are pending I/O transactions. It is not clear that is works 100% of
the time under SMP, but since the bt_cmds() that are sent after other
CPUs are started are not critical, the driver will function until I
can figure out why this is the case.
Oh, I forgot to mention: this driver also works on FreeBSD/alpha (big
thanks to Andrew Gallatin). And there is a 2.2.x version available for
those who stubbornly refuse to upgrade.
Networks Tigon 1 and Tigon 2 chipsets. There are a _lot_ of OEM'ed
gigabit ethernet adapters out there which use the Alteon chipset so
this driver covers a fair amount of hardware. I know that it works with
the Alteon AceNIC, 3Com 3c985 and Netgear GA620, however it should also
work with the DEC/Compaq EtherWORKS 1000, Silicon Graphics Gigabit
ethernet board, NEC Gigabit Ethernet board and maybe even the IBM and
and Sun boards. The Netgear board is the cheapest (~$350US) but still
yields fairly good performance.
Support is provided for jumbo frames with all adapters (just set the
MTU to something larger than 1500 bytes), as well as hardware multicast
filtering and vlan tagging (in conjunction with the vlan support in
-current, which I should merge into -stable soon). There are some hooks
for checksum offload support, but they're turned off for now since
FreeBSD doesn't have an officially sanctioned way to support checksum
offloading (yet).
I have not added the 'device ti0' entry to GENERIC since the driver
with all the firmware compiled in is quite large, and it doesn't really
fit into the category of generic hardware.
Repeatedly-tripped-over-by: Vallo Kallaste <vallo@matti.ee>
When VINUMDEBUG is set, free any memory found still
allocated.
Only log errors if DEBUG_EXITFREE is set.
board versions with no BIOS. Separate mailbox interrupts from
IOCB interrupts. Read OUTMAILBOX5 while RISC_INT is active- not
after you clear it (potential race condition). Clear out older broken
BIG_ENDIAN goop. Don't negotiate narrow/async for LVD busses at startup
if already in LVD mode. Note usage of presumptive 1040C revision. For
all the LIP, PDB Changed, Loop UP/DOWN async events, mark fw state
as unknown as well as marking the need to do a getpdb on targets- after
a LIP for certain the f/w has to do PRLI/PLOGI for all targets again
and marking f/w state as unknown gives us a fighting chance to (start
to) hold up for that to complete.
that will SBusify an isp header or the lun/target portions of a request IOCB-
and have these only valid iff __sparc__ (no non-sparc SBus machine that *I*
know about).
free_vinum: Wait for daemon to stop by checking the
vinum_conf.flags & VF_DAEMONOPEN.
vinum_modevent:
When compiled with VINUMDEBUG, check if we have
forgotten to free any memory, and log an error if we
have.
vinumopen: Allow open of an empty subdisk (otherwise we can't
initialize it).
plexes_used and volumes_used. Now these fields are only informative,
and the <object>_allocated count is used for searches, etc. This also
required checking the object state before doing things with the
presumed object.
Problems-reported-by: Kiril Mitev <kiril@ideaglobal.com>
VINUM_<object>CONFIG: return ENXIO rather than EFAULT if an object
doesn't exist.
plexes_used and volumes_used. Now these fields are only informative,
and the <object>_allocated count is used for searches, etc. This also
required checking the object state before doing things with the
presumed object.
Problems-reported-by: Kiril Mitev <kiril@ideaglobal.com>
vinum_scandisk: increment drive use count when we find a good one.
plexes_used and volumes_used. Now these fields are only informative,
and the <object>_allocated count is used for searches, etc. This also
required checking the object state before doing things with the
presumed object.
Problems-reported-by: Kiril Mitev <kiril@ideaglobal.com>
Remove unused (and braindead) functions volume_index, plex_index,
sd_index and drive_index.
Add a flag VF_CREATED for volumes. VF_NEWBORN was being used in two
capacities, and they clashed, my Lord, they clashed.
find_object: restructure the search loop as a result of the change in
variable use.
Decrement object use count in the remove_<object> functions, not in
the free_<object> functions, which are often called with partially
initialized (and uncounted) objects.
plexes_used and volumes_used. Now these fields are only informative,
and the <object>_allocated count is used for searches, etc. This also
required checking the object state before doing things with the
presumed object.
Problems-reported-by: Kiril Mitev <kiril@ideaglobal.com>
Well, better late than newer, but things has been hectic
around here, sorry for the long delay.
DMA support has been added to the ATA disk driver.
This only works on Intel PIIX3/4, Acer Aladdin and Promise controllers.
The promise support works without the BIOS on the board,
and timing modes are set to support up to UDMA speed. This
solves the problems with having more than one promise controller
in the same system.
There is support for "generic" DMA, that might work on other
controllers, but now you have been warned :)
More chipset specific code will come soon, I have to find testers
with the approbiate HW, more on that when I have it ready.
The system now uses its own major numbers, please run MAKEDEV
with the devices you need (ad?, acd?, afd?, ast?).
For now the disk driver will also attach to the old wd major
so one can at least boot without this step, but be warned, this
will eventually go away. The bootblocks will have to be changed
before one can boot directly from an "ad" device though.
Fixed problems:
All known hang problems should be solved
The probe code has been sligthly changed, this should solve
the reports I have lying around (I hope).
Hangs when accessing ata & atapi device on the same channel simultaniously.
A real braino in ata_start caused this, fixed.
As usual USE AT YOUR OWN RISK!!, this is still pre alpha level code.
Especially the DMA support can hose your disk real bad if anything
goes wrong, agaiin you have been warned :)
But please tell me how it works for you!
Enjoy!
-Søren
longjmp. I suspect that the occasional double panic may be the result
of incorrect parameters to longjmp. This happens, of course, like the
entire file, only with -DVINUMDEBUG.
give_sd_to_plex: Don't set Raid-5 subdisk state here.
config_subdisk: handle the name parameter correctly when the subdisk
was referenced in a previous plex definition. The
name parameter must come first.
Handle autosizing relatively correctly. There is
still a danger of losing drive space if problems
occur with an autosized subdisk.
Set state to empty, not up, when complete. This also
solves a nagging problem about enforcing the need to
initialize RAID-5 plexes.
config_plex: handle the name parameter correctly when the plex
was referenced in a previous volume definition. The
name parameter must come first.
Handle initial state better.
update_plex_config:
Calculate the trim factor for RAID-5 plexes correctly.
Set the number of down subdisks correctly when reading
from disk config.
settings you've just sent them and return random values if you follow
the set by a get. This causes problems when you latter run a Tag-enabled
command when you've command tagged mode off.
Recognize aic7895 controllers that have been "acquired" by a RAIDPort
card as normal aic7895s.
Recognize the aic7815 Raid Parity/Memory controller chip and notify
the user that it's RAID functionality will be ignored.
Don't mess with the IRQMS bit in the host control register unless
we are an aic7770 chip.
Use calling context to determine if the card is already paused when
we update the target message request bit field in controller scratch
ram. Looking at the paused bit in the HCNTRL register opened up a
race condition.
Insert delays in the target message request update routine as a temporary
work around for what looks like a chip bug. I'm still investigating this
one.
Fix the Abort/Abort Tag/BDR handler to pull its message from the message
buffer in our softc instead of attempting to get it from a register on
the controller. The message is never recorded by the controller in the
new message scheme.
Don't rely on having an SCB when a BDR occurs. We can issue these during
invalid reconnects to.
Fix a few cases where we were restarting the sequencer but then still
falling out of a switch statement to unpause the sequencer again.
This could cause us to mess up sequencer state if it generated another
pausing interrupt between the time of the restart and unpause.
Kill the 'transceiver settle' loop during card initialization. I
failed to realize that a controller that is not connected to any
cables will never settle or enable the SCSI transceivers at all.
The correct solution is to monitor the IOERR interrupt which indicates
that the transceiver state has changed (UW<->LVD).
Modify the aic7xxx assembler to properly echo input when stdin is not
a tty.
remove the splbio() around the call to launch read requests.
launch_requests:
Move the splbio() protection outside the entire launch_loop. The
previous location was causing problems with IDE drives, where the
call to the strategy routine often did not complete until after
complete_rqe deallocated the request structure.
Solution-independently-found-by: Russell Neeper <r-neeper@tamu.edu>
Problem-reported-by: Vallo Kallaste <vallo@matti.ee>
John Saunders <john@nlc.net.au>
Bernd Walter <ticso@cicely.de> (maybe)
Check for partition types FS_VINUM and FS_UNUSED. Accept both, but
complain about FS_UNUSED. At a later date, only FS_VINUM will be
accepted.
Threatened-since: over a year
Add a flag `force' (VF_FORCECONFIG) to force name changes of
existing drives.
config_drive:
If the drive already has a vinum label, and name doesn't match the
specified drive, do it anyway if the 'force' flag is specified.
finish_config:
Reset the `force' flag.
Continually-tripped-over-by: Karl Pielorz <kpielorz@tdx.co.uk>
give_sd_to_drive:
If the drive is down, take the subdisk down and don't try to fix
things.
update_plex_config:
Don't try to update the config parameters of a plex which isn't
fully configured (state plex_init or plex_unallocated).
Correctly calculate the amount to trim off a striped or RAID-5 plex
whose size is not a multiple of the stripe size.
compiled with or without debugging support. This enables us to catch
(fatal) mismatches between the kernel and userland.
Coalesce flags VINUM_DISKCONFIG and VINUM_READING_CONFIG. They did
essentially the same thing.
Add VINUM_BIGDRIVE for pretending we have macho hardware.
pretends that each drive is 100 times as large as it really is. Not
for use at home.
Coalesce flags VINUM_DISKCONFIG and VINUM_READING_CONFIG. They did
essentially the same thing.
This solved a problem where 'vinum resetconfig' only reset half
the drives.
Reported-by: Brad Knowles <blk@skynet.be>
Karl Pielorz <kpielorz@tdx.co.uk>
Change the super device. We now have three super devices:
1. The normal superdevice used by vinum(8).
2. The superdevice used by vinum(8) when compiled with debug support.
3. The superdevice used by the daemon.
This method allows vinum(8) to determine debug mismatches. Also check
correctly for the device type. The old code did not check all bits of
the minor number.
Reported-by: a cast of thousands, most recently by Brad Knowles
<blk@skynet.be>.
gave yet another internal register layout model for what is
*still* the same architecture. I hope they saved billyuns of gates
'coz otherwise this is *really* annoying.
thwank in register layout goop). A different mboxcmd approach. Some PDB change
infrastructure. Some better management of loopdown/loopup events (keep them
distinct from resource starvation for simq freeze/unfreeze actions).
The old VN device broke in -4.x when the definition of B_PAGING
changed. This patch fixes this plus implements additional capabilities.
The new VN device can be backed by a file ( as per normal ), or it can
be directly backed by swap.
Due to dependencies in VM include files (on opt_xxx options) the new
vn device cannot be a module yet. This will be fixed in a later commit.
This commit delimitted by tags {PRE,POST}_MATT_VNDEV
chip int. and ext. clock synchronisation). Fixed workaround for
transmit threshold underrun. Added volatile keyword to CSR_READ_* and
CSR_WRITE_* macroses. Added DELAYs to eliminate randomness caused
by processor speed. Fixed all TXCON and RXCON registers to be accessed
only when chip is idle, as manual told. Changed epic_init_phy to
drop link by isolating and going loopback, should should force link
partner to restart autonegotiation.
PR: kern/10535, kern/9742, kern/10575
Submitted by: Peter Jeremy, David Greenman
a wierd double-queue arrangement.. It always empties the if_snd queue
then puts the transmit packets into a different queue that is limited
by the number of TX descriptors and does it's own discards...
This should stop the boot-time XXX warning anyway.
