dragonfly.git
2 years agotcp: Allow per-tcpcb keepintvl and keepcnt
Sepherosa Ziehau [Fri, 16 Sep 2011 07:53:37 +0000 (15:53 +0800)]
tcp: Allow per-tcpcb keepintvl and keepcnt

- Expose tcp_keepcnt
- Add t_keepintvl, t_keepcnt and t_maxidle to tcpcb.  t_keepintvl and
  t_keepcnt is initialized to tcp_keepintvl and tcp_keepcnt respectively.
  t_maxidle is initialized to t_keepcnt * t_keepintvl
- The accepted socket's t_keepintvl, t_keepcnt and t_maxidle is inherited
  from the listen socket
- Add IPPROTO_TCP/TCP_KEEPINTVL and IPPROTO_TCP.TCP_KEEPCNT socket options
  to get and set t_keepintvl and t_keepcnt.  TCP_KEEPINTVL's unit is
  milliseconds, which is as same as the unit of the sysctl node
  net.inet.tcp.keepintvl
- Remove no longer used tcp_maxidle

2 years agotcp: Allow per-tcpcb keepidle
Sepherosa Ziehau [Fri, 16 Sep 2011 07:13:22 +0000 (15:13 +0800)]
tcp: Allow per-tcpcb keepidle

- Add t_keepidle to tcpcb, it is initialized to tcp_keepidle
- The accepted socket's t_keepidle is inherited from the listen socket
- Add IPPROTO_TCP/TCP_KEEPIDLE socket option to get and set t_keepidle.
  The unit is milliseconds, which is as same as the unit of the sysctl
  node net.inet.tcp.keepidle

2 years agotcp: Allow per-tcpcb keepinit
Sepherosa Ziehau [Fri, 16 Sep 2011 06:41:13 +0000 (14:41 +0800)]
tcp: Allow per-tcpcb keepinit

- Add t_keepinit to tcpcb, it is initialized to tcp_keepinit
- The accepted socket's t_keepinit is inherited from the listen socket
- Add IPPROTO_TCP/TCP_KEEPINIT socket option to get and set t_keepinit.
  The unit is milliseconds, which is as same as the unit of the sysctl
  node net.inet.tcp.keepinit

2 years agotcp: Expose tcp_keepcnt through sysctl net.inet.tcp.keepcnt
Sepherosa Ziehau [Fri, 16 Sep 2011 06:06:46 +0000 (14:06 +0800)]
tcp: Expose tcp_keepcnt through sysctl net.inet.tcp.keepcnt

2 years agokernel - Fix various memory & swap leaks in tmpfs
Matthew Dillon [Thu, 15 Sep 2011 06:25:03 +0000 (23:25 -0700)]
kernel - Fix various memory & swap leaks in tmpfs

* tmpfs was using a vref() instead of a vget() (cache_vget() to be precise)
  which does not necessarily reactivate an inactive vnode.  This can cause
  the vnode to languish in the vnode cache even when the underlying file
  has been removed and no longer has any open() descriptors.

* tmpfs_nrmdir(), tmpfs_nremove(), and tmpfs_nrename() (for target replacement
  renames) was effected.

Reported-by: sephe and others, tested by sephe, dillon
2 years agoMachIntrABI: Field rename; no functional changes
Sepherosa Ziehau [Wed, 14 Sep 2011 13:32:37 +0000 (21:32 +0800)]
MachIntrABI: Field rename; no functional changes

2 years agoi386/files: isa/clock.c is clean enough to compile w/ -Werror
Sepherosa Ziehau [Wed, 14 Sep 2011 13:31:12 +0000 (21:31 +0800)]
i386/files: isa/clock.c is clean enough to compile w/ -Werror

2 years agoopen_memstream.3: A little bit of cleanup.
Sascha Wildner [Tue, 13 Sep 2011 02:33:35 +0000 (04:33 +0200)]
open_memstream.3: A little bit of cleanup.

2 years agolibc: Prototype and manual page for open_memstream.
Venkatesh Srinivas [Mon, 12 Sep 2011 20:35:15 +0000 (13:35 -0700)]
libc: Prototype and manual page for open_memstream.

2 years agoSync zoneinfo database with tzdata2011j from elsie.nci.nih.gov
Sascha Wildner [Mon, 12 Sep 2011 14:02:12 +0000 (16:02 +0200)]
Sync zoneinfo database with tzdata2011j from elsie.nci.nih.gov

africa:         8.32 -> 8.33
australasia:    8.25 -> 8.27

* africa: Four instances of 2:44:45 changed to 2:45 (one in a comment,
    one each in zones Africa/Dar_es_Salaam, Africa/Kampala, Africa/Nairobi)
    (thanks to Jules Descartes)

* australasia: Samoa uses DST in 2011-2012 (thanks to Laupue Raymond Hughes)

2 years agoMachIntrABI: Split vectorctl into intr_setup and intr_teardown
Sepherosa Ziehau [Mon, 12 Sep 2011 12:57:34 +0000 (20:57 +0800)]
MachIntrABI: Split vectorctl into intr_setup and intr_teardown

2 years agoUPDATING: Add some words about the ISA only drivers I removed recently.
Sascha Wildner [Mon, 12 Sep 2011 11:53:09 +0000 (13:53 +0200)]
UPDATING: Add some words about the ISA only drivers I removed recently.

2 years agolibc: Initial implementation of open_memstream.
Venkatesh Srinivas [Mon, 12 Sep 2011 05:12:47 +0000 (22:12 -0700)]
libc: Initial implementation of open_memstream.

open_memstream allows wrapping a buffer via the FILE * interface; the
buffer is dynamically allocated and automatically expanding. The interface
appeared in POSIX 2008.

This implementation doesn't validate seek offsets and doesn't currently
handle SEEK_END.

2 years agokernel: mmap -- Return EINVAL for len = 0 mmap.
Venkatesh Srinivas [Mon, 12 Sep 2011 03:46:07 +0000 (20:46 -0700)]
kernel: mmap -- Return EINVAL for len = 0 mmap.

Closes bug: 2131

2 years agoUpdate fmemopen from NetBSD.
Venkatesh Srinivas [Mon, 12 Sep 2011 02:33:22 +0000 (19:33 -0700)]
Update fmemopen from NetBSD.

NetBSD's version of fmemopen() was more correct than our own wrt error
handling and handling seeks at the tail of buffers. We now pass the ATF
fmemopen test set, with the exception of test08; that problem is related to
malloc return for SIZE_MAX requests.

Changes:
* Modified NetBSD's fmemopen to work with our libc's private FILE struct
* Modified fmemopen_read() to use bcopy() rather than a byte-at-a-time loop

2 years agoSome cleanup in the ndis manpages.
Sascha Wildner [Sat, 10 Sep 2011 18:13:00 +0000 (20:13 +0200)]
Some cleanup in the ndis manpages.

2 years agondis(4): Remove comment that doesn't apply to us.
Sascha Wildner [Sat, 10 Sep 2011 18:12:24 +0000 (20:12 +0200)]
ndis(4): Remove comment that doesn't apply to us.

2 years agokernel: Fix UP remove the "race averted" msg in callout_terminate().
Sascha Wildner [Fri, 9 Sep 2011 16:42:16 +0000 (18:42 +0200)]
kernel: Fix UP remove the "race averted" msg in callout_terminate().

2 years agoFix the VKERNEL/VKERNEL64 build.
Sascha Wildner [Thu, 8 Sep 2011 17:16:42 +0000 (19:16 +0200)]
Fix the VKERNEL/VKERNEL64 build.

Stuff dt_lock into <machine/segments.h>.

2 years agokernel - Add missing prototype for last commit
Matthew Dillon [Wed, 7 Sep 2011 22:54:54 +0000 (15:54 -0700)]
kernel - Add missing prototype for last commit

* Add missing prototype for last commit

2 years agokernel - Fix panic related to kqueue-based timers
Matthew Dillon [Wed, 7 Sep 2011 22:48:07 +0000 (15:48 -0700)]
kernel - Fix panic related to kqueue-based timers

* Fix a panic related to kqueue-based timers.  When a knote is destroyed it
  is possible for the callout_stop() to race a callout on another cpu that
  is blocked on a lock.

* Introduce callout_terminate() which stops a callout and ensures that no
  function calls for the callout is still in progress before returning.
  The kevent code now uses this function instead of callout_stop().

* We can't have this assurance for the normal callout_stop() function because
  it can deadlock a backend stuck waiting for a lock when a frontend holding
  that lock calls callout_stop().

Reported-by: Antonio Huete Jimenez <tuxillo@quantumachine.net>
2 years agoSync ndis(4) and tools with FreeBSD and hook it all back into the build.
Sascha Wildner [Wed, 7 Sep 2011 20:28:06 +0000 (22:28 +0200)]
Sync ndis(4) and tools with FreeBSD and hook it all back into the build.

It supports NDIS 5.x drivers, that means Windows XP and Server 2003.
The latter is an educated guess by me. No testing has actually been
done using Windows Server 2003 drivers.

It is tested on i386 with the following PCI adapters:

* D-Link DWL-G520+ (Texas Instruments ACX111 chip), aka acx(4).

* Linksys WMP600N (Ralink RT2860 chip).

* Planex GW-DS54GR (Realtek RTL8185 chip) which was kindly donated by
  sephe.

Thanks to Max Herrgard <herrgard@gmail.com> for testing it on x86_64
with a CNet CWP-854 (Ralink RT2561T chip), aka ral(4).

Note that this port has the following caveats:

* PCI adapters need hw.ioapic_enable=0 in /boot/loader.conf but worked
  great then. Without it, the box will completely freeze after a while.
  This has been observed on both my own testing box as well as on a
  different box by Max Herrgard. The cause is yet unknown. :-(

* PC Card adapters are untested (they might just work).

* USB adapters are not yet stable at all. Panics might ensue.

Taken-from: FreeBSD

2 years agokernel: Add dt_lock (for GDT and LDT).
Sascha Wildner [Wed, 7 Sep 2011 20:10:53 +0000 (22:10 +0200)]
kernel: Add dt_lock (for GDT and LDT).

2 years agokernel: Add m_extadd().
Sascha Wildner [Thu, 30 Dec 2010 08:16:56 +0000 (09:16 +0100)]
kernel: Add m_extadd().

2 years agoFix some typos: in to -> into.
Sascha Wildner [Wed, 7 Sep 2011 14:11:08 +0000 (16:11 +0200)]
Fix some typos: in to -> into.

2 years agoetc/rc.d/wpa_supplicant: Fix broken check for ndis(4) interfaces.
Sascha Wildner [Wed, 7 Sep 2011 13:26:53 +0000 (15:26 +0200)]
etc/rc.d/wpa_supplicant: Fix broken check for ndis(4) interfaces.

With the new wlan stack, we need to check a sysctl to determine if it
is an ndis(4) interface because it is now wlanX that gets passed to
the script.

Depending on this, wpa_supplicant(8) will be started with either
"-D bsd" or "-D ndis".

Reported-by: Max Herrgard <herrgard@gmail.com>
Taken-from:  FreeBSD

2 years agoinstall.1: One more markup fix.
Sascha Wildner [Wed, 7 Sep 2011 09:00:53 +0000 (11:00 +0200)]
install.1: One more markup fix.

2 years agoinstall.1: A little cleanup.
Sascha Wildner [Wed, 7 Sep 2011 08:53:18 +0000 (10:53 +0200)]
install.1: A little cleanup.

2 years agoAdd a -D option to install(1)
François Tigeot [Wed, 7 Sep 2011 07:37:22 +0000 (09:37 +0200)]
Add a -D option to install(1)

This option is implemented for compatibility with the NetBSD version of
install(1). Even though its code is fully operating, it is only useful
when used in combination with the non-implemented -M NetBSD flag.

Thus, it is effectively a dummy as of the present commit.

2 years agokernel: Add linker_file_foreach() (taken from FreeBSD).
Sascha Wildner [Wed, 29 Dec 2010 16:39:38 +0000 (17:39 +0100)]
kernel: Add linker_file_foreach() (taken from FreeBSD).

2 years ago<net/ethernet.h>: Add ETHER_MAX_FRAME (taken from FreeBSD).
Sascha Wildner [Wed, 29 Dec 2010 15:55:59 +0000 (16:55 +0100)]
<net/ethernet.h>: Add ETHER_MAX_FRAME (taken from FreeBSD).

2 years agomachintr.h: Add header guardian
Sepherosa Ziehau [Mon, 5 Sep 2011 11:52:13 +0000 (19:52 +0800)]
machintr.h: Add header guardian

2 years agoMachIntrABI: Remove unused setvar/getvar interfaces
Sepherosa Ziehau [Mon, 5 Sep 2011 11:42:41 +0000 (19:42 +0800)]
MachIntrABI: Remove unused setvar/getvar interfaces

2 years agomachintr.h: Remove unnecessary inclusion
Sepherosa Ziehau [Mon, 5 Sep 2011 11:21:24 +0000 (19:21 +0800)]
machintr.h: Remove unnecessary inclusion

2 years agobfq.4: Fix and improve a bit.
Sascha Wildner [Sat, 3 Sep 2011 18:18:32 +0000 (20:18 +0200)]
bfq.4: Fix and improve a bit.

2 years agodsched.4: Add a reference to bfq(4).
Sascha Wildner [Sat, 3 Sep 2011 18:17:13 +0000 (20:17 +0200)]
dsched.4: Add a reference to bfq(4).

2 years agopsm.4: Use %O for URLs. We don't have %U yet.
Sascha Wildner [Sat, 3 Sep 2011 18:16:25 +0000 (20:16 +0200)]
psm.4: Use %O for URLs. We don't have %U yet.

2 years agousb.4: Add a reference to lgue(4).
Sascha Wildner [Fri, 2 Sep 2011 20:12:30 +0000 (22:12 +0200)]
usb.4: Add a reference to lgue(4).

2 years agokernel: Use NULL for DRIVER_MODULE()'s evh & arg (which are pointers).
Sascha Wildner [Fri, 2 Sep 2011 20:07:03 +0000 (22:07 +0200)]
kernel: Use NULL for DRIVER_MODULE()'s evh & arg (which are pointers).

2 years agonetif/e1000: Add a missing include file to the Makefile.
Sascha Wildner [Fri, 2 Sep 2011 18:41:12 +0000 (20:41 +0200)]
netif/e1000: Add a missing include file to the Makefile.

Reported-by: Miro Svetlik <miro@svetlik.be>
2 years agoacpi/pcib: Use DEFINE_CLASS_1 to define acpi_pcib driver
Sepherosa Ziehau [Fri, 2 Sep 2011 11:48:11 +0000 (19:48 +0800)]
acpi/pcib: Use DEFINE_CLASS_1 to define acpi_pcib driver

This avoids code duplication and maintenance headache

2 years agopci: Declare pcib_driver
Sepherosa Ziehau [Fri, 2 Sep 2011 11:47:56 +0000 (19:47 +0800)]
pci: Declare pcib_driver

2 years agoThird parameter of VFS_STATFS is no longer struct proc *
François Tigeot [Thu, 1 Sep 2011 21:48:59 +0000 (23:48 +0200)]
Third parameter of VFS_STATFS is no longer struct proc *

2 years agogdb: Remove previous local modification
John Marino [Thu, 1 Sep 2011 18:17:20 +0000 (20:17 +0200)]
gdb: Remove previous local modification

Due to a previous effort that saw binutils support DragonFly, the
config.bfd file no longer needs to be modified.  Unfortunately the
official binutils version wasn't modified the same way as it was in
base, so the auto-merge didn't catch that there were two versions of
each DragonFly entry.  Luckily the good version came first, so this
is simply a cleanup.

Also the README.DRAGONFLY file was updated to include a new set of
local modifications introduced with the FAKE_PROCESS_ID changes.

2 years agokgdb: Tweak to avoid extra inferior
John Marino [Tue, 12 Jul 2011 22:57:53 +0000 (00:57 +0200)]
kgdb: Tweak to avoid extra inferior

Since one inferior exists when kgdb launches, the code was changed to
update the pid of the first inferior rather than to add it outright.

2 years agogdb: Improve concept of FAKE_PROCESS_ID to fix JIT crash
John Marino [Tue, 12 Jul 2011 22:06:43 +0000 (00:06 +0200)]
gdb: Improve concept of FAKE_PROCESS_ID to fix JIT crash

The fix for 7.2 seems to break on gdb 7.3 with addition of JIT
initialization which assumes at least one inferior defined.  The problem
is that this inferior isn't defined before the JIT routine is run.

Changing the fake process ID from zero has a lot of risk (miss some,
accidently change zero's that need to remain zero, etc), but it seems to
be the only way to fix it cleanly.

2 years agogdb: Adjust makefiles for gdb 7.3
John Marino [Tue, 12 Jul 2011 18:42:26 +0000 (20:42 +0200)]
gdb: Adjust makefiles for gdb 7.3

2 years agoMerge branch 'vendor/GDB'
John Marino [Thu, 1 Sep 2011 17:50:06 +0000 (19:50 +0200)]
Merge branch 'vendor/GDB'

Conflicts:
contrib/gdb-7/gdb/charset.c

2 years agoImport gdb 7.3 into vendor branch
John Marino [Wed, 31 Aug 2011 21:22:16 +0000 (23:22 +0200)]
Import gdb 7.3 into vendor branch

3 years agompt(4): Properly lock mpt_timeout().
Sascha Wildner [Wed, 31 Aug 2011 16:24:44 +0000 (18:24 +0200)]
mpt(4): Properly lock mpt_timeout().

I had removed it with an assertion (like in FreeBSD) as part of the
last sync I did, but I had overlooked that FreeBSD initializes their
callout with callout_init_mtx().

For now, just put back locking like it was.

Reported-by: G.Isenmann
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue2127>

3 years agomfi(4): Remove another debug message.
Sascha Wildner [Wed, 31 Aug 2011 09:49:44 +0000 (11:49 +0200)]
mfi(4): Remove another debug message.

3 years agoioapic/x86_64: Rearrange pin information for ioapic low level intrhandler
Sepherosa Ziehau [Wed, 31 Aug 2011 09:09:06 +0000 (17:09 +0800)]
ioapic/x86_64: Rearrange pin information for ioapic low level intrhandler

- Remove unused field, thus reduce the information size from 32 bytes
  to 16bytes.
- Struct, field and macro renaming

3 years agointr/x86_64: Define ISA_IRQ_CNT; use it consistently instead of magic number
Sepherosa Ziehau [Wed, 31 Aug 2011 08:37:22 +0000 (16:37 +0800)]
intr/x86_64: Define ISA_IRQ_CNT; use it consistently instead of magic number

3 years agointr/x86_64: Define IDT_HWI_VECTORS
Sepherosa Ziehau [Wed, 31 Aug 2011 08:11:35 +0000 (16:11 +0800)]
intr/x86_64: Define IDT_HWI_VECTORS

This macro defines how many IDT vectors that could be setup as hardware
interrupts.  Use this macro instead of MAX_HARDINTS and APIC_INTMAPSIZE.

3 years agoicu/x86_64: Remove unnecessary header inclusion
Sepherosa Ziehau [Wed, 31 Aug 2011 06:58:02 +0000 (14:58 +0800)]
icu/x86_64: Remove unnecessary header inclusion

3 years agomfi(4): Remove some debug messages.
Sascha Wildner [Wed, 31 Aug 2011 06:53:31 +0000 (08:53 +0200)]
mfi(4): Remove some debug messages.

3 years agomkdir(1): Don't reference memory beyond the end of the path
Sepherosa Ziehau [Wed, 31 Aug 2011 06:24:05 +0000 (14:24 +0800)]
mkdir(1): Don't reference memory beyond the end of the path

3 years agoLINT/LINT64: Add SAFE_NO_RNG.
Sascha Wildner [Tue, 30 Aug 2011 19:51:31 +0000 (21:51 +0200)]
LINT/LINT64: Add SAFE_NO_RNG.

3 years agoioapic/i386: Rearrange pin information for ioapic low level intrhandler
Sepherosa Ziehau [Mon, 29 Aug 2011 11:54:51 +0000 (19:54 +0800)]
ioapic/i386: Rearrange pin information for ioapic low level intrhandler

- Remove unused field, thus reduce the information size from 32 bytes
  to 16bytes.
- Struct, field and macro renaming

3 years agointr/i386: Define ISA_IRQ_CNT; use it consistently instead of magic number
Sepherosa Ziehau [Sun, 28 Aug 2011 08:18:41 +0000 (16:18 +0800)]
intr/i386: Define ISA_IRQ_CNT; use it consistently instead of magic number

3 years agoSync zoneinfo database with tzdata2011i from elsie.nci.nih.gov
Sascha Wildner [Mon, 29 Aug 2011 11:08:07 +0000 (13:08 +0200)]
Sync zoneinfo database with tzdata2011i from elsie.nci.nih.gov

africa:         8.31 -> 8.32
australasia:    8.23 -> 8.25
europe:         8.33 -> 8.34
iso3166.tab:    8.9  -> 8.11
northamerica:   8.42 -> 8.48
zone.tab:       8.45 -> 8.47

* africa: Add Africa/Juba (South Sudan) zone.

* australasia: Samoa moves from east to west of the international date line
    (changes from UTC-11 to UTC+13). It will skip December 30, 2011.
    Thanks to David Zuelke and Mark Sim-Smith.

* europe: Use KALT as abbreviation for Europe/Kalinigrad from spring 2011
    forward (thanks to Yoshito Umaoka).

* iso3166.tab: Add SS (South Sudan) entry.

* northamerica:
    - Newfoundland changes time of day for DST start and end to 2:00 a.m.
      starting fall 2011 and forward (thanks to Michael Pelley).
    - Change America/Resolute use of EST to 2006-2007 only (thanks to
      Chris Walton).
    - Assume Metlakatla abandoned use of daylight saving as a result of
      its 1983 vote.
    - Avoid unnecessary use of US rule in America/Honolulu; fix two
      typos in America/Honolulu comments.
    - Adjust America/Goose_Bay as well as America/St_Johns from fall 2011
      forward (thanks to Chris Walton).

* zone.tab:
    - Change America/Resolute entry to Central Standard Time.
      (thanks to Chris Walton).
    - Add SS (South Sudan, Africa/Juba) entry.

3 years agobce(4): Flush BCE_PCICFG_INT_ACK_CMD in interrupt handler
Sepherosa Ziehau [Mon, 29 Aug 2011 09:37:36 +0000 (17:37 +0800)]
bce(4): Flush BCE_PCICFG_INT_ACK_CMD in interrupt handler

This could avoid possible spurious interrupts due to the PCI write
to BCE_PCICFG_INT_ACK_CMD is still pending after we leave interrupt
handler.

Obtained-from: Linux bnx2

3 years agoatkbd/x86_64: Do the same AT controller check on x86_64 as on i386
Sepherosa Ziehau [Mon, 29 Aug 2011 07:18:16 +0000 (15:18 +0800)]
atkbd/x86_64: Do the same AT controller check on x86_64 as on i386

DragonFly-bug:
    http://bugs.dragonflybsd.org/issue2052
    http://bugs.dragonflybsd.org/issue2065

3 years agomfi(4): Fix a bug in the disk initialization code.
Sascha Wildner [Mon, 29 Aug 2011 03:19:54 +0000 (05:19 +0200)]
mfi(4): Fix a bug in the disk initialization code.

This ultimately caused a panic when kldunloading.

Reported-by: Tomi Juvonen <tomi.juvonen@kenou.biz>
3 years agokernel: Bring in safe(4), for SafeNet crypto accelerators.
Sascha Wildner [Sun, 28 Aug 2011 22:39:54 +0000 (00:39 +0200)]
kernel: Bring in safe(4), for SafeNet crypto accelerators.

Yet untested.

Taken-from: FreeBSD

3 years agodsched - increase max size of diskctx
Alex Hornung [Sun, 28 Aug 2011 21:57:45 +0000 (21:57 +0000)]
dsched - increase max size of diskctx

3 years agoadd missing DSCHED_BFQ option
Alex Hornung [Sun, 28 Aug 2011 21:51:26 +0000 (21:51 +0000)]
add missing DSCHED_BFQ option

 * fixes LINT build

Reported-by: Sascha Wildner
3 years agodm - Add device type for future raid1
Alex Hornung [Sun, 28 Aug 2011 08:45:12 +0000 (08:45 +0000)]
dm - Add device type for future raid1

Submitted-by: Adam Hoka
Sponsored-by: Google Summer of Code
3 years agodm - Move the target unbusy to the dm_table_destroy function
Adam Hoka [Tue, 21 Jun 2011 01:23:21 +0000 (03:23 +0200)]
dm - Move the target unbusy to the dm_table_destroy function

Sponsored-by: Google Summer of Code
3 years agodm - Add method to get inactive table size
Adam Hoka [Sun, 28 Aug 2011 08:37:19 +0000 (08:37 +0000)]
dm - Add method to get inactive table size

Sponsored-by: Google Summer of Code
3 years agosystat: Change the way ifstats are displayed.
Peter Avalos [Sun, 28 Aug 2011 04:00:29 +0000 (21:00 -0700)]
systat:  Change the way ifstats are displayed.

-Default to displaying bits/s for the rate, and bytes for the totals.
-kilo, mega, giga, tera are all powers of 1000, not 1024.
-Add two new commands: autobyte and autobit which will automatically
change the scale factor in bytes or bits, respectively.  auto defaults
to displaying bits.

3 years agosystat: Sync ifstat code with FreeBSD.
Peter Avalos [Sun, 28 Aug 2011 01:53:25 +0000 (18:53 -0700)]
systat:  Sync ifstat code with FreeBSD.

-Add a tera factor.
-Change factors from binary (1024) to SI (1000).
-Allow for larger scale factors.
-Drop useless initializers and includes.
-Use C99 initializers and enums.
-Remove trailing whitespace.

Obtained-from:   FreeBSD

3 years agodsched_bfq - Add documentation
Brills Peng [Sun, 28 Aug 2011 01:06:24 +0000 (01:06 +0000)]
dsched_bfq - Add documentation

Sponsored-by: Google Summer of Code
3 years agodsched_bfq - A budget fair-queuing dsched policy
Brills Peng [Sat, 27 Aug 2011 18:19:49 +0000 (18:19 +0000)]
dsched_bfq - A budget fair-queuing dsched policy

 * dsched_bfq is a budget fair queuing scheduling policy for the dsched
   framework.

 * NOTE: this scheduler is still highly experimental and
   work-in-progress, it's not recommended for widespread use (yet).
   There are several well-known issues, such as a possible deadlock on
   unloading the module.

Sponsored-by: Google Summer of Code
3 years agodsched_as - A simple & experimental AS
Brills Peng [Sat, 27 Aug 2011 17:27:04 +0000 (17:27 +0000)]
dsched_as - A simple & experimental AS

 * dsched_as is a purely experimental anticipatory scheduler. It serves
   no real-life purpose other than testing and experimenting.

 * Don't use this scheduling policy for anything other than
   experimentation!!!

Sponsored-By: Google Summer of Code
3 years agosystat: Don't revert to bits if >1024GB.
Peter Avalos [Sat, 27 Aug 2011 21:08:04 +0000 (14:08 -0700)]
systat:  Don't revert to bits if >1024GB.

3 years agokernel - Fix ptrace signal races between child and parent
Matthew Dillon [Sat, 27 Aug 2011 19:57:53 +0000 (12:57 -0700)]
kernel - Fix ptrace signal races between child and parent

* Fix several ptrace signal races which could cause the parent's wait()
  function to block forever (or until the parent is killed).  Generally
  speaking we have to use the process token to interlock the wakeup()/tsleep()
  sequence used by the ptrace code.

* Fix a MP race in the wait code, we must hold the child process's token
  when manipulating its p_flag field.

Reported-by: corecode, tuxillo
3 years agodsched - Bump maximum tdio size to 384 bytes
Brills Peng [Sat, 27 Aug 2011 11:42:08 +0000 (11:42 +0000)]
dsched - Bump maximum tdio size to 384 bytes

Sponsored-by: Google Summer of Code
3 years agodsched - Add request polling wrapper
Brills Peng [Sat, 27 Aug 2011 11:24:35 +0000 (11:24 +0000)]
dsched - Add request polling wrapper

 * Add a request polling emulation layer to dsched. This emulated
   request polling as if a disk driver would poll for requests instead
   of requests being actively pushed down.

 * The policy->polling_func() callback is called whenever a BIO
   completes.

 * A field in the diskctx that shows the current tag queue depth and the
   maximum tag queue depth (currently fixed value of 32) are used in the
   policies using request polling directly and is not enforced in the
   dsched layer. That is, a policy using request polling emulation
   should take care of not having (many) more BIOs in flight than
   max_tag_queue_depth.

Sponsored-by: Google Summer of Code
3 years agodsched - Add debugging & fix rare problem conditions
Alex Hornung [Sat, 27 Aug 2011 10:51:24 +0000 (10:51 +0000)]
dsched - Add debugging & fix rare problem conditions

 * Add a bunch of debugging to see whether a particular tdio was
   initialized and to which policy it belongs.

 * Reorder some locking to ensure the whole switch of policy is
   protected as expected.

 * Make sure the tdio from the newest policy is used when there
   are tdios for several scheduling policies in the tdctx->tdio_list.

Reported-by: Brills Peng
3 years agofq - fix a (possible) deadlock
Alex Hornung [Fri, 26 Aug 2011 00:29:17 +0000 (00:29 +0000)]
fq - fix a (possible) deadlock

 * A deadlock can occur if a lock is held around the strategy call when
   the completion is synchronous and ends up in the destruction of a
   tdio, as both will acquire the same diskctx lock.

 * Refactor the code around strategy calls so that the bios are first
   prepared, under the protection of the lock, and inserted into a local
   array. Once that is ready and the bios are ready and taken out of
   their respective queues, the locks are dropped and the strategy call
   occurs locklessly.

Reported-by: Antonio Huete (tuxillo@)
3 years agohammer(8): Adjust status message (Dedup-real -> Dedup).
Sascha Wildner [Thu, 25 Aug 2011 16:59:38 +0000 (18:59 +0200)]
hammer(8): Adjust status message (Dedup-real -> Dedup).

3 years agosh.1: Nuke unneeded .Pp
Sascha Wildner [Thu, 25 Aug 2011 16:31:57 +0000 (18:31 +0200)]
sh.1: Nuke unneeded .Pp

3 years agomfi(4): Add forgotten brace.
Sascha Wildner [Wed, 24 Aug 2011 15:13:09 +0000 (17:13 +0200)]
mfi(4): Add forgotten brace.

3 years agomfi(4): Add a panic to catch future mistakes.
Sascha Wildner [Wed, 24 Aug 2011 06:26:30 +0000 (08:26 +0200)]
mfi(4): Add a panic to catch future mistakes.

It is not reached with the current code but will prevent future
mistakes in case the mfi_identifiers[] array is not set up
properly.

3 years agomfi(4): Style fix.
Sascha Wildner [Wed, 24 Aug 2011 03:48:38 +0000 (05:48 +0200)]
mfi(4): Style fix.

3 years agoFrom NetBSD revision 1.11:
Justin C. Sherrill [Wed, 24 Aug 2011 01:48:15 +0000 (18:48 -0700)]
From NetBSD revision 1.11:

"Do proper input validation without penalizing performance."

Suggested-by: Joerg Sonnenberger <joerg@britannica.bec.de>
3 years agoFrom NetBSD revision 1.10:
Justin C. Sherrill [Wed, 24 Aug 2011 01:40:40 +0000 (18:40 -0700)]
From NetBSD revision 1.10:

"set errno on overflow return."

Suggested-by: Joerg Sonnenberger <joerg@britannica.bec.de>
3 years agoFrom NetBSD revision 1.9:
Justin C. Sherrill [Wed, 24 Aug 2011 01:37:29 +0000 (18:37 -0700)]
From NetBSD revision 1.9:

"provisional fix for CVS-2011-2895, buffer overflow when uncompressing"

Suggested-by: Joerg Sonnenberger <joerg@britannica.bec.de>
3 years agomly(4): Fix comment indentation.
Sascha Wildner [Mon, 22 Aug 2011 20:29:06 +0000 (22:29 +0200)]
mly(4): Fix comment indentation.

3 years agocam(4): Add a quirk to turn off SYNC CACHE for mly(4).
Sascha Wildner [Mon, 22 Aug 2011 20:22:41 +0000 (22:22 +0200)]
cam(4): Add a quirk to turn off SYNC CACHE for mly(4).

This fixes mounting a HAMMER file system from an mly(4) disk on
the AcceleRAID 160 I tried this with.

I think it will also fix the following issue that was reported
for the AcceleRAID 170 on FreeBSD:

http://lists.freebsd.org/pipermail/freebsd-hardware/2008-December/005676.html

The problem with fixing this was that the mly(4) driver modifies
the SCSI inquiry data and uses it for its own purposes (to show
the disk type and status). So I had to make these fields a bit
more recognizable to the quirks code because the previous strings
were too generic for my taste.

3 years agomly(4): Reduce diffs with FreeBSD and update CVS IDs.
Sascha Wildner [Mon, 22 Aug 2011 16:38:26 +0000 (18:38 +0200)]
mly(4): Reduce diffs with FreeBSD and update CVS IDs.

3 years agoLINT: Sort descriptions better than I did in the last commit.
Sascha Wildner [Mon, 22 Aug 2011 12:14:24 +0000 (14:14 +0200)]
LINT: Sort descriptions better than I did in the last commit.

3 years agolongrun(4): Remove from Makefile, it is not a module.
Sascha Wildner [Mon, 22 Aug 2011 11:37:04 +0000 (13:37 +0200)]
longrun(4): Remove from Makefile, it is not a module.

3 years agokernel: Bring back longrun(4) which was lost in 90e8a35b.
Sascha Wildner [Mon, 22 Aug 2011 11:31:42 +0000 (13:31 +0200)]
kernel: Bring back longrun(4) which was lost in 90e8a35b.

FreeBSD had moved it to its own file (from identcpu.c) but that wasn't
brought in.

3 years agomly(4): Perform some minor cleanup.
Sascha Wildner [Mon, 22 Aug 2011 07:17:49 +0000 (09:17 +0200)]
mly(4): Perform some minor cleanup.

* Add dependencies for pci and cam.

* Fix a typo (MHZ -> MHz)

* Cleanup debug printf.

* Remove unused define from the Makefile and add MLY_DEBUG.

3 years agosh(1): Extend documentation about subshells.
Peter Avalos [Sun, 21 Aug 2011 22:12:11 +0000 (15:12 -0700)]
sh(1): Extend documentation about subshells.

Because sh executes commands in subshell environments without forking in
more and more cases, it makes sense to describe subshell environments
more precisely using ideas from POSIX, together with some
DragonFly-specific items.

In particular, the hash and times builtins may not behave as if their
state is copied for a subshell environment while leaving the parent
shell environment unchanged.

Obtained-from:   FreeBSD 223909

3 years agosh: include <limits.h> instead of non-standard <sys/limits.h>
Peter Avalos [Sun, 21 Aug 2011 22:10:34 +0000 (15:10 -0700)]
sh:  include <limits.h> instead of non-standard <sys/limits.h>

Obtained-from:   FreeBSD 223580

3 years agosh(1): Improve documentation of shell patterns:
Peter Avalos [Sun, 21 Aug 2011 22:07:23 +0000 (15:07 -0700)]
sh(1): Improve documentation of shell patterns:

* Shell patterns are also for ${var#pat} and the like.
* An '!' by itself will not trigger pathname generation so do not call
  it a meta-character, even though it has a special meaning directly
  after an '['.
* Character ranges are locale-dependent.
* A '^' will complement a character class like '!' but is non-standard.

Obtained-from:   FreeBSD 223522

3 years agosh(1): Document the case command better.
Peter Avalos [Sun, 21 Aug 2011 22:02:29 +0000 (15:02 -0700)]
sh(1): Document the case command better.

Obtained-from:  FreeBSD 223517