ugen and ugenif share the same /dev/ugenN.xx namespace in such a way
that the device unit number does not necessarily match the /dev/ugenN.xx
unit number (N). If you ONLY have ugen devices, it happens to work out
and devpubd scripts can be extremely naive. If you ONLY have ugenif
devices, it also happens to work out, but your devpubd scripts have to
slightly more informed. If you have a mix of ugen AND ugenif devices,
though, you're pretty much out of luck.

So, this change adds a "ugen-unit" device property which devpubd scripts
can query to determine which /dev/ugenN.xx nodes a given ugen or ugenif
device is using.

Define a "flags 1" config directive for ugenif, which is similar to ugen's,
but rather forces the ugenif to match at the *lowest* match priority rather
than the highest. This allows ugenif to claim only otherwise unclaimed
interfaces.

Undo the DPRINTFN part of the last commit. It breaks i386 at least
when KERNHIST_LOG casts the pointer to uintmax_t.

Kernel ALL/i386 compiles again.

Debug printf tidy up, KNF comma,space nits.

Don't call versioned stuff "old". Follow the naming convention for versioning
and name them after the last version of the OS they appeared on.

various debug updates for some usb drivers

- several new *_DEBUG_DEFAULT options that allow usb debug values to
be set to a default that is non-zero:
EHCI_DEBUG_DEFAULT, UGEN_DEBUG_DEFAULT, URTWN_DEBUG_DEFAULT,
UMS_DEBUG_DEFAULT, and USB_DEBUG_DEFAULT
- ugen debug uses fewer usbhist lines for the same info
- ums.c converted from printf() to usbhist

ugen(4): Make sure opened is initialized in ugenopen.

Otherwise the error branch is based on garbage.

Fix USBDEBUG build on ILP32

ugen(4): convert to USBHIST style debugging.

Change the kqueue filterops::f_isfd field to filterops::f_flags, and
define a flag FILTEROP_ISFD that has the meaning of the prior f_isfd.
Field and flag name aligned with OpenBSD.

This does not constitute a functional or ABI change, as the field location
and size, and the value placed in that field, are the same as the previous
code, but we're bumping __NetBSD_Version__ so 3rd-party module source code
can adapt, as needed.

NetBSD 9.99.89

ugen(4): Keep fields null when not allocated; kassert on close.

This avoids silent leaks in DIAGNOSTIC kernels.

ugen(4): Use cv_wait loop for draining reference count on detach.

- Should be no need to use cv_timedwait because all users have now
been given a wakeup (previously writers were not, so we relied on
the timeouts to work out).

- Need to run this in a loop or else a spurious wakeup could cause us
to free data structures before the users have actually drained.

ugen(4): Use cv_broadcast to wake all I/O operations on detach.

Nothing prevents two concurrent reads or two concurrent writes on any
particular ugen endpoint, as far as I can tell, and we need to wake
all of them, so use cv_broadcast rather than cv_signal on detach.

XXX It's not clear to me that cv_signal in the xfer completion
callbacks is correct either: any one consumer might use less than the
full buffer. So I think either we should use cv_broadcast, or
consumers that don't use the whole buffer need to issue cv_signal too
to wake up another consumer even if we want to avoid a thundering
herd.

ugen(4): Issue explicit wakeup on detach for OUT endpoints too.

Writers can be blocked in cv_timedwait_sig too.

While here, fix comment: aborting the pipes does not cause all
waiters to wake, because the xfer completion callbacks sometimes skip
the notification. We should maybe change that, but this is a simpler
fix to ensure everyone waiting on I/O is woken to notice sc_dying.

ugen(4): Ensure we close pipes on detach.

Calling vdevgone has the effect of VOP_REVOKE -> spec_node_revoke ->
VOP_CLOSE -> spec_close -> cdev_close -> ugenclose, but:

(a) the flags passed to VOP_CLOSE don't have FREAD or FWRITE, and
(b) ugenclose has no effect when sc_dying is set anyway.

So create another path into the close logic, ugen_do_close. We have
to do this in detach _after_ the references have drained because we
may free buffers that other users are still using while the reference
count is nonzero.

ugen(4): Refuse non-forced detach with EBUSY if endpoints are open.

Sprinkle some comments.

ugen(4): Prevent ugenopen while ugen_set_config is in progress.

(except on the control endpoint)

Although we hold the kernel lock (which we should eventually change),
we may sleep in usbd_set_config_no at which point ugenopen might
happen and start making use of endpoint state which we'll stomp all
over once usbd_set_config_no returns. Setting sc_is_open[endpt]
while we wait prevents this.

ugen(4): Fix race of ugenopen against itself.

Even though we have the kernel lock held, a sleep during kmem_alloc
or usbd_open_pipe could allow another ugenopen to run concurrently
before we have marked the endpoint opened.

To avoid this, mark the endpoint open immediately (while we still
have the kernel lock held and before any sleeps, so there is no
TOCTOU error here), and then revert on unwind in the event of
failure.

ugen(4): Sprinkle KERNEL_LOCKED_P assertions around sc_is_open.

Use sel{record,remove}_knote().

branches: 1.157.2;
Fix ugen detach after partial attach.

While here, register null pmf handler even for partially attached
devices so they don't needlessly interfere with suspend.

Reported-by: syzbot+5a091d2e62da20b77259@syzkaller.appspotmail.com

Fix sloppy mistakes in previous.

1. Give the offset of the rbnode, not some other random members to
overwrite with garbage.

2. Don't try to unlock a mutex at NULL.

3. Make sure all paths out after ugenif_acquire go via
ugenif_release.

Share unit numbering for ugen and ugenif.

This way putting ugenif in kernel config actually works to wire it to
the /dev/ugenN.MM device nodes in userland.

Not a fully fleshed out solution to the ugen problem -- there's no
way for a userland driver to kick out a kernel driver and take over,
but this will let us, e.g., use uhidev(4) for Yubikey OTP/U2F/FIDO2
but ugen(4), with pcscd(8), for Yubikey CCID.

Fix various MP-safety issues while here (still not MPSAFE, but more
progress).

Convert DIAGNOSTIC prints to KASSERTs.

Hold the lock over access to the data structures it covers.

Still not MPSAFE, but progress.

Remove UGEN_ASLP microoptimization.

cv_signal already has this microoptimization.

While here, make the lock cover the relevant things we're issuing
cv_signal about -- progress toward real MP-safety.

KNG

revert the 0x% -> %# change for fixed width formats pointed out by uwe.

PR/55068: sc.dying: Fix printf formats:
- no %s/%p for kernel log
- 0x% -> %#
- always %j for kernel log

reading usbdi.c it looks like usbd_get_config_descriptor() can actually
return NULL, so check for this.
I got NULL pointer dereference here with a device showing:
[ 303.732632] ugen0: autoconfiguration error: setting configuration index 0 failed

localify

branches: 1.146.2;
remove explicit 'extern struct cfdriver <my>_cd;' and use ioconf.h

Rename the MODULE_*_HOOK() macros to MODULE_HOOK_*() as briefly
discussed on irc.

NFCI intended.

Ride the earlier kernel bump - it;s getting crowded.

Remove (mostly useless) usb_detach_{broadcast,wait} and replace with
cv_{broadcast,timedwait}

Really should loop on conditon.

Normalize all the compat hooks' names to the form

<subsystem>_<function>_<version>_hook

NFCI

XXX Note that although this introduces a change in the kernel-to-
XXX module interface, we are NOT bumping the kernel version number.
XXX We will bump the version number once the interface stabilizes.

Merge the [pgoyette-compat] branch

Enfore USB timeout on ugen(4) write operations

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: 1.139.2; 1.139.4;
Fix last:

Since config(1) could not distinguish between device and
interface attachments, it was generating only the latter.
Thus devices without their own driver wouldn't match the
ugen driver anymore.

Fix this by using a different device name for interface attachments.

Attach uftdi to each interface found in the device separately.
This allows for other drivers (e.g. ugen) to attach to some of
the other interfaces.

Allow ugen to attach only to some of the interfaces found in a device.

PR kern/52931 Kernel panics with Atheros usb wireless interface

Audit the flags to usbd_create_xfer so that USBD_FORCE_SHORT_XFER is
supplied wherever such a transfer is setup. We can drop
USBD_SHORT_XFER_OK as it has not bearing on number of TDs

Use C99 initializer for filterops

Mostly done with spatch with touchups for indentation

@@
expression a;
identifier b,c,d;
identifier p;
@@
const struct filterops p =
- { a, b, c, d
+ {
+ .f_isfd = a,
+ .f_attach = b,
+ .f_detach = c,
+ .f_event = d,
};

remove redundant checks against sc and sc->sc_dying.
check sc_dying in more places.

branches: 1.134.10;
KNF. Remove extra spaces. No functional change.

Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

KNF

Provide a ugen_clear_endpoints declaration

Only clear the endpoint information in ugen_set_interface only if setting
the new altno suceeds.

Avoids the null de-ref in PR/50597 and PR/50810

Remove always true conditional

One more s/0/NULL/

s/0/NULL/

branches: 1.126.2;
Include opt_usb.h to make "options UGEN_DEBUG" work.

Don't nest structure definitions.

branches: 1.124.2; 1.124.4;
Add d_discard to all struct cdevsw instances I could find.

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

branches: 1.123.2;
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.

branches: 1.122.2;
- need opt_usb.h if depending on USB_DEBUG
- remove trailing whitespace
- add missing KERNEL_RCSID

Fix some error branches in ugen.

There remains some cruft that should perhaps be better organized, but
at least this should reduce some memory leaks in screw cases, and at
least this does fix panics when plugging in and unplugging a USB
device with a botched configuration (a beaglebone with a hosed sd
card).

branches: 1.120.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

Remove duplicate break. PR 46368 from Henning Petersen.

pull down from usbmp branch:

- rename usb_detach_{wake,waitup}() to usb_detach_{wake,waitup}old()
- use some c99 struct .initialisers

remove any remnants of freebsd/openbsd code.

Revert previous due to active usbmp branch(es).

Adjust-away inconsistent and trailing whitespace.

branches: 1.114.2;
Don't double clfree() when closing an interrupt endpoint.
From Geoff C. Wing in PR#37934.

Prevent poll and kqueue on the control endpoint.
Addresses PR#33352.

(The control endpoint doesn't support pipes, so this would be of dubious
usefulness even if it didn't expose bugs.)

Implement timeouts when blocking the calling process with tsleep.

Fixes PR kern/33452

branches: 1.111.8;
Stop using the compatibility macros USB_ATTACH(), USB_DETACH(),
USB_MATCH(), et cetera. These files produce the same assembly
(according to objdump -d) before and after the change

Tune ugen(4) isochronous parameters to better support high speed pipes.

Fixes PR#43159.

Justification:

UGEN_NISORFRMS, previously 4, is now 8 because that's the number of
microframes per frame on a high speed bus. This allows data to be
transfered every microframe if necessary.

UGEN_NISOREQS, previously 6, is now 4. This is harder to justify,
but I didn't think the total number of buffer frames needed to be
much more than it was.

UGEN_NISOFRAMES, previously 500, is now (UGEN_NISORFRMS * UGEN_NISOREQS)
(32 with values as above). This is all the more we've actually used.

branches: 1.109.2;
Replace USB_GET_SC() instances with USB_GET_SC_OPEN(). The latter
is exactly like the former, except it returns with ENXIO if sc is
NULL instead of continuing on to use it. Most of the time this is
handled by the fd layer and you get EBADF, but sometimes it's
possible to race into the device method for whatever reason and
you get a kernel panic. I have no idea what the "whatever reason"
might be.

branches: 1.108.2;
defflag UGEN_BULK_RA_WB is no more.
Hi pooka

unifdef -D UGEN_BULK_RA_WB

Default behaviour unchanged, the feature must still be explicitly
enabled for a ugen fd.

Simplify device-activation hooks.

Add a simple kvm integer toggle to control ugen priority and override
the autoconf flags value. This helps to change the priority at
runtime. (XXX: is there a better way to do this?)

ugen_get_cdesc() can return NULL. When this happens, return the
karmic EINVAL error instead of crashing.

fix up USB drivers printing of autoconf information

1. expand the USB_ATTACH_SETUP macro (requested by jmcneill)

2. reorder the attach function so that the first thing it does is print
newlines.

3. after this, we can call usbd_devinfo_alloc(), which polls the device
allowing a context switch, and aprint_normal() the device information.

this avoids problems where autoconf messages are getting mixed up.

Putting a device into the unconfigured state by an ioctl seems legitimate,
so check for a non-NULL configuration descriptor before dereferencing.
Should fix a crash reported by Nicolas Joly per PR kern/41048.
(It still doesn't look good that the ioctl which unconfigures the device
returns EIO -- either it is legitimate or it isn't -- but since this
is a pullup candidate I don't dare to change user visible behaviour.)

branches: 1.101.2;
Change major()/minor() to return 32-bit types again, called
devmajor_t/devminor_t, as proposed on tech-kern.
This avoids 64-bit arithmetics and 64-bit printf formats in parts
of the kernel where it is not really useful, and helps clarity.

make this compile

branches: 1.99.6; 1.99.8;
Split device_t and softc for all USB device drivers, and related cosmetic
changes.

Matthias Drochner kindly reviewed this patch, and tested ums, ubt, uaudio
and ral. I tested umass myself.

branches: 1.98.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.97.2; 1.97.4;
Welcome to 4.99.55:

- Add a lot of missing selinit() and seldestroy() calls.

- Merge selwakeup() and selnotify() calls into a single selnotify().

- Add an additional 'events' argument to selnotify() call. It will
indicate which event (POLL_IN, POLL_OUT, etc) happen. If unknown,
zero may be used.

Note: please pass appropriate value of 'events' where possible.
Proposed on: <tech-kern>

branches: 1.96.2; 1.96.6;
Add missing call to pmf_deregister() in the USB_DETACH routine.

branches: 1.95.2;
Merge jmcneill-pm branch.

branches: 1.94.2;
Do not "return 1" from kqfilter for errors. That value is passed
directly to the userland caller and results in a mysterious EPERM.
Instead, return EINVAL or something else sensible depending on the
case.

branches: 1.93.2;
aprintify

branches: 1.92.14; 1.92.16; 1.92.22;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.91.4;
The manpage promises not to touch the device on USB_GET_DEVICEINFO.
Make it so.

Fix debug format strings (found on amd64).

branches: 1.89.2;
Restore compatibility of USB_DEVICEINFO ioctl and reads from /dev/usb with
NetBSD 3.x. Patch from Stephan Thesing provided in
http://mail-index.netbsd.org/current-users/2006/03/21/0002.html, with some
modifications by me.
See also
http://mail-index.netbsd.org/current-users/2006/08/29/0017.html

The code is conditionally compiled depending on COMPAT_30.

Also fix a leak of struct usb_event in usbread() introduced while converting
on-stack variables to dynamic allocation.

Reviewed by martin@.

branches: 1.88.2;
__unused removal on arguments; approved by core.

sigh, another missing __unused.

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

branches: 1.85.2; 1.85.4;
add missing initializer

Add UGEN_BULK_RA_WB, which allows users of ugen(4) to request read
ahead and write behind, improving performance for the Universal
Software Radio Peripheral (USRP) used with GNU Radio.

Enable UGEN_BULK_RA_WB in GENERIC and GENERIC_LAPTOP; behavior is
unchanged unless the new ioctl is called.

This code was written by Joanne Mikkelson under funding from DARPA's
ACERT program.

ok'd by christos@, tested by Berndt Josef Wulf

stack police: Don't allocate large buffers on the stack for I/O. Put the
buffer in the softc instead.

branches: 1.82.2;
Coverity CID 1354: Add a KASSERT to convince coverity that there is no NULL
pointer dereference.

Coverity CID 1355: Add a KASSERT to convince coverity that this is not NULL.

Use device_unit().

branches: 1.79.2; 1.79.4; 1.79.6;
merge yamt-uio_vmspace branch.

- use vmspace rather than proc or lwp where appropriate.
the latter is more natural to specify an address space.
(and less likely to be abused for random purposes.)
- fix a swdmover race.

branches: 1.78.2; 1.78.4; 1.78.6;
merge ktrace-lwp.

Add a new match level, UMATCH_HIGHEST.

Normally a ugen device only attaches if no other driver wants the device.
Add the ability to force ugen to attach with very high priority if "flags 1"
is specified. This can be used with the vendor and product locators to
force ugen to be used for certain devices.
Similarly, uhid only attaches if no other HID driver (ums or ukbd) wants it.
Again, "flags 1" will force uhid to attach anyway.

branches: 1.75.2; 1.75.8;
PR-30566: Poll must not return <sys/errno.h> values.
Start with those places I can easily test.

- const poisoning
- eliminate variable shadowing

Allow multiple opens of the control endpoint. This is unproblematic since
the control endpoint is in some sense always open to the device.

From wulf@ping.net.au.

Don't keep the devinfo string on the stack, instead use malloc/free.
This should cure some rare stack overflows.

Copyright maintenance.

nuke trailing whitespace

branches: 1.69.4; 1.69.6;
Allow interrupt output devices. From FreeBSD.

Yes, some devices return incorrect lengths in their string descriptors. Rather
than losing, do what Windows does: just request the maximum size, and allow a
shorter response. Obsoletes the need for UQ_NO_STRINGS, and therefore these
"quirks" are removed.

use bounded string ops (snprintf, strl*)

branches: 1.66.2;
Cast a printf() arg so that USB_DEBUG compiles on multiple platforms.

branches: 1.65.2;
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

si_ -> sel_

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 breaks after returns, unreachable returns and returns after
returns(!).

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.

Get rid of trailing white space.

branches: 1.58.8;
Prefix structure members to protect them against clashes with eg. c++ keywords.
Suggested by Alfred Perlstein, from FreeBSD, ok'd by augustss

Give usbd_do_request_flags() an extra argument for the timeout.

More whitespace fixes from FreeBSD.

Whitespace fixes (from FreeBSD).

Make a typedef for struct proc to make portingeasier.

Don't allow setting the configuration when an endpoint is open. From
FreeBSD.

Only clear stall if endpoint is stalled. From FreeBSD.

More test for NULL pipes.

add RCSIDs

Add commented out match level feature.

branches: 1.48.2;
make ugen use updated frlengths.

Spell 'occurred' with two 'r's.

correct debug messages.

branches: 1.45.2; 1.45.4; 1.45.6;
Don't try to access a device that is being disconnected when generating
the detach event. Fixes (I hope) PR 11713 from itohy@netbsd.org (ITOH Yasufumi).

Improve a debug message.

Tell usbd_set_config_no() to be verbose. It's nice to know whyit fails.

ANSIfy last patch.

Add isoc support. From FreeBSD via Berndt Josef Wulf <wulf@ping.net.au>

Bring the coding style into the 80s, i.e., get rid of __P and use
ANSI prototypes and declarations.

Add a comment.

branches: 1.38.2;
Change my email address.

Change (almost) all static to Static. The symbol `Static' can then be defined
to `' or `static' depending on if you want to debug or not.

Generate better error codes on bulk write.

Distinguish between device and interface classes.
(I finally found a document that said that they were different.)

Use NULL instead of 0.

Generate usb events on attach and detach.

Add an argument to usbd_open_pipe_intr() to specify the polling interval
for an interrupt pipe in case we don't what what the descriptor suggests.

Change the way the initial configuration value is picked.
Idea from Matthias Drochner <M.Drochner@fz-juelich.de>

Cosmetic changes and some small improvements. From FreeBSD and Nick Hibma.

A few more purely stylistic changes that I missed in the last round.

A number of stylistic changes to increase readability (many suggested
by Nick Hibma):
use NULL not 0
declare all local definitions static
rename s/usbd_request/usbd_xfer/ s/reqh/xfer/
rename s/r/err/
use implicit test for no err
KNF

Add a few more tests for safety.

Make sure read() and write() fails on the control pipe.

branches: 1.25.2; 1.25.4;
Merge in a large batch of changes from Nick Hibma <hibma@skylink.it> so
the USB stack compiles on FreeBSD again.

Add an event mechanism so that a userland process can watch devices come
and go.

branches: 1.23.2;
Change the internal API to allow DMA buffers to be pre-allocated by
the device driver instead of happening automagically in the HC driver.
This affects both the HC-USBD interface as well as the USBD-device
interface.
This change will allow DMA buffers to be reused e.g. in isochronous
traffic.

Add isochronous support to the UHCI driver (not for OHCI yet).

Change the way the `struct device' base part of all driver softc are
declared and accessed to make it more portable. Idea from Nick Hibma, FreeBSD.
No functional changes.

Change the way the direction is extracted from the endpoint descriptor.
No functional changes to the drivers. From Nick Hibma, FreeBSD.

Change some 'struct device' to 'bdevice'. From FreeBSD.

Change a type name.

Make sure to mark the device as dying already in the (de)activate routine.
This avoids access to it before the detach routine has blown it away.

Move more of the transfer completion processing to HC independent code.
Fix some problems with transfer abort & timeout.

Use the right type for the size argument when calling usbd_bulk_transfer().

Make some small changes to make it compile on OpenBSD.

Some changes from FreeBSD (no functional differences).

Change it so that a pipe can be open RW instead of just R or W.
This makes close() work properly, but it is still not ideal. Perhaps
there should be different device nodes for input and output on to
and endpoint with the same number?
Pay attention to the SHORT_XFER_OK ioctl().

Totally redo the way device detach is done. It now uses a kernel event
thread and the config detach method.
Squish a number of space leaks on detach.

branches: 1.11.4;
Various little fixes from the FreeBSD version.

Fix some pastos.

Add an ugly workaround for a bug (feature) in the NetBSD open()/close()
protocol.

Fix bug in reading from interrupt pipe.
Make read and write on bulk pipes interruptible.

Don't set configuration unnecessarily, some (broken) devices seem to break
if you do.

Make it possible to specify the request flags when issuing a raw USB request.

Merge changes to make the USB stack work with FreeBSD. The original
diffs from Nick Hibma <n_hibma@freebsd.org>, but with substantial
changes from me.
XXX Not tested on FreeBSD yet.

Move initialization of sce around.

Take care of some lines > 80 chars.

Improvement to the ugen driver.
Better error checking.
Some code rearrengment.

Add a generic USB driver. It allows easy access to descriptors, the
control pipe via ioctl() and read() and write() access to bulk and
interrupt pipes.

Define a "flags 1" config directive for ugenif, which is similar to ugen's,
but rather forces the ugenif to match at the *lowest* match priority rather
than the highest. This allows ugenif to claim only otherwise unclaimed
interfaces.

Undo the DPRINTFN part of the last commit. It breaks i386 at least
when KERNHIST_LOG casts the pointer to uintmax_t.

Kernel ALL/i386 compiles again.

Debug printf tidy up, KNF comma,space nits.

Don't call versioned stuff "old". Follow the naming convention for versioning
and name them after the last version of the OS they appeared on.

various debug updates for some usb drivers

- several new *_DEBUG_DEFAULT options that allow usb debug values to
be set to a default that is non-zero:
EHCI_DEBUG_DEFAULT, UGEN_DEBUG_DEFAULT, URTWN_DEBUG_DEFAULT,
UMS_DEBUG_DEFAULT, and USB_DEBUG_DEFAULT
- ugen debug uses fewer usbhist lines for the same info
- ums.c converted from printf() to usbhist

ugen(4): Make sure opened is initialized in ugenopen.

Otherwise the error branch is based on garbage.

Fix USBDEBUG build on ILP32

ugen(4): convert to USBHIST style debugging.

Change the kqueue filterops::f_isfd field to filterops::f_flags, and
define a flag FILTEROP_ISFD that has the meaning of the prior f_isfd.
Field and flag name aligned with OpenBSD.

This does not constitute a functional or ABI change, as the field location
and size, and the value placed in that field, are the same as the previous
code, but we're bumping __NetBSD_Version__ so 3rd-party module source code
can adapt, as needed.

NetBSD 9.99.89

ugen(4): Keep fields null when not allocated; kassert on close.

This avoids silent leaks in DIAGNOSTIC kernels.

ugen(4): Use cv_wait loop for draining reference count on detach.

- Should be no need to use cv_timedwait because all users have now
been given a wakeup (previously writers were not, so we relied on
the timeouts to work out).

- Need to run this in a loop or else a spurious wakeup could cause us
to free data structures before the users have actually drained.

ugen(4): Use cv_broadcast to wake all I/O operations on detach.

Nothing prevents two concurrent reads or two concurrent writes on any
particular ugen endpoint, as far as I can tell, and we need to wake
all of them, so use cv_broadcast rather than cv_signal on detach.

XXX It's not clear to me that cv_signal in the xfer completion
callbacks is correct either: any one consumer might use less than the
full buffer. So I think either we should use cv_broadcast, or
consumers that don't use the whole buffer need to issue cv_signal too
to wake up another consumer even if we want to avoid a thundering
herd.

ugen(4): Issue explicit wakeup on detach for OUT endpoints too.

Writers can be blocked in cv_timedwait_sig too.

While here, fix comment: aborting the pipes does not cause all
waiters to wake, because the xfer completion callbacks sometimes skip
the notification. We should maybe change that, but this is a simpler
fix to ensure everyone waiting on I/O is woken to notice sc_dying.

ugen(4): Ensure we close pipes on detach.

Calling vdevgone has the effect of VOP_REVOKE -> spec_node_revoke ->
VOP_CLOSE -> spec_close -> cdev_close -> ugenclose, but:

(a) the flags passed to VOP_CLOSE don't have FREAD or FWRITE, and
(b) ugenclose has no effect when sc_dying is set anyway.

So create another path into the close logic, ugen_do_close. We have
to do this in detach _after_ the references have drained because we
may free buffers that other users are still using while the reference
count is nonzero.

ugen(4): Refuse non-forced detach with EBUSY if endpoints are open.

Sprinkle some comments.

ugen(4): Prevent ugenopen while ugen_set_config is in progress.

(except on the control endpoint)

Although we hold the kernel lock (which we should eventually change),
we may sleep in usbd_set_config_no at which point ugenopen might
happen and start making use of endpoint state which we'll stomp all
over once usbd_set_config_no returns. Setting sc_is_open[endpt]
while we wait prevents this.

ugen(4): Fix race of ugenopen against itself.

Even though we have the kernel lock held, a sleep during kmem_alloc
or usbd_open_pipe could allow another ugenopen to run concurrently
before we have marked the endpoint opened.

To avoid this, mark the endpoint open immediately (while we still
have the kernel lock held and before any sleeps, so there is no
TOCTOU error here), and then revert on unwind in the event of
failure.

ugen(4): Sprinkle KERNEL_LOCKED_P assertions around sc_is_open.

Use sel{record,remove}_knote().

branches: 1.157.2;
Fix ugen detach after partial attach.

While here, register null pmf handler even for partially attached
devices so they don't needlessly interfere with suspend.

Reported-by: syzbot+5a091d2e62da20b77259@syzkaller.appspotmail.com

Fix sloppy mistakes in previous.

1. Give the offset of the rbnode, not some other random members to
overwrite with garbage.

2. Don't try to unlock a mutex at NULL.

3. Make sure all paths out after ugenif_acquire go via
ugenif_release.

Share unit numbering for ugen and ugenif.

This way putting ugenif in kernel config actually works to wire it to
the /dev/ugenN.MM device nodes in userland.

Not a fully fleshed out solution to the ugen problem -- there's no
way for a userland driver to kick out a kernel driver and take over,
but this will let us, e.g., use uhidev(4) for Yubikey OTP/U2F/FIDO2
but ugen(4), with pcscd(8), for Yubikey CCID.

Fix various MP-safety issues while here (still not MPSAFE, but more
progress).

Convert DIAGNOSTIC prints to KASSERTs.

Hold the lock over access to the data structures it covers.

Still not MPSAFE, but progress.

Remove UGEN_ASLP microoptimization.

cv_signal already has this microoptimization.

While here, make the lock cover the relevant things we're issuing
cv_signal about -- progress toward real MP-safety.

KNG

revert the 0x% -> %# change for fixed width formats pointed out by uwe.

PR/55068: sc.dying: Fix printf formats:
- no %s/%p for kernel log
- 0x% -> %#
- always %j for kernel log

reading usbdi.c it looks like usbd_get_config_descriptor() can actually
return NULL, so check for this.
I got NULL pointer dereference here with a device showing:
[ 303.732632] ugen0: autoconfiguration error: setting configuration index 0 failed

localify

branches: 1.146.2;
remove explicit 'extern struct cfdriver <my>_cd;' and use ioconf.h

Rename the MODULE_*_HOOK() macros to MODULE_HOOK_*() as briefly
discussed on irc.

NFCI intended.

Ride the earlier kernel bump - it;s getting crowded.

Remove (mostly useless) usb_detach_{broadcast,wait} and replace with
cv_{broadcast,timedwait}

Really should loop on conditon.

Normalize all the compat hooks' names to the form

<subsystem>_<function>_<version>_hook

NFCI

XXX Note that although this introduces a change in the kernel-to-
XXX module interface, we are NOT bumping the kernel version number.
XXX We will bump the version number once the interface stabilizes.

Merge the [pgoyette-compat] branch

Enfore USB timeout on ugen(4) write operations

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: 1.139.2; 1.139.4;
Fix last:

Since config(1) could not distinguish between device and
interface attachments, it was generating only the latter.
Thus devices without their own driver wouldn't match the
ugen driver anymore.

Fix this by using a different device name for interface attachments.

Attach uftdi to each interface found in the device separately.
This allows for other drivers (e.g. ugen) to attach to some of
the other interfaces.

Allow ugen to attach only to some of the interfaces found in a device.

PR kern/52931 Kernel panics with Atheros usb wireless interface

Audit the flags to usbd_create_xfer so that USBD_FORCE_SHORT_XFER is
supplied wherever such a transfer is setup. We can drop
USBD_SHORT_XFER_OK as it has not bearing on number of TDs

Use C99 initializer for filterops

Mostly done with spatch with touchups for indentation

@@
expression a;
identifier b,c,d;
identifier p;
@@
const struct filterops p =
- { a, b, c, d
+ {
+ .f_isfd = a,
+ .f_attach = b,
+ .f_detach = c,
+ .f_event = d,
};

remove redundant checks against sc and sc->sc_dying.
check sc_dying in more places.

branches: 1.134.10;
KNF. Remove extra spaces. No functional change.

Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

KNF

Provide a ugen_clear_endpoints declaration

Only clear the endpoint information in ugen_set_interface only if setting
the new altno suceeds.

Avoids the null de-ref in PR/50597 and PR/50810

Remove always true conditional

One more s/0/NULL/

s/0/NULL/

branches: 1.126.2;
Include opt_usb.h to make "options UGEN_DEBUG" work.

Don't nest structure definitions.

branches: 1.124.2; 1.124.4;
Add d_discard to all struct cdevsw instances I could find.

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

branches: 1.123.2;
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.

branches: 1.122.2;
- need opt_usb.h if depending on USB_DEBUG
- remove trailing whitespace
- add missing KERNEL_RCSID

Fix some error branches in ugen.

There remains some cruft that should perhaps be better organized, but
at least this should reduce some memory leaks in screw cases, and at
least this does fix panics when plugging in and unplugging a USB
device with a botched configuration (a beaglebone with a hosed sd
card).

branches: 1.120.2;
merge the jmcneill-usbmp branch. many thanks to jared for the
initial work, and every one else who has tested things for me.
this is largely my fault at this point :-)

the main changes are something like:

- usbd_bus_methods{} gains a get_lock() to enable the
host controller to provide a lock for the USB code.
if the lock isn't provided, old-style protection is
(partially) applied.

- ehci/ohci/uhci have been converted to the new
interfaces, including mutex/cv/etc conversion.

- usbdivar.h contains a discussion about locking and
what locks are held for which method calls. more
to come for usbdi(9) here.

- audio drivers (uaudio, umidi, auvitek) have been
properly SMPified now that USB is ready.

- scsi drivers have been modified to take the kernel
lock explicitly before calling into scsi code.

- usb pipes are associated with a lock, that is the
same as the controller lock. (this could be split
up further in the future.)

- several usbfoo_locked() or usbfoo_unlocked()
functions have been added to the usbdi(9) to
enable functionality with or without the USB
lock (per controller) already being held.

the TODO.usbmp file has specific details on what is left to
do, including what device-specific changes should be done now
that the whole framework is ready.

Remove duplicate break. PR 46368 from Henning Petersen.

pull down from usbmp branch:

- rename usb_detach_{wake,waitup}() to usb_detach_{wake,waitup}old()
- use some c99 struct .initialisers

remove any remnants of freebsd/openbsd code.

Revert previous due to active usbmp branch(es).

Adjust-away inconsistent and trailing whitespace.

branches: 1.114.2;
Don't double clfree() when closing an interrupt endpoint.
From Geoff C. Wing in PR#37934.

Prevent poll and kqueue on the control endpoint.
Addresses PR#33352.

(The control endpoint doesn't support pipes, so this would be of dubious
usefulness even if it didn't expose bugs.)

Implement timeouts when blocking the calling process with tsleep.

Fixes PR kern/33452

branches: 1.111.8;
Stop using the compatibility macros USB_ATTACH(), USB_DETACH(),
USB_MATCH(), et cetera. These files produce the same assembly
(according to objdump -d) before and after the change

Tune ugen(4) isochronous parameters to better support high speed pipes.

Fixes PR#43159.

Justification:

UGEN_NISORFRMS, previously 4, is now 8 because that's the number of
microframes per frame on a high speed bus. This allows data to be
transfered every microframe if necessary.

UGEN_NISOREQS, previously 6, is now 4. This is harder to justify,
but I didn't think the total number of buffer frames needed to be
much more than it was.

UGEN_NISOFRAMES, previously 500, is now (UGEN_NISORFRMS * UGEN_NISOREQS)
(32 with values as above). This is all the more we've actually used.

branches: 1.109.2;
Replace USB_GET_SC() instances with USB_GET_SC_OPEN(). The latter
is exactly like the former, except it returns with ENXIO if sc is
NULL instead of continuing on to use it. Most of the time this is
handled by the fd layer and you get EBADF, but sometimes it's
possible to race into the device method for whatever reason and
you get a kernel panic. I have no idea what the "whatever reason"
might be.

branches: 1.108.2;
defflag UGEN_BULK_RA_WB is no more.
Hi pooka

unifdef -D UGEN_BULK_RA_WB

Default behaviour unchanged, the feature must still be explicitly
enabled for a ugen fd.

Simplify device-activation hooks.

Add a simple kvm integer toggle to control ugen priority and override
the autoconf flags value. This helps to change the priority at
runtime. (XXX: is there a better way to do this?)

ugen_get_cdesc() can return NULL. When this happens, return the
karmic EINVAL error instead of crashing.

fix up USB drivers printing of autoconf information

1. expand the USB_ATTACH_SETUP macro (requested by jmcneill)

2. reorder the attach function so that the first thing it does is print
newlines.

3. after this, we can call usbd_devinfo_alloc(), which polls the device
allowing a context switch, and aprint_normal() the device information.

this avoids problems where autoconf messages are getting mixed up.

Putting a device into the unconfigured state by an ioctl seems legitimate,
so check for a non-NULL configuration descriptor before dereferencing.
Should fix a crash reported by Nicolas Joly per PR kern/41048.
(It still doesn't look good that the ioctl which unconfigures the device
returns EIO -- either it is legitimate or it isn't -- but since this
is a pullup candidate I don't dare to change user visible behaviour.)

branches: 1.101.2;
Change major()/minor() to return 32-bit types again, called
devmajor_t/devminor_t, as proposed on tech-kern.
This avoids 64-bit arithmetics and 64-bit printf formats in parts
of the kernel where it is not really useful, and helps clarity.

make this compile

branches: 1.99.6; 1.99.8;
Split device_t and softc for all USB device drivers, and related cosmetic
changes.

Matthias Drochner kindly reviewed this patch, and tested ums, ubt, uaudio
and ral. I tested umass myself.

branches: 1.98.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.97.2; 1.97.4;
Welcome to 4.99.55:

- Add a lot of missing selinit() and seldestroy() calls.

- Merge selwakeup() and selnotify() calls into a single selnotify().

- Add an additional 'events' argument to selnotify() call. It will
indicate which event (POLL_IN, POLL_OUT, etc) happen. If unknown,
zero may be used.

Note: please pass appropriate value of 'events' where possible.
Proposed on: <tech-kern>

branches: 1.96.2; 1.96.6;
Add missing call to pmf_deregister() in the USB_DETACH routine.

branches: 1.95.2;
Merge jmcneill-pm branch.

branches: 1.94.2;
Do not "return 1" from kqfilter for errors. That value is passed
directly to the userland caller and results in a mysterious EPERM.
Instead, return EINVAL or something else sensible depending on the
case.

branches: 1.93.2;
aprintify

branches: 1.92.14; 1.92.16; 1.92.22;
Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.91.4;
The manpage promises not to touch the device on USB_GET_DEVICEINFO.
Make it so.

Fix debug format strings (found on amd64).

branches: 1.89.2;
Restore compatibility of USB_DEVICEINFO ioctl and reads from /dev/usb with
NetBSD 3.x. Patch from Stephan Thesing provided in
http://mail-index.netbsd.org/current-users/2006/03/21/0002.html, with some
modifications by me.
See also
http://mail-index.netbsd.org/current-users/2006/08/29/0017.html

The code is conditionally compiled depending on COMPAT_30.

Also fix a leak of struct usb_event in usbread() introduced while converting
on-stack variables to dynamic allocation.

Reviewed by martin@.

branches: 1.88.2;
__unused removal on arguments; approved by core.

sigh, another missing __unused.

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

branches: 1.85.2; 1.85.4;
add missing initializer

Add UGEN_BULK_RA_WB, which allows users of ugen(4) to request read
ahead and write behind, improving performance for the Universal
Software Radio Peripheral (USRP) used with GNU Radio.

Enable UGEN_BULK_RA_WB in GENERIC and GENERIC_LAPTOP; behavior is
unchanged unless the new ioctl is called.

This code was written by Joanne Mikkelson under funding from DARPA's
ACERT program.

ok'd by christos@, tested by Berndt Josef Wulf

stack police: Don't allocate large buffers on the stack for I/O. Put the
buffer in the softc instead.

branches: 1.82.2;
Coverity CID 1354: Add a KASSERT to convince coverity that there is no NULL
pointer dereference.

Coverity CID 1355: Add a KASSERT to convince coverity that this is not NULL.

Use device_unit().

branches: 1.79.2; 1.79.4; 1.79.6;
merge yamt-uio_vmspace branch.

- use vmspace rather than proc or lwp where appropriate.
the latter is more natural to specify an address space.
(and less likely to be abused for random purposes.)
- fix a swdmover race.

branches: 1.78.2; 1.78.4; 1.78.6;
merge ktrace-lwp.

Add a new match level, UMATCH_HIGHEST.

Normally a ugen device only attaches if no other driver wants the device.
Add the ability to force ugen to attach with very high priority if "flags 1"
is specified. This can be used with the vendor and product locators to
force ugen to be used for certain devices.
Similarly, uhid only attaches if no other HID driver (ums or ukbd) wants it.
Again, "flags 1" will force uhid to attach anyway.

branches: 1.75.2; 1.75.8;
PR-30566: Poll must not return <sys/errno.h> values.
Start with those places I can easily test.

- const poisoning
- eliminate variable shadowing

Allow multiple opens of the control endpoint. This is unproblematic since
the control endpoint is in some sense always open to the device.

From wulf@ping.net.au.

Don't keep the devinfo string on the stack, instead use malloc/free.
This should cure some rare stack overflows.

Copyright maintenance.

nuke trailing whitespace

branches: 1.69.4; 1.69.6;
Allow interrupt output devices. From FreeBSD.

Yes, some devices return incorrect lengths in their string descriptors. Rather
than losing, do what Windows does: just request the maximum size, and allow a
shorter response. Obsoletes the need for UQ_NO_STRINGS, and therefore these
"quirks" are removed.

use bounded string ops (snprintf, strl*)

branches: 1.66.2;
Cast a printf() arg so that USB_DEBUG compiles on multiple platforms.

branches: 1.65.2;
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

si_ -> sel_

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 breaks after returns, unreachable returns and returns after
returns(!).

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.

Get rid of trailing white space.

branches: 1.58.8;
Prefix structure members to protect them against clashes with eg. c++ keywords.
Suggested by Alfred Perlstein, from FreeBSD, ok'd by augustss

Give usbd_do_request_flags() an extra argument for the timeout.

More whitespace fixes from FreeBSD.

Whitespace fixes (from FreeBSD).

Make a typedef for struct proc to make portingeasier.

Don't allow setting the configuration when an endpoint is open. From
FreeBSD.

Only clear stall if endpoint is stalled. From FreeBSD.

More test for NULL pipes.

add RCSIDs

Add commented out match level feature.

branches: 1.48.2;
make ugen use updated frlengths.

Spell 'occurred' with two 'r's.

correct debug messages.

branches: 1.45.2; 1.45.4; 1.45.6;
Don't try to access a device that is being disconnected when generating
the detach event. Fixes (I hope) PR 11713 from itohy@netbsd.org (ITOH Yasufumi).

Improve a debug message.

Tell usbd_set_config_no() to be verbose. It's nice to know whyit fails.

ANSIfy last patch.

Add isoc support. From FreeBSD via Berndt Josef Wulf <wulf@ping.net.au>

Bring the coding style into the 80s, i.e., get rid of __P and use
ANSI prototypes and declarations.

Add a comment.

branches: 1.38.2;
Change my email address.

Change (almost) all static to Static. The symbol `Static' can then be defined
to `' or `static' depending on if you want to debug or not.

Generate better error codes on bulk write.

Distinguish between device and interface classes.
(I finally found a document that said that they were different.)

Use NULL instead of 0.

Generate usb events on attach and detach.

Add an argument to usbd_open_pipe_intr() to specify the polling interval
for an interrupt pipe in case we don't what what the descriptor suggests.

Change the way the initial configuration value is picked.
Idea from Matthias Drochner <M.Drochner@fz-juelich.de>

Cosmetic changes and some small improvements. From FreeBSD and Nick Hibma.

A few more purely stylistic changes that I missed in the last round.

A number of stylistic changes to increase readability (many suggested
by Nick Hibma):
use NULL not 0
declare all local definitions static
rename s/usbd_request/usbd_xfer/ s/reqh/xfer/
rename s/r/err/
use implicit test for no err
KNF

Add a few more tests for safety.

Make sure read() and write() fails on the control pipe.

branches: 1.25.2; 1.25.4;
Merge in a large batch of changes from Nick Hibma <hibma@skylink.it> so
the USB stack compiles on FreeBSD again.

Add an event mechanism so that a userland process can watch devices come
and go.

branches: 1.23.2;
Change the internal API to allow DMA buffers to be pre-allocated by
the device driver instead of happening automagically in the HC driver.
This affects both the HC-USBD interface as well as the USBD-device
interface.
This change will allow DMA buffers to be reused e.g. in isochronous
traffic.

Add isochronous support to the UHCI driver (not for OHCI yet).

Change the way the `struct device' base part of all driver softc are
declared and accessed to make it more portable. Idea from Nick Hibma, FreeBSD.
No functional changes.

Change the way the direction is extracted from the endpoint descriptor.
No functional changes to the drivers. From Nick Hibma, FreeBSD.

Change some 'struct device' to 'bdevice'. From FreeBSD.

Change a type name.

Make sure to mark the device as dying already in the (de)activate routine.
This avoids access to it before the detach routine has blown it away.

Move more of the transfer completion processing to HC independent code.
Fix some problems with transfer abort & timeout.

Use the right type for the size argument when calling usbd_bulk_transfer().

Make some small changes to make it compile on OpenBSD.

Some changes from FreeBSD (no functional differences).

Change it so that a pipe can be open RW instead of just R or W.
This makes close() work properly, but it is still not ideal. Perhaps
there should be different device nodes for input and output on to
and endpoint with the same number?
Pay attention to the SHORT_XFER_OK ioctl().

Totally redo the way device detach is done. It now uses a kernel event
thread and the config detach method.
Squish a number of space leaks on detach.

branches: 1.11.4;
Various little fixes from the FreeBSD version.

Fix some pastos.

Add an ugly workaround for a bug (feature) in the NetBSD open()/close()
protocol.

Fix bug in reading from interrupt pipe.
Make read and write on bulk pipes interruptible.

Don't set configuration unnecessarily, some (broken) devices seem to break
if you do.

Make it possible to specify the request flags when issuing a raw USB request.

Merge changes to make the USB stack work with FreeBSD. The original
diffs from Nick Hibma <n_hibma@freebsd.org>, but with substantial
changes from me.
XXX Not tested on FreeBSD yet.

Move initialization of sce around.

Take care of some lines > 80 chars.

Improvement to the ugen driver.
Better error checking.
Some code rearrengment.

Add a generic USB driver. It allows easy access to descriptors, the
control pipe via ioctl() and read() and write() access to bulk and
interrupt pipes.

Undo the DPRINTFN part of the last commit. It breaks i386 at least
when KERNHIST_LOG casts the pointer to uintmax_t.

Kernel ALL/i386 compiles again.

Debug printf tidy up, KNF comma,space nits.

Don't call versioned stuff "old". Follow the naming convention for versioning
and name them after the last version of the OS they appeared on.

various debug updates for some usb drivers

- several new *_DEBUG_DEFAULT options that allow usb debug values to
be set to a default that is non-zero:
EHCI_DEBUG_DEFAULT, UGEN_DEBUG_DEFAULT, URTWN_DEBUG_DEFAULT,
UMS_DEBUG_DEFAULT, and USB_DEBUG_DEFAULT
- ugen debug uses fewer usbhist lines for the same info
- ums.c converted from printf() to usbhist

ugen(4): Make sure opened is initialized in ugenopen.

Otherwise the error branch is based on garbage.

Fix USBDEBUG build on ILP32

ugen(4): convert to USBHIST style debugging.

Change the kqueue filterops::f_isfd field to filterops::f_flags, and
define a flag FILTEROP_ISFD that has the meaning of the prior f_isfd.
Field and flag name aligned with OpenBSD.

This does not constitute a functional or ABI change, as the field location
and size, and the value placed in that field, are the same as the previous
code, but we're bumping __NetBSD_Version__ so 3rd-party module source code
can adapt, as needed.

NetBSD 9.99.89

ugen(4): Keep fields null when not allocated; kassert on close.

This avoids silent leaks in DIAGNOSTIC kernels.

ugen(4): Use cv_wait loop for draining reference count on detach.

- Should be no need to use cv_timedwait because all users have now
been given a wakeup (previously writers were not, so we relied on
the timeouts to work out).

- Need to run this in a loop or else a spurious wakeup could cause us
to free data structures before the users have actually drained.

ugen(4): Use cv_broadcast to wake all I/O operations on detach.

Nothing prevents two concurrent reads or two concurrent writes on any
particular ugen endpoint, as far as I can tell, and we need to wake
all of them, so use cv_broadcast rather than cv_signal on detach.

XXX It's not clear to me that cv_signal in the xfer completion
callbacks is correct either: any one consumer might use less than the
full buffer. So I think either we should use cv_broadcast, or
consumers that don't use the whole buffer need to issue cv_signal too
to wake up another consumer even if we want to avoid a thundering
herd.

ugen(4): Issue explicit wakeup on detach for OUT endpoints too.

Writers can be blocked in cv_timedwait_sig too.

While here, fix comment: aborting the pipes does not cause all
waiters to wake, because the xfer completion callbacks sometimes skip
the notification. We should maybe change that, but this is a simpler
fix to ensure everyone waiting on I/O is woken to notice sc_dying.

ugen(4): Ensure we close pipes on detach.

Calling vdevgone has the effect of VOP_REVOKE -> spec_node_revoke ->
VOP_CLOSE -> spec_close -> cdev_close -> ugenclose, but:

(a) the flags passed to VOP_CLOSE don't have FREAD or FWRITE, and
(b) ugenclose has no effect when sc_dying is set anyway.

So create another path into the close logic, ugen_do_close. We have
to do this in detach _after_ the references have drained because we
may free buffers that other users are still using while the reference
count is nonzero.

ugen(4): Refuse non-forced detach with EBUSY if endpoints are open.

Sprinkle some comments.

ugen(4): Prevent ugenopen while ugen_set_config is in progress.

(except on the control endpoint)

Although we hold the kernel lock (which we should eventually change),
we may sleep in usbd_set_config_no at which point ugenopen might
happen and start making use of endpoint state which we'll stomp all
over once usbd_set_config_no returns. Setting sc_is_open[endpt]
while we wait prevents this.

ugen(4): Fix race of ugenopen against itself.

Even though we have the kernel lock held, a sleep during kmem_alloc
or usbd_open_pipe could allow another ugenopen to run concurrently
before we have marked the endpoint opened.

To avoid this, mark the endpoint open immediately (while we still
have the kernel lock held and before any sleeps, so there is no
TOCTOU error here), and then revert on unwind in the event of
failure.

ugen(4): Sprinkle KERNEL_LOCKED_P assertions around sc_is_open.

Use sel{record,remove}_knote().

branches: 1.157.2;
Fix ugen detach after partial attach.

While here, register null pmf handler even for partially attached
devices so they don't needlessly interfere with suspend.

Reported-by: syzbot+5a091d2e62da20b77259@syzkaller.appspotmail.com

Fix sloppy mistakes in previous.

1. Give the offset of the rbnode, not some other random members to
overwrite with garbage.

2. Don't try to unlock a mutex at NULL.

3. Make sure all paths out after ugenif_acquire go via
ugenif_release.

Share unit numbering for ugen and ugenif.

This way putting ugenif in kernel config actually works to wire it to
the /dev/ugenN.MM device nodes in userland.

Not a fully fleshed out solution to the ugen problem -- there's no
way for a userland driver to kick out a kernel driver and take over,
but this will let us, e.g., use uhidev(4) for Yubikey OTP/U2F/FIDO2
but ugen(4), with pcscd(8), for Yubikey CCID.

Fix various MP-safety issues while here (still not MPSAFE, but more
progress).

Convert DIAGNOSTIC prints to KASSERTs.

Hold the lock over access to the data structures it covers.

Still not MPSAFE, but progress.

Remove UGEN_ASLP microoptimization.

cv_signal already has this microoptimization.

While here, make the lock cover the relevant things we're issuing
cv_signal about -- progress toward real MP-safety.

KNG

revert the 0x% -> %# change for fixed width formats pointed out by uwe.

PR/55068: sc.dying: Fix printf formats:
- no %s/%p for kernel log
- 0x% -> %#
- always %j for kernel log

reading usbdi.c it looks like usbd_get_config_descriptor() can actually
return NULL, so check for this.
I got NULL pointer dereference here with a device showing:
[ 303.732632] ugen0: autoconfiguration error: setting configuration index 0 failed

localify

branches: 1.146.2;
remove explicit 'extern struct cfdriver <my>_cd;' and use ioconf.h

Rename the MODULE_*_HOOK() macros to MODULE_HOOK_*() as briefly
discussed on irc.

NFCI intended.

Ride the earlier kernel bump - it;s getting crowded.

Remove (mostly useless) usb_detach_{broadcast,wait} and replace with
cv_{broadcast,timedwait}

Really should loop on conditon.

Normalize all the compat hooks' names to the form

<subsystem>_<function>_<version>_hook

NFCI

XXX Note that although this introduces a change in the kernel-to-
XXX module interface, we are NOT bumping the kernel version number.
XXX We will bump the version number once the interface stabilizes.

Merge the [pgoyette-compat] branch

Enfore USB timeout on ugen(4) write operations

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: 1.139.2; 1.139.4;
Fix last:

Since config(1) could not distinguish between device and
interface attachments, it was generating only the latter.
Thus devices without their own driver wouldn't match the
ugen driver anymore.

Fix this by using a different device name for interface attachments.

Attach uftdi to each interface found in the device separately.
This allows for other drivers (e.g. ugen) to attach to some of
the other interfaces.

Allow ugen to attach only to some of the interfaces found in a device.

PR kern/52931 Kernel panics with Atheros usb wireless interface

Audit the flags to usbd_create_xfer so that USBD_FORCE_SHORT_XFER is
supplied wherever such a transfer is setup. We can drop
USBD_SHORT_XFER_OK as it has not bearing on number of TDs

Use C99 initializer for filterops

Mostly done with spatch with touchups for indentation

@@
expression a;
identifier b,c,d;
identifier p;
@@
const struct filterops p =
- { a, b, c, d
+ {
+ .f_isfd = a,
+ .f_attach = b,
+ .f_detach = c,
+ .f_event = d,
};

remove redundant checks against sc and sc->sc_dying.
check sc_dying in more places.

branches: 1.134.10;
KNF. Remove extra spaces. No functional change.

Merge nick-nhusb

- API / infrastructure changes to support memory management changes.
- Memory management improvements and bug fixes.
- HCDs should now be MP safe
- conversion to KERNHIST based debug
- FS/LS isoc support on ehci(4).
- conversion to kmem(9)
- Some USB 3 support - mostly from Takahiro HAYASHI (t-hash).
- interrupt transfers now get proper DMA operations
- general bug fixes
- kern/48308
- uhub status notification improvements
- umass(4) probe fix (applied to HEAD already)
- ohci(4) short transfer fix

KNF

Provide a ugen_clear_endpoints declaration

Only clear the endpoint information in ugen_set_interface only if setting
the new altno suceeds.

Avoids the null de-ref in PR/50597 and PR/50810

Remove always true conditional