branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.14;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

drop my name from the advertising clause.

branches: 1.2.10;
remove my copyright.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.14; 1.1.32; 1.1.50;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

bwi(4): Add support for Nintendo Wii WLAN.

Adapt the bwi(4) driver to support SDIO attachment and driving TX/RX using
PIO instead of DMA since the latter is not supported on SDIO busses.

branches: 1.5.36; 1.5.42;
Add driver for Broadcom 802.11a/b/g/n/ac SDIO wireless devices, based on
the OpenBSD bwfm(4) driver.

I could not test this on any hardware yet, as it does not attach as-is on
my Raspberry PI 3.

debug ld_sdmmc with LD_SDMMC_DEBUG instead of SDMMC_DEBUG

branches: 1.3.6;
Use "opt_sdmmc.h" for SDMMC_DEBUG, etc.

branches: 1.2.2; 1.2.8; 1.2.16; 1.2.20;
Remove trailing whitespace,
from KIYOHARA Takashi on current-users@

branches: 1.1.2; 1.1.4; 1.1.6;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

Prevent kernel panic when corrupted data is read
from a super frame.

Match another chip version found in RPI zero 2 W.
Fix frame parser and add more verbose checks.
Handle transmit sequence counter.

Work around transmit scheduler issue by always allowing
a send space of 1 packet. This avoids a permanent stall on
RPI zero 2 W.

There is currently no way to correctly identify the
chip used on RPI zero 2 W,

Use device-path, not fdt-path.

Fix bug in protocol parser that often caused fatal 'checksum error'.
Defer power save setting to interface start.
More verbose on errors.
Allow build without FDT.

Trailing whitespace

Add OpenBSD to code to optionally load CLM (Country Locale Matrix) files.

Use SMBIOS system product instead of the string "netbsd,generic-acpi"
for firmware loading on ACPI systems.

fix proplib deprecation

branches: 1.25.4;
Rename of_match_compat_data() to of_compatible_match(). Similarly,
rename of_search_compatible() to of_compatible_lookup().

Standardize on of_compatible_match() for driver matching, and adapt
all call sites.

Continue parsing frames after empty payload.
Add diagnostic messages.

branches: 1.23.2;
Sort includes. Nix trailing whitespace.

No functional change intended.

Need <sys/kmem.h> for kmem_*.

Currently accidentally side-loaded by <sys/pcq.h>.

clear all interrupts, not just those we expect from the hostintmask.

this removes the final hard hang i have seen in pinebookpro wifi,
though one may still need to 'ifconfig bwfm0 down up' occasionally,
so we still have bugs to fix here (the hang is usually associated
with 'checksum error' from bwfm/sdio.)

only ask for SDPCMD_INTSTATUS_HMB_SW_MASK and SDPCMD_INTSTATUS_CHIPACTIVE
interrupts, not all of them. we only ack these ones.

mostly fixes pinebookpro wifi hard hangs. still is problematic and can
trigger interrupt storm that appears as a hard hang without NET_MPSAFE,
and a follow up, less clearly right, change will reduce that to a soft
hang of the interface that can be cleared with 'ifconfig bwfm0 down up',
and even often recovers itself now.

Make this work on big endian machines

discussed with thorpej@, and it seems it's better to allocate the bss
alignment buffer statically (as part of softc) just to be sure to not
affect performance

constify, and make static where possible

Fix races in sdmmc tasks and teach ld@sdmmc to abort xfers on detach.

- Teach sdmmc_add_task to queue it only if not already queued.
- Remove now-redundant logic to avoid repeated queueing elsewhere.
- Teach sdmmc_del_task to wait until task has completed.
- Call sdmmc_del_task in various needful places.
- Replace abuse of pcq by a lock and a tailq.
(pcq is multi-producer, _single_-consumer, but there are potentially
multiple consumers here and really only one producer.)
- Teach ld_sdmmc to abort xfers on detach.
(Mechanism is kinda kludgey but it'll do for now; any effort one is
tempted to spend overhauling this should be spent overhauling sdmmc
to support proper asynchronous commands.)
- Make sure ld_sdmmc_discard either returns failure or eventually calls
ldenddiscard.

XXX Currently ld_sdmmc_detach aborts xfers _before_ ldbegindetach has
has committed to detaching or not. This is currently necessary to
avoid a deadlock because ldbegindetach waits for xfers to drain --
which strikes me as wrong; ldbegindetach shouldn't wait for anything,
and should only make the decision to commit to detaching or not so
the caller can decide whether to abort xfers before we actually wait
for them in ldenddetach.

XXX pullup -- although this changes some kernel symbols (sdmmc_add_task
and sdmmc_del_task), it shouldn't affect any existing modules; the only
module that uses sdmmc is ld_sdmmc.kmod, which is `.if 0' in the build
so there shouldn't be any of them floating around.

add entry for BCM43362, found on Cubietruck
ok jmcneill@

- Refactor the firmware loading code into a set of common routines in the
core driver that understand how to find model-specific files before the
generic files. This greatly simplifies the firmware loading procedure
in each of the bus front ends. Adding additional firmware file types
(such as CLM files) much simpler.
- Make firmware image selection table-driven rather than a set of switch()
and if() statements.
- Add several additional firmware image entries.

Adopt <net/if_stats.h>.

branches: 1.12.2;
size check was backwards.

Use correct firmware for BCM43456

Follow the Linux driver an use the FDT "compatible" property to build a
filename for the nvram config file, fall back to the standard filename.

E.g.

# ofctl -p /
[Caching 123 nodes and 1093 properties]
#address-cells 00000001 ........ ........ ........ 1
#size-cells 00000001 ........ ........ ........ 1
compatible 73696e6f 766f6970 2c627069 2d6d322d "sinovoip,bpi-m2-
0010: 7a65726f 00...... ........ ........ zero"
0015: 616c6c77 696e6e65 722c7375 6e38692d "allwinner,sun8i-
0025: 68322d70 6c757300 ........ ........ h2-plus"
interrupt-parent 00000001 ........ ........ ........ ....
model 42616e61 6e612050 69204250 492d4d32 "Banana Pi BPI-M2
0010: 2d5a6572 6f00.... ........ ........ -Zero"
name 00...... ........ ........ ........ ""
serial-number 30326330 30303432 65636431 36376566 02c00042ecd167ef
0010: 00...... ........ ........ ........ .


-rw-r--r-- 1 root wheel 875 Nov 2 12:06 brcmfmac43430-sdio.AP6212.txt
lrwxr-xr-x 1 root wheel 29 Dec 30 16:19 brcmfmac43430-sdio.sinovoip,bpi-m2-zero.txt -> brcmfmac43430-sdio.AP6212.txt
-rw-r--r-- 1 root wheel 874 Jun 30 2019 brcmfmac43430-sdio.raspberrypi,3-model-b.txt
-rw-r--r-- 1 root wheel 1864 Jun 30 2019 brcmfmac43455-sdio.raspberrypi,3-model-b-plus.txt
lrwxr-xr-x 1 root wheel 29 Dec 30 11:24 brcmfmac43455-sdio.raspberrypi,4-model-b-plus.txt -> brcmfmac43455-sdio.raspberrypi,3-model-b-plus.txt

More code from OpenBSD
no need to splnet() when enqueing packets
explicit structure padding
make internal functions static

also prepare for GPIO interrupts.

Simplyfy sdmmc_io_set_blocklen function signature by dropping the
extra softc pointer. Aligns with OpenBSD.

white space police.

Use correct function to verify if a task has been queued. Avoids race
that can corrupt the task queue.

Don't pass empty mbufs to the network stack.

From OpenBSD:
- All the missing pieces (firmware load, chip setup, protocol handling)
TX queue and interrupt handling via sdmmc_task.
Fix locking.
Fix packet parsing.
Add parser for original firmware config files.

branches: 1.3.2; 1.3.8;
sync with openbsd bwfm to some extent.

add a txcheck
set chip active/passive for more kinds of chips
add wrapper around setting active/passive
detect chip RAM
make bwfm_rx take an mbuf

Outsource setting the backplane window into a specific function so it
can be called and reused in different places.

From OpenBSD.

branches: 1.1.2; 1.1.4;
Add driver for Broadcom 802.11a/b/g/n/ac SDIO wireless devices, based on
the OpenBSD bwfm(4) driver.

I could not test this on any hardware yet, as it does not attach as-is on
my Raspberry PI 3.

Fixup bwfm(4) register bit regarding SDIO device suspend/resume.

From OpenBSD if_bwfm_sdio.h r1.3

branches: 1.2.4; 1.2.10;
More register definitions.

From OpenBSD:
- All the missing pieces (firmware load, chip setup, protocol handling)
TX queue and interrupt handling via sdmmc_task.
Fix locking.
Fix packet parsing.
Add parser for original firmware config files.

branches: 1.1.2; 1.1.6;
bwi(4): Add support for Nintendo Wii WLAN.

Adapt the bwi(4) driver to support SDIO attachment and driving TX/RX using
PIO instead of DMA since the latter is not supported on SDIO busses.

ld(4): Convert blkno argument for sc_dump() to daddr_t

PR kern/59153

(1) For backends that accept 64-bit block address, i.e.,
nvme(4), virtio(4), aac(4), iop(4), and mainbus(usermode/4),
this should enable to dump beyond 2Gi blocks.

(2) sdmmc(4) backend allows to dump up to the last block.

(3) For other backends, block address is handled as `int`.
Some of them may support blocks up to 4Gi, but I do not have
enough time to examine datasheets. So, continue to reject >2Gi
blocks as before.

XXX
This is KABI change, and cannot be pulled up into netbsd-{10,9}.

XXX
Compile-test only (for amd64/ALL) due to lack of large SSDs ;)

Thanks mlelstv@ for discussion and careful review!!

sdmmc: Add support for SD card caches.

SD physical 6.0 specification introduced Application Performance Class 2
(A2), which adds support for drive caches and command queueing.

Add support for enabling and flushing the cache when this feature is
present.

branches: 1.43.2;
ld@sdmmc(4): Hack around deadlock in cache sync on detach.

Yanking a card triggers the sdmmc discovery task, which runs in the
sdmmc task thread, to detach any attached child devices.

Detaching ld@sdmmc triggers a cache flush (via ldbegindetach ->
disk_begindetach -> ld_lastclose -> ld_flush -> ioctl DIOCCACHESYNC),
which is implemented by scheduling a task to do sdmmc_mem_flush_cache
and then waiting for it to complete.

The sdmmc_mem_cache_flush is done by an sdmmc task so it happens
after all previously scheduled I/O operations -- that way the cache
flush doesn't complete until the previously scheduled I/O operations
are complete.

However, when the cache flush task is issued from the discovery task,
this doesn't work, because the cache flush task can't start until the
discovery task has returned -- but the discovery task won't return
until the cache flush task has completed.

To work around this deadlock, which usually happens only when the
device has been yanked anyway so further I/O would be lost anyway,
just do the cache flush synchronously in DIOCCACHESYNC if we're
running in the task thread.

This isn't quite right -- implementation details of the task thread
shouldn't bleed into ld@sdmmc, and running the cache sync _before_
any subsequently scheduled I/O tasks is asking for trouble -- but it
should serve to avoid the deadlock in PR kern/57870 until we can fix
a host of concurrency bugs in sdmmc by fixing the locking scheme and
running discovery in a separate thread from tasks.

XXX pullup-10

branches: 1.42.4;
sdmmc: ld: Include the card type (SD card, MMC) in the disk description.

Remove unnecessary wait in ldbegindetach.

Like disk_begindetach, ldbegindetach only commits to detaching but
doesn't wait for existing xfers to drain; it is up to the driver to
abort them, once we are committed, and then ldenddetach to wait for
them to drain.

Sort #includes. Nix trailing whitespace.

No functional change intended.

Omit needless <sys/pcq.h>.

Fix races in sdmmc tasks and teach ld@sdmmc to abort xfers on detach.

- Teach sdmmc_add_task to queue it only if not already queued.
- Remove now-redundant logic to avoid repeated queueing elsewhere.
- Teach sdmmc_del_task to wait until task has completed.
- Call sdmmc_del_task in various needful places.
- Replace abuse of pcq by a lock and a tailq.
(pcq is multi-producer, _single_-consumer, but there are potentially
multiple consumers here and really only one producer.)
- Teach ld_sdmmc to abort xfers on detach.
(Mechanism is kinda kludgey but it'll do for now; any effort one is
tempted to spend overhauling this should be spent overhauling sdmmc
to support proper asynchronous commands.)
- Make sure ld_sdmmc_discard either returns failure or eventually calls
ldenddiscard.

XXX Currently ld_sdmmc_detach aborts xfers _before_ ldbegindetach has
has committed to detaching or not. This is currently necessary to
avoid a deadlock because ldbegindetach waits for xfers to drain --
which strikes me as wrong; ldbegindetach shouldn't wait for anything,
and should only make the decision to commit to detaching or not so
the caller can decide whether to abort xfers before we actually wait
for them in ldenddetach.

XXX pullup -- although this changes some kernel symbols (sdmmc_add_task
and sdmmc_del_task), it shouldn't affect any existing modules; the only
module that uses sdmmc is ld_sdmmc.kmod, which is `.if 0' in the build
so there shouldn't be any of them floating around.

Whitespace police

branches: 1.36.4;
Provide CID data (manufacturer/product id/product name) as disk type.

Defer DIOCCACHESYNC to the sdmmc task queue so they are serialized with other requests.

branches: 1.34.2; 1.34.4;
make the sc_discard interface for the ld backend asynchronous and
signal completion through new callback lddiscardend. Use a standard
struct buf to pass disk address and range instead of two off_t values.

make lddiscard synchronous again. This is a requirement of the current
ffs discard code.

Defer sdmmc discard operations to the sdmmc task queue. Fixes a panic
introduced by ld.c r1.102.

While ld(4) is MP safe, many backends are not.

Add a flag for backends that are MP safe. Take KERNEL_LOCK when calling
into a backend that doesn't have the flag set. Do the same for the
discard routine.

Fixes PR 52462.

branches: 1.31.2;
Add support for eMMC 4.5's optional cache feature. If a cache is present,
and the host controller reports the SMC_CAPS_POLLING capability (needed
to flush cache at shutdown), it will be automatically enabled and used.

Decrypt comment about the attach thread. No functional change.

The config_* APIs are not MPSAFE, so make sure the deferred attach
thread holds KERNEL_LOCK.

Add discard support to ld@sdmmc using the ERASE (CMD38) command.

Fix a race between ld_sdmmc_start and ld_sdmmc_dobio that could result in
tasks getting lost from the task queue. The symptom of this is a NULL
deref in ld_sdmmc_start since the code assumes that a task will always be
available from the pool.

This changes the code to use pcq(9) instead of a TAILQ to manage the free
task list.

branches: 1.26.4;
Ensure that the task is returned to the free task list in an error branch.

Add missing includes

Support retry when error. like wd(4).

Modularize the ld driver and all of its attachments. Ensure that all
parents are capable of rescan (or otherwise provide a means of attaching
children post-initialization).

modify ldattach() to have default strategy as a parameter

branches: 1.21.2;
remove unused include sys/rndsource.h

Add small command queue to reduce latency between ld driver and
sdmmc I/O thread.

Print the selected transfer mode if known.

use mutex locking for MP safety.

Don't timeout waiting for ld_sdmmc_dobio in the IO thread to start.

XXX why does it take >1s to get scheduled?

need to call lddone with splbio and kernel lock held

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

branches: 1.14.2;
debug ld_sdmmc with LD_SDMMC_DEBUG instead of SDMMC_DEBUG

Be a bit more verbose about errors, also pass through error code
from lower layer like other drivers.

branches: 1.12.4;
Pass the device name in, so we can debug what deferred drivers did not work.

branches: 1.11.2;
Display the whole CID register upon attach as "<MID:OID:PNM:PRV:PSN:MDT>"
instead of just "<PNM>". Probably should figure out a way to stuff this
into the device properties database too, but not everything has
pseudo-device drvctl enabled.

Also, drop trailing whitespace.

branches: 1.10.6;
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.

Use "opt_sdmmc.h" for SDMMC_DEBUG, etc.

Protect ldattach ops with config_pending_intr(9)/config_pending_decr(9)
so that "boot device: ld0" detection works properly even on slower zaurus.

branches: 1.7.8; 1.7.12;
Don't pull in the whole uvm(9) API to access only PAGE_SIZE and
some other constants. These are provided by sys/param.h now.

Print bus width and clock for feature(High speed support).

Print product name. like sd(4), wd(4) and etc...

branches: 1.4.2; 1.4.4; 1.4.6;
Always sector size is treated as 512 bytes.

branches: 1.3.2;
It is avoided that the error occurs when the card attaches it,
when wedge is supported.

branches: 1.2.2;
struct cfdata * -> cfdata_t, no functional changes intended.

branches: 1.1.2; 1.1.4;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sys: Drop redundant NULL check before m_freem(9)

m_freem(9) safely has accepted NULL argument at least since 4.2BSD:
https://www.tuhs.org/cgi-bin/utree.pl?file=4.2BSD/usr/src/sys/sys/uipc_mbuf.c

Compile-tested on amd64/ALL.

Suggested by knakahara@

branches: 1.8.32;
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.

remove M_CANFAIL flag for malloc(9) - it was completely ignored, so had
actually no effect

branches: 1.6.2;
m_free -> m_freem, m_copyback could have added mbufs in the chain

branches: 1.5.10; 1.5.16;
- Use aprint*() instead of printf() in xxx_attach().
- Add missing aprint_naive("\n");
- KNF

branches: 1.4.4;
netbt: rename some attach/detach functions to have _pcb suffix, so
we could use standard attach/detach naming for pr_usrreq functions.
No functional change.

branches: 1.3.10;
split device_t/softc for all remaining drivers.
replace "struct device *" with "device_t".
use device_xname(), device_unit(), etc.

branches: 1.2.2; 1.2.4; 1.2.10; 1.2.18; 1.2.28;
struct cfdata * -> cfdata_t, no functional changes intended.

branches: 1.1.2; 1.1.4;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

Use the SCF_NO_STOP flag to disable AUTO_CMD12_ENABLE
instead of checking the SD_IO_RW_EXTENDED command op code.

Also add this logic to sunxi_mmc.

sdhc: Do not set auto cmd12 enable for SDIO RW extended ops

sdhc: Disable bus power for 32-bit register accesses.

SDHC_FLAG_32BIT_ACCESS shouldn't prohibit clearing the SD Bus Power bit in
the Power Control Register, that's what SDHC_FLAG_NO_PWR0 is for.

Should help with card initialization issues at boot on controllers that
set the 32-bit access flag.

branches: 1.121.2;
sdhc: Remove delay in sdhc_wait_state

Replace the 10us delay loop in sdhc_wait_state with a microuptime based
timeout test.

sdhc: Remove now unused sc_write_delay

branches: 1.119.2;
If the ACPI flag "non-removable" is present and true for an SD/MMC
slot, then set a new flag on that slot to indicate that the media
is non-removable. Make sdhc_card_detect always return true for a
slot if its non-removable media flag is set.

This change lets the kernel automatically configure the
permanently-installed MMC slot on the NXP LX2160-based HoneyComb
LX2 board.

sdmmc: add support for optional delay after register write

branches: 1.117.2;
Select DMA mode after programming the ADMA base address register(s).

The Arasan SDHCI 8.9a found in the Xilinx Zynq-7000 SoC requires this
sequence to avoid sporadic transfer errors.

Disable ADMA2 on Ricoh SDHCI controllers.

PR# 57015

sdhc: Retire SDHC_FLAG_USE_ADMA2 flag.

ADMA2 support in sdhc is mature now, so no need for it to be opt-in.

sdhc: avoid taking adaptive mutex while holding spin mutex

the clock_bus pre- and post- callbacks used in sdhc_bus_clock_ddr()
are called with a spin mutex held, and the new sdhc@acpi ends up
calling into (sleeping) ACPI code, triggering lockdebug assertion.

introduce an adaptive mutex that is held around these callbacks,
and reduce the spin mutex held time in sdhc_bus_clock_ddr().

sdhc: High speed support capability flag applies to eMMC too.

s/suport/support/ in comment.

Merge thorpej-cfargs2.

branches: 1.110.4;
Use unsigned to avoid undefined behavior in hwrite[12](). Found by kUBSan.

branches: 1.109.2; 1.109.4;
Merge thorpej-cfargs branch:

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

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

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

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

...and a sentinel value CFARG_EOL.

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

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

branches: 1.108.2;
Enable block count only for count > 0.
Don't enable autostop when command sets new flag SCF_NO_STOP.

branches: 1.107.2;
Identify SDHC 4.1 and 4.2. From {DragonFly,Free}BSD.

Avoid warnings for tautological shifts as sole conditional.

Skip setting power when the voltage doesn't change.
Also increase some timeouts.

Add SDHC flags.

+ SDHC_FLAG_BROKEN_ADMA2_ZEROLEN
Broken ADMA2 Zero length descriptor.
Can't 64K Byte data transfer.
+ SDHC_FLAG_NO_1_8_V
Support no 1.8V Supply.
Disable UHS-I bus speed mode (SDR50, DDR50, SDR104).

branches: 1.103.2;
If switching to fixed sampling clock, do not return an error to the sdmmc layer.

Add vendor callback for post-bus clock ops and add SDHC_FLAG_STOP_WITH_TC flag

branches: 1.101.6;
fix problem for ESDHC/USDHC due to change of r1.96

on ESDHC/USDHC, even if the iosize is less than SDHC_HOST_CTL_VERSION,
specver must be an appropriate value.

branches: 1.100.4;
Add a vendor callback for setting signal voltage.

sdhc(4): hardware reset support for Intel eMMC controller

sdhc(4), sdmmc(4): Added MMC HS DDR52 support.

branches: 1.97.2;
Add flags SDHC_FLAG_NO_AUTO_STOP and SDHC_FLAG_NO_BUSY_INTR.

Don't access SDHC_HOST_CTL_VERSION, if iosize less than this.

Use 1.65-1.95 voltage window for 1.8V support.

branches: 1.94.2;
Please more delay, if timeout in sdhc_wait_intr().

add support iMX6 uSDHC
- some UHS-I/SDR104 card are not stable
- eMMC doesn't work yet

fix typo. use pmask for SDHC_PRESENT_STATE.

Also wait for DAT when sending RSP_BSY commands. Fixes PR 50388.

fix timeout path in soft reset when building with SDHC_DEBUG

allow vendor specific code to hook into bus width changes

support hiding command timeout messages with a new command flag and use this
when probing for cards. Should fix PR 50302.

The iMX uSDHC controller doesn't have an interrupt error status flag,
the bit position is reserved and reads as value 0.
- Fake the flag if any of the error bits is set.
- uSDHC supports the 32bit access, the 16bit path doesn't need that quirk.

Don't test flags from error value in status value. Instead test the
error interrupt status.

fix DPRINTF parameters

Ignore data inhibit for commands that do not use the DAT line.
Do a soft reset when the inhibit condition persists for better
error recovery.

Simplify interrupt handling and print errors reported by the controller.

Add more specific debug messages for timeout errors.

Protect the whole tuning operation including the register setup.

Add a quirk for Ricoh 5U823 controller. Operation with a 100MHz
bus clock for SDR50 seems to be unstable, reduce frequency one notch
(effectively down to 66MHz with divisor = 3).

don't hold intr_lock while calling sdhc_execute_tuning

support re-tuning modes 1 and 2

Implement SDHC sampling clock tuning procedure.

lock more hardware accesses

Add SDHC_FLAG_POLL_CARD_DET flag which lets the bus glue request polling
instead of standard SDHC card insert / detach events.

Add support for DDR50 transfer modes.

better error message

don't select SDR12 mode if we are <= 400 kHz

dont confuse signaling voltage and bus voltage

let SDHC_FLAG_HOSTCAPS override SDHC_CAPABILITIES2 as well

add support for UHS-I modes on capable 3.0+ controllers

bus_dma_segment_t ds_addr is bus_addr_t not paddr_t (another case of it)

bus_dma_segment_t ds_addr is bus_addr_t not paddr_t

Be more verbose about error conditions.

correct (the commented out) calculation of bus frequency.

return MMC_OCR_HCS bit from host_ocr if the controller supports high-speed mode

Simplify sdhc(4) locking

print "platform DMA" instead of "SDMA" when using external DMA engine

Add ADMA2 support, which enables scatter gather DMA for data transfers on
controllers that support it.

Remove unnecessary #include

Add a SDHC_FLAG_NO_TIMEOUT quirk to handle spurious timeouts on Tegra K1
during data transfers. While here, increase the soft timeout for DMA
transfers from 1s to 3s.

Use C99 designated initializers.

No functional change.

support 8-bit mode for SDHC 3.0+ controllers

print some useful information at attach time

Add SDHC_FLAG_SINGLE_POWER_WRITE flag, that tells the driver to update
the SDHC_POWER_CTL register with a single write rather than in multiple
steps. Required for Tegra K1 SDHC.

support SDHC 4.0, allow for overriding clk base frequency

Add a new SDHC_FLAG, SDHC_FLAG_EXTDMA_DMAEN, which request that the
SDHC_DMA_ENABLE bit be set in the command, even if we're using an
external DMA engine. Needed by the upcoming DMA support for AM335x
(beaglebone).

eSDHC has no DMA_BOUNDARY in BLOCK_SIZE register.
So clear multi segment DMA transfer support bit.

Don't clear other bits. pq3sdhc is set more bits.

eSDHC has non standard Host Controller Version Register offset.

branches: 1.51.2;
pass sdhc_softc instead of sdhc_host to sc_vendor_transfer_data_dma, since the vendor specific code has no visibility into struct sdhc_host and can easily do device_private(sdhc_softc->sc_dev)

Fix divisor calculation for SDHC 3.0.
1/1 -> 0
1/2 -> 1
1/4 -> 2
...
1/2*N -> N

correct clk div calculation for SD 3.0 controllers

Don't check slot type for pre-3.0 controllers, the bits were reserved at
the time. If an older controller advertises 1.8V support, just use it.
This fixes high-speed mode support for me on BeagleBone Black.

When doing vendor dma transfer wait for SDHC_TRANSFER_COMPLETE.

From jmcneill@ with some layout changes from me.

Avoid NULL deref when sdhc(4) attaches via config_interrupts().

Allow sdhc(4) to use a non-integrated DMA engine.

branches: 1.44.2;
eSDHC Controller doesn't have "Host SDMA Buffer Boundary" bit field

branches: 1.43.10;
add SDHC_FLAG_NO_HS_BIT flag to prevent sdhc driver from setting the SDHC_HIGH_SPEED bit in SDHC_HOST_CTL, required to get many high speed cards working on rpi

If the command protocol calls for a trailing busy signal,
wait (up to 10 seconds) for it to be released.

Find correct slot index in sdhc_cfprint().

SDHC_VOLTAGE_SUPP_1_8V is only valid in the presence of SDHC_EMBEDDED_SLOT.
Fixes probing of full-size 1GB Toshiba card made in 2008 that fails to
respond with a 1.8V supply. (JMB389 (in TS-RDF1) host claims
SDHC_VOLTAGE_SUPP_1_8V but does not claim SDHC_EMBEDDED_SLOT.)

SMC_CAPS_DMA without SMC_CAPS_MULTI_SEG_DMA means we as a host driver only
can support one segment per operation. This is not the case with SDHCI SDMA,
the only SDHCI DMA mechanism we currently support in sdhc(4). Should fix
corrupt data on DMA-enabled SDHCI controllers claiming conformance to a SDHCI
Specification other than 1.0.

Add support for SDHCI 3.0 bus clock generation, which unlike 1.0 and 2.0
has support for non-power-of-two divisors from 1 to 1023. Additionally
restore finding the correct power-of-two divisor for standard-compliant
pre-3.0 SDHCI parts. Also, calculate minimum divided clock for the
clkmsk-quirk parts correctly.

XXX Will future standard-compliant SDHCI versions have a SDHCI 3.0
backwards-compatible clock generator, or just SDHCI 1.0/2.0?

If we get a SDMA boundary-crossing interrupt at the end of the final
segment, the transfer is not implicitly complete; we still need to wait
for the transfer complete interrupt so that we can issue new commands
successfully.

This fixes the problem my Ricoh R5C822 was having where all writes after
a larger (64KiB) write were failing because we attempted to issue CMD13
(to check the status of the aformentioned write) too soon afterwards.

I suspect this fix to apply to other Ricoh SDHCI controllers I've heard
people have had trouble with. Sorry I didn't look at this sooner.

Change sdhc_detach so that it detaches all host controllers at once.
This should make multiple slot controllers, for example those with
a controller on more than one PCI/CardBus BAR, detach with fewer bugs.
Tested with as-of-yet-uncommited sdhc_pci changes on a single-host
ExpressCard JMicron JMB38[89].

Implement TI AM335x's SDHC reset quirk. Beaglebone SDHC works now!

On the AM335x, we first must wait for the controller to acknowledge
the reset; then we can wait for the reset to complete.

I believe this quirk also applies to the OMAP4 ES, but I don't have
one of those to test and we don't seem to have an obvious conditional
for it anyway.

This quirk may work for controllers that don't require it too, but I
am nervous about doing it by default because if we miss the reset
acknowledgement, then we'll just time out even though everything is
really hunky-dory.

Also, for all sdhc, don't bother writing 0 in sdhc_soft_reset while
waiting for the reset to complete; there is no need.

ok matt

Add sc_vendor_bus_clock hook
From jmcneill.

Avoid the 16-bit read in sdhc_host_found for 32-bit-only devices.

This change makes sdhc_host_found allocate hp up front before showing
the identification, in order to avoid having to open-code a copy of
HREAD2 before hp is available.

Support omap3 SDHC driver. tested on OVERO only.

Don't use const foo const as type, one const is enough.

branches: 1.30.2;
Support deciphering SDHCv3 clock rates.

Make PIO on normal SDHC devices work correctly on big-endian machines.
Add locking around interrupt manipulation (it should now be MP safe).

Only SDHC 1.0 controllers can do that fucky MULTI_SEG stuff.

sdhc(4) currently only implements support for the SDHC 1.0 DMA mechanism, so
only enable in that case.

MULTI_SEG does not for ESDHC. (Don't know how it works at all on any SDHC
but ...)

Responses are actually in host order (except SCR which is return in
big endian so that's convert to host order).

Provide a method for attachments to specify capabilites.

Add use of watermark register when PIO to an ESDHC. After every kill or
drain of watermask words, pause a bit to give time for the fifo to recover.
Always the command response in BE byteorder. Rewrite __bitfield to deal
with this.

If there was an error in 32-bit mode, just set ERROR_INTERRUPT otherwise
see if matched anything we care about.

Handle interrupt acknowledgement in the SDHC_FLAG_32BIT_ACCESS case in
the same way as non-SDHC_FLAG_32BIT_ACCESS case.

bus_space_*_stream_N() functions are not universally available.
Provite alternate implementation for when they are unavailable.

Completely rework (and enable) SDHC 1.0 DMA data transfer.

While without a large physically-contiguous buffer the performance
suffers severly, this should still be better than PIO.

No need to panic on failure to find a divisor.
Additionally, add commented out calculation of actual clock frequency.

SDHC can have multiple slots, note which slot a specific sdmmc(4) attaches to.

SDHCI byte swaps the BE response on the wire into LE registers.
As we always want response data in LE, use bus_space_read_stream.
Additonally, read response data in 1 or 4 4-byte chunks, instead of
one 4-byte chunk or 15 1-byte chunks.

Always enable and set block count register. We never do infinite transfers.
The old way seemed to be causing issues on a R5C822, at least with DMA enabled.

Add a flag for the lack of LED_ON in HOST_CTL (ti omap3 doesn't do that).

Do not read past array end, found by gcc -O3.
This could cause to HWRITE4() a bad value, but maybe last 2 bytes are
probably ignored by hardware anyway.

Added Ricoh 5U822/5U823 SD/MMC Controller support.

Support 32-bit only access to the SDHC registers.
Add support for FreeScale "Enhanced" SDHC port.
Add support for CGM mode (XLP and BCM2835 (Arason)).

branches: 1.10.2;
Use "opt_sdmmc.h" for SDHC_DEBUG.

branches: 1.9.2; 1.9.6;
Add some inital changes for the Freescale eSDHC.

Support change Open-drain/Push-pull by bus_rod().

branches: 1.7.2;
Added sdhc at cardbus support.

branches: 1.6.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.5.2;
According to SD Host Controller Simplified Specification Version 2.00, 2.2.10.
Host Control Register (Offset 028h), the "Data Transfer Width" bit is in Host
Control Register (Offset 028h), not Power Control Register (Offset 029h).

Expand PMF_FN_* macros.

branches: 1.3.2;
Typo in comment.

fix crash with LOCKDEBUG+DIAGNOSTIC kernels:

sdhc_bus_clock() calls mutex_enter().
then in DIAGNOSTIC section, sdhc_bus_clock() calls sdhc_card_detect().
sdhc_card_detect() calls mutex_enter() on the same mutex again.
when LOCKDEBUG is enabled, you get

Mutex error: lockdebug_wantlock: locking against myself

branches: 1.1.2; 1.1.4; 1.1.6;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

Identify SDHC 4.1 and 4.2. From {DragonFly,Free}BSD.

Add SDHC flags.

+ SDHC_FLAG_BROKEN_ADMA2_ZEROLEN
Broken ADMA2 Zero length descriptor.
Can't 64K Byte data transfer.
+ SDHC_FLAG_NO_1_8_V
Support no 1.8V Supply.
Disable UHS-I bus speed mode (SDR50, DDR50, SDR104).

branches: 1.19.6; 1.19.12;
fix problem for ESDHC/USDHC due to change of r1.96

on ESDHC/USDHC, even if the iosize is less than SDHC_HOST_CTL_VERSION,
specver must be an appropriate value.

branches: 1.18.10;
add support iMX6 uSDHC
- some UHS-I/SDR104 card are not stable
- eMMC doesn't work yet

support re-tuning modes 1 and 2

Implement SDHC sampling clock tuning procedure.

add support for UHS-I modes on capable 3.0+ controllers

Add ADMA2 support, which enables scatter gather DMA for data transfers on
controllers that support it.

support SDHC 4.0, allow for overriding clk base frequency

eSDHC has non standard Host Controller Version Register offset.

branches: 1.11.12; 1.11.14;
Add a few more register bits from SDHCI 3.0.
Also, add comment to denote an ESDHC bit.

Make whitespace following preprocessor keywords consistent
with the predominating form in this file (that is, a space).

branches: 1.9.2;
Support deciphering SDHCv3 clock rates.

Add use of watermark register when PIO to an ESDHC. After every kill or
drain of watermask words, pause a bit to give time for the fifo to recover.
Always the command response in BE byteorder. Rewrite __bitfield to deal
with this.

Add DMA boundary field shift and mask.

Added Ricoh 5U822/5U823 SD/MMC Controller support.

branches: 1.5.2;
Add XLP specific CGM mode for SDHC_CLOCK_CTL.

Add some more FreeScale ESDHC defintions.

branches: 1.3.2; 1.3.6;
Add some inital changes for the Freescale eSDHC.

Add stuff that the Freescale PowerQUICC3 ESDHC controllers have.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.12; 1.1.14; 1.1.16;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sdhc: Remove now unused sc_write_delay

branches: 1.35.2;
If the ACPI flag "non-removable" is present and true for an SD/MMC
slot, then set a new flag on that slot to indicate that the media
is non-removable. Make sdhc_card_detect always return true for a
slot if its non-removable media flag is set.

This change lets the kernel automatically configure the
permanently-installed MMC slot on the NXP LX2160-based HoneyComb
LX2 board.

sdmmc: add support for optional delay after register write

branches: 1.33.2;
Disable ADMA2 on Ricoh SDHCI controllers.

PR# 57015

sdhc: Retire SDHC_FLAG_USE_ADMA2 flag.

ADMA2 support in sdhc is mature now, so no need for it to be opt-in.

Add SDHC flags.

+ SDHC_FLAG_BROKEN_ADMA2_ZEROLEN
Broken ADMA2 Zero length descriptor.
Can't 64K Byte data transfer.
+ SDHC_FLAG_NO_1_8_V
Support no 1.8V Supply.
Disable UHS-I bus speed mode (SDR50, DDR50, SDR104).

Add vendor callback for post-bus clock ops and add SDHC_FLAG_STOP_WITH_TC flag

branches: 1.29.12;
Add a vendor callback for setting signal voltage.

sdhc(4): hardware reset support for Intel eMMC controller

branches: 1.27.2;
Add flags SDHC_FLAG_NO_AUTO_STOP and SDHC_FLAG_NO_BUSY_INTR.

branches: 1.26.2;
add support iMX6 uSDHC
- some UHS-I/SDR104 card are not stable
- eMMC doesn't work yet

allow vendor specific code to hook into bus width changes

Add a quirk for Ricoh 5U823 controller. Operation with a 100MHz
bus clock for SDR50 seems to be unstable, reduce frequency one notch
(effectively down to 66MHz with divisor = 3).

Add SDHC_FLAG_POLL_CARD_DET flag which lets the bus glue request polling
instead of standard SDHC card insert / detach events.

let SDHC_FLAG_HOSTCAPS override SDHC_CAPABILITIES2 as well

Simplify sdhc(4) locking

Add ADMA2 support, which enables scatter gather DMA for data transfers on
controllers that support it.

Add a SDHC_FLAG_NO_TIMEOUT quirk to handle spurious timeouts on Tegra K1
during data transfers. While here, increase the soft timeout for DMA
transfers from 1s to 3s.

Add SDHC_FLAG_SINGLE_POWER_WRITE flag, that tells the driver to update
the SDHC_POWER_CTL register with a single write rather than in multiple
steps. Required for Tegra K1 SDHC.

support SDHC 4.0, allow for overriding clk base frequency

Add a new SDHC_FLAG, SDHC_FLAG_EXTDMA_DMAEN, which request that the
SDHC_DMA_ENABLE bit be set in the command, even if we're using an
external DMA engine. Needed by the upcoming DMA support for AM335x
(beaglebone).

branches: 1.15.2;
pass sdhc_softc instead of sdhc_host to sc_vendor_transfer_data_dma, since the vendor specific code has no visibility into struct sdhc_host and can easily do device_private(sdhc_softc->sc_dev)

Allow sdhc(4) to use a non-integrated DMA engine.

branches: 1.13.12;
add SDHC_FLAG_NO_HS_BIT flag to prevent sdhc driver from setting the SDHC_HIGH_SPEED bit in SDHC_HOST_CTL, required to get many high speed cards working on rpi

Change sdhc_detach so that it detaches all host controllers at once.
This should make multiple slot controllers, for example those with
a controller on more than one PCI/CardBus BAR, detach with fewer bugs.
Tested with as-of-yet-uncommited sdhc_pci changes on a single-host
ExpressCard JMicron JMB38[89].

Implement TI AM335x's SDHC reset quirk. Beaglebone SDHC works now!

On the AM335x, we first must wait for the controller to acknowledge
the reset; then we can wait for the reset to complete.

I believe this quirk also applies to the OMAP4 ES, but I don't have
one of those to test and we don't seem to have an obvious conditional
for it anyway.

This quirk may work for controllers that don't require it too, but I
am nervous about doing it by default because if we miss the reset
acknowledgement, then we'll just time out even though everything is
really hunky-dory.

Also, for all sdhc, don't bother writing 0 in sdhc_soft_reset while
waiting for the reset to complete; there is no need.

ok matt

Add sc_vendor_bus_clock hook
From jmcneill.

Support omap3 SDHC driver. tested on OVERO only.

branches: 1.8.2;
Provide a method for attachments to specify capabilites.

Add a flag for the lack of LED_ON in HOST_CTL (ti omap3 doesn't do that).

branches: 1.6.2;
Add more FLAGS

branches: 1.5.2; 1.5.6;
Add some inital changes for the Freescale eSDHC.

branches: 1.4.2;
Added sdhc at cardbus support.

branches: 1.3.2;
A pointer typedef entails trading too much flexibility to declare const
and non-const types, and the kernel uses both const and non-const
PMF qualifiers and device suspensors, so change the pmf_qual_t and
device_suspensor_t typedefs from "pointers to const" to non-pointer,
non-const types.

branches: 1.2.2;
Expand PMF_FN_* macros.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sdmmc.c: add missing newlines to several device_printf(9) calls

branches: 1.43.12;
Merge thorpej-cfargs2.

branches: 1.42.2;
Don't crash on detach where interlock == NULL.

branches: 1.41.2;
Merge thorpej-cfargs branch:

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

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

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

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

...and a sentinel value CFARG_EOL.

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

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

branches: 1.40.4;
Fix races in sdmmc tasks and teach ld@sdmmc to abort xfers on detach.

- Teach sdmmc_add_task to queue it only if not already queued.
- Remove now-redundant logic to avoid repeated queueing elsewhere.
- Teach sdmmc_del_task to wait until task has completed.
- Call sdmmc_del_task in various needful places.
- Replace abuse of pcq by a lock and a tailq.
(pcq is multi-producer, _single_-consumer, but there are potentially
multiple consumers here and really only one producer.)
- Teach ld_sdmmc to abort xfers on detach.
(Mechanism is kinda kludgey but it'll do for now; any effort one is
tempted to spend overhauling this should be spent overhauling sdmmc
to support proper asynchronous commands.)
- Make sure ld_sdmmc_discard either returns failure or eventually calls
ldenddiscard.

XXX Currently ld_sdmmc_detach aborts xfers _before_ ldbegindetach has
has committed to detaching or not. This is currently necessary to
avoid a deadlock because ldbegindetach waits for xfers to drain --
which strikes me as wrong; ldbegindetach shouldn't wait for anything,
and should only make the decision to commit to detaching or not so
the caller can decide whether to abort xfers before we actually wait
for them in ldenddetach.

XXX pullup -- although this changes some kernel symbols (sdmmc_add_task
and sdmmc_del_task), it shouldn't affect any existing modules; the only
module that uses sdmmc is ld_sdmmc.kmod, which is `.if 0' in the build
so there shouldn't be any of them floating around.

Add and use sdmmc_pause to avoid long-term busy waits.
Add sdio abort function.
Additional error messages.
Print parameters for SDIO devices.
Minor cosmetics.

Add SDHC flags.

+ SDHC_FLAG_BROKEN_ADMA2_ZEROLEN
Broken ADMA2 Zero length descriptor.
Can't 64K Byte data transfer.
+ SDHC_FLAG_NO_1_8_V
Support no 1.8V Supply.
Disable UHS-I bus speed mode (SDR50, DDR50, SDR104).

From OpenBSD:
- support block length per function
- add functions to read/write regions
Decode (but not use) SDIO tuple in CIS.
Fix locking.
Add more SDIO defines (partially from version 3.0).

branches: 1.36.4;
Increase kthread priority to PRI_SOFTBIO.

branches: 1.35.2;
correct whitespace in attach message

branches: 1.34.12;
sdhc(4): hardware reset support for Intel eMMC controller

branches: 1.33.2; 1.33.4;
ignore timeouts for APP_CMD prefix as for the application command itself.

Add some event counters to track transfer sizes.

Send an explicit CMD12 (stop transmission) when there was an error
in multi-sector I/O.

The SDHC spec has a complex flowchart describing when an explicit
CMD12 is necessary, so we probably use it too often.

protect card detection with mutex.

Add support for DDR50 transfer modes.

use mutex locking for MP safety.

Be more verbose about errors.

sdmmc_task_thread isn't MPSAFE so instead of just wrapping the callback
in KERNEL_LOCK/KERNEL_UNLOCK_ONE, remove KTHREAD_MPSAFE. While here, use
PRI_BIO instead of PRI_NONE for the task thread priority. Since this is
how all transfers are queued, and we have a 1 second timeout on the task
being picked up off the queue (!), better not to rely on a PRI_NONE thread.

need to call lddone with splbio and kernel lock held

fix to simulate multi-segment dma transfer for pq3sdhc(4).

branches: 1.23.2;
Replace callout_stop with callout_halt

In order to call callout_destroy for a callout safely, we have to ensure
the function of the callout is not running and pending. To do so, we should
use callout_halt, not callout_stop.

Discussed with martin@ and riastradh@.

branches: 1.22.4;
Add a driver for Realtek RTS5209/RTS5229 Card Reader.
Ported from OpenBSD.

Pass the device name in, so we can debug what deferred drivers did not work.

branches: 1.20.2;
We can't block our interrupt while sdmmc_detach_card is called, as
it can issue commands, and some host drivers have interrupt-driven
command completion. Prevents assertion (or deadlock) upon detach
of sdhc(4) with inserted card.

Revert sdmmc.c r1.18 (previous commit to this file).
It seems to have caused regressions to detaching drivers from cards at
the appropriate time.

DETACH_FORCE to sdmmc_card_detach really means the SD/SDIO card is gone,
and can not be abused to mean that the host controller disappered.

Only use DETACH_FORCE as flags argument to sdmmc_card_detach when the
host controller has disappeared or is detaching.
XXX should pass flags from sdmmc_detach through to the task thread ending
to differentiate host controller disappearance from mere detachment.

It's also useful to compile-test before committing.

When debugging, it's useful to know what exactly failed.

branches: 1.15.2;
Print 'Standard Function Interface Code" in sdmmc_print(), if not SD_IO_SFIC_NO_STANDARD.

Whitespace.

If necessary, unload and destroy dmamap upon detach.

Use "opt_sdmmc.h" for SDMMC_DEBUG, etc.

Print error if couldn't enable card.

fix my license notice.

branches: 1.9.4; 1.9.8;
if poll card detect is enabled, don't show "couldn't enable card".

- Don't switch MMC high-speed timing, if host controller isn't supported.
- Only check EXT_CSD STRUCTURE version when CSD version is 3.
- initialize width at sdmmc_function_alloc().

Don't switch SD high-speed timing, if host controller isn't supported.

branches: 1.6.2; 1.6.4;
Fix read CIS. And pass to attach args the standard function interface code.

Support High-Speed mode.

Use DMA bounce buffer, if DMA buffer is making by multiple segments. A lot
of host controllers do not support to two or more segments.

Clear SMF_CARD_PRESENT, if failed to attach.

branches: 1.2.2;
- mention MMC SPI mode.
- support SD 4bit bus width mode.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.10; 1.1.12;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sdmmc: Capture lan_nid and expose sdmmc_cisptr

LAN NID contains the MAC address for networking adapters.

Device drivers may want to processor vendor specific tuple codes, so
expose sdmmc_cisptr to help this.

sdmmc: Decode CISTPL_FUNCE of type PCMCIA_TPLFE_TYPE_LAN_NID

branches: 1.8.26; 1.8.32;
Whitespace police

Remove debug printf.

From OpenBSD:
- support block length per function
- add functions to read/write regions
Decode (but not use) SDIO tuple in CIS.
Fix locking.
Add more SDIO defines (partially from version 3.0).

branches: 1.5.4; 1.5.10;
Don't print an error when we find a CIS tuple code in the vendor-unique
range (80h-8Fh).

Use "opt_sdmmc.h" for SDMMC_DEBUG, etc.

branches: 1.3.8; 1.3.12;
Fix build failure. s/const static/static const/.

Fix read CIS. And pass to attach args the standard function interface code.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.10; 1.1.12; 1.1.14;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

Use DMA for SD_IO_RW_EXTENDED.

Use the SCF_NO_STOP flag to disable AUTO_CMD12_ENABLE
instead of checking the SD_IO_RW_EXTENDED command op code.

Also add this logic to sunxi_mmc.

sdmmc: Use block mode for multi block SDIO transfers.

Change from OpenBSD:
https://github.com/openbsd/src/commit/5e0782b17abc2070c235843e37de911ce3f98ef8

So far the SDIO stack issued one transfer for every 64 byte to be
copied. This severely limits the speed over the bus. By using
block mode we can issue block-sized transfers, which bumps the
transfer size to at least 512 bytes. By using multi-block mode
we can copy up to 511 blocks per transfer.

branches: 1.21.28;
Fix error message. No functional change, both commands use the same
bit to select read or write mode.

Fix races in sdmmc tasks and teach ld@sdmmc to abort xfers on detach.

- Teach sdmmc_add_task to queue it only if not already queued.
- Remove now-redundant logic to avoid repeated queueing elsewhere.
- Teach sdmmc_del_task to wait until task has completed.
- Call sdmmc_del_task in various needful places.
- Replace abuse of pcq by a lock and a tailq.
(pcq is multi-producer, _single_-consumer, but there are potentially
multiple consumers here and really only one producer.)
- Teach ld_sdmmc to abort xfers on detach.
(Mechanism is kinda kludgey but it'll do for now; any effort one is
tempted to spend overhauling this should be spent overhauling sdmmc
to support proper asynchronous commands.)
- Make sure ld_sdmmc_discard either returns failure or eventually calls
ldenddiscard.

XXX Currently ld_sdmmc_detach aborts xfers _before_ ldbegindetach has
has committed to detaching or not. This is currently necessary to
avoid a deadlock because ldbegindetach waits for xfers to drain --
which strikes me as wrong; ldbegindetach shouldn't wait for anything,
and should only make the decision to commit to detaching or not so
the caller can decide whether to abort xfers before we actually wait
for them in ldenddetach.

XXX pullup -- although this changes some kernel symbols (sdmmc_add_task
and sdmmc_del_task), it shouldn't affect any existing modules; the only
module that uses sdmmc is ld_sdmmc.kmod, which is `.if 0' in the build
so there shouldn't be any of them floating around.

Be less noisy for some commands.

Simplyfy sdmmc_io_set_blocklen function signature by dropping the
extra softc pointer. Aligns with OpenBSD.

Add and use sdmmc_pause to avoid long-term busy waits.
Add sdio abort function.
Additional error messages.
Print parameters for SDIO devices.
Minor cosmetics.

SD_IO_RW_EXTENDED is a data transfer command, so set ADTC flag instead of AC

From OpenBSD:
- support block length per function
- add functions to read/write regions
Decode (but not use) SDIO tuple in CIS.
Fix locking.
Add more SDIO defines (partially from version 3.0).

branches: 1.14.4;
remove M_CANFAIL flag for malloc(9) - it was completely ignored, so had
actually no effect

branches: 1.13.2; 1.13.4;
- Get SDIO reset working (from OpenBSD).
- After switching bus width, notify the host controller of the change.

support hiding command timeout messages with a new command flag and use this
when probing for cards. Should fix PR 50302.

lock around interrupt registration.
don't release lock around interrupt handlers, the code cannot
handle an unstable handler list.

Add support for DDR50 transfer modes.

use mutex locking for MP safety.

cosmetic change only.

branches: 1.7.6; 1.7.24;
Use "opt_sdmmc.h" for SDMMC_DEBUG, etc.

branches: 1.6.4; 1.6.8;
- Don't switch MMC high-speed timing, if host controller isn't supported.
- Only check EXT_CSD STRUCTURE version when CSD version is 3.
- initialize width at sdmmc_function_alloc().

branches: 1.5.2; 1.5.4;
Fix read CIS. And pass to attach args the standard function interface code.

Remove unnecessary Tab.

Support High-Speed mode.

branches: 1.2.2; 1.2.4; 1.2.6;
Convert tsleep(&lbolt) to kpause(). Make ltsleep/mtsleep on lbolt
illegal. I examined all places where lbolt is referenced to make
sure there were pointer aliases of it passed to tsleep, but put a
KASSERT in m/ltsleep() just to be sure.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

Avoid changing signedness bit with << in sdmmc_ioreg.h

Reported by <prlw1>

Missing commit for sdio abort function.

From OpenBSD:
- support block length per function
- add functions to read/write regions
Decode (but not use) SDIO tuple in CIS.
Fix locking.
Add more SDIO defines (partially from version 3.0).

branches: 1.3.2;
Define SD_ARG_CMD52_WRITE macro correctly. Found by KUBSan. This macro is
used in sdmmc_io_write_1() and sdmmc_io_reset.

branches: 1.2.60;
Fix read CIS. And pass to attach args the standard function interface code.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.10; 1.1.12; 1.1.14;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

Don't stop sd card initialization early when bus clock is already set
to the target rate.

From jmnceill@ and tested by me on beagle-v.

sdmmc: Add support for SD card caches.

SD physical 6.0 specification introduced Application Performance Class 2
(A2), which adds support for drive caches and command queueing.

Add support for enabling and flushing the cache when this feature is
present.

branches: 1.75.6;
sdmmc: Only check chipset WP status for SD cards.

The sdmmc_chip_write_protect callback returns the write protect switch
status from the controller (SDWP#). This signal does not exist for eMMC;
instead, write protect is signaled using card registers (CSD). So lets
skip asking the chipset for WP status on eMMC cards for each write
request.

branches: 1.74.8;
Use unsigned to avoid undefined behavior in sdmmc_mem_sd_switch().
Found by kUBSan.

Be less verbose normally and more when debugging.

branches: 1.72.6;
If the controller doesn't support switch func (opcode 6) then skip
setting this but continue with other settings. This allows us to use
a card, albeit at a lower speed.

Be less noisy for some commands.

Whitespace police

Add and use sdmmc_pause to avoid long-term busy waits.
Add sdio abort function.
Additional error messages.
Print parameters for SDIO devices.
Minor cosmetics.

branches: 1.68.2;
If setting HS_TIMING fails, keep trying slower speeds instead of bailing
out.

If a mem function fails to initialize, set the error flag so sdmmc doesn't try to use it anyway.

Wait for data ready after eMMC flush cache commands

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: 1.64.2; 1.64.4;
Fix uninitialized variable use:
if there is an error, or if we are using a SPI controller,
sdmmc_mem_send_op_cond() doens't assign a value to *ocrp,
but it is used unconditionally in sdmmc_mem_enable() to see if we can switch
to low voltage.

In sdmmc_mem_send_op_cond(), if the new ocr is not returned by the
card for whatever reason, set *ocrp to the orig value.

For SD cards, send the SET_WR_BLK_ERASE_COUNT app command before a
multi-block write to improve write performance.

make the sc_discard interface for the ld backend asynchronous and
signal completion through new callback lddiscardend. Use a standard
struct buf to pass disk address and range instead of two off_t values.

make lddiscard synchronous again. This is a requirement of the current
ffs discard code.

branches: 1.61.2;
Add support for eMMC 4.5's optional cache feature. If a cache is present,
and the host controller reports the SMC_CAPS_POLLING capability (needed
to flush cache at shutdown), it will be automatically enabled and used.

Revert part of previous; the SD card spec recommends not to issue a DISCARD
command to the file system area.

Read SD status register and print card status when a new SD card is found:

sdmmc0: SD card status: 4-bit, C10, U1, V10

If the SD status register reports discard support, set the DISCARD arg to
the ERASE operation to let the card know that the host doesn't care if the
erase is performed.

Add discard support to ld@sdmmc using the ERASE (CMD38) command.

Give a hint to controllers in the command flags if we are performing a
transfer with an SDHC capable card. If the controller needs to adjust
command args, it can use this hint to understand how it is encoded.

branches: 1.56.4;
Remove a test that prevents tuning from happening on eMMC devices.

sdhc(4): hardware reset support for Intel eMMC controller

sdhc(4), sdmmc(4): Added MMC HS DDR52 support.

sdmmc(4): Use EXT_CSD[HS_TIMING] definitions.

branches: 1.52.2;
Fix incorrect sector counts with MMC.

branches: 1.51.2;
Call the second sdmmc_mem_send_if_cond() only where it's necessary.

This makes SMC_CAPS_SPI_MODE devices (currently evbsh3 only) work again.
"Maybe ok" from nonaka@.

Be graceful about command timeouts when probing.

Add some event counters to track transfer sizes.

After setting HS_TIMING value for HS200 or later, send repeated SEND_STATUS
command until the device is no longer busy or the SWITCH_ERROR bit is set.

support hiding command timeout messages with a new command flag and use this
when probing for cards. Should fix PR 50302.

eMMC fixes

Add support for sampling clock tuning, required for some UHS modes and
MMC HS200.

ODROID-C1 has a bug where the card is not power-cycled when the board is
reset. If you had previously switched to 1.8V signaling level, upon reboot
the card will still be in 1.8V mode and you cannot detect it with an S18R
request.

A card in 1.8V mode will report UHS modes though, so if the card reports
SDR50, DDR50, or SDR104 capabilities, and the previous S18R request failed
to switch, use this as an opportunity to re-enable UHS support in the
subsystem and host controller drivers.

instead of trying to handle cards already in 1.8V mode, request host controller to switch to 3.3V signaling before starting the init sequence

Handle cards that are already in low voltage (1.8V) mode. Test correct
bit when determining the best SD transfer mode.

clear SMF_UHS_MODE flag before enabling card

fix SDMMC_DEBUG build

Add support for DDR50 transfer modes.

use mutex locking for MP safety.

more debug output

Add support for eMMC 5.0 HS200 timings.

Add basic UHS-I support. SDR50 and SDR104 are supported, but not DDR50.

fix to simulate multi-segment dma transfer for pq3sdhc(4).

Return an actual error. Reported by:
http://www.m00nbsd.net/ae123a9bae03f7dde5c6d654412daf5a.html#Report-4

Fix high capacity (> 2GB) eMMC support, from OpenBSD.

branches: 1.31.4; 1.31.6;
Add a driver for Realtek RTS5209/RTS5229 Card Reader.
Ported from OpenBSD.

Turn a few __unused into __diagused

branches: 1.29.4;
Add support for the valid card types in eMMC v4.4 (needed by beaglebone
black).

Fix typo.

fix off-by-one in switch function argument validation.

Correctly read the 512-bit-wide big-endian Switch Function Status register.
Some of this could/will also be useful for the SD Status register.

The Card Command Class (CCC) field is valid in both SD_CSD_CSDVER_1_0
and SD_CSD_CSDVER_2_0.

Fix variable name for message of printf.

branches: 1.23.2;
Fix a delay to a correct value.

Responses are actually in host order (except SCR which is return in
big endian so that's convert to host order).

Add use of watermark register when PIO to an ESDHC. After every kill or
drain of watermask words, pause a bit to give time for the fifo to recover.
Always the command response in BE byteorder. Rewrite __bitfield to deal
with this.

branches: 1.20.2;
Use "opt_sdmmc.h" for SDMMC_DEBUG, etc.

Fix sdmmmc_mem_decode_scr for BE machines.

fix my license notice.

branches: 1.17.4; 1.17.8;
use MMC_CSD_CSDVER_EXT_CSD.

- Don't switch MMC high-speed timing, if host controller isn't supported.
- Only check EXT_CSD STRUCTURE version when CSD version is 3.
- initialize width at sdmmc_function_alloc().

Don't switch SD high-speed timing, if host controller isn't supported.

branches: 1.14.2; 1.14.4;
Don't pull in the whole uvm(9) API to access only PAGE_SIZE and
some other constants. These are provided by sys/param.h now.

Support High-Speed mode.

Use DMA bounce buffer, if DMA buffer is making by multiple segments. A lot
of host controllers do not support to two or more segments.

Print bus width and clock for feature(High speed support).

Fix build failure.
A new variable width and value had been forgotten since r1.7.

Clean up return/error in sdmem_mem_send_scr().

Fix SCR data.

Support MMC 4,8-bit mode.
It tested only 4bit on Marvell Sheevaplug.

Sync with POSTREAD before memcpy.

Initialise error variable to zero. If all the if's fail it exists the function
sdmmc_init returning the then still uninitialised variable.

gcc can't know if one of the if's is bound to happen.

Detected with -DDEBUG -DDIAGNOSTIC -DKGDB - ...

branches: 1.4.2;
- mention MMC SPI mode.
- support SD 4bit bus width mode.

branches: 1.3.2; 1.3.4;
Always sector size is treated as 512 bytes.

branches: 1.2.2;
SDHC card was able to be used.

branches: 1.1.2; 1.1.4; 1.1.6;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

Add SDHC flags.

+ SDHC_FLAG_BROKEN_ADMA2_ZEROLEN
Broken ADMA2 Zero length descriptor.
Can't 64K Byte data transfer.
+ SDHC_FLAG_NO_1_8_V
Support no 1.8V Supply.
Disable UHS-I bus speed mode (SDR50, DDR50, SDR104).

branches: 1.9.14;
sdhc(4): hardware reset support for Intel eMMC controller

sdhc(4), sdmmc(4): Added MMC HS DDR52 support.

branches: 1.7.2; 1.7.4;
Add support for sampling clock tuning, required for some UHS modes and
MMC HS200.

Add support for DDR50 transfer modes.

Add basic UHS-I support. SDR50 and SDR104 are supported, but not DDR50.

branches: 1.4.14; 1.4.32;
#include <sys/bus.h>, not <machine/bus.h>.

branches: 1.3.2;
Add bus_rod. It use to change to Data Transfer mode.

branches: 1.2.2;
- mention MMC SPI mode.
- support SD 4bit bus width mode.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.10; 1.1.12;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sdmmc: Add Nintendo Wii WLAN device ID

branches: 1.8.26; 1.8.32;
Add product ID for Broadcom BCM43455

Whitespace police

Fix typo

Add Broadcom devices

branches: 1.4.4;
Add ReakTek RTL8189FTV SDIO 802.11 interface.

branches: 1.3.2;
Add vendor REALTEK and product RTL8703BS

branches: 1.2.28;
add Ricoh MMC reader

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8; 1.1.14; 1.1.22; 1.1.32; 1.1.36;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sdmmc: regen sdmmcdevs.h

branches: 1.8.26; 1.8.32;
regen

regen

regen

regen

branches: 1.4.4;
Regen.

branches: 1.3.2;
Regen

branches: 1.2.2; 1.2.4; 1.2.6; 1.2.8; 1.2.14; 1.2.72;
regen

Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sdmmc: Add support for SD card caches.

SD physical 6.0 specification introduced Application Performance Class 2
(A2), which adds support for drive caches and command queueing.

Add support for enabling and flushing the cache when this feature is
present.

branches: 1.34.34; 1.34.40;
s/static inline/static __inline/g for consistency.

branches: 1.33.2;
For SD cards, send the SET_WR_BLK_ERASE_COUNT app command before a
multi-block write to improve write performance.

branches: 1.32.2;
Add support for eMMC 4.5's optional cache feature. If a cache is present,
and the host controller reports the SMC_CAPS_POLLING capability (needed
to flush cache at shutdown), it will be automatically enabled and used.

Read SD status register and print card status when a new SD card is found:

sdmmc0: SD card status: 4-bit, C10, U1, V10

If the SD status register reports discard support, set the DISCARD arg to
the ERASE operation to let the card know that the host doesn't care if the
erase is performed.

Add discard support to ld@sdmmc using the ERASE (CMD38) command.

branches: 1.29.6;
sdhc(4): hardware reset support for Intel eMMC controller

sdmmc(4): Added EXT_CSD[HS_TIMING] definitions.

sdmmc(4): Change EXT_CSD[CARD_TYPE] HS DDR 52 MHz definition.

Because it has been difficult to understand from the definition is DDR.

sdmmc(4): Remove unused EXT_CSD[CARD_TYPE] definitions.

sdmmc(4): Added comments at EXT_CSD[CARD_TYPE] definitions.

sdmmc(4): Added 4 and 8 bit mode DDR definitions at EXT_CSD[BUS_WIDTH].

sdmmc(4): Fix cell type in comments.

From JEDEC Standard No.84-B51, 7.4. Extended CSD register.

branches: 1.22.2;
Use 1.65-1.95 voltage window for 1.8V support.

branches: 1.21.2;
After setting HS_TIMING value for HS200 or later, send repeated SEND_STATUS
command until the device is no longer busy or the SWITCH_ERROR bit is set.

eMMC fixes

Add support for sampling clock tuning, required for some UHS modes and
MMC HS200.

Add support for DDR50 transfer modes.

Add support for eMMC 5.0 HS200 timings.

Add basic UHS-I support. SDR50 and SDR104 are supported, but not DDR50.

Fix high capacity (> 2GB) eMMC support, from OpenBSD.

branches: 1.14.12;
Add support for the valid card types in eMMC v4.4 (needed by beaglebone
black).

Correctly read the 512-bit-wide big-endian Switch Function Status register.
Some of this could/will also be useful for the SD Status register.

branches: 1.12.2;
Fix comments about __bitfield.

Responses are actually in host order (except SCR which is return in
big endian so that's convert to host order).

Add use of watermark register when PIO to an ESDHC. After every kill or
drain of watermask words, pause a bit to give time for the fifo to recover.
Always the command response in BE byteorder. Rewrite __bitfield to deal
with this.

Add SD_STATUS (ACMD13) opcode.

branches: 1.8.2;
Remove suplurfious ++

branches: 1.7.4; 1.7.8;
use MMC_CSD_CSDVER_EXT_CSD.

- Don't switch MMC high-speed timing, if host controller isn't supported.
- Only check EXT_CSD STRUCTURE version when CSD version is 3.
- initialize width at sdmmc_function_alloc().

branches: 1.5.2; 1.5.4;
Support High-Speed mode.

branches: 1.4.2;
- mention MMC SPI mode.
- support SD 4bit bus width mode.

branches: 1.3.2; 1.3.4; 1.3.6; 1.3.8; 1.3.10; 1.3.12;
fix build broken.
Pointed by Takeshi Nakayama.

Added some command definitions.

Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.

sdmmc: Capture lan_nid and expose sdmmc_cisptr

LAN NID contains the MAC address for networking adapters.

Device drivers may want to processor vendor specific tuple codes, so
expose sdmmc_cisptr to help this.

sdmmc: Add support for SD card caches.

SD physical 6.0 specification introduced Application Performance Class 2
(A2), which adds support for drive caches and command queueing.

Add support for enabling and flushing the cache when this feature is
present.

branches: 1.36.18; 1.36.24;
define NO_STOP flag

branches: 1.35.2;
Fix races in sdmmc tasks and teach ld@sdmmc to abort xfers on detach.

- Teach sdmmc_add_task to queue it only if not already queued.
- Remove now-redundant logic to avoid repeated queueing elsewhere.
- Teach sdmmc_del_task to wait until task has completed.
- Call sdmmc_del_task in various needful places.
- Replace abuse of pcq by a lock and a tailq.
(pcq is multi-producer, _single_-consumer, but there are potentially
multiple consumers here and really only one producer.)
- Teach ld_sdmmc to abort xfers on detach.
(Mechanism is kinda kludgey but it'll do for now; any effort one is
tempted to spend overhauling this should be spent overhauling sdmmc
to support proper asynchronous commands.)
- Make sure ld_sdmmc_discard either returns failure or eventually calls
ldenddiscard.

XXX Currently ld_sdmmc_detach aborts xfers _before_ ldbegindetach has
has committed to detaching or not. This is currently necessary to
avoid a deadlock because ldbegindetach waits for xfers to drain --
which strikes me as wrong; ldbegindetach shouldn't wait for anything,
and should only make the decision to commit to detaching or not so
the caller can decide whether to abort xfers before we actually wait
for them in ldenddetach.

XXX pullup -- although this changes some kernel symbols (sdmmc_add_task
and sdmmc_del_task), it shouldn't affect any existing modules; the only
module that uses sdmmc is ld_sdmmc.kmod, which is `.if 0' in the build
so there shouldn't be any of them floating around.

Simplyfy sdmmc_io_set_blocklen function signature by dropping the
extra softc pointer. Aligns with OpenBSD.

Add and use sdmmc_pause to avoid long-term busy waits.
Add sdio abort function.
Additional error messages.
Print parameters for SDIO devices.
Minor cosmetics.

Add SDHC flags.

+ SDHC_FLAG_BROKEN_ADMA2_ZEROLEN
Broken ADMA2 Zero length descriptor.
Can't 64K Byte data transfer.
+ SDHC_FLAG_NO_1_8_V
Support no 1.8V Supply.
Disable UHS-I bus speed mode (SDR50, DDR50, SDR104).

From OpenBSD:
- support block length per function
- add functions to read/write regions
Decode (but not use) SDIO tuple in CIS.
Fix locking.
Add more SDIO defines (partially from version 3.0).

branches: 1.30.4;
Add SCF_NEED_BOUNCE command flag. Can be used by a driver to keep track of
which command(s) need data transfers to go through DMA bounce buffers.

branches: 1.29.4;
make the sc_discard interface for the ld backend asynchronous and
signal completion through new callback lddiscardend. Use a standard
struct buf to pass disk address and range instead of two off_t values.

make lddiscard synchronous again. This is a requirement of the current
ffs discard code.

branches: 1.28.2;
Add support for eMMC 4.5's optional cache feature. If a cache is present,
and the host controller reports the SMC_CAPS_POLLING capability (needed
to flush cache at shutdown), it will be automatically enabled and used.

Revert part of previous; the SD card spec recommends not to issue a DISCARD
command to the file system area.

Read SD status register and print card status when a new SD card is found:

sdmmc0: SD card status: 4-bit, C10, U1, V10

If the SD status register reports discard support, set the DISCARD arg to
the ERASE operation to let the card know that the host doesn't care if the
erase is performed.

Add discard support to ld@sdmmc using the ERASE (CMD38) command.

Give a hint to controllers in the command flags if we are performing a
transfer with an SDHC capable card. If the controller needs to adjust
command args, it can use this hint to understand how it is encoded.

branches: 1.23.6;
sdhc(4): hardware reset support for Intel eMMC controller

sdmmc(4): Add MMC HS DDR52 timing support bit and use __BIT() macro.

branches: 1.21.2; 1.21.4;
Add some event counters to track transfer sizes.

support hiding command timeout messages with a new command flag and use this
when probing for cards. Should fix PR 50302.

Send an explicit CMD12 (stop transmission) when there was an error
in multi-sector I/O.

The SDHC spec has a complex flowchart describing when an explicit
CMD12 is necessary, so we probably use it too often.

Add support for DDR50 transfer modes.

Add basic UHS-I support. SDR50 and SDR104 are supported, but not DDR50.

fix to simulate multi-segment dma transfer for pq3sdhc(4).

branches: 1.15.4; 1.15.6;
Add a driver for Realtek RTS5209/RTS5229 Card Reader.
Ported from OpenBSD.

branches: 1.14.2; 1.14.4;
Replace integer constant SDMMC_MAXNSEGS with the formula said value was
calculated from.

Fix transposition typo in comment.

Use "opt_sdmmc.h" for SDMMC_DEBUG, etc.

branches: 1.11.4; 1.11.8;
#include <sys/bus.h>, not <machine/bus.h>.

- Don't switch MMC high-speed timing, if host controller isn't supported.
- Only check EXT_CSD STRUCTURE version when CSD version is 3.
- initialize width at sdmmc_function_alloc().

Don't switch SD high-speed timing, if host controller isn't supported.

branches: 1.8.2; 1.8.4;
Support High-Speed mode.

Use DMA bounce buffer, if DMA buffer is making by multiple segments. A lot
of host controllers do not support to two or more segments.

Print bus width and clock for feature(High speed support).

Back out previous change. "automagically" is NOT a typo.

+ Add macro SMC_CAPS_8BIT_MODE for 8bit mode support.
+ Typo. s/automagically/automatically/.

branches: 1.3.2;
- mention MMC SPI mode.
- support SD 4bit bus width mode.

branches: 1.2.2; 1.2.4;
Always sector size is treated as 512 bytes.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8;
Added SD/MMC support from OpenBSD.
tested on i386, amd64 at current-users ML by pgoyette@.
tested on zaurus by myself.