Sascha Wildner [Fri, 10 Dec 2010 12:04:36 +0000 (13:04 +0100)]
Bring in an igb(4) manual page (from FreeBSD).
Sascha Wildner [Fri, 10 Dec 2010 10:30:20 +0000 (11:30 +0100)]
manual pages: Punctuation and whitespace cleanup.
Matthew Dillon [Fri, 10 Dec 2010 07:53:44 +0000 (23:53 -0800)]
kernel - Fix MP system call performance regression
* The userland scheduler was unconditionally calling lwkt_switch()
via userexit() (i.e. on every system call), creating unnecessary
overhead and possibly also triggering a bsd4 scheduler event
requiring a common spinlock.
* Rearrange the code slightly to reduce instances where lwkt_switch()
is called. We want to try to keep instances where a higher priority
LWKT thread is potentially runnable or when the LWKT fairq accumulator
for the current thread has been exhausted.
* This removes system call overhead multiplication on MP systems. For
example, on a 48-core box system call overhead when all 48 cpus are
busy doing getuid() loops went from 10uS back down to 270nS (which
is near the single-cpu test results).
Matthew Dillon [Fri, 10 Dec 2010 07:12:20 +0000 (23:12 -0800)]
kernel - Add support for up to 63 cpus & 512G of ram for 64-bit builds (3)
* Correct compile-time errors in vkernel64 build.
Matthew Dillon [Fri, 10 Dec 2010 07:11:32 +0000 (23:11 -0800)]
kernel - Simplify AP CPU reporting on x86-64
* Simplify the AP CPU kprintf()s to reduce console spew.
Matthew Dillon [Fri, 10 Dec 2010 07:10:50 +0000 (23:10 -0800)]
kernel - Add support for up to 63 cpus & 512G of ram for 64-bit builds (2)
* Fix mis-translation in previous commit
Samuel J. Greear [Fri, 10 Dec 2010 06:35:33 +0000 (06:35 +0000)]
style(9) - Prefer a lack of case '?' fallthrough for getopt
* This also coincides with the example in getopt(3)
Discussed-with: swildner@
Samuel J. Greear [Fri, 10 Dec 2010 06:16:57 +0000 (06:16 +0000)]
Merge branch 'malloc_constructor'
Samuel J. Greear [Fri, 10 Dec 2010 06:16:48 +0000 (06:16 +0000)]
Merge branch 'GCI2010'
Samuel J. Greear [Fri, 10 Dec 2010 04:58:22 +0000 (04:58 +0000)]
vkernel - Add usage()
* Adds a new argument -h, which prints extended usage information
* Prints typical usage information (same as manpage) when no arguments are
specified or an invalid argument is given. Otherwise, error output is the
same as before.
* Convert to sysexits(3)
Sponsored-By: Google Code-In
Matthew Dillon [Thu, 9 Dec 2010 22:09:17 +0000 (14:09 -0800)]
kernel - Add support for up to 63 cpus & 512G of ram for 64-bit builds.
* Increase SMP_MAXCPU to 63 for 64-bit builds.
* cpumask_t is 64 bits on 64-bit builds now. It remains 32 bits on 32-bit
builds.
* Add #define's for atomic_set_cpumask(), atomic_clear_cpumask, and
atomic_cmpset_cpumask(). Replace all use cases on cpu masks with
these functions.
* Add CPUMASK(), BSRCPUMASK(), and BSFCPUMASK() macros. Replace all
use cases on cpu masks with these functions.
In particular note that (1 << cpu) just doesn't work with a 64-bit
cpumask.
Numerous bits of assembly also had to be adjusted to use e.g. btq instead
of btl, etc.
* Change __uint32_t declarations that were meant to be cpu masks to use
cpumask_t (most already have).
Also change other bits of code which work on cpu masks to be more agnostic.
For example, poll_cpumask0 and lwp_cpumask.
* 64-bit atomic ops cannot use "iq", they must use "r", because most x86-64
do NOT have 64-bit immediate value support.
* Rearrange initial kernel memory allocations to start from KvaStart and
not KERNBASE, because only 2GB of KVM is available after KERNBASE.
Certain VM allocations with > 32G of ram can exceed 2GB. For example,
vm_page_array[]. 2GB was not enough.
* Remove numerous mdglobaldata fields that are not used.
* Align CPU_prvspace[] for now. Eventually it will be moved into a
mapped area. Reserve sufficient space at MPPTDI now, but it is
still unused.
* When pre-allocating kernel page table PD entries calculate the number
of page table pages at KvaStart and at KERNBASE separately, since
the KVA space starting at KERNBASE caps out at 2GB.
* Change kmem_init() and vm_page_startup() to not take memory range
arguments. Instead the globals (virtual_start and virtual_end) are
manipualted directly.
Sascha Wildner [Thu, 9 Dec 2010 05:12:16 +0000 (06:12 +0100)]
periodic(8): Adjust the bsdstats script to use drill(1) instead of dig(1).
The check_dns() check is still wrong which is due to the setting of
IFS but until scrappy@ reports back, I won't touch that. For the
moment, it will succeed, but would also not fail in check_dns() if
the network was down.
Reported-by: Tim Darby <t+dfbsd@timdarby.net>
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1927>
Sascha Wildner [Thu, 9 Dec 2010 04:10:26 +0000 (05:10 +0100)]
mfiutil(8): Fix a in the interval argument to the 'patrol auto' command.
Sascha Wildner [Thu, 9 Dec 2010 04:09:03 +0000 (05:09 +0100)]
swapcache.8: Fix typos.
Sascha Wildner [Thu, 9 Dec 2010 04:08:33 +0000 (05:08 +0100)]
devattr(8): Fix weird style and perform some general cleanup.
Sascha Wildner [Wed, 8 Dec 2010 15:42:06 +0000 (16:42 +0100)]
Fix installworld.
Venkatesh Srinivas [Wed, 8 Dec 2010 06:55:34 +0000 (22:55 -0800)]
bitstring(3): Describe bit_fls and bit_nsearch.
Venkatesh Srinivas [Wed, 8 Dec 2010 06:44:17 +0000 (22:44 -0800)]
bitstring.h: Add bit_fls to complement bit_ffs and bit_nsearch.
Venkatesh Srinivas [Wed, 8 Dec 2010 06:37:25 +0000 (22:37 -0800)]
Import vm_map_findspace(9) manpage from FreeBSD and add flags/align arguments.
Sascha Wildner [Wed, 8 Dec 2010 02:38:38 +0000 (03:38 +0100)]
Kernel configs: Move twa(4) to the right place, too.
Sascha Wildner [Wed, 8 Dec 2010 02:35:57 +0000 (03:35 +0100)]
Clean up and sort the RAID sections in our configs.
Sascha Wildner [Wed, 8 Dec 2010 00:50:07 +0000 (01:50 +0100)]
kernel: Serial numbers have to be either 2 chars minimum or not be a blank.
The hptiop(4) controller's logical disks have a " " (single space)
serial number, which caused /dev/serno/_ to be created.
Sascha Wildner [Wed, 8 Dec 2010 00:28:05 +0000 (01:28 +0100)]
hptiop(4): Fix device description.
Sascha Wildner [Tue, 7 Dec 2010 23:54:45 +0000 (00:54 +0100)]
hptiop(4): Add the hptiop(4) RAID driver.
The driver supports the following adapters:
* HighPoint RocketRAID 3020
* HighPoint RocketRAID 3120
* HighPoint RocketRAID 3122
* HighPoint RocketRAID 3220
* HighPoint RocketRAID 3320
* HighPoint RocketRAID 3410
* HighPoint RocketRAID 3510
* HighPoint RocketRAID 3511
* HighPoint RocketRAID 3520
* HighPoint RocketRAID 3521
* HighPoint RocketRAID 3522
* HighPoint RocketRAID 3540
* HighPoint RocketRAID 4320
It was tested with the 3120.
Many thanks to FreeBSD, from which this code was obtained.
YONETANI Tomokazu [Tue, 7 Dec 2010 01:46:06 +0000 (10:46 +0900)]
periodic/security: fix warnings on a system with no packages installed
Nolan Lum [Tue, 7 Dec 2010 06:03:30 +0000 (06:03 +0000)]
devattr - a tool to expose the libdevattr functionality
Matthew Dillon [Tue, 7 Dec 2010 03:27:46 +0000 (19:27 -0800)]
kernel - Enable e1000 / igb netif in module build
* Add if_igb to the module build, via e1000.
Matthew Dillon [Tue, 7 Dec 2010 03:24:30 +0000 (19:24 -0800)]
kernel - 64-bit memory & ncpus work
* Create a macro and bsr function for cpu masks in preparation for
making cpumask_t 64 bits on 64-bit builds. cpumask_t is still
32 bits in this commit.
* Expand SMP_MAXCPUS to 31 for 64-bit builds in preparation for
eventually expanding it to 63.
* Fix an array overflow when probing ncpus on 64-bit.
* Increase the boot-time kmap entries to accomodate large numbers of
cpus.
* Clean up a few places where integer overflows could occur on
64-bit machines with large amounts of ram.
For the moment we are limited to 32G. This will change in a
followup commit.
Matthew Dillon [Mon, 6 Dec 2010 22:03:42 +0000 (14:03 -0800)]
kernel - Fix lapic mask for > 16 apic ids
* The lapic mask in the MP startup was only allowing for 16 apic ids,
adjust the mask to allow for all 256.
* Supermicro now boots w/12 cpus and probably 24. More work required
to get it to boot w/48.
Sascha Wildner [Mon, 6 Dec 2010 20:21:53 +0000 (21:21 +0100)]
amr(4): Return the correct error.
error was not being set for two functions, and if these failed, 0 was
incorrectly returned.
Sascha Wildner [Mon, 6 Dec 2010 18:12:30 +0000 (19:12 +0100)]
aac(4) & mfi(4): Fix masking bug.
When masking direct and processor devices during an inquiry, properly
preserve the upper bits of the first data byte.
Taken-from: FreeBSD
Sascha Wildner [Mon, 6 Dec 2010 17:57:55 +0000 (18:57 +0100)]
mfi(4): Shorten some lines.
Taken-from: FreeBSD
Samuel J. Greear [Sun, 5 Dec 2010 05:09:15 +0000 (05:09 +0000)]
kernel - Provide descriptions for many net.inet sysctl's
* Switch net.inet.ip.dispatch_* to RD from RW
Submitted-by: Taras Klaskovsky
Sponsored-by: Google Code-In
Venkatesh Srinivas [Sun, 5 Dec 2010 04:40:22 +0000 (20:40 -0800)]
kernel - mlockall: Return EINVAL for invalid flags combinations.
Reported by: swildner
Sascha Wildner [Sun, 5 Dec 2010 04:01:00 +0000 (05:01 +0100)]
Add a missing file for building libiconv into the kernel.
Sascha Wildner [Sun, 5 Dec 2010 03:43:36 +0000 (04:43 +0100)]
Make more things (HPFS, *_ICONV) buildable into the kernel.
Samuel J. Greear [Sun, 5 Dec 2010 03:24:28 +0000 (03:24 +0000)]
test - Add test case for prior select(2)/EOF bug
* This test reproduces the problem fixed in commit:
57b24f4ee66aaaa59b54e9577b93253cf435672f
Issue: http://bugs.dragonflybsd.org/issue1809
Submitted-by: Johannes Hofmann
Sascha Wildner [Sun, 5 Dec 2010 00:05:35 +0000 (01:05 +0100)]
Regenerate the pciconf(8) database
...from the following files:
Hart: Jan 22, 2008 (version 671)
Boemler: Dec 4, 2010
Mares: Nov 24, 2010
Sascha Wildner [Sat, 4 Dec 2010 21:00:49 +0000 (22:00 +0100)]
Fix a typo and add PMAP_DEBUG to LINT.
YONETANI Tomokazu [Sat, 4 Dec 2010 13:05:42 +0000 (22:05 +0900)]
kernel - remove extra crit_exit()
Those are leftover from
080c00e6fcafedf719b52e5de548a005201b0218 .
Sascha Wildner [Sat, 4 Dec 2010 02:20:29 +0000 (03:20 +0100)]
installer: Fix a bug when /boot (HAMMER) or / (UFS) is are to be encrypted.
The installer will warn that this is not possible, but previously left the
encryption flag set. Clear it.
Matthew Dillon [Fri, 3 Dec 2010 21:37:32 +0000 (13:37 -0800)]
docs - Expand the swapcache(8) manual page
* Add a description for static and dynamic wear leveling and write-combining
Sascha Wildner [Fri, 3 Dec 2010 19:18:23 +0000 (20:18 +0100)]
kernel: Remove some PC98 remains.
These were either not caught when we removed support, or they came back
in later.
Matthew Dillon [Fri, 3 Dec 2010 16:48:05 +0000 (08:48 -0800)]
kernel - Remove unneeded critical sections from VM code, add pmap asserts
* Various bits of VM code now only need vm_token and no longer need
a critical section.
* Add pmap_page_assertzero() assertions in a couple of places when
PMAP_DEBUG is enabled.
* Add PMAP_DEBUG as a global kernel config option.
* Be a bit more conservative and clear PG_ZERO for all page table
pages and not just terminal page table pages. When such pages
are finally freed they will be zero again due to the way the
pmap code works but don't make that assumption.
Matthew Dillon [Fri, 3 Dec 2010 16:38:01 +0000 (08:38 -0800)]
kernel - Document ordering requirement in lwkt_cpusync_remote2()
* Add a cpu_ccfence() to document necessary ordering but assume
that the atomic op following handles the actual requirement.
This is essentially a nop.
Nolan Lum [Wed, 1 Dec 2010 21:49:53 +0000 (16:49 -0500)]
Add sha256/512 support to libcrypt, change default
* Add sha256 and sha512 support to libcrypt.
* Change the default password hash to sha256.
* All old passwords will of course still works, but all new ones will be
created using sha256.
Sponsored-by: Google Code-In
Sascha Wildner [Fri, 3 Dec 2010 13:34:47 +0000 (14:34 +0100)]
amr(4): Sync with FreeBSD.
This adds support for various new devices (full list is in the manual
page), among other things.
Reported-by: Tomaz Borstnar <tomaz.borstnar@over.net>
Jonas Trollvik <jontro@gmail.com>
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue305>
Sascha Wildner [Fri, 3 Dec 2010 13:21:03 +0000 (14:21 +0100)]
mfi(4): Add missing dev_ops_remove_minor().
Matthew Dillon [Fri, 3 Dec 2010 02:11:29 +0000 (18:11 -0800)]
kernel - Fix bug in x86-64 version of bzeront()
* The x86-64 bzeront code was subtracting the wrong value from the count
when looping. Note that prior to this commit the code was not being
used anyway.
* Rearrange the way movnti support is detected and enable for x86-64.
Alex Hornung [Thu, 2 Dec 2010 11:03:11 +0000 (11:03 +0000)]
dm - undo my pointless moving of dm.h
* There was no need to move dm.h, we can just use <dev/disk/dm/dm.h>.
* And since we can do the above, we can also get rid of the strange
Makefile CFLAGS magic I added.
* Also undo the removal on upgrade of /usr/include/dev/disk/dm/dm.h and
instead remove sys/dm.h
Reported-by: Sascha Wildner
Sascha Wildner [Thu, 2 Dec 2010 11:50:33 +0000 (12:50 +0100)]
Remove <dev/disk/dm/dm.h> via make upgrade. It's <sys/dm.h> now.
Sascha Wildner [Thu, 2 Dec 2010 10:06:37 +0000 (11:06 +0100)]
style.9: Somehow, half of the usage() example was missing. Fix it.
Noticed-by: sjg
Samuel J. Greear [Thu, 2 Dec 2010 09:34:13 +0000 (09:34 +0000)]
kernel - Provide descriptions for many net.inet6 sysctl's
Submitted-by: Taras Klaskovsky
Sponsored-by: Google Code-In
Sascha Wildner [Wed, 1 Dec 2010 21:18:19 +0000 (22:18 +0100)]
mfi(4): Make mfi_linux.ko compile (maybe it even works).
I haven't tested this yet.
Sascha Wildner [Wed, 1 Dec 2010 19:51:56 +0000 (20:51 +0100)]
vga/vesa: Use hexdump() instead of cooking our own.
Sascha Wildner [Wed, 1 Dec 2010 19:04:36 +0000 (20:04 +0100)]
Add a hexdump(9) manual page (taken from FreeBSD).
Matthew Dillon [Wed, 1 Dec 2010 17:55:39 +0000 (09:55 -0800)]
kernel - Fix mbuf header exhaustion w/ degenerate unix domain socket case
* Increase the number of mbuf headers to accomodate a degenerate
fd-passing case with chains of unix domain sockets, one fd per
socket.
* This does not address other mbuf exhaustion issues which can crop
up in a user ddos attack due to the default number of fds a user
is allowed to own.
Reported-by: Venkatesh Srinivas <me@endeavour.zapto.org>
Matthew Dillon [Wed, 1 Dec 2010 17:53:08 +0000 (09:53 -0800)]
kernel - Fix deep stack recursion in unix domain socket garbage collection
* A chain of unix domain sockets with the previous fd passed into the
next one can cause a kernel stack overflow.
* Flatten the portion of the gc which would otherwise recurse by placing
the fp on a list instead of calling a nested closef(), then running
the closes at the top level.
Reported-by: Venkatesh Srinivas <me@endeavour.zapto.org>
Alex Hornung [Wed, 1 Dec 2010 15:25:06 +0000 (15:25 +0000)]
installer - encrypted root fix for new modules
* With the new modularized dm, dm_target_crypt will have to be loaded
explicitly if the module_path is not reachable in the moment the crypt
target is needed.
* Simply write dm_target_crypt_load="YES" to the installed loader.conf
so that there are no nasty surprises.
Reported-by: Sascha Wildner
Alex Hornung [Wed, 1 Dec 2010 10:17:55 +0000 (10:17 +0000)]
dm - Modularize it fully
* Move each of the targets (except error and zero) into a subdirectory
in targets/.
* Change the macro DM_TARGET_MODULE to also let the module depend on dm.
* Add a new macro DM_TARGET_BUILTIN, which must be used instead of
DM_TARGET_MODULE for targets that are compiled into the main dm.ko
instead of a separate module. Not doing this will prevent dm.ko from
unloading.
* Enable automatic loading of target modules if they aren't builtin or
loaded yet. Automatically loaded modules can be unloaded via kldunload
safely via a little hack (see source).
* NOTE: automatic loading doesn't always work. /boot must be mounted
already, so if you are planning on encrypting your swap or similar
stuff, you need to *EXPLICITLY* load dm_target_crypt.ko via the
loader.
* NOTE: PLEASE NOTE THE NOTE ABOVE!
* Fix a little issue with dm_target_crypt that would print "unloaded"
even if it wouldn't unload.
* Change the name of dm_target_stripe to dm_target_striped, since the
target is actually called "striped".
* Move dm.h to sys/sys/dm.h
* Get rid of files.dm, a leftover from the import
Sascha Wildner [Wed, 1 Dec 2010 08:29:48 +0000 (09:29 +0100)]
Unbreak buildworld.
Matthew Dillon [Wed, 1 Dec 2010 06:01:15 +0000 (22:01 -0800)]
kernel - Replace broken mp lock in bounce pages swi_vm() with spinlock
* Replace old non-MPSAFE crit section and mplock code with a spinlock
* x86-64 was not registering swi_vm(), add that code in.
Reported-by: Venkatesh Srinivas <me@endeavour.zapto.org>
Matthew Dillon [Wed, 1 Dec 2010 05:59:52 +0000 (21:59 -0800)]
kernel - Fix NFS race
* The nfsnode code could in very rare circumstances double-vput a vnode,
creating havoc.
Venkatesh Srinivas [Wed, 1 Dec 2010 03:39:50 +0000 (19:39 -0800)]
kernel - vm_mmap: Remove Advertising clause and DragonFly CVSID.
Sascha Wildner [Wed, 1 Dec 2010 02:44:50 +0000 (03:44 +0100)]
mmcsd(4): Enhance readability by using a variable for the buf pointer.
Sascha Wildner [Wed, 1 Dec 2010 01:44:29 +0000 (02:44 +0100)]
hammer.8: Fix typo.
Sascha Wildner [Wed, 1 Dec 2010 01:28:24 +0000 (02:28 +0100)]
mfi(4): Add device statistics.
Stathis Kamperis [Tue, 30 Nov 2010 16:31:35 +0000 (18:31 +0200)]
condvar - Constify
Sascha Wildner [Tue, 30 Nov 2010 18:27:34 +0000 (19:27 +0100)]
mfi(4): Remove the unused MFI_DECODE_LOG option.
Sascha Wildner [Tue, 30 Nov 2010 16:49:07 +0000 (17:49 +0100)]
mfi(4): Add the mfi(4) RAID driver and the mfiutil(8) configuration tool.
The driver supports the following adapters:
* LSI MegaRAID SAS 1078
* LSI MegaRAID SAS 8408E
* LSI MegaRAID SAS 8480E
* LSI MegaRAID SAS 9260
* Dell PERC5
* Dell PERC6
* IBM ServeRAID M5015 SAS/SATA
* IBM ServeRAID-MR10i
* Intel RAID Controller SROMBSAS18E
It was tested with the LSI MegaRAID SAS 8408E card.
Many thanks to FreeBSD, from which this code was obtained.
Sascha Wildner [Mon, 29 Nov 2010 02:38:14 +0000 (03:38 +0100)]
kernel: Bring in hexdump() from FreeBSD.
Sascha Wildner [Tue, 30 Nov 2010 15:28:23 +0000 (16:28 +0100)]
mlockall.2: Fix month name and remove trailing whitespace.
Samuel J. Greear [Tue, 30 Nov 2010 05:01:38 +0000 (05:01 +0000)]
Merge branch 'GCI2010'
Samuel J. Greear [Tue, 30 Nov 2010 04:42:32 +0000 (04:42 +0000)]
kernel - Provide descriptions for lwkt.* and debug.* sysctl's
* Take the opportunity to chop a few dead sysctl's as well
Submitted-by: Taras Klaskovsky
Sponsored-by: Google Code-In
Joe Talbott [Tue, 30 Nov 2010 02:11:17 +0000 (21:11 -0500)]
Merge branch 'master' of git://crater.dragonflybsd.org/dragonfly
Joe Talbott [Tue, 30 Nov 2010 02:03:20 +0000 (21:03 -0500)]
unp_connect() - Use euid/egid for access checks via VOP_EACCESS().
Fixes socket access for set[ug]id applications by using the euid/egid
rather than the uid/gid.
Reported-By: Tim Darby <t+dfbsd@timdarby.net>
Matthew Dillon [Mon, 29 Nov 2010 19:38:33 +0000 (11:38 -0800)]
kernel - Run pfil hooks for bridged packets
* The pfil hooks were not being run in all cases for bridged packets.
Submitted-by: Rumko <rumcic@gmail.com>
Venkatesh Srinivas [Mon, 29 Nov 2010 19:23:38 +0000 (11:23 -0800)]
kernel - tmpfs - Remove mplock in tmpfs_read().
Venkatesh Srinivas [Mon, 29 Nov 2010 16:46:27 +0000 (08:46 -0800)]
Import mlockall(2)/munlockall(2) manpage from OpenBSD and modify for Dfly.
Alex Hornung [Mon, 29 Nov 2010 08:02:08 +0000 (08:02 +0000)]
dm - access io_cnt atomically
* Now that we are using shared locks at some points, only update the
io_cnt atomically, otherwise we can mess things up.
Alex Hornung [Fri, 26 Nov 2010 14:23:12 +0000 (15:23 +0100)]
dm - misc minor cleanup
* remove dead code, weird white spaces and fix comments.
Alex Hornung [Fri, 26 Nov 2010 14:08:33 +0000 (15:08 +0100)]
dm - remove convoluted string comparison methods
Alex Hornung [Fri, 26 Nov 2010 14:01:33 +0000 (15:01 +0100)]
dm - minor cleanup
* Fix a bit of indenting, add a comment.
* Remove useless and pointless (just as redundant as using both of those
words) code.
Alex Hornung [Fri, 26 Nov 2010 13:42:12 +0000 (14:42 +0100)]
dm - use shared/exclusive lockmgr for table access
* Use shared/exclusive lockmgr locks for table access instead of this
random mess of exclusive lockmgr locks, reference counts and condvars.
Alex Hornung [Fri, 26 Nov 2010 13:28:38 +0000 (14:28 +0100)]
dm - properly encapsulate disable_dev
Alex Hornung [Fri, 26 Nov 2010 13:27:48 +0000 (14:27 +0100)]
condvar - add cv_has_waiters()
Samuel J. Greear [Sun, 28 Nov 2010 07:14:55 +0000 (07:14 +0000)]
nmalloc - Remove startup locking
* A spinlock is no longer needed for malloc_init, it is run as a constructor
now, as such it is run before any other threads are started.
Samuel J. Greear [Sun, 28 Nov 2010 07:10:27 +0000 (07:10 +0000)]
cdefs - Use proper GCC version
Matthew Dillon [Sun, 28 Nov 2010 06:43:00 +0000 (22:43 -0800)]
kernel - Document an odd case in vm_page.c
* Document cases from the buffer cache where a vm_page's clean/dirty bits
might be modified without busying the related vm_page. vm_token is still
held through the operation.
Matthew Dillon [Sun, 28 Nov 2010 06:40:13 +0000 (22:40 -0800)]
HAMMER VFS - Fix HAMMER_INODE_XDIRTY flags / state assertion
* The HAMMER_INODE_XDIRTY flag was not being immediately synchronized
with the state of ip->rec_tree in one case and the delay could create
a window of opportunity where an assertion would get hit.
* Remove the window of opportunity, the assertion no longer gets hit.
Reported-by: YONETANI Tomokazu <qhwt.dfly@les.ath.cx>
Matthew Dillon [Sun, 28 Nov 2010 06:34:56 +0000 (22:34 -0800)]
kernel - x86-64 - normalize the location of the pcb/md_regs and remove gd_rsp0
* Normalize the location of the pcb and related user process register
frame for traps and system calls.
* The pcb is now 16-byte aligned at the top of the kernel thread's stack
and md_regs is placed just below it.
* We no longer reserve 16 bytes for a 8086 emulation frame.
* Change the md_regs assignment in the trap code to an assertion that the
value is correct.
* Remove gd_rsp0, use the value from common_tss.tss_rsp0 instead (which
is the same).
Matthew Dillon [Sun, 28 Nov 2010 06:28:52 +0000 (22:28 -0800)]
kernel - Hold vm_token() for certain buf/bio/vm_page interactions
* Hold the vm_token when messing around with page valid and dirty bits
in buffer cache operations. Such race conditions are possible but would
be very rare.
* Fix inconsistent use of bio2.bio_offset when swapcache intercepts an
I/O operation. When this case occurs the normal filesystem VOP_STRATEGY
function will not be called and any assumption that bio2.bio_offset
will represent a block device offset will be invalid.
To deal with this case swapcache will re-set bio_offset to NOOFFSET.
Filesystems use this value to determine that no block device offset
has been cached for a logical file block.
Reported-by: Jan Lentfer <Jan.Lentfer@web.de> (bio_offset issue)
Matthew Dillon [Sun, 28 Nov 2010 06:27:10 +0000 (22:27 -0800)]
kernel - Fix MP race in signotify_remote()
* In rare cases it is possible for a process to hop multiple cpus while
a signal is trying to chase the process. Adjust signotify_remote() to
deal with the case.
Matthew Dillon [Sun, 28 Nov 2010 06:20:04 +0000 (22:20 -0800)]
kernel - Fix NFS stall and likely also x86-64 seg-fault issue
* The nfs node hash code was not MPSAFE due to its use of a global
hash table. This could lead to a stall condition due to the
global hash table losing track of its manual lock. For now use a
token to interlock hash table access and replace the manual lock
with a lockmgr lock.
* This appears to have also fixed the long-standing random seg-fault
issue with x86-64, where a buildworld loop would seg-fault every once
in a while for no apparent reason. /usr/src on my test box has always
been NFS mounted but was never considered a possible contributor to
the problem.
If NFS loses track of related nfs nodes or races operations on underlying
vnodes while expecting their storage to remain stable it is possible
for a random vnode to become corrupt. How this could translate to a
completely random seg-fault is not entirely understood but I surmise
the unrelated cc1's binary mmap could become corrupt.
Reported-by: Thomas Nikolajsen <thomas.nikolajsen@mail.dk> (nfs issue)
Samuel J. Greear [Sun, 28 Nov 2010 02:17:56 +0000 (02:17 +0000)]
kernel - Provide descriptions for vm.stats.* sysctl's
Submitted-by: Taras Klaskovsky
Sponsored-by: Google Code-In
Samuel J. Greear [Sat, 27 Nov 2010 10:12:06 +0000 (10:12 +0000)]
nmalloc - Optimize initialization function out of hot path
* Add a __constructor macro to cdefs.h
* Turn malloc_init into a gcc constructor
* Modify ifconfig to use the cdefs macro instead of its own
Alex Hornung [Thu, 25 Nov 2010 11:38:22 +0000 (11:38 +0000)]
dmirror - Add Matt's design notes
Alex Hornung [Fri, 23 Jul 2010 19:54:11 +0000 (20:54 +0100)]
dmirror - initial code (layout)
* Initial code laying out basic functionality and design except for
resynchronization.
* NOTE: this code is faaaaaaaaaaaar away from being usable. It really is
only a layout.
* NOTE: did I mention that this code is far away from being usable and
still needs a lot of love? The only reason I'm committing this, is so
it doesn't linger forever on my personal repo.
* If anyone feels like it, *PLEASE* work on dmirror!!!!
Alex Hornung [Thu, 25 Nov 2010 10:33:24 +0000 (10:33 +0000)]
dm - Major cleanup & module work
* Add module code to all the targets and get rid of the hackish global
target insertion.
* Clean up all sorts of dead code and crappy whitespaces. There isn't
much point in maintaining whitespaces just to have a small diff to
NetBSD since our diff is quite big now anyways.
Alex Hornung [Thu, 25 Nov 2010 08:18:15 +0000 (08:18 +0000)]
mmap - add mmap offset randomization
* mmap offset randomization, based upon OpenBSD's implementation in
sys/uvm/uvm_map.c and sys/uvm/uvm_mmap.c.
* The randomization is controllable via the sysctl vm.randomize_mmap,
which by default is 0 (off).
Obtained-from: OpenBSD