Why NOOP? Because CFQ is wasting CPU cycles for no performance gain. There should
be more left for the system now, and identical block device performance.
Devtmpfs in config enabled for Lucid/Maverick and other 2010 Linux support.
Added modules for CPU governors for the Smarttop board since power management is
useful.
* Features:
- Low metadata overhead (just 4 bytes per object)
- O(1) Alloc/Free - except when we have to call system page allocator to
get additional memory.
- Very low fragmentation: In all tests, xvmalloc memory usage is within 12%
of "Ideal".
- Pool based allocator: Each pool can grow and shrink.
- It maps pages only when required. So, it does not hog vmalloc area which
is very small on 32-bit systems.
SLUB allocator could not be used due to fragmentation issues:
http://code.google.com/p/compcache/wiki/AllocatorsComparison
Data here shows kmalloc using ~43% more memory than TLSF and xvMalloc
is showed ~2% more space efficiency than TLSF (due to smaller metadata).
Creating various kmem_caches can reduce space efficiency gap but still
problem of being limited to low memory exists. Also, it depends on
allocating higher order pages to reduce fragmentation - this is not
acceptable for ramzswap as it is used under memory crunch (its a swap
device!).
SLOB allocator could not be used do to reasons mentioned here:
http://lkml.org/lkml/2009/3/18/210
* Implementation:
It uses two-level bitmap search to find free list containing block of
correct size. This idea is taken from TLSF (Two-Level Segregate Fit)
allocator and is well explained in its paper (see [Links] below).
* Limitations:
- Poor scalability: No per-cpu data structures (work in progress).
[Links]
1. Details and Performance data:
http://code.google.com/p/compcache/wiki/xvMallochttp://code.google.com/p/compcache/wiki/xvMallocPerformance
2. TLSF memory allocator:
home: http://rtportal.upv.es/rtmalloc/
paper: http://rtportal.upv.es/rtmalloc/files/MRBC_2008.pdf
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
* Features:
- Low metadata overhead (just 4 bytes per object)
- O(1) Alloc/Free - except when we have to call system page allocator to
get additional memory.
- Very low fragmentation: In all tests, xvmalloc memory usage is within 12%
of "Ideal".
- Pool based allocator: Each pool can grow and shrink.
- It maps pages only when required. So, it does not hog vmalloc area which
is very small on 32-bit systems.
SLUB allocator could not be used due to fragmentation issues:
http://code.google.com/p/compcache/wiki/AllocatorsComparison
Data here shows kmalloc using ~43% more memory than TLSF and xvMalloc
is showed ~2% more space efficiency than TLSF (due to smaller metadata).
Creating various kmem_caches can reduce space efficiency gap but still
problem of being limited to low memory exists. Also, it depends on
allocating higher order pages to reduce fragmentation - this is not
acceptable for ramzswap as it is used under memory crunch (its a swap
device!).
SLOB allocator could not be used do to reasons mentioned here:
http://lkml.org/lkml/2009/3/18/210
* Implementation:
It uses two-level bitmap search to find free list containing block of
correct size. This idea is taken from TLSF (Two-Level Segregate Fit)
allocator and is well explained in its paper (see [Links] below).
* Limitations:
- Poor scalability: No per-cpu data structures (work in progress).
[Links]
1. Details and Performance data:
http://code.google.com/p/compcache/wiki/xvMallochttp://code.google.com/p/compcache/wiki/xvMallocPerformance
2. TLSF memory allocator:
home: http://rtportal.upv.es/rtmalloc/
paper: http://rtportal.upv.es/rtmalloc/files/MRBC_2008.pdf
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Wed, 24 Mar 2010 08:04:11 +0000 (17:04 +0900)
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
Picked out manually from 5a0e3ad6af8660be21ca98a971cd00f331318c05 so we don't break further ramzswap updates when cherry-picking
Currently, we crash (issue BUG_ON) if backing swap
disk size is zero. This can happen is user specified
an extended partition or simply a bad disk as backing
swap. A crash is really an unpleasant surprise to user
for such trivial problems.
Now, we check for this condition and simply fail device
initialization if this is the case.
Additional cleanups:
* use static for all functions
* remove extra newline between functions
* memset backing_swap_name to NULL on device reset
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ramzswap_free_page() already handles the case for zero filled
pages. So, remove redundant logic for the same in ramzswap_write().
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently, we return 0 if create_device() fails and 1 otherwise.
Now, proper error code is returned from create_device() and the
same is propagated as module error code from ramzswap_init().
Also added some cleanups for ramzswap_init(), improving function
structure.
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
ramzswap block size needs to be set equal to PAGE_SIZE to
avoid receiving any unaligned block I/O requests (happens
due to readahead logic during swapon). These unaligned
accesses produce unnecessary I/O errors, scaring users.
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Make sure we flush block device before freeing all metadata
during reset ioctl.
Signed-off-by: Nitin Gupta <ngupta@vflar.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
64-bit stats corruption was observed when ramzswap was
used on SMP systems. To prevent this, use separate spinlock
to protect these stats.
Also, replace stat_*() with rzs_stat*() to avoid possible
conflict with core kernel code.
Eventually, these will be converted to per-cpu counters
if this driver finds use on large scale systems and this
locking is found to affect scalability.
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
If create_device is failed, it can't free gendisk and request_queue of
preceding devices. It cause memory leak.
This patch fixes it.
Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Remove d-cache hack in ramzswap driver that was needed
to workaround a bug in ARM version of update_mmu_cache()
which caused stale data in d-cache to be transferred to
userspace. This bug was fixed by git commit:
787b2faadc4356b6c2c71feb42fb944fece9a12f
This also brings down one entry in TODO file.
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Creates RAM based block devices (/dev/ramzswapX) which can be
used (only) as swap disks. Pages swapped to these are compressed
and stored in memory itself.
The module is called ramzswap.ko. It depends on:
- xvmalloc memory allocator (compiled with this driver)
- lzo_compress.ko
- lzo_decompress.ko
See ramzswap.txt for usage details.
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Added SWP_BLKDEV flag to distinguish block and regular file backed
swap devices. We could also check if a swap is entire block device,
rather than a file, by:
S_ISBLK(swap_info_struct->swap_file->f_mapping->host->i_mode)
but, I think, simply checking this flag is more convenient.
Signed-off-by: Nitin Gupta <ngupta@vflare.org>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Acked-by: Nigel Cunningham <nigel@tuxonice.net>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As a side effect of not doing this, the V4L driver would happily go off and touch a framebuffer address that was invalid, and there may have been a slight SNAFU with a divide by zero on the VPU framebuffer.
Arnaud Patard from Mandriva noticed these a long while ago, but now here they are.
BUG: these are temporary hacks, let's try and work out why line_length is 0..
