Commit Graph

89 Commits

Author SHA1 Message Date
ad d991fcb3b6 More changes to improve kern_descrip.c.
- Avoid atomics in more places.
- Remove the per-descriptor mutex, and just use filedesc_t::fd_lock.
  It was only being used to synchronize close, and in any case we needed
  to take fd_lock to free the descriptor slot.
- Optimize certain paths for the <NDFDFILE case.
- Sprinkle more comments and assertions.
- Cache more stuff in filedesc_t.
- Fix numerous minor bugs spotted along the way.
- Restructure how the open files array is maintained, for clarity and so
  that we can eliminate the membar_consumer() call in fd_getfile().  This is
  mostly syntactic sugar; the main functional change is that fd_nfiles now
  lives alongside the open file array.

Some measurements with libmicro:

- simple file syscalls are like close() are between 1 to 10% faster.
- some nice improvements, e.g. poll(1000) which is ~50% faster.
2009-05-24 21:41:25 +00:00
lukem dadffc77c0 Fix sign-compare issues. 2009-04-12 06:36:12 +00:00
christos 4fe86cf5b0 64 bit dev_t 2008-12-29 00:59:08 +00:00
gmcgarry 0e0358051d Cast f_type to signed integer before comparison with negative values. 2008-09-09 00:56:23 +00:00
lukem 93505eb7ec Add back rev 1.83. (hi christos!) 2008-07-23 00:57:53 +00:00
christos 380fafb740 a better attempt on printing dtype_misc 2008-07-22 22:58:04 +00:00
lukem 98e5374ccb Remove the \n and tabs from the __COPYRIGHT() strings.
Tweak to use a consistent format.
2008-07-21 14:19:20 +00:00
christos eaa1a34779 handle NULL better. 2008-04-04 21:15:30 +00:00
christos 03829a56bb elementary handling for different types of DTYPE_FOO.
XXX: DTYPE_CRYPTO is unused.
XXX: should we change fd_clone() to pass a DTYPE_FOO?
2008-04-04 21:14:08 +00:00
christos 6238528d0f catch up with struct filedesc changes. 2008-04-04 18:27:00 +00:00
he 7fa93d44f6 Please repeat after me: I shall not define _KERNEL before including
<sys/types.h> in user-land programs.  Doing so causes build problems
for lots of our ports due to differences in how <sys/bswap.h> ends up
defining the swap functions/macros.
2008-03-23 23:01:21 +00:00
ad 5456139476 Catch up with recent file descriptor changes. 2008-03-22 02:39:06 +00:00
perry 8b0f9554ff convert __attribute__s to applicable cdefs.h macros 2007-12-15 19:44:37 +00:00
dyoung ca1b22492a Bug fix: filed.fd_ofiles is always a pointer to the open files
table, so always read the table from kernel virtual memory at that
address.  Do not ever try to read the table from filed0.fd_dfiles,
because it just isn't necessary, and sometimes we read a stale
table from there.
2007-09-09 01:26:55 +00:00
yamt 2416177358 tweak for user.h cleanup.
while i'm here, remove unnecessary #include.
2006-05-11 11:56:38 +00:00
mrg 9ce2c843f8 maxfiles -> fstat_maxfiles. (this _KERNEL definition code needs to not
clash with kernel headers....)
2006-05-11 01:21:23 +00:00
christos 62bee0ffa5 PR/32788: KAMADA Ken'ichi: add support for tmpfs 2006-02-10 16:01:45 +00:00
christos 553881914b WARNS=3, knf, ansi prototypes 2005-07-17 07:36:26 +00:00
atatat a92869d7ef Cast fst.fileid to unsigned long and print it as such. Inode numbers
are never negative.
2004-12-14 03:09:24 +00:00
christos 09c258b188 Add ptyfs. 2004-12-12 22:41:03 +00:00
itojun b95181fc9b NI_WITHSCOPEID was not picked up by IETF standardization process. 2004-11-16 05:59:32 +00:00
dsl e833833eb1 Add (unsigned char) cast to ctype functions 2004-10-30 17:29:47 +00:00
christos 6bd1d6d4db Replace the statfs() family of system calls with statvfs().
Retain binary compatibility.
2004-04-21 01:05:31 +00:00
aymeric 5f3a5a8279 fix the way the ext2fs-specific size and mode variables are fetched:
they are indirectly gotten from kernel memory and as such should be read
with KVM_READ.
This fixes fstat core dumps when some process holds a reference in a Linux
ext2fs filesystem.
Thanks to Aleksy Schubert (alx.at.mimuw.dot.edu.dot.pl) for debugging this
with me.
2004-04-02 14:22:05 +00:00
fvdl 49ec36daa1 Don't compare integers and NULL. 2003-10-21 02:16:59 +00:00
agc 89aaa1bb64 Move UCB-licensed code from 4-clause to 3-clause licence.
Patches provided by Joel Baker in PR 22365, verified by myself.
2003-08-07 11:13:06 +00:00
itojun 3c95d355e5 use bounded string op 2003-07-15 05:56:55 +00:00
fvdl 42614ed3f3 Add support for UFS2. UFS2 is an enhanced FFS, adding support for
64 bit block pointers, extended attribute storage, and a few
other things.

This commit does not yet include the code to manipulate the extended
storage (for e.g. ACLs), this will be done later.

Originally written by Kirk McKusick and Network Associates Laboratories for
FreeBSD.
2003-04-02 10:39:19 +00:00
christos 0fcef5865c undo mallocvar.h changes! 2003-02-02 02:35:58 +00:00
tron 3f39a3d045 Include "sys/mallocvar.h" with "_KERNEL" defined in time to avoid
build failure because MALLOC_DECLARE() is not defined.
2003-02-01 18:31:09 +00:00
martin 8010344d66 Make it compile again on LP64 archs.
XXX - we *realy* need some format string macros for size_t/off_t!
2003-01-20 06:20:37 +00:00
dsl 7e2ddd8ef0 Use kvm_getprocs2 so independant of kernel version
(change aggreed by christos)
2003-01-19 18:14:06 +00:00
yamt ebc708387a add VT_LFS case. 2002-10-26 06:03:50 +00:00
jdolecek 879d35cda1 merge rev. 1.50.2.1 (made by lukem):
rudimentary support for for displaying DTYPE_KQUEUE fds
2002-10-23 22:37:11 +00:00
simonb ba31699a83 Add an extra column for inode numbers, and pull back the mode field to
just 5 columns if in "numeric" mode.  Other fields should dynamically
resize, but not today...
2002-02-12 03:28:20 +00:00
jdolecek 305085a35d for (new)pipes, print 'n' if it's in nonblocking mode, and 'a' if it's
set to async
2001-11-25 18:49:02 +00:00
lukem 99147a7648 remove #include <ufs/ufs/quota.h> where it was just to appease
<ufs/ufs/inode.h>, since the latter now includes the former.  leave the former
in source that obviously uses specific bits of it (for completeness.)
2001-10-26 05:56:06 +00:00
jdolecek ac2147b9f2 remove accidentaly left debug printf
change the pipe descriptor output to include 'r' or 'w' to explicitly
indicate it's open for reading or writing, similarily to "regular" descriptors
2001-09-28 20:29:04 +00:00
jdolecek 92f9fd86ff the descriptor flags may also contain other stuff besides FWRITE|FREAD,
so do '& FWRITE' instead of '== FWRITE' in the check in pipe-related code
2001-09-28 20:16:43 +00:00
jdolecek 298d03ceba fix comment 2001-07-08 20:24:05 +00:00
wiz f3f6c5b675 `accessible' only has one `a'. 2001-06-19 12:52:20 +00:00
lukem a4224d7f92 convert if ... else if ... tree to a switch 2001-06-17 08:27:57 +00:00
jdolecek 28dbb7357f Add support for DTYPE_PIPE pipes (a.k.a NEW_PIPE) 2001-06-16 12:08:05 +00:00
mrg 7276c20511 don't #define _LKM to work around header lossage; it is no longer necessary. 2001-06-05 09:20:12 +00:00
lukem c2b84904ae use %ll_ instead of the less standard %q_ 2001-01-04 23:05:54 +00:00
enami a5989caf7a Redo previous a bit different way so that mount point of top layer is shown
instead of lowest one and ``fstat file-on-upperlayer'' works.
2000-08-14 09:17:11 +00:00
enami 1dccbbfbd2 Print lower vnode if vnode is on a layered filesystem. 2000-08-14 06:03:21 +00:00
sommerfeld 080fdae8fc Bring into line with recent ktrace changes 2000-05-27 15:02:04 +00:00
jdolecek 1371d1b72d add support for msdosfs 2000-02-04 11:02:00 +00:00
jdolecek 9200b8eabf move the mapping of vnode type to file type to separate function getftype()
and use it where appropriate
2000-02-04 10:35:46 +00:00