to search the QINFIFO to remove any possible command that is waiting
otherwise our abort request may not be held up still waiting for the
first command to complete.
I/O port address of most devices is not contiguos, a return value of
probe routine is not so useful for detecting conflict. The return
value was too big, and kernel sometimes detected conflict even though
two devices are not conflict in I/O address between them.
Suggested by: Chiharu Shibata <chi@rd.njk.co.jp>
Fix a few panics during error recovery:
1) Stupid mistake in the "no SCB match handler" where I was using the wrong
variable (busy_scbid instead of scb_index).
2) Unbusy the target of an abort request if the command we are trying to
abort is an untagged transaction. If we don't, we get a fatal NO_MATCH_BUSY
condition which "should never happen".
3) When an abort completes, turn off ahc->in_timeout or else the next timeout
will hit the protective "scb timesout again" panic.
4) Fix a typo that caused the requeued "abort" SCB to have its TAG_ENB and
disconnect bits to be cleared (missing ~) so that devices would complain
about overlapped commands.
Be sure to turn off the unexpected busfree interrupt after we do a bus
reset since we are expecting the bus to go free in that case.
Return XS_TIMEOUT instead of XS_DRIVERSTUFFUP in certain scenarios. XS_TIMEOUT
allows for retries, XS_DRIVERSTUFFUP does not.
Allow commands with SDTR and WDTR negotiation to be tagged. The SCSI II spec
says that you probably should not do this for fear of hitting bogus devices.
The driver did this in the past for almost two years without any problem,
and not doing it causes problems during error recovery to a tag capable device
as the number of openings is higher than two and we'll start sending it
tagged commands causing "overlapped commands attempted" type errors. The
real fix needs to happen in the generic SCSI layer which can limit the
number and type of transactions to a device during error recovery efficiently.
Give ourselves at least 100ms to perform a request sense instead of relying
on the original timeout to be long enough to complete this new command as
well as the one that generated the condition.
Removed some redundant code.
host DMAs. The additional test to ensure that the DMA has stopped is also
unnecessary since we've already waited for the DMA to complete.
Update my copyright for the new year.
(since T_DIRECT just incidentally happens to be equal 0). This causes
more harm than it would do good. Instead , get it at the uk driver.
Reviewed by: obrien@NUXI.com (David O'Brien)
set it in the first place, independent of whether sin->sin_port
is set.
The result is that diverted packets that are being forwarded
will be diverted once and only once on the way in (ip_input())
and again, once and only once on the way out (ip_output()) -
twice in total. ICMP packets that don't contain a port will
now also be diverted.
with <= 100 usec between each character arrival time. This didn't happen
until rev.1.75 of clock.c because DELAY(100) used to delay for closer to
80 usec than 100 usec, and the minimum time between character arrivals is
87.8 usec at the maximum supported speed of 115200 bps 8N1.
Clear DCD timestamp flag on close (the input timestamp flag is already
cleared).
key "print scrn".
It used to stop at the first non-open vty, now it skips the non-open
ones and thereby enable one to cycle around all open vty by pressing
"print scrn".
I have code to calibrate the overhead fairly accurately, but there
is little point in using it since it is most accurate on machines
where an estimate of 0 works well. On slow machines, the accuracy
of DELAY() has a large variance since it is limited by the resolution
of getit() even if the initial delay is calibrated perfectly.
Use fixed point and long longs to speed up scaling in DELAY().
The old method slowed down a lot when the frequency became variable.
Assume the default frequency for short delays so that the fixed
point calculation can be exact.
Fast scaling is only important for small delays. Scaling is done
after looking at the counter and outside the loop, so it doesn't
decrease accuracy or resolution provided it completes before the
delay is up. The comment in the code is still confused about this.
- don't uselessly initialize the fifo "DMA" bit at attach time.
- initialize the fifo "DMA" bit at open time. Without this, the device
interrupts for every character received, reducing input performance
to that of an 8250.
- don't uselessly initialize the fifo trigger level to 8 (scaled to
256) at attach time.
- don't scale the fifo trigger level to 512 bytes. The driver's pseudo-
dma buffer has size 256, so it can't handle bursts of size 512 or 256.
It should be able to handle the second lowest ftl (2 scaled to 64).
- don't reset the fifos in siostop(). Reset triggers a hardware bug
involving wedging of the output interrupt bit This workaround
unfortunately requires ESP support to be configured.
slices in sd_open() after a media change when the previous sd_open()
discards the previous slices and then fails. sd_open() just handles
media changes poorly and fails too often.
If we can, use timeouts instead of DELAYs when dealing with a bus reset.
This prevents us from holding up the whole machine for a noticible amount
of time (especially for a real time app).
Make a pass over the timeout/error handling code. Aborts are more
reliable. We actually handle parity errors correctly now instead of
locking up the bus. Added code to properly clean up disconnected SCBs
down on the card during error handling. Improved robustness in several
areas.
If we are using defaults, but are an Ultra card, negotiate at 20MHz instead
of 10.
Don't attempt to handle any commands for 100ms after a reset has occured.
This is the minimum time before a target will respond to selection. Also
disable the busfree interrupt before doing a bus reset. This prevents the
driver from getting confused by an "unexpected busfree".
Expand the boundaries of a pause disabled region to close of possible race
condition.
Revert a portion of the DMA code to fix false overruns.
Add a missing "add_scb_to_free_list" so we don't leak SCBs.
The limit is now only used by init, so it may as well be "infinite".
Don't use RLIM_INFINITY, since setrlimit() doesn't allow setting
that value. Use maxfiles instead of RLIM_INFINITY for the hard
limit for the same reason.
Similarly for the maxprocesses limits (use the "infinite" value of
maxproc instead if MAXUPRC and RLIM_INFINITY).
NOFILES, MAXUPRC, CHILD_MAX and OPEN_MAX are no longer used in
/usr/src and should go away. Their values are almost guaranteed to
be wrong now that login.conf exists, so anything that uses the values
is broken. Unfortunately, there are probably a lot of ports that
depend on them being defined.
The global limits maxfilesperproc and maxprocperuid should go away
too.
Put obsolete GATEWAY option back in opt_defunct.h. It's the only
significant option that has gone away since 2.1.6, so warning about
it might be useful.
SEM* and SHM*. These are already supported in the options files. I
mostly used the default value plus 1. This ensures that the LINT kernel
depends on the options headers.
("Hey! Who made _you_ the keeper of all things BSDish?!") but this has
bugged me for a long time, and now that I finally have the chance
to hack on it (and test the results), I'll take my chances. I can also
point to other BSD implementations for precedents if you put my back to
the wall.
The only thing that's changed is how the messages are formatted. Now,
instead of having this:
aha0 at 0x330-0x333 irq 11 drq 5 on isa
(aha0:3:0): "HP C1553A 9503" type 1 removable SCSI 2
st0(aha0:3:0): Sequential-Access density code 0x24, variable blocks, write-enabled
(aha0:3:1): "HP C1553A 9503" type 8 removable SCSI 2
ch0(aha0:3:1): Medium-Changer 6 slot(s) 1 drive(s) 0 arm(s) 0 i/e-slot(s)
We have this:
aha0 at 0x330-0x333 irq 11 drq 5 on isa
scbus0 at aha0 bus 0
st0 at scbus0 target 3 lun 0
st0: <HP C1553A 9503> type 1 removable SCSI 2
st0: Sequential-Access density code 0x24, variable blocks, write-enabled
ch0 at scbus0 target 3 lun 1
ch0: <HP C1553A 9503> type 8 removable SCSI 2
ch0: Medium-Changer 6 slot(s) 1 drive(s) 0 arm(s) 0 i/e-slot(s)
Which is (to me anyway) is a lot more pleasant to look at. (Call me
crazy -- g'head: you know you wanna -- but the previous messages remind
me of Linux. Ever see the output from the linux device probes? It's a mess
of copyright notices, version numbers/dates, author e-mail addresses and
other crap. Let's not go there, okay? Bleh.)
Notice that devices are now specified in terms of the scsi bus they
live on rather than the adapter. This better reflects the contents
of the kernel config file (if you use wired-down device specifications
anyway) and removes some ambiguity that may arise if you have a multi-
channel adapter with more than one bus.
Also, sc_print_addr() now generates messages like this:
st0 at scbus0 target 3 lun 0: NOT READY asc:3a,0 Medium not present
instead of this:
st0(aha0:3:0): NOT READY asc:3a,0 Medium not present
I also added a quirk entry for the HP Superstore 12000e 6 tape DAT
autoloader, which needs SC_MORE_LUS in order for the changer device
to be properly probed and attached. (I'm working on a chcontrol utility
to manipulate the changer on this drive which should hopefully be general
enough to work with other changers too. If you want the prototype I have
now, it's at ftp://skynet.ctr.columbia.edu/pub/freebsd/changer.c.)
Remaining bugs:
- The 'foodev0: yadda yadda yadda' bits should probably be printed entirely
by the device-specific subdriver attach code instead of half by the
scsi_device_attach() routine and half by the device specific attach
routine like it is now.
- The wired-down device specifications in the kernel config file should
be used to control bus/device probing to some extent rather than just
for choosing names for devices we find. If the config says there's a
device at scbus0 target 0 lun 0 called sd0, we should look there and
check for a device that can be managed by the sd driver. If we don't
find one, we should probably complain that there's no device there or
that there is a device but of the wrong type. Once all the devices from
the wired down list have been probed, the code can then autodetect and
autoattach any devices that remain unassigned.
- Apparently some tape changers (hi Ulf!) return 'not ready/medium not
present' when the magazine is loaded but a tape has not been put in the
drive yet. This causes an open(/dev/ch0) to fail and prevents you from
using the changer.c utility to load the first tape into the drive. My
HP changer does not behave this way. The workaround is to manually load
a tape into the drive before attempting to use the changer program, but
you can get in trouble if you accidentally eject a tape without loading
a new one and you're at a remote location: you won't be able to load
any tapes anymore. I'm not sure what the correct software solution is
for this but ideally there should be one.
- I should not be doing this: I'm the NIS guru, not the SCSI guru.
(This is not my beautiful code. How did I get here? My god: what
have I done?)
