Fix some misuse of the | operator. In particular | has higher
precedence than ?:
ok guenther@ krw@ miod@

Remove some includes include-what-you-use claims don't
have any direct symbols used. Tested for indirect use by compiling
amd64/i386/sparc64 kernels.

ok tedu@ deraadt@

remove the second void * argument on tasks.

when workqs were introduced, we provided a second argument so you
could pass a thing and some context to work on it in. there were
very few things that to

remove the second void * argument on tasks.

when workqs were introduced, we provided a second argument so you
could pass a thing and some context to work on it in. there were
very few things that took advantage of the second argument, so when
i introduced pools i suggested removing it. since tasks were meant
to replace workqs, it was requested that we keep the second argument
to make porting from workqs to tasks easier.

now that workqs are gone, i had a look at the use of the second
argument again and found only one good use of it (vdsp(4) on sparc64
if you're interested) and a tiny handful of questionable uses. the
vast majority of tasks only used a single argument. i have since
modified all tasks that used two args to only use one, so now we
can remove the second argument.

so this is a mechanical change. all tasks only passed NULL as their
second argument, so we can just remove it.

ok krw@

show more ...

be less confusing to the compiler when setting up the sge.

mark the interrupt handler mpsafe, and drop the kernel lock in the scs_cmd
paths. take it again when going back to other parts of the kernel.

tested by and ok kettenis@

remove uneeded proc.h includes
ok mpi@ kspillner@

tasks dont need a mutex and a semaphore to protect against multiple uses
of the task structure like workq tasks did.

tested on an fc929

move mpi(4) from workq to taskq

ok dlg@

Some reallocarray() use; review Jean-Philippe Ouellet, patrick keshishian
ok tedu

add a size argument to free. will be used soon, but for now default to 0.
after discussions with beck deraadt kettenis.

print the board name and firmware revision like we do on mpii and
nvme and other stuff.

use lemtohXX and htolemXX as much as possible

when reading from the chip you use letoh32, not htole32.

jmatthew@ agrees with me

use htolemXX and lemtohXX in the scsi_cmd paths.

apart from the flags handling in sgls, this shrinks the io hot path
on sparc64 and powerpc a lot. its pretty much the same on
i386/amd64/alpha.

mpi scatter gather lists are 24 bytes long and next to each other in
memory. that means you cant do 8 byte loads and stores on the sg_addr
member cos it wont be 8 byte aligned half the time which mak

mpi scatter gather lists are 24 bytes long and next to each other in
memory. that means you cant do 8 byte loads and stores on the sg_addr
member cos it wont be 8 byte aligned half the time which makes
strict alignment archs (ie, the fun ones ones) upset.

annotate the sge as being 4 byte aligned. replace the sg_addr member
with sg_addr_lo and sg_addr_hi.

replace htole64 assignment of the sg addr with a wrapper that does
the right thing with a couple of htolem32 calls.

generated code shrinks again.

show more ...

factor out the setting of the request context field so mpi_start does it
on behalf of all its callers.

use htolem32 and lemtoh32 to handle the message context.

both save bytes.

bzero to memset

bcopy to memcpy

data transfers suit STREAMING dma maps.

mikeb pointed at that the interrupt status register read in mpi_intr was
important, as per r1.167.

we dont need to read the interrupt status register to know if we have
any work to do in the interrupt handler, or to clear it. the relevant bits
indicate whether there's work on the doorbell and clea

we dont need to read the interrupt status register to know if we have
any work to do in the interrupt handler, or to clear it. the relevant bits
indicate whether there's work on the doorbell and clear when there isnt.
we need to read the doorbell if there is work to do, so lets just go that
straight away anyway.

get rid of bus_space_barriers in the io path. barriers are for enforcing
ordering. the doorbell reads and writes dont depend on any other register
values so ordering isnt applicable here.

show more ...

treat DVAs as 64bits all the time so we can avoid ugly casts and shifts in
the code.

several tweaks to make mpi(4) work for vmware emulated sas adapters.

1. vmware advertises more scsi targets than command slots, so the maths
we did for openings gave each target 0 openings. always a

several tweaks to make mpi(4) work for vmware emulated sas adapters.

1. vmware advertises more scsi targets than command slots, so the maths
we did for openings gave each target 0 openings. always advertise at least
16 openings.

2. if we cant configure the ATA queue depth, dont fail to attach the
controller whole.

finally, improve the error reporting during attach so its more obvious
where things fail.

mostly figured out by jmatthew@

show more ...

Use sg_addr instead of sg_lo_addr, leftovers from last commit.

ok dlg

Make sure we don't sleep on autoconf.

ok mikeb