Add a new option to edquota(8) - "-f", which will
allow to limit the prototype quota distribution (-p)
to a single filesystem. Useful when initializing
quotas on a newly added disk.
peter [Fri, 28 Sep 2001 04:37:08 +0000 (04:37 +0000)]
Unwind some more macros. NFSMADV() was kinda silly since it was right
next to equivalent m_len adjustments. Move the nfsm_subs.h macros
into groups depending on which phase they are used in, since that
affects the error recovery requirements. Collect some of the common error
checking into a single macro as preparation for unwinding some more.
Have nfs_rephead return a value instead of secretly modifying args.
Remove some unused function arguments that were being passed around.
Clarify nfsm_reply()'s error handling (I hope).
Stop using linux_getpgid(). The implementation at this time is
broken and fixing it only creates a duplicate of what is already
in the FreeBSD kernel. Therefore, map the syscall directly to
getpgid().
PR: kern/21402
Submitted by: Christian Weisgerber <naddy@mips.inka.de>
While here, redefine the second entry for setpgid() so that we
don't need a stub. This is achieved by giving the second instance
the type NODEF.
Stop using linux_getpgid(). The implementation at this time is
broken and fixing it only creates a duplicate of what is already
in the FreeBSD kernel. Therefore, map the syscall directly to
getpgid().
PR: kern/21402
Submitted by: Christian Weisgerber <naddy@mips.inka.de>
Make the NODEF type usable. A syscall of type NODEF will only
have its entry in the syscall table added. Nothing else is
done. This differs from type NOPROTO in that NOPROTO adds a
definition to syscall.h besides adding a sysent. A syscall can
now have multiple entries without conflict. Note that the
argssize is fixed and depends on the syscall name.
luigi [Fri, 28 Sep 2001 00:05:11 +0000 (00:05 +0000)]
namei.h: move "struct componentname" definition outside "struct nameidata",
and provide a valid STDC/C++ definition for function NDINIT
queue.h libkern.h: put explicit casts from void * in insque, remque and memset
(for the records, these changes are necessary to let the files
compile with g++, which is used to build a FreeBSD module
for "Click" -- see www.pdos.lcs.mit.edu/click/ .
Given that they have zero impact on our code, it is worthwhile
to have them in.
Add device driver support for the Broadcom BCM570x family of gigabit
ethernet controllers. This adds support for the 3Com 3c996-T, the
SysKonnect SK-9D21 and SK-9D41, and the built-in gigE NICs on
Dell PowerEdge 2550 servers. The latter configuration hauls ass:
preliminary measurements show TCP speeds of over 900Mbps using
only normal size frames.
TCP/IP checksum offload, jumbo frames and VLAN tag insertion/stripping
are supported, as well as interrupt moderation.
Still need to fix autonegotiation support for 1000baseSX NICs, but
beyond that, driver is pretty solid.
luigi [Thu, 27 Sep 2001 23:44:27 +0000 (23:44 +0000)]
Two main changes here:
+ implement "limit" rules, which permit to limit the number of sessions
between certain host pairs (according to masks). These are a special
type of stateful rules, which might be of interest in some cases.
See the ipfw manpage for details.
+ merge the list pointers and ipfw rule descriptors in the kernel, so
the code is smaller, faster and more readable. This patch basically
consists in replacing "foo->rule->bar" with "rule->bar" all over
the place.
I have been willing to do this for ages!
- Fix some minor whitespace nits.
- Move the SPECIAL_FLAG #define up next to the NOHOLDER #define and fix a
little nit that caused it to be defined as -(sizeof (struct thread) + 1)
instead of -2.
RFC2349 (http://www.hypermail.org/rfcs/rfc2349.html) adds support
for negotiation of timeout and file size to the tftp protocol. This
is required by some firmware like EFI boot managers (at least on
HP i2000 Itanium servers) in order to boot an image using tftp. The
attached patch implements the RFC, and in doing so also implements
RFC2347; a generic tftp option extension.
Disable the check in icu_setup() to see if a handler was already used as
the current interrupt thread routines will guarantee the condition this is
checking for at a higher level but inthand_add() and inthand_remove() as
they currently exist don't satisfy this condition. (Which does need to be
fixed but which will take a bit more work.) This fixes shared interrupts.
Basically rewrite the sortq() routine which is used by 'lpc clean' and
'lpc tclean'. In some obscure cases, the previous version could cause a
valid user job to be removed (by 'clean'), due to invalid assumptions in
the sort routine. This was a rare problem, unless ctlinfo.c is compiled
with 'LEAVE_TMPCF_FILES' turned on (to check what that rtn was doing).
Reviewed by: Lack of outcry on -audit and freebsd-print@bostonradio.org
MFC after: 10 days
We need to build ``.depend'' early in the "build-tools" for the GCC
"build-tools". If we do not do this, the "depend" stage of
"buildworld" will build ``.depend'' and it will record the wrong
library and header dependencies (DESTDIR=${WORLDTMP}). Even worse,
the "all" stage may clobber build-architecture-format build tools
built in the "build-tools" stage with target-architecture-format ones.
Gate low level console output on mtx_lock_spin(&sio_lock), if the
sio_lock has been initialized. This prevents the low level console
output (kernel printf) from clobbering the sio settings if the system
happens to be in the middle of comstart().
luigi [Thu, 27 Sep 2001 11:53:22 +0000 (11:53 +0000)]
Remove unused (and duplicate) struct ip_opts which is never used,
not referenced in Stevens, and does not compile with g++.
There is an equivalent structure, struct ipoption in ip_var.h
which is actually used in various parts of the kernel, and also referenced
in Stevens.
Bill Fenner also says:
... if you want the trivia, struct ip_opts was introduced
in in.h SCCS revision 7.9, on 6/28/1990, by Mike Karels.
struct ipoption was introduced in ip_var.h SCCS revision 6.5,
on 9/16/1985, by... Mike Karels.
NULL commit due to misunderstanding "cvs commit -m <blah>". The
previous log message "/home/brooks/ng_gif.message" should have read:
Add a pair of new netgraph nodes. The ng_gif node is like ng_ether
except that it works on gif interfaces. The ng_gif_demux node attaches
to an ng_gif node to allow separate processing or discarding of
different types of encapsulated traffic.
Unconditionally use basename.c source vs. only doing this if the libc we
are linking against does not have basename(). There is a buffer overflow
bug in lib/libc/gen/basename.c rev 1.1. There is no way for us to test
what revision of basename() we have in libc, thus this change.
o Re-enable support of system file flags in jail() by adding back the
PRISON_ROOT to the suser_xxx() check. Since securelevels may now
be raised in specific jails, use of system flags can still be
restricted in jail(), but in a more configurable way.
o Users of jail() expecting system flags (such as schg) to restrict
jail()'s should be sure to set the securelevel appropriately in
jail()'s.
o This fixes activities involving automated system flag removal in
jail(), including installkernel and friends.
o When performing a securelevel check as part of securelevel_ge() or
securelevel_gt(), determine first if a local securelevel exists --
if so, perform the check based on imax(local, global). Otherwise,
simply use the global value.
o Note: even though local securelevels might lag below the global one,
if the global value is updated to higher than local values, maximum
will still be used, making the global dominant even if there is local
lag.
o Modify kern.securelevel MIB entry to return a local securelevel, if
one is present in the current jail, otherwise, to return the global
securelevel.
o If the securelevel is being updated, require that it be greater than
the maximum of local and global, if a local securelevel exists,
otherwise, just maximum of the global. If there is a local
securelevel, update the local one instead of the global one.
o Note: this does allow local securelevels to lag behind the global one
as long as the local one is not updated following a global increase.
o Modify access control checks in linux_iopl() to use securelevel_gt()
rather than direct variable checks. (Yet another API to perform
direct hardware I/O.)
o Modify pcvt ioctl() security checks to use securelevel_gt() instead
of direct securelevel variable checks. (Yet another API to perform
direct hardware I/O using KDENABIO.)
o Modify IPFW and DUMMYNET administrative setsockopt() calls to use
securelevel_gt() to check the securelevel, rather than direct access
to the securelevel variable.
o Modify static settime() to accept the proc * for the process requesting
a time change, and callers so that they provide td->td_proc.
o Modify settime() to use securevel_gt() for securelevel checking.
The "cat - -" feature was broken by the last commit.
Restore the code that avoided closing and reopening
stdin. This is also required by POSIX. As a bonus,
enable multiple stdin reads with the -benstv flags,
by resetting the EOF condition on stdin.
Make uio_yield() a global. Call uio_yield() between chunks
in vn_rdwr_inchunks(), allowing other processes to gain an exclusive
lock on the vnode. Specifically: directory scanning, to avoid a race to the
root directory, and multiple child processes coring simultaniously so they
can figure out that some other core'ing child has an exclusive adv lock and
just exit instead.
This completely fixes performance problems when large programs core. You
can have hundreds of copies (forked children) of the same binary core all
at once and not notice.
The arg parameter is passed by value in Linux, but not in FreeBSD.
We still have to account for a copyin. Make sure the copyin will
succeed by passing the FreeBSD syscall a pointer to userspace,
albeit one that's automagically mapped into kernel space.
Add a standard hack in the spirit of PCI_ENABLE_IO_MODES to allow systems
with weird PCI-PCI bridge configurations to work. Defining
PCI_ALLOW_UNSUPPORTED_IO_RANGE causes the sanity checks to pass even
with out of range values.
o So, when <dd> e-mailed me and said that the comment was inverted
for securelevel_ge() and securelevel_gt(), I was a little surprised,
but fixed it. Turns out that it was the code that was inverted, during
a whitespace cleanup in my commit tree. This commit inverts the
checks, and restores the comment.