Hiten Pandya [Sun, 6 Jun 2004 13:30:33 +0000 (13:30 +0000)]
Do not use the installed include files, instead, set the include path
to ${.CURDIR}/../../include; so that the right version is used.
Hiten Pandya [Sun, 6 Jun 2004 11:49:54 +0000 (11:49 +0000)]
Now that we have clients that use m_getcl(9), set the default mcl_pool_max
pool size for mbuf+clusters to '10' instead of '0', to improve performace.
Obtained-from: FreeBSD, rev. 1.51.2.17 (RELENG_4)
Hiten Pandya [Sun, 6 Jun 2004 10:58:31 +0000 (10:58 +0000)]
Discard the first 1024 bytes of output as suggested by
http://citeseer.nj.nec.com/fluhrer01weaknesses.html and
http://citeseer.nj.nec.com/531224.html.
Obtained-from: FreeBSD rev. 1.7 src/lib/libc/gen/arc4random.c
Jeffrey Hsu [Sun, 6 Jun 2004 05:59:44 +0000 (05:59 +0000)]
Close race condition in accept(2).
Chris Pressey [Sun, 6 Jun 2004 05:48:01 +0000 (05:48 +0000)]
Fix typo in error message - there is no system call named "lchflags."
Jeffrey Hsu [Sun, 6 Jun 2004 05:38:58 +0000 (05:38 +0000)]
Detect and foil optimistic ACK attack with forced slow-start
as suggested by Azcorra, Bernardos, and Soto.
Matthew Dillon [Sat, 5 Jun 2004 19:57:40 +0000 (19:57 +0000)]
Add the MSFBUF API. MSFBUFs are like SFBUFs but they manage ephermal
multi-page mappings instead of single-page mappings. MSFBUFs have the
same caching and page invalidation optimizations that SFBUFs have and are
considered to be SMP-friendly.
Whereas XIO manages pure page lists, MSFBUFs manage KVA mappings of pure
page lists.
This initial commit just gets the basic API operational. The roadmap for
future work includes things like better interactions with third-party XIOs,
mapping user buffers into the kernel (extending the xio_init_ubuf() API into
the MSFBUF API), and allowing higher level subsystems to pass previously
released MSFBUFs as a hint to speed-up regeneration. We also need to come
up with a way to overload additional sets of MSFBUFs representing smaller
chunks of memory on top of the same KVA space in order to efficiently use
our KVA reservation when dealing with subsystems like the buffer cache.
MSFBUFs will eventually replace the KVA management in the BUF/BIO, PIPE,
and other subsystems which create fake linear mappings with pbufs. The
general idea for BUF/BIO will be to use XIO and MSFBUFs to avoid KVA
mapping file data through the nominal I/O path. XIO will be the primary I/O
buffer mechanism while MSFBUFs will be used when things like UFS decide they
need a temporary mapping.
This is a collaborative work between Hiten Pandya <hmp@leaf.dragonflybsd.org>
and Matthew Dillon <dillon@backplane.com>.
Matthew Dillon [Sat, 5 Jun 2004 18:04:47 +0000 (18:04 +0000)]
M_NOWAIT is just wrong in the init code. The allocation must succeed.
Use M_WAITOK.
Matthew Dillon [Sat, 5 Jun 2004 16:34:07 +0000 (16:34 +0000)]
Rearrange the kern_getcwd() procedure to return the base of the string
rather then relocating the string.
Also fix two bugs: (1) the original bcopy was copying data beyond the end of
the buffer ([bp, bp+buflen] exceeds the buffer), and (2), the uap->buflen
checks must be made in __getcwd(), before the kernel tries to malloc() space.
Hiten Pandya [Sat, 5 Jun 2004 15:02:05 +0000 (15:02 +0000)]
Add a manual page describing the LSI Fusion family of devices with a
few changes by me.
Requested-by: David Rhodus <drhodus@machdep.com>
Joerg Sonnenberger [Sat, 5 Jun 2004 13:52:29 +0000 (13:52 +0000)]
em(4) assumes that bus_dmamap_destroy of bus_dmamap_load_mbuf maps
unconditionally succeds w/o previous bus_dmamap_unload. This assumption
is currently true because bus_dmamap_load_mbuf doesn't support bounce
pages. Nevertheless is that an implementation detail which em(4) should
not depend on.
Discussed-with: Hiten Pandya <hmp@backplane.com>
Joerg Sonnenberger [Sat, 5 Jun 2004 13:09:00 +0000 (13:09 +0000)]
Fix a small bug in the last commit. ether_ifdetach has to be called
explicitly if bus_setup_intr fails.
Eirik Nygaard [Sat, 5 Jun 2004 10:27:58 +0000 (10:27 +0000)]
Swap order of first and second argument in bcopy, forgot this when changing
from strncpy.
Eirik Nygaard [Sat, 5 Jun 2004 10:04:38 +0000 (10:04 +0000)]
Use MAXPATHLEN as the size of the checkpoint filename sysctl variable instead
of a hardcoded value.
Eirik Nygaard [Sat, 5 Jun 2004 10:02:07 +0000 (10:02 +0000)]
Print the entire path to where the checkpoint file is being created to avoid
confusion if the application changes it's working dir.
Pointers by: Matthew Dillon <dillon@apollo.backplane.com>
Eirik Nygaard [Sat, 5 Jun 2004 09:58:15 +0000 (09:58 +0000)]
Split the __getcwd syscall into a kernel and an userland part, so it can be
used in the kernel as well.
Pointers by: Matthew Dillon <dillon@apollo.backplane.com>
Matthew Dillon [Sat, 5 Jun 2004 05:34:57 +0000 (05:34 +0000)]
Allow dup_sockaddr() to block, otherwise the code becomes non-deterministic.
Since we are a thread now this should be fairly safe.
Chris Pressey [Sat, 5 Jun 2004 03:23:14 +0000 (03:23 +0000)]
Fix typo.
Reported-by: Brad Harvell
Matthew Dillon [Fri, 4 Jun 2004 20:35:39 +0000 (20:35 +0000)]
async syscall work: The async syscall code got dated by recent LWKT
changes, set mp_abort_port and clear MSGF_DONE as appropriate.
If a system call returns EASYNC, record the message in p->p_sysmsgq so we
can run them down in exit1().
In exit1(), run down any asynch system calls that are still running. Note
that this commit does not implement abort support (yet).
Get rid of lwkt_port->mp_refs, it was not being used and it is likely never
going to be used (reference counting LWKT is hazzardous anyway since it
doesn't really fit the access model).
Add lwkt_checkmsg() to support some of the rearranged async syscall code.
Matthew Dillon [Fri, 4 Jun 2004 20:29:02 +0000 (20:29 +0000)]
Remove dtom(). dtom() is no longer supported (precursor requirement for
eventual mbuf allocator replacement).
Matthew Dillon [Fri, 4 Jun 2004 20:27:32 +0000 (20:27 +0000)]
Use a normal malloc() for PCB allocations instead of the really aweful mbuf
hacks that were previously being used (which also gets rid of the remaining
dtom() calls).
Get rid of the packet/queue overloading and instead malloc() a queue header
for queueing packets. The old code depended on dtom() to figure out the
mbuf associated with the packet and we are removing dtom().
Give NS a default usrreq so it doesn't panic the system on boot, but note
that NS will remain broken until the usrreq can be tied into the existing
code.
Review and testing by: Hiten Pandya and David Rhodus
Hiten Pandya [Fri, 4 Jun 2004 17:30:26 +0000 (17:30 +0000)]
Re-arrange the 'Executable' column for the '-l' option so that long
pathnames are printed correctly.
Suggested-by: Matthew Dillon <dillon at backplane.com>
Joerg Sonnenberger [Fri, 4 Jun 2004 16:32:11 +0000 (16:32 +0000)]
Continue cleaning em(4).
Merge PCI resource allocation / deallocation into em_attach and em_detach.
Change the order of allocations in em_attach to work without splimp()
protection.
Remove adapter->unit, adapter->next and adapter->prev.
Use em_detach for the error handling in em_attach as well.
Move the reset of if_timer into em_stop.
Use adapter->hw.mac_addr instead of arpcom's entry.
Hiten Pandya [Fri, 4 Jun 2004 11:21:53 +0000 (11:21 +0000)]
Allow top(1) to toggle display of processes+threads or *only* threads
by adding a '-O' command line argument and run-time toggle.
This is useful when one wants to only monitor activity of current
active threads.
Document this new functionality and update the patches accordingly.
Requested-by: David Rhodus <drhodus@machdep.com>
Hiten Pandya [Fri, 4 Jun 2004 07:45:46 +0000 (07:45 +0000)]
Deprecate use of m_act, which is an alias of m_nextpkt; just use
m_nextpkt instead.
Matthew Dillon [Fri, 4 Jun 2004 06:56:12 +0000 (06:56 +0000)]
Add MPF_INT (for mpipe_init()), which allows one to specify that an MPIPE
allocation request may use the interrupt memory reserve.
Hiten Pandya [Fri, 4 Jun 2004 05:06:40 +0000 (05:06 +0000)]
Flush cached access mode after modifying a files attributes for NFSv3.
It's likely that modifying the attiributes will affect the file's
accessibility
Fix created by Ian Dowse and Peter Edwards of the FreeBSD Project.
Test case available in PR# 44336 in FreeBSD GNATS.
Matthew Dillon [Fri, 4 Jun 2004 04:59:57 +0000 (04:59 +0000)]
Use MPIPE instead of misusing m_getclr() (i.e. m_get) for NETNS's
struct nspcb. This also gets rid of a dtom() call.
Note: the changes compile but are untested.
Matthew Dillon [Fri, 4 Jun 2004 04:32:23 +0000 (04:32 +0000)]
Use MPIPE instead of the really hackish use of m_get() and mtod()/dtom()
for the 'tcptemp' structure.
Matthew Dillon [Fri, 4 Jun 2004 04:30:55 +0000 (04:30 +0000)]
Remove dtom() in unused (#ifndef notdef'd) code. Well, just remove the whole
of the related #ifdef notdef'd sections.
Matthew Dillon [Fri, 4 Jun 2004 04:25:07 +0000 (04:25 +0000)]
TEST
Matthew Dillon [Fri, 4 Jun 2004 03:57:41 +0000 (03:57 +0000)]
Use MPIPE instead of (the really aweful improper use of) mbuf_alloc() for
struct ipq. This also gets rid of a nasty dtoc().
Matthew Dillon [Fri, 4 Jun 2004 01:46:49 +0000 (01:46 +0000)]
ANSIfication. No functional changes.
Matthew Dillon [Thu, 3 Jun 2004 21:03:32 +0000 (21:03 +0000)]
ms_cmd has changed to a union, update the test code.
Hiten Pandya [Thu, 3 Jun 2004 19:26:36 +0000 (19:26 +0000)]
Add a manual page which describes the vn_fullpath(9) function.
Adapted from the manual page in the FreeBSD CVS repository.
Joerg Sonnenberger [Thu, 3 Jun 2004 18:30:04 +0000 (18:30 +0000)]
Make pr_input use variadic arguments for anything but the first mbuf.
Add a warning for NSIP, it is likely to be broken.
Hiten Pandya [Thu, 3 Jun 2004 18:17:43 +0000 (18:17 +0000)]
Lock manipulation of the 'exec_res_list', i.e., the list of resident
executables.
Requested-by:
Joerg Sonnenberger <joerg@bec.de>, Matthew Dillon <dillon@backplane.com>
Hiten Pandya [Thu, 3 Jun 2004 18:12:41 +0000 (18:12 +0000)]
Surround a multi-line conditional block with braces for readability.
Suggested-by: Matthew Dillon <dillon at backplane.com>
Hiten Pandya [Thu, 3 Jun 2004 18:09:33 +0000 (18:09 +0000)]
Remove the compat macro textvp_fullpath(), and use vn_fullpath()
directly.
Discussed-with: Matthew Dillon <dillon at backplane.com>
Hiten Pandya [Thu, 3 Jun 2004 16:37:05 +0000 (16:37 +0000)]
Cleanup the manual page:
* add an AUTHORS section.
* use the '.Pa' command to markup a filename.
* remove hard sentence breaks and an empty section command.
Hiten Pandya [Thu, 3 Jun 2004 16:28:15 +0000 (16:28 +0000)]
VM Resident Executables update:
Kernel Part:
* decrement the exec_res_id counter in exec_sys_unregister, otherwise
we will have an incorrect count of currently resident executables.
* add a structure called 'xresident' which stores enough information
about resident executables, so that we can export it to userland
via sysctl; the structure resides in sys/resident.h.
* add a sysctl node, called 'vm.resident' which enumerates though
exec_res_list, i.e., the queue which stores information about the
executables that are currently resident
* move the exec_res_list queue initialization and a MALLOC_DEFINE
near the top of the file, right after the include files; and also
initialize the exec_res_id counter to '0'.
NOTE! The 'vm.resident' sysctl node is only available to root!
Userland Part:
* add a '-l' option to resident(8), so that it displays the list
of executables that are currently memory-resident.
* update the program usage string so that it is in par with the
functionality offered by the program.
* update the manual page.
Requested-by: many...
Discussed-with: Matthew Dillon <dillon@backplane.com>
Joerg Sonnenberger [Thu, 3 Jun 2004 16:21:22 +0000 (16:21 +0000)]
Make pr_domain and pr_usrreqs pointers to const. The general stack is not
supposed to change this and it allows us to make all this structures
const like they should be. If a network protocol wants to modify them,
they can still either modify the private version directly or create a copy
and change that.
Hiten Pandya [Thu, 3 Jun 2004 15:54:14 +0000 (15:54 +0000)]
Cleanup the textvp_fullpath() function; summary of changes:
* rename it to vn_fullpath(), and move prototype to sys/vnode.h;
for now, keep a compat macro called textvp_fullpath(); this
macro is temporary until I change linprocfs and procfs to play
with vn_fullpath().
* rename last argument to 'freebuf', instead of 'retfreebuf'.
* if the second argument (vnode pointer) is NULL, fall back
and use the passed-in process' p->p_textvp.
* make sure the client passes a process pointer, otherwise
return EINVAL.
A manual page will be added real soon.
Hiten Pandya [Thu, 3 Jun 2004 15:40:22 +0000 (15:40 +0000)]
Quotactl(2) should set the uid correctly, based on the QUOTA type supplied.
Previously, it would set the uid to p->p_ucred->cr_ruid unconditionally
without checking if the QUOTA type is GRPQUOTA or USRQUOTA; thus allowing
a user with UID X to access the groupquota of gid X.
Inspired-by: PR# 33940 (by Vladimir B. Grebenschikov) [FreeBSD GNATS]
Hiten Pandya [Thu, 3 Jun 2004 15:29:43 +0000 (15:29 +0000)]
Remove some long gone #defines, PFCLUSTER_BEHIND and PFCLUSTER_AHEAD;
This change was inspired by a similar change made in FreeBSD's HEAD
a few months ago by:
Alan Cox <alc at freebsd.org>
No operational changes.
Joerg Sonnenberger [Thu, 3 Jun 2004 15:04:52 +0000 (15:04 +0000)]
Change pr_output's signature to take two fixed arguments and possible
additional varidic arguments. Change the various protocols accordingly.
This is not the signature NetBSD uses (only struct mbuf * + variadic),
but since all protocols use the socket argument (beside NS), it is more
reasonable to keep that static as well.
While at it, fix the pr_soport in ip6protosw to match the protosw.
Joerg Sonnenberger [Thu, 3 Jun 2004 13:34:37 +0000 (13:34 +0000)]
Move extern declaration to file scope to fix warning
Joerg Sonnenberger [Thu, 3 Jun 2004 13:30:26 +0000 (13:30 +0000)]
Don't cast away the const before dereferencing.
Joerg Sonnenberger [Thu, 3 Jun 2004 13:12:24 +0000 (13:12 +0000)]
Hide unused functions to silence GCC
Joerg Sonnenberger [Thu, 3 Jun 2004 13:09:07 +0000 (13:09 +0000)]
Hide unused function under #ifdef SMP
Eirik Nygaard [Thu, 3 Jun 2004 10:00:06 +0000 (10:00 +0000)]
Add ref counting to the checkpoint handler and unload function so an unload
will not happen when a program is being checkpointed.
Matthew Dillon [Wed, 2 Jun 2004 19:31:02 +0000 (19:31 +0000)]
Clean up some misuses of bp->b_dev after a strategy function has completed
(the field cannot be used after biodone() has been called). Add a separate
dev_t argument to diskerr() to take care of the issue and get rid of some
FD error reporting hacks at the same time.
Reported-by: David Rhodus
Matthew Dillon [Wed, 2 Jun 2004 17:18:43 +0000 (17:18 +0000)]
ANSIfication and cleanup. No functional changes.
Submitted-by: Tim Wickberg <me@k9mach3.org>
Jeffrey Hsu [Wed, 2 Jun 2004 16:41:28 +0000 (16:41 +0000)]
Allow an inp control block to be inserted on multiple wildcard hash tables.
Hiten Pandya [Wed, 2 Jun 2004 15:25:01 +0000 (15:25 +0000)]
Plug a memory leak when the kernel initialiazes config_devtab resources
in resource_new_name().
PR: kern/33344 (FreeBSD GNATS)
Patch-by: David Xu <davidxu at freebsd.org>
Eirik Nygaard [Wed, 2 Jun 2004 14:43:04 +0000 (14:43 +0000)]
Change mbug allocation flags from M_ to MB_ to avoid confusion with malloc
flags.
Requested by: Jeffrey Hsu
Joerg Sonnenberger [Wed, 2 Jun 2004 14:19:11 +0000 (14:19 +0000)]
Include ns.h to get prototyp for ns_cksum
Joerg Sonnenberger [Wed, 2 Jun 2004 14:00:12 +0000 (14:00 +0000)]
Improve the way error message from ALART are printed.
Joerg Sonnenberger [Wed, 2 Jun 2004 13:56:34 +0000 (13:56 +0000)]
set_lapic_isrloc depends on APIC_IO for the prototype to exist, it isn't
used otherwise.
Joerg Sonnenberger [Wed, 2 Jun 2004 13:49:46 +0000 (13:49 +0000)]
Use volatile and __DEVOLATILE to silence gcc warnings.
Joerg Sonnenberger [Wed, 2 Jun 2004 11:46:11 +0000 (11:46 +0000)]
Use __DECONST to silence GCC. Make some private routines static.
Hiten Pandya [Wed, 2 Jun 2004 07:34:29 +0000 (07:34 +0000)]
Merge from FreeBSD, RELENG_4 branch, revision 1.250.2.26.
--- original commit message ---
Log:
There is a comma missing in the table initializing the
pmap_prefault_pageorder array. This has two effects:
1. The resulting bogus contents of the array thwarts part of
the optimization effect pmap_prefault() is supposed to have.
2. The resulting array is only 7 elements long (auto-sized), while
pmap_prefault() expects it to be the intended 8 elements. So
this function in fact accesses memory beyond the end of the array.
Fortunately though, if the data at this location is out of bounds
it will be ignored.
This bug dates back more than 6 years. It has been introduced
in revision 1.178.
Submitted by: Uwe Doering <gemini@geminix.org>
PR: 67460
--- original commit message ---
Matthew Dillon [Tue, 1 Jun 2004 22:24:10 +0000 (22:24 +0000)]
VM86 calls on some BIOSs, apparently mainly VESA calls, use 8254 timers that
we are using. Restore our timers after making a vm86 call. This is a hack,
but it seems to work.
Testing by: Sascha Wildner <saw@online.de>
Matthew Dillon [Tue, 1 Jun 2004 22:19:30 +0000 (22:19 +0000)]
ANSIfication. No operational changes.
Submitted-by: Tim Wickberg <me@k9mach3.org>
Matthew Dillon [Tue, 1 Jun 2004 20:49:08 +0000 (20:49 +0000)]
Add a pfil_has_hooks() inline to shortcut calls to pfil_run_hooks(),
allowing us to include PFIL_HOOKS without introducing any performance impact
when no hooks are being used. This is in prep for making PFIL_HOOKS
permanent.
Matthew Dillon [Tue, 1 Jun 2004 20:34:01 +0000 (20:34 +0000)]
Add RLIMIT_POSIXLOCKS support to csh/tcsh.
Joerg Sonnenberger [Tue, 1 Jun 2004 17:35:58 +0000 (17:35 +0000)]
The const call is linted, use proper cast to silence GCC
Joerg Sonnenberger [Tue, 1 Jun 2004 17:35:03 +0000 (17:35 +0000)]
Fix some const warnings
Joerg Sonnenberger [Tue, 1 Jun 2004 17:31:17 +0000 (17:31 +0000)]
Initialize the magic cookie using real numbers instead of a multi-character
literal.
Joerg Sonnenberger [Tue, 1 Jun 2004 17:30:30 +0000 (17:30 +0000)]
Explicitly cast-away volatile since it should be save here.
Joerg Sonnenberger [Tue, 1 Jun 2004 17:27:44 +0000 (17:27 +0000)]
Explicitly cast away the volatile for conversions and argument passings.
Deploy __DEVOLATILE.
Joerg Sonnenberger [Tue, 1 Jun 2004 17:12:15 +0000 (17:12 +0000)]
Use struct thread for kernel threads, not struct proc.
Hiten Pandya [Tue, 1 Jun 2004 14:00:50 +0000 (14:00 +0000)]
Correct spelling.
Submitted-by: Tim Wickberg <me at k9mach3.org>
Hiten Pandya [Tue, 1 Jun 2004 11:36:53 +0000 (11:36 +0000)]
Remove '-*- nroff -*-'.
Lets make a habit not to put editor-related magic into our base
files. They are redundant and useless.
Chris Pressey [Tue, 1 Jun 2004 01:15:59 +0000 (01:15 +0000)]
Style(9) cleanup:
- Remove `register' keywords.
- Single space after `switch'.
- No functional changes.
Jeffrey Hsu [Mon, 31 May 2004 23:53:14 +0000 (23:53 +0000)]
Panic in udp_output() if a socket is found in an inconsistent state.
Eirik Nygaard [Mon, 31 May 2004 20:48:41 +0000 (20:48 +0000)]
Remove unused error variable.
Submitted-by: Dheeraj Reddy <dheerajs@comcast.net>
Matthew Dillon [Mon, 31 May 2004 17:52:30 +0000 (17:52 +0000)]
Since RNDC does not yet support 'restart', use RCNG to restart named
instead of rolling our own.
Suggested-by: Brad Harvell <dfbrad@whatzit.org>
Matthew Dillon [Mon, 31 May 2004 17:51:47 +0000 (17:51 +0000)]
Fix the path to named's pid file.
Submitted-by: Brad Harvell <dfbrad@whatzit.org>
Matthew Dillon [Mon, 31 May 2004 17:50:39 +0000 (17:50 +0000)]
When doing a restart, sleep for 0.1 seconds after the kill to avoid racing
the kill in wait_for_pids() (which imposes a 2 second delay when the race
is lost).
Matthew Dillon [Mon, 31 May 2004 17:38:08 +0000 (17:38 +0000)]
In the root fs search use the correct unit number when checking for
device existinance. The new device infrastructure often registers disk
unist separately, so looking up unit 0 when only unit N exists will not
succeed. We might want to later augment this with the slice and partition
number (which are 0 at the time the search is made prior to the real lookup),
but this should work for now.
Submitted-by: Hiroki Sato <hrs@allbsd.org>
Matthew Dillon [Mon, 31 May 2004 17:06:41 +0000 (17:06 +0000)]
Two unused arguments were recently removed from pmap_init() without
fixing the call or the prototype. Fix those.
Reported-by: walt <wa1ter@myrealbox.com>
Hiten Pandya [Mon, 31 May 2004 14:59:34 +0000 (14:59 +0000)]
Do not use the display function if the -o (opaque) or -x (hexdump)
arguments were provided to sysctl(8).
The display function is used when a sysctl node is registered as
`special' in sysctl(8), so instead of showing the normal "var:val"
output, it would format the values instead. E.g. vm.loadavg.
Hiten Pandya [Mon, 31 May 2004 11:43:49 +0000 (11:43 +0000)]
Remove an unimplemented advisory function, pmap_pageable(); there is
no pmap implementation in existance that requires it implemented.
Discussed-with: Alan Cox <alc at freebsd.org>,
Matthew Dillon <dillon at backplane.com>
Hiten Pandya [Mon, 31 May 2004 01:51:23 +0000 (01:51 +0000)]
Remove a stale comment: PG_DIRTY and PG_FILLED were removed in
revisions 1.17 and 1.12 respectively.
Noticed-by: Alan Cox <alc at freebsd.org>
Hiten Pandya [Sat, 29 May 2004 09:53:58 +0000 (09:53 +0000)]
Force commit to clarify that the previous revision should have been
credited to:
YONETANI Tomokazu <qhwt+dragonfly-commit at les.ath.cx>
Hiten Pandya [Sat, 29 May 2004 09:51:43 +0000 (09:51 +0000)]
Adjust include path so that the machine.h in ${.OBJDIR} is used
instead of the one in src/contrib/top.
Hiten Pandya [Sat, 29 May 2004 05:11:15 +0000 (05:11 +0000)]
First pass at updating top(1):
* add command line option '-T' for displaying kernel threads
such as pagedaemon, tcp_thread, and so on; the threads are
now marked '<thread>' in the top(1) display.
* cleanup manipulation of COMMAND strings, insipired by a
similar change in NetBSD's top(1).
* add a toggle for system processes, called 'S' so we dont
need to restart top(1) to {un}display system processes.
* change some sprintf's to snprintfs for safety.
* update the sort/ordering top(1) commands so that it correctly
displays kernel threads by priority time used etc.
* add a 'thr' sort order function so that kernel threads can be
collected together in the top(1) display; this is useful when
running it in batch mode.
* fix the screen width issue for terminals that have more than
80 columns; this was necessary for showing threads properly in
SMP mode.
* add ability for top(1) to sort kernel threads their priority
(note, thread priorities are stored differently).
NOTE! top(1) is now using the contrib-patching framework just like
we do with GCC-3 so that it will be simple to update the base code
in the future if required.
CREDITS
Thanks go to Matthew Dillon <dillon@backplane.com> for reviewing the
patches and supplying me with an SMP machine to test the changes.
Matthew Dillon [Sat, 29 May 2004 02:47:50 +0000 (02:47 +0000)]
Make sysctl_kern_proc() iterate through available cpus to retrieve the
list of threads installed on each one rather then just retrieve the threads
on the cpu it happens to have been called from.
To accomplish this, sysctl_kern_proc() now uses the new LWKT API to
migrate itself to each target cpu in order to safely access the thread
list on that cpu, ending up on the same cpu it started from. The overhead
is nasty to be sure, but who cares when only 'top' and 'ps' and related
non-critical utilities use the feature?
David Rhodus [Fri, 28 May 2004 20:30:18 +0000 (20:30 +0000)]
oops.
make `pkg_add -r XFree86` work again by changing the Latest path back.
Matthew Dillon [Fri, 28 May 2004 08:37:34 +0000 (08:37 +0000)]
Add lwkt_setcpu_self(), a function which migrates the current thread to
the specified cpu. This will soon be used by sysctl_kern_proc() to
collect thread information across all available cpus (because it is only
legal to manipulate a thread on the cpu it belongs to).
Yes, you heard that right and, yes, the overhead is nasty... one whole
microsecond per cpu at least, possibly even two. But who cares for
something like 'ps'?
In-conversation-with: Hiten Pandya <hmp@freebsd.org>
Matthew Dillon [Thu, 27 May 2004 19:21:49 +0000 (19:21 +0000)]
Initialize the FP unit earlier in the AP boot sequence. This solves
reported problems on MP systems not booting with mmxopt=1 .
Reported-by: Many. Thanks to David Rhodus for testing and debugging!
Matthew Dillon [Thu, 27 May 2004 18:15:43 +0000 (18:15 +0000)]
Attach bind-9.2.4rc4 to the base system. Rip out bind-8 binaries and add
in / adjust for most bind-9 binaries. Note that some pre-generated header
files have been hacked into /usr/src/usr.sbin/named and might have to be
regenerated from the original distribution when incorporating updates to
bind. But, for the most part, the build is clean.
Clean up rc.d/named and etc/defaults/rc.conf, add a README file to
/etc/namedb describing setup requirements, and change the default
named.conf file and RCNG variables to (if named is enabled) run named in
a chrooted environment as user 'bind'. Adapt the 'getroot' script that
I have used for many years to provide a better root.zone master methodology
instead of the root hints file that has been traditionally shipped.
Adjust the mtree files and distrib-dirs target to create the necessary
infrastructure in /etc/namedb for handling a chrooted named.
NOTE: bind-8 had 'ndc' and 'named-xfer'. bind-8 has rndc and there is no
named-xfer (it is builtin to named). nslookup still exists but is deprecated
in favor of 'dig', and the old dnskeygen/dnsquery mechanism has been replaced
by the newer dnssec-* mechanism.
Matthew Dillon [Thu, 27 May 2004 00:38:58 +0000 (00:38 +0000)]
Bring in the fictitious page wiring bug fixes from FreeBSD-5. Make additional
major changes to the APIs to clean them up (so this commit is substantially
different than what was committed to FreeBSD-5).
Obtained-from: Alan Cox <alc@cs.rice.edu> (FreeBSD-5)
Matthew Dillon [Wed, 26 May 2004 23:34:25 +0000 (23:34 +0000)]
Mask bits properly for pte_prot() in case it is called with additional
VM_PROT_ bits.
Fix a wired memory leak bug in pmap_enter(). If a page wiring change is
made and the page has already been faulted in for read access, and a
write-fault occurs, pmap_enter() was losing track of the wiring count in
the pmap when it tried to optimize the RO->RW case in the page table.
This prevented the page table page from being freed and led to a memory leak.
The case is easily reproducable if you attempt to wire the data/bss crossover
page in a program (typically just declare a global variable in a small program
and mlock() its page, then exit without munlock()ing). 4K is lost each time
the program is run.
Matthew Dillon [Wed, 26 May 2004 20:04:07 +0000 (20:04 +0000)]
ANSIfication and cleanup. No functional changes.
Submitted-by: Tim Wickberg <me@k9mach3.org>
Matthew Dillon [Wed, 26 May 2004 19:11:08 +0000 (19:11 +0000)]
Correct a bug in the last commit, udev must be assigned from vp->v_udev, not
vp->v_rdev.
Matthew Dillon [Wed, 26 May 2004 19:10:00 +0000 (19:10 +0000)]
Cleanup warnings. No operational changes. Note that the original ACPICA
character device interface was already disabled prior to this commit.
Matthew Dillon [Wed, 26 May 2004 19:09:04 +0000 (19:09 +0000)]
Cleanup warnings. No operational changes.
Hiten Pandya [Wed, 26 May 2004 14:12:34 +0000 (14:12 +0000)]
Consolidate SYSCTL_DECL(_kern_ipc), move it to sys/sysctl.h as
a common second-level OID.
No operational changes.