Adopt <net/if_stats.h>.

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.

Use ETHER_LOCK()/ETHER_UNLOCK() for all ethernet drivers to protect ec_multi*.

Whitespace fix (mainly tabify).

-No functional change:
- Simplify struct ethercom's pointer near ETHER_FIRST_MULTI().
- Simplify MII structure initialization.
- u_int*_t -> uint*_t.
- KNF

Remove very old IFF_NOTRAILERS flag.

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macppc/if_gm.c (superseded by gem(4))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: 1.59.2;
Implement the BPF direction filter (BIOC[GS]DIRECTION). It provides backward
compatibility with BIOC[GS]SEESENT ioctl. The userland interface is the same
as FreeBSD.

This change also fixes a bug that the direction is misunderstand on some
environment by passing the direction to bpf_mtap*() instead of checking
m->m_pkthdr.rcvif.

It's not required to include net/bpfdesc.h. Remove it.

branches: 1.57.8; 1.57.14;
Move bpf_mtap and if_ipackets++ on Rx of each driver to percpuq if_input

The benefits of the change are:
- We can reduce codes
- We can provide the same behavior between drivers
- Where/When if_ipackets is counted up
- Note that some drivers still update packet statistics in their own
way (periodical update)
- Moved bpf_mtap run in softint
- This makes it easy to MP-ify bpf

Proposed on tech-kern and tech-net

branches: 1.56.2;
Introduce m_set_rcvif and m_reset_rcvif

The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.

No functional change.

Introduce softint-based if_input

This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.

This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.

To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.

Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html

Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!

Add missing splx() found by brainy.

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

branches: 1.52.4;
Merge tls-earlyentropy branch into HEAD.

branches: 1.51.10;
Second part of PR/46998. The following is taken from the PR with a slight
edit from me.

The TX buffer size is not a function of the SEEQ chip, it is an arbitrary
driver threshold, so I've renamed the define accordingly and added a new
one to specify how many buffers are used (in my port I allow multiple
packets to be in flight at once, so have used #ifndef, such that the
makefile can override).

Comment corrected, and make use of the above defines.

Missing delay(1) added, otherwise the 20,000 timeout loop is dependent
on the speed of your processor. Matches ea_stoptx logic now.

The FIFO empty check does nothing if the previous mode was 'read', but is
required before changing the BUFCODE (per 80C04 datasheet page 19,
note [2]). Then the mode is set to write, so a second FIFO empty check is
needed incase the previous mode was read.

Treat m0 as a pointer not an integer.

Remove double write of the NULL packet header. Either do
memset/ea_writebuf or two writes to SEEQ_BUFWIN, but not both.

The calculation of nextpacket (for hdr[]) assumes bufstart = 0, and puts
the packet header pointing in the wrong place when it isn't.

The setting of CFG2_OUTPUT is done in ea_init(), so doing it in ea_rxinit
is duplicated code.

Split softc/device_t
Use device_xname

From chuq

Rename eatxpacket to ea_txpacket for consistency.

First part of PR/46998

branches: 1.48.2;
PR/46576: Robert Sprowson: Shutdown doesn't disable TX/RX interrupts in
SEEQ8005 driver, plus misc white-space and 0->NULL fixes.

PR port-acorn32/46435: type mismatch, padbuf should be unsigned char.

branches: 1.46.2;
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: 1.45.8; 1.45.12;
Push the bpf_ops usage back into bpf.h. Push the common ifp->if_bpf
check into the inline functions as well the fourth argument for
bpf_attach.

branches: 1.44.2; 1.44.4;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

branches: 1.43.4;
use aprint_*_dev and device_xname

branches: 1.42.16;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.41.2; 1.41.6;
Constify: LLADDR -> CLLADDR. I'm aiming here to make it easier to
identify sockaddr_dl abuse that remains in the kernel, especially
the potential for overwriting memory past the end of a sockaddr_dl
with, e.g., memcpy(LLADDR(), ...).

branches: 1.40.2; 1.40.10; 1.40.14;
Use a char* variable for doing pointer arithmetic with.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.38.26;
merge ktrace-lwp.

branches: 1.37.4;
nuke trailing whitespace

branches: 1.36.4; 1.36.6;
When adding/deleting multicast addresses, only whack the address
filter if the interface is marked RUNNING.

Fixes kern/27678.

branches: 1.35.2; 1.35.6;
Use a properly zero'd buffer to pad the packet to ETHER_MIN_LEN
XXX should it be ETHER_MIN_LEN-ETHER_CRC_LEN ?

Quieten lint a little.

When copying a block with an odd size to or from the card, be very careful
not to overrun the end of it on the host side, since that might just slip over
a page boundary and cause an embarrassing kernel panic.

branches: 1.32.8; 1.32.10;
don't need <sys/types.h> when including <sys/param.h>

add/cleanup RCSID

Add support for feeding entropy to rnd(4).

branches: 1.29.2; 1.29.6;
bcopy -> strcpy

bcmp -> memcmp

Add support for handling full-size 802.1Q packets.

ea_mc_reset_8004(): Rather than doing the CRC32 ourselves, call
ether_crc32_be() to do it for us.

Fix bogons in previous commit:
* Remove harmful debugging code from seeq8005_attach().
* Re-insert variable declarations in ea_mc_reset_8004(), since they aren't
dead yet.

Attampt support for 8-bit mode. I believe this should work, but I need a more
flexible bus_space on arm26 before I can test it.

No need to pull in INET or NS headers -- that's all done by ether_ioctl() now.

In ea_writembuf(), reset the DMA pointer for each mbuf we copy, since
odd-sized mbufs will confuse things. This makes transmission of odd-sized
packets work.

Add a bit of voodoo code from the old "eb" driver that seems to prevent the
80C04 returning garbage when we read from it. This driver can now handle flood
pings without dropping packets.

Tweak product ID printing in light of the 80C04A data sheet.

Split out Rx and Tx interrupt handlers into their own functions for clarity.

On an 80C04, try not to start reads at addresses of the form xx,ea, xx,ee, or
xx,f0. This appears to heavily alleviate, but not to eliminate entirely,
the problems I've been seeing with garbage being read from the rx buffer.
I suspect the real solution lies elsewhere.

Print a warning message when the DMA FIFO fails to fill or empty before we
time out.

Print the address within the Rx buffer whenever anything goes wrong: it seems
to be significant.

Clean up debugging printfs a little.

Add some extra diagnostics:
* Use the timer to timoe out transmit operations.
* Spot when the "next packet" pointer falls outside the recieve buffer and
reset the interface.
* Don't reset the interface when we get a bad packet (unless there's
something else wrong as well).

Comment and #ifdef cleanup.

Tweak handling of multicast on 80C04. I think we need to set the match mode
to "multicast" for it to work, but I could be wrong.

Also make ALLMULTI work on 80C04, but this is untested.

Re-arrange board memory to suit the current (crude) transmit code.

Add ea_writembuf() from mark's code, and re-work eatxpacket() to use it.

Enormous pile of changes from mark's (uncommitted) work on this driver.
About the only bit of his code not here is the transmit routines, which I'll
merge in separately.

Also a few bug-fixes, so (for instance) multicast on an 8005 doesn't
immediately fall back to IFF_ALLMULTI.

Change all the register definitions to match the names using in mark's
improved version of the driver.

Handle IFF_ALLMULTI correctly. We can now cope with multicast, but for
some reason DAD fails when starting IPv6.

branches: 1.8.2;
Print the newline after the memory test. This makes it clear that the delay's
related to this driver rather than the next device to be attached.

ALTQ'ify.

branches: 1.6.2;
Move bpfattach()/bpfdetach() calls into ether_ifattach()/ether_ifdetach().

First attempt at multicast stuff. I suspect it doesn't work.
Use ether_ioctl -- saves 300 bytes of text.

Move the check for "promisc + unicast + not for us" into ether_input(),
and change Ethernet drivers to always pass all received frames to
ether_input() (with a few exceptions, which are documented in the
code).

Substantial overhaul, factoring out of common code, removal of redundancy
and general cleanup. Still not a beautiful driver, but one I'd be willing to
introduce to my parents now.

Add rudimentary 80C04 support (basically saying "Oh look, an 80C04!").
Move printing the initial ":" into the board driver, like i82586.c does.
Don't bother printing the amount of RAM, as it's always 64 KB.

Split the arm26 Ether3 (ea) driver into an MI driver for the SEEQ 8005 chip,
and a front-end driver for the Ether3. Only semantic change is to remove
ea_claimirq() and ea_releaseirq() on the grounds that the seem too spurious
to warrant a callback to the front-end.

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.

Use ETHER_LOCK()/ETHER_UNLOCK() for all ethernet drivers to protect ec_multi*.

Whitespace fix (mainly tabify).

-No functional change:
- Simplify struct ethercom's pointer near ETHER_FIRST_MULTI().
- Simplify MII structure initialization.
- u_int*_t -> uint*_t.
- KNF

Remove very old IFF_NOTRAILERS flag.

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: 1.59.2;
Implement the BPF direction filter (BIOC[GS]DIRECTION). It provides backward
compatibility with BIOC[GS]SEESENT ioctl. The userland interface is the same
as FreeBSD.

This change also fixes a bug that the direction is misunderstand on some
environment by passing the direction to bpf_mtap*() instead of checking
m->m_pkthdr.rcvif.

It's not required to include net/bpfdesc.h. Remove it.

branches: 1.57.8; 1.57.14;
Move bpf_mtap and if_ipackets++ on Rx of each driver to percpuq if_input

The benefits of the change are:
- We can reduce codes
- We can provide the same behavior between drivers
- Where/When if_ipackets is counted up
- Note that some drivers still update packet statistics in their own
way (periodical update)
- Moved bpf_mtap run in softint
- This makes it easy to MP-ify bpf

Proposed on tech-kern and tech-net

branches: 1.56.2;
Introduce m_set_rcvif and m_reset_rcvif

The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.

No functional change.

Introduce softint-based if_input

This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.

This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.

To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.

Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html

Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!

Add missing splx() found by brainy.

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

branches: 1.52.4;
Merge tls-earlyentropy branch into HEAD.

branches: 1.51.10;
Second part of PR/46998. The following is taken from the PR with a slight
edit from me.

The TX buffer size is not a function of the SEEQ chip, it is an arbitrary
driver threshold, so I've renamed the define accordingly and added a new
one to specify how many buffers are used (in my port I allow multiple
packets to be in flight at once, so have used #ifndef, such that the
makefile can override).

Comment corrected, and make use of the above defines.

Missing delay(1) added, otherwise the 20,000 timeout loop is dependent
on the speed of your processor. Matches ea_stoptx logic now.

The FIFO empty check does nothing if the previous mode was 'read', but is
required before changing the BUFCODE (per 80C04 datasheet page 19,
note [2]). Then the mode is set to write, so a second FIFO empty check is
needed incase the previous mode was read.

Treat m0 as a pointer not an integer.

Remove double write of the NULL packet header. Either do
memset/ea_writebuf or two writes to SEEQ_BUFWIN, but not both.

The calculation of nextpacket (for hdr[]) assumes bufstart = 0, and puts
the packet header pointing in the wrong place when it isn't.

The setting of CFG2_OUTPUT is done in ea_init(), so doing it in ea_rxinit
is duplicated code.

Split softc/device_t
Use device_xname

From chuq

Rename eatxpacket to ea_txpacket for consistency.

First part of PR/46998

branches: 1.48.2;
PR/46576: Robert Sprowson: Shutdown doesn't disable TX/RX interrupts in
SEEQ8005 driver, plus misc white-space and 0->NULL fixes.

PR port-acorn32/46435: type mismatch, padbuf should be unsigned char.

branches: 1.46.2;
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: 1.45.8; 1.45.12;
Push the bpf_ops usage back into bpf.h. Push the common ifp->if_bpf
check into the inline functions as well the fourth argument for
bpf_attach.

branches: 1.44.2; 1.44.4;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

branches: 1.43.4;
use aprint_*_dev and device_xname

branches: 1.42.16;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.41.2; 1.41.6;
Constify: LLADDR -> CLLADDR. I'm aiming here to make it easier to
identify sockaddr_dl abuse that remains in the kernel, especially
the potential for overwriting memory past the end of a sockaddr_dl
with, e.g., memcpy(LLADDR(), ...).

branches: 1.40.2; 1.40.10; 1.40.14;
Use a char* variable for doing pointer arithmetic with.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.38.26;
merge ktrace-lwp.

branches: 1.37.4;
nuke trailing whitespace

branches: 1.36.4; 1.36.6;
When adding/deleting multicast addresses, only whack the address
filter if the interface is marked RUNNING.

Fixes kern/27678.

branches: 1.35.2; 1.35.6;
Use a properly zero'd buffer to pad the packet to ETHER_MIN_LEN
XXX should it be ETHER_MIN_LEN-ETHER_CRC_LEN ?

Quieten lint a little.

When copying a block with an odd size to or from the card, be very careful
not to overrun the end of it on the host side, since that might just slip over
a page boundary and cause an embarrassing kernel panic.

branches: 1.32.8; 1.32.10;
don't need <sys/types.h> when including <sys/param.h>

add/cleanup RCSID

Add support for feeding entropy to rnd(4).

branches: 1.29.2; 1.29.6;
bcopy -> strcpy

bcmp -> memcmp

Add support for handling full-size 802.1Q packets.

ea_mc_reset_8004(): Rather than doing the CRC32 ourselves, call
ether_crc32_be() to do it for us.

Fix bogons in previous commit:
* Remove harmful debugging code from seeq8005_attach().
* Re-insert variable declarations in ea_mc_reset_8004(), since they aren't
dead yet.

Attampt support for 8-bit mode. I believe this should work, but I need a more
flexible bus_space on arm26 before I can test it.

No need to pull in INET or NS headers -- that's all done by ether_ioctl() now.

In ea_writembuf(), reset the DMA pointer for each mbuf we copy, since
odd-sized mbufs will confuse things. This makes transmission of odd-sized
packets work.

Add a bit of voodoo code from the old "eb" driver that seems to prevent the
80C04 returning garbage when we read from it. This driver can now handle flood
pings without dropping packets.

Tweak product ID printing in light of the 80C04A data sheet.

Split out Rx and Tx interrupt handlers into their own functions for clarity.

On an 80C04, try not to start reads at addresses of the form xx,ea, xx,ee, or
xx,f0. This appears to heavily alleviate, but not to eliminate entirely,
the problems I've been seeing with garbage being read from the rx buffer.
I suspect the real solution lies elsewhere.

Print a warning message when the DMA FIFO fails to fill or empty before we
time out.

Print the address within the Rx buffer whenever anything goes wrong: it seems
to be significant.

Clean up debugging printfs a little.

Add some extra diagnostics:
* Use the timer to timoe out transmit operations.
* Spot when the "next packet" pointer falls outside the recieve buffer and
reset the interface.
* Don't reset the interface when we get a bad packet (unless there's
something else wrong as well).

Comment and #ifdef cleanup.

Tweak handling of multicast on 80C04. I think we need to set the match mode
to "multicast" for it to work, but I could be wrong.

Also make ALLMULTI work on 80C04, but this is untested.

Re-arrange board memory to suit the current (crude) transmit code.

Add ea_writembuf() from mark's code, and re-work eatxpacket() to use it.

Enormous pile of changes from mark's (uncommitted) work on this driver.
About the only bit of his code not here is the transmit routines, which I'll
merge in separately.

Also a few bug-fixes, so (for instance) multicast on an 8005 doesn't
immediately fall back to IFF_ALLMULTI.

Change all the register definitions to match the names using in mark's
improved version of the driver.

Handle IFF_ALLMULTI correctly. We can now cope with multicast, but for
some reason DAD fails when starting IPv6.

branches: 1.8.2;
Print the newline after the memory test. This makes it clear that the delay's
related to this driver rather than the next device to be attached.

ALTQ'ify.

branches: 1.6.2;
Move bpfattach()/bpfdetach() calls into ether_ifattach()/ether_ifdetach().

First attempt at multicast stuff. I suspect it doesn't work.
Use ether_ioctl -- saves 300 bytes of text.

Move the check for "promisc + unicast + not for us" into ether_input(),
and change Ethernet drivers to always pass all received frames to
ether_input() (with a few exceptions, which are documented in the
code).

Substantial overhaul, factoring out of common code, removal of redundancy
and general cleanup. Still not a beautiful driver, but one I'd be willing to
introduce to my parents now.

Add rudimentary 80C04 support (basically saying "Oh look, an 80C04!").
Move printing the initial ":" into the board driver, like i82586.c does.
Don't bother printing the amount of RAM, as it's always 64 KB.

Split the arm26 Ether3 (ea) driver into an MI driver for the SEEQ 8005 chip,
and a front-end driver for the Ether3. Only semantic change is to remove
ea_claimirq() and ea_releaseirq() on the grounds that the seem too spurious
to warrant a callback to the front-end.

Use ETHER_LOCK()/ETHER_UNLOCK() for all ethernet drivers to protect ec_multi*.

Whitespace fix (mainly tabify).

-No functional change:
- Simplify struct ethercom's pointer near ETHER_FIRST_MULTI().
- Simplify MII structure initialization.
- u_int*_t -> uint*_t.
- KNF

Remove very old IFF_NOTRAILERS flag.

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!)

Implement the BPF direction filter (BIOC[GS]DIRECTION). It provides backward
compatibility with BIOC[GS]SEESENT ioctl. The userland interface is the same
as FreeBSD.

This change also fixes a bug that the direction is misunderstand on some
environment by passing the direction to bpf_mtap*() instead of checking
m->m_pkthdr.rcvif.

It's not required to include net/bpfdesc.h. Remove it.

branches: 1.57.8; 1.57.14;
Move bpf_mtap and if_ipackets++ on Rx of each driver to percpuq if_input

The benefits of the change are:
- We can reduce codes
- We can provide the same behavior between drivers
- Where/When if_ipackets is counted up
- Note that some drivers still update packet statistics in their own
way (periodical update)
- Moved bpf_mtap run in softint
- This makes it easy to MP-ify bpf

Proposed on tech-kern and tech-net

branches: 1.56.2;
Introduce m_set_rcvif and m_reset_rcvif

The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.

No functional change.

Introduce softint-based if_input

This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.

This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.

To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.

Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html

Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!

Add missing splx() found by brainy.

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

branches: 1.52.4;
Merge tls-earlyentropy branch into HEAD.

branches: 1.51.10;
Second part of PR/46998. The following is taken from the PR with a slight
edit from me.

The TX buffer size is not a function of the SEEQ chip, it is an arbitrary
driver threshold, so I've renamed the define accordingly and added a new
one to specify how many buffers are used (in my port I allow multiple
packets to be in flight at once, so have used #ifndef, such that the
makefile can override).

Comment corrected, and make use of the above defines.

Missing delay(1) added, otherwise the 20,000 timeout loop is dependent
on the speed of your processor. Matches ea_stoptx logic now.

The FIFO empty check does nothing if the previous mode was 'read', but is
required before changing the BUFCODE (per 80C04 datasheet page 19,
note [2]). Then the mode is set to write, so a second FIFO empty check is
needed incase the previous mode was read.

Treat m0 as a pointer not an integer.

Remove double write of the NULL packet header. Either do
memset/ea_writebuf or two writes to SEEQ_BUFWIN, but not both.

The calculation of nextpacket (for hdr[]) assumes bufstart = 0, and puts
the packet header pointing in the wrong place when it isn't.

The setting of CFG2_OUTPUT is done in ea_init(), so doing it in ea_rxinit
is duplicated code.

Split softc/device_t
Use device_xname

From chuq

Rename eatxpacket to ea_txpacket for consistency.

First part of PR/46998

branches: 1.48.2;
PR/46576: Robert Sprowson: Shutdown doesn't disable TX/RX interrupts in
SEEQ8005 driver, plus misc white-space and 0->NULL fixes.

PR port-acorn32/46435: type mismatch, padbuf should be unsigned char.

branches: 1.46.2;
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: 1.45.8; 1.45.12;
Push the bpf_ops usage back into bpf.h. Push the common ifp->if_bpf
check into the inline functions as well the fourth argument for
bpf_attach.

branches: 1.44.2; 1.44.4;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

branches: 1.43.4;
use aprint_*_dev and device_xname

branches: 1.42.16;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.41.2; 1.41.6;
Constify: LLADDR -> CLLADDR. I'm aiming here to make it easier to
identify sockaddr_dl abuse that remains in the kernel, especially
the potential for overwriting memory past the end of a sockaddr_dl
with, e.g., memcpy(LLADDR(), ...).

branches: 1.40.2; 1.40.10; 1.40.14;
Use a char* variable for doing pointer arithmetic with.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.38.26;
merge ktrace-lwp.

branches: 1.37.4;
nuke trailing whitespace

branches: 1.36.4; 1.36.6;
When adding/deleting multicast addresses, only whack the address
filter if the interface is marked RUNNING.

Fixes kern/27678.

branches: 1.35.2; 1.35.6;
Use a properly zero'd buffer to pad the packet to ETHER_MIN_LEN
XXX should it be ETHER_MIN_LEN-ETHER_CRC_LEN ?

Quieten lint a little.

When copying a block with an odd size to or from the card, be very careful
not to overrun the end of it on the host side, since that might just slip over
a page boundary and cause an embarrassing kernel panic.

branches: 1.32.8; 1.32.10;
don't need <sys/types.h> when including <sys/param.h>

add/cleanup RCSID

Add support for feeding entropy to rnd(4).

branches: 1.29.2; 1.29.6;
bcopy -> strcpy

bcmp -> memcmp

Add support for handling full-size 802.1Q packets.

ea_mc_reset_8004(): Rather than doing the CRC32 ourselves, call
ether_crc32_be() to do it for us.

Fix bogons in previous commit:
* Remove harmful debugging code from seeq8005_attach().
* Re-insert variable declarations in ea_mc_reset_8004(), since they aren't
dead yet.

Attampt support for 8-bit mode. I believe this should work, but I need a more
flexible bus_space on arm26 before I can test it.

No need to pull in INET or NS headers -- that's all done by ether_ioctl() now.

In ea_writembuf(), reset the DMA pointer for each mbuf we copy, since
odd-sized mbufs will confuse things. This makes transmission of odd-sized
packets work.

Add a bit of voodoo code from the old "eb" driver that seems to prevent the
80C04 returning garbage when we read from it. This driver can now handle flood
pings without dropping packets.

Tweak product ID printing in light of the 80C04A data sheet.

Split out Rx and Tx interrupt handlers into their own functions for clarity.

On an 80C04, try not to start reads at addresses of the form xx,ea, xx,ee, or
xx,f0. This appears to heavily alleviate, but not to eliminate entirely,
the problems I've been seeing with garbage being read from the rx buffer.
I suspect the real solution lies elsewhere.

Print a warning message when the DMA FIFO fails to fill or empty before we
time out.

Print the address within the Rx buffer whenever anything goes wrong: it seems
to be significant.

Clean up debugging printfs a little.

Add some extra diagnostics:
* Use the timer to timoe out transmit operations.
* Spot when the "next packet" pointer falls outside the recieve buffer and
reset the interface.
* Don't reset the interface when we get a bad packet (unless there's
something else wrong as well).

Comment and #ifdef cleanup.

Tweak handling of multicast on 80C04. I think we need to set the match mode
to "multicast" for it to work, but I could be wrong.

Also make ALLMULTI work on 80C04, but this is untested.

Re-arrange board memory to suit the current (crude) transmit code.

Add ea_writembuf() from mark's code, and re-work eatxpacket() to use it.

Enormous pile of changes from mark's (uncommitted) work on this driver.
About the only bit of his code not here is the transmit routines, which I'll
merge in separately.

Also a few bug-fixes, so (for instance) multicast on an 8005 doesn't
immediately fall back to IFF_ALLMULTI.

Change all the register definitions to match the names using in mark's
improved version of the driver.

Handle IFF_ALLMULTI correctly. We can now cope with multicast, but for
some reason DAD fails when starting IPv6.

branches: 1.8.2;
Print the newline after the memory test. This makes it clear that the delay's
related to this driver rather than the next device to be attached.

ALTQ'ify.

branches: 1.6.2;
Move bpfattach()/bpfdetach() calls into ether_ifattach()/ether_ifdetach().

First attempt at multicast stuff. I suspect it doesn't work.
Use ether_ioctl -- saves 300 bytes of text.

Move the check for "promisc + unicast + not for us" into ether_input(),
and change Ethernet drivers to always pass all received frames to
ether_input() (with a few exceptions, which are documented in the
code).

Substantial overhaul, factoring out of common code, removal of redundancy
and general cleanup. Still not a beautiful driver, but one I'd be willing to
introduce to my parents now.

Add rudimentary 80C04 support (basically saying "Oh look, an 80C04!").
Move printing the initial ":" into the board driver, like i82586.c does.
Don't bother printing the amount of RAM, as it's always 64 KB.

Split the arm26 Ether3 (ea) driver into an MI driver for the SEEQ 8005 chip,
and a front-end driver for the Ether3. Only semantic change is to remove
ea_claimirq() and ea_releaseirq() on the grounds that the seem too spurious
to warrant a callback to the front-end.

Whitespace fix (mainly tabify).

-No functional change:
- Simplify struct ethercom's pointer near ETHER_FIRST_MULTI().
- Simplify MII structure initialization.
- u_int*_t -> uint*_t.
- KNF

Remove very old IFF_NOTRAILERS flag.

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!)

Implement the BPF direction filter (BIOC[GS]DIRECTION). It provides backward
compatibility with BIOC[GS]SEESENT ioctl. The userland interface is the same
as FreeBSD.

This change also fixes a bug that the direction is misunderstand on some
environment by passing the direction to bpf_mtap*() instead of checking
m->m_pkthdr.rcvif.

It's not required to include net/bpfdesc.h. Remove it.

branches: 1.57.8; 1.57.14;
Move bpf_mtap and if_ipackets++ on Rx of each driver to percpuq if_input

The benefits of the change are:
- We can reduce codes
- We can provide the same behavior between drivers
- Where/When if_ipackets is counted up
- Note that some drivers still update packet statistics in their own
way (periodical update)
- Moved bpf_mtap run in softint
- This makes it easy to MP-ify bpf

Proposed on tech-kern and tech-net

branches: 1.56.2;
Introduce m_set_rcvif and m_reset_rcvif

The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.

No functional change.

Introduce softint-based if_input

This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.

This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.

To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.

Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html

Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!

Add missing splx() found by brainy.

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

branches: 1.52.4;
Merge tls-earlyentropy branch into HEAD.

branches: 1.51.10;
Second part of PR/46998. The following is taken from the PR with a slight
edit from me.

The TX buffer size is not a function of the SEEQ chip, it is an arbitrary
driver threshold, so I've renamed the define accordingly and added a new
one to specify how many buffers are used (in my port I allow multiple
packets to be in flight at once, so have used #ifndef, such that the
makefile can override).

Comment corrected, and make use of the above defines.

Missing delay(1) added, otherwise the 20,000 timeout loop is dependent
on the speed of your processor. Matches ea_stoptx logic now.

The FIFO empty check does nothing if the previous mode was 'read', but is
required before changing the BUFCODE (per 80C04 datasheet page 19,
note [2]). Then the mode is set to write, so a second FIFO empty check is
needed incase the previous mode was read.

Treat m0 as a pointer not an integer.

Remove double write of the NULL packet header. Either do
memset/ea_writebuf or two writes to SEEQ_BUFWIN, but not both.

The calculation of nextpacket (for hdr[]) assumes bufstart = 0, and puts
the packet header pointing in the wrong place when it isn't.

The setting of CFG2_OUTPUT is done in ea_init(), so doing it in ea_rxinit
is duplicated code.

Split softc/device_t
Use device_xname

From chuq

Rename eatxpacket to ea_txpacket for consistency.

First part of PR/46998

branches: 1.48.2;
PR/46576: Robert Sprowson: Shutdown doesn't disable TX/RX interrupts in
SEEQ8005 driver, plus misc white-space and 0->NULL fixes.

PR port-acorn32/46435: type mismatch, padbuf should be unsigned char.

branches: 1.46.2;
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: 1.45.8; 1.45.12;
Push the bpf_ops usage back into bpf.h. Push the common ifp->if_bpf
check into the inline functions as well the fourth argument for
bpf_attach.

branches: 1.44.2; 1.44.4;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

branches: 1.43.4;
use aprint_*_dev and device_xname

branches: 1.42.16;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.41.2; 1.41.6;
Constify: LLADDR -> CLLADDR. I'm aiming here to make it easier to
identify sockaddr_dl abuse that remains in the kernel, especially
the potential for overwriting memory past the end of a sockaddr_dl
with, e.g., memcpy(LLADDR(), ...).

branches: 1.40.2; 1.40.10; 1.40.14;
Use a char* variable for doing pointer arithmetic with.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.38.26;
merge ktrace-lwp.

branches: 1.37.4;
nuke trailing whitespace

branches: 1.36.4; 1.36.6;
When adding/deleting multicast addresses, only whack the address
filter if the interface is marked RUNNING.

Fixes kern/27678.

branches: 1.35.2; 1.35.6;
Use a properly zero'd buffer to pad the packet to ETHER_MIN_LEN
XXX should it be ETHER_MIN_LEN-ETHER_CRC_LEN ?

Quieten lint a little.

When copying a block with an odd size to or from the card, be very careful
not to overrun the end of it on the host side, since that might just slip over
a page boundary and cause an embarrassing kernel panic.

branches: 1.32.8; 1.32.10;
don't need <sys/types.h> when including <sys/param.h>

add/cleanup RCSID

Add support for feeding entropy to rnd(4).

branches: 1.29.2; 1.29.6;
bcopy -> strcpy

bcmp -> memcmp

Add support for handling full-size 802.1Q packets.

ea_mc_reset_8004(): Rather than doing the CRC32 ourselves, call
ether_crc32_be() to do it for us.

Fix bogons in previous commit:
* Remove harmful debugging code from seeq8005_attach().
* Re-insert variable declarations in ea_mc_reset_8004(), since they aren't
dead yet.

Attampt support for 8-bit mode. I believe this should work, but I need a more
flexible bus_space on arm26 before I can test it.

No need to pull in INET or NS headers -- that's all done by ether_ioctl() now.

In ea_writembuf(), reset the DMA pointer for each mbuf we copy, since
odd-sized mbufs will confuse things. This makes transmission of odd-sized
packets work.

Add a bit of voodoo code from the old "eb" driver that seems to prevent the
80C04 returning garbage when we read from it. This driver can now handle flood
pings without dropping packets.

Tweak product ID printing in light of the 80C04A data sheet.

Split out Rx and Tx interrupt handlers into their own functions for clarity.

On an 80C04, try not to start reads at addresses of the form xx,ea, xx,ee, or
xx,f0. This appears to heavily alleviate, but not to eliminate entirely,
the problems I've been seeing with garbage being read from the rx buffer.
I suspect the real solution lies elsewhere.

Print a warning message when the DMA FIFO fails to fill or empty before we
time out.

Print the address within the Rx buffer whenever anything goes wrong: it seems
to be significant.

Clean up debugging printfs a little.

Add some extra diagnostics:
* Use the timer to timoe out transmit operations.
* Spot when the "next packet" pointer falls outside the recieve buffer and
reset the interface.
* Don't reset the interface when we get a bad packet (unless there's
something else wrong as well).

Comment and #ifdef cleanup.

Tweak handling of multicast on 80C04. I think we need to set the match mode
to "multicast" for it to work, but I could be wrong.

Also make ALLMULTI work on 80C04, but this is untested.

Re-arrange board memory to suit the current (crude) transmit code.

Add ea_writembuf() from mark's code, and re-work eatxpacket() to use it.

Enormous pile of changes from mark's (uncommitted) work on this driver.
About the only bit of his code not here is the transmit routines, which I'll
merge in separately.

Also a few bug-fixes, so (for instance) multicast on an 8005 doesn't
immediately fall back to IFF_ALLMULTI.

Change all the register definitions to match the names using in mark's
improved version of the driver.

Handle IFF_ALLMULTI correctly. We can now cope with multicast, but for
some reason DAD fails when starting IPv6.

branches: 1.8.2;
Print the newline after the memory test. This makes it clear that the delay's
related to this driver rather than the next device to be attached.

ALTQ'ify.

branches: 1.6.2;
Move bpfattach()/bpfdetach() calls into ether_ifattach()/ether_ifdetach().

First attempt at multicast stuff. I suspect it doesn't work.
Use ether_ioctl -- saves 300 bytes of text.

Move the check for "promisc + unicast + not for us" into ether_input(),
and change Ethernet drivers to always pass all received frames to
ether_input() (with a few exceptions, which are documented in the
code).

Substantial overhaul, factoring out of common code, removal of redundancy
and general cleanup. Still not a beautiful driver, but one I'd be willing to
introduce to my parents now.

Add rudimentary 80C04 support (basically saying "Oh look, an 80C04!").
Move printing the initial ":" into the board driver, like i82586.c does.
Don't bother printing the amount of RAM, as it's always 64 KB.

Split the arm26 Ether3 (ea) driver into an MI driver for the SEEQ 8005 chip,
and a front-end driver for the Ether3. Only semantic change is to remove
ea_claimirq() and ea_releaseirq() on the grounds that the seem too spurious
to warrant a callback to the front-end.

-No functional change:
- Simplify struct ethercom's pointer near ETHER_FIRST_MULTI().
- Simplify MII structure initialization.
- u_int*_t -> uint*_t.
- KNF

Remove very old IFF_NOTRAILERS flag.

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!)

Implement the BPF direction filter (BIOC[GS]DIRECTION). It provides backward
compatibility with BIOC[GS]SEESENT ioctl. The userland interface is the same
as FreeBSD.

This change also fixes a bug that the direction is misunderstand on some
environment by passing the direction to bpf_mtap*() instead of checking
m->m_pkthdr.rcvif.

It's not required to include net/bpfdesc.h. Remove it.

branches: 1.57.8; 1.57.14;
Move bpf_mtap and if_ipackets++ on Rx of each driver to percpuq if_input

The benefits of the change are:
- We can reduce codes
- We can provide the same behavior between drivers
- Where/When if_ipackets is counted up
- Note that some drivers still update packet statistics in their own
way (periodical update)
- Moved bpf_mtap run in softint
- This makes it easy to MP-ify bpf

Proposed on tech-kern and tech-net

branches: 1.56.2;
Introduce m_set_rcvif and m_reset_rcvif

The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.

No functional change.

Introduce softint-based if_input

This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.

This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.

To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.

Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html

Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!

Add missing splx() found by brainy.

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

branches: 1.52.4;
Merge tls-earlyentropy branch into HEAD.

branches: 1.51.10;
Second part of PR/46998. The following is taken from the PR with a slight
edit from me.

The TX buffer size is not a function of the SEEQ chip, it is an arbitrary
driver threshold, so I've renamed the define accordingly and added a new
one to specify how many buffers are used (in my port I allow multiple
packets to be in flight at once, so have used #ifndef, such that the
makefile can override).

Comment corrected, and make use of the above defines.

Missing delay(1) added, otherwise the 20,000 timeout loop is dependent
on the speed of your processor. Matches ea_stoptx logic now.

The FIFO empty check does nothing if the previous mode was 'read', but is
required before changing the BUFCODE (per 80C04 datasheet page 19,
note [2]). Then the mode is set to write, so a second FIFO empty check is
needed incase the previous mode was read.

Treat m0 as a pointer not an integer.

Remove double write of the NULL packet header. Either do
memset/ea_writebuf or two writes to SEEQ_BUFWIN, but not both.

The calculation of nextpacket (for hdr[]) assumes bufstart = 0, and puts
the packet header pointing in the wrong place when it isn't.

The setting of CFG2_OUTPUT is done in ea_init(), so doing it in ea_rxinit
is duplicated code.

Split softc/device_t
Use device_xname

From chuq

Rename eatxpacket to ea_txpacket for consistency.

First part of PR/46998

branches: 1.48.2;
PR/46576: Robert Sprowson: Shutdown doesn't disable TX/RX interrupts in
SEEQ8005 driver, plus misc white-space and 0->NULL fixes.

PR port-acorn32/46435: type mismatch, padbuf should be unsigned char.

branches: 1.46.2;
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: 1.45.8; 1.45.12;
Push the bpf_ops usage back into bpf.h. Push the common ifp->if_bpf
check into the inline functions as well the fourth argument for
bpf_attach.

branches: 1.44.2; 1.44.4;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

branches: 1.43.4;
use aprint_*_dev and device_xname

branches: 1.42.16;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.41.2; 1.41.6;
Constify: LLADDR -> CLLADDR. I'm aiming here to make it easier to
identify sockaddr_dl abuse that remains in the kernel, especially
the potential for overwriting memory past the end of a sockaddr_dl
with, e.g., memcpy(LLADDR(), ...).

branches: 1.40.2; 1.40.10; 1.40.14;
Use a char* variable for doing pointer arithmetic with.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.38.26;
merge ktrace-lwp.

branches: 1.37.4;
nuke trailing whitespace

branches: 1.36.4; 1.36.6;
When adding/deleting multicast addresses, only whack the address
filter if the interface is marked RUNNING.

Fixes kern/27678.

branches: 1.35.2; 1.35.6;
Use a properly zero'd buffer to pad the packet to ETHER_MIN_LEN
XXX should it be ETHER_MIN_LEN-ETHER_CRC_LEN ?

Quieten lint a little.

When copying a block with an odd size to or from the card, be very careful
not to overrun the end of it on the host side, since that might just slip over
a page boundary and cause an embarrassing kernel panic.

branches: 1.32.8; 1.32.10;
don't need <sys/types.h> when including <sys/param.h>

add/cleanup RCSID

Add support for feeding entropy to rnd(4).

branches: 1.29.2; 1.29.6;
bcopy -> strcpy

bcmp -> memcmp

Add support for handling full-size 802.1Q packets.

ea_mc_reset_8004(): Rather than doing the CRC32 ourselves, call
ether_crc32_be() to do it for us.

Fix bogons in previous commit:
* Remove harmful debugging code from seeq8005_attach().
* Re-insert variable declarations in ea_mc_reset_8004(), since they aren't
dead yet.

Attampt support for 8-bit mode. I believe this should work, but I need a more
flexible bus_space on arm26 before I can test it.

No need to pull in INET or NS headers -- that's all done by ether_ioctl() now.

In ea_writembuf(), reset the DMA pointer for each mbuf we copy, since
odd-sized mbufs will confuse things. This makes transmission of odd-sized
packets work.

Add a bit of voodoo code from the old "eb" driver that seems to prevent the
80C04 returning garbage when we read from it. This driver can now handle flood
pings without dropping packets.

Tweak product ID printing in light of the 80C04A data sheet.

Split out Rx and Tx interrupt handlers into their own functions for clarity.

On an 80C04, try not to start reads at addresses of the form xx,ea, xx,ee, or
xx,f0. This appears to heavily alleviate, but not to eliminate entirely,
the problems I've been seeing with garbage being read from the rx buffer.
I suspect the real solution lies elsewhere.

Print a warning message when the DMA FIFO fails to fill or empty before we
time out.

Print the address within the Rx buffer whenever anything goes wrong: it seems
to be significant.

Clean up debugging printfs a little.

Add some extra diagnostics:
* Use the timer to timoe out transmit operations.
* Spot when the "next packet" pointer falls outside the recieve buffer and
reset the interface.
* Don't reset the interface when we get a bad packet (unless there's
something else wrong as well).

Comment and #ifdef cleanup.

Tweak handling of multicast on 80C04. I think we need to set the match mode
to "multicast" for it to work, but I could be wrong.

Also make ALLMULTI work on 80C04, but this is untested.

Re-arrange board memory to suit the current (crude) transmit code.

Add ea_writembuf() from mark's code, and re-work eatxpacket() to use it.

Enormous pile of changes from mark's (uncommitted) work on this driver.
About the only bit of his code not here is the transmit routines, which I'll
merge in separately.

Also a few bug-fixes, so (for instance) multicast on an 8005 doesn't
immediately fall back to IFF_ALLMULTI.

Change all the register definitions to match the names using in mark's
improved version of the driver.

Handle IFF_ALLMULTI correctly. We can now cope with multicast, but for
some reason DAD fails when starting IPv6.

branches: 1.8.2;
Print the newline after the memory test. This makes it clear that the delay's
related to this driver rather than the next device to be attached.

ALTQ'ify.

branches: 1.6.2;
Move bpfattach()/bpfdetach() calls into ether_ifattach()/ether_ifdetach().

First attempt at multicast stuff. I suspect it doesn't work.
Use ether_ioctl -- saves 300 bytes of text.

Move the check for "promisc + unicast + not for us" into ether_input(),
and change Ethernet drivers to always pass all received frames to
ether_input() (with a few exceptions, which are documented in the
code).

Substantial overhaul, factoring out of common code, removal of redundancy
and general cleanup. Still not a beautiful driver, but one I'd be willing to
introduce to my parents now.

Add rudimentary 80C04 support (basically saying "Oh look, an 80C04!").
Move printing the initial ":" into the board driver, like i82586.c does.
Don't bother printing the amount of RAM, as it's always 64 KB.

Split the arm26 Ether3 (ea) driver into an MI driver for the SEEQ 8005 chip,
and a front-end driver for the Ether3. Only semantic change is to remove
ea_claimirq() and ea_releaseirq() on the grounds that the seem too spurious
to warrant a callback to the front-end.

Remove very old IFF_NOTRAILERS flag.

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!)

Implement the BPF direction filter (BIOC[GS]DIRECTION). It provides backward
compatibility with BIOC[GS]SEESENT ioctl. The userland interface is the same
as FreeBSD.

This change also fixes a bug that the direction is misunderstand on some
environment by passing the direction to bpf_mtap*() instead of checking
m->m_pkthdr.rcvif.

It's not required to include net/bpfdesc.h. Remove it.

branches: 1.57.8; 1.57.14;
Move bpf_mtap and if_ipackets++ on Rx of each driver to percpuq if_input

The benefits of the change are:
- We can reduce codes
- We can provide the same behavior between drivers
- Where/When if_ipackets is counted up
- Note that some drivers still update packet statistics in their own
way (periodical update)
- Moved bpf_mtap run in softint
- This makes it easy to MP-ify bpf

Proposed on tech-kern and tech-net

branches: 1.56.2;
Introduce m_set_rcvif and m_reset_rcvif

The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.

No functional change.

Introduce softint-based if_input

This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.

This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.

To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.

Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html

Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!

Add missing splx() found by brainy.

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

branches: 1.52.4;
Merge tls-earlyentropy branch into HEAD.

branches: 1.51.10;
Second part of PR/46998. The following is taken from the PR with a slight
edit from me.

The TX buffer size is not a function of the SEEQ chip, it is an arbitrary
driver threshold, so I've renamed the define accordingly and added a new
one to specify how many buffers are used (in my port I allow multiple
packets to be in flight at once, so have used #ifndef, such that the
makefile can override).

Comment corrected, and make use of the above defines.

Missing delay(1) added, otherwise the 20,000 timeout loop is dependent
on the speed of your processor. Matches ea_stoptx logic now.

The FIFO empty check does nothing if the previous mode was 'read', but is
required before changing the BUFCODE (per 80C04 datasheet page 19,
note [2]). Then the mode is set to write, so a second FIFO empty check is
needed incase the previous mode was read.

Treat m0 as a pointer not an integer.

Remove double write of the NULL packet header. Either do
memset/ea_writebuf or two writes to SEEQ_BUFWIN, but not both.

The calculation of nextpacket (for hdr[]) assumes bufstart = 0, and puts
the packet header pointing in the wrong place when it isn't.

The setting of CFG2_OUTPUT is done in ea_init(), so doing it in ea_rxinit
is duplicated code.

Split softc/device_t
Use device_xname

From chuq

Rename eatxpacket to ea_txpacket for consistency.

First part of PR/46998

branches: 1.48.2;
PR/46576: Robert Sprowson: Shutdown doesn't disable TX/RX interrupts in
SEEQ8005 driver, plus misc white-space and 0->NULL fixes.

PR port-acorn32/46435: type mismatch, padbuf should be unsigned char.

branches: 1.46.2;
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: 1.45.8; 1.45.12;
Push the bpf_ops usage back into bpf.h. Push the common ifp->if_bpf
check into the inline functions as well the fourth argument for
bpf_attach.

branches: 1.44.2; 1.44.4;
Redefine bpf linkage through an always present op vector, i.e.
#if NBPFILTER is no longer required in the client. This change
doesn't yet add support for loading bpf as a module, since drivers
can register before bpf is attached. However, callers of bpf can
now be modularized.

Dynamically loadable bpf could probably be done fairly easily with
coordination from the stub driver and the real driver by registering
attachments in the stub before the real driver is loaded and doing
a handoff. ... and I'm not going to ponder the depths of unload
here.

Tested with i386/MONOLITHIC, modified MONOLITHIC without bpf and rump.

branches: 1.43.4;
use aprint_*_dev and device_xname

branches: 1.42.16;
machine/{bus,cpu,intr}.h -> sys/{bus,cpu,intr}.h

branches: 1.41.2; 1.41.6;
Constify: LLADDR -> CLLADDR. I'm aiming here to make it easier to
identify sockaddr_dl abuse that remains in the kernel, especially
the potential for overwriting memory past the end of a sockaddr_dl
with, e.g., memcpy(LLADDR(), ...).

branches: 1.40.2; 1.40.10; 1.40.14;
Use a char* variable for doing pointer arithmetic with.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

branches: 1.38.26;
merge ktrace-lwp.

branches: 1.37.4;
nuke trailing whitespace

branches: 1.36.4; 1.36.6;
When adding/deleting multicast addresses, only whack the address
filter if the interface is marked RUNNING.

Fixes kern/27678.

branches: 1.35.2; 1.35.6;
Use a properly zero'd buffer to pad the packet to ETHER_MIN_LEN
XXX should it be ETHER_MIN_LEN-ETHER_CRC_LEN ?

Quieten lint a little.

When copying a block with an odd size to or from the card, be very careful
not to overrun the end of it on the host side, since that might just slip over
a page boundary and cause an embarrassing kernel panic.

branches: 1.32.8; 1.32.10;
don't need <sys/types.h> when including <sys/param.h>

add/cleanup RCSID

Add support for feeding entropy to rnd(4).

branches: 1.29.2; 1.29.6;
bcopy -> strcpy

bcmp -> memcmp

Add support for handling full-size 802.1Q packets.

ea_mc_reset_8004(): Rather than doing the CRC32 ourselves, call
ether_crc32_be() to do it for us.

Fix bogons in previous commit:
* Remove harmful debugging code from seeq8005_attach().
* Re-insert variable declarations in ea_mc_reset_8004(), since they aren't
dead yet.

Attampt support for 8-bit mode. I believe this should work, but I need a more
flexible bus_space on arm26 before I can test it.

No need to pull in INET or NS headers -- that's all done by ether_ioctl() now.

In ea_writembuf(), reset the DMA pointer for each mbuf we copy, since
odd-sized mbufs will confuse things. This makes transmission of odd-sized
packets work.

Add a bit of voodoo code from the old "eb" driver that seems to prevent the
80C04 returning garbage when we read from it. This driver can now handle flood
pings without dropping packets.

Tweak product ID printing in light of the 80C04A data sheet.

Split out Rx and Tx interrupt handlers into their own functions for clarity.

On an 80C04, try not to start reads at addresses of the form xx,ea, xx,ee, or
xx,f0. This appears to heavily alleviate, but not to eliminate entirely,
the problems I've been seeing with garbage being read from the rx buffer.
I suspect the real solution lies elsewhere.

Print a warning message when the DMA FIFO fails to fill or empty before we
time out.

Print the address within the Rx buffer whenever anything goes wrong: it seems
to be significant.

Clean up debugging printfs a little.

Add some extra diagnostics:
* Use the timer to timoe out transmit operations.
* Spot when the "next packet" pointer falls outside the recieve buffer and
reset the interface.
* Don't reset the interface when we get a bad packet (unless there's
something else wrong as well).

Comment and #ifdef cleanup.

Tweak handling of multicast on 80C04. I think we need to set the match mode
to "multicast" for it to work, but I could be wrong.

Also make ALLMULTI work on 80C04, but this is untested.

Re-arrange board memory to suit the current (crude) transmit code.

Add ea_writembuf() from mark's code, and re-work eatxpacket() to use it.

Enormous pile of changes from mark's (uncommitted) work on this driver.
About the only bit of his code not here is the transmit routines, which I'll
merge in separately.

Also a few bug-fixes, so (for instance) multicast on an 8005 doesn't
immediately fall back to IFF_ALLMULTI.

Change all the register definitions to match the names using in mark's
improved version of the driver.

Handle IFF_ALLMULTI correctly. We can now cope with multicast, but for
some reason DAD fails when starting IPv6.

branches: 1.8.2;
Print the newline after the memory test. This makes it clear that the delay's
related to this driver rather than the next device to be attached.

ALTQ'ify.

branches: 1.6.2;
Move bpfattach()/bpfdetach() calls into ether_ifattach()/ether_ifdetach().

First attempt at multicast stuff. I suspect it doesn't work.
Use ether_ioctl -- saves 300 bytes of text.

Move the check for "promisc + unicast + not for us" into ether_input(),
and change Ethernet drivers to always pass all received frames to
ether_input() (with a few exceptions, which are documented in the
code).

Substantial overhaul, factoring out of common code, removal of redundancy
and general cleanup. Still not a beautiful driver, but one I'd be willing to
introduce to my parents now.

Add rudimentary 80C04 support (basically saying "Oh look, an 80C04!").
Move printing the initial ":" into the board driver, like i82586.c does.
Don't bother printing the amount of RAM, as it's always 64 KB.

Split the arm26 Ether3 (ea) driver into an MI driver for the SEEQ 8005 chip,
and a front-end driver for the Ether3. Only semantic change is to remove
ea_claimirq() and ea_releaseirq() on the grounds that the seem too spurious
to warrant a callback to the front-end.

Move bpf_mtap and if_ipackets++ on Rx of each driver to percpuq if_input

The benefits of the change are:
- We can reduce codes
- We can provide the same behavior between drivers
- Where/When if_ipackets is counted up
- Note that some drivers still update packet statistics in their own
way (periodical update)
- Moved bpf_mtap run in softint
- This makes it easy to MP-ify bpf

Proposed on tech-kern and tech-net

Introduce m_set_rcvif and m_reset_rcvif

The API is used to set (or reset) a received interface of a mbuf.
They are counterpart of m_get_rcvif, which will come in another
commit, hide internal of rcvif operation, and reduce the diff of
the upcoming change.

No functional change.

Introduce softint-based if_input

This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.

This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.

To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.

Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html

Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!

Add missing splx() found by brainy.