Matthew Dillon [Tue, 23 Jun 2009 23:10:34 +0000 (16:10 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Tue, 23 Jun 2009 23:08:26 +0000 (16:08 -0700)]
kernel diskmbr handling - Detect maxed out slice ds_size field.
If the DOS slice field is found to be maxed out (0xFFFFFFFFU), use the
actual media size for calculations instead of ds_size. This allows
disks > 2TB to have a conventional slice table and still contain a
disklabel64 which covers the actual size of the media.
Matthew Dillon [Tue, 23 Jun 2009 23:07:15 +0000 (16:07 -0700)]
disklabel - expand warning if the media is too large.
Instead of just warning that the media is too large and exiting, also
now suggest that disklabel64 be used.
Matthew Dillon [Tue, 23 Jun 2009 23:05:51 +0000 (16:05 -0700)]
fdisk - allow media sizes larger then 2TB
If the media size is larger then 2TB set the 32 bit slice logical
size field, ds_size, to its maximum value of 0xFFFFFFFFU instead of
overflowing the field.
Sascha Wildner [Tue, 23 Jun 2009 18:38:00 +0000 (20:38 +0200)]
Add sili(4) and drm(4) manpages (from OpenBSD and NetBSD, respectively).
Thanks to Hasso Tepper for helping with the drm(4) one.
Sascha Wildner [Tue, 23 Jun 2009 06:36:58 +0000 (08:36 +0200)]
make upgrade: Remove GNU cpio's info file, too.
Hasso Tepper [Mon, 22 Jun 2009 19:14:14 +0000 (22:14 +0300)]
Make all G4X chips use the new framecount register.
The G45 docs indicate that all G4X chips use the new framecount register.
Intel agrees with my reading of the docs, make it so for all G4X chips.
The new register also has a 32 bit width as opposed to 24 bits. Fix
things up so that the counters roll over properly.
Obtained-from: FreeBSD
Hasso Tepper [Mon, 22 Jun 2009 19:12:25 +0000 (22:12 +0300)]
krealloc() behaves identically to kmalloc when passed a NULL object pointer.
If an error does occur we would have left max_context with an incorrect
value.
Obtained-from: FreeBSD
Hasso Tepper [Mon, 22 Jun 2009 19:11:37 +0000 (22:11 +0300)]
Don't panic if drm_rmmap is called with a NULL map pointer.
Obtained-from: FreeBSD
Hasso Tepper [Mon, 22 Jun 2009 19:09:32 +0000 (22:09 +0300)]
Don't try to setup interrupts for drivers that don't support them.
This causes sis and probably a couple of other drivers to panic and fail.
Obtained-from: FreeBSD
Hasso Tepper [Mon, 22 Jun 2009 18:55:03 +0000 (21:55 +0300)]
Check offsets for R300_ZB_ZPASS_ADDR.
Obtained-from: FreeBSD
Hasso Tepper [Mon, 22 Jun 2009 18:53:56 +0000 (21:53 +0300)]
Add regs required for occlusion queries support.
Obtained-from: FreeBSD
Hasso Tepper [Mon, 22 Jun 2009 18:51:42 +0000 (21:51 +0300)]
Add support for RV790 (HD 4890) asics.
Obtained-from: FreeBSD
Hasso Tepper [Mon, 22 Jun 2009 18:50:04 +0000 (21:50 +0300)]
Radeon microcode handling cleanup.
A little more cleanup from AMD, if we don't have the right microcode
there is no reason to mess with the chip.
Obtained-from: FreeBSD
Sepherosa Ziehau [Mon, 22 Jun 2009 10:42:34 +0000 (18:42 +0800)]
MP table: Don't mix lapic/ioapic/imcr configuration information extraction
Sepherosa Ziehau [Mon, 22 Jun 2009 10:24:17 +0000 (18:24 +0800)]
Merge mptable_check() into mptable_probe()
Sepherosa Ziehau [Sun, 21 Jun 2009 08:34:34 +0000 (16:34 +0800)]
Use madt_iterate_entries() in madt_pass2()
Sepherosa Ziehau [Sun, 21 Jun 2009 07:58:53 +0000 (15:58 +0800)]
ACPI MADT: When locating lapic address, take Local APIC Address Override
into consideration.
Sepherosa Ziehau [Sun, 21 Jun 2009 07:25:36 +0000 (15:25 +0800)]
ACPI MADT: Check existance of BSP in madt_check()
Sepherosa Ziehau [Sun, 21 Jun 2009 06:53:21 +0000 (14:53 +0800)]
ACPI MADT: Add madt_iterate_entries()
Sepherosa Ziehau [Sun, 21 Jun 2009 05:41:28 +0000 (13:41 +0800)]
Use mptable_iterate_entries() in mptable_pass2()
Sepherosa Ziehau [Sun, 21 Jun 2009 04:37:23 +0000 (12:37 +0800)]
Pull IMCR setting out of mptable_pass2()
Sepherosa Ziehau [Sun, 21 Jun 2009 04:30:00 +0000 (12:30 +0800)]
Let mptable_pass2() call mptable_default()
Sepherosa Ziehau [Sun, 21 Jun 2009 04:17:05 +0000 (12:17 +0800)]
mptable_default() only configures ioapic stuffs
Sepherosa Ziehau [Sun, 21 Jun 2009 04:08:50 +0000 (12:08 +0800)]
Use mptable_iterate_entries() in mptable_pass1()
Sepherosa Ziehau [Sat, 20 Jun 2009 12:49:49 +0000 (20:49 +0800)]
MP table: Remove global variables for HT fixup
Sepherosa Ziehau [Sat, 20 Jun 2009 12:18:53 +0000 (20:18 +0800)]
NAPICID 16 -> 256.
This also makes sure that we could actually do 32 (MAXCPU) CPUs
instead of 16.
Sepherosa Ziehau [Sat, 20 Jun 2009 11:19:52 +0000 (19:19 +0800)]
mptable_lapic_enumerate(): Fix comment
Sepherosa Ziehau [Sat, 20 Jun 2009 10:59:08 +0000 (18:59 +0800)]
Rearrange mptable_lapic_enumerate() by factoring out mptable_lapic_default()
Sepherosa Ziehau [Sat, 20 Jun 2009 10:29:27 +0000 (18:29 +0800)]
Add APIC_ID to extract apic id from local apic id field
Sepherosa Ziehau [Sat, 20 Jun 2009 10:18:51 +0000 (18:18 +0800)]
Add lapic_init() to map local apic
Sepherosa Ziehau [Sat, 20 Jun 2009 09:35:58 +0000 (17:35 +0800)]
MP table: Pull lapic related information extraction into seperate function
Sepherosa Ziehau [Sat, 20 Jun 2009 03:53:36 +0000 (11:53 +0800)]
MP table: Do preliminary checks before claiming that the MP table is valid
Sepherosa Ziehau [Fri, 19 Jun 2009 13:40:55 +0000 (21:40 +0800)]
Put mp_nbusses and bus_data under APIC_IO
Sepherosa Ziehau [Fri, 19 Jun 2009 12:19:18 +0000 (20:19 +0800)]
Staticize mp_nbusses
Sepherosa Ziehau [Fri, 19 Jun 2009 11:54:15 +0000 (19:54 +0800)]
Remove bsp_apic_ready
Sepherosa Ziehau [Wed, 17 Jun 2009 11:54:22 +0000 (19:54 +0800)]
Correct the logical_cpus calculation in mptable_hyperthread_fixup()
The original code extracts logical_cpus directly from the output of
cpuid(eax=1), however,
According to Intel's #253668, section 7.7
"(CPUID.1:EBX[23:16]) -- Indicates the maximum number of addressable
ID for logical processors in a physical package."
According to AMD's #25481 CPUID Fn0000_0001_EBX
"LogicalProcessorCount is the number of threads per CPU core times the
number of CPU cores per processor."
So what we originally saved in logical_cpus is actually the upper limit
of the logical CPUs per phyical CPU package. It may work correctly if
there is only one logical CPU per physical CPU package detected in MP
table. But if more than one logical CPUs per phyical CPU package are
detected in the MP table, then the missing logical CPUs will not be
detected later, since the code assumed that logical_cpus is the number
of logical CPUs missed between two detected logical CPUs.
We try to fix the problem by calculating the distances between two nearest
APIC IDs. If all such distances for all detect logical CPUs are same,
then the logical_cpus is set the calculated distance. The value extracted
from cpuid(eax=1) just serves as the upper limit of the logical_cpus now.
Reported-and-Tested-by: Robert Luciani <rluciani@gmail.com>
Peter Avalos [Mon, 22 Jun 2009 02:06:50 +0000 (16:06 -1000)]
Merge branch 'vendor/LIBARCHIVE'
Peter Avalos [Mon, 22 Jun 2009 02:04:39 +0000 (16:04 -1000)]
Add README.DELETED to libarchive's vendor branch.
Peter Avalos [Sun, 21 Jun 2009 22:00:54 +0000 (12:00 -1000)]
Merge branch 'vendor/LIBARCHIVE'
Peter Avalos [Sun, 21 Jun 2009 21:48:22 +0000 (11:48 -1000)]
Merge branch 'vendor/LIBARCHIVE' (early part)
Peter Avalos [Sun, 21 Jun 2009 21:45:01 +0000 (11:45 -1000)]
Rename libarchive-2/ to libarchive/. No need for it to be versioned.
Peter Avalos [Sun, 21 Jun 2009 21:44:30 +0000 (11:44 -1000)]
Remove old versions of libarchive.
Peter Avalos [Sun, 21 Jun 2009 21:32:18 +0000 (11:32 -1000)]
Remove GNU cpio.
Sascha Wildner [Sun, 21 Jun 2009 21:24:35 +0000 (23:24 +0200)]
pnpinfo(8): Oops, add missing semicolons.
Sascha Wildner [Sun, 21 Jun 2009 19:06:56 +0000 (21:06 +0200)]
diff(1): Commit our patches to contrib/ and get rid of -I-.
Peter Avalos [Sun, 21 Jun 2009 18:42:58 +0000 (08:42 -1000)]
Merge branch 'master' of /home/www-data/gitweb/dragonfly
Peter Avalos [Sun, 21 Jun 2009 18:40:42 +0000 (08:40 -1000)]
Merge branch 'vendor/LESS'
Peter Avalos [Sun, 21 Jun 2009 18:31:16 +0000 (08:31 -1000)]
Upgrade to less-429.
* The -D option with one number now means use the normal background color.
* Don't change permissions on history file if it is not a regular file.
* Fix non-ANSI-compliant code that caused problems with some compilers.
* Fix binary file detection in UTF-8 mode.
* Fix display problems with long lines on "ignaw" terminals.
* Fix problem interrupting the line number calculation for initial prompt.
* Fix SGR emulation when dealing with multiple attributes (eg.
bold+underline).
* Fix highlight bug when searching for underlined/overstruck text.
* New "&" command allows filtering of lines based on a pattern.
* Status column now displays a search match, even if the matched string is
scrolled off screen because -S is in effect.
* Improve behavior of -F option.
* Allow CSI character (0x9B) to work in UTF-8 mode.
* Output carriage return at startup in case terminal doesn't default to
column 1.
* Fix bug in '' (quote, quote) command after G command.
Peter Avalos [Sun, 21 Jun 2009 18:27:17 +0000 (08:27 -1000)]
Add README.DELETED to the vendor/LESS branch.
Peter Avalos [Sun, 21 Jun 2009 18:26:30 +0000 (08:26 -1000)]
Merge branch 'vendor/LESS'
Sascha Wildner [Sun, 21 Jun 2009 18:08:31 +0000 (20:08 +0200)]
pnpinfo(8): isolation_protocol() takes no arguments.
While here, clean up a bit, fix a couple of warnings and raise WARNS to 6.
Peter Avalos [Sun, 21 Jun 2009 18:05:12 +0000 (08:05 -1000)]
Move less-4/ to less/. No need for versioned directories.
Peter Avalos [Sun, 21 Jun 2009 18:02:47 +0000 (08:02 -1000)]
Remove old versions of less.
Peter Avalos [Sun, 26 Apr 2009 04:46:03 +0000 (04:46 +0000)]
Add MD, optimized versions of string functions for amd64.
Obtained-from: FreeBSD
Peter Avalos [Sun, 26 Apr 2009 04:23:41 +0000 (04:23 +0000)]
Add amd64 assembly versions of div(3), ldiv(3), and lldiv(3).
Obtained-from: FreeBSD
Peter Avalos [Sun, 26 Apr 2009 03:57:31 +0000 (03:57 +0000)]
Update some of the amd64 parts of libc:
* Change the syscall stub branch orders so that the static branch
prediction will assume that syscalls will succeed rather than fail.
* Add error handling for PSEUDO syscalls.
* Add END() macros for Assembler code. This causes the symbol
table to include the size of the function.
* ldexp: Fix a typo in the !__GNUC__ case and remove an obsolete
comment.
* setjmp/longjmp changes:
-Save and restore the control part of the MXCSR in addition to the
i387 control word to ensure that the two are consistent.
-Don't clobber the FPU exception flags in longjmp. C99 requires them
to remain unchanged.
Obtained-from: FreeBSD
Peter Avalos [Fri, 24 Apr 2009 04:10:45 +0000 (04:10 +0000)]
Sync with i386's _fpmath.h:
* Define LDBL_NBIT to be a mask indicating the position of the integer
bit in a long double.
* Add an alternative view of the bits in an 80-bit long double (64+16
instead of 32+32+15+1).
Obtained-from: FreeBSD
Peter Avalos [Sun, 21 Jun 2009 02:43:11 +0000 (16:43 -1000)]
Define an END() macro for use in i386 and amd64 assembly code.
Obtained-from: FreeBSD
Peter Avalos [Sun, 21 Jun 2009 02:35:27 +0000 (16:35 -1000)]
Remove dead code (_ARCH_INDIRECT) from <machine/asm.h>.
Peter Avalos [Sun, 21 Jun 2009 08:07:22 +0000 (08:07 +0000)]
Allow pre-2.3 systems to build the new version of file.
Add a few things under COMPILE_ONLY so systems without strtof and/or
daylight can still build the host tool.
Matthew Dillon [Sun, 21 Jun 2009 02:40:31 +0000 (19:40 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sun, 21 Jun 2009 02:29:57 +0000 (19:29 -0700)]
HAMMER VFS - Implement inode number matching against the directory hash
Change hammer_create_inode() and hammer_alloc_objid() to take the upper
10 bits of the filename hash and attempt to use them as the lower 10 bits
of the inode number.
hammer_alloc_objid() now utilizes a small two-level 1024-bit bitmap to
assign the low 10 bits of the inode number, on a per-directory basis.
If a collision occurs a nearby value is used instead.
The idea is for inode numbers to have approximately the same linearity
as the directory read scan order, so programs such as cpio, tar, cpdup,
find, and ls which do recursive scans of the directory tree get much
better locality of reference for stat() and other related system calls.
Based-on: Numerous discussions with people, esp corecode.
Matthew Dillon [Sun, 21 Jun 2009 02:25:26 +0000 (19:25 -0700)]
cpdup - Do not create a .tmp file and rename if the target does not exist.
* When copying if the target does not exist use the target's actual name
instead of the name of a temporary file.
This helps filesystems (aka HAMMER) which use hashed directories. Inode
numbers can be selected by the FS on the basis of the file-name in
order to maintain directory-order linearity. If the copy program creates
the target file with one name and then renames it to the real name,
however, the feature is lost and we get one big scramble again.
Peter Avalos [Sun, 21 Jun 2009 00:48:42 +0000 (14:48 -1000)]
Merge branch 'vendor/FILE'
Peter Avalos [Sat, 20 Jun 2009 23:57:23 +0000 (13:57 -1000)]
Upgrade to file-5.03.
Changes can be viewed in the ChangeLog file.
Matthew Dillon [Sat, 20 Jun 2009 23:53:21 +0000 (16:53 -0700)]
HAMER VFS - Cache additional b-tree nodes in directory.
* Cache inode record cursors in the directory at dip->cache[2] to reduce
cpu overhead when iterating a directory.
* Cache file object record cursors in the file's governing directory at
dip->cache[3] after reading the last data record in a file, to reduce
cpu overhead when chain-reading files.
* Also add vfs.hammer.stats_btree_root_iterations, a counter which
is incremented whenever a B-Tree search is initiated from the root
node.
Peter Avalos [Sat, 20 Jun 2009 22:20:03 +0000 (12:20 -1000)]
Merge branch 'vendor/FILE'
Peter Avalos [Sat, 20 Jun 2009 22:18:24 +0000 (12:18 -1000)]
Add README.DELETED to file's vendor branch.
Peter Avalos [Sat, 20 Jun 2009 22:07:26 +0000 (12:07 -1000)]
Merge branch 'vendor/FILE'
Peter Avalos [Sat, 20 Jun 2009 22:00:13 +0000 (12:00 -1000)]
Merge branch 'vendor/FILE' (early part)
Peter Avalos [Sat, 20 Jun 2009 21:57:20 +0000 (11:57 -1000)]
Move file-4/ to file/. We don't need a versioned directory.
Peter Avalos [Sat, 20 Jun 2009 21:56:41 +0000 (11:56 -1000)]
Remove old versions of file.
Matthew Dillon [Sat, 20 Jun 2009 21:49:12 +0000 (14:49 -0700)]
HAMMER VFS - Add hinting capability to block allocator, hint B-Tree
* A hammer_off_t can now be supplied to the blockmap allocator as a hint.
* Use the hinting mechanism to better-localize B-Tree node allocations
and meta-data updates.
Matthew Dillon [Sat, 20 Jun 2009 21:07:33 +0000 (14:07 -0700)]
iostat - Increase the tps field width from 3 to 4.
* Standard workloads now easily push the tps rate into the 4 digits.
Increase the tps field with to 4.
Peter Avalos [Sat, 20 Jun 2009 21:04:55 +0000 (11:04 -1000)]
Merge branch 'vendor/OPENSSH'
Peter Avalos [Sat, 20 Jun 2009 18:55:46 +0000 (08:55 -1000)]
Upgrade to OpenSSH-5.2p1.
Security:
* This release changes the default cipher order to prefer the AES CTR
modes and the revised "arcfour256" mode to CBC mode ciphers that are
susceptible to CPNI-957037 "Plaintext Recovery Attack Against SSH".
* This release also adds countermeasures to mitigate CPNI-957037-style
attacks against the SSH protocol's use of CBC-mode ciphers. Upon
detection of an invalid packet length or Message Authentication
Code, ssh/sshd will continue reading up to the maximum supported
packet length rather than immediately terminating the connection.
This eliminates most of the known differences in behaviour that
leaked information about the plaintext of injected data which formed
the basis of this attack. We believe that these attacks are rendered
infeasible by these changes.
New features:
* Added a -y option to ssh(1) to force logging to syslog rather than
stderr, which is useful when running daemonised (ssh -f)
* The sshd_config(5) ForceCommand directive now accepts commandline
arguments for the internal-sftp server.
* The ssh(1) ~C escape commandline now support runtime creation of
dynamic (-D) port forwards.
* Support the SOCKS4A protocol in ssh(1) dynamic (-D) forwards.
(bz#1482)
* Support remote port forwarding with a listen port of '0'. This
informs the server that it should dynamically allocate a listen
port and report it back to the client. (bz#1003)
* sshd(8) now supports setting PermitEmptyPasswords and
AllowAgentForwarding in Match blocks
Bug and documentation fixes
* Repair a ssh(1) crash introduced in openssh-5.1 when the client is
sent a zero-length banner (bz#1496)
* Due to interoperability problems with certain
broken SSH implementations, the eow@openssh.com and
no-more-sessions@openssh.com protocol extensions are now only sent
to peers that identify themselves as OpenSSH.
* Make ssh(1) send the correct channel number for
SSH2_MSG_CHANNEL_SUCCESS and SSH2_MSG_CHANNEL_FAILURE messages to
avoid triggering 'Non-public channel' error messages on sshd(8) in
openssh-5.1.
* Avoid printing 'Non-public channel' warnings in sshd(8), since the
ssh(1) has sent incorrect channel numbers since ~2004 (this reverts
a behaviour introduced in openssh-5.1).
* Avoid double-free in ssh(1) ~C escape -L handler (bz#1539)
* Correct fail-on-error behaviour in sftp(1) batchmode for remote
stat operations. (bz#1541)
* Disable nonfunctional ssh(1) ~C escape handler in multiplex slave
connections. (bz#1543)
* Avoid hang in ssh(1) when attempting to connect to a server that
has MaxSessions=0 set.
* Multiple fixes to sshd(8) configuration test (-T) mode
* Several core and portable OpenSSH bugs fixed: 1380, 1412, 1418,
1419, 1421, 1490, 1491, 1492, 1514, 1515, 1518, 1520, 1538, 1540
* Many manual page improvements.
Peter Avalos [Sat, 20 Jun 2009 18:29:52 +0000 (08:29 -1000)]
Add README.DELETED to the OpenSSH's vendor branch.
Peter Avalos [Sat, 20 Jun 2009 18:21:37 +0000 (08:21 -1000)]
Merge branch 'vendor/OPENSSH'
Peter Avalos [Sat, 20 Jun 2009 18:14:52 +0000 (08:14 -1000)]
Merge branch 'vendor/OPENSSH' (early part)
Peter Avalos [Sat, 20 Jun 2009 18:01:56 +0000 (08:01 -1000)]
Move openssh-5/ to openssh/. We don't need a versioned directory.
Peter Avalos [Sat, 20 Jun 2009 18:01:16 +0000 (08:01 -1000)]
Remove old versions of OpenSSH.
Sascha Wildner [Thu, 18 Jun 2009 21:58:29 +0000 (23:58 +0200)]
libmd: Raise WARNS to 2 and fix a warning.
Sascha Wildner [Thu, 18 Jun 2009 23:00:38 +0000 (01:00 +0200)]
Raise WARNS level of several libraries.
Sascha Wildner [Sat, 20 Jun 2009 17:41:08 +0000 (19:41 +0200)]
Add sys/boot/pc32/boot0sio to the build.
Also mention it in boot0cfg.8 (taken from FreeBSD).
While here, also do some minor cleanup of the manual page.
Matthew Dillon [Sat, 20 Jun 2009 17:31:39 +0000 (10:31 -0700)]
HAMMER UTILITY - reorg cache, add -C, document blockmap and show command.
* Reorganize the way the local block cache works (in the utility, not the
filesystem). Add the -C option to specify cache size and read-ahead.
* Document the blockmap command and enhance the documentation for the show
command.
Matthew Dillon [Sat, 20 Jun 2009 00:04:31 +0000 (17:04 -0700)]
HAMMER (and kernel) - Fix cpu-bound kernel thread issue.
* HAMMER now calls a new kernel function, lwkt_user_yield(), in its
ioctl-based loops (such as the reblocker).
* Add a new LWKT function called lwkt_user_yield(). This function
causes a kernel thread to yield at user priority (instead of kernel
priority).
This function also deals with a nasty issue related to the MP lock.
A cpu-bound kernel thread holding the MP lock can prevent other
cpus from serving interrupt threads which also need the MP lock.
Detect the condition and release the MP lock for 10uS to give the
other cpus a chance to pick it up. This is a bad hack but it
actually does work.
* Move passive_release() from MD code to kern/lwkt_thread.c and add
an inline for the passive release recovery function. Adjust all
platforms to use the new API instead of rolling the same code for
each platform.
Reported-by: Many, but especially Hasso Tepper <hasso@estpak.ee>
Matthew Dillon [Fri, 19 Jun 2009 17:25:12 +0000 (10:25 -0700)]
HAMMER UTIL - Add -C cachesize option to improve the 'show' command.
* Allows the default in-memory cache size of 16m to be changed.
Matthew Dillon [Fri, 19 Jun 2009 17:20:35 +0000 (10:20 -0700)]
HAMMER UTIL - Add minor verboseness to "show"
Matthew Dillon [Fri, 19 Jun 2009 16:04:22 +0000 (09:04 -0700)]
AHCI - Correct prdbc warning.
* This field is only updated by hardware for non-NCQ packets, don't
spew warnings if it is zero for NCQ packets.
Reported-by: Sepherosa Ziehau <sepherosa@gmail.com>
Matthew Dillon [Fri, 19 Jun 2009 07:02:23 +0000 (00:02 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Fri, 19 Jun 2009 06:59:19 +0000 (23:59 -0700)]
SILI - Minor fixes and adjustments.
* Aligned scratch space for READ LOG
* Better exclusivity code. Minor fix to only copy-back the rfis
if the CCB was on-chip, instead of unconditionally.
* Probe all ports reported by the PM, we'll have to adjust for
'fake' disks at the end of the PM port chain more specifically.
Matthew Dillon [Fri, 19 Jun 2009 06:49:07 +0000 (23:49 -0700)]
AHCI - Intel fixes, error processing fixes.
* Allocate separate scratch space to enforce a reasonable alignment
for the READ LOG page.
* Fix numerous issues with READ LOG / NCQ error processing. READ LOG
wasn't working at all before.
* Intel fixes. Move the SDBE interrupt from the AsyncNotify-only
code to the general code. Intel parts seem to use this interrupt
(and not DPE) to signal completion for NCQ commands, so we get
timeouts if we don't use it generally.
* Rearrange a great deal of code in a manner similar to how the
Sili driver was arranged. Primarily: ATA_F_AUTOSENSE to
acquire D2H rfis's more reliably, and ATA_F_EXCLUSIVE. More work
needs to be done. Also move the PM probe code to the PM source
module.
* Add code to deal with D2H ERR fises. These generate a DHRS
interrupt but only those which post an ERR in the task file
stop command processing. This is not well documented in the
AHCI specs.
Sascha Wildner [Thu, 18 Jun 2009 13:39:50 +0000 (15:39 +0200)]
Sync zoneinfo database with tzdata2009j from elsie.
asia: 8.34 -> 8.35
As Bangladesh's start of DST on June 19 is now official, simplify the
last change.
Sascha Wildner [Thu, 18 Jun 2009 13:38:24 +0000 (15:38 +0200)]
Remove trailing whitespace.
Michael Neumann [Thu, 18 Jun 2009 10:32:07 +0000 (12:32 +0200)]
priv: Narrow down privileges
Michael Neumann [Thu, 18 Jun 2009 10:01:19 +0000 (12:01 +0200)]
priv: Narrow down privileges
Leave PRISON_ROOT for now despite having a separate privilege for
the prison case. PRISON_ROOT will be removed in a later step, once
all jail privileges have been determined.
Michael Neumann [Thu, 18 Jun 2009 09:56:32 +0000 (11:56 +0200)]
priv: Narrow down privileges
Matthew Dillon [Thu, 18 Jun 2009 07:56:32 +0000 (00:56 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Thu, 18 Jun 2009 07:52:45 +0000 (00:52 -0700)]
SILI - Fix port re-insertion probe problems when coming out of DET_INIT
* When a port is pulled we enter into a listen state where we program
SCTL to the DET_INIT state to send a continuous stream of COMRESETs.
When a port is plugged back in we clear the DET_INIT state. However,
we were not waiting nearly long enough for the PHY to renegotiate!
It can take up to 4 seconds!
Add a polling loop to wait for the PHY to renegotiate before we start
the DEVRESET sequence. If we start the sequence too early the PHY
will still be unstable and the DEVRESET sequence or something following
it will likely fail with a fatal error interrupt (usually error code 4).
* Reduce the physical port insertion delay to 7 seconds. Leave the
PM port insertion delay at 10 seconds.