which is automatically included during kernel config, and add comments
to individual machine-dependant majors.* files to assign new MI majors
in MI file.
Range 0-191 is reserved for machine-specific assignments, range
192+ are MI assignments.
Follows recent discussion on tech-kern@
* Implement read for ulpt.
* If the device is not opened for reading, occasionally drain any
data the printer might have (but don't hammer the printer with reads).
* Lower the buffer size to one page.
The driver seems to work with more printers now.
>- Move PMAGB-BA and PMAG-JA (which I have) out of "may be incorrect" entries.
>- Put driver name "tfb" for PMAG-JA (which will appear in future pmax-wscons).
removing assumptions that there are only two B channels and by
adding support for a varying number of channels.
Due to this, rename previously used isdn identified "bri" to "isdnif",
which better describes the current situation.
* auich_calibrate() assumes that the AC97 part is in its reset state. To
ensure this, call audio_attach_mi() after auich_calibrate().
* Explicitly support 12000Hz and 24000Hz. (Why is there a discrete list at
all?)
* Fix an obvious recording bug -- we were acking the wrong interrupt.
* Ensure that we don't get an interrupt during the AC97 speed probe by clearing
the "interrupt on completion" bit in the DMA setup.
* Separate the code to set the default parameters into a new function,
audio_set_defaults(). Make it use audiosetinfo(), which properly initializes
the block size and whatnot. Use this in both audioattach() and the
/dev/audio case of audio_open().
* Do not force a reinitialization when /dev/sound is opened.
* Do all of the block size sanity checks in auto_init_ringbuffer(), not in
both audio_calc_blksize() and audiosetinfo().
* Fix a bug in audiosetinfo() that caused the block size to not be recalculated
immediately if we set it to 0.
* For AUDIO_GET[IO]OFFS, modify the deltablks calculation so that it gives us
the number of block boundaries crossed.
interface controllers (of varying intelligence levels).
Contributed by Wasabi Systems, Inc. Primarily written by Steve Woodford,
with some modification by me.
to pci_attach_args. The latter is allocated on the stack during auto-
configuration and so will not be valid after that time.
It's amazing how the old code worked for so long. I guess pci_attach_args
is allocated deep in pid#0's kernel stack on most platforms.
fit in the number of bits used for the port still does something.
This fixes PR pkg/18741 for ac97-based hardware. Other audio drivers
might need a similar fix.
- if_bcereg.h should only have register definitions, so declarations
of softc structure etc. should be in if_bce.c.
- The types of DMA descriptors should be u_int32_t, not unsigned long.
- netinet headers are not required here.
- Values passed via bce_tx_ring should also be byte-swapped.
- byte-swapping is not needed for bus_space access.
from me:
- KNF
- mark all magic numbers with /* MAGIC */ so they can be identified at
some point. there are 33 in total though many appear to be the same
value, or related (eg, the 6 emac points.)
leaves these issues:
- RX pre-packet headers need to be byte-swapped or not?
- PAGE_SIZE bytes are allocated for both TX and RX DMA ring descriptors,
but they should be 1024 (== sizeof(struct bce_dma_slot) * N[TR]XDESC).
- the mcast filter.
it should be there, but in practice, on some systems an ugly race condition
rears its head: SCSI commands are issued before the FC thread ever runs, and
fail from then on forever. Yuck.
The driver backend is capable of supporting also ISA cards (no DMA)
and primary rate (PRI) cards in addition to the basic rate ones,
but I don't any to test on right now, so we don't support those
currently.
This code was originally written by Juha-Matti Liukkonen <jml@cubical.fi>
of Cubical Solutions Ltd. for FreeBSD, and was ported to NetBSD by
myself for the same company.
CCB is returned to the caller.
- Make code paths that can use twe_ccb_alloc_wait() use it, and assert
that a CCB is always returned from that function.
- Assert that a CCB is always returned when allocating the reserved CCB
for an AEN fetch.
reserved CCB. This means that all remaining callers of twe_param_get*()
are called from a valid thread context, and thus have no need to use a
reserved CCB. This will allow for further cleanup in a future commit.
1) Don't wait for DRQ on an IDENTIFY command -- if it's not set when we see
BSY clear, abort the command and ignore the drive. (Do this by testing
for DRQ in the read/write cases in __wdccommand_intr().)
2) Don't wait for DRQ to deassert when we finish an IDENTIFY (or any other
non-block command that reads data) -- we don't do this for block I/O, and
empirically it doesn't clear on my CF cards at all, causing a pointless 1s
delay.
3) Add comments to some of the delay()s, and add missing ones in wdcreset()
and the WDCC_RECAL in the so-called "pre-ATA" probe.
4) Slightly simplify the reset sequence -- we were doing an extra I/O.
5) Modify the register writability test to make sure that registers are not
overlapped -- this can happen in some weird cases with a missing device 1.
6) Check the error register value after the reset -- if it's not 01h or 81h,
as appropriate (see ATA spec), punt.
Tested with a number of ATA-only, ATAPI-only, mixed ATA-ATAPI, CF, and IDE
disk configurations.
Also remove the SINGLE_DRIVE nonsense again.
- Make AENs use the generic code table stuff.
- Add a few more AEN codes (from FreeBSD).
- Correct the context of a few AEN codes (some were listed as
"unit context" when they're really "port context").
- Add a queue of AENs that management tools in userspace can poll
(from FreeBSD).
* introduce fsetown(), fgetown(), fownsignal() - this sets/retrieves/signals
the owner of descriptor, according to appropriate sematics
of TIOCSPGRP/FIOSETOWN/SIOCSPGRP/TIOCGPGRP/FIOGETOWN/SIOCGPGRP ioctl; use
these routines instead of custom code where appropriate
* make every place handling TIOCSPGRP/TIOCGPGRP handle also FIOSETOWN/FIOGETOWN
properly, and remove the translation of FIO[SG]OWN to TIOC[SG]PGRP
in sys_ioctl() & sys_fcntl()
* also remove the socket-specific hack in sys_ioctl()/sys_fcntl() and
pass the ioctls down to soo_ioctl() as any other ioctl
change discussed on tech-kern@
of struct wsdisplay_softc. Unused entries are NULL, and there were
a lot of places where we did not checked that the caller did not
requested an operation on a NULL entry.
While we are there, one bit of KNFification, and make return code more
consistent by always returning the same code (EINVAL) when a screen
number higer than the maximum is requested.
of both devices are written, and device 0 will respond for device 1 if
device 1 isn't present. Pointed out by Quentin Garnier.
So the only way to know if device 1 is there or not is to send a command, which
is done later.
Detecting the second device here isn't important and won't speed up the probe.
We just need to know if there is at last one device on the bus.
Fix PR kern/22869 from Julio M. Merino Vidal.
call ata_raid_check_component() immediately, and panics since DMA setup
isn't done yet. So, defer the call until attach stage is almost done.
Tested with Promise TX2000.
most polling.
2) Clean up some goofiness in pciide -- get rid of the whole "candisable" path
(it's gratuitous) and simplify the code by calling pciide_map_compat_intr(),
*_set_modes() and wdc_print_modes() from central locations.
3) Add a register writability and register ghost test to eliminate phantom
drives more quickly.
another function, always doing a page 0 MODE SENSE to get the block
descriptor if we use READ CAPACITY, and use SMS_DBD on the page 4/5 MODE
SENSE. This does one extra command in some cases, but it separates and
simplifies the code a little.
Why do we prefer READ CAPACITY over READ FORMAT CAPACITIES? Two reasons:
1) It's much older and is much less likely to have had its command code
abused, and is thus "safer" to try first. 2) ALL of my USB flash readers
and pen drives screw up their capacity descriptors -- mostly off-by-one
errors in the size (they return the maximum LBA number instead, a la READ
CAPACITY, which has *never* been how READ FORMAT CAPACITIES was documented
in the MMC spec), and one returns the "no media" code on slots that have
media inserted (despite returning almost-correct data otherwise)!
F*** me with a chainsaw.
This wasn't an error, on this chipset we have the SATA controller on function
0 of the IDE controller, not the pcib bridge.
Fix provided by Stephen Degler.
don't activate any quirks for VM_6 variant
problem reported by Lars Heidieker on current-users
also don't activate any quirks for VM_3 - VM_5 variants, neither
VE_2 - VE_4, in anticipation that the more modern variants don't need
the quirks
voltage sense wired. So, disable it and force the card to 5V by default.
Also, recode the hack to use the "direct Vcc" feature of the chip, letting
it manage the voltage directly, as this is supported on both the 296 and
396.
controls. As it was before, the master volume control in all mixer programs
did nothing. It seems that this codec is not correctly wired in (almost?)
all motherboards.
Patch by Quentin Garnier. Fixes first part of PR kern/22548.
have proper definitions for offsets of APM DMA registers.
NULL out round_buffersize and round_blocksize in audio_hw_if. We
don't seem to have any special requirements and audio(9) already
provides enough rounding.
we start polling the BMCR again. Greg Woods noted on tech-net@
that the DP83840A manual states that there should be a 500us delay
between asserting software reset and attempting MII serial operations.
I've also noted that a DP83815 can get into a bad state on cable
removal and reinsertion if we do not delay here.
This may well address PR/16346, and I seem to recall occasional
reports of auto-negotiation and flaky kinds of errors that this
might also alleviate.
descriptor sizes if larger mbuf clusters are in use -- currently
commented out because, for some reason, transmit doesn't work at
all with large mbuf cluster sizes.
Note that using actual 9K frames is a _lot_ less efficient than
using frames that are a little smaller than 8K so that they fit
in one or two pages (instead of two or three) and avoid the extra
chained descriptor with only 1K of data in it, too. TCP will
segment things just fine for an 8K total packet size, just as it
does for 1500-byte packets.
On my system, using "mtu 8000" with ifconfig, I get an almost 50%
performance boost for TCP (930Mbit/sec instead of 660Mbit/sec) over
1500-byte frames. "mtu 9000" yields only 770Mbit/sec.
* We were never calling the callback function when we got a UFI CCI(!).
* We were passing a bogus residual count on non-UFI CCIs.
* After a REQUEST SENSE, the UFI CCI may include the ASC/ASCQ from the sense
information -- don't consider this a failure.
In addition, remove a horrible hack that was causing us to drop sense
information on the floor and return empty INQUIRY responses, often in the case
where there is a UNIT ATTENTION pending during the INQUIRY. (Some drives send
the data and then fail, some don't.)
This eliminates the need for quirks for the Y-E DATA FlashBuster.
These changes work on:
umass0: SAMSUNG SFD-321U [FW 0C], rev 1.10/2.10, addr 2
umass0: using UFI over CBI with CCI
sd0 at atapibus1 drive 0: <SMSC, USB FDC, 2.10> disk removable
umass0: TEAC TEAC FD-05PUB, rev 1.10/0.00, addr 2
umass0: using UFI over CBI with CCI
sd0 at atapibus0 drive 0: <TEAC, FD-05PUB, 1026> disk removable
umass0: Y-E DATA FlashBuster-U, rev 1.00/3.04, addr 2
umass0: using UFI over CBI with CCI
sd0 at atapibus0 drive 0: <Y-E DATA, USB-FDU, 3.04> disk removable
support it, we'll just get an ILLEGAL REQUEST and ignore it. Furthermore,
on the PQI Travel Flash at least, it works just fine without. These seem to
have been added in omnibus changes for particular devices, along with other
"quirks" that are now gone. Adios, we won't cry for you.
ready. This avoids gratuitously starting the motor on floppy and CD-ROM
drives, and eliminates the need for the audio playing test in cdopen().
Therefore, also remove PQUIRK_NOSTARTUNIT.
allocation length a little more precisely -- add the space for the header in
cd_mode_sense(). Also delete the XS_CTL_SILENT, since we really do want to
see errors.
Lastly, add a similar wrapper for mode select, simplifying the callers
slightly.
if our other MODE SENSEs fail. Use this code for the optical device case,
at least for now. (We could query the optical media type and do a table
lookup for the geometry, but why bother? Actually, why bother with geometry
at all, but I digress...)
the "bus type" for this.
Merge all the code in the SCSI and ATAPI backends for "cd" devices. All of
the mode page handling and whatnot is general to SCSI MMC devices, and should
never have been separated to begin with. This fixes a variety of problems,
and adds load/unload support for SCSI-attached devices.
have to adjust for whether he have a "small" or "large" mode sense response.
sd_scsibus_getcache(), sd_scsibus_setache(): Set SMS_DBD so that we don't
have to deal with block descriptors here.
sd_scsibus_mode_sense(): Add a new parameter, "int *big", which tells the
caller whether it's a small or large header.
sd_scsibus_get_parms(), sd_scsibus_getcache(), sd_scsibus_setcache(): Use
the "big" value to find the data following the mode sense header correctly.
sd_scsibus_mode_sense(): Disable PQUIRK_NOBIGMODESENSE for now, as the
reasons for it were dubious and most likely fixed now.
sd_scsibus_get_parms(): Now that we're actually locating the right data,
check the returned page code to see if it's correct before using the mode
page info.
sd_scsibus_get_parms(): Get the rotational rate from the flex geometry page
too, if present.
XXX sd_scsibus_get_{opt,simplifiedparms}() need some work.
PCIC_VENDOR_NONE: New.
PCIC_VENDOR_CIRRUS_*: Collapse the 2 chips into one vendor ID.
pcic_ident_ok(): Check the ID revision field -- if it's 0, punt.
pcic_vendor(): Check the ID revision field -- if it's 0, or the ID register
is all-1s, assume there is no chip present. (Previously this would return
"Unknown controller" -- which, AFAICT, *never* resulted in a working device.)
Do the Cirrus check only after verifying that we got the Intel ID.
pcic_attach(): Use a priori knowledge of the Cirrus chips to determine the
number of sockets rather than trying (unsuccessfully) to probe. Also, just
blast all of PCIC_INTR -- we do this in pcic_deactivate_card() anyway.
* Use the "block descriptor length" to find the beginning of the mode page
data. This is critical, since some devices do not return a block descriptor
at all. (XXX There is disgusting structure hackery here.)
* Always use scsipi_size() (i.e. READ CAPACITY) to get the total disk size.
* Get the rotational rate from the rigid geometry page.
* Don't read the block size from the block descriptor if there was none.
* Get the block size from the flexible geometry page if possible.
* If we don't like the way page 4 smells, try page 5, rather than just punting.
My Imation USB FlashGO! adapter responds to a Get Max Lun request with a
stall. With uhci, this does the expected thing. With ohci, it was returning
a "data underrun" error because we weren't setting "buffer rounding" (i.e.
USBD_SHORT_XFER_OK), and the underrun was taking priority. This happened with
both the ohci in a Mac cube and the builtin ohci on the S3C2410.
So, set USBD_SHORT_XFER_OK on the Get Max Lun. Now I get a stall reported and
umass attaches correctly.
* Remove some redundant code.
* ahd: Don't initiate negotiation for a discovery xs when the negotiation goal
is set. Just looking at the auto_negotiate mask is enough, and it can
cause an endless loop if the request gets requeued (happens with a
of a PPR negotiation, which ends with a busfree).
want the NO_ESTIMATE flag here, because if the source doesn't have it,
rnd_add_data will *use the time the data was added directly as entropy*,
which is not a good idea either for a hardware RNG that works like this
one (periodically polled by the driver) or, really, like any other that I've
seen, because even for "interrupt when ready" RNGs the data rate (and thus
interrupt time) will vary only slightly for a given RNG configuration over
time, if at all.
The data we add _is_ counted into the pool estimage because we directly
pass in an entropy estimate.
The overloading of the NO_ESTIMATE flag with this meaning seems, to me, like
a design bug (albeit easily fixed) in the rng interface; but until that's
changed, we need to use it here.
charge/discharge rates was incorrect - the value printed is actually
in A, W, Ah, or Wh, not in mA, mW, mAh, or mWh (assuming as we must
that the period is interpreted as a decimal point, not a thousands
separator).
FreeBSD ahd driver:
Fix a race condition in the flushing of commands that
have completed across the bus but not to the host before
processing of an exception condition (busfree, bus reset,
etc.). When flushing the controller of completed commands,
we also look for packetized commands that have completed
with good status and are stored in the "good status fifo".
The hardware will post to the good status fifo even if
data for that command is still active in a FIFO. In
one particular failure case, a command outstanding on the
bus reconnected, transferred data into a FIFO, and provided
good status while the host driver was processing an expected
busfree event (PPR message negotiation). This resulted in
an entry in the good status fifo that we completed, but
since the sequencer was paused, the data in the data FIFO
for this command had never been transferred to the host.
Once the busfree processing was complete, the sequencer
was unpaused, and the data completed its transfer to the
host. In some instances, the client for the data was notified
of the completion and attempted to view the data before
it arrived. This case only occurred during FreeBSD's
multi-target probe of the SCSI bus while some devices are
negotiating to go packetized and some devices are already
running in packetized.
The fix is to run and FIFOs active with a context in the
good status fifo to completion before completing the command
to the SCSI layer. This requies duplicating the FIFO rundown
operations in the host driver that would usually be handled
by the firmware, but there is no other alternative.
Don't blindly shutdown the SCB dma engine when restarting
the sequencer. We may be killing an operation that is
not supposed to be cancelled. The cases where we need to
shutdown these dma engines are already handled elsewhere in
the driver.
Fix a few more ahd_in?() -> ahd_in?_scbram() instances.
FreeBSD ahd driver:
aic79xx.h:
Add softc flag to indicate that we have seen at
least one selection since the last bus reset or
SE/LVD bus change.
aic79xx.c:
Fix a few style nits.
In ahd_update_pending_scbs(), only touch card registers
once we have found an SCB that needs to be updated.
This removes lots of clutter from PCI traces taken of
error recovery performed by the driver.
Short circuit the first selection iocell workaround handler
if we've run once since the last bus reset or iocell change.
This also removes clutter from PCI traces.
Note if completions are pending in the qoutfifo when we dump
card state.
FreeBSD ahd driver:
Correct a typo in a comment.
Add a comment in ahd_clear_critical_sections() about
our need to leave ENBUSFREE set in SIMODE1 while single
stepping.
Re-arrange some delay loops so that we always perform
a read after any register write and before the delay.
This should make the delay loop more accurate.
When completing message processing for a packetized
commention, return the controller to a state where
invalid non-packetized phases will still cause protocol
violations. These are the same operations as those
performed in the clear_target_state routine in the
firmware.
Now that we have a chip with working ABORTPENDING
support (the 7901B), comment out the automatic use
of this feature until we can adequately test it.
The previous checkin updated the bug mask for the
7901B so this code was exercised.
When resetting the bus, perform an ahd_flush_device_writes()
call so that our reset assertion delay is acurately
timed from when the reset bit is written to the controller.
FreeBSD ahd driver:
aic79xx.h:
Add a 7901A specific feature definition.
aic79xx_pci.c:
Split out the general aic790X setup into it's own
setup handler that works on single and dual controllers.
Adjust all other PCI setup handlers to initialize the
chips basic features and type before calling the generic
handler.
Turn off a few Rev B workarounds that are not required
on the 7901B.
FreeBSD ahd driver:
Add parenthesis so that we get all of the bits all
of the contents of the CCSCBCTL register into our
local varaible. The other bits are used in later tests.
This avoids a potential deadlock in ahd_run_qoutfifo()
if we happen to catch the DMA engine in just the right
state.
FreeBSD ahd driver:
When single stepping, only leave ENBUSFREE enabled
if it was already enabled. We don't want to set it
when it shouldn't be set, we just don't want to
inadvertantly turn it off. This should fix a recent
report of the aic7xxx driver repeatedly complaining of
"unexpected busfree while idle" in one configuration.
FreeBSD ahd driver:
aic79xx.seq:
Remove an old comment that no longer applies.
Fix a jump in our unexpected non-packetized phase
handler to use an explicit lable. The old code
had a hardcoded jump offset that was off by one
instruction.
FreeBSD ahd driver:
Work around SCSI spec violation by the Quantum Atlas 10K.
This drive delays going async after receiving a WDTR
message. We now send an SDTR message after a WDTR even
if our goal is to go async. This should work even for
confused devices.
If we get an unexpected busfree when attempting a WDTR
or SDTR, only set the goal negotiation parameters we were
trying to negotiate to off. This means that should a WDTR
message fail, we will still try an SDTR if our goal is
non-async.
Fix a few more places where we were looking at goal.period
instead of goal.offset for determining if we should be
negotiating sync. This should not have any impact on
our behavior, but the offset is more definitive and should
be used.
FreeBSD ahd driver:
Fix a reported case of severe data corruption:
aic79xx.h:
aic79xx.reg:
Return the SCB_TAG field to 16byte alignment.
It seems that on some PCI systems, SCBs are not
transferred correctly to the controller with
the previous placement of the SCB_TAG field.
Approved by: re (rwatson)
FreeBSD ahd driver:
Add 7901B support.
Sort IDs based on chip type.
Remove IROC IDs. We'll switch to using the IROC masks
if/when we want to start attaching to IROC controllers.
Approved by: RE
FreeBSD ahd driver:
aic7770.c:
aic79xx.c:
aic79xx.h:
aic79xx_pci.c:
aic7xxx.c:
aic7xxx.h:
aic7xxx_pci.c:
Switch ah?_reset() to take an additional "reinit" argument.
Use this instead of init_level to determin if the chip
should be fully reinitialized after a chip reset. This
is required so that ah?_shutdown() can reset the chip
without side-effects.
FreeBSD ahd driver:
aic79xx.c:
Use the special LUNLEN_SINGLE_LEVEL constant for
post Rev A4 hardware for single byte luns. Without
this change, Rev B hardware would place the single
byte of lun data in byte 0 of the lun structure when
it should be in byte 1. Since there are few if any
devices on the market that support multiple luns in
target mode, the corrupted lun field (which was only
corrupted for non-zero luns) wasn't hurting us.
Approved by: re (rwatson)
FreeBSD ahd driver:
Correct/Simplify ignore wide residue message handling
aic79xx.c:
In ahd_handle_ign_wide_residue():
o Use SCB_XFERLEN_ODD SCB field to determine transfer
"oddness" rather than the DATA_COUNT_ODD logic.
SCB_XFERLEN_ODD is toggled on every ignore wide
residue message so that multiple ignore wide residue
messages for the same transaction are properly supported.
o If the sg list has been exausted, the sequencer
doesn't bother to update the residual data count
since it is known to be zero. Perform the zeroing
manually before calculating the remaining data count.
o Use multibyte in/out macros instead of shifting/masking
by hand.
aic79xx_inline.h:
In ahd_setup_scb_common(), setup the SCB_XFERLEN_ODD field.
aic79xx.reg:
Use the SCB_TASK_ATTRIBUTE field as a bit field in the
non-packetized case. We currently only define one bit,
SCB_XFERLEN_ODD.
Remove the ODD_SEG bit field that was used to carry the odd
transfer length information through the SG cache. This
is obviated by SCB_XFERLEN_ODD field.
Remove the DATA_COUNT_ODD scratch ram byte that was used
dynamicaly compute data transfer oddness. This is obviated
by SCB_XFERLEN_ODD field.
aic79xx.seq:
Remove all updates to the DATA_COUNT_ODD scratch ram field.
Remove all uses of ODD_SEG. These two save quite a few
sequencer instructions.
Use SCB_XFERLEN_ODD to validate the end of transfer
ignore wide residue message case.
FreeBSD ahd driver:
Change hadling of the Rev. A packetized lun output bug
to be more efficient by having the sequencer copy the
single byte of valid lun data into the long lun field.
aic79xx.c:
Memset our hardware SCB to 0 so that untouched
fields don't confuse diagnostic output. With the
old method for handling the Rev A bug, if the long
lun field was not 0, this could result in bogus
lun information being sent to drives.
Use the same SCB transfer size for all chip types
now that the long lun is not DMA'ed to the chip.
aic79xx.seq:
Add code to copy lun information for Rev.A hardware.
aic79xx_inline.h:
Remove host update of the long_lun field on every
packetized command.
to the FreeBSD ahd driver:
Correct spelling errors.
Switch to handling bad SCSI status as a sequencer interrupt
instead of having the kernel proccess these failures via
the completion queue. This is done because:
o The old scheme required us to pause the sequencer and clear
critical sections for each SCB. It seems that these pause
actions, if coincident with a sequencer FIFO interrupt, would
result in a FIFO interrupt getting lost or directing to the
wrong FIFO. This caused hangs when the driver was stressed
under high "queue full" loads.
o The completion code assumed that it was always called with
the sequencer running. This may not be the case in timeout
processing where completions occur manually via
ahd_pause_and_flushwork().
o With this scheme, the extra expense of clearing critical
sections is avoided since the sequencer will only self pause
once all pending selections have cleared and it is not in
a critical section.
aic79xx.c
Add code to handle the new BAD_SCB_STATUS sequencer
interrupt code. This just redirects the SCB through
the already existing ahd_complete_scb() code path.
Remove code in ahd_handle_scsi_status() that paused
the sequencer, made sure that no selections where
pending, and cleared critical sections. Bad
status SCBs are now only processed when all of these
conditions are true.
aic79xx.reg:
Add the BAD_SCB_STATUS sequencer interrupt code.
aic79xx.seq:
When completing an SCB upload to the host, if
we are doing this because the SCB contains non-zero
SCSI status, defer completing the SCB until there
are no pending selection events. When completing
these SCBs, use the new BAD_SCB_STATUS sequencer
interrupt. For all other uploaded SCBs (currently
only for underruns), the SCB is completed via the
normal done queue. Additionally, keep the SCB that
is currently being uploaded on the COMPLETE_DMA_SCB
list until the dma is completed, not just until the
DMA is started. This ensures that the DMA is restarted
properly should the host disable the DMA transfer for
some reason.
In our RevA workaround for Maxtor drives, guard against
the host pausing us while trying to pause I/O until the
first data-valid REQ by clearing the current snapshot
so that we can tell if the transfer has completed prior
to us noticing the REQINIT status.
In cfg4data_intr, shave off an instruction before getting
the data path running by adding an entrypoint to the
overrun handler to also increment the FIFO use count.
In the overrun handler, be sure to clear our LONGJMP
address in both exit paths.
Perform a few sequencer optimizations.
aic79xx.c:
Print the full path from the SCB when a packetized
status overrun occurs.
Remove references to LONGJMP_SCB which is being
removed from firmware usage.
Print the new SCB_FIFO_USE_COUNT field in the
per-SCB section of ahd_dump_card_state(). The
SCB_TAG field is now re-used by the sequencer,
so it no longer makes sense to reference this
field in the kernel driver.
aic79xx.h:
Re-arrange fields in the hardware SCB from largest
size type to smallest. This makes it easier to
move fields without changing field alignment.
The hardware scb tag field is now down near the
"spare" portion of the SCB to facilitate reuse
by the sequencer.
aic79xx.reg:
Remove LONGJMP_ADDR.
Rearrange SCB fields to match aic79xx.h.
Add SCB_FIFO_USE_COUNT as the first byte
of the SCB_TAG field.
aic79xx.seq:
Add a per-SCB "Fifos in use count" field and use
it to determine when it is safe (all data posted)
to deliver status back to the host. The old method
involved polling one or both FIFOs to verify that
the current task did not have pending data. This
makes running down the GSFIFO very cheap, so we
will empty the GSFIFO in one idle loop pass in
all cases.
Use this simplification of the completion process
to prune down the data FIFO teardown sequencer for
packetized transfers. Much more code is now shared
between the data residual and transfer complete cases.
Correct some issues in the packetized status handler.
It used to be possible to CLRCHN our FIFO before status
had fully transferred to the host. We also failed to
handle NONPACKREQ phases that could occur should a CRC
error occur during transmission of the status data packet.
Correct a few big endian issues:
aic79xx.c:
aic79xx_inline.h:
aic79xx_pci.c:
aic79xx_osm.c:
o Always get the SCB's tag via the SCB_GET_TAG acccessor
o Add missing use of byte swapping macros when touching
hscb fields.
o Don't double swap SEEPROM data when it is printed.
Correct a big-endian bug. We cannot assign a
o When assigning a 32bit LE variable to a 64bit LE
variable, we must be explict about how the words
of the 64bit LE variable are initialized. Cast to
(uint32_t*) to do this.
aic79xx.c:
In ahd_clear_critical_section(), hit CRLSCSIINT
after restoring the interrupt masks to avoid what
appears to be a glitch on SCSIINT. Any real SCSIINT
status will be persistent and will immidiately
reset SCSIINT. This clear should only get rid of
spurious SCSIINTs.
This glitch was the cause of the "Unexpected PKT busfree"
status that occurred under high queue full loads
Call ahd_fini_scbdata() after shutdown so that
any ahd_chip_init() routine that might access
SCB data will not access free'd memory.
Reset the bus on an IOERR since the chip doesn't
seem to reset to the new voltage level without
this.
Change offset calculation for scatter gather maps
so that the calculation is correct if an integral
multiple of sg lists does not fit in the allocation
size.
Adjust bus dma tag for data buffers based on 39BIT
addressing flag in our softc.
Use the QFREEZE count to simplify ahd_pause_and_flushworkd().
We can thus rely on the sequencer eventually clearing ENSELO.
In ahd_abort_scbs(), fix a bug that could potentially
corrupt sequencer state. The saved SCB was being
restored in the SCSI mode instead of the saved mode.
It turns out that the SCB did not need to be saved at all
as the scbptr is already restored by all subroutines
called during this function that modify that register.
aic79xx.c:
aic79xx.h:
aic79xx_pci.c:
Add support for parsing the seeprom vital product
data. The VPD data are currently unused.
aic79xx.h:
aic79xx.seq:
aic79xx_pci.c:
Add a firmware workaround to make the LED blink
brighter during packetized operations on the H2A.
aic79xx_inline.h:
The host does not use timer interrupts, so don't
gate our decision on whether or not to unpause
the sequencer on whether or not a timer interrupt
is pending.
