename() directory link count fix
tmpfs: show a non-zero size for directories
tmpfs: small fixlets
separate the kswapd and direct reclaim code paths
add shrink_all_memory() for swsusp
partial fix for aic7xxx error recovery
remove page->virtual for !WANT_PAGE_VIRTUAL
sterner dirty-memory clamping
Fix for stuck tasks in mempool_alloc()
IO schduling for indirect blocks
remove lock_kernel() from ext2_readdir()
undefine contif_page_data for discontigmem
ia32 NUMA: per-node ZONE_NORMAL
extended barrier primitives
Use RCU for dcache
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to email@example.com
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 : Mon Sep 23 2002 - 22:00:34 EST