The sparc64/sparc64/pmap.c implementations of pmap_remove(),
pmap_protect(), and pmap_copy() have optimizations for regions
larger than PMAP_TSB_THRESH (which works out to 16MB). This
caused a panic in tsb_foreach for kernel mappings, since
pm->pm_tsb is NULL in that case. This fix teaches tsb_foreach
to use the kernel's tsb in that case.
Connect up a QUEUE FULL event with CAM and adjust openings.
Unfortunately, the QUEUE FULL event only tells you Bus && Target.
It doesn't tell you lun. In order for the XPT_REL_SIMQ action to
work, we have to have a real lun. But which one? For now, just
iterate over MPT_MAX_LUNS.
Practically speaking, this is only going to be happening for lower
quality SAS or SATA drives behind the SAS controller, which means
only lun 0, so it's not so bad.
Helpful Reminder Nagging from: John Baldwin, Fred Whiteside
andre [Thu, 21 Sep 2006 19:14:51 +0000 (19:14 +0000)]
Move the initialization of the hardware capabilities in em_init_locked()
before em_setup_transmit_structures() as it needs this information to
properly set up TSO parameters.
atomic_fetchadd_int is used by mb_free_ext(), but it
returns the previous value that the "add" effected (In
this case we are adding -1), afterwhich we compare it
to '0'... to see if we free the mbuf... we should
be comparing it to '1'... Note that this only effects
when there is contention since there is a first part
to the comparison that checks to see if its '1'. So
this bug would only crop up if two CPU's are trying
to free the same mbuf refcount at the same time. This
will happen in SCTP but I doubt can happen in TCP or
UDP.
PR: N/A
Submitted by: rrs
Reviewed by: gnn,sam
Approved by: gnn,sam
Vendor import of OpenBSM 1.0 alpha 11, with the following change history
notes since the last import:
OpenBSM 1.0 alpha 11
- Reclassify certain read/write operations as having no class rather than the
fr/fw class; our default classes audit intent (open) not operations (read,
write).
- Introduce AUE_SYSCTL_WRITE event so that BSD/Darwin systems can audit reads
and writes of sysctls as separate events. Add additional kernel
environment and jail events for FreeBSD.
- Break AUDIT_TRIGGER_OPEN_NEW into two events, AUDIT_TRIGGER_ROTATE_USER
(issued by the user audit(8) tool) and AUDIT_TRIGGER_ROTATE_KERNEL (issued
by the kernel audit implementation) so that they can be distinguished.
- Disable rate limiting of rotate requests; as the kernel doesn't retransmit
a dropped request, the log file will otherwise grow indefinitely if the
trigger is dropped.
- Improve auditd debugging output.
- Fix a number of threading related bugs in audit_control file reading
routines.
- Add APIs au_poltostr() and au_strtopol() to convert between text
representations of audit_control policy flags and the flags passed to
auditon(A_SETPOLICY) and retrieved from auditon(A_GETPOLICY).
- Add API getacpol() to return the 'policy:' entry from audit_control, an
extension to the Solaris file format to allow specification of policy
persistent flags.
- Update audump to print the audit_control policy field.
- Update auditd to read the audit_control policy field and set the kernel
policy to match it when configuring/reconfiguring. Remove the -s and -h
arguments as these policies are now set via the configuration file. If a
policy line is not found in the configuration file, continue with the
current default of setting AUDIT_CNT.
- Fix bugs in the parsing of large execve(2) arguments and environmental
variable tokens; increase maximum parsed argument and variable count.
- configure now detects strlcat(), used by policy-related functions.
- Reference token and record sample files added to test tree.
Replace system call thr_getscheduler, thr_setscheduler, thr_setschedparam
with rtprio_thread, while rtprio system call is for process only, the new
system call rtprio_thread is responsible for LWP.
Introduce a new method ipv6if which attemptes to figure out if an
interface is an IPv6 interface.
Use this method to decide if we should attempt to configure an interface
with an IPv6 address in pccard_ether. The mechanism pccard_ether uses
to do this is unsuited to the task because it assumes the list of
interfaces it is passed is the full list of IPv6 interfaces and makes
decissions based on that. This is at least a step in the right
direction and is probably about as much as we can MFC safely.
kan [Thu, 21 Sep 2006 01:37:02 +0000 (01:37 +0000)]
Use __builtin_va_start instead of __builtin_stdarg_start. GCC4 obsoletes
the former and __builtin_va_start was present in all GCC version 3.1 and
later.
In the spirit of nanoBSD and PicoBSD, add TinyBSD.
Submitted by jmeloatfreebsdbrasil,com-br
(Jean Milanez Melo)
As PicoBSD becomes slightly less useful, TinyBSD fills the gap below nanoBSD.
Based on The Open Group Base Specifications Issue 6 IEEE Std 1003.1, our
current implementation of df(1) is does not properly format the output under
certain conditions. Right now -kP and -Pk are not the same thing. Further,
when we set the BLOCKSIZE environment variable, we use "1k" instead of "1024",
making the header display incorrectly.
To quote the specification:
"When both the -k and -P options are specified, the following header line
shall be written (in the POSIX locale):
"Filesystem 1024-blocks Used Available Capacity Mounted on\n"
- If -P has been specified, check to make sure that -k has not already been
specified, if so, simply break instead of clobbering the previous blocksize
- Use 1024 instead of 1k to make the header POSIX compliant
Flushing all IPv4 routes when an interface is removed or unconfigured
makes no sense. Remove the undocumented removable_route_flush feature
from pccard_ether.
Add a new 'bce_mgmt_init_locked' function to enable the minimal parts
of the chip to let ASF/IPMI firmware to respond to IPMI after attaching
and when the chip is down. David looked at it but could really say
what they right minimal config. stuff would be. It's not documented.
I figured this out via trial and error.
Allow hw.mfi.event_locale/hw.mfi.event_class to be set via loader.
If an event doesn't match the criteria then don't print it. Some
events are not saved in the log (<0 class events).
Bring the i386 linux mmap code more into line with how linux (2.4.x)
behaves. This fixes a lot of test which failed before. For amd64 there
are still some problems, but without any testers which apply patches
and run some predefines tests we can't do more ATM.
Submitted by: Marcin Cieslak <saper@SYSTEM.PL> (minor fixups by myself)
Tested with: LTP
Remove MAC_DEBUG + MPRINTF debugging from System V IPC. This no longer
appears to be serving a useful purpose, as it was used during initial
development of MAC support for System V IPC.
MFC after: 1 month
Obtained from: TrustedBSD Project
Suggested by: Christopher dot Vance at SPARTA dot com
Remove MAC_DEBUG label counters, which were used to debug leaks and
other problems while labels were first being added to various kernel
objects. They have outlived their usefulness.
MFC after: 1 month
Suggested by: Christopher dot Vance at SPARTA dot com
Obtained from: TrustedBSD Project
Rather than allocating all buffer memory for the completed BSM record
when allocating the record in the first place, allocate the final buffer
when closing the BSM record. At that point, more size information is
available, so a sufficiently large buffer can be allocated.
This allows the kernel to generate audit records in excess of
MAXAUDITDATA bytes, but is consistent with Solaris's behavior. This only
comes up when auditing command line arguments, in which case we presume
the administrator really does want the data as they have specified the
policy flag to gather them.
Obtained from: TrustedBSD Project
MFC after: 3 days
Revert back to always using *(int *)arg for now. While this
is incorrect, and causes endianness bugs on 64-bit big-endian
machines (sparc64), it's the best choice for now, as many of
these IOCTLs are used inside the kernel, and bogusly pass an
argument as "int *" which results in unaligned access panics
on sparc64 when attempting to dereference them via *(intptr_t *).
(Several of us are working on a real fix, which is uneasy.)
Fix the glitch introduced in rev. 1.93. In softdep_sync_metadata(),
switch by worklist type contains two for() loops, for D_INDIRDEP and
D_PAGEDEP. On error, these loops are exited by break, where the switch
actually shall be leaved. Use goto instead of break to reach the error
handling code.
Add missing parentheses to fix a segmentation fault that is easily
reproducable with `jot -s " " 400 1 | column -t'. The bug was present
in the the original CSRG 'column -t' added in 1989.
Fix races between tty.c and sessrele() / doenterpgrp() / leavepgrp(). The tty
code is still under giant lock, but the session/pgrp release code just used
proctree_locks. This explains why moving the proctree_lock in sys/kern/tty.c
rev. 1.258 did fix the panics in our SMP systems.
This should also fix some race panics with revoked ttys.
Fix the bug in rev. 1.134. In devfs_allocv_drop_refs(), when not_found == 2
and drop_dm_lock is true, no unlocking shall be attempted. The lock is
already dropped and memory is freed.
Found with: Coverity Prevent(tm)
CID: 1536
Approved by: pjd (mentor)
Fix the buggy rev. 1.117. dagmembers are only initialized for dlopen'ed
dso that are actually loading. If dso a.so depends on b.so, then dlsym
with handle from dlopen("b.so") will fail unconditionally.
Correct implementation shall use the Obj_Entry.needed list to walk
dependencies DAG.
Test provided by: jkim
Tested (prev. version) by: jkim, Nicolas Blais <nb_root at videotron ca>, h.blanke at chello nl
Pointy hat to: kib
Approved by: kan (mentor)
Back out revision 1.8. The only thing it did was to break LEDs (and
some other less noticeable things) on little-endian 64-bit platforms
such as amd64.
PR: kern/101931
Discussed with: emax, marius
MFC after: 1 day
Do not try to call keyboard callback unless keyboard is active and busy.
This should fix 'kbdcontrol -K < /dev/console' panic on sparc64 with sunkbd(4).
sam [Mon, 18 Sep 2006 16:49:15 +0000 (16:49 +0000)]
o move ath hal os glue code from the hal to the driver: this code was
part of the hal distribution early on when the hal was built for
each os but it's been portable for a long time so move the os-specific
code out (and off the vendor branch)
o correct the copyright on ah_osdep.?; it was mistakenly given a
restricted license and not a dual-bsd/gpl license
o remove the module api definition as it was never used
o fixup include paths for move of ah_osdep.h