Joerg Sonnenberger [Thu, 10 Feb 2005 00:20:09 +0000 (00:20 +0000)]
Remove stale local variable ifq.
Joerg Sonnenberger [Thu, 10 Feb 2005 00:09:17 +0000 (00:09 +0000)]
Use ifp->if_snd directly. Use IFQ_SET_MAXLEN.
Joerg Sonnenberger [Thu, 10 Feb 2005 00:08:38 +0000 (00:08 +0000)]
Don't assign &ifp->if_snd to a temporary variable, it's type will change
soon. Use ifp directly for BPF_MTAP.
Joerg Sonnenberger [Thu, 10 Feb 2005 00:07:23 +0000 (00:07 +0000)]
Directly use ifp->if_snd, it's type will change soon.
Joerg Sonnenberger [Wed, 9 Feb 2005 22:46:58 +0000 (22:46 +0000)]
Add -DIN_TARGET_LIB to prevent libobjc from depending on in-tree GCC code.
This removes the undefined x86_field_alignment from the DSO.
Noticed-by: eirikn
Max Okumoto [Wed, 9 Feb 2005 22:10:08 +0000 (22:10 +0000)]
Cleanup white space (removed spc between tabs)
David Rhodus [Wed, 9 Feb 2005 21:01:50 +0000 (21:01 +0000)]
Connect our 1:1 threading library to the build.
Max Okumoto [Wed, 9 Feb 2005 20:55:15 +0000 (20:55 +0000)]
Changed renamed macros *_BRACKET to *_BRACE.
Max Okumoto [Wed, 9 Feb 2005 20:40:38 +0000 (20:40 +0000)]
Untangle VarFind and, while rewriting most of the function, fix the
intendation. (A large part of the function was already at intendation 8).
Cool! much cleaner.
Taken-from: FreeBSD
Author: harti
Max Okumoto [Wed, 9 Feb 2005 20:26:53 +0000 (20:26 +0000)]
Move VarCreate() and VarDestroy() close to top of file, and add white
space at begining of some functions.
Max Okumoto [Wed, 9 Feb 2005 11:52:18 +0000 (11:52 +0000)]
Removed uneeded white space.
Max Okumoto [Wed, 9 Feb 2005 06:05:41 +0000 (06:05 +0000)]
- Move Error case to top of elseif chain.
- Do not initialize startc, at top of function, but where a know
valid value can be assigned.
Max Okumoto [Wed, 9 Feb 2005 06:03:05 +0000 (06:03 +0000)]
- Split out error case where input to Var_Parse() is "$", only a
dollar sign.
- Make default case an explicit error instead of falling out of the
switch statment into another error case.
Matthew Dillon [Wed, 9 Feb 2005 02:51:04 +0000 (02:51 +0000)]
Fix a case that can prevent the vnlru_proc vnode recycler from operating.
The vnode recycler is not allowed to recycle 'internal' directory nodes
in the namecache topology. i.e. if the path A/B/C/D is cached, then the
recycler is not allowed to remove A, B, or C. The recycler checks
this condition by checking the vnode's v_holdcnt. The recycler also tries
to recycle vnodes with little or no cached data rather then the vnodes
with a great deal of cached data.
However, the buffer cache now uses v_holdcnt as well. If the buffer cache
is large enough and maxvnodes is small enough the recycler can wind up not
finding any vnodes to recycle.
The fix is clean out the buffers with the candidate vnode and then make a
final v_holdcnt check.
Reported-by: Gary Allan <dragonfly@gallan.plus.com>
Jeffrey Hsu [Tue, 8 Feb 2005 23:00:33 +0000 (23:00 +0000)]
Eliminate conditional check for initialized 'fp' on error in kern_sendfile().
Jeffrey Hsu [Tue, 8 Feb 2005 22:56:19 +0000 (22:56 +0000)]
Now that 'so_pcb' is properly declared as a 'void *', remove a layer of
indirection and directly use 'so->so_pcb' in place of 'sotoinpcb(so)'.
Joerg Sonnenberger [Tue, 8 Feb 2005 15:54:01 +0000 (15:54 +0000)]
Set so->so_pcb directly to NULL to avoid lvalue cast.
Joerg Sonnenberger [Tue, 8 Feb 2005 15:51:24 +0000 (15:51 +0000)]
Include pci_private.h to get pci_class_to_string.
Joerg Sonnenberger [Tue, 8 Feb 2005 14:31:16 +0000 (14:31 +0000)]
- convert to bus_space macros
- netgraph cleanup
- bus_alloc_resource_any
- if_printf insteed of printf("%s...", ifp_if_xname);
- use M_WAITOK in arc_init, it's called only from arc_attach
Mostly-obtained-from: FreeBSD
David Rhodus [Tue, 8 Feb 2005 04:07:57 +0000 (04:07 +0000)]
Remove call to readdir() that was missed durning the libc namespace
cleanup. We now make use of _readidir_unlocked().
First reported by : Joerg Anslik <joerg@anslik.de>
Bug hunter and fix provided by : YONETANI Tomokazu <qhwt+dfly@les.ath.cx>
Max Okumoto [Tue, 8 Feb 2005 02:13:13 +0000 (02:13 +0000)]
- Removed declaration of VarModify() since it is defined before it is used.
Done to sync with FreeBSD.
Max Okumoto [Tue, 8 Feb 2005 02:11:28 +0000 (02:11 +0000)]
Change whitespace to match with FreeBSD. But to reduce diffs, but
we really need to do a complete file sweep and clean them all up.
Max Okumoto [Tue, 8 Feb 2005 02:07:03 +0000 (02:07 +0000)]
- Add brace around switch statement.
- Cleanup white space in var.h
Matthew Dillon [Mon, 7 Feb 2005 23:32:26 +0000 (23:32 +0000)]
Up the initial interrupt configuration hook delay to 20 seconds before
reporting a potential problem.
Matthew Dillon [Mon, 7 Feb 2005 21:33:36 +0000 (21:33 +0000)]
Add support for pure kernel thread stack frames. Pure kernel threads do not
have a PCB, they simply push the register state (depending on the situation)
onto the stack and save the stack pointer.
Max Okumoto [Mon, 7 Feb 2005 20:46:19 +0000 (20:46 +0000)]
More white space diff reduction with FreeBSD.
Matthew Dillon [Mon, 7 Feb 2005 20:39:01 +0000 (20:39 +0000)]
gdb-6 uses /dev/kmem exclusively for kernel addresses when gdb'ing a live
kernel, but the globaldata mapping is outside the bounds of kernel_map.
Make sure that the globaldata mapping is visible to it.
Max Okumoto [Mon, 7 Feb 2005 20:11:54 +0000 (20:11 +0000)]
- Reduce white space diffs with FreeBSD.
Max Okumoto [Mon, 7 Feb 2005 20:09:45 +0000 (20:09 +0000)]
- Reduce white space diffs after harti commited my change to FreeBSD.
- Removed extra scope.
Liam J. Foy [Mon, 7 Feb 2005 18:00:46 +0000 (18:00 +0000)]
- Remove two unnecessary headers
Max Okumoto [Sun, 6 Feb 2005 23:58:23 +0000 (23:58 +0000)]
- Use a temp variable to simplify long condition in if-statement.
Max Okumoto [Sun, 6 Feb 2005 23:56:21 +0000 (23:56 +0000)]
- Eliminate duplicate code.
Max Okumoto [Sun, 6 Feb 2005 23:54:55 +0000 (23:54 +0000)]
- cleanup Var_Parse() by removing negtive logic in if statement.
Joerg Sonnenberger [Sun, 6 Feb 2005 23:26:43 +0000 (23:26 +0000)]
Don't cast lvavalues.
Max Okumoto [Sun, 6 Feb 2005 23:22:04 +0000 (23:22 +0000)]
- Remove redundent code. Why check for isupper before a
case statement where you only have upper case values.
Max Okumoto [Sun, 6 Feb 2005 23:17:16 +0000 (23:17 +0000)]
- Define new typedef VarModifyProc and use it to clarify the code.
Max Okumoto [Sun, 6 Feb 2005 06:16:40 +0000 (06:16 +0000)]
Moved WARNS?=6 out of subdir Makefiles in to src/bin/Makefile.inc.
Subdir that need lower values should set WARNS instead.
Max Okumoto [Sun, 6 Feb 2005 06:12:32 +0000 (06:12 +0000)]
Move WARNS?=6 into src/bin/Makefile.inc instead of having it set
in all the subdir Makefiles. Subdirectories that fail WARNS=6
should have lower values to override.
Joerg Sonnenberger [Sun, 6 Feb 2005 02:00:23 +0000 (02:00 +0000)]
Add missing */.
Noticed-by: corecode
Joerg Sonnenberger [Sat, 5 Feb 2005 23:23:25 +0000 (23:23 +0000)]
Remove the conditionalized FreeBSD 5 code. Keep the capability assignment,
even if we currently don't use it.
Joerg Sonnenberger [Sat, 5 Feb 2005 23:16:42 +0000 (23:16 +0000)]
During init time, we can savely allocate the mbuf cluster with
MB_WAIT. Pass the flag to em_get_buf to allow this.
Joerg Sonnenberger [Sat, 5 Feb 2005 23:04:28 +0000 (23:04 +0000)]
Add a macro SYSCTL_SET_CHILDREN. Use it to avoid lhs cast.
Joerg Sonnenberger [Sat, 5 Feb 2005 22:54:49 +0000 (22:54 +0000)]
- Add support for DT_FLAGS.
- Define various things from the most recent ELF spec.
rtld.c, 1.83
rtld.h, 1.28
map_object.c, 1.14
sys/elf_common.h, 1.12
Joerg Sonnenberger [Sat, 5 Feb 2005 22:20:20 +0000 (22:20 +0000)]
Do initialise fp to NULL explicitly, the last comment wasn't enough.
The first conditional can fail and the variable points to some random
place.
Liam J. Foy [Sat, 5 Feb 2005 16:07:08 +0000 (16:07 +0000)]
- Remove mode argument from open().
Chris Pressey [Sat, 5 Feb 2005 01:05:12 +0000 (01:05 +0000)]
Raise WARNS to 6:
- Move the print() function above the display() function so that it can
be inlined by the compiler.
- Un-nest extern declarations; put them into the shared header file.
- Initialize some local variables.
Max Okumoto [Sat, 5 Feb 2005 00:58:47 +0000 (00:58 +0000)]
- Reduce diffs from FreeBSD.
Max Okumoto [Fri, 4 Feb 2005 23:32:09 +0000 (23:32 +0000)]
- make a temporary copy of the name because Var_Subst insists on
writing into the string
Max Okumoto [Fri, 4 Feb 2005 23:18:42 +0000 (23:18 +0000)]
- Pass NULL to brk_string() when we don't need the number
of result strings.
Joerg Sonnenberger [Fri, 4 Feb 2005 23:16:28 +0000 (23:16 +0000)]
Readd the copystr for f_mntfromname of root. It wasn't meant to be
deleted, but didn't change normal system behaviour. In case of an
unclean root filesystem, it prevented the remounting.
Max Okumoto [Fri, 4 Feb 2005 23:09:06 +0000 (23:09 +0000)]
- Fix syntax error. That I added. :-P
Max Okumoto [Fri, 4 Feb 2005 22:35:42 +0000 (22:35 +0000)]
- In main.c use NULL instead of 0
- In str.c check if brk_string() is passed a null pointer in
store_argc. If so then don't return argc.
Taken-from: FreeBSD
Joerg Sonnenberger [Fri, 4 Feb 2005 22:28:23 +0000 (22:28 +0000)]
Stop caring about GCC versions between 2.5 and 2.7, they are
historic artefacts.
Remove a double define of __func__ from the !LINT case.
Define __func__ to NULL for GCC before 2.7 and non ISO C99 compilers.
Since all GCC versions above 2.95 support __restrict, include them
in the check too. Remove the weak defines for !__STDC__, we don't
support such a setting.
If _C99_SOURCE is set, don't include POSIX / BSD / XSI extensions.
Partly-submitted-by: Andreas Hauser <andy@splashground.de>
Max Okumoto [Fri, 4 Feb 2005 22:06:01 +0000 (22:06 +0000)]
- Constify Var_Delete()
- And more white space work
Max Okumoto [Fri, 4 Feb 2005 21:53:47 +0000 (21:53 +0000)]
- Removed unneeded variable 'internal'.
Max Okumoto [Fri, 4 Feb 2005 21:45:36 +0000 (21:45 +0000)]
- Revert some white space changes, to keep us in sync with FreeBSD.
- Move Buf_Clear() to end of file (why?) to keep us in sycn with FreeBSD.
Chris Pressey [Fri, 4 Feb 2005 21:32:55 +0000 (21:32 +0000)]
Raise WARNS to 6:
- ANSIfy function arguments.
Chris Pressey [Fri, 4 Feb 2005 21:30:42 +0000 (21:30 +0000)]
Raise WARNS to 6.
Max Okumoto [Fri, 4 Feb 2005 21:24:06 +0000 (21:24 +0000)]
More white space insertion. Why does putting a blank line at the
begining of functions without variables help readablity?
Max Okumoto [Fri, 4 Feb 2005 21:16:44 +0000 (21:16 +0000)]
Put back spaces at begining of functions that have no variables. Silly
but it keeps us in sync with FreeBSD.
Chris Pressey [Fri, 4 Feb 2005 20:00:25 +0000 (20:00 +0000)]
Raise WARNS to 6:
- Constify some function arguments and a variable.
Matthew Dillon [Fri, 4 Feb 2005 19:16:00 +0000 (19:16 +0000)]
The last commit was not sufficient. Rework the code a bit to make it
sufficient.
Reported-by: EM <EM1897@aol.com>
Max Okumoto [Fri, 4 Feb 2005 06:45:16 +0000 (06:45 +0000)]
Back out constify of Parse_File(). It is assigned to curFile.fname
and you cannot const this because it's passed to free() later on.
Matthew Dillon [Fri, 4 Feb 2005 03:32:37 +0000 (03:32 +0000)]
Revert the last device_print_child() change, it was too confusing to
display the device name after the attach because recursion can insert a
lot of other messages in there.
Instead, if -v is specified, generate a line before and after the attach
suitably marked [tentative] or [attached!]. This also has the advantage
of bracketing any recursion that occurs.
Keep the bootverbose printing of the device chain but get rid of the "ATTACH "
on the front. It's obvious what it is and the attach message follows
directly now.
YONETANI Tomokazu [Fri, 4 Feb 2005 03:04:40 +0000 (03:04 +0000)]
s/ich_arg/ich_desc/
Matthew Dillon [Fri, 4 Feb 2005 02:57:20 +0000 (02:57 +0000)]
One of the last things the system does before it tries to mount root is
wait for interrupt-driven configuration hooks to finish. This is one of
the common places where the system locks up due to e.g. blown interrupt
routing.
Wait up to 60 seconds for interrupt driven hooks to complete. Generate
a warning every 10 seconds listing the hooks that have not yet completed.
Name all the hook structures to make the listing meaningful. After 60
seconds, give up and continue booting. The system might actually
be useable enough to aid in debugging depending on which device failed.
Matthew Dillon [Fri, 4 Feb 2005 02:52:18 +0000 (02:52 +0000)]
Annotate the class byte with a class name in the bootverbose pci "found->"
bus logging code. Numbers are fine, but when there are thousands of them
on the screen you need a bit of human readable context.
Jeffrey Hsu [Fri, 4 Feb 2005 01:49:30 +0000 (01:49 +0000)]
Temporarily disable non-working Path MTU discovery pending real fix.
Joerg Sonnenberger [Fri, 4 Feb 2005 01:33:48 +0000 (01:33 +0000)]
Do not depend on existence of _end symbol in obj_from_addr, use
obj->mapbase and obj->mapsize instead.
rtld.c, 1.92
Obtained-from: FreeBSD
Joerg Sonnenberger [Fri, 4 Feb 2005 01:23:16 +0000 (01:23 +0000)]
If we change obj_rtld.path after initialising __progname, make sure we
change __progname to point at the new storage otherwise it ends up
pointing at freed memory which leads to confusing garbled error messages.
rtld.c, 1.95
Obtained-from: FreeBSD
Matthew Dillon [Fri, 4 Feb 2005 01:14:27 +0000 (01:14 +0000)]
m_clalloc() was improperly assuming that an mcl malloc would always succeed
when in fact it is allowed not to (because M_NULLOK is specified).
Reported-by: EM <EM1897@aol.com>
Joerg Sonnenberger [Fri, 4 Feb 2005 01:06:05 +0000 (01:06 +0000)]
give out a little more information in case of a missing dependency
rtld.c, 1.96
Obtained-from: FreeBSD
Joerg Sonnenberger [Fri, 4 Feb 2005 00:33:49 +0000 (00:33 +0000)]
Remove -DFREEBSD_ELF.
Verified-by: md5
Joerg Sonnenberger [Fri, 4 Feb 2005 00:24:23 +0000 (00:24 +0000)]
No need to zero fill memory, mmapped anonymously. Kernel will
return pre-zeroed pages itself.
i386/reloc.c, 1.12
Obtained-from: FreeBSD
Matthew Dillon [Thu, 3 Feb 2005 23:54:22 +0000 (23:54 +0000)]
Generate more useful -v information on the console during device attach.
The complete device chain is output prior to each attach.
The normal device_print_child() is now moved from before the attach
to after the attach so the correct resource information gets reported,
especially the correct IRQ.
Max Okumoto [Thu, 3 Feb 2005 23:41:05 +0000 (23:41 +0000)]
Remove duplicate #include of <sys/sysctl.h>
Joerg Sonnenberger [Thu, 3 Feb 2005 23:28:24 +0000 (23:28 +0000)]
Provide a mechanism for dumping relocation information.
Setting the LD_DUMP_REL_PRE or LD_DUMP_REL_POST environment variables
cause rtld-elf to output a table of all relocations.
This is useful for debugging.
debug.c, 1.3
rtld.c, 1.85
rtld.h, 1.31
Obtained-from: FreeBSD
Joerg Sonnenberger [Thu, 3 Feb 2005 22:10:57 +0000 (22:10 +0000)]
Remove the default VINUMDEBUG option for now. I'll make most of the
debug code available via KTR later.
Joerg Sonnenberger [Thu, 3 Feb 2005 22:07:12 +0000 (22:07 +0000)]
Forced commit to note that I'm removing the default VINUMDEBUG for now.
Most debug informations will be made available via KTR again.
Joerg Sonnenberger [Thu, 3 Feb 2005 22:05:27 +0000 (22:05 +0000)]
strcmp() needs string.h
Joerg Sonnenberger [Thu, 3 Feb 2005 22:03:31 +0000 (22:03 +0000)]
exit() needs stdlib.h
Joerg Sonnenberger [Thu, 3 Feb 2005 20:48:08 +0000 (20:48 +0000)]
Use M_ZERO instead of manual bzero.
Max Okumoto [Thu, 3 Feb 2005 18:43:08 +0000 (18:43 +0000)]
Cleanup whitespace at end of file.
Max Okumoto [Thu, 3 Feb 2005 18:36:26 +0000 (18:36 +0000)]
Cleanup some white space.
Joerg Sonnenberger [Thu, 3 Feb 2005 17:44:21 +0000 (17:44 +0000)]
Remove debug flag which slipped into last commit.
Joerg Sonnenberger [Thu, 3 Feb 2005 17:40:15 +0000 (17:40 +0000)]
Don't define _Bool for GCC 3.0 and later. It was added was builtin
on November 13 2000, which is why before the gcc 3.0 release.
Joerg Sonnenberger [Thu, 3 Feb 2005 17:28:40 +0000 (17:28 +0000)]
Fully separate the kcore initialisation and the kinfo wrapper.
Call kcore_wrapper_open if you want the kinfo wrapper, otherwise
kcore_open. This fixes the segfaults in pstat.
Reported-by: Sven Willenberger <sven@dmv.com>
Joerg Sonnenberger [Thu, 3 Feb 2005 17:11:54 +0000 (17:11 +0000)]
No namespace pollution in sys/cdefs.h. Rename VM_CACHELINE_SIZE
to __VM_CACHELINE_SIZE and use the protected attribut names.
Joerg Sonnenberger [Thu, 3 Feb 2005 12:58:44 +0000 (12:58 +0000)]
Read eaddr in two parts (32 bit read and 16 bit read). This fixes a
compiler warning too.
Tested-by: <jkanowitz@snet.net>
Joerg Sonnenberger [Thu, 3 Feb 2005 11:20:31 +0000 (11:20 +0000)]
GCC 1.x is dead.
Joerg Sonnenberger [Thu, 3 Feb 2005 09:20:27 +0000 (09:20 +0000)]
Free temporary buffer in the buffer overflow case too.
David Rhodus [Thu, 3 Feb 2005 02:47:26 +0000 (02:47 +0000)]
Add information about return code for immutable files.
Noted by : Guy Harris <guy@alum.mit.edu>
David Rhodus [Thu, 3 Feb 2005 02:43:44 +0000 (02:43 +0000)]
Add information about return code for immutable files.
Noted by : Guy Harris <guy@alum.mit.edu>
Joerg Sonnenberger [Wed, 2 Feb 2005 21:34:19 +0000 (21:34 +0000)]
Don't use the statfs field f_mntonname in filesystems. For the userland
export code, it can synthesized from mnt_ncp.
For debugging code, use f_mntfromname, it should be enough to find
culprit. The vfs_unmountall doesn't use code_fullpath to avoid problems
with resource allocation and to make it more likely that a call from ddb
succeds.
Change getfsstat and fhstatfs to not show directories outside a chroot
path, with the exception of the filesystem counting the chroot root itself.
Liam J. Foy [Wed, 2 Feb 2005 20:42:15 +0000 (20:42 +0000)]
- Remove unnecessary header (time.h)
- Tidy up some sentences
- Remove white space
Matthew Dillon [Wed, 2 Feb 2005 20:36:09 +0000 (20:36 +0000)]
Fix a hard-to-find bugaboo in the struct file list sysctl. The code was
not properly calculating the required buffer size because it was ignoring
the fact that descriptor tables are shared between threaded processes. Simply
adding fp->f_count for the fp's in the filehead list doesn't do the job if
a system is running threaded processes.
Reported-by: "Erik P. Skaalerud" <erik@pentadon.com>
Chris Pressey [Wed, 2 Feb 2005 20:27:51 +0000 (20:27 +0000)]
Raise WARNS to 6:
- Rename a local variable to avoid conflict with the name of a global;
- Mark some function arguments as unused;
- Make local functions static and fix up function prototypes;
- Use better logic for checking return value from snprintf().
Joerg Sonnenberger [Wed, 2 Feb 2005 18:46:01 +0000 (18:46 +0000)]
Don't cast int ==> unsigned char ==> char, int ==> char is enough.
Submitted-by: Andreas Hauser <andy@splashground.de>
Liam J. Foy [Wed, 2 Feb 2005 15:21:21 +0000 (15:21 +0000)]
- Reduce the size of msg (1024 -> 80)
- Remove double '' when writing a warning
- Remove unnecessary check
Hiroki Sato [Wed, 2 Feb 2005 15:10:55 +0000 (15:10 +0000)]
Query A records before AAAA records in getaddrinfo() when AF_UNSPEC
is specified. Some broken DNS servers return NXDOMAIN against
non-existent AAAA queries, even when it should return NOERROR
with empty return records. This is a problem for an IPv4/IPv6 dual
stack node since the NXDOMAIN returned by the first query of
an AAAA record makes it give up querying the A record. Also, this
behavior has been recognized as a potential denial-of-service attack.
Note that although the query order has been changed, the result
linked-list of (struct addrinfo) set by getaddrinfo() is still
in order of AF_INET6 -> AF_INET.
Reference: http://www.kb.cert.org/vuls/id/714121