Andrew R. Reiter [Wed, 24 Apr 2002 18:06:18 +0000 (18:06 +0000)]
- Remove atm_attributes_pool and the relating atm_allocate() and atm_free()
calls associated with the pool and the objects allocated out from the
pool.
- Insert atm_attributes_zone which is a uma_zone that is used just as the
atm_attributes_pool was (including the max objects value). Also, used
the appropriate zalloc and zfree's where necesary.
ahc_eisa.c:
ahc_pci.c:
Conform to new aic7xxx IRQ API.
Adapt to aic7xxx_freebsd -> aic7xxx_osm changes.
aic7770.c:
Disable card generated interrupt early in our probe for
"extra safety"
Commonize some seeprom code with the PCI side of the driver.
aic7xxx.c:
Correctly initialize a few scratch ram locations during
a sequencer restart. This avoids spurious sequencer ram
parity errors in some configurations.
Include the softc in ahc_update_residual calls. We need it
for some diagnostics in this code path.
Flag a data overrun on an auto-request sense failure as a
CAM_AUTOSENSE_FAIL rather than a CAM_DATA_RUN_ERR.
Force a renegotiation after noticing a parity error. This
covers targets that lose our negotiation settings but don't
bother to give us a unit attention condition. This can happen
if a target fails during a reselection of us during a cable
pull.
Convert some code to using constants.
Fix some typos.
Correct target mode message loop handling. ahc_clear_msg_state
was not clearing the "need to go to message out phase" bit once
our loop was over.
Simplify some abort handling code.
Include tag information in target mode immediate notify events.
When shutting down EISA controllers, don't EISA BIOS settings in
the high portions of scratch ram. This fixes warm boot issues on
some systems.
Save a bit of space by only allocating the SCBs that we can use.
Avoid some code paths in ahc_abort_scbs() if we are currently
acting as a target.
Correctly cleanup stranded SCBs in the card's SCB array. These
are SCBs who's mapping has already been torn down by code that
aborted the SCB by seeing it in another list first.
Add a comment about some potential bus reset issues for target
mode on Twin (EISA only) controllers.
aic7xxx.h:
Cleanup the hardware scb definitions a bit.
Allocate a ful 256 byte scb mapping index. This simplifies
the lookup code since the table covers all possible (and potentially
bogus) values.
Make AHC_DEBUG work again.
aic7xxx.reg:
Updates to hardware SCB definition.
New definitions for target mode fixes.
aic7xxx.seq:
In target mode, initialize SAVED_LUN just after we receive
the identify message. It may be required in the error recovery
path when a normal cdb packet (includes lun) is not sent up to
the host for processing.
Respond to irregular messages during a selection in target mode.
Defer looking for space for a cdb packet until we are about to
enter command phase. We want to be able to handle irregular messages
even if we would otherwise return QUEUE_FULL or BUSY.
Add support for sending Ignore Wide Residue messages as a target.
In the disable disconnect case in target mode, set our transfer
rate correctly once data are availble.
aic7xxx_93cx6.c:
aic7xxx_93cx6.h:
Add the ability to write and erase the seeprom.
aic7xxx_inline.h:
Correct Big Endian handling of large cdb sizes (> 12 bytes).
Adaptec to changes in the calc_residual API.
Correct a target mode bug where we always attempted to service
the input queue even if no progress could be made due to lack
of ATIOs.
aic7xxx_osm.c:
Adaptec to new IRQ mapping API. The new API allows the core
to only enable our IRQ mapping once it is safe (sufficient
initialization) to do so.
Slap bootverbose protection around some diagnostics.
Only attempt DT phases if we are wide.
aic7xxx_osm.h:
Enable big endian support.
Adjust for IRQ API change.
aic7xxx_pci.c:
Be more careful about relying on subvendor 9005 information.
We now only trust it for HBAs. This should allow the driver
to attach to some MBs where the subvendor/device information
does not follow the Adaptec spec.
Only enable interrupts on the card once we are fully setup.
Disable external SCB ram usage on the aic7895. I have not
been able to make it 100% reliable.
Adjust to seeprom routines being properly prefixed with "ahc".
Fix a few bugs in the external SCB ram probing routine. We
need to clear any parity errors we've triggered during the
probe to avoid future, fatal, interrupts.
If we detect an invalid cable combination, pretent there are
no cable at all. This will enable all of the terminators
which is probably the safest configuration we can "guess".
Warner Losh [Wed, 24 Apr 2002 15:30:11 +0000 (15:30 +0000)]
o Work around bugs in the powerof2 macro: It thinks that 0 is a power of
2, but that's not the case. This fixes the case where there were slots
in the PIR table that had no bits set, but we assumed they did and used
strange results as a result.
o Map invalid INTLINE registers to 255 in pci_cfgreg.c. This should allow
us to remove the bogus checks in MI code for non-255 values.
I put these changes out for review a while ago, but no one responded
to them, so into current they go.
This should help us work better on machines that don't route
interrupts in the traditional way.
Ruslan Ermilov [Wed, 24 Apr 2002 13:03:44 +0000 (13:03 +0000)]
GC bsd.sgml.mk. Nothing in the src/ tree uses it, and two doc/
instances that still use it are unconnected from the build and
have corresponding DocBook replacements.
Continue the cleanup preparations for UFS2 (& GEOM):
Use only one filedescriptor. Open in R/O or R/W based in the '-N' option.
Make the filedescriptor a global variable instead of passing it around
as semi-global variable(s).
Remove the undocumented ability to specify type without '-T' option.
Replace fatal() with straight err(3)/errx(3). Save calls to strerror()
where applicable. Loose the progname variable.
Get the sense of the cpgflag test correct so we only issue warnings if
people specify cpg and can't get that. It can be argued that this
should be an error.
Remove the check to see if the disk is mounted: Open for writing
would fail if it were mounted.
Attempt to get the sectorsize and mediasize with the generic disk
ioctls, fall back to disklabel and /etc/disktab as we can.
Notice that on-disk labels still take precedence over /etc/disktab,
this is probably wrong, but not as wrong as the entire concept of
/etc/disktab is.
Ruslan Ermilov [Wed, 24 Apr 2002 11:26:19 +0000 (11:26 +0000)]
The install.debug and reinstall.debug targets are needed solely
to build kernel and kernel modules so stop supporting them in
bsd.subdir.mk and reimplement them in kern.post.mk and kmod.mk
as special versions of the install and reinstall targets, and
only define them if DEBUG is also defined (when debug versions
are really built).
Scott Long [Wed, 24 Apr 2002 05:12:50 +0000 (05:12 +0000)]
Prepare for a major update to the aac driver:
Update the aac driver with the new crashdump api.
Protect sync fibs with a mutex.
Align all DMA buffers on a PAGE_SIZE boundary.
Add support for loading files other than /boot/loader. Useful if you're
loader is screwed and you want loader.old.
Rewrite the scaled down printf so it actually works right, and add support
for more formats.
Clean up elf loading to not make assumptions about the ordering of sections.
Don't load the symbol table; this is only needed for loading kernels and we
load the loader.
Mike Barcroft [Wed, 24 Apr 2002 01:26:11 +0000 (01:26 +0000)]
Rearrange <netinet/in.h> so that it is easier to conditionalize
sections for various standards. Conditionalize sections for various
standards. Use standards conforming spelling for types in the
sockaddr_in structure.
David E. O'Brien [Tue, 23 Apr 2002 23:59:51 +0000 (23:59 +0000)]
Back out the NO_RCMNDS knob.
1. The committer refused to respond to questions over the commit.
2. The servers rlogind, rshd, rexecd were not wrapped.
3. "rcmnds" as an abbreviation gets an order of magnitude less hits on
Google than the much more well known "rcmds".
David E. O'Brien [Tue, 23 Apr 2002 22:16:41 +0000 (22:16 +0000)]
Change the name of the 'bin' distribution to 'base'.
This is done since it contains much more than /bin, and also gets in the
way when making a combined install+fixit CD.
1) Proberly conditionalize PAM "last login" printout.
2) For "copyright" case #ifdef HAVE_LOGIN_CAP was placed on too big block,
narrow it down.
3) Don't check the same conditions twice (for "copyright" and "welcome"),
put them under single block.
4) Print \n between "copyright" and "welcome" as our login does.
Ruslan Ermilov [Tue, 23 Apr 2002 09:24:00 +0000 (09:24 +0000)]
Do not spam initial ${CHROOTDIR} world with Texinfo and HTML docs.
For rerelease, run "make -DNOCLEAN world" instead of simple and
insufficient in some cases "make all install". This is especially
true for cross-arch "make release"s which we don't (yet) support.
Ruslan Ermilov [Tue, 23 Apr 2002 09:03:56 +0000 (09:03 +0000)]
Merge bsd.obj.mk's version of the _SUBDIR target with bsd.subdir.mk.
Ensure all standard targets honor SUBDIR. Now `make obj' descends into
SUBDIRs even if NOOBJ is set (some descendants may still need an object
directory, but we do not have such precedents). Now `make install' in
non-bsd.subdir.mk makefiles runs `afterinstall' target _after_ `install'
in SUBDIRs, like we do in bsd.subdir.mk. Nothing depended on the wrong
order anyway.
Fixed `distribute' targets (except for the bsd.subdir.mk version) so that
they do not depend on _SUBDIR; `distribute' calls `install' which already
depends on _SUBDIR.
De-standardize `maninstall', otherwise manpages would be installed twice.
(To be revised later.)
Tim J. Robbins [Tue, 23 Apr 2002 07:15:09 +0000 (07:15 +0000)]
Add the -t option, which allows tab stop positions to be specified similarly
to the way expand(1) allows. This brings unexpand(1) up to SUSv3 conformance,
and it now passes all relevant parts of the GNU textutils test suite.
Juli Mallett [Tue, 23 Apr 2002 02:56:16 +0000 (02:56 +0000)]
- printf shouldn't bail out if a conversion fails, it should just keep
processing them.
- \c escape to immediately stop output (similar to echo's \c)
- \0NNN should be allowed for octal character escapes (instead of just \NNN)
- %b conversion, which is like %s but interprets \n \t etc. inside the
string is missing.
And I may not be any poet, but in lieu of an in-tree regression test:
ref5% ./printf '%s%b%b%c%s%d\n' 'PR' '\0072' '\t' '3' '56' 0x10
PR: 35616
Improve the error message the user sees if the startdaemon routine cannot
connect() to the socket for lpd. Tell them this error probably means that
the master 'lpd' process is not running.
Implement new printcap option of "rc" aka "remote.resend_copies".
This is a boolean option, and if it is specified in a print queue
for a remote host, it causes lpd to resend the data file for each
copy the user requested on 'lpr -#n'. This is useful for network
printers which accept lpd-style jobs, but which ignore the control
file (and thus they ignore any request for multiple copies).
PR: 25635
Reviewed by: short review on freebsd-audit
MFC after: 6 days
David E. O'Brien [Tue, 23 Apr 2002 00:05:48 +0000 (00:05 +0000)]
Do not use 'ps -e' for entropy gathering. It uses /proc/*/mem to rummage
around *user* memory to extract the environment variable strings. This
is problematic for us.
Maxime Henrion [Mon, 22 Apr 2002 23:03:03 +0000 (23:03 +0000)]
Do our best to determine if the user is attempting an NFS mount when
the filesystem type isn't given in the command line. In the case of
an IPv6 address containing ':', one must use the '@' separator for it
to be properly parsed (mount_nfs(8) still needs fixing at the moment
though).
John Polstra [Mon, 22 Apr 2002 16:15:16 +0000 (16:15 +0000)]
Add missing splx calls in bge_tick. These don't make any functional
difference in -current, but they are important for -stable where
they are missing too.
Usage style sweep: spell "usage" with a small 'u'.
Also change one case of blatant __progname abuse (several more remain)
This commit does not touch anything in src/{contrib,crypto,gnu}/.
Comment out Kirks io-request priority hack until we can do this in a
civilized way which doesn't cause grief.
The problem is that it is not generally safe to cast a "struct bio
*" to a "struct buf *". Things like ccd, vinum, ata-raid and GEOM
constructs bio's which are not entrails of a struct buf.
Also, curthread may or may not have anything to do with the I/O request
at hand.
The correct solution can either be to tag struct bio's with a
priority derived from the requesting threads nice and have disksort
act on this field, this wouldn't address the "silly-seek syndrome"
where two equal processes bang the diskheads from one edge to the
other of the disk repeatedly.
Alternatively, and probably better: a sleep should be introduced
either at the time the I/O is requested or at the time it is completed
where we can be sure to sleep in the right thread.
The sleep also needs to be in constant timeunits, 1/hz can be practicaly
any sub-second size, at high HZ the current code practically doesn't
do anything.
Add ia64_sal_init_state(). This function will initialize the machine
check handling. In its current form, it only determines the largest
amount of state information it can get from SAL and allocates a region
7 memory block for it.
The next steps involve:
o get and log any unconsumed (NVM stored) error records across
reboots,
o register an OS_MCA handler and enable machine checks.