]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/log
FreeBSD/FreeBSD.git
21 years agoMake atm WARNS=6 clean. The changes are mostly:
Hartmut Brandt [Mon, 28 Jul 2003 08:14:27 +0000 (08:14 +0000)]
Make atm WARNS=6 clean. The changes are mostly:

 - remove some instances of __P()
 - use real prototypes and un-K&R function headers
 - constify where necessary (mostly strings and structures containing
   strings)
 - make functions and variables static that need not to be global
 - tag unused function parameters as __unused

Testing: a fresh universe

21 years agoAdd a PATH_INQ flag, PIM_NO_6_BYTE, which indicates the SIM never wishes to
Nate Lawson [Mon, 28 Jul 2003 06:15:59 +0000 (06:15 +0000)]
Add a PATH_INQ flag, PIM_NO_6_BYTE, which indicates the SIM never wishes to
receive 6 byte commands.  Add a check for this flag to da(4) and cd(4) so
that they honor it.  This is a quick workaround for many devices (especially
USB) that require da(4) quirks to operate.  The more complete approach is
to finish the new transport code which will be aware of the SCSI version a
transport implements.

MFC after: 1 day

21 years agoNone of the "alloc" functions used by UMA assume that Giant is held any
Alan Cox [Mon, 28 Jul 2003 02:29:07 +0000 (02:29 +0000)]
None of the "alloc" functions used by UMA assume that Giant is held any
longer.  (If they still need it, e.g., contigmalloc(), they acquire it
themselves.)  Therefore, we need not acquire Giant in slab_zalloc().

21 years agoauth.conf seems to be being gradually subsumed by user classes
Robert Watson [Mon, 28 Jul 2003 02:28:51 +0000 (02:28 +0000)]
auth.conf seems to be being gradually subsumed by user classes
and PAM configuration.  Remove the line concerning "auth_list"
from the template, since it's referenced only in the tinyware
password command, and only #ifdef KERBEROS, which isn't defined
in tinyware.  Add a comment about auth.conf being on the way
out the door.  The one remaining consumer of auth.conf is
crypt(3).

21 years agoThe LP_ETH_10_100_CF entry needs to be tagged as a DL100019.
Warner Losh [Mon, 28 Jul 2003 00:07:58 +0000 (00:07 +0000)]
The LP_ETH_10_100_CF entry needs to be tagged as a DL100019.

Submitted by: Scott Renfro

21 years agoReset the per-CPU unique value at boot and clear it in the PCB of the
Marcel Moolenaar [Sun, 27 Jul 2003 23:45:48 +0000 (23:45 +0000)]
Reset the per-CPU unique value at boot and clear it in the PCB of the
child when forking. This provides a consistent initial state.
Note that cpu_set_upcall() does not clear the per-CPU unique value as
it is followed by a call to set_mcontext(), which sets it accordingly.

21 years agoMake pmap_pvo_allocf() callable without Giant.
Alan Cox [Sun, 27 Jul 2003 20:57:53 +0000 (20:57 +0000)]
Make pmap_pvo_allocf() callable without Giant.

21 years agoAdd -dynamic to natd if dhcp is used for the natd interface.
Martin Blapp [Sun, 27 Jul 2003 20:34:30 +0000 (20:34 +0000)]
Add -dynamic to natd if dhcp is used for the natd interface.
Kill natd in stop().

Reviewed by: mtm

21 years agoPass the file descriptor index down to vn_open.
Poul-Henning Kamp [Sun, 27 Jul 2003 20:09:13 +0000 (20:09 +0000)]
Pass the file descriptor index down to vn_open.

If the method vector was replaced and we got the "special return code"
smile and trust that whatever happened below DTRT.

21 years agoPass the fdidx argument from vn_open{_cred}() onto VOP_OPEN()
Poul-Henning Kamp [Sun, 27 Jul 2003 20:05:36 +0000 (20:05 +0000)]
Pass the fdidx argument from vn_open{_cred}() onto VOP_OPEN()

21 years agoRemove GIANT_REQUIRED from kmem_alloc().
Alan Cox [Sun, 27 Jul 2003 18:31:32 +0000 (18:31 +0000)]
Remove GIANT_REQUIRED from kmem_alloc().

21 years agoSome Typos fix
Sebastien Gioria [Sun, 27 Jul 2003 18:10:24 +0000 (18:10 +0000)]
Some Typos fix

21 years agoAdd fdidx argument to vn_open() and vn_open_cred() and pass -1 throughout.
Poul-Henning Kamp [Sun, 27 Jul 2003 17:04:56 +0000 (17:04 +0000)]
Add fdidx argument to vn_open() and vn_open_cred() and pass -1 throughout.

21 years agoCall the new argument "fdidx" that is more precise than "fd".
Poul-Henning Kamp [Sun, 27 Jul 2003 17:03:20 +0000 (17:03 +0000)]
Call the new argument "fdidx" that is more precise than "fd".

21 years agoTry a lot harder to get dependancies right. This involves some ugly
Mark Murray [Sun, 27 Jul 2003 16:49:10 +0000 (16:49 +0000)]
Try a lot harder to get dependancies right. This involves some ugly
looking ${.OBJDIR} work that has the up-side of actually working
in upgrade and make -jN cases.

This needs to be revisited further, and it is conceivable that
the ${.OBJDIR} stuff can be simplified, but the sheer number of
edge cases and other causes make this Hard(tm). For now, this works.

21 years agoip6fw does not handle ESP correctly
Hajimu UMEMOTO [Sun, 27 Jul 2003 16:21:10 +0000 (16:21 +0000)]
ip6fw does not handle ESP correctly

PR: kern/54874
Submitted by: JINMEI Tatuya <jinmei@shuttle.wide.toshiba.co.jp>
MFC after: 1 week

21 years agoRespect BUS_DMA_ZERO in iommu_dvmamem_alloc().
Thomas Moestl [Sun, 27 Jul 2003 15:19:45 +0000 (15:19 +0000)]
Respect BUS_DMA_ZERO in iommu_dvmamem_alloc().

21 years agoUse the BUS_DMA_ZERO flag.
Maxime Henrion [Sun, 27 Jul 2003 14:38:54 +0000 (14:38 +0000)]
Use the BUS_DMA_ZERO flag.

21 years agoNow that we can call kmem_malloc without Giant it should be safe
David Malone [Sun, 27 Jul 2003 14:19:23 +0000 (14:19 +0000)]
Now that we can call kmem_malloc without Giant it should be safe
to do mbuf allocation without Giant, so remove the GIANT_REQUIRED
from mb_alloc in the M_TRYWAIT case.

21 years agoEnable the safe, harp, hatm and patm driver documentations for pc98.
Yoshihiro Takahashi [Sun, 27 Jul 2003 14:08:54 +0000 (14:08 +0000)]
Enable the safe, harp, hatm and patm driver documentations for pc98.

21 years agoDocument the new BUS_DMA_ZERO flag.
Maxime Henrion [Sun, 27 Jul 2003 14:05:29 +0000 (14:05 +0000)]
Document the new BUS_DMA_ZERO flag.

21 years agoChanged the data types of three index variables, two of them
Robert Drehmel [Sun, 27 Jul 2003 14:01:33 +0000 (14:01 +0000)]
Changed the data types of three index variables, two of them
in the `video_state' structure, to larger ones (from u_char to
u_short).  Each can now hold values at least as large as the
size of the array it is meant to point into.

This eliminates warnings printed by GCC 3.3.1 and hence makes
pcvt compilable using -Werror.

21 years agoUse the BUS_DMA_ZERO flag.
Maxime Henrion [Sun, 27 Jul 2003 14:00:02 +0000 (14:00 +0000)]
Use the BUS_DMA_ZERO flag.

21 years agoUse the BUS_DMA_ZERO flag instead of bzero()'ing DMA memory.
Maxime Henrion [Sun, 27 Jul 2003 13:56:03 +0000 (13:56 +0000)]
Use the BUS_DMA_ZERO flag instead of bzero()'ing DMA memory.

21 years ago- Introduce a new busdma flag BUS_DMA_ZERO to request for zero'ed
Maxime Henrion [Sun, 27 Jul 2003 13:52:10 +0000 (13:52 +0000)]
- Introduce a new busdma flag BUS_DMA_ZERO to request for zero'ed
  memory in bus_dmamem_alloc().  This is possible now that
  contigmalloc() supports the M_ZERO flag.
- Remove the locking of Giant around calls to contigmalloc() since
  contigmalloc() now grabs Giant itself.

21 years agoTry a lot harder to get dependancies right. This involves some ugly
Mark Murray [Sun, 27 Jul 2003 13:17:31 +0000 (13:17 +0000)]
Try a lot harder to get dependancies right. This involves some ugly
looking ${.OBJDIR} work that has the up-side of actually working
in upgrade and make -jN cases.

This needs to be revisited further, and it is conceivable that
the ${.OBJDIR} stuff can be simplified, but the sheer number of
edge cases and other causes make this Hard(tm). For now, this works.

21 years agoFix some off-by-one errors dealing with limits of server names,
Tim J. Robbins [Sun, 27 Jul 2003 11:41:38 +0000 (11:41 +0000)]
Fix some off-by-one errors dealing with limits of server names,
usernames, workgroup names and passwords. We can now connect to
servers with 15-character NetBIOS names. (Some versions of Windows
use semi-random 15-char names by default.)

PR: 46902

21 years agoReserve space for the trailing null byte in the srvname member of
Tim J. Robbins [Sun, 27 Jul 2003 11:36:00 +0000 (11:36 +0000)]
Reserve space for the trailing null byte in the srvname member of
struct smb_vc_info.

PR: 46902

21 years agoAdd the -n maxrec option as an alias for -maxrec for compatibility with
Tim J. Robbins [Sun, 27 Jul 2003 10:53:28 +0000 (10:53 +0000)]
Add the -n maxrec option as an alias for -maxrec for compatibility with
System V and consistency with other utilities. Document the new form
instead of the old form in the manual page.

PR: 54661

21 years agoUse pmap_zero_page() to zero pages instead of bzero() because
Maxime Henrion [Sun, 27 Jul 2003 10:41:33 +0000 (10:41 +0000)]
Use pmap_zero_page() to zero pages instead of bzero() because
they haven't been vm_map_wire()'d yet.

21 years agoSimplify sigwait code a bit by using a waitset and removing oldsigmask.
David Xu [Sun, 27 Jul 2003 06:46:34 +0000 (06:46 +0000)]
Simplify sigwait code a bit by using a waitset and removing oldsigmask.

Reviewed by: deischen

21 years agoAllow vm_object_reference() on kernel_object without Giant.
Alan Cox [Sun, 27 Jul 2003 05:43:58 +0000 (05:43 +0000)]
Allow vm_object_reference() on kernel_object without Giant.

21 years agoChanged the type of the variable `qidx' from u_int8_t to int,
Robert Drehmel [Sun, 27 Jul 2003 00:13:49 +0000 (00:13 +0000)]
Changed the type of the variable `qidx' from u_int8_t to int,
mainly to quiet a warning emitted by GCC 3.3 about comparing
a variable to a value which is larger than the former can hold.

The value was checked to make sure the `np->squeue' array is
not accessed behind its boundary.
This worked due to possibly accidental truncation when
(np->squeueput + 1) was larger than or equal to MAX_START (256)
when it was assigned to `qidx'.

`qidx' is used to hold the next position in the start queue
for an insertion.  The new type was chosen because some other
code in the function ncr_freeze_devq() also uses plain integers
to hold those indices.

Wrapped the line after the closing parenthesis of an `if'
condition.

21 years agoUse M_WAITOK instead of M_WAIT in sppp_attach().
Gary Jennejohn [Sat, 26 Jul 2003 21:58:06 +0000 (21:58 +0000)]
Use M_WAITOK instead of M_WAIT in sppp_attach().

21 years agoAcquire Giant rather than asserting it is held in contigmalloc(). This is
Alan Cox [Sat, 26 Jul 2003 21:48:46 +0000 (21:48 +0000)]
Acquire Giant rather than asserting it is held in contigmalloc().  This is
a prerequisite to removing further uses of Giant from UMA.

21 years agoAdd notice about the local change in r1.25
Martin Blapp [Sat, 26 Jul 2003 21:29:28 +0000 (21:29 +0000)]
Add notice about the local change in r1.25

21 years agoFix a couple of bugs in the resume handler. Don't call the if_init
John Polstra [Sat, 26 Jul 2003 20:54:57 +0000 (20:54 +0000)]
Fix a couple of bugs in the resume handler.  Don't call the if_init
function unless the device is configured up.  Without this fix, the
device ends up in the RUNNING state even though it is configured down.
Also, check the RUNNING flag before calling the if_start function, in
case the if_init function failed for one reason or another.

21 years agoDefine the stop command to do nothing when only a specific
Mike Makonnen [Sat, 26 Jul 2003 18:58:46 +0000 (18:58 +0000)]
Define the stop command to do nothing when only a specific
interface is to be stopped. In this case we don't want to go
beyond releasing the lease.

21 years agoWhen stopping a dhcp interface, don't just release the lease on
Mike Makonnen [Sat, 26 Jul 2003 18:50:53 +0000 (18:50 +0000)]
When stopping a dhcp interface, don't just release the lease on
the address, also kill the dhclient process. Instead of doing the
release in the stop command, move it to the precmd stage and allow
rc.subr(8) to automatically kill the dhclient process by leaving the
stop command undefined.

Noticed by: mbr

21 years agoAdd Socket Communications Low Power 10/100 CF Ethernet card.
Warner Losh [Sat, 26 Jul 2003 17:52:09 +0000 (17:52 +0000)]
Add Socket Communications Low Power 10/100 CF Ethernet card.

Obtained from: NetBSD (information, not code)

21 years agoSync to 1.61 of pccarddevs
Warner Losh [Sat, 26 Jul 2003 17:46:43 +0000 (17:46 +0000)]
Sync to 1.61 of pccarddevs

21 years agoMFNetBSD:
Warner Losh [Sat, 26 Jul 2003 17:45:45 +0000 (17:45 +0000)]
MFNetBSD:
1.183: gmcgarry; Add Sharp PALDIO 611S
1.184: martin; Socket LP 10/100 CF

Obtained from: NetBSD

21 years agoo Fix usage(): remove '-l', add missed '-f', sort.
Maxim Konovalov [Sat, 26 Jul 2003 15:29:10 +0000 (15:29 +0000)]
o Fix usage(): remove '-l', add missed '-f', sort.

21 years agoInline a function that gcc refused to inline. This function was used
Hartmut Brandt [Sat, 26 Jul 2003 14:57:26 +0000 (14:57 +0000)]
Inline a function that gcc refused to inline. This function was used
only in one place and it just served as semantic sugar.

21 years agoSilence a gcc-warning. Do this by inlining the macro-call. This is
Hartmut Brandt [Sat, 26 Jul 2003 14:20:37 +0000 (14:20 +0000)]
Silence a gcc-warning. Do this by inlining the macro-call. This is
not very nice - the compiler should just silently optimize away the
unused else clause.

21 years agoThe /rescue system is now enabled by default again, so remove note
Simon L. B. Nielsen [Sat, 26 Jul 2003 13:04:40 +0000 (13:04 +0000)]
The /rescue system is now enabled by default again, so remove note
saying that it is disabled.

Approved by: ceri (mentor)

21 years agoSet mc_len to sizeof(mcontext_t), otherwise it is an invalid context.
David Xu [Sat, 26 Jul 2003 12:58:28 +0000 (12:58 +0000)]
Set mc_len to sizeof(mcontext_t), otherwise it is an invalid context.

21 years agoRemove references to the '-l' option in synopsis. The rest of the
Simon L. B. Nielsen [Sat, 26 Jul 2003 12:57:56 +0000 (12:57 +0000)]
Remove references to the '-l' option in synopsis.  The rest of the
description of this option was removed in v. 1.22.

PR: docs/54880
Submitted by: Lukas Ertl <l.ertl@univie.ac.at>
Approved by: ceri (mentor)

21 years agoMinor mdoc(7) cleanup, based on the PR below.
Simon L. B. Nielsen [Sat, 26 Jul 2003 12:41:44 +0000 (12:41 +0000)]
Minor mdoc(7) cleanup, based on the PR below.

PR: docs/54826
Submitted by: Lukas Ertl <l.ertl@univie.ac.at>
Reviewed by: ru
Approved by: ceri (mentor)

21 years agoFix "write from stdin".
Pierre Beyssac [Sat, 26 Jul 2003 12:14:58 +0000 (12:14 +0000)]
Fix "write from stdin".

Reviewed by: sos
MFC after: 1 week

21 years agoFixed wrong function names in the error message that was translated to
Bruce Evans [Sat, 26 Jul 2003 11:53:40 +0000 (11:53 +0000)]
Fixed wrong function names in the error message that was translated to
C in rev.1.32 so that it doesn't take 2 commits for all branches to fix
this message.

21 years agoRemove prototype of ia64_pa_access(). The function has been moved to
Marcel Moolenaar [Sat, 26 Jul 2003 10:13:30 +0000 (10:13 +0000)]
Remove prototype of ia64_pa_access(). The function has been moved to
mem.c where it's been made static.

21 years agoEnable wake up GPE before shutdown, not only for sleeping.
Takanori Watanabe [Sat, 26 Jul 2003 09:54:17 +0000 (09:54 +0000)]
Enable wake up GPE before shutdown, not only for sleeping.

21 years agoAvoid using __aligned(16). Instead define the jmp_buf in terms of
Marcel Moolenaar [Sat, 26 Jul 2003 08:03:43 +0000 (08:03 +0000)]
Avoid using __aligned(16). Instead define the jmp_buf in terms of
long doubles. This gives us 16-byte alignment. Add a CTASSERT for
the size of the jmp_buf to detect ABI breakages.

21 years agoAdd a "int fd" argument to VOP_OPEN() which in the future will
Poul-Henning Kamp [Sat, 26 Jul 2003 07:32:23 +0000 (07:32 +0000)]
Add a "int fd" argument to VOP_OPEN() which in the future will
contain the filedescriptor number on opens from userland.

The index is used rather than a "struct file *" since it conveys a bit
more information, which may be useful to in particular fdescfs and /dev/fd/*

For now pass -1 all over the place.

21 years agoUnbreak ia64 builds now -Werror is enabled again. Avoid obsolete
Marcel Moolenaar [Sat, 26 Jul 2003 07:23:25 +0000 (07:23 +0000)]
Unbreak ia64 builds now -Werror is enabled again. Avoid obsolete
memory operand construct.

21 years agoGuard against MLEN growing larger than a uint8_t due to MSIZE grwoing to a
Scott Long [Sat, 26 Jul 2003 07:23:24 +0000 (07:23 +0000)]
Guard against MLEN growing larger than a uint8_t due to MSIZE grwoing to a
value of 512 in LINT.  This keeps gcc from complaining.

21 years agoBah, I go and get all preachy, and then commit from the wrong branch. Revert
Scott Long [Sat, 26 Jul 2003 06:47:40 +0000 (06:47 +0000)]
Bah, I go and get all preachy, and then commit from the wrong branch.  Revert
the use of '-j8'.

Apologies to: ru

21 years agoFix the release build with a forgotten '-f'.
Scott Long [Sat, 26 Jul 2003 05:57:52 +0000 (05:57 +0000)]
Fix the release build with a forgotten '-f'.

21 years agoFixes for 64 bit cleanliness. The length arg to sysctl is a pointer to
Peter Wemm [Sat, 26 Jul 2003 04:03:18 +0000 (04:03 +0000)]
Fixes for 64 bit cleanliness.  The length arg to sysctl is a pointer to
size_t not int.

Reviewed by: bp

21 years agoUse #include <string.h> rather than <strings.h> so that the strdup()
Peter Wemm [Sat, 26 Jul 2003 04:02:22 +0000 (04:02 +0000)]
Use #include <string.h> rather than <strings.h> so that the strdup()
prototype gets brought into scope.  This is a big deal for 64 bit
systems where the default return value of 'int' is != pointer.

Reviewed by: bp

21 years agoGulp ... call kmem_malloc() without Giant.
Alan Cox [Sat, 26 Jul 2003 03:55:32 +0000 (03:55 +0000)]
Gulp ... call kmem_malloc() without Giant.

21 years agoFix some recently introduced warnings. 'Declaration does not declare
Peter Wemm [Sat, 26 Jul 2003 03:47:56 +0000 (03:47 +0000)]
Fix some recently introduced warnings.  'Declaration does not declare
anything' etc.

21 years agoTurn -Werror back on again. I've tested with/without
Peter Wemm [Sat, 26 Jul 2003 03:46:39 +0000 (03:46 +0000)]
Turn -Werror back on again.  I've tested with/without
invariants/witness/etc on i386, sparc64, amd64 and alpha for GENERIC.
Lint probably still needs fixing, as do a couple of other drivers
that have broken recently and not been noticed.

21 years agoFix typo.
David Xu [Sat, 26 Jul 2003 02:36:50 +0000 (02:36 +0000)]
Fix typo.

21 years agoChange the inline limit switch to something that exists in gcc-3.2.
Peter Wemm [Sat, 26 Jul 2003 02:27:50 +0000 (02:27 +0000)]
Change the inline limit switch to something that exists in gcc-3.2.
Unfortunately, it has different units.

21 years agoRevise and improve ntfs_subr.c 1.30: read only a single cluster at a time
Tim J. Robbins [Sat, 26 Jul 2003 02:21:49 +0000 (02:21 +0000)]
Revise and improve ntfs_subr.c 1.30: read only a single cluster at a time
in ntfs_writentvattr_plain and ntfs_readntvattr_plain, and purge the boot
block from the buffer cache if isn't exactly one cluster long. These two
changes work around the same buffer cache bug that ntfs_subr.c 1.30 tried
to, but in a different way. This may decrease throughput by reading smaller
amounts of data from the disk at a time, but may increase it by avoiding
bogus writes of clean buffers.
Problem (re)reported by Karel J. Bosschaart on -current.

21 years agoRevert previous commit. We don't use setjmp()/longjmp() for context
Marcel Moolenaar [Fri, 25 Jul 2003 22:36:48 +0000 (22:36 +0000)]
Revert previous commit. We don't use setjmp()/longjmp() for context
switching anymore, so there's no need to save and restore GP. This
change breaks threaded applications linked against libc_r. Pull the
tier 2 card again: relink. This will link against libthr instead.

21 years agorevision 1.51 of vm/uma_core.c modified uma_large_malloc() to acquire
Alan Cox [Fri, 25 Jul 2003 22:26:43 +0000 (22:26 +0000)]
revision 1.51 of vm/uma_core.c modified uma_large_malloc() to acquire
Giant when needed.

21 years agopowerpc isn't tinderbox-ready yet.
Dag-Erling Smørgrav [Fri, 25 Jul 2003 22:17:28 +0000 (22:17 +0000)]
powerpc isn't tinderbox-ready yet.

21 years agoUse __FBSDID().
David E. O'Brien [Fri, 25 Jul 2003 21:19:19 +0000 (21:19 +0000)]
Use __FBSDID().

Brought to you by: a boring talk at Ottawa Linux Symposium

21 years agoUse __FBSDID().
David E. O'Brien [Fri, 25 Jul 2003 21:10:19 +0000 (21:10 +0000)]
Use __FBSDID().

Brought to you by: a boring talk at OLS

21 years agoAdd support for the M_ZERO flag to contigmalloc().
Maxime Henrion [Fri, 25 Jul 2003 21:02:25 +0000 (21:02 +0000)]
Add support for the M_ZERO flag to contigmalloc().

Reviewed by: jeff

21 years agoMake this 64 bit clean. Use size_t for sysctl() length pointer args.
Peter Wemm [Fri, 25 Jul 2003 20:56:57 +0000 (20:56 +0000)]
Make this 64 bit clean. Use size_t for sysctl() length pointer args.

21 years agoRemove alpha vtophys() hack from if_bgereg.h and clean up unneeded
Bill Paul [Fri, 25 Jul 2003 20:33:43 +0000 (20:33 +0000)]
Remove alpha vtophys() hack from if_bgereg.h and clean up unneeded
#includes formerly required to pull in vtophys support, since we
don't need them anymore.

21 years agoConvert bge(4) to use busdma. I have not tested this on anything
Bill Paul [Fri, 25 Jul 2003 19:42:44 +0000 (19:42 +0000)]
Convert bge(4) to use busdma. I have not tested this on anything
besides x86 yet since I don't have access to a non-x86 FreeBSD
box at the moment. Volunteers welcome.

21 years agoFix for 64 bit platforms. sysctl's length args are pointers to
Peter Wemm [Fri, 25 Jul 2003 19:17:46 +0000 (19:17 +0000)]
Fix for 64 bit platforms.  sysctl's length args are pointers to
size_t, not int.  This could be fatal where size_t is long.

Reviewed by: bp

21 years agoMFi386 revision 1.416
Alan Cox [Fri, 25 Jul 2003 18:58:39 +0000 (18:58 +0000)]
MFi386 revision 1.416
 Add vm object locking to pmap_prefault().

Note: powerpc and sparc64 do not implement this function.

21 years agoDiscussions on src-committers with alfred and bde pointed out that
Garance A Drosehn [Fri, 25 Jul 2003 18:40:36 +0000 (18:40 +0000)]
Discussions on src-committers with alfred and bde pointed out that
the "do-nothing" versions of __RCSID(), __RCSID_SOURCE(), __SCCSID(),
and __COPYRIGHT() were not strictly correct.  They should not expand
into [nothing], because the ';' which follows them would then cause
a syntax error (in a strict C compiler, if not gcc...).

So, change the do-nothing versions of those macros to use the
'struct __hack' tactic, as was already used with __FBSDID().

Approved by: discussions with bde
MFC after: 1 week

21 years agoFix amd64 build by deleting trailing whitespace. Doh!
Peter Wemm [Fri, 25 Jul 2003 18:22:53 +0000 (18:22 +0000)]
Fix amd64 build by deleting trailing whitespace.  Doh!

21 years agoState that in -exec ... ; the semicolon should be quoted if invoked from
Jens Schweikhardt [Fri, 25 Jul 2003 17:32:43 +0000 (17:32 +0000)]
State that in -exec ... ; the semicolon should be quoted if invoked from
a shell.

PR: docs/54667
Submitted by: Patrick Alken <pa59@cornell.edu>
MFC after: 3 days

21 years agoFix output from an error message.
Nate Lawson [Fri, 25 Jul 2003 17:11:15 +0000 (17:11 +0000)]
Fix output from an error message.
Use sysctl -n instead of sed

Submitted by: Scott Lambert <lambert@lambertfam.org>

21 years agoAdd wrapper for kqueue() to keep track of the allocated fd and allow it to
Mark Peek [Fri, 25 Jul 2003 17:02:33 +0000 (17:02 +0000)]
Add wrapper for kqueue() to keep track of the allocated fd and allow it to
be closed. This fixes a file descriptor leak when closing a kqueue() fd.

Reviewed by: deischen
MFC after: 1 week

21 years agoAdd an informational debugging printf of the maximum time spent in
Nate Lawson [Fri, 25 Jul 2003 16:49:46 +0000 (16:49 +0000)]
Add an informational debugging printf of the maximum time spent in
EcEventWait().  If you get AE_HARDWARE_NO_RESPONSE errors, please enable
this info by setting hw.acpi.verbose=1 in loader.conf.

21 years agoBacked out the micro-optimization in 1.4. It was to help gcc-2.6.3
Bruce Evans [Fri, 25 Jul 2003 15:54:23 +0000 (15:54 +0000)]
Backed out the micro-optimization in 1.4.  It was to help gcc-2.6.3
on i486's (and probably i386's), but it has had very little effect
since gcc-2.7 or gcc-2.95.  With gcc-3.3, it gave a small
pessimization for at least i386's, athlon-xp's and pentium4's, a
small optimization (I think) for pentium1's, and made no difference
for i386's.  (movzbl is best for all the later processors, and the
micro-optimization was to stop it being used on i486's.)

21 years agoNew release notes: bge(4) 5705 support.
Hiroki Sato [Fri, 25 Jul 2003 14:08:32 +0000 (14:08 +0000)]
New release notes:  bge(4) 5705 support.

Update release notes:  rescue(7) man page.

MFC:  safe(4).

21 years agoMake the PC98 small boot floppy usable again.
Ruslan Ermilov [Fri, 25 Jul 2003 13:39:11 +0000 (13:39 +0000)]
Make the PC98 small boot floppy usable again.

Spotted by: nyan
Reviewed by: nyan

21 years agoWhen deciding whether to download the microcode or not look at the API rather
Hartmut Brandt [Fri, 25 Jul 2003 12:40:03 +0000 (12:40 +0000)]
When deciding whether to download the microcode or not look at the API rather
than at the vendor. We have three different Fore cards and only the PCA200
need the microcode. Look also at the RAM address and load the code only if
it is not zero. A zero RAM address means either a bug in the driver or
this is a interface created by harp(4) in which case fatm(4) handles the
microcode issue.

21 years agoPrint the offending SPANS message only if printing is enabled.
Hartmut Brandt [Fri, 25 Jul 2003 12:32:08 +0000 (12:32 +0000)]
Print the offending SPANS message only if printing is enabled.

21 years agoAn u_int8_t can never be bigger than 255, so remove a useless check.
Maxime Henrion [Fri, 25 Jul 2003 12:23:25 +0000 (12:23 +0000)]
An u_int8_t can never be bigger than 255, so remove a useless check.

Spotted by: GCC

21 years agoRemove a duplicate entry.
Maxime Henrion [Fri, 25 Jul 2003 11:14:03 +0000 (11:14 +0000)]
Remove a duplicate entry.

21 years agoo Do not overwrite saved interrupt priority level by alloc_hash(),
Maxim Konovalov [Fri, 25 Jul 2003 09:59:16 +0000 (09:59 +0000)]
o Do not overwrite saved interrupt priority level by alloc_hash(),
  use a separate variable.
o Restore interrupt priority level before return (no-op in HEAD).

Spotted by: Don Bowman <don@sandvine.com>
MFC after: 5 days

21 years agoThe RELEASE9 knob is now OBE.
Ruslan Ermilov [Fri, 25 Jul 2003 08:55:51 +0000 (08:55 +0000)]
The RELEASE9 knob is now OBE.

OK'ed by: jhb

21 years agoMoved the boot floppy generation code out from release.9 to release.10.
Ruslan Ermilov [Fri, 25 Jul 2003 08:54:41 +0000 (08:54 +0000)]
Moved the boot floppy generation code out from release.9 to release.10.
The release.9 target is now responsible only for generation of MFS root
file systems, that are built for all architectures, even those that do
not provide a floppy installation option.  The release.10 target is now
responsible for creation of a set of boot, MFS root, and fixit floppies,
and the NO_FLOPPIES variable now affects only this target.

Also, replaced the FIXIT_TARGET variable with a check of whether the
*FIXITSIZE variables are present for a given architecture, similar to
how this is done when generating boot floppies.

Discussed with: jhb

21 years agoAdd support for VBR and CBR PVCs for IP over ATM.
Hartmut Brandt [Fri, 25 Jul 2003 08:35:26 +0000 (08:35 +0000)]
Add support for VBR and CBR PVCs for IP over ATM.

Submitted by: Vincent Jardin <vjardin@wanadoo.fr>
MFC after: 2 weeks

21 years agoMake library WARNS=6 clean. The problems have been: alignment on sparc64
Hartmut Brandt [Fri, 25 Jul 2003 08:22:08 +0000 (08:22 +0000)]
Make library WARNS=6 clean. The problems have been: alignment on sparc64
and one of the usual sizeof(in_addr_t) == sizeof(u_long) bugs.

21 years agoDue to a gcc bug, it doesn't like local variables with names like 'sin'.
Hartmut Brandt [Fri, 25 Jul 2003 08:15:09 +0000 (08:15 +0000)]
Due to a gcc bug, it doesn't like local variables with names like 'sin'.
Rename this for the moment. Also fix a sparc64 alignment warning.

21 years agoMake ilmid WARNS=6 clean. The problem were a couple of unused function
Hartmut Brandt [Fri, 25 Jul 2003 08:09:18 +0000 (08:09 +0000)]
Make ilmid WARNS=6 clean. The problem were a couple of unused function
arguments and missing consts.

21 years agoRemove __aligned(16) from the definition of struct _ia64_fpreg. It's
Marcel Moolenaar [Fri, 25 Jul 2003 08:02:24 +0000 (08:02 +0000)]
Remove __aligned(16) from the definition of struct _ia64_fpreg. It's
a non-standard construct. Instead, redefine struct _ia64_fpreg as a
union and put a long double in it. On ia64 and for LP64, this is
defined by the ABI to have 16-byte alignment. For ILP32 a long double
has 4-byte alignment, but we don't support ILP32.

Note that the in-memory image of a long double does not match the in-
memory image of spilled FP registers. This means that one cannot use
the fpr_flt field to interpet the bits. For this reason we continue
to use an aggregate type.

21 years agoRemove INVARIANT* and WITNESS. This makes the simulator much more
Marcel Moolenaar [Fri, 25 Jul 2003 07:52:20 +0000 (07:52 +0000)]
Remove INVARIANT* and WITNESS. This makes the simulator much more
pleasant to use.