The following 8 patches against 2.6.20-mm2 create a zone called ZONE_MOVABLE
that is only usable by allocations that specify both __GFP_HIGHMEM and
__GFP_MOVABLE. This has the effect of keeping all non-movable pages within a
single memory partition while allowing movable allocations to be satisfied
from either partition. The patches may be applied with the list-based
anti-fragmentation patches that groups pages together based on mobility.
The size of the zone is determined by a kernelcore= parameter specified at
boot-time. This specifies how much memory is usable by non-movable
allocations and the remainder is used for ZONE_MOVABLE. Any range of pages
within ZONE_MOVABLE can be released by migrating the pages or by reclaiming.
When selecting a zone to take pages from for ZONE_MOVABLE, there are two
things to consider. First, only memory from the highest populated zone is
used for ZONE_MOVABLE. On the x86, this is probably going to be ZONE_HIGHMEM
but it would be ZONE_DMA on ppc64 or possibly ZONE_DMA32 on x86_64. Second,
the amount of memory usable by the kernel will be spread evenly throughout
NUMA nodes where possible. If the nodes are not of equal size, the amount of
memory usable by the kernel on some nodes may be greater than others.
By default, the zone is not as useful for hugetlb allocations because they are
pinned and non-migratable (currently at least). A sysctl is provided that
allows huge pages to be allocated from that zone. This means that the huge
page pool can be resized to the size of ZONE_MOVABLE during the lifetime of
the system assuming that pages are not mlocked. Despite huge pages being
non-movable, we do not introduce additional external fragmentation of note as
huge pages are always the largest contiguous block we care about.
Credit goes to Andy Whitcroft for catching a large variety of problems during
review of the patches.
This patch creates an additional zone, ZONE_MOVABLE. This zone is only usable
by allocations which specify both __GFP_HIGHMEM and __GFP_MOVABLE. Hot-added
memory continues to be placed in their existing destination as there is no
mechanism to redirect them to a specific zone.
[y-goto@jp.fujitsu.com: Fix section mismatch of memory hotplug related code]
[akpm@linux-foundation.org: various fixes] Signed-off-by: Mel Gorman <mel@csn.ul.ie> Cc: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add __GFP_MOVABLE for callers to flag allocations from high memory that may be migrated
It is often known at allocation time whether a page may be migrated or not.
This patch adds a flag called __GFP_MOVABLE and a new mask called
GFP_HIGH_MOVABLE. Allocations using the __GFP_MOVABLE can be either migrated
using the page migration mechanism or reclaimed by syncing with backing
storage and discarding.
An API function very similar to alloc_zeroed_user_highpage() is added for
__GFP_MOVABLE allocations called alloc_zeroed_user_highpage_movable(). The
flags used by alloc_zeroed_user_highpage() are not changed because it would
change the semantics of an existing API. After this patch is applied there
are no in-kernel users of alloc_zeroed_user_highpage() so it probably should
be marked deprecated if this patch is merged.
Note that this patch includes a minor cleanup to the use of __GFP_ZERO in
shmem.c to keep all flag modifications to inode->mapping in the
shmem_dir_alloc() helper function. This clean-up suggestion is courtesy of
Hugh Dickens.
Additional credit goes to Christoph Lameter and Linus Torvalds for shaping the
concept. Credit to Hugh Dickens for catching issues with shmem swap vector
and ramfs allocations.
[akpm@linux-foundation.org: build fix]
[hugh@veritas.com: __GFP_ZERO cleanup] Signed-off-by: Mel Gorman <mel@csn.ul.ie> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Christoph Lameter <clameter@sgi.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
do_generic_mapping_read currently samples the i_size at the start and doesn't
do so again unless it needs to call ->readpage to load a page. After
->readpage it has to re-sample i_size as a truncate may have caused that page
to be filled with zeros, and the read() call should not see these.
However there are other activities that might cause ->readpage to be called on
a page between the time that do_generic_mapping_read samples i_size and when
it finds that it has an uptodate page. These include at least read-ahead and
possibly another thread performing a read.
So do_generic_mapping_read must sample i_size *after* it has an uptodate page.
Thus the current sampling at the start and after a read can be replaced with
a sampling before the copy-out.
The same change applied to __generic_file_splice_read.
Note that this fixes any race with truncate_complete_page, but does not fix a
possible race with truncate_partial_page. If a partial truncate happens after
do_generic_mapping_read samples i_size and before the copy_out, the nuls that
truncate_partial_page place in the page could be copied out incorrectly.
I think the best fix for that is to *not* zero out parts of the page in
truncate_partial_page, but rather to zero out the tail of a page when
increasing i_size.
Signed-off-by: Neil Brown <neilb@suse.de> Cc: Jens Axboe <jens.axboe@oracle.com> Acked-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Merge branch 'drm-patches' of ssh://master.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6
* 'drm-patches' of ssh://master.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm: add idr_init to drm_stub.c
drm: fix problem with SiS typedef with sisfb enabled.
Merge branch 'drm-patches' of ssh://master.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6
* 'drm-patches' of ssh://master.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm: convert drawable code to using idr
drm: convert drm context code to use Linux idr
Merge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc: (209 commits)
[POWERPC] Create add_rtc() function to enable the RTC CMOS driver
[POWERPC] Add H_ILLAN_ATTRIBUTES hcall number
[POWERPC] xilinxfb: Parameterize xilinxfb platform device registration
[POWERPC] Oprofile support for Power 5++
[POWERPC] Enable arbitary speed tty ioctls and split input/output speed
[POWERPC] Make drivers/char/hvc_console.c:khvcd() static
[POWERPC] Remove dead code for preventing pread() and pwrite() calls
[POWERPC] Remove unnecessary #undef printk from prom.c
[POWERPC] Fix typo in Ebony default DTS
[POWERPC] Check for NULL ppc_md.init_IRQ() before calling
[POWERPC] Remove extra return statement
[POWERPC] pasemi: Don't auto-select CONFIG_EMBEDDED
[POWERPC] pasemi: Rename platform
[POWERPC] arch/powerpc/kernel/sysfs.c: Move NUMA exports
[POWERPC] Add __read_mostly support for powerpc
[POWERPC] Modify sched_clock() to make CONFIG_PRINTK_TIME more sane
[POWERPC] Create a dummy zImage if no valid platform has been selected
[POWERPC] PS3: Bootwrapper support.
[POWERPC] powermac i2c: Use mutex
[POWERPC] Schedule removal of arch/ppc
...
David S. Miller [Tue, 17 Jul 2007 00:05:11 +0000 (17:05 -0700)]
[SERIAL] SUNHV: Fix jerky console on LDOM guests.
Mixing putchar() and write() hvcalls does not work %100
correctly. But we should be using write() all the time
if we can, even from ->start_tx(), anyways.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Mon, 16 Jul 2007 23:50:36 +0000 (16:50 -0700)]
[SPARC64]: Fix race between MD update and dr-cpu add.
We need to make sure the MD update occurs before we try to
process dr-cpu configure requests. MD update and dr-cpu
were being processed by seperate threads so that did not
happen occaisionally.
Fix this by executing all domain services data packets from
a single thread, in order.
This will help simplify some other things as well.
Signed-off-by: David S. Miller <davem@davemloft.net>
The SCSI code can be compiled modular, but BLK_DEV_BSG currently cannot,
and depends on the SCSI layer. So make sure that it depends on the SCSI
layer being compiled in, not just available as a module.
Jay Cliburn [Sun, 15 Jul 2007 16:03:29 +0000 (11:03 -0500)]
atl1: reorder atl1_main functions
Reorder functions in atl1_main into more logical groupings to make the
code easier to follow. This patch is large, but it's harmless; it neither
adds nor removes any functionality whatsoever.
Signed-off-by: Jay Cliburn <jacliburn@bellsouth.net> Signed-off-by: Jeff Garzik <jeff@garzik.org>
myri10ge: Remove nonsensical limit in the tx done routine
Remove nonsensical limit in the tx done routine. Specifically,
the loop will always terminate after processing <= 1 rings worth
of frames, as the mcp index is not refetched, so the removed
conditional could never be true.
Signed-off-by: Brice Goglin <brice@myri.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Frank Blaschka [Thu, 12 Jul 2007 10:51:34 +0000 (12:51 +0200)]
s390: scatter-gather for inbound traffic in qeth driver
For large incoming packets > PAGE_SIZE/2 qeth creates a fragmented skb
by adding pointers to qdio pages to the fragment list of the skb.
This avoids allocating big chunks of consecutive memory. Also copying
data from the qdio buffer to the skb is economized.
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com> Signed-off-by: Ursula Braun <braunu@de.ibm.com> Signed-off-by: Frank Pavlic <fpavlic@de.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Thomas Klein [Wed, 11 Jul 2007 14:32:00 +0000 (16:32 +0200)]
eHEA: Introducing support vor DLPAR memory add
This patch adds support for DLPAR memory add to the eHEA driver. To detect
whether memory was added the driver uses its own memory mapping table and
checks for kernel addresses whether they're located in already known memory
sections. If not the function ehea_rereg_mrs() is triggered which performs
a rebuild of the mapping table and a re-registration of the global memory
region.
Signed-off-by: Thomas Klein <tklein@de.ibm.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Fix a potential NULL pointer dereference in free_shared_mem() in drivers/net/s2io.c
This patch fixes a potential null dereference bug where we dereference
nic before a null check. This patch simply moves the dereferencing
after the null check.
Signed-off-by: Micah Gruber < micah.gruber@gmail.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
Ingo Molnar reports complete breakage with his e1000 card (no
networking, card reports transmit timeouts), and bisected it down to
this commit. Let's figure out what went wrong, but not keep breaking
machines until we do.
Cc: Ingo Molnar <mingo@elte.hu> Cc: Olaf Kirch <olaf.kirch@oracle.com> Cc: David Miller <davem@davemloft.net> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Merge branch 'drm-patches' of ssh://master.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6
* 'drm-patches' of ssh://master.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm: remove core typedefs from the ioc32 wrappers
drm: remove sarea typedefs
drm: detypedef the hashtab and more of sman
drm: de-typedef sman
drm: detypedeffing continues...
drm: detypef waitlist/freelist/buf_entry/device_dma/drm_queue structs
drm: drop drm_vma_entry_t, drm_magic_entry_t
drm: drop drm_buf_t typedef
drm: fixup other drivers for typedef removals
drm: remove drm_file_t, drm_device_t and drm_head_t typedefs
drm: remove a bunch of typedefs on the userspace interface
r300: updates register header
radeon: add support for vblank on crtc2
drm: cleanup list initialisation
drm: fix typo on code drm getsarea
drm: remove DRM_GETSAREA and replace with drm_getsarea function
drm: cleanup use of Linux list handling macros
caglar@zangetsu linux-2.6 $ make
CHK include/linux/version.h
CHK include/linux/utsrelease.h
CALL scripts/checksyscalls.sh
CHK include/linux/compile.h
LD drivers/block/built-in.o
CC [M] drivers/block/cciss.o
drivers/block/cciss.c: In function `cciss_ioctl':
drivers/block/cciss.c:1173: warning: passing arg 2 of `scsi_cmd_ioctl' from incompatible pointer type
drivers/block/cciss.c:1173: warning: passing arg 3 of `scsi_cmd_ioctl' makes pointer from integer without a cast
drivers/block/cciss.c:1173: warning: passing arg 4 of `scsi_cmd_ioctl' makes integer from pointer without a cast
drivers/block/cciss.c:1173: error: too few arguments to function `scsi_cmd_ioctl'
...
make[2]: *** [drivers/block/cciss.o] Hata 1
make[1]: *** [drivers/block] Hata 2
make: *** [drivers] Hata 2
Teach LDM about a new field encountered with Windows Vista.
This fixes LDM for people using Vista who have disabled drive letter
assignment from one or more volumes. Doing this introduces a so far
unknown field in the LDM database in the VOL5 VBLK structure which
causes the LDM driver to fail to parse the VBLK structure and hence LDM
fails to parse the disk altogether. This patch teaches the driver about
this field.
Thanks got to Ashton Mills <amills@iinet.com.au> for reporting the
problem and working with me on getting it fixed. It is now working for
him.
Signed-off-by: Anton Altaparmakov <aia21@cantab.net> CC: Richard Russon <ldm@flatcap.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Randy Dunlap noticed that the recent comment clarifications from Andrew
had somehow gotten duplicated. Quoth Andrew: "hm, that could have been
some late-night reject-fixing."
Fix it up.
Cc: From: Andrew Morton <akpm@linux-foundation.org> Cc: Randy Dunlap <randy.dunlap@oracle.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We had a merge issue with the "dentry" field going away from the
kobject, and being replaced by a sysfs_dirent field (named "sd")
instead. That broke the BSG compile.
Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2: (32 commits)
[PATCH] ocfs2: zero_user_page conversion
ocfs2: Support xfs style space reservation ioctls
ocfs2: support for removing file regions
ocfs2: update truncate handling of partial clusters
ocfs2: btree support for removal of arbirtrary extents
ocfs2: Support creation of unwritten extents
ocfs2: support writing of unwritten extents
ocfs2: small cleanup of ocfs2_write_begin_nolock()
ocfs2: btree changes for unwritten extents
ocfs2: abstract btree growing calls
ocfs2: use all extent block suballocators
ocfs2: plug truncate into cached dealloc routines
ocfs2: simplify deallocation locking
ocfs2: harden buffer check during mapping of page blocks
ocfs2: shared writeable mmap
ocfs2: factor out write aops into nolock variants
ocfs2: rework ocfs2_buffered_write_cluster()
ocfs2: take ip_alloc_sem during entire truncate
ocfs2: Add "preferred slot" mount option
[KJ PATCH] Replacing memset(<addr>,0,PAGE_SIZE) with clear_page() in fs/ocfs2/dlm/dlmrecovery.c
...
Merge branch 'bsg' of git://git.kernel.dk/data/git/linux-2.6-block
* 'bsg' of git://git.kernel.dk/data/git/linux-2.6-block: (25 commits)
bsg: Kconfig updates
bsg: add SCSI transport-level request support
bsg: add bidi support
add a struct request pointer to the request structure
bsg: fix the deadlock on discarding done commands
bsg: fix a blocking read bug
bsg: minor bug fixes
improve bsg device allocation
bind bsg to all SCSI devices
bsg: bind bsg to request_queue instead of gendisk
bsg: add a request_queue argument to scsi_cmd_ioctl()
bsg: simplify __bsg_alloc_command failpath
bsg: add cheasy error checks for sysfs stuff
Add queue resizing support
Replace s32, u32 and u64 with __s32, __u32 and __u64 in bsg.h for userspace
bsg: silence a bogus gcc warning
bsg: style cleanup
bsg: use u32 etc instead of uint32_t
bsg: add SG_IO to SG v4
bsg: replace SG v3 with SG v4
...
Merge branch 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block
* 'for-linus' of git://git.kernel.dk/data/git/linux-2.6-block:
splice: direct splicing updates ppos twice
more ACSI removal
umem: Fix match of pci_ids in umem driver
umem: Remove references to dead CONFIG_MM_MAP_MEMORY variable
remove the documentation for the legacy CDROM drivers
Merge branch 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6
* 'master' of master.kernel.org:/pub/scm/linux/kernel/git/davem/sparc-2.6: (26 commits)
[SPARC64]: Fix UP build.
[SPARC64]: dr-cpu unconfigure support.
[SERIAL]: Fix console write locking in sparc drivers.
[SPARC64]: Give more accurate errors in dr_cpu_configure().
[SPARC64]: Clear cpu_{core,sibling}_map[] in smp_fill_in_sib_core_maps()
[SPARC64]: Fix leak when DR added cpu does not bootup.
[SPARC64]: Add ->set_affinity IRQ handlers.
[SPARC64]: Process dr-cpu events in a kthread instead of workqueue.
[SPARC64]: More sensible udelay implementation.
[SPARC64]: SMP build fixes.
[SPARC64]: mdesc.c needs linux/mm.h
[SPARC64]: Fix build regressions added by dr-cpu changes.
[SPARC64]: Unconditionally register vio_bus_type.
[SPARC64]: Initial LDOM cpu hotplug support.
[SPARC64]: Fix setting of variables in LDOM guest.
[SPARC64]: Fix MD property lifetime bugs.
[SPARC64]: Abstract out mdesc accesses for better MD update handling.
[SPARC64]: Use more mearningful names for IRQ registry.
[SPARC64]: Initial domain-services driver.
[SPARC64]: Export powerd facilities for external entities.
...
* master.kernel.org:/pub/scm/linux/kernel/git/lethal/sh-2.6: (68 commits)
sh: sh-rtc support for SH7709.
sh: Revert __xdiv64_32 size change.
sh: Update r7785rp defconfig.
sh: Export div symbols for GCC 4.2 and ST GCC.
sh: fix race in parallel out-of-tree build
sh: Kill off dead mach.c for hp6xx.
sh: hd64461.h cleanup and added comments.
sh: Update the alignment when 4K stacks are used.
sh: Add a .bss.page_aligned section for 4K stacks.
sh: Don't let SH-4A clobber SH-4 CFLAGS.
sh: Add parport stub for SuperIO ports.
sh: Drop -Wa,-dsp for DSP tuning.
sh: Update dreamcast defconfig.
fb: pvr2fb: A few more __devinit annotations for PCI.
fb: pvr2fb: Fix up section mismatch warnings.
sh: Select IPR-IRQ for SH7091.
sh: Correct __xdiv64_32/div64_32 return value size.
sh: Fix timer-tmu build for SH-3.
sh: Add cpu and mach links to CLEAN_FILES.
sh: Preliminary support for the SH-X3 CPU.
...
FAT12 entry is 12bits, so it needs 2 phase to update the value. And
writer and reader access it without any lock, so reader can get the
half updated value.
This fixes the long standing race condition by adding a global
spinlock to only FAT12 for avoiding any impact against FAT16/32.
Andrew Morton [Mon, 16 Jul 2007 06:42:03 +0000 (23:42 -0700)]
cpwatchdog build fix
sparc64:
drivers/sbus/char/cpwatchdog.c: In function `wd_toggleintr':
drivers/sbus/char/cpwatchdog.c:523: error: implicit declaration of function `readb'
drivers/sbus/char/cpwatchdog.c:533: error: implicit declaration of function `writeb'
drivers/sbus/char/cpwatchdog.c: In function `wd_pingtimer':
drivers/sbus/char/cpwatchdog.c:545: error: implicit declaration of function `readw'
drivers/sbus/char/cpwatchdog.c: In function `wd_starttimer':
drivers/sbus/char/cpwatchdog.c:584: error: implicit declaration of function `writew'
drivers/sbus/char/cpwatchdog.c: In function `wd_init':
drivers/sbus/char/cpwatchdog.c:767: error: implicit declaration of function `ioremap'
drivers/sbus/char/cpwatchdog.c:767: warning: assignment makes pointer from integer without a cast
drivers/sbus/char/cpwatchdog.c: In function `wd_cleanup':
drivers/sbus/char/cpwatchdog.c:849: error: implicit declaration of function `iounmap'
Cc: "David S. Miller" <davem@davemloft.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Robert P. J. Day [Mon, 16 Jul 2007 06:42:01 +0000 (23:42 -0700)]
NLS: Remove obsolete Makefile entries
Since the corresponding source files no longer exist, remove the
irrelevant Makefile entries for them.
Signed-off-by: Robert P. J. Day <rpjday@mindspring.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Rusty Russell [Mon, 16 Jul 2007 06:42:00 +0000 (23:42 -0700)]
permit mempool_free(NULL)
Christian Borntraeger points out that mempool_free() doesn't noop when
handed NULL. This is inconsistent with the other free-like functions
in the kernel.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Cc: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a patch that speeds up statfs. It is very simple - the "overhead"
calculation, which takes a huge amount of time for large filesystems, never
changes unless the size of the filesystem itself changes. That means we can
store it in memory and only recalculate if the filesystem has been resized
(almost never).
It also fixes a minor problem that we never update the on-disk superblock free
blocks/inodes counts until the filesystem is unmounted. While not fatal, we
may as well update that on disk when we have the information, and it makes
things like debugfs and dumpe2fs report a bit more accurate info.
Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com> Signed-off-by: Andreas Dilger <adilger@clusterfs.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a patch that speeds up statfs. It is very simple - the "overhead"
calculation, which takes a huge amount of time for large filesystems, never
changes unless the size of the filesystem itself changes. That means we can
store it in memory and only recalculate if the filesystem has been resized
(almost never).
It also fixes a minor problem that we never update the on-disk superblock free
blocks/inodes counts until the filesystem is unmounted. While not fatal, we
may as well update that on disk when we have the information, and it makes
things like debugfs and dumpe2fs report a bit more accurate info.
Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com> Signed-off-by: Andreas Dilger <adilger@clusterfs.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a patch that speeds up statfs. It is very simple - the "overhead"
calculation, which takes a huge amount of time for large filesystems, never
changes unless the size of the filesystem itself changes. That means we can
store it in memory and only recalculate if the filesystem has been resized
(almost never).
It also fixes a minor problem that we never update the on-disk superblock free
blocks/inodes counts until the filesystem is unmounted. While not fatal, we
may as well update that on disk when we have the information, and it makes
things like debugfs and dumpe2fs report a bit more accurate info.
Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com> Signed-off-by: Andreas Dilger <adilger@clusterfs.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
vsprintf.c: optimizing, part 2: base 10 conversion speedup, v2
Optimize integer-to-string conversion in vsprintf.c for base 10. This is
by far the most used conversion, and in some use cases it impacts
performance. For example, top reads /proc/$PID/stat for every process, and
with 4000 processes decimal conversion alone takes noticeable time.
Using code from
http://www.cs.uiowa.edu/~jones/bcd/decimal.html
(with permission from the author, Douglas W. Jones)
binary-to-decimal-string conversion is done in groups of five digits at
once, using only additions/subtractions/shifts (with -O2; -Os throws in
some multiply instructions).
On i386 arch gcc 4.1.2 -O2 generates ~500 bytes of code.
This patch is run tested. Userspace benchmark/test is also attached.
I tested it on PIII and AMD64 and new code is generally ~2.5 times
faster. On AMD64:
# ./vsprintf_verify-O2
Original decimal conv: .......... 151 ns per iteration
Patched decimal conv: .......... 62 ns per iteration
Testing correctness 12895992590592 ok... [Ctrl-C]
# ./vsprintf_verify-O2
Original decimal conv: .......... 151 ns per iteration
Patched decimal conv: .......... 62 ns per iteration
Testing correctness 26025406464 ok... [Ctrl-C]
More realistic test: top from busybox project was modified to
report how many us it took to scan /proc (this does not account
any processing done after that, like sorting process list),
and then I test it with 4000 processes:
#!/bin/sh
i=4000
while test $i != 0; do
sleep 30 &
let i--
done
busybox top -b -n3 >/dev/null
on unpatched kernel:
top: 4120 processes took 102864 microseconds to scan
top: 4120 processes took 91757 microseconds to scan
top: 4120 processes took 92517 microseconds to scan
top: 4120 processes took 92581 microseconds to scan
on patched kernel:
top: 4120 processes took 75460 microseconds to scan
top: 4120 processes took 66451 microseconds to scan
top: 4120 processes took 67267 microseconds to scan
top: 4120 processes took 67618 microseconds to scan
The speedup comes from much faster generation of /proc/PID/stat
by sprintf() calls inside the kernel.
Signed-off-by: Douglas W Jones <jones@cs.uiowa.edu> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
vsprintf.c: optimizing, part 1 (easy and obvious stuff)
* There is no point in having full "0...9a...z" constant vector,
if we use only "0...9a...f" (and "x" for "0x").
* Post-decrement usually needs a few more instructions, so use
pre decrement instead where makes sense:
- while (i < precision--) {
+ while (i <= --precision) {
* if base != 10 (=> base 8 or 16), we can avoid using division
in a loop and use mask/shift, obtaining much faster conversion.
(More complex optimization for base 10 case is in the second patch).
Overall, size vsprintf.o shows ~80 bytes smaller text section
with this patch applied.
Signed-off-by: Douglas W Jones <jones@cs.uiowa.edu> Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
isapnp: remove pointless check of 'type' against 0 in isapnp_read_tag()
In drivers/pnp/isapnp/core.c::isapnp_read_tag() there is a test of 'type'
being == 0 a bit down in the function. That test doesn't make any sense.
If 'type' could indeed be NULL, then the test happens way too late as we'd
already have tried to dereference the pointer earlier and looking at the
callers it also turns out that there is no way type can ever actually be
NULL.
So the test is completely pointless and should just be removed.
kernel/printk.c: document possible deadlock against scheduler
kernel/printk.c: document possible deadlock against scheduler
The printk's comment states that it can be called from every context,
which might lead to false illusion that it could be called from everywhere
without any restrictions.
This is however not true - a call to printk() could deadlock if called from
scheduler code (namely from schedule(), wake_up(), etc) on runqueue lock
when it tries to wake up klogd. Document this.
Ed L. Cashin [Mon, 16 Jul 2007 06:41:50 +0000 (23:41 -0700)]
docs: static initialization of spinlocks is OK
Static initialization of spinlocks is preferable to dynamic initialization
when it is practical. This patch updates documentation for consistency
with comments in spinlock_types.h.
Signed-off-by: Ed L. Cashin <ecashin@coraid.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Alan Cox [Mon, 16 Jul 2007 06:41:47 +0000 (23:41 -0700)]
genericserial: remove bogus optimisation check and dead code paths
We've been using the 'new locking' for a long time now so it seems
pointless keeping the old one around. Remove it and undo the macros it
uses back into real code for readability. Remove the bogus 'no termios
change' checks.
Signed-off-by: Alan Cox <alan@redhat.com> Cc: Morten Helgesen <morten@sourcepoet.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Lots of serial drivers check and optimise for setting the termios values to
the ones they were before. This is pointless and the check is wrong
anyway. Remove the checks on the serial drivers. If we ever do need such
a check put it back in the tty layer instead _once_!
Signed-off-by: Alan Cox <alan@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Rusty Russell [Mon, 16 Jul 2007 06:41:46 +0000 (23:41 -0700)]
modules: remove modlist_lock
Now we always use stop_machine for module insertion or deletion, we no
longer need the modlist_lock: merely disabling preemption is sufficient to
block against list manipulation. This avoids deadlock on OOPSen where we
can potentially grab the lock twice.
Bug: 8695 Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Cc: Ingo Molnar <mingo@elte.hu> Cc: Tobias Oed <tobiasoed@hotmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Change cancel_work_sync() and cancel_delayed_work_sync() to return a boolean
indicating whether the work was actually cancelled. A zero return value means
that the work was not pending/queued.
Without that kind of change it is not possible to avoid flush_workqueue()
sometimes, see the next patch as an example.
Also, this patch unifies both functions and kills the (unlikely) busy-wait
loop.
This looks as if the first 2 functions differ in "type" of their argument
which is not true any longer, nowadays the difference is the behaviour.
The semantics of cancel_rearming_delayed_work(dwork) was changed
significantly, it doesn't require that dwork rearms itself, and cancels dwork
synchronously.
Rename it to cancel_delayed_work_sync(). This matches cancel_delayed_work()
and cancel_work_sync(). Re-create cancel_rearming_delayed_work() as a simple
inline obsolete wrapper, like cancel_rearming_delayed_workqueue().
The UMSDOS filesystem was removed back in 2.6.11, but some tiny bits stuck
around. This patch removes the few remaining leftovers. The only things
left behind after this are the entries in the CREDITS file and the ioctl
number in Documentation/ioctl-number.txt as documentation.
This third (hopefully final) version of the patch doesn't edit the
arch/um/config.release file, since Jeff Dike pointed out to me that it
should die completely, and asked me to remove it from my patch as he'll
send in a seperate patch removing the file completely.
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com> Acked-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
generic bug: use show_regs() instead of dump_stack()
The current generic bug implementation has a call to dump_stack() in case a
WARN_ON(whatever) gets hit. Since report_bug(), which calls dump_stack(),
gets called from an exception handler we can do better: just pass the
pt_regs structure to report_bug() and pass it to show_regs() in case of a
warning. This will give more debug informations like register contents,
etc... In addition this avoids some pointless lines that dump_stack()
emits, since it includes a stack backtrace of the exception handler which
is of no interest in case of a warning. E.g. on s390 the following lines
are currently always present in a stack backtrace if dump_stack() gets
called from report_bug():
Dave Jones [Mon, 16 Jul 2007 06:41:38 +0000 (23:41 -0700)]
Add -Werror-implicit-function-declaration
Add -Werror-implicit-function-declaration
This makes builds fail sooner if something is implicitly defined instead
of having to wait half an hour for it to fail at the linking stage.
Signed-off-by: Dave Jones <davej@redhat.com> Cc: Sam Ravnborg <sam@ravnborg.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Niels de Vos [Mon, 16 Jul 2007 06:41:35 +0000 (23:41 -0700)]
parport_pc: it887x fix
The IO port range requested by parport_pc.c:sio_ite_8872_probe is too small.
The IO-ports of ttyS1 (0x2f8) will be missconfigured by the ITE-chip. The ITE
starts looking for the chip a 0x2a0. An IO-portrange of 32 will not overwrite
the ports of ttyS1. Therefore register 0x60 should be written with
0xe5000000, enabling the ITE and setting IO-portsize to 32 bytes.
Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Andrea Arcangeli [Mon, 16 Jul 2007 06:41:33 +0000 (23:41 -0700)]
make seccomp zerocost in schedule
This follows a suggestion from Chuck Ebbert on how to make seccomp
absolutely zerocost in schedule too. The only remaining footprint of
seccomp is in terms of the bzImage size that becomes a few bytes (perhaps
even a few kbytes) larger, measure it if you care in the embedded.
Signed-off-by: Andrea Arcangeli <andrea@cpushare.com> Cc: Andi Kleen <ak@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Andrea Arcangeli [Mon, 16 Jul 2007 06:41:32 +0000 (23:41 -0700)]
move seccomp from /proc to a prctl
This reduces the memory footprint and it enforces that only the current
task can enable seccomp on itself (this is a requirement for a
strightforward [modulo preempt ;) ] TIF_NOTSC implementation).
Signed-off-by: Andrea Arcangeli <andrea@cpushare.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
projects / linux-2.6-omap-h63xx.git / log