merge ktrace-lwp.

Part 1 of a cleanup pass over the SCSI subsystem. The aim is to name
everything "scsi_*", since we really are talking about the SCSI command
set, ATAPI transport not withstanding. Improve the names of many structures,
and prepend "SCSI_" onto all SCSI command opcodes. Place items described
by the SCSI Primary Commands document into scsi_spc.h.

branches: 1.4.22; 1.4.30; 1.4.32;
Install "dev/scsipi/scsipi_cd.h" again because it is e.g. used by the
CD player include in version 2.x of the "kdemultimedia" package.

Only install headers which are actually used by our userland. This
saves about 2.2MB under /usr/include/dev/. Discussed on tech-kern@
recently.

I HOPE to get the list right. The headers I left in are ones
used for MI tools and those whose usage I discovered by grep over tree sources.
Feel free to put needed includes back in if you encounter anything which
should not be removed from lists.

branches: 1.2.6;
add ses.h to list of includes

branches: 1.1.14;
Rework the way kernel include files are installed. In the new method,
as with user-land programs, include files are installed by each directory
in the tree that has includes to install. (This allows more flexibility
as to what gets installed, makes 'partial installs' easier, and gives us
more options as to which machines' includes get installed at any given
time.) The old SYS_INCLUDES={symlinks,copies} behaviours are _both_
still supported, though at least one bug in the 'symlinks' case is
fixed by this change. Include files can't be build before installation,
so directories that have includes as targets (e.g. dev/pci) have to move
those targets into a different Makefile.

Use SCSI/ATAPI common definition for MODE_{SELECT,SENSE}{,_BIG}. Define
functions to send theses commands in scsipi_base.c and use them instead
of ad-hoc commands setups.

branches: 1.3.26;
Backout previous change, and rather, remove all protection but the one
actually currently required. Suggested by Charles M. Hannum.

Fix or add protection for mutiple inclusion.

add support for an ATAPI attachment for 'sd'.
fix 'cd' driver's NCD_SCSI bogosity (was using testing wrong macro!)
clean up in various ways:
* make common atapi_mode_{sense,select}() functions.
* put ATAPI data structures in more sensible headers, split up by
device type.
* include headers a bit more carefully.
* pass flags to attachment-specific cd functions, and use them.
* get rid of SCSI bits in scsipi_base.h's scsipi_make_xs(), move
them into the correct place in scsi_base.c.
* fix minor typo in struct name in scsipiconf.h (which was apparently
never used except in a #define later in the same file).
* use __attribute__ to force 4-byte alignment for xs command store,
so that architectures trying to bus_space_write_multi_N() (where
N > 1) that data to a controller won't lose.
* clean up a few comments in typos, and make a few #defines easier to
understand/maintain.
* rename cd_link.h to cdvar.h (via repository copy). This is exactly
what a 'var' file is supposed to be.

- add or adjust /* FALLTHROUGH */ where appropriate
- add __unreachable() after functions that can return but won't in
this case, and thus can't be marked __dead easily

branches: 1.29.20;
src is too big these days to tolerate superfluous apostrophes. It's
"its", people!

branches: 1.28.44;
Remove clause 3 and 4 from TNF licenses

branches: 1.27.2; 1.27.4;
use aprint_*_dev and device_xname

branches: 1.26.70;
merge ktrace-lwp.

- Sprinkle const
- Avoid variable shadowing.
- Eliminate some caddr_t abuse.

Part 1 of a cleanup pass over the SCSI subsystem. The aim is to name
everything "scsi_*", since we really are talking about the SCSI command
set, ATAPI transport not withstanding. Improve the names of many structures,
and prepend "SCSI_" onto all SCSI command opcodes. Place items described
by the SCSI Primary Commands document into scsi_spc.h.

branches: 1.23.4; 1.23.6;
Change the way bustype_cmd is used. Rather than having it be responsible for
calling scsipi_make_xs() and scsipi_execute_xs(), instead push these into
scsipi_command. Make bustype_cmd and PHOLD/PRELE be called from
scsipi_execute_xs(). This allows us to create a xfer structure -- possibly on
the stack -- and call scsipi_execute_xs() directly.

Make the xxstart() functions reentrant again, as some drivers HBA can call
scsipi_done() from their scsipi_request().
For this, add a struct scsipi_xfer * argument to scsipi_command().
If not NULL scsipi_command() will use this to enqueue this xfer, otherwise
it'll try to allocate a new one. This scsipi_xfer has to be allocated
and initialised by scsipi_make_xs() or equivalent.
In xxstart(), allocate a scsipi_xfer using scsipi_make_xs(), and if not NULL,
dequeue the buffer before calling scsipi_command(). This makes sure that
scsipi_command() will not fail, and also makes sure that xxstart() won't
be called again between the BUFQ_PEEK() and BUFQ_GET().

Fix "dequeued wrong buf" panics reported by Juergen Hannken-Illjes in
private mail and Andreas Wrede on current-users@.
Thanks to Jason Thorpe and Chuck Silver for review, and Andreas Wrede for
testing the patch.

Improve handling of memory shortage, to fix problems like:
sd3(mpt0:0:1:0): unable to allocate scsipi_xfer
sd3: not queued, error 12
Havard Eidnes's analysis of this problem is that the scsipi_xfer pool is
competing for resources with other pools, including the the inode and vnode
pools which can grow quite large.

*_scsipi_cmd(): don't biodone the buffer if scsipi_make_xs() fails, let the
caller deal with the problem
start function of block devices drivers: dequeue the buffer after the
scsipi_command() call. If scsipi_command() fails with ENOMEM don't dequeue
the buffer, and schedule a callout to call the start function after
some delay.
scsipi_init(): prime the scsipi_xfer_pool with one page. This ensure that
there is always some scsipi_xfer to play with. If scsipi_command() fails
because of pool_get(), we're sure there will be resources available later,
when the pending commands have completed.

Reviewed by Jason Thorpe and Havard Eidnes.
Todo: remove the "unable to allocate scsipi_xfer" and "not queued, error %d"
printfs, but I choose to keep them for now, to help make sure the code does
what it should.

Use ANSI function decls and make use of static.

branches: 1.19.2;
Decrease xs_retries before retrying aborted commands, and report EIO if it
reaches 0. Avoids looping on aborded command in some special cases.

branches: 1.18.16;
don't need <sys/types.h> when including <sys/param.h>

add RCSIDs

branches: 1.16.2;
Use SCSI/ATAPI common definition for MODE_{SELECT,SENSE}{,_BIG}. Define
functions to send theses commands in scsipi_base.c and use them instead
of ad-hoc commands setups.

Pull up the thorpej_scsipi branch to main branch.
This is a completely rewritten scsipi_xfer execution engine, and the
associated changes to HBA drivers. Overview of changes & features:
- All xfers are queued in the mid-layer, rather than doing so in an
ad-hoc fashion in individual adapter drivers.
- Adapter/channel resource management in the mid-layer, avoids even trying
to start running an xfer if the adapter/channel doesn't have the resources.
- Better communication between the mid-layer and the adapters.
- Asynchronous event notification mechanism from adapter to mid-layer and
peripherals.
- Better peripheral queue management: freeze/thaw, sorted requeueing during
recovery, etc.
- Clean separation of peripherals, adapters, and adapter channels (no more
scsipi_link).
- Kernel thread for each scsipi_channel makes error recovery much easier
(no more dealing with interrupt context when recovering from an error).
- Mid-layer support for tagged queueing: commands can have the tag type
set explicitly, tag IDs are allocated in the mid-layer (thus eliminating
the need to use buggy tag ID allocation schemes in many adapter drivers).
- support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command
will be requeued, or a REQUEST SENSE will be sent as appropriate.

Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge

branches: 1.14.2; 1.14.14;
Cleanup the scsipi_xfer flags:
- `flags' is now gone, replaced with `xs_control' and `xs_status'.
- Massive cleanup of the control flags. Now we explicitly say that
a job is to complete asynchronously, rather than relying on side-effects,
and use a new flag to now that device discovery is being performed.
- Do SCSI device discovery interrupt-driven.

If scsipi_make_xs() fails, make sure to flag the buffer as error'd,
if provided, and biodone() it. Callers of scsipi_command() expect
this to be done.

Go to splbio() when calling scsipi_free_xs(), just like the SCSI counterpart.
Fixes PR #7853, Paul Dokas <dokas@cs.umn.edu>.

branches: 1.11.4; 1.11.6; 1.11.8;
In atapi_wdc.c, issue a 'REQUEST SENSE' command when appropriate.
Return XS_SENSE when the full sense info has been retrieved, or
XS_SHORTSENSE if only the sense key was available (from the error register)
Make atapi_interpret_sense() deal with this, and call scsipi_interpret_sense()
for XS_SENSE. (XXX sd_interpret_sense() and the ioctl code needs to be made
aware of XS_SHORTSENSE too ! sense hanlding for these is now less broken for
devices that support 'REQUEST SENSE')
All the ATAPI devices I have access to seems to honnor the SENSE_REQUEST
command, but I suspect some ATAPI devices will not (althouh it's mandatory).
The code should be able to deal with this, but is untested ...

Merge bouyer-ide

branches: 1.9.2;
Assign my copyrights to TNF.

part of fix for kern/3835: use of enumerated returns from target sense handlers

Clarify that Sense Key 1 is a "RECOVERED ERROR"- not just an error.

Use new Sense Key defines (for clarity)

add support for an ATAPI attachment for 'sd'.
fix 'cd' driver's NCD_SCSI bogosity (was using testing wrong macro!)
clean up in various ways:
* make common atapi_mode_{sense,select}() functions.
* put ATAPI data structures in more sensible headers, split up by
device type.
* include headers a bit more carefully.
* pass flags to attachment-specific cd functions, and use them.
* get rid of SCSI bits in scsipi_base.h's scsipi_make_xs(), move
them into the correct place in scsi_base.c.
* fix minor typo in struct name in scsipiconf.h (which was apparently
never used except in a #define later in the same file).
* use __attribute__ to force 4-byte alignment for xs command store,
so that architectures trying to bus_space_write_multi_N() (where
N > 1) that data to a controller won't lose.
* clean up a few comments in typos, and make a few #defines easier to
understand/maintain.
* rename cd_link.h to cdvar.h (via repository copy). This is exactly
what a 'var' file is supposed to be.

EACCESS->EROFS, for the write on readonly medium error.

Cosmetic changes to keep coding style consistency in this directory;

- Indent with tab of width 8.
- Use four column to indent continuation line.
- Fold long line if possible.
- Use return (xx) instead of return xx.
- Compare pointer against NULL instead of testing like boolean.
- Delete whitespace at the end of line.
- Delete whitespace in front of function call operator.
- Delete whitespace after cast.
- Dereference a pointer to function explicitly.
- Add an empty line after local variable declaration.
- Use NULL instead of (char *)0.
- Dont use block for single statement.

branches: 1.2.2;
Merge scsipi branch in the mainline. This add support for ATAPI devices
(currently only CD-ROM drives on i386). The sys/dev/scsipi system provides 2
busses to which devices can attach (scsibus and atapibus). This needed to
change some include files and structure names in the low level scsi drivers.

branches: 1.1.2;
file atapi_base.c was initially added on branch bouyer-scsipi.

Set PQUIRK_ONLYBIG in the wdc-atapi frontend, obviating the need to ever test
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.

branches: 1.10.22;
Use SCSI/ATAPI common definition for MODE_{SELECT,SENSE}{,_BIG}. Define
functions to send theses commands in scsipi_base.c and use them instead
of ad-hoc commands setups.

branches: 1.9.26;
Those tabs sure are excitable!!

Add support for ATA CD changer devices like the NEC CDR-251.
Evidence of managerial coding removed by Victor T. Cleaner (thorpej)

Backout previous change, and rather, remove all protection but the one
actually currently required. Suggested by Charles M. Hannum.

Fix or add protection for mutiple inclusion.

add support for an ATAPI attachment for 'sd'.
fix 'cd' driver's NCD_SCSI bogosity (was using testing wrong macro!)
clean up in various ways:
* make common atapi_mode_{sense,select}() functions.
* put ATAPI data structures in more sensible headers, split up by
device type.
* include headers a bit more carefully.
* pass flags to attachment-specific cd functions, and use them.
* get rid of SCSI bits in scsipi_base.h's scsipi_make_xs(), move
them into the correct place in scsi_base.c.
* fix minor typo in struct name in scsipiconf.h (which was apparently
never used except in a #define later in the same file).
* use __attribute__ to force 4-byte alignment for xs command store,
so that architectures trying to bus_space_write_multi_N() (where
N > 1) that data to a controller won't lose.
* clean up a few comments in typos, and make a few #defines easier to
understand/maintain.
* rename cd_link.h to cdvar.h (via repository copy). This is exactly
what a 'var' file is supposed to be.

Don't define AUDIO_PLAY twice (accidently add in last commit).

Cosmetic changes to keep coding style consistency in this directory;

- Indent with tab of width 8.
- Use four column to indent continuation line.
- Fold long line if possible.
- Use return (xx) instead of return xx.
- Compare pointer against NULL instead of testing like boolean.
- Delete whitespace at the end of line.
- Delete whitespace in front of function call operator.
- Delete whitespace after cast.
- Dereference a pointer to function explicitly.
- Add an empty line after local variable declaration.
- Use NULL instead of (char *)0.
- Dont use block for single statement.

branches: 1.2.2;
Merge scsipi branch in the mainline. This add support for ATAPI devices
(currently only CD-ROM drives on i386). The sys/dev/scsipi system provides 2
busses to which devices can attach (scsibus and atapibus). This needed to
change some include files and structure names in the low level scsi drivers.

branches: 1.1.2;
file atapi_cd.h was initially added on branch bouyer-scsipi.

Remove this file; ATAPI_READ_FORMAT_CAPACITIES is the same as
READ_FORMAT_CAPACITIES; atapi_flex_geometry_page should probably be the same as
scsi_disk's page_flex_geometry.

Remove PQUIRK_BYTE5_ZERO.

branches: 1.4.22;
Use SCSI/ATAPI common definition for MODE_{SELECT,SENSE}{,_BIG}. Define
functions to send theses commands in scsipi_base.c and use them instead
of ad-hoc commands setups.

branches: 1.3.26;
Backout previous change, and rather, remove all protection but the one
actually currently required. Suggested by Charles M. Hannum.

Fix or add protection for mutiple inclusion.

add support for an ATAPI attachment for 'sd'.
fix 'cd' driver's NCD_SCSI bogosity (was using testing wrong macro!)
clean up in various ways:
* make common atapi_mode_{sense,select}() functions.
* put ATAPI data structures in more sensible headers, split up by
device type.
* include headers a bit more carefully.
* pass flags to attachment-specific cd functions, and use them.
* get rid of SCSI bits in scsipi_base.h's scsipi_make_xs(), move
them into the correct place in scsi_base.c.
* fix minor typo in struct name in scsipiconf.h (which was apparently
never used except in a #define later in the same file).
* use __attribute__ to force 4-byte alignment for xs command store,
so that architectures trying to bus_space_write_multi_N() (where
N > 1) that data to a controller won't lose.
* clean up a few comments in typos, and make a few #defines easier to
understand/maintain.
* rename cd_link.h to cdvar.h (via repository copy). This is exactly
what a 'var' file is supposed to be.

merge ktrace-lwp.

Part 1 of a cleanup pass over the SCSI subsystem. The aim is to name
everything "scsi_*", since we really are talking about the SCSI command
set, ATAPI transport not withstanding. Improve the names of many structures,
and prepend "SCSI_" onto all SCSI command opcodes. Place items described
by the SCSI Primary Commands document into scsi_spc.h.

Backing out changes to clean up scsipi. I was pointed out there were
problems i hadn't seen. To prevent lossage i'd decided to back off all
changes and let them be reviewed on tech-kern.

Part of the cleanup of sys/scsipi's use of types; rename all u_int* to
uint* and change the u_long's to uint32_t's where possible. Note that the
iocl definitions/hooks have to be ulong (or u_long) or they'll bomb out.

branches: 1.2.16; 1.2.24; 1.2.26;
Various quirks for the ATAPI OnStream DI-30, mostly from the FreeBSD driver.
Many thanks to Chris Pinnock for giving me remote access to his hardware.

branches: 1.1.2; 1.1.4;
Snapshot of ATAPI tapes support. Known to be able to read/write to tape with:
st0 at atapibus0 drive 1: <Seagate STT8000A, , 5.51> type 1 sequential
removable
Major changes may still happen in order to properly support other ATAPI
tape drives.

PR kern/56403

Fix kernel freeze for wdc(4) variants with ATAC_CAP_NOIRQ:

(1) Change ata_xfer_ops:c_poll from void to int function. When it returns
ATAPOLL_AGAIN, let ata_xfer_start() iterate itself again.

(2) Let wdc_ata_bio_poll() return ATAPOLL_AGAIN until ATA_ITSDONE is
achieved.

A similar change has been made for mvsata(4) (see mvsata_bio_poll()),
and no functional changes for other devices.

This is how the drivers worked before jdolecek-ncq branch was merged.

Note that this changes are less likely to cause infinite recursion:

(1) wdc_ata_bio_intr() called from wdc_ata_bio_poll() asserts ATA_ITSDONE
in its error handling paths via wdc_ata_bio_done().

(2) Return value from c_start (= wdc_ata_bio_start()) is checked in
ata_xfer_start().

Therefore, errors encountered in ata_xfer_ops:c_poll and c_start routines
terminate the recursion for wdc(4). The situation is similar for mvsata(4).

Still, there is a possibility where ata_xfer_start() takes long time to
finish a normal operation. This can result in a delayed response for lower
priority interrupts. But, I've never observed such a situation, even when
heavy thrashing takes place for swap partition in wd(4).

"Go ahead" by jdolecek@.

Merge thorpej-cfargs2.

branches: 1.139.8;
Merge thorpej-cfargs branch:

Simplify and make extensible the config_search() / config_found() /
config_attach() interfaces: rather than having different variants for
which arguments you want pass along, just have a single call that
takes a variadic list of tag-value arguments.

Adjust all call sites:
- Simplify wherever possible; don't pass along arguments that aren't
actually needed.
- Don't be explicit about what interface attribute is attaching if
the device only has one. (More simplification.)
- Add a config_probe() function to be used in indirect configuiration
situations, making is visibly easier to see when indirect config is
in play, and allowing for future change in semantics. (As of now,
this is just a wrapper around config_match(), but that is an
implementation detail.)

Remove unnecessary or redundant interface attributes where they're not
needed.

There are currently 5 "cfargs" defined:
- CFARG_SUBMATCH (submatch function for direct config)
- CFARG_SEARCH (search function for indirect config)
- CFARG_IATTR (interface attribte)
- CFARG_LOCATORS (locators array)
- CFARG_DEVHANDLE (devhandle_t - wraps OFW, ACPI, etc. handles)

...and a sentinel value CFARG_EOL.

Add some extra sanity checking to ensure that interface attributes
aren't ambiguous.

Use CFARG_DEVHANDLE in MI FDT, OFW, and ACPI code, and macppc and shark
ports to associate those device handles with device_t instance. This
will trickle trough to more places over time (need back-end for pre-OFW
Sun OBP; any others?).

branches: 1.138.4;
fix use-after-free for ata xfer on bio submission found by KASAN

driver ata_bio hooks read parts of the xfer after ata_exec_xfer()
call in order to determine return value, change so that the hook
doesn't return any value - callers do not care already,
as all I/O requests are asynchronous

this problem was uncovered by recent change for wd(4) to not hold
wd mutex during ata_bio call, the interrupt for the xfer might
thus actually fire immediately

adjust also ata_exec_command driver hooks similarily - remove all
completion and waiting logic from drivers, upper layer ata code
using AT_WAIT/AT_POLL changed to call ata_wait_cmd() itself

PR kern/55169 by Nick Hudson

branches: 1.137.2;
fix deadlock in wdcwait() when xfer timeout happens while the atabus
thread sleeps in wdcwait() - check current lwp rather than relying
on global ATACH_TH_RUN channel flag

should fix the hang part of the problem reported in
http://mail-index.netbsd.org/netbsd-users/2020/03/12/msg024249.html

thanks to Paul Ripke for providing extensive debugging info

C99 initializers for scsipi_bustype. No functional change intended.

the number of possible ATAPI devices on an ATA bus is not always 2,
it is however many devices the underlying ATA bus can have (eg. 1 for SATA),
so initialize the scsipi chan_ntargets from the ATA ch_ndrives.
this fixes a memory read overrun detected by KASAN.
discussed with mlelstv@ and jdolecek@

branches: 1.134.2;
in many device attach paths, allocate memory with M_WAITOK instead of M_NOWAIT
and remove code to handle failures that can no longer happen.

branches: 1.133.4;
pass correct status + error to *_atapi_phase_complete(), so that
the function is actually able to recognize when there was an error;
tested via reading a cd0 device in QEMU with ejected cdrom

bug was introduced with jdolecek-ncq branch

fixes PR kern/53724 by Andreas Gustafsson

hold channel lock during whole ata_dmaerr()/ata_downgrade_mode() -
according to code inspection this is safe, none of the set_modes
hooks execute anything which would be taking the lock

adresses PR kern/53714 by Andreas Gustafsson

Merge jdolecek-ncqfixes branch

- ata_xfer's are dynamicall allocated as needed using a pool, no longer
limited to number of possible openings supported by controller; dump
and recovery paths use dedicated pre-allocated storage
- moved callouts and condvars from ata_xfer to queue or channel, so that
ata_xfer does not need special initialization
- slot allocation now done when xfer is being activated, uncoupled
from memory allocation; active slots are no longer tracked by controller
code
- channel and drive reset is done always via the atabus thread, and
now executes with channel locked the whole time
- NCQ recovery moved to shared function, and run via the thread also
- added some workarounds for buggy error recovery AHCI emulation in QEMU
and Parallels

designed to primarily fix kern/52614, but might also help with kern/47041
and kern/53183

Make wdcdebug_atapi_mask configurable like the other debug variables.

branches: 1.129.2; 1.129.4; 1.129.6;
reintroduce ATACH_IRQ_WAIT flag for attachments using wdcintr(), only
process the interrupt when the flag is set - this fixes spurious interrupt
during post-reset drive setup in wdc_ata_bio_start(), and wdc_atapi_start()

while those functions set WDCTL_IDS, this seems to be ignored by certain
(maybe all) PCI-IDE controllers; usually the implicit KERNEL_LOCK() would
prevent the interrupt anyway, but not when the start routine is started
from the atabus thread, which doesn't take it

fixes 'panic: wdc_ata_bio_intr: bad state' reported on current-users
by Chavdar Ivanov

revert the logic in wdc_atapi_intr() for wdc_wait_for_unbusy() to what it
was before NCQ merge; it got broken during the efford to remove ch_status
and ch_error on the branch

fixes atapi timeouts in vbox and with real harware reported separately
by Abhinav Upadhyay, Pault Goyette, Chavdar Ivanov, and Rares
Aioanei; with a bit of luck it could also fix PR kern/52605 and/or PR
kern/52606 by Martin Husemann

don't initialize tfd twice.

initialize tfd

Initialize tfd even when not waiting for DSC.

Merge support for SATA NCQ (Native Command Queueing) from jdolecek-ncq branch

ATA subsystem was changed to support several outstanding commands, and use
NCQ xfers if supported by both the controller and the disk, including NCQ
error recovery. Set NCQ high priority for BPRIO_TIMECRITICAL xfers
if supported. Added FUA support.

Done some work towards MP-safe, all ATA code tsleep()/wakeup() replaced
by condvars, and switched most code from spl* to mutexes (separate
wd(4) and ata channel lock).

Introduced new option WD_CHAOS_MONKEY to facilitate testing of error
handling, fixed several uncovered issues. Also fixed several problems
with kernel dump to wd(4) disk.

Tested with ahcisata(4), mvsata(4), siisata(4), piixide(4) on amd64,
with and without port multiplier, both disk and ATAPI devices; other
drivers and archs mechanically adjusted and compile-tested. NCQ is
supported for ahcisata(4) and siisata(4) for any controller, for
mvsata(4) only Gen IIe ones for now. Also enabled ATAPI support in
mvsata(4).

Thanks to Matt Thomas for initial ATA infrastructure patch, and
Jonathan A.Kollasch for siisata(4) NCQ changes and general testing.

Also fixes PR kern/43169 (wd(4)); and PR kern/11811, PR kern/47041,
PR kern/51979 (kernel dump)

branches: 1.123.4;
Make scsipi framework MPSAFE.

Data structures are now protected by a per-adapter mutex at IPL_BIO
that is created by the scsibus or atapibus instance when the adapter
is configured.
The enable reference counter and the channel freeze counter which are
currently used by HBA code before the adapter is configured, are made
atomic.
The target drivers are now all tagged as D_MPSAFE.

Almost all HBA drivers still require the kernel lock to present,
so all callbacks into HBA code are still protected by kernel lock
unless the driver is tagged as SCSIPI_ADAPT_MPSAFE.

TODO: refactor sd and cd to use dksubr.

branches: 1.122.2;
move scsipi_strvis -> libkern:strnvisx()
change the prototype to match userland
fix sizes of strings passed to it

Pass a correct residual size to wdcbit_bucket() that drains extra bytes.

Seems broken (but almost harmless?) since rev 1.101 committed ~10 years ago:
http://mail-index.netbsd.org/source-changes/2006/09/07/0043.html

Remove extra spaces.

branches: 1.119.2; 1.119.16;
Move a check for the drive type to the right place. This bug would cause
ATA drives to not be detected when the PATA (or emulated PATA) bus is shared
with an ATAPI device.
Should fix the problem reported by various peoples.
Thanks to Onno van der Linden and vanoven@free.fr for pointing out the
code in error.

Apply back changes that were reverted on Jul 24 and Jul 26 (general ata/wdc
cleanup and SATA PMP support), now that I'm back to fix the fallouts.

Revert, with intention of restoring in a less invasive way, the SATA Port
Multiplier code.

ok christos@

Revert dsl@'s changes of Sun, 15 Jul 2012 10:55:35 +0000 and
Sun, 15 Jul 2012 10:56:50 +0000, excepting the kernel version bump.
First step in reverting regressions to ata(4) subsystem during the addition of
port multiplier support.

Some namespace protection (and add greppablity).
Prefix the DRIVE_ and DRIVET_ constants from atavar.h with ATA_.
Don't use an enum for drive_type - you don't know how big it will be.
Move driver_type to avoid implicit structure padding (esp on arm).
This change is purely lexical and mechanical.

Update to 6.99.9 - this wasn't done when the SATA PMP changes
were made - I'm sure they warranted a bump.

Add sata Port MultiPlier (PMP) support to the ata bus layer,
as described in
http://mail-index.netbsd.org/tech-kern/2012/06/23/msg013442.html
PMP support in integrated to the atabus layer.
struct ata_channel's ch_drive[] is not dynamically allocated, and ch_ndrive
(renamed to ch_ndrives) closely reflects the size of the ch_drive[] array.
Add helper functions atabus_alloc_drives() and atabus_free_drives()
to manage ch_drive[]/ch_ndrives.
Add wdc_maxdrives to struct wdc_softc so that bus front-end can specify
how much drive they really support (master/slave or single).
ata_reset_drive() callback gains a uint32_t *sigp argument which,
when not NULL, will contain the signature of the device being reset.
While there, some cosmetic changes:
- added a drive_type enum to ata_drive_datas, and stop encoding the
probed drive type in drive_flags (we were out of drive flags anyway).
- rename DRIVE_ATAPIST to DRIVE_ATAPIDSCW to better reflect what this
really is
- remove ata_channel->ata_drives, it's redundant with the pointer in
ata_drive_datas
- factor out the interpretation of SATA signatures in sata_interpet_sig()

propagate these changes to the ATA HBA drivers, and add support for PMP
to ahcisata(4) and siisata(4).

Thanks to:
- Protocase (http://www.protocase.com/) which provided a system
with lots of controllers, SATA PMP and drive slots
- Conservation Genomics Laboratory, Department of Biology, New Mexico State
University for hosting the above system
- Brook Milligan, who set up remote access and has been very responsive
when SATA cable move was needed

Add a bustype_async_event_xfer_mode() callback to scsipi_bustype (which can
be NULL), so that transport-specific details of transfer mode setting/printing
can be handled more easily.
Move scsipi_async_event_xfer_mode() and scsipi_print_xfer_mode() to
scsi_base.c and split in parallel scsi and FC/SAS parts.
size of struct scsipi_bustype has changed, welcome to 6.99.5

branches: 1.112.4; 1.112.8;
Fix obvious condition snafu

branches: 1.111.2;
Spell "enough" properly.

branches: 1.110.2; 1.110.4;
Remove closes 3 & 4 from my licence. Lots of thanks to Soren Jacobsen
for the booring work !

cpu_intr_p() doesn't account for software interrupts (e.g. callouts) so
we can't use it here. Rssurect ATACH_TH_RUN, backing out
src/sys/dev/ata/ata.c 1.101
src/sys/dev/ata/ata_wdc.c 1.90
src/sys/dev/ata/atavar.h 1.77
src/sys/dev/ic/wdc.c 1.255
src/sys/dev/scsipi/atapi_wdc.c 1.108
Should fix kern/39927 and kern/39725.

branches: 1.108.2; 1.108.4;
Kill ATACH_TH_RUN and use cpu_intr_p() instead.

branches: 1.107.4; 1.107.6; 1.107.10;
Split device_t and softc for atapibus(4).

Split device_t and softc for ATA devices, as well as wd(4). Other
cosmetic changes where appropriate.

branches: 1.105.12; 1.105.16;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.104.6; 1.104.18; 1.104.20; 1.104.24;
Fix for atapi devices that have 32 bit databusses selected. The devices
won't error out but will return garbage data. Implemented by generalising
the odd-check for PIO selection to check for 4 byte alignment in offset and
size.

Note that this check could be relaxed by checking if the ATAPI/ide bus
width is 16 for the device but its unknown to me if that information is
available here.

branches: 1.103.2;
In gcc4 structures are differently aligned compared to gcc3. Lots of
structures can thus be misaligned for DMA when declared on stack.

The current atapi code accepts the misaligned dma's only screw's them up.
This small patch disables DMA for the rare transactions that are
misaligned. Note that all bulk-transfers are aligned due to page sizes.

See PR kern/34689

A little effort against kernel bloat....
Exclude ATA DMA support if no ATA DMA capable drivers are compiled in.

branches: 1.101.2; 1.101.4;
Add PIOBM (busmastering transfer using ATA PIO mode) support.
The PIOBM is used by only one driver (will be added later,
stay tuned) and intruduce an attribute "ata_piobm" so that
it will be conditionally compiled in.
The "ata_dma" (busmastering transfer using ATA DMA mode) and
"ata_udma" (busmastering transfer using ATA Ultra DMA mode)
attributes are also added for consistency, but unused for now.

Make this build with GCC 4.x.

branches: 1.99.2;
void casts to functions whose return values are ignored.

branches: 1.98.2; 1.98.6; 1.98.8;
Use device_is_active() rather than testing dv_flags for DVF_ACTIVE
directly.

branches: 1.97.2; 1.97.4; 1.97.6;
merge ktrace-lwp.

tighten the autoconf constraints by passing the atapi/ata/ata_hl attribute
to config_found()
(after some cleanup we might be able to kill SCSIPI_BUSTYPE_*)

branches: 1.95.2;
- When starting an ATA or ATAPI transfer, handle the case where (*dma_init)()
returns EINVAL, indicating that DMA cannot be done for this transfer.
Fall back to PIO in this case.
- Add a geodeide_dma_init() routine that checks to make sure that transfers
start on a 16 byte boundary, returning EINVAL if not. Works around a chip
bug that causes a hard system hang.

Problem reported and patch tested by Erik Fair.

Fix typos.

- Sprinkle const
- Avoid variable shadowing.
- Eliminate some caddr_t abuse.

branches: 1.92.2;
nuke trailing whitespace

Backing out changes to clean up scsipi. I was pointed out there were
problems i hadn't seen. To prevent lossage i'd decided to back off all
changes and let them be reviewed on tech-kern.

Part of the cleanup of sys/scsipi's use of types; rename all u_int* to
uint* and change the u_long's to uint32_t's where possible. Note that the
iocl definitions/hooks have to be ulong (or u_long) or they'll bomb out.

branches: 1.89.4; 1.89.6;
atastart() (called only at splbio(), and from interrupts) can change
drive_flags, to make sure all drive_flags manipulations are done at
splbio().

Move most of wdc_softc into a new atac_softc structure that contains
info common to all types of ATA controllers.

- Split the register handles out of struct wdc_channel into a separate
wdc_regs structure, and array of which (indexed per channel) is pointed
to by struct wdc_softc.
- Move the resulting wdc_channel structure to atavar.h and rename it to
ata_channel. Rename the corresponding flags.
- Add a "ch_ndrive" member to struct ata_channel, which indicates the
maximum number of drives that can be present on the channel. For now,
this is always 2. Add an ATA_MAXDRIVES constant that places an upper
limit on this value, also currently 2.

WDCDEBUG -> ATADEBUG.

Don't bother with bits that tell of the presence of optional callbacks;
just check the function pointers for NULL.

Move wdcstart() to ata.c and rename it to atastart().

Move wdc_exec_xfer() to ata.c and rename it ata_exec_xfer().

Move wdc_addref() and wdc_delref() to ata.c and rename them to
ata_*.

- Move wdc_xfer_pool, wdc_get_xfer(), wdc_free_xfer() to ata.c, and rename
to ata_*.
- Use a static initializer for the ata_xfer_pool.

Move wdc_probe_caps() to ata.c and rename it ata_probe_caps().

Rename some constants:
WDC_COMPLETE -> ATACMD_COMPLETE
WDC_QUEUED -> ATACMD_QUEUED
WDC_TRY_AGAIN -> ATACMD_TRY_AGAIN

Rename "struct wdc_command" to "struct ata_command".

Make datain_pio and dataout_pio function pointers in wdc_softc, which can
be overridden by the backend if desired. Add experimental code to wdc_pcmcia
to use this in memory-mapped mode, disabled by default.

Change the behavior when a transfer mode setting is rejected. For PIO, fall
back to mode 3. For DMA, fall back to mode 0. These are the minimums for
ATAPI. For UDMA, fail.

Add two helper functions -- wdc_datain_pio() and wdc_dataout_pio() -- which
encapsulate the logic for the various methods of transferring data. Use these
throughout.

Change wdc_kill_pending() to take a struct ata_drive_datas * as argument,
and kill only pending requests for this drive.
Implement a DRIVE_WAITDRAIN flag, which will cause the active command to
be killed once complete.
Other minor fixes.
Now it's possible to detach a ATA or ATAPI device from ioctl even when
a dd on the raw char partition is running.

Add
struct ata_xfer *active_xfer
to ata_queue. Now the active xfer isn't the head of the queue any more,
this makes a few things easier (this will also help for tagged queuing
support).
Remove the WDCF_ACTIVE flag, test active_xfer != NULL instead.
clean up wdc_free_xfer() and kill_xfer().
Clean up wdc_reset_channel(), and make it issue a ATAPI_SOFT_RESET if the
active command is ATAPI.
In wdc_atapi_get_params(), use AT_WAIT | AT_POLL for ATAPI_SOFT_RESET,
so that we'll use tsleep() instead of delay().
In wdc_atapi_start(), call wdc_dmawait() at the right place.

Implement an atabus control device, and define some ATA bus control
IOCTLS. Implement ATABUSIORESET, which will reset the given ATA bus.

Implement asynchronous channel reset.
Use this to reset the channel before doing a dump, instead of the hack in
wdc_exec_xfer() based on C_POLL. This hack was causing problems on
controllers with a shared queue, because we now can have C_POLL set during
concurent channels probes (problem found and analysed on sparc64 by
Martin Husemann).
This should even make core dumps marginally more reliable on ATA drives.

Add a delay(10) after re-enabling interrupts in the control register.
Some controllers/drives (e.g. SataLink 3114 with WD Raptor) require
it. Should fix kern/23808 by Chris Gilbert, patch suplied by Chris Gilbert
on tech-kern, extended to all places enabling interrupts by me.

branches: 1.69.2;
Clear drive_flags when wdc_atapi_get_params() fails. Avoid a panic later
when probing a slave-only bus in some cases.

More structure member namespace cleanup: thread -> ch_thread

More wdc_channel structure member namespace cleanup:
- channel -> ch_channel
- wdc -> ch_wdc

Rename "struct channel_softc" to "struct wdc_channel".

Tidy this file up.

Prepend "wdc_" to wait_for_drq, wait_for_unbusy, and wait_for_ready.

Rename:
- wdc_xfer to ata_xfer
- channel_queue to ata_queue
and move them to <dev/ata/atavar.h> so they can be used by non-wdc ATA
controllers. Clean up the member names of these structures while at it.

There are some cards that map the ATA control and IDE DMA registers
in a different fashion. Individually, they have the same functionality,
but their layout is different. An example of such a chipset is
the Promise 203xx.

To be able to deal with this, transform the cmd and dma bus_space handles
into an array of handles, each seperately created with bus_space_subregion.
The code generated by using the extra indirection shouldn't change much,
since the extra indirection is negated by having the offset calculation
already done in bus_space_subregion. E.g.

bus_space_write_4(tag, handle, offset, value)

becomes

bus_space_write_4(tag, handles[offset], 0, value)

Reviewed by Manuel Bouyer. Tested on wdc_isa, wdc_pcmcia, viaide, piixide (i386)
and on cmdide (sparc64).

freese->freeze, as pointed out by Frederick Bruck.

Clear chp->ch_drive[i].drive_flags if we didn't find a driver for the
drive.

Make the ATA mid-layer appears as atabus, as proposed in
http://mail-index.netbsd.org/tech-kern/2003/09/25/0006.html
This adds a device (atabus) between IDE controllers and wd or atapibus, to
have each ATA channel show up in the device tree. Later there will be atabus
devices in /dev, so that we can do IOCTL on them.
Each atabus has its own kernel thread, to handle operations that needs polling,
e.g. reset and others.

Device probing on each bus it defered to the atabus thread creation.
This allows to do the reset and basic device probes in parallel, which reduce
boot time on systems with several pciide controllers.

Remove references to University of California from my copyright notices.

1) Use config_interrupts() to attach IDE and ATAPI drives. This eliminates
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.

Set PQUIRK_ONLYBIG in the wdc-atapi frontend, obviating the need to ever test
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.

branches: 1.55.6;
remove trailing \n in panic(). approved perry.

enouth -> enough.

Do PIO transfers for len < 4 bytes. From Matthias Drochner in private mail.

branches: 1.52.2;
If the data len if not a multiple of 2 bytes, force PIO mode. DMA can't
deal with such xfers, and can wedge the system with some controllers.
It's a bug to request such xfers for ATAPI, but as the request may come
from userland we have to protect against it.

branches: 1.51.2;
More copyright fixes, pointed out by Thomas. Thanks !

Implement mstohz() as discussed on tech-kern, and use it in SCSI drivers
to convert xs->timeout to callout() parameter.

Don't clear the adapt structure of the attaching channel - this nukes
all function pointers passed in from the adapter driver.

This partly fixes PR 13480, i.e. the FREECOM CD driver works now in pcmcia
adapters.

The remaining issue (timing problems with slow cards and cardbus bridges)
is probably the cause of several other PRs too.

IF we change the ATA modes, also reset the device, so that we can retry
the command with the new mode.

Update my copyrigth.

Clean up attach of wd/atapibus:
kill ata_atapi_attach. Change atapibus to use a struct scsipi_channel instead
of ata_atapi_attach as attach arch. Create a ata_device, compatible with
scsipi_channel, to attach wd.

For ATAPI tape drives, poll for DSC (using a callout), to get the real
status of the command, and make sure the drive is ready for the next one.

add RCSIDs

branches: 1.43.2;
Some ATAPI CD occasionally reports "ATA illegal indication length" for
a request sense command, but the sense is properly transfered anyway.
Ignore the error for this case.

Add explicit support for IDE and SCSI adaptors which don't support interrupts.
On such adaptors, all transfers are done in polling mode.

OK'ed by Manuel on tech-kern.

delint: can't have an empty case item at the end of a switch; need a ; at least

Pull up the thorpej_scsipi branch to main branch.
This is a completely rewritten scsipi_xfer execution engine, and the
associated changes to HBA drivers. Overview of changes & features:
- All xfers are queued in the mid-layer, rather than doing so in an
ad-hoc fashion in individual adapter drivers.
- Adapter/channel resource management in the mid-layer, avoids even trying
to start running an xfer if the adapter/channel doesn't have the resources.
- Better communication between the mid-layer and the adapters.
- Asynchronous event notification mechanism from adapter to mid-layer and
peripherals.
- Better peripheral queue management: freeze/thaw, sorted requeueing during
recovery, etc.
- Clean separation of peripherals, adapters, and adapter channels (no more
scsipi_link).
- Kernel thread for each scsipi_channel makes error recovery much easier
(no more dealing with interrupt context when recovering from an error).
- Mid-layer support for tagged queueing: commands can have the tag type
set explicitly, tag IDs are allocated in the mid-layer (thus eliminating
the need to use buggy tag ID allocation schemes in many adapter drivers).
- support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command
will be requeued, or a REQUEST SENSE will be sent as appropriate.

Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge

branches: 1.39.2;
Force certain commands (i.e. DVD authentication) to PIO mode. They are
explicitly not supported in DMA mode, and generally return an `illegal field
in CDB' error.
This should have been done ages ago.

If the SET_FEATURE for piomode is rejected, fallback to mode 0 instead of
reporting an error.
This fixes the "piomode error (0x4)" problem with some ATAPI ZIP drives
reported on port-i386.

remove include of <vm/vm.h>

branches: 1.36.2;
Add a callback (*irqack), for controllers that need special action to ack
the interrupt once it has been ack'd on the drive.

branches: 1.35.2;
Sync copyrigth notice

Let the device thet gets an atapibus attached specify how to kill pending
transfers by giving a function pointer.
The old method always called wdc specific code.

- DMA code cleanup: pciide_dma_finish() doesn't stop/unload the current DMA op
if an IRQ was not detected, unless the force flag was given. Use this to
detect if the IRQ was for us (closer to shared IRQ for controllers which
don't have their own IRQ handler in pciide.c) and to poll for DMA xfer.
Also makes the timeout recovery code simpler.
- ATAPI cleanup: don't call controller-specific functions from atapiconf.c
(wdc_*), so that it's possible to attach an atapibus to something else
than a wdc/pciide (Hi Lennart :).
Overload struct scsi_adapter with struct atapi_adapter, defined
as struct scsi_adapter + atapi-specific callbacks. scsipi_link still points
to an scsi_adapter, atapi code casts it to atapi_adapter if needed.
Move atapi_softc to atapiconf.h so that it can be used by the underlying
controller code (e.g. atapi_wdc.c).
Add an atapi-specific callback *atapi_probedev(), which probe a drive
in a controller-specific way, allocate the sc_link and fills in the
ataparams if needed. It then calls atapi_probedev() (from atapiconf.c)
to do the generic initialisations and attach the device.
- While I'm there merge and centralise the state definitions in atavar.h.
It should now be possible to use a common ata/atapi routine to set the
drive's modes (will do later).

New callout mechanism with two major improvements over the old
timeout()/untimeout() API:
- Clients supply callout handle storage, thus eliminating problems of
resource allocation.
- Insertion and removal of callouts is constant time, important as
this facility is used quite a lot in the kernel.

The old timeout()/untimeout() API has been removed from the kernel.

Cancel requested command if controller isn't active.

Improve the downgrade logic:
- If UDMA 2 is failing try UDMA 1 first, it helps in some cases
- downgrade if we get an error in the first 4000 xfers, or if we get
4 errors in 4000 xfers if the first 4000 went without troubles.

While I'm there commit a local change I have since some time to get my CD
probed: issue a "blanck" IDENTIFY before the one used to detect slave ghosts,
with my drive the first IDENTIFY following a controller reset fails with an
aborted command ...

Pass xfer->c_bcount instead of sc_xfer->datalen to the device, so that
it gets the rigth len for a request sense (from OpenBSD).

Cancel active transfers on aic/wdc detach.
Also makes LS-120 drive works for me again.

branches: 1.27.2; 1.27.4; 1.27.6;
Cleanup the scsipi_xfer flags:
- `flags' is now gone, replaced with `xs_control' and `xs_status'.
- Massive cleanup of the control flags. Now we explicitly say that
a job is to complete asynchronously, rather than relying on side-effects,
and use a new flag to now that device discovery is being performed.
- Do SCSI device discovery interrupt-driven.

Allow to detach wdc, atapibus, wd and cd.

Add a missing 'return' in previous commit.
Don't use C_NEEDDONE anymore (it's safe to always call scsipi_done()).

- Add some debug printf (WDCPROBE) in _wdcreset_wait(), I've needed these
2 times in the past
- Set up timeout per xfer instead of per interrupt. This helps with
PIO transfer (we would call timeout()/untimout() several times for a
transfer).
- If we missed an interrupt for a PIO transfer, reset and restart it
immedialy, don't try to recover and continue. If we missed an interrupt we
may have lost a read/write cycle on the IDE bus. If this happens 1) we
corrupt data and 2) we enter an interrupt loop at the end of the xfer, as
the drive has some more data to read/write, but the host thinks the xfer is
done.
This last change fix the (or at last some of the) 'lookup after lost interrupt'
some peoples have been experiencing.

Add patch from Manuel Bouyer -- his mail message said:

By code inspection I found 2 bugs in the ATAPI code, one may be the cause
of your problem: A counter is not reset when issuing the request_sense
command. This is c_skip, the counter used to track the offset in the data
buffer when a data tranfer needs multiple phases. The effect of this is that
the sense data transfered may be stored outside the sense buffer (sense buffer
+ some, potentially several KB, offset). This can only occur in PIO mode,
DMA is not affected.
This doesn't occur in "normal" use because when reading a data CD, either
the sense is issued for a non-data command (in which case c_skip stays to
0), or an error occured and no data has been transfered, and c_skip is still 0.
I can't see a case where a data READ/WRITE could cause a sense tranfer without
error.

The second problem is that b_resid can be set to a false value (resulting of
the sense tranfer and not the data transfer). Again this is not a problem with
usual data tranfers because both values ends up being 0 when no error occurs.

If setting to pio mode < 3 fails, try to set pio mode 3 before returning
EIO. The spec says ATAPI devices should support "PIO 3 or better".
They are supposed to support less as well. Setting the device to a highter
mode than the controller shoul'nt be a problem, and this is likely what
happens with legaty ISA controllers.
Solve problem reported by Ruey-Shyang Guo.

Fix a long-standing bug with the ACAP_* flags:
- the cap field is a u_int8_t, so none of the defined flags would fit in.
Looks like nobody had a drive using 16 bytes commands.
- the ACAP_DRQ_* flags are all wrong. Just remove them and use the definitions
from ata/atareg.h, there's no need to duplicate theses. The effect of this
was that we were always polling for the command phase, even for drives
with interrupt DRQ. This didn't break until the code was changed to support
shared interrupts.

Should fix the lookup problems or 'boot hangs' reported by some users, and
kern/7111.

branches: 1.20.2;
- change the interrupt routines to take a 3rd arguments, set to 1 if we
are called from the interrupt or timeout handler, 0 otherwise.
- use this to know if we can busy-wait for wait_for_unbusy or wait_for_ready
This fixes a bug where CDs withot the DRQ_INTR capability would not busy-wait
for the CMDOUT phase.
While I'm there change 2 delay() to DELAY() for consistency, and
garbage-collect some old code from wdcintr() which has been ifdef'd out
for some time now.

Avoid busy-waiting when possible if we are not polling (real IRQ or timeout
callback). Shared PCI IRQ should now work (but still untested).

Only define WDCDEBUG if not already define (e.g. by kernel option)

If the REQUEST_SENSE command fails, issue a reset instead of XS_DRIVER_STUFFUP.
If REQUEST_SENSE succeeds but doens't return any data, return XS_SHORTSENSE
as the sense field will just contain a copy or the error register.

For IDE/ATAPI protocols botches, reset the bus and return XS_RESET
instead of XS_DRIVER_STUFFUP.

Cosmetic changes, from Soren S. Jorvan.

Make this compile again after the void * arith changes.

Avoid arithmetic on `void *' since that's not ANSI C.

Add a new quirk flags, "ADEV_NOSENSE", for devices that don't
handle properly the request sense command. Add <FX320S, , q01> as being
sense-unfriendly in the quirk table.

Keep track of DMA errors, and downgrade the transert mode (UDMA ->DMA,
DMA->PIO) in case of 2 consecutive errors. Don't downgrade if the
PIO/DMA/UDMA modes were forced by a config flag.

in wdc_softc: access the per-channel data via a pointer array instead of
an array of fixed-sized channel_softc elements. This way IDE controllers
which more than 1 channel (pciide) can extend the channel data easily
for private needs.
To avoid the double dereference at runtime, change the argument of
wdcstart() to the channel data pointer instead of the array index.

Adapt to the new scsipi_adapter interface.

In atapi_wdc.c, issue a 'REQUEST SENSE' command when appropriate.
Return XS_SENSE when the full sense info has been retrieved, or
XS_SHORTSENSE if only the sense key was available (from the error register)
Make atapi_interpret_sense() deal with this, and call scsipi_interpret_sense()
for XS_SENSE. (XXX sd_interpret_sense() and the ioctl code needs to be made
aware of XS_SHORTSENSE too ! sense hanlding for these is now less broken for
devices that support 'REQUEST SENSE')
All the ATAPI devices I have access to seems to honnor the SENSE_REQUEST
command, but I suspect some ATAPI devices will not (althouh it's mandatory).
The code should be able to deal with this, but is untested ...

Move the "warning: bcount value is %d after io" message in a WDCDEBUG_PRINT.
Ifdef DIAGNOSTIC, print the above message only if bcount is < 0.

Add a new drive flag, DRIVE_MODE, set when the drive properly reported its
PIO/DMA modes. Don't try a SET_FEATURE if it didn't report its mode.

In debug message, always print controller:channel:drive

Properly separate DEBUF_XFERS and DEBUG_FUNCS: DEBUF_XFERS is for data
transfers (may be called often), where DEBUG_FUNCS is for setup functions
(not used for normal contitions).

pciide.c: don't define WDCDEBUG, so it compiles on alpha. Correct a bogon
in the printing of DMA mode (piix3/4 only)
others: set the debug_mask to 0, so that debug messages are turned off by
default but can be easily turned on.
Reset drive_flags to 0 for unconfigured devices, so that they are ignored
later. For configured devices, reset state to 0 after probe/attach.

Merge bouyer-ide

branches: 1.1.2;
file atapi_wdc.c was initially added on branch bouyer-ide.

branches: 1.1.2;
file atapicd.c was initially added on branch bouyer-scsipi.

sys: Use membar_release/acquire around reference drop.

This just goes through my recent reference count membar audit and
changes membar_exit to membar_release and membar_enter to
membar_acquire -- this should make everything cheaper on most CPUs
without hurting correctness, because membar_acquire is generally
cheaper than membar_enter.

sys: Membar audit around reference count releases.

If two threads are using an object that is freed when the reference
count goes to zero, we need to ensure that all memory operations
related to the object happen before freeing the object.

Using an atomic_dec_uint_nv(&refcnt) == 0 ensures that only one
thread takes responsibility for freeing, but it's not enough to
ensure that the other thread's memory operations happen before the
freeing.

Consider:

Thread A Thread B
obj->foo = 42; obj->baz = 73;
mumble(&obj->bar); grumble(&obj->quux);
/* membar_exit(); */ /* membar_exit(); */
atomic_dec -- not last atomic_dec -- last
/* membar_enter(); */
KASSERT(invariant(obj->foo,
obj->bar));
free_stuff(obj);

The memory barriers ensure that

obj->foo = 42;
mumble(&obj->bar);

in thread A happens before

KASSERT(invariant(obj->foo, obj->bar));
free_stuff(obj);

in thread B. Without them, this ordering is not guaranteed.

So in general it is necessary to do

membar_exit();
if (atomic_dec_uint_nv(&obj->refcnt) != 0)
return;
membar_enter();

to release a reference, for the `last one out hit the lights' style
of reference counting. (This is in contrast to the style where one
thread blocks new references and then waits under a lock for existing
ones to drain with a condvar -- no membar needed thanks to mutex(9).)

I searched for atomic_dec to find all these. Obviously we ought to
have a better abstraction for this because there's so much copypasta.
This is a stop-gap measure to fix actual bugs until we have that. It
would be nice if an abstraction could gracefully handle the different
styles of reference counting in use -- some years ago I drafted an
API for this, but making it cover everything got a little out of hand
(particularly with struct vnode::v_usecount) and I ended up setting
it aside to work on psref/localcount instead for better scalability.

I got bored of adding #ifdef __HAVE_ATOMIC_AS_MEMBAR everywhere, so I
only put it on things that look performance-critical on 5sec review.
We should really adopt membar_enter_preatomic/membar_exit_postatomic
or something (except they are applicable only to atomic r/m/w, not to
atomic_load/store_*, making the naming annoying) and get rid of all
the ifdefs.

Merge thorpej-cfargs2.

branches: 1.92.8;
Merge thorpej-cfargs branch:

Simplify and make extensible the config_search() / config_found() /
config_attach() interfaces: rather than having different variants for
which arguments you want pass along, just have a single call that
takes a variadic list of tag-value arguments.

Adjust all call sites:
- Simplify wherever possible; don't pass along arguments that aren't
actually needed.
- Don't be explicit about what interface attribute is attaching if
the device only has one. (More simplification.)
- Add a config_probe() function to be used in indirect configuiration
situations, making is visibly easier to see when indirect config is
in play, and allowing for future change in semantics. (As of now,
this is just a wrapper around config_match(), but that is an
implementation detail.)

Remove unnecessary or redundant interface attributes where they're not
needed.

There are currently 5 "cfargs" defined:
- CFARG_SUBMATCH (submatch function for direct config)
- CFARG_SEARCH (search function for indirect config)
- CFARG_IATTR (interface attribte)
- CFARG_LOCATORS (locators array)
- CFARG_DEVHANDLE (devhandle_t - wraps OFW, ACPI, etc. handles)

...and a sentinel value CFARG_EOL.

Add some extra sanity checking to ensure that interface attributes
aren't ambiguous.

Use CFARG_DEVHANDLE in MI FDT, OFW, and ACPI code, and macppc and shark
ports to associate those device handles with device_t instance. This
will trickle trough to more places over time (need back-end for pre-OFW
Sun OBP; any others?).

branches: 1.91.22;
The atapibus detach path did hold the channel mutex while calling into autoconf,
which would trigger a panic when unplugging a USB ATAPI CDROM.

Align detach code for scsibus and atapibus to fix this.

Also avoid races when detaching devices by replacing callout_stop with
callout_halt.

branches: 1.90.8;
reference count adapter mutex possibly shared by multiple channels.

fix error in atapibusdetach, when a child device cannot be detached,
keep atapibus instance alive.

Make scsipi framework MPSAFE.

Data structures are now protected by a per-adapter mutex at IPL_BIO
that is created by the scsibus or atapibus instance when the adapter
is configured.
The enable reference counter and the channel freeze counter which are
currently used by HBA code before the adapter is configured, are made
atomic.
The target drivers are now all tagged as D_MPSAFE.

Almost all HBA drivers still require the kernel lock to present,
so all callbacks into HBA code are still protected by kernel lock
unless the driver is tagged as SCSIPI_ADAPT_MPSAFE.

TODO: refactor sd and cd to use dksubr.

branches: 1.88.2;
Sync with scsiconf.c. (use aprint_normal(9) for "not configured")

branches: 1.87.6;
Don't probe beyond chan_ntargets in atapi_probe_bus in the "all" case.

PR/48626: ahci_atapi_probe_device panic with kmemguard

branches: 1.86.2; 1.86.4;
Take the kernel lock in atapibusdetach just like atapibuschilddet.

Fixes kassert in scsipi_lookup_periph when I press the power button
on one of my laptops (and maybe another one) to power it off.

ok mrg

Expand struct scsipi_bustype {} in a ABI-backward-compatible way to
pass more informations about the bus:
- bustype_type has 2 different bytes, one holding the existing
SCSIPI_BUSTYPE_* (scsi, atapi, ata), and one for a per-SCSIPI_BUSTYPE_*
subtype. Introduce macros to build or extract bustype_type.
- for SCSIPI_BUSTYPE_SCSI, define subtypes for parallel SCSI, Fibre Channel,
SAS and USB, to specify the transport method. SCSIPI_BUSTYPE_SCSI_PSCSI
is 0 so that bustype_type value doesn't change for existing code
- for non-SCSIPI_BUSTYPE_SCSI busses there's no defined subtype yet,
so the bustype_type value doesn't change.
- provide scsi_fc_bustype, scsi_sas_bustype and scsi_usb_bustype
along with scsi_bustype to be used by bus driver where appropriate
- scsipi_print_xfer_mode(): more existing code under a
(SCSIPI_BUSTYPE_SCSI, SCSIPI_BUSTYPE_SCSI_PSCSI) case, as
sync/wide parameters only make sense for parallel SCSI.
For (SCSIPI_BUSTYPE_SCSI, SCSIPI_BUSTYPE_SCSI_FC) and
(SCSIPI_BUSTYPE_SCSI, SCSIPI_BUSTYPE_SCSI_SAS), only print
tagged queing status if enabled. Just be silent for other
bustypes.

This change is prompted by this problem:
right now, FC (e.g. isp(4)) and SAS (e.g. mfi(4)) don't
do anything for ADAPTER_REQ_SET_XFER_MODE, and especially never
call scsipi_async_event(ASYNC_EVENT_XFER_MODE), so sd(4) always
runs untagged. Doing a scsipi_async_event(ASYNC_EVENT_XFER_MODE) with
appropriate parameters is enough to enable tagged queuing,
but then scsipi will print:
sd0: async, 8-bit transfers, tagged queueing
which is harmless (async, 8-bit transfers doens't make sense on SAS anyway)
but will confuse users. With this change scsipi will only print:
sd0: tagged queueing
which is correct.

In the long run, knowning the underlying transport in scsipi will
allow better handling of device which are not parallel SCSI.

Another change adding an extra callback to struct scsipi_bustype {}
will come (so that scsipi_print_xfer_mode(), which is SCSI-specific,
can be moved out of scsipi_base, and split into per-subtype callback),
but this will break kernel ABI and so is not suitable for
netbsd-6, so will be commmited later. The above is enough to get
tagged queuing on FC and SAS in netbsd-6.

take the kernel lock during detach of atapibus as well.

branches: 1.83.8; 1.83.12; 1.83.14;
Update scsiverbose module to use module_autoload() rather than module_load().
Load the module right before each attempt to use its features, and let the
module subsystem handle unloading.

Extract SCSIVERBOSE into a kernel module. The module can be builtin
by defining 'options SCSIVERBOSE' in the kernel config file (no change
from current behavior), or it can be loaded at boot time on those
architectures that support the boot loader's "load" command.

The module is built for all architectures, whether or not SCSI or
atapi support exists.

Remove superfluous activation hooks.

Remove closes 3 & 4 from my licence. Lots of thanks to Soren Jacobsen
for the booring work !

struct cfdata * -> cfdata_t, no functional changes intended.

Detach atapibus(4), scsibus(4), cd(4), and sd(4) during shutdown.
Destroy sd->sc_callout in sddetach(). Delete some dead code in
cddetach().

branches: 1.77.4; 1.77.12; 1.77.18;
Split device_t and softc for atapibus(4).

fix a double-free bug introduced by the following change. PR/38179.

revision 1.75
date: 2008/01/29 17:26:57; author: dyoung; state: Exp; lines: +34 -15
Use device_t. Add a handler for child detachment. Now I can detach
cd0 at atapibus0 without getting a panic when atapibus0 detaches,
later.

branches: 1.75.2; 1.75.6;
Use device_t. Add a handler for child detachment. Now I can detach
cd0 at atapibus0 without getting a panic when atapibus0 detaches,
later.

Merge jmcneill-pm branch.

branches: 1.73.2; 1.73.4;
aprintify

branches: 1.72.22; 1.72.24; 1.72.30;
__unused removal on arguments; approved by core.

- sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386

branches: 1.70.8; 1.70.10;
Use device_private().

branches: 1.69.4; 1.69.6; 1.69.8; 1.69.10; 1.69.12;
merge ktrace-lwp.

s/locdesc_t/int/g

branches: 1.67.2;
- Sprinkle const
- Avoid variable shadowing.
- Eliminate some caddr_t abuse.

a round of autoconf cleanup:
-convert submatch() style functions (passed to config_search() or
config_found_sm()) to the locator passing variants
-pass interface attributes in some cases
-make submatch() functions look uniformly as far as possible
-avoid macros which just hide cfdata members, and reduce dependencies
on "locators.h"

Const poison scsipi_dtype().

Use ANSI function decls and make use of static.

Remove the Fuji quirk from here, too.

Make the ATA mid-layer appears as atabus, as proposed in
http://mail-index.netbsd.org/tech-kern/2003/09/25/0006.html
This adds a device (atabus) between IDE controllers and wd or atapibus, to
have each ATA channel show up in the device tree. Later there will be atabus
devices in /dev, so that we can do IOCTL on them.
Each atabus has its own kernel thread, to handle operations that needs polling,
e.g. reset and others.

Device probing on each bus it defered to the atabus thread creation.
This allows to do the reset and basic device probes in parallel, which reduce
boot time on systems with several pciide controllers.

Merge the geometry and cache handling code for all direct access and optical
devices, as it's general to all SCSI MMC devices. In the process, remove
PQUIRK_NO_FLEX_PAGE.

Remove PQUIRK_BYTE5_ZERO.

Exorcise PQUIRK_NODOORLOCK.

Do a START UNIT only if the TEST UNIT READY reports that the device is not
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.

branches: 1.57.2;
Add a quirk to allow my NEO Jukebox to work again.

Add PQUIRK_NO_FLEX_PAGE for Fujitsu MO MCJ3230AP.
This drive returns bogus geometry (0 heads, 0 sectors).

Use aprint_normal() in cfprint routines.

As in scsiconf.c, don't print the numberic device type in the attach message.

Add trailing ; to CFATTACH_DECL.

Use CFATTACH_DECL().

Declare all cfattach structures const.

Introduce a new routine, config_match(), which invokes the
cfattach->ca_match function in behalf of the caller. Use it
rather than invoking cfattach->ca_match directly.

Force the initial probes to happen within the newly forked off kthread.
This eliminates problems where the underlying interrupt handler isn't the
specific layer calling scsipi_complete() for a given scsi transaction.
This avoids deadlocks where the kthread that called the autoconf routines
to configure a scsibus shouldn't be the one put to sleep waiting on a
scsipi_complete (only the scsibus's kthread should be doing that).

To avoid jitter this will force the scsibus's to probe in the order they
run through autoconf (so machines with multiple bus's don't move sd* devices
around on every reboot).

Add a chan_name to struct scsipi_channel, holding the channel's name.
Set this to dv_xname for scsibus and atapibus.
Set the name of the kernel thread to chan_name instead of controller's
name:channel number (so that we can use this name for controller-specific
threads).

The FujiFilm USB storage drive doesn't have a flex geom page and
doesn't know REQUEST SENSE.

Update my copyrigth.

No need to include atavar.h here.

Clean up attach of wd/atapibus:
kill ata_atapi_attach. Change atapibus to use a struct scsipi_channel instead
of ata_atapi_attach as attach arch. Create a ata_device, compatible with
scsipi_channel, to attach wd.

don't need <sys/types.h> when including <sys/param.h>

add RCSIDs

Pass the correct pointer to atapibusprint().

branches: 1.40.2; 1.40.4;
Use SCSI/ATAPI common definition for MODE_{SELECT,SENSE}{,_BIG}. Define
functions to send theses commands in scsipi_base.c and use them instead
of ad-hoc commands setups.

Pull up the thorpej_scsipi branch to main branch.
This is a completely rewritten scsipi_xfer execution engine, and the
associated changes to HBA drivers. Overview of changes & features:
- All xfers are queued in the mid-layer, rather than doing so in an
ad-hoc fashion in individual adapter drivers.
- Adapter/channel resource management in the mid-layer, avoids even trying
to start running an xfer if the adapter/channel doesn't have the resources.
- Better communication between the mid-layer and the adapters.
- Asynchronous event notification mechanism from adapter to mid-layer and
peripherals.
- Better peripheral queue management: freeze/thaw, sorted requeueing during
recovery, etc.
- Clean separation of peripherals, adapters, and adapter channels (no more
scsipi_link).
- Kernel thread for each scsipi_channel makes error recovery much easier
(no more dealing with interrupt context when recovering from an error).
- Mid-layer support for tagged queueing: commands can have the tag type
set explicitly, tag IDs are allocated in the mid-layer (thus eliminating
the need to use buggy tag ID allocation schemes in many adapter drivers).
- support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command
will be requeued, or a REQUEST SENSE will be sent as appropriate.

Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge

The ZiO! MMC adapter doesn't have a flex geometry page.

Add quirks for Panasonic MultiMediaCard adapter (attaches via USB).

branches: 1.36.2;
constify

Let the device thet gets an atapibus attached specify how to kill pending
transfers by giving a function pointer.
The old method always called wdc specific code.

- DMA code cleanup: pciide_dma_finish() doesn't stop/unload the current DMA op
if an IRQ was not detected, unless the force flag was given. Use this to
detect if the IRQ was for us (closer to shared IRQ for controllers which
don't have their own IRQ handler in pciide.c) and to poll for DMA xfer.
Also makes the timeout recovery code simpler.
- ATAPI cleanup: don't call controller-specific functions from atapiconf.c
(wdc_*), so that it's possible to attach an atapibus to something else
than a wdc/pciide (Hi Lennart :).
Overload struct scsi_adapter with struct atapi_adapter, defined
as struct scsi_adapter + atapi-specific callbacks. scsipi_link still points
to an scsi_adapter, atapi code casts it to atapi_adapter if needed.
Move atapi_softc to atapiconf.h so that it can be used by the underlying
controller code (e.g. atapi_wdc.c).
Add an atapi-specific callback *atapi_probedev(), which probe a drive
in a controller-specific way, allocate the sc_link and fills in the
ataparams if needed. It then calls atapi_probedev() (from atapiconf.c)
to do the generic initialisations and attach the device.
- While I'm there merge and centralise the state definitions in atavar.h.
It should now be possible to use a common ata/atapi routine to set the
drive's modes (will do later).

Change a printf() to a panic() since the kernel is going to die on the next
line anyway.

atapiprint() does not exist.

Add ADEV_NOSENSE quirk for HITACHI CDR-7730.

Patch sent by Hume Smith in kern/9489.

Nobody said no to adding a pointer to original scsi inquiry data
to the scsibus attach args. Make sure it's nulled for ATAPI. Also,
for scsiconf.c, modify SENA's quirk entry.

Cancel active transfers on aic/wdc detach.
Also makes LS-120 drive works for me again.

branches: 1.28.2; 1.28.4; 1.28.6;
Cleanup the scsipi_xfer flags:
- `flags' is now gone, replaced with `xs_control' and `xs_status'.
- Massive cleanup of the control flags. Now we explicitly say that
a job is to complete asynchronously, rather than relying on side-effects,
and use a new flag to now that device discovery is being performed.
- Do SCSI device discovery interrupt-driven.

Allow to detach wdc, atapibus, wd and cd.

Add another CD to the quirk table.

branches: 1.25.2; 1.25.4; 1.25.6;
Revert to 1.23, this was not supposed to be checked in yet.

For CDIOCCLOSE, ignore media changes, as some drive seems to return this
info here.

Another CD drive that doesn't support REQUEST SENSE, from Patrick Welche.

Add a mitsumi CD that can't handle REQUEST SENSE command. From
Johan Danielsson <joda@pdc.kth.se>.

Move test for SDEV_NOSTARTUNIT quirk from sd.c to scsipi_start().
Add a SDEV_NOSTARTUNIT quirk entry for BCD-16X 1997-04-25", "", "VER 2.2"
CD-rom (from Michael Santos).

"Memorex CRW-2642" CD-R can't properly handle REQUEST SENSE command.
Add it to the quick table, per PR kern/6844.

Change quirk entry: "FX320S", "", " q01" to: "FX320S", "", "q01". This
space does seem to matter...

Add a new quirk flags, "ADEV_NOSENSE", for devices that don't
handle properly the request sense command. Add <FX320S, , q01> as being
sense-unfriendly in the quirk table.

Leave some room for the string-terminating char in temp. buffers.
Pointed out by Scott Presnell (thanks !).

When allocating a device's scsipi_link, initialize the pending_xfers
queue.

Add a reference to the adapter before probing the bus, and delete it
once we are done probing.

Make this file compile again with -DATAPI_DEBUG_PROBE.

Merge bouyer-ide

kill the last remnants of __BROKEN_INDIRECT_CONFIG. (only the pica port
used it, and it's non-working and apparently slated for replacement.)

Improve generation of default disklabels:
-store printable product ID in cd's and sd's softc, use it as "typename"
-for this, add a "destination buffer length" argument to scsipi_strvis()
-return ATAPI device type for ATAPI devices

Use the pool allocator for scsipi_xfer structures.

branches: 1.9.2; 1.9.4;
add support for an ATAPI attachment for 'sd'.
fix 'cd' driver's NCD_SCSI bogosity (was using testing wrong macro!)
clean up in various ways:
* make common atapi_mode_{sense,select}() functions.
* put ATAPI data structures in more sensible headers, split up by
device type.
* include headers a bit more carefully.
* pass flags to attachment-specific cd functions, and use them.
* get rid of SCSI bits in scsipi_base.h's scsipi_make_xs(), move
them into the correct place in scsi_base.c.
* fix minor typo in struct name in scsipiconf.h (which was apparently
never used except in a #define later in the same file).
* use __attribute__ to force 4-byte alignment for xs command store,
so that architectures trying to bus_space_write_multi_N() (where
N > 1) that data to a controller won't lose.
* clean up a few comments in typos, and make a few #defines easier to
understand/maintain.
* rename cd_link.h to cdvar.h (via repository copy). This is exactly
what a 'var' file is supposed to be.

Adjust for changes to config.

Add a few quirk entries from OpenBSD.

Add a quirk for a sanyo CD (From OpenBSD).

Add NOCAPACITY quirk entry for MATSHITA CR-574, rev 1.02.

branches: 1.4.2;
Cosmetic changes to keep coding style consistency in this directory;

- Indent with tab of width 8.
- Use four column to indent continuation line.
- Fold long line if possible.
- Use return (xx) instead of return xx.
- Compare pointer against NULL instead of testing like boolean.
- Delete whitespace at the end of line.
- Delete whitespace in front of function call operator.
- Delete whitespace after cast.
- Dereference a pointer to function explicitly.
- Add an empty line after local variable declaration.
- Use NULL instead of (char *)0.
- Dont use block for single statement.

Added a quirk entry for the SANYO CRD-254P (from OpenBSD), and
one for the MATSHITA CR-574 (from Josef Wulf <wulf@ping.net.au> via OpenBSD).

branches: 1.2.2;
Merge scsipi branch in the mainline. This add support for ATAPI devices
(currently only CD-ROM drives on i386). The sys/dev/scsipi system provides 2
busses to which devices can attach (scsibus and atapibus). This needed to
change some include files and structure names in the low level scsi drivers.

branches: 1.1.2;
file atapiconf.c was initially added on branch bouyer-scsipi.

Remove closes 3 & 4 from my licence. Lots of thanks to Soren Jacobsen
for the booring work !

branches: 1.22.4;
Split device_t and softc for atapibus(4).

branches: 1.21.48;
This file needs to know the size of "struct device", so include
<sys/device.h> here. This should allow the sparc GENERIC_SUN4U and
sparc64 kernels to build sata_subr.c again.

branches: 1.20.20; 1.20.22;
merge ktrace-lwp.

branches: 1.19.12;
Change the way bustype_cmd is used. Rather than having it be responsible for
calling scsipi_make_xs() and scsipi_execute_xs(), instead push these into
scsipi_command. Make bustype_cmd and PHOLD/PRELE be called from
scsipi_execute_xs(). This allows us to create a xfer structure -- possibly on
the stack -- and call scsipi_execute_xs() directly.

Make the xxstart() functions reentrant again, as some drivers HBA can call
scsipi_done() from their scsipi_request().
For this, add a struct scsipi_xfer * argument to scsipi_command().
If not NULL scsipi_command() will use this to enqueue this xfer, otherwise
it'll try to allocate a new one. This scsipi_xfer has to be allocated
and initialised by scsipi_make_xs() or equivalent.
In xxstart(), allocate a scsipi_xfer using scsipi_make_xs(), and if not NULL,
dequeue the buffer before calling scsipi_command(). This makes sure that
scsipi_command() will not fail, and also makes sure that xxstart() won't
be called again between the BUFQ_PEEK() and BUFQ_GET().

Fix "dequeued wrong buf" panics reported by Juergen Hannken-Illjes in
private mail and Andreas Wrede on current-users@.
Thanks to Jason Thorpe and Chuck Silver for review, and Andreas Wrede for
testing the patch.

De-__P.

Protect against multiple inclusion.

Remove an unused prototype.

branches: 1.14.16; 1.14.18;
Update my copyrigth.

Clean up attach of wd/atapibus:
kill ata_atapi_attach. Change atapibus to use a struct scsipi_channel instead
of ata_atapi_attach as attach arch. Create a ata_device, compatible with
scsipi_channel, to attach wd.

branches: 1.12.2;
Use SCSI/ATAPI common definition for MODE_{SELECT,SENSE}{,_BIG}. Define
functions to send theses commands in scsipi_base.c and use them instead
of ad-hoc commands setups.

Pull up the thorpej_scsipi branch to main branch.
This is a completely rewritten scsipi_xfer execution engine, and the
associated changes to HBA drivers. Overview of changes & features:
- All xfers are queued in the mid-layer, rather than doing so in an
ad-hoc fashion in individual adapter drivers.
- Adapter/channel resource management in the mid-layer, avoids even trying
to start running an xfer if the adapter/channel doesn't have the resources.
- Better communication between the mid-layer and the adapters.
- Asynchronous event notification mechanism from adapter to mid-layer and
peripherals.
- Better peripheral queue management: freeze/thaw, sorted requeueing during
recovery, etc.
- Clean separation of peripherals, adapters, and adapter channels (no more
scsipi_link).
- Kernel thread for each scsipi_channel makes error recovery much easier
(no more dealing with interrupt context when recovering from an error).
- Mid-layer support for tagged queueing: commands can have the tag type
set explicitly, tag IDs are allocated in the mid-layer (thus eliminating
the need to use buggy tag ID allocation schemes in many adapter drivers).
- support for QUEUE FULL and CHECK CONDITION status in mid-layer; the command
will be requeued, or a REQUEST SENSE will be sent as appropriate.

Just before the merge syssrc has been tagged with thorpej_scsipi_beforemerge

branches: 1.10.6;
Let the device thet gets an atapibus attached specify how to kill pending
transfers by giving a function pointer.
The old method always called wdc specific code.

- DMA code cleanup: pciide_dma_finish() doesn't stop/unload the current DMA op
if an IRQ was not detected, unless the force flag was given. Use this to
detect if the IRQ was for us (closer to shared IRQ for controllers which
don't have their own IRQ handler in pciide.c) and to poll for DMA xfer.
Also makes the timeout recovery code simpler.
- ATAPI cleanup: don't call controller-specific functions from atapiconf.c
(wdc_*), so that it's possible to attach an atapibus to something else
than a wdc/pciide (Hi Lennart :).
Overload struct scsi_adapter with struct atapi_adapter, defined
as struct scsi_adapter + atapi-specific callbacks. scsipi_link still points
to an scsi_adapter, atapi code casts it to atapi_adapter if needed.
Move atapi_softc to atapiconf.h so that it can be used by the underlying
controller code (e.g. atapi_wdc.c).
Add an atapi-specific callback *atapi_probedev(), which probe a drive
in a controller-specific way, allocate the sc_link and fills in the
ataparams if needed. It then calls atapi_probedev() (from atapiconf.c)
to do the generic initialisations and attach the device.
- While I'm there merge and centralise the state definitions in atavar.h.
It should now be possible to use a common ata/atapi routine to set the
drive's modes (will do later).

Cancel active transfers on aic/wdc detach.
Also makes LS-120 drive works for me again.

branches: 1.7.6; 1.7.12; 1.7.14; 1.7.16;
Merge bouyer-ide

branches: 1.6.2;
Backout previous change, and rather, remove all protection but the one
actually currently required. Suggested by Charles M. Hannum.

Fix or add protection for mutiple inclusion.

add support for an ATAPI attachment for 'sd'.
fix 'cd' driver's NCD_SCSI bogosity (was using testing wrong macro!)
clean up in various ways:
* make common atapi_mode_{sense,select}() functions.
* put ATAPI data structures in more sensible headers, split up by
device type.
* include headers a bit more carefully.
* pass flags to attachment-specific cd functions, and use them.
* get rid of SCSI bits in scsipi_base.h's scsipi_make_xs(), move
them into the correct place in scsi_base.c.
* fix minor typo in struct name in scsipiconf.h (which was apparently
never used except in a #define later in the same file).
* use __attribute__ to force 4-byte alignment for xs command store,
so that architectures trying to bus_space_write_multi_N() (where
N > 1) that data to a controller won't lose.
* clean up a few comments in typos, and make a few #defines easier to
understand/maintain.
* rename cd_link.h to cdvar.h (via repository copy). This is exactly
what a 'var' file is supposed to be.

Cosmetic changes to keep coding style consistency in this directory;

- Indent with tab of width 8.
- Use four column to indent continuation line.
- Fold long line if possible.
- Use return (xx) instead of return xx.
- Compare pointer against NULL instead of testing like boolean.
- Delete whitespace at the end of line.
- Delete whitespace in front of function call operator.
- Delete whitespace after cast.
- Dereference a pointer to function explicitly.
- Add an empty line after local variable declaration.
- Use NULL instead of (char *)0.
- Dont use block for single statement.

branches: 1.2.2;
Merge scsipi branch in the mainline. This add support for ATAPI devices
(currently only CD-ROM drives on i386). The sys/dev/scsipi system provides 2
busses to which devices can attach (scsibus and atapibus). This needed to
change some include files and structure names in the low level scsi drivers.

branches: 1.1.2;
file atapiconf.h was initially added on branch bouyer-scsipi.

s/assised/assisted/ in comment.

branches: 1.355.2;
Add quirk for devices that does not handle READ_DISCINFO

s/Ramdom/Random/ in comments.

fix various typos, mainly in comments.

fix mainly same typos as in my previous commit but outside sys/dev/dm.

Limit buffer size for device capabilities requests as a work-around for PR
kern/56109.

branches: 1.350.2;
PR/55986: Ryo Onodera: DK_BUSY must have mask as second argument. Make cd.c
consistent by also using __BIT()

branches: 1.349.2;
Avoid buffer overflow when copying from bounce buffer.
Fixes PR 54810

Don't use uninitialized pointer in split bounce buffer case and
free a partially allocated bounce buffer on error.

s/settng/setting/

s/implicitely/implicitly/

s/occurence/occurrence/

s/parition/partition/

constify

branches: 1.343.2;
Allow open of RAWPART even when no medium is loaded.
Keep errors silent if no medium is loaded.

Fixes PR kern/55104

branches: 1.342.4;
Rename min/max -> uimin/uimax for better honesty.

These functions are defined on unsigned int. The generic name
min/max should not silently truncate to 32 bits on 64-bit systems.
This is purely a name change -- no functional change intended.

HOWEVER! Some subsystems have

#define min(a, b) ((a) < (b) ? (a) : (b))
#define max(a, b) ((a) > (b) ? (a) : (b))

even though our standard name for that is MIN/MAX. Although these
may invite multiple evaluation bugs, these do _not_ cause integer
truncation.

To avoid `fixing' these cases, I first changed the name in libkern,
and then compile-tested every file where min/max occurred in order to
confirm that it failed -- and thus confirm that nothing shadowed
min/max -- before changing it.

I have left a handful of bootloaders that are too annoying to
compile-test, and some dead code:

cobalt ews4800mips hp300 hppa ia64 luna68k vax
acorn32/if_ie.c (not included in any kernels)
macppc/if_gm.c (superseded by gem(4))

It should be easy to fix the fallout once identified -- this way of
doing things fails safe, and the goal here, after all, is to _avoid_
silent integer truncations, not introduce them.

Maybe one day we can reintroduce min/max as type-generic things that
never silently truncate. But we should avoid doing that for a while,
so that existing code has a chance to be detected by the compiler for
conversion to uimin/uimax without changing the semantics until we can
properly audit it all. (Who knows, maybe in some cases integer
truncation is actually intended!)

branches: 1.341.4; 1.341.6;
The atapibus detach path did hold the channel mutex while calling into autoconf,
which would trigger a panic when unplugging a USB ATAPI CDROM.

Align detach code for scsibus and atapibus to fix this.

Also avoid races when detaching devices by replacing callout_stop with
callout_halt.

branches: 1.340.4; 1.340.6;
ignore a failed scspi_prevent when accessing the RAW_PART as before.

branches: 1.339.2;
Fix memory leak.

Found by maxv@

really keep request queued after resource shortage

cdrestart takes a softc parameter, not a periph.

add comment about "missing" dk_start.

appease gcc.

Refactored sd and cd to use common disk subroutines.

Make scsipi framework MPSAFE.

Data structures are now protected by a per-adapter mutex at IPL_BIO
that is created by the scsibus or atapibus instance when the adapter
is configured.
The enable reference counter and the channel freeze counter which are
currently used by HBA code before the adapter is configured, are made
atomic.
The target drivers are now all tagged as D_MPSAFE.

Almost all HBA drivers still require the kernel lock to present,
so all callbacks into HBA code are still protected by kernel lock
unless the driver is tagged as SCSIPI_ADAPT_MPSAFE.

TODO: refactor sd and cd to use dksubr.

Avoid calling bufq_free() from critical code sections.

branches: 1.331.2;
Use _align(2) workaround for PR kern/51141. This fixes odd CD length
reporting.

Use C99-style initializers for struct dkdriver.

Convert sys/dev to use <sys/rndsource.h>.

We have three sets of DTYPE_ constants in the kernel:
altq Drop Type
disklabel Disk Type
file Descriptor Type
(not to mention constants that contain the string DTYPE).
Let's make them two, by changing the disklabel one to be DisK TYPE since the
other disklabel constants seem to do that. Not many userland programs use
these constants (and the ones that they do are mostly in ifdefs). They will
be fixed shortly.

make more drivers use disk_ioctl, and add a dev parameter to it so that
we can merge the "easy" disklabel ioctls to it. Ultimately all this will
go do dk_ioctl once all the drivers have been converted.

forgot this one. disk_set_info includes the functionality of disk_blocksize.

branches: 1.325.2;
src is too big these days to tolerate superfluous apostrophes. It's
"its", people!

Use uint32_t for blocksize as per other uses and to fix warnings

Merge tls-earlyentropy branch into HEAD.

Add d_discard to all struct cdevsw instances I could find.

All have been set to "nodiscard"; some should get a real implementation.

Add d_discard to all struct bdevsw instances I could find.

I've set them all to nodiscard. Some of them (wd, dk, vnd, ld,
raidframe, maybe cgd) should be implemented for real.

Prevent detaching an open device unless forced.

Fix PR kern/48550 by aligning the single instance of scsipi_read_cd_cap_data
that we found misaligned in the wild so far properly for the ahcisata
driver. Also point at PR kern/48754 for the real issue.

branches: 1.318.2;
PR kern/48550: additional initialization and sanity checking on the reported
blocksize of the medium.

Change (mostly mechanically) every cdevsw/bdevsw I can find to use
designated initializers.

I have not built every extant kernel so I have probably broken at
least one build; however I've also found and fixed some wrong
cdevsw/bdevsw entries so even if so I think we come out ahead.

Turn a few __unused into __diagused

Mark a potentially unused variable

Don't return EPASSTHROUGH when succeeding in ODIOCGDINFO or ODIOCGDEFLABEL

Remove unused variables

Small patch to clarify last_lba and now using size confusingly. No functional
change.

branches: 1.311.2;
phase 1 of disk geometry cleanup:
- centralize the geometry -> plist code so that we don't have
n useless copies of it.

Prompted by PR kern/47646, zero the value buffer before running the
GET_CONFIGURATION command and check for an unexpectedly large feature
length answer afterwards.

branches: 1.309.2;
When ejecting a medium, invalidate the in core disklabel - it is not
meaningfull anymore. This makes the following cdclose() use silent
mode and finally fixes PR kern/43785.

In cdopen: make the test for media presence always silent. In all cases
requiring a message, we will most likely get that from the spinup attempt
anyway.
This avoids the spurious "Check Condition on CDB, Not Ready, Medium Not
Present, Tray Closed" messages at boot/shutdown time.

Expand struct scsipi_bustype {} in a ABI-backward-compatible way to
pass more informations about the bus:
- bustype_type has 2 different bytes, one holding the existing
SCSIPI_BUSTYPE_* (scsi, atapi, ata), and one for a per-SCSIPI_BUSTYPE_*
subtype. Introduce macros to build or extract bustype_type.
- for SCSIPI_BUSTYPE_SCSI, define subtypes for parallel SCSI, Fibre Channel,
SAS and USB, to specify the transport method. SCSIPI_BUSTYPE_SCSI_PSCSI
is 0 so that bustype_type value doesn't change for existing code
- for non-SCSIPI_BUSTYPE_SCSI busses there's no defined subtype yet,
so the bustype_type value doesn't change.
- provide scsi_fc_bustype, scsi_sas_bustype and scsi_usb_bustype
along with scsi_bustype to be used by bus driver where appropriate
- scsipi_print_xfer_mode(): more existing code under a
(SCSIPI_BUSTYPE_SCSI, SCSIPI_BUSTYPE_SCSI_PSCSI) case, as
sync/wide parameters only make sense for parallel SCSI.
For (SCSIPI_BUSTYPE_SCSI, SCSIPI_BUSTYPE_SCSI_FC) and
(SCSIPI_BUSTYPE_SCSI, SCSIPI_BUSTYPE_SCSI_SAS), only print
tagged queing status if enabled. Just be silent for other
bustypes.

This change is prompted by this problem:
right now, FC (e.g. isp(4)) and SAS (e.g. mfi(4)) don't
do anything for ADAPTER_REQ_SET_XFER_MODE, and especially never
call scsipi_async_event(ASYNC_EVENT_XFER_MODE), so sd(4) always
runs untagged. Doing a scsipi_async_event(ASYNC_EVENT_XFER_MODE) with
appropriate parameters is enough to enable tagged queuing,
but then scsipi will print:
sd0: async, 8-bit transfers, tagged queueing
which is harmless (async, 8-bit transfers doens't make sense on SAS anyway)
but will confuse users. With this change scsipi will only print:
sd0: tagged queueing
which is correct.

In the long run, knowning the underlying transport in scsipi will
allow better handling of device which are not parallel SCSI.

Another change adding an extra callback to struct scsipi_bustype {}
will come (so that scsipi_print_xfer_mode(), which is SCSI-specific,
can be moved out of scsipi_base, and split into per-subtype callback),
but this will break kernel ABI and so is not suitable for
netbsd-6, so will be commmited later. The above is enough to get
tagged queuing on FC and SAS in netbsd-6.

Fix typos in comments.

OK by wiz@

branches: 1.305.2;
Entropy-pool implementation move and cleanup.

1) Move core entropy-pool code and source/sink/sample management code
to sys/kern from sys/dev.

2) Remove use of NRND as test for presence of entropy-pool code throughout
source tree.

3) Remove use of RND_ENABLED in device drivers as microoptimization to
avoid expensive operations on disabled entropy sources; make the
rnd_add calls do this directly so all callers benefit.

4) Fix bug in recent rnd_add_data()/rnd_add_uint32() changes that might
have lead to slight entropy overestimation for some sources.

5) Add new source types for environmental sensors, power sensors, VM
system events, and skew between clocks, with a sample implementation
for each.

ok releng to go in before the branch due to the difficulty of later
pullup (widespread #ifdef removal and moved files). Tested with release
builds on amd64 and evbarm and live testing on amd64.

branches: 1.304.2;
Replace memset with incorrect size argument with M_ZERO.

branches: 1.303.4;
Fix size reporting for DVD+R/DL and BluRay's; the value was trunced due to a
32 bit trunc due to a lacking type cast. The number of sectors for such media
can be more than 1<<32-1.

branches: 1.302.2;
cd_size: if we fake a size (and I realy have no idea why this would be a
good idea), at least set up all values to the fake values, as the caller
expects.
Should fix PR kern/39904, though if noone can find out why the fake value
would be needed, we should change it to just return 0 as suggested in the PR.

If we find a device in non-2048 byte/block mode and manage to switch
it over, we need to adjust the capacity - so just read it again.

Make cdclose() silent, if we only have the raw partition open (e.g. when
probing for media and not finding any)

Add a DIOCTUR (test unit ready) ioctl, to allow userland to easily
(and silently) check for presence of media. While this would be easy
to implement via SCIOCCOMMAND, our kernel has quirks for this
command, which otherwise would not be readily available to userland
applications.

branches: 1.298.2;
various aprint_* fixes.

branches: 1.297.2;
Some usb devices come with an internal emulated umass CD drive (containing
windows drivers). I have such a device that has 0 features - avoid wrappig
features_len to a very big unsiged 32bit number in this case.

Delete do-nothing device-activation hooks.

Remove uarea swap-out functionality:

- Addresses the issue described in PR/38828.
- Some simplification in threading and sleepq subsystems.
- Eliminates pmap_collect() and, as a side note, allows pmap optimisations.
- Eliminates XS_CTL_DATA_ONSTACK in scsipi code.
- Avoids few scans on LWP list and thus potentially long holds of proc_lock.
- Cuts ~1.5k lines of code. Reduces amd64 kernel size by ~4k.
- Removes __SWAP_BROKEN cases.

Tested on x86, mips, acorn32 (thanks <mpumford>) and partly tested on
acorn26 (thanks to <bjh21>).

Discussed on <tech-kern>, reviewed by <ad>.

Significantly bump time for SCSI track/disc closure. DVD-RW when used
sequentially can significantly take more time.

Add support for DIOCGDISKINFO to disk like device drivers. Change
partutil.c::getdiskinfo to use it to get disk geometry info.
Use DIOCGWEDGEINFO ioctl to get information about partition size, if disk
driver doesn't support it use old DIOCGDINFO. This patch adds support for
wedge like devices(lvm logical volumes, ZFS zvol partitions) to newfs and
other tools.

No objections on tech-userlevel@.

Detach atapibus(4), scsibus(4), cd(4), and sd(4) during shutdown.
Destroy sd->sc_callout in sddetach(). Delete some dead code in
cddetach().

Implement MMC_TRACKINFO_DATA and MMC_TRACKINFO_AUDIO detection for cdrom and
dvdrom devices. These two compatibility codes were forgotten when those were
introduced.

bzero -> memset

Fix long standing problem with CD/DVD writing with `dd'. If done this way the
device will remain in a wait-for-more-stuff-to-be-written state delaying other
SCSI calls, esp. some read-size or read-layout calls, until its told there is
no more comming.

To fix this, on last close explicitly flush the caches releasing the device
from this mode and allowing other SCSI commands to reach it.

This fixes PR 40528

'boot -z' bogons

branches: 1.287.2;
buildfix: re-adapt to major()/minor() returning a 32bit value.

g/c BUFQ_FOO() macros and use bufq_foo() directly.

make this compile

Add ISO partition detection enabling auto-detection of iso9660 and UDF
partitions on optical media like CD/DVD/BD but also on all other media if
there is no NetBSD disklabel or MBR label.

Also fix cd's readdisklabel arguments so the ioctl's arrive at the right
device (!) and update its default label to make more sense.

branches: 1.283.2; 1.283.4;
Remove unreachable code introduced by rev 1.224. Fixes PR#38973.

branches: 1.282.2;
use device_lookup_private to get softc

branches: 1.281.2;
add support for drvctl properties

Fix SSP kernel builds.

Implement write support for the MMC framework. This consists of a setting
up of write parameters call and a call to perform a series of operations on
these devices.

Note that the MMC framework interface is still not exposed to userland
unless the expose_mmc flag is set. This is to prevent applications to build
on a possibly still moving target. This flag will eventually be removed.

branches: 1.278.2;
getiobuf(false, NULL) -> getiobuf(NULL, false)

Extend the MMC framework to also create sensible reports on audio discs.

Fixup CD and DVD drives that appear to be on crack when reporting features.
They state f.e. that a recordable CD-R is rewritable or completely forget
that the fixed packet size formatted CD-RW is strict overwrite rewritable
and not randomly rewritable.

Make the SCSI sense code "Logical Unit Not Ready, Operation In Progress"
(Sense code 0x04, 0x07) not a fatal error anymore but retry command later.
This sense code can be returned on a CD-MRW write that encounters a bad
block resulting in the drive being busy relocating it. During that time it
can return this sense code to indicate its busy for a while.

Also reduce waiting time for the other "Logical Unix Not Ready, Long Write
In Progress" to half a second.

Remove clause 3 and 4 from TNF licenses

branches: 1.273.2; 1.273.4;
Split device_t for cd(4).

branches: 1.272.6;
Merge vmlocking2 to head.

branches: 1.271.2;
Merge jmcneill-pm branch.

branches: 1.270.2; 1.270.4;
Pullup fixes from UDF write development project. This patch fixes a
conceptional bug in mmc_discinfo, improves DVD+R feature detection and
prevents corruption of mmc_trackinfo readout on some devices that return
short trackinfo structures.

branches: 1.269.4;
Merge disk init changes from the vmlocking branch. These seperate init /
destroy of 'struct disk' from attach / detach.

branches: 1.268.2; 1.268.4; 1.268.6;
Be sure not to make requests over MAXPHYS when employing bounce buffers.
It was previously possible to round up one sector length if the start block
were not properly aligned.

Resolves PR kern/36716. This may also fix the following reported issues
on port-sgimips:
http://mail-index.netbsd.org/port-sgimips/2005/09/12/0000.html
http://mail-index.netbsd.org/port-sgimips/2005/09/14/0000.html

While here, be sure to put back allocated bufs, which previously appear
to have been leaked. Further, remove the dead write bounce code which never
got executed anyway.

No complaints on tech-kern.

branches: 1.267.4;
It's not a good idea for device drivers to modify b_flags, as they don't
need to understand the locking around that field. Instead of setting
B_ERROR, set b_error instead. b_error is 'owned' by whoever completes
the I/O request.

Replace some uses of lockmgr().

branches: 1.265.2;
Merge some of the less invasive changes from the vmlocking branch:

- kthread, callout, devsw API changes
- select()/poll() improvements
- miscellaneous MT safety improvements

The cd ioctl definitions are not condusive to being called from within
the kernel (eg from withthe linux compat code).
Add CDIOCREADSUBCHANNEL_BUF and CDIOREADTOCENTRIES_BUF which are like the
existing ioctls withoutthe _BUF, except that the data area immediately
follows the control structure (and is fixed length).
Modify scsipi/cd.c so that it only ever has one 'struct cd_formatted_toc'
allocated on the stack (maybe this info ought to be in the softc)
structure since even one 800 byte on-stack buffer isn't overly friendly.
cdplay(1) still seems able to read the track-list of an ATAPI cd.

Remove a load of unnecessary casts now that 'addr' is 'void *'.

branches: 1.262.2; 1.262.4;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.261.2;
The previous fix for PR 34202 was wrong (some sizeof used the union, not
the command block) - but instead of fixing this, we can now revert it
completley: the real fix has been applied to ../ic/wdc.c in rev. 1.244,
removing the alignement constraints.

branches: 1.260.2;
Fix PR kern/34202 differently, by aligning the variables "the traditional
way".

- Call disk_blocksize(9) when we determine the physical block size of the
media. (Will this ever be anything other than 2048?)
- bounds_check_with_label() has been fixed, so put the correct partition
size in the default disklabel.

__unused removal on arguments; approved by core.

Implement/add bufq strategy setting by dkctl for the CD class devices.

Though its hardly advisable to change the CD strategy setting to anything
other than `disksort' or `prioscan' it ought to be possible.

Implement another class of `unit not ready' sense handling that is not
fatal.. A `long write in progress' is a retry again later command that is
issued when a device returns immediately after a write request but needs
some time before it can handle read requests.

- don't allocate huge arrays on the stack
- no bogus ; after }; in block statements!

- sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386

Also print blksize with %u - this is what I get for eyeballing
a patch instead of applying it directly. PR#33966.

Print an unsigned parameter with %lu, not %ld. From David A. Holland
in PR#33966.

Use last track info only if it gives a sane value. Fixes PR#34688.

branches: 1.250.2;
Oeps! I overlooked the size argument of the SCSI call. It ought to be equal
to the number of bytes we are reading: READ_DISCINFO_BIGSIZE

sorry.

The size of struct scsipi_read_discinfo_data is not even due to the data[1]
pseudo array in the structure. The scsi command issued used
sizeof(scsipi_read_discinfo_data) but included thus the last byte that
wasn't going to be inspected anyway. Using the constant
READ_DISCINFO_BIGSIZE fixes the bug.

Even though SCSI adapters *should* accept odd lengths, the Sun U10
(sparc64's) crashes in bus_space_read_multi_stream_2(). That explains the
crash.

NOTE: somewhere in NetBSD/sparc64's atapibus support there is the
assumption on the even size; this needs to be fixed! (Hi Sparc64 portmaster
:-) )

branches: 1.248.2;
Use an unsigned int for the blksize on read_cd_capacity.

Fix panic on reading/writing to a (raw) CD device and encountering an
error. It now correctly sets bp->b_resid to the full size of the buffer.
The failed SCSI read/write command allways contains the complete buffer.

When encountering a read (or write) error, the scsipi stack sets the
xs->resid to zero since the command has been accepted by the drive. The
scsi command set does not allow a partial read or write to be performed and
will signal a success or an error.

Remove two spurious empty lines in functions

Fix fencing bug on CD disk size; the comment was wrong, its the *length* of
the device and since we count from zero its equal to the number of sectors
and thus one higher than the last sector.

This fixes the read/write problems on the (raw) device where the last
sector was not readable/writable.

Add missing initializer.

Fix incorrect `loast possible lba' reporting, remove some empty lines and
add mmc classification for some still rare HD DVD device types.

branches: 1.242.6;
Use device_private().

Use device_unit().

branches: 1.240.2; 1.240.4; 1.240.6;
Add name of the processed feature like the others have.

Add support for SCSI MMC feature 0x0024: hardware assisted defect
management to be passed in mmc_discinfo's device capabilities.

When a device reports support for this the hardware will automatically
remap sectors on read- or write-errors. CD-MRW, DVD+MRW, DVD-RAM and BR-RE
support these.

branches: 1.238.2;
Adding two Ioctl's to scsipi's cd.c to abstract SCSI MMC devices. The
ioctl's and their structures are currenly hidden from inclusion for normal
userland applications to allow the MMC abstraction interface to mature
first.

Its useage is mainly for applications dedicated to handling and processing
SCSI MMC compliant devices and their media. Examples of these are CDs, DVDs
and comparable optical devices but also some magnetic swapable devices that
present themselves as a SCSI MMC compliant device.

Its current use is currently the to be committed in-kernel UDF filingsystem
and its userland tools.

branches: 1.237.2;
fix typo.

remove B_EINTR as it isn't used anymore.

- add simple functions to allocate/free a buffer for i/o.
- make bufpool static.

branches: 1.234.2;
Improve disc size reporting. The current discsize thats reported is the
current recorded extent and not the complete maximum extent of the disc.

An empty CD/DVD disc is seen as having an invalid length and given 800 Mb
size. This makes it possible to write to a CD-R using `dd' as some tend to
do.

However when an extent is recorded the initial recorded length is seen as
valid and is thus set as absolute upperbound to read and write actions
preventing extention of the disc.

Committed after consultation on tech-kern with a positive advice by Manuel
Bouyer.

delete extraneous verbiage.

Clean up cdgetdisklabel. It was jumping over a return...

merge ktrace-lwp.

Fix illogical use of extra variable. It was mainly to get a better
readability in KNF. This solution gives better readability without the
extra variable.

Second stage of read TOC modifications. Introducing the various forms of
TOC reading. Note that the external interfaces haven't changed. Only the
formatted toc is requestable.

The read msinfo command hasn't been changed in this patch though it could
become more clever using the CD_TOC_MSINFO format.

- change the way to specify a bufq strategy. (by string rather than by number)
- rather than embedding bufq_state in driver softc,
have a pointer to the former.
- move bufq related functions from kern/subr_disk.c to kern/subr_bufq.c.
- rename method to strategy for consistency.
- move some definitions which don't need to be exposed to the rest of kernel
from sys/bufq.h to sys/bufq_impl.h.
(is it better to move it to kern/ or somewhere?)
- fix some obvious breakage in dev/qbus/ts.c. (not tested)

Cleanup and fix cd_size() function; it was returning wrong results.
Also factor out read_cd_capacity()

Implement DIOCCACHESYNC for SCSI/ATAPI cd/mmc recordable devices found by
scsipi/cd.c by issueing the SYNCHRONISE CACHES scsi/atapi call as defined
per MMC standard.

Document executed SCSI commands with the IOCTL's to prevent confusion.

Always allow open() on the raw partition to succeed, even if any of the
SCSI command in cdopen() fails. Fix PR kern/30288 by Pavel Cahyna.

branches: 1.223.2;
- Sprinkle const
- Avoid variable shadowing.
- Eliminate some caddr_t abuse.

fix more SSD_RCODE_VALID misuse introduced in cleanup

introduce a function to drain bufq and use it where appropriate.

branches: 1.220.2;
nuke trailing whitespace

Part 1 of a cleanup pass over the SCSI subsystem. The aim is to name
everything "scsi_*", since we really are talking about the SCSI command
set, ATAPI transport not withstanding. Improve the names of many structures,
and prepend "SCSI_" onto all SCSI command opcodes. Place items described
by the SCSI Primary Commands document into scsi_spc.h.

Backing out changes to clean up scsipi. I was pointed out there were
problems i hadn't seen. To prevent lossage i'd decided to back off all
changes and let them be reviewed on tech-kern.

Fix LP64 problems introduced by my u_long->uint32_t conversion.

Part of the cleanup of sys/scsipi's use of types; rename all u_int* to
uint* and change the u_long's to uint32_t's where possible. Note that the
iocl definitions/hooks have to be ulong (or u_long) or they'll bomb out.

As part of cleaning up sys/scsipi, replace all u_char by uint8_t and
replace all `short' with int16_t.

As in revision 1.213, not all information is returned in read-TOC format 0. In order to add a new generic read-toc ioctl, i renamed the origional to _f0 to indicate its only format 0. The routines can then be changed one at a time to the new generic calling. This patch fixes two calls i'd overseen.... silly me forgot to recompile :(

The code asumes all programs want to know is TOC response format 0. Support
for TOC response format 1 and 2 are mandatory on CD/DVD too and provide
more information.

Next an IOCTL needs to be implemented that can read all TOC formats in a
generic way. This is pending.

branches: 1.212.2; 1.212.4;
USe more appropriate macro/struct names for READ/WRITE (6) and
READ/WRITE (10).

move buffer queue related stuffs from buf.h to their own header, bufq.h.

Fix debug message output args to match yamt's src/sys/sys/buf.h changes.

Standardize some variable names and the calling pattern for scsipi_command().
Use void pointer casts.

Remove the "xfer" argument to scsipi_command().

In places where we've already called scsipi_make_xs(), call scsipi_execute_xs()
directly rather than going through scsipi_command().

Do not manipulate xs->bp in "generic" code -- do it only in the psw_done
routine. As part of this, pass down our pre-parsed error code -- though this
interface will probably change later to accomodate better error handling.

Make the xxstart() functions reentrant again, as some drivers HBA can call
scsipi_done() from their scsipi_request().
For this, add a struct scsipi_xfer * argument to scsipi_command().
If not NULL scsipi_command() will use this to enqueue this xfer, otherwise
it'll try to allocate a new one. This scsipi_xfer has to be allocated
and initialised by scsipi_make_xs() or equivalent.
In xxstart(), allocate a scsipi_xfer using scsipi_make_xs(), and if not NULL,
dequeue the buffer before calling scsipi_command(). This makes sure that
scsipi_command() will not fail, and also makes sure that xxstart() won't
be called again between the BUFQ_PEEK() and BUFQ_GET().

Fix "dequeued wrong buf" panics reported by Juergen Hannken-Illjes in
private mail and Andreas Wrede on current-users@.
Thanks to Jason Thorpe and Chuck Silver for review, and Andreas Wrede for
testing the patch.

Fix comment: xxstart() can also be called from xxrestart()

Improve handling of memory shortage, to fix problems like:
sd3(mpt0:0:1:0): unable to allocate scsipi_xfer
sd3: not queued, error 12
Havard Eidnes's analysis of this problem is that the scsipi_xfer pool is
competing for resources with other pools, including the the inode and vnode
pools which can grow quite large.

*_scsipi_cmd(): don't biodone the buffer if scsipi_make_xs() fails, let the
caller deal with the problem
start function of block devices drivers: dequeue the buffer after the
scsipi_command() call. If scsipi_command() fails with ENOMEM don't dequeue
the buffer, and schedule a callout to call the start function after
some delay.
scsipi_init(): prime the scsipi_xfer_pool with one page. This ensure that
there is always some scsipi_xfer to play with. If scsipi_command() fails
because of pool_get(), we're sure there will be resources available later,
when the pending commands have completed.

Reviewed by Jason Thorpe and Havard Eidnes.
Todo: remove the "unable to allocate scsipi_xfer" and "not queued, error %d"
printfs, but I choose to keep them for now, to help make sure the code does
what it should.

Use ANSI function decls and make use of static.

Some older devices do not understand the `disable block descriptor' bit in
the mode sense request. So fall back on mode sense data including a block
descriptor section.

See also sd.c rev. 1.214. Again: should we bother trying DBD at all?

branches: 1.200.2;
Backout an obvious bug introduced in rev. 1.145.

store a i/o priority hint in struct buf for buffer queue discipline.

Spell address with two d's. Inspired by similar changes in OpenBSD,
originating from Jonathon Gray and forwarded by jmc@openbsd.

Merge the geometry and cache handling code for all direct access and optical
devices, as it's general to all SCSI MMC devices. In the process, remove
PQUIRK_NO_FLEX_PAGE.

Don't be silent if there is no media present.

Make sure the "raw partition" can always be opened again.

In the test for whether to start the unit, I used the wrong error code.
ENODEV is only returned when we get "medium not present," which we can fail
immediately on. All other "not ready" cases return EIO.

Oops. Fix a typo in cd_mode_select().

Do a START UNIT only if the TEST UNIT READY reports that the device is not
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.

Attempt to deal with Martin's weirdass Sun drive by setting the mode sense
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.

Eliminate the separate ATAPI and SCSI attachments for "cd".

Nuke some printf()s.

Set PQUIRK_ONLYBIG in the wdc-atapi frontend, obviating the need to ever test
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.

Add a cast to avoid an integer overflow.
Fixes playing (at least some) DVDs.
Patch from Tsubai Masanari (tsubai at iri.co.jp) -- thanks a lot!

If a CD drive reports != 2048 byte block size, try to change it into
2048 byte mode. Fixes PR kern/22090.
Reviewed by Manuel Bouyer.

branches: 1.185.2;
Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.

Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records. The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.

Bump the kernel rev up to 1.6V

Change bounds_check_with_label() to take a pointer to the disk structure,
rather than the label itself. This paves the way for some future changes.

Change return type of readdisklabel() to const char *
I hope I've found all the correct places!

Check RAW_PART against the media size instead of the disklabel.
Add the media size in 512-byte sectors to the softc, to avoid
some 64 bit computations. Bump the capacity stored in softcs
for disks to 64 bits.

use PRId64 to printf bp->b_blkno, which is of type daddr_t

Add a new BUF_INIT() macro which initializes b_dep and b_interlock, and
use it. This fixes a few places where either b_dep or b_interlock were
not properly initialized.

Make the buffer cache code MP-safe.

Test callout_pending(), not callout_active(), and eliminate now-unnecessary
callout_deactivate() calls.

Being able to make image backups of your whole disk, not to mention not
causing certain ata disks to lock up by reading beyond the last block,
beats catering to broken devices.

bounds_check_with_label() RAW_PART too.

The Double-Semi-Colon Police.

Use the correct byte positions to read data returned by DVD_LU_SEND_RPC_STATE.

Another sizeof -> 4 + 2048 fix for dvd_read_manufact after malloc changes

In dvd_read_disckey, sizeof(buf) is no longer '4 + 2048' after the
malloc change. Replace 'sizeof(buf)' with '4 + 2048' -- this makes
dvd_read_disckey work properly again.

Don't waste too much kernel stack space on (infrequent) ioctl operations,
use malloc instead for temp space.

Fix signed/unsigned comparison warnings.

implement separate read/write disk statistics:
- disk_unbusy() gets a new parameter to tell the IO direction.
- struct disk_sysctl gets 4 new members for read/write bytes/transfers.
when processing hw.diskstats, add the read&write bytes/transfers for
the old combined stats to attempt to keep backwards compatibility.

unfortunately, due to multiple bugs, this will cause new kernels and old
vmstat/iostat/systat programs to fail. however, the next time this is
change it will not fail again.

this is just the kernel portion.

merge kqueue branch into -current

kqueue provides a stateful and efficient event notification framework
currently supported events include socket, file, directory, fifo,
pipe, tty and device changes, and monitoring of processes and signals

kqueue is supported by all writable filesystems in NetBSD tree
(with exception of Coda) and all device drivers supporting poll(2)

based on work done by Jonathan Lemon for FreeBSD
initial NetBSD port done by Luke Mewburn and Jason Thorpe

remove all vesitages of dk_establish().

Merge the gehenna-devsw branch into the trunk.

This merge changes the device switch tables from static array to
dynamically generated by config(8).

- All device switches is defined as a constant structure in device drivers.

- The new grammer ``device-major'' is introduced to ``files''.

device-major <prefix> char <num> [block <num>] [<rules>]

- All device major numbers must be listed up in port dependent majors.<arch>
by using this grammer.

- Added the new naming convention.
The name of the device switch must be <prefix>_[bc]devsw for auto-generation
of device switch tables.

- The backward compatibility of loading block/character device
switch by LKM framework is broken. This is necessary to convert
from block/character device major to device name in runtime and vice versa.

- The restriction to assign device major by LKM is completely removed.
We don't need to reserve LKM entries for dynamic loading of device switch.

- In compile time, device major numbers list is packed into the kernel and
the LKM framework will refer it to assign device major number dynamically.

Remove the old device buffer queue interface.

Approved by: Jason R. Thorpe <thorpej@wasabisystems.com>