Don't use almost perfectly pessimal cluster allocation. Allocation
of the the first cluster in a file (and, if the allocation cannot be
continued contiguously, for subsequent clusters in a file) was randomized
in an attempt to leave space for contiguous allocation of subsequent
clusters in each file when there are multiple writers. This reduced
internal fragmentation by a few percent, but it increased external
fragmentation by up to a few thousand percent.
Use simple sequential allocation instead. Actually maintain the fsinfo
sequence index for this. The read and write of this index from/to
disk still have many non-critical bugs, but we now write an index that
has something to do with our allocations instead of being modified
garbage. If there is no fsinfo on the disk, then we maintain the index
internally and don't go near the bugs for writing it.
Allocating the first free cluster gives a layout that is almost as good
(better in some cases), but takes too much CPU if the FAT is large and
the first free cluster is not near the beginning.
The effect of this change for untar and tar of a slightly reduced copy
of /usr/src on a new file system was:
Before (msdosfs 4K-clusters):
untar: 459.57 real untar from cached file (actually a pipe)
tar: 342.50 real tar from uncached tree to /dev/zero
Before (ffs2 soft updates 4K-blocks 4K-frags)
untar: 39.18 real
tar: 29.94 real
Before (ffs2 soft updates 16K-blocks 2K-frags)
untar: 31.35 real
tar: 18.30 real
After (msdosfs 4K-clusters):
untar 54.83 real
tar 16.18 real
All of these times can be improved further.
With multiple concurrent writers or readers (especially readers), the
improvement is smaller, but I couldn't find any case where it is
negative. 342 seconds for tarring up about 342 MB on a ~47MB/S partition
is just hard to unimprove on. (This operation would take about 7.3
seconds with reasonably localized allocation and perfect read-ahead.)
However, for active file systems, 342 seconds is closer to normal than
the 16+ seconds above or the 11 seconds with other changes (best I've
measured -- won easily by msdosfs!). E.g., my active /usr/src on ffs1
is quite old and fragmented, so reading to prepare for the above
benchmark takes about 6 times longer than reading back the fresh copies
of it.
Reviewed by: kib, bp (slightly different version)
Tested by: Yuri Pankov <yuri.pankov at gmail dot com>,
Jiawei Ye <leafy7382 at gmail dot com>
Approved by: re (kensmith)
MFp4 122896
- reduce cpu usage by as much as 25% (40% -> 30) by doing txq reclaim more efficiently
- use mtx_trylock when trying to grab the lock to avoid spinning during long encap loop
- add per-txq reclaim task
- if mbufs were successfully re-claimed try another pass
- track txq overruns with sysctl
- Add controller id for Intel 82801I (ICH9).
PR: kern/114399
Submitted by: Michael Fuckner <michael@fuckner.net>
- MSI support. Disable by default due to various issues with too many
broken hardwares. MSI can be enabled through device.hints(5) or
kenv(8) by setting "hint.pcm.%d.msi=1".
Partially submitted by: kevlo
YAMAMOTO Taku <taku@tackymt.homeip.net>
Tested by: joel, kevlo, YAMAMOTO Taku
General style, white space, and comment cleanup; move to ANSI C
prototypes, don't use register, etc. Synchronize structure and
layout to the IPv4 versions of these functions to a greater extent,
making visual comparison easier.
Remove now stale or incorrect comments.
Enable full lock assertions, and correct one exception handling
case where the wrong label was jumped to.
When all the other drivers were converted to scheduling a taskqueue to
do the heavy lifting of the 'mii_tick' function, rue was left behind.
Implement this in a naive way. Reports from the field show this makes
the driver functional with some locking issues, as opposed to an
instant panic. Those will be addressed in a later version of the
driver.
Fix a regression in IPv4 multicast join path (IP_ADD_MEMBERSHIP).
With the in_mcast.c code, if an interface for an IPv4 multicast join was
not specified, and a route did not exist for the specified group in the
unicast forwarding tables, the join would be rejected with the error
EADDRNOTAVAIL.
This change restores the old behaviour whereby if no interface is specified,
and no route exists for the group destination, the IPv4 address list is
walked to find a non-loopback, multicast-capable interface to satisfy
the join request.
This should resolve problems with starting multicast services during
system boot or when a default forwarding entry does not exist.
Eliminate the special case handling of OBJT_DEVICE objects in
vm_fault_additional_pages() that was introduced in revision 1.47. Then
as now, it is unnecessary because dev_pager_haspage() returns zero for
both the number of pages to read ahead and read behind, producing the
same exact behavior by vm_fault_additional_pages() as the special case
handling.
MFp4:
- Plug memory leak.
- Respect underlying vnode's properties rather than assuming that
the user want root:wheel + 0755. Useful for using tmpfs(5) for
/tmp.
- Use roundup2 and howmany macros instead of rolling our own version.
- Try to fix fsx -W -R foo case.
- Instead of blindly zeroing a page, determine whether we need a pagein
order to prevent data corruption.
- Fix several bugs reported by Coverity.
Submitted by: Mingyan Guo <guomingyan gmail com>, Howard Su, delphij
Coverity ID: CID 2550, 2551, 2552, 2557
Approved by: re (tmpfs blanket)
Fix a bug of retrieving configuration ROM.
- Handle directories and leaves other than unit directories and text leaves
correctly.
- Now we can retrieve CROM of iSight correctly.
Approved by: re (hrs)
Tested by: flz
MFC after: 3 days
Actual code shows several problems in ia32 LDT handling:
- When a LDT entry changes, the old one is freed while it is still
referenced by gdt and ldtr. This can lead to disruptive behaviours in
particular on SMP machines.
- When a LDT entry changes, it is assumed that the only one entity sharing
the same LDT are threads in the same proc. It doesn't take in account
edge cases where two processes share the same VM (rfork'ed ones, for
example).
This patch addresses these two problems and addictionally it fixes the
usage of refcount switching back it to the old manually-grown refcount
(since in this case would be faster).
Diagnosed by: tegge
Tested by: pho (a former version)
Reviewed by: kib
Approved by: jeff (mentor)
Approved by: re
Minor UDPv4 cleanup: capitalize comment, move statistics update after mbuf
free to be consistent with other error handling, and release socket buffer
lock before freeing mbufs and statistics updates rather than after.
When a cached page is reactivated in vm_fault(), update the counter that
tracks the total number of reactivated pages. (We have not been
counting reactivations by vm_fault() since revision 1.46.)
Trivial differences with the proposed merged BSD usbdevs file merged
in. These are exclusively in the name of the company for this round.
No new devices have been added, but the MITEL entry has been
eliminated because nothing uses it. You won't see any difference
unless you have USBVERBOSE defined for the kernel.
Before doing compile_re() which needs a parameter to identify
whether we should ignore case, determine the flag by calling
compile_flags() first. Also, make sure that we obtain an
initialized cmd->u.s buffer before processing further. We
may want to refine this solution later, but for now, make
the changes in order to unbreak world build after a sed(1)
with rev. 1.29 of compile.c is installed.
Forced commit to document a cast correction needed to get
clean compiles using either old or new API. I should
have just committed this one-line fix separately. Apologies.
Make test suite work with libarchive 1.3.1: Take advantage of
ARCHIVE_VERSION_STAMP to selectively disable tests that don't
apply to that version; new "skipping()" function reports skipped
tests; modify final summary to report component test failures and
skips.
Note: I don't currently intend to MFC the test suite itself;
anyone interested should just checkout and use this version
of the test suite, which should work for any library version.
New "version stamp" simplifies determining the exact version
of libarchive being used. I've been taking advantage of this
with a recent round of updates to libarchive_test so that it
can test older and newer versions of the library.
Fix some problems with lock_profiling in sx locks:
- Adjust lock_profiling stubs semantic in the hard functions in order to be
more accurate and trustable
- Disable shared paths for lock_profiling. Actually, lock_profiling has a
subtle race which makes results caming from shared paths not completely
trustable. A macro stub (LOCK_PROFILING_SHARED) can be actually used for
re-enabling this paths, but is currently intended for developing use only.
- Use homogeneous names for automatic variables in hard functions regarding
lock_profiling
- Style fixes
- Add a CTASSERT for some flags building
Discussed with: kmacy, kris
Approved by: jeff (mentor)
Approved by: re
I4B header files were repo-copied from sys/i386/include/ to
sys/i4b/include/ so they will be available to all architectures
once I4B compiles on those.
I4B header files were repo-copied from sys/i386/include/ to
sys/i4b/include/ so they will be available to all architectures
once I4B compiles on those.
I4B header files were repo-copied from sys/i386/include/ to
sys/i4b/include/ so they will be available to all architectures
once I4B compiles on those.
peter [Fri, 6 Jul 2007 00:47:44 +0000 (00:47 +0000)]
Fix warnings.
nxge: cast page size fragments down to (int). If the vm's demand paging
PAGE_SIZE is ever too big for that, we've got far bigger problems.
ofw: move va_start() a little earlier. gcc-4.2 doesn't like us modifying
the last arg before the va_start().
re(4) devices requires an external EEPROM. Depending on models it
would be 93C46(1Kbit) or 93C56(2Kbit). One of differences between them
is number of address lines required to access the EEPROM. For example,
93C56 EEPROM needs 8 address lines to read/write data. If 93C56
recevied premature end of required number of serial clock(CLK) to set
OP code/address of EEPROM, the result would be unexpected behavior.
Previously it tried to detect 93C46, which requires 6 address lines,
and then assumed it would be 93C56 if read data was not expected
value. However, this approach didn't work in some models/situations
as 93C56 requries 8 address lines to access its data. In order to fix
it, change EEPROM probing order such that 93C56 is detected reliably.
While I'm here change hard-coded address line numbers with defined
constant to enhance readability.
In preparation for 7.0 privilege cleanup, clean up style:
- Sort copyrights by date.
- Re-wrap, and in some cases, fix comments.
- Fix tabbing, white space, remove extra blank lines.
- Remove commented out debugging printfs.
Allow the LACP state to be queried from userland which at the moment is the
actor and partner peer info. Print out the active aggregator and per port data
in verbose mode from ifconfig.
This was needed during the IPSEC->FAST_IPSEC->IPSEC transition
period to not break the build after picking up netipsec header
files. Now that the FAST_IPSEC kernel option is gone and the
default is IPSEC again those defines are superfluous.
Remove netkey directory from cscope/TAGs generation and replace
it with netipsec now that KAME IPsec is gone.
While here add missing netinet6 directories.
Add comments about the ports needed to be able to run those targets.
peter [Thu, 5 Jul 2007 07:46:33 +0000 (07:46 +0000)]
Quiet framelen uninitialized warning. I think it was a false alarm.
If check_fhdr() returns false, the frame_ok variable should protect any
meaningful evaluations of framelen.
peter [Thu, 5 Jul 2007 06:59:14 +0000 (06:59 +0000)]
Quiet warnings. These do not appear to be actually used uninitialized,
but gcc's optimizer isn't smart enough to see that. Pre-initializing
seems harmless enough.
Add quirks, derived from the linux quirks table for:
o Adonics Cable 205
o Aiptek PocketCAM 3Mega
o Belkin USB2SCSI
o Casio QV DigiCam
o CCYU EasyDisk ED1064
o Desknote UCR-61S2B
o Epson Stylus Photo 875DC Card Reader
o Epson Stylus Photo 895 Card Reader
o Feiya 5-in-1 Card Reader
o Hitachi Dvd-CAM DZ-MV100A Camcorder
o HP CD-WRiter+ CD-4e
o Insystem Storage Adapter v2
o Kyocera Finecam S3x
o Kyocera Finecam S4
o Kyocera Finecam S5
o Kyocera Finecam L3
o Lexar USB CF Reader
o MindAtWork Digital Wallet
o Minolta Dimage F300
o Minolta Dimage E223
o Minsumi USB Fdd
o Netac USB-CF-Card
o NetChip USB Clik! 40
o Onspec MDCFE-B USB CF Reader
o Onspec SIIG/Datafab Memory Stick + CF Reader/Writer
o Onspec Datafab-based Reader
o Onspec PNY/Datafab CF+SM Reader
o Onspec SimpleTech/Datafab CF+SM Reader
o Onspec MDSM-b Reader
o Onspec USB To CF + SM Combo (LC1)
o Onspec ImageMate SDDR55
o Panasonic LS-120 Camera
o Samsung Techwin Digimax 410
o Shuttle eUSB SmartMedia / CompactFlash Adapter
o Skanhex MD 7425 Camera
o Skanhex SX 520z Camera
o Sony Memorystick NW-MS7
o Sony Portable USB Hardrive V2
o Sony Memorystick PEG N760c
o Sony Memorystick MSC-U03
o TREK/IBM USB memory key
o Trumpion T33520 USB Flash Card Controller
o Trumpion MP3 Player
o Vivtar Vivicam 35Xx
o WinMaxGroup USB Flash Disk 64M-C
o Zoran Digital Camera EX-20 DSC
and maybe a few others...
Submitted by: Vaidas Damosevicius and flz
PR: 79893
Reviewed by: njl, flz
Approved by: re (blanket)