This function is doing to many clever things at the same time under
too many various conditions.
Start to make things clearer by refactoring: break out the
finalization of the previous asynchronous request to its own
function mmc_finalize_areq(). We can get rid of the default
assignment of status and let the call deal with this.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
With the coexisting __mmc_start_request(), mmc_start_request()
and __mmc_start_req() it is a bit confusing that mmc_start_req()
actually does not start a normal request, but an asynchronous
request.
Rename it to mmc_start_areq() to make it explicit what the
function is doing, also fix the kerneldoc for this function
while we're at it.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
commit 64b12a68a9
"mmc: core: fix prepared requests while doing bkops"
is fixing a bug in the wrong way. A bug in the MMCI
device driver is fixed by amending the MMC core.
Thinking about it: what the pre- and post-callbacks
are doing is to essentially map and unmap SG lists
for DMA transfers. Why would we not be able to do that
just because a BKOPS command is sent inbetween?
Having to unprepare/prepare the next asynchronous
request for DMA seems wrong.
Looking the backtrace in that commit we can see what
the real problem actually is:
mmci_data_irq() is calling mmci_dma_unmap() twice
which is goung to call arm_dma_unmap_sg() twice
and v7_dma_inv_range() twice for the same sglist
and that will crash.
This happens because a request is prepared, then
a BKOPS is sent. The IRQ completing the BKOPS command
goes through mmci_data_irq() and thinks that a DMA
operation has just been completed because
dma_inprogress() reports true. It then proceeds to
unmap the sglist.
But that was wrong! dma_inprogress() should NOT be
true because no DMA was actually in progress! We had
just prepared the sglist, and the DMA channel
dma_current has been configured, but NOT started!
Because of this, the sglist is already unmapped when
we get our actual data completion IRQ, and we are
unmapping the sglist once more, and we get this crash.
Therefore, we need to revert this solution pushing
the problem to the core and causing problems, and
instead augment the implementation such that
dma_inprogress() only reports true if some DMA has
actually been started.
After this we can keep the request prepared during the
BKOPS and we need not unprepare/reprepare it.
Fixes: 64b12a68a9 ("mmc: core: fix prepared requests while doing bkops")
Cc: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Earlier the mmc_set_signal_voltage() existed, but since it has been renamed
to mmc_set_uhs_voltage(), we can now use that name instead.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Shawn Lin <shawn.lin@rock-chips.com>
Tested-by: Jan Glauber <jglauber@cavium.com>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
The mmc_set_signal_voltage() function is used for SD/SDIO when switching to
1.8V for UHS mode. To clarify this let's do the following changes.
- We are always providing MMC_SIGNAL_VOLTAGE_180 as the signal_voltage
parameter to the function. Then, let's just remove the parameter as it
serves no purpose.
- Rename the function to mmc_set_uhs_voltage().
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Shawn Lin <shawn.lin@rock-chips.com>
Tested-by: Jan Glauber <jglauber@cavium.com>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
The mmc_set_signal_voltage() function is used for SD/SDIO when switching to
1.8V for UHS mode. Therefore let's remove the redundant code dealing with
MMC_SIGNAL_VOLTAGE_330.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Shawn Lin <shawn.lin@rock-chips.com>
Tested-by: Jan Glauber <jglauber@cavium.com>
Tested-by: Stefan Wahren <stefan.wahren@i2se.com>
A significant amount of functions and other definitions are available
through the public mmc card.h header file. Let's slim down this public mmc
interface, as to prevent users from abusing it, by moving some of the
functions/definitions to private mmc header files.
This change concentrates on moving the functions into private mmc headers,
following changes may continue with additional clean-ups.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Shawn Lin <shawn.lin@rock-chips.com>
In the MMC subsystem, we see such initializers that only clears the
first member explicitly.
For example,
struct mmc_request mrq = {NULL};
sets the first member (.sbc) to NULL explicitly. However, this is
an unstable form because we may insert a non-pointer member at the
top of the struct mmc_request in the future. (if we do so, the
compiler will spit warnings.)
So, using a designated initializer is preferred coding style. The
expression above is equivalent to:
struct mmc_request mrq = { .sbc = NULL };
Of course, this does not express our intention. We want to fill
all struct members with zeros. Please note struct members are
implicitly zero-cleared unless otherwise specified in the initializer.
After all, the most reasonable (and stable) form is:
struct mmc_request mrq = {};
Do likewise for mmc_command, mmc_data as well.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Commit e0097cf5f2 ("mmc: queue: Fix queue thread wake-up") did not go far
enough. mmc_wait_for_data_req_done() still contains some problems and can
be further simplified. First it should not touch
context_info->is_waiting_last_req because that is a wake-up control used by
the owner of the context. Secondly, it should always return when one of its
wake-up conditions is met because, again, that is contolled by the owner of
the context.
While the current block driver does not have an issue, these problems were
exposed during testing of the Software Command Queue patches.
Fixes: e0097cf5f2 ("mmc: queue: Fix queue thread wake-up")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Tested-by: Harjani Ritesh <riteshh@codeaurora.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
BUG_ONs doesn't help anything except for stop the system from
running. If it occurs, it implies we should deploy proper error
handling for that. So this patch is gonna discard these meaningless
BUG_ONs and deploy error handling if needed.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The only time the driver sleeps expecting to be woken upon the arrival of
a new request, is when the dispatch queue is empty. The only time that it
is known whether the dispatch queue is empty is after NULL is returned
from blk_fetch_request() while under the queue lock.
Recognizing those facts, simplify the synchronization between the queue
thread and the request function. A couple of flags tell the request
function what to do, and the queue lock and barriers associated with
wake-ups ensure synchronization.
The result is simpler and allows the removal of the context_info lock.
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Harjani Ritesh <riteshh@codeaurora.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The void (*pre_req) callback in the struct mmc_host_ops vtable
is passing an argument "is_first_req" indicating whether this is
the first request or not.
None of the in-kernel users use this parameter: instead, since
they all just do variants of dma_map* they use the DMA cookie
to indicate whether a pre* callback has already been done for
a request when they decide how to handle it.
Delete the parameter from the callback and all users, as it is
just pointless cruft.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Acked-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
In cases when a speed mode change is requested for mmc cards, a CMD6 is
sent by calling __mmc_switch() during the card initialization. The CMD6
leads to the card entering a busy period. When that is completed, the host
must parse the CMD6 status to find out whether the change of the speed mode
succeeded.
To enable the mmc core to poll the card by using CMD13 to find out when the
busy period is completed, it's reasonable to make sure polling is done by
having the mmc host and the mmc card, being configured to operate at the
same selected bus speed timing.
Therefore, let's extend __mmc_switch() to take yet another parameter, which
allow its callers to update the bus speed timing of the mmc host. In this
way, __mmc_switch() also becomes capable of reading and validating the CMD6
status by sending a CMD13, in cases when that's desired.
If __mmc_switch() encounters a failure, we make sure to restores the old
bus speed timing for the mmc host, before propagating the error code.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
There were several instances of code using the
enum mmc_blk_status by arbitrarily converting it to an int and
throwing it around to different functions. This makes the code
hard to understand to may give rise to strange errors.
Especially the function prototype mmc_start_req() had to be
modified to take a pointer to an enum mmc_blk_status and the
function pointer .err_check() inside struct mmc_async_req
needed to return an enum mmc_blk_status.
In every case: instead of assigning the block layer error code
to an int, use the enum, also change the signature of all
functions actually passing this enum to use the enum.
To make it possible to use the enum everywhere applicable, move
it to <linux/mmc/core.h> so that all code actually using it can
also see it.
An interesting case was encountered in the MMC test code which
did not return a enum mmc_blk_status at all in the .err_check
function supposed to check whether asynchronous requests worked
or not: instead it returned a normal -ERROR or even the test
frameworks internal error codes.
The test code would also pass on enum mmc_blk_status codes as
error codes inside the test code instead of converting them
to the local RESULT_* codes.
I have tried to fix all instances properly and run some tests
on the result.
Cc: Chunyan Zhang <zhang.chunyan@linaro.org>
Cc: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The MMC_CAP2_NO_PRESCAN_POWERUP was invented to avoid running the power up
sequence, mmc_power_up(), during ->probe() of the mmc host driver, but
instead defer this to the mmc detect work. This is especially useful for
those hosts that suffers from a long initialization time, as this time
would otherwise add up to the total boot time.
However, due to the introduction of runtime PM of mmc host devices in the
mmc core, this behaviour changed a bit. More precisely, it caused the mmc
core to runtime resume the host device during ->probe() of the host driver.
In cases like the rtsx_usb_sdmmc, runtime resuming the device may be costly
and thus affecting the total boot time.
To improve this behaviour when using MMC_CAP2_NO_PRESCAN_POWERUP, let's
postpone also calling mmc_power_off() when starting the host. This change
allows the mmc core to avoid runtime resuming the device, as it don't need
to claim the host for that execution path.
Cc: Ritesh Raj Sarraf <rrs@researchut.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Per spec, block size should always be 512 bytes for dual rate mode,
so any attempts to switch the block size under dual rate mode should
be neglected.
Signed-off-by: Ziyuan Xu <xzy.xu@rock-chips.com>
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
A host controller driver exposes its capability using caps flag
MMC_CAP_CMD_DURING_TFR. A driver with that capability can accept requests
that are marked mrq->cap_cmd_during_tfr = true. Then the driver informs the
upper layers when the command line is available for further commands by
calling mmc_command_done(). Because of that, the driver will not then
automatically send STOP commands, and it is the responsibility of the upper
layer to send a STOP command if it is required.
For requests submitted through the mmc_wait_for_req() interface, the caller
sets mrq->cap_cmd_during_tfr = true which causes mmc_wait_for_req() in fact
not to wait. The caller can then send commands that do not use the data
lines. Finally the caller can wait for the transfer to complete by calling
mmc_wait_for_req_done() which is now exported.
For requests submitted through the mmc_start_req() interface, the caller
again sets mrq->cap_cmd_during_tfr = true, but mmc_start_req() anyway does
not wait. The caller can then send commands that do not use the data
lines. Finally the caller can wait for the transfer to complete in the
normal way i.e. calling mmc_start_req() again.
Irrespective of how a cap_cmd_during_tfr request is started,
mmc_is_req_done() can be called if the upper layer needs to determine if
the request is done. However the appropriate waiting function (either
mmc_wait_for_req_done() or mmc_start_req()) must still be called.
The implementation consists primarily of a new completion
mrq->cmd_completion which notifies when the command line is available for
further commands. That completion is completed by mmc_command_done().
When there is an ongoing data transfer, calls to mmc_wait_for_req() will
automatically wait on that completion, so the caller does not have to do
anything special.
Note, in the case of errors, the driver may call mmc_request_done() without
calling mmc_command_done() because mmc_request_done() always calls
mmc_command_done().
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
In most cases the 'card->erase_size' is power of 2, then the round_up/down()
function is more efficient than '%' operation when the 'card->erase_size' is
power of 2.
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Tested-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
In order to clean up the mmc_erase() function and do some optimization
for erase size alignment, factor out the guts of erase size alignment
into mmc_align_erase_size() function.
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Tested-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
In cases when the host->max_busy_timeout isn't specified, the calculated
number of maximum discard sectors defaults to UINT_MAX. This may cause a
too long timeout for a discard request.
Avoid this by using a default maximum erase timeout of 60s, used when we
calculate the maximum number of sectors that are allowed to be discarded
per request.
Do note that the minimum number of sectors to be discarded is still at
least one "preferred erase size".
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Shawn Lin <shawn.lin@rock-chips.com>
When mmc host HW supports busy signalling (using R1B as response), don't
use the host->max_busy_timeout as the limitation when deciding the max
discard sectors, which we inform the generic BLOCK layer about. Instead,
let's use at least one preferred erase size as the max discard sectors.
In cases when the host controller supports HW busy signalling and the
timeout for the erase operation doesn't exceed the max_busy_timeout, we
keep the R1B response, otherwise we prevent the host from doing HW busy
detection by converting to a R1 response.
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Host drivers which needs to valdiate for non-supported MMC
commands and returnn error code for such requests.
To improve and simplify the behaviour, let's invent MMC_CAP2_NO_MMC
which these host drivers can set to tell the mmc core to skip sending MMC
commands during card initialization.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
There are host drivers which needs to valdiate for non-supported SD
commands and returnn error code for such requests.
To improve and simplify the behaviour, let's invent MMC_CAP2_NO_SD
which these host drivers can set to tell the mmc core to skip sending SD
commands during card initialization.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
If available, eMMC stack uses HC_ERASE_GRP_SIZE as the preferred erase
size. As some high capacity eMMC (64MB) reports this size to 512kB, the
discard operations end up taking very long time.
Improve the behaviour by instead calculating the preferred erase size
based on the eMMC size. In this way the discard operations becomes faster.
Signed-off-by: Gwendal Grignou <gwendal@chromium.org>
[Ulf: Updated changelog and improved comment in code]
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Controllers use data strobe line to latch data from devices
under hs400 mode, but not for cmd line. So since emmc 5.1, JEDEC
introduces enhanced strobe mode for latching cmd response from
emmc devices to host controllers. This new feature is optional,
so it depends both on device's cap and host's cap to decide
whether to use it or not.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Reviewed-by: Jaehoon Chung <jh80.chung@samsung.com>
Tested-by: Douglas Anderson <dianders@chromium.org>
Tested-by: Jaehoon Chung <jh80.chung@samsung.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
008GE0 Toshiba mmc in some Intel Baytrail tablets responds to
MMC_SEND_EXT_CSD in 450-600ms.
This patch will...
() Increase the long read time quirk timeout from 300ms to 600ms. Original
author of that quirk says 300ms was only a guess and that the number
may need to be raised in the future.
() Add this specific MMC to the quirk
Signed-off-by: Matt Gumbel <matthew.k.gumbel@intel.com>
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The eMMC HW reset may be implemented either via the host ops ->hw_reset()
callback or through DT and the eMMC pwrseq. Additionally some eMMC cards
don't support HW reset.
To allow a reset to be done for the different combinations of mmc hosts
and eMMC/MMC cards, let's implement a fallback via trying a regular power
cycle. This improves the mmc block layer retry mechanism of failing I/O
requests.
Signed-off-by: Gwendal Grignou <gwendal@chromium.org>
[Ulf: Rewrote changelog]
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
This patch provides some tracepoints for the lifecycle of a mmc request
from starting to completion to help with performance analysis of MMC
subsystem.
Signed-off-by: Baolin Wang <baolin.wang@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Print the error code when the tuning command fails. This allows the
reason for the failure to be reported, which aids debugging.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Improve mmc_of_parse_voltage()'s return values so that drivers can tell
whether a voltage-range specification was present, and whether it has
been successfully parsed, or there was an error while parsing.
We return a negative errno when parsing fails, zero if no voltage-range
specification is present, or one if a voltage-range specification is
successfully parsed.
No users need modifying as no users check the return value.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Each time a driver such as sdhci-esdhc-imx is probed, we get a info
printk complaining that the DT voltage-ranges property has not been
specified.
However, the DT binding specifically says that the voltage-ranges
property is optional. That means we should not be complaining that
DT hasn't specified this property: by indicating that it's optional,
it is valid not to have the property in DT.
Silence the warning if the property is missing.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The bus width is sometimes the actual bus width, and sometimes indices
to different arrays encoding the bus width. In my debugging case "2"
could mean 8-bit as well as 4-bit, which was extremly confusing. Let's
use the human-readable actual bus width in all places.
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
In linux/mmc/host.h, mmc_card_is_removable() is already defined.
There is no reason that it doesn't use.
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The mmc workqueue is an ordered workqueue, allowing only one work to
execute per given time. As this workqueue is used for card detection, the
conseqeunce is that cards will be detected one by one waiting for each
other.
Moreover, most of the time spent during card initialization is waiting for
the card's internal firmware to be ready. From a CPU perspective this
typically means waiting for a completion variable to be kicked via an
IRQ-handler or waiting for a sleep timer to finish.
This behaviour of detecting/initializing cards is sub-optimal, especially
for SOCs having several controllers/cards.
Let's convert to use the system_freezable_wq for the mmc detect works.
This enables several works to be executed simultaneously and thus also
cards to be detected like so.
Tests on UX500, which holds two eMMC cards and an SD-card (actually also
an SDIO card, currently not detected), shows a significant improved
behaviour due to this change.
Before this change, both the eMMC cards waited for the SD card to be
initialized as its detect work entered the workqueue first. In some cases,
depending on the characteristic of the SD-card, they got delayed 1-1.5 s.
Additionally for the second eMMC, it needed to wait for the first eMMC to
be initialized which added another 120-190 ms.
Converting to the system_freezable_wq, removed these delays and made both
the eMMC cards available far earlier in the boot sequence.
Selecting the system_freezable_wq, in favour of for example the system_wq,
is because we need card detection mechanism to be disabled once userspace
are frozen during system PM. Currently the mmc core deal with this via PM
notifiers, but following patches may utilize the behaviour of the
system_freezable_wq, to simplify the use of the PM notifiers.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Alan Cooper <alcooperx@gmail.com>
Tested-by: Shawn Lin <shawn.lin@rock-chips.com>
This patch introduce a new MMC_CAP2_NO_SDIO cap used to tell the mmc
core to not send SDIO specific commands.
Signed-off-by: Carlo Caione <carlo@endlessm.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Turn the informative message about no vmmc/vqmmc regulator found in
debug one. There is no need to indicate that something optional is
missing. Moreover, it can bring confusion, people who doesn't know
it is optional may consider these messages as warnings or errors.
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
_mmc_detect_card_removed() validates that the card is removable, but when
being called via the bus_ops ->detect() callbacks, the validation is
redundant as it's already done in mmc_rescan().
Move the validation of a removable card to the mmc_detect_card_removed()
API, which is where it's applicable, to allow the blk error recovery path
to get the response a bit earlier.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Instead of checking for "#ifdef" directly in the code, let's invent a pair
of mmc core functions to deal with register/unregister the MMC PM notifier
block. Implement stubs for these functions when CONFIG_PM_SLEEP is unset,
as in that case the PM notifiers isn't used.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
As mmc_claim_host() invokes pm_runtime_get_sync() for the mmc host device,
it's important that the host is kept claimed for *all* accesses to it via
the host_ops callbacks.
In mmc_rescan(), the ->card_event() and the ->get_cd() callback are being
invoked without claiming the host, let's fix this.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The ->card_event() callback may be called when re-scan is disabled and for
non-removable cards, which both cases are unnecessary.
Instead let's move the call later in mmc_rescan() where these constraints
have been validated.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
This adds logic to the MMC core to set VQMMC. This is expected to be
called by MMC drivers like dw_mmc as part of (or instead of) their
start_signal_voltage_switch() callback.
A few notes:
* When setting the signal voltage to 3.3V we do our best to make VQMMC
and VMMC match. It's been reported that this makes some old cards
happy since they were tested back in the day before UHS when VQMMC
and VMMC were provided by the same regulator. A nice side effect of
this is that we don't end up on the hairy edge of VQMMC (2.7V),
which some EEs claim is a little too close to the minimum for
comfort.
This is done in two steps. At first we try to find a VQMMC within
a 0.3V tolerance of VMMC and if this is not supported by the
supplying regulator we try to find a suitable voltage within the
whole 2.7V-3.6V area of the spec.
* The two step approach is currently necessary, as the used
regulator_set_voltage_triplet(min, target, max) uses a simple
implementation that just tries two basic steps:
regulator_set_voltage(target, max);
regulator_set_voltage(min, target);
So with only one step with 2.7-3.6V borders, if a suitable voltage
is a bit below VMMC, we would directly get the lowest 2.7V
which some boards (like Rockchips) don't like at all.
* When setting the signal voltage to 1.8V or 1.2V we aim for that
specific voltage instead of picking the lowest one in the range.
* We very purposely don't print errors in mmc_regulator_set_vqmmc().
There are cases where the MMC core will try several different
voltages and we don't want to pollute the logs.
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
We will shortly need the calculation of an ocr-bit to the actual
voltage in a second place too, so move it from mmc_regulator_set_ocr
to a common function mmc_ocrbitnum_to_vdd to make that possible.
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
MMC_CLKGATE was once invented to save power by gating the bus clock at
request inactivity. At that time it served its purpose. The modern way to
deal with power saving for these scenarios, is by using runtime PM.
Nowadays, several host drivers have deployed runtime PM, but for those
that haven't and which still cares power saving at request inactivity,
it's certainly time to deploy runtime PM as it has been around for several
years now.
To simplify code to mmc core and thus decrease maintenance efforts, this
patch removes all code related to MMC_CLKGATE.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Some sdio wifi chips will not work properly if we try to start new
sdio-rw requests while the device is signalling that it is busy.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
As mmc_claim_host() invokes pm_runtime_get_sync() for the mmc host device,
it's important that the host is kept claimed for *all* accesses to it via
the host_ops callbacks.
In some code paths for SDIO, particularly related to the PM support,
mmc_power_off|up() is invoked without keeping the host claimed. Let's fix
these.
Moreover, mmc_start|stop_host() also invokes mmc_power_off|up() without
claiming the host, let's fix these as well.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Kishon Vijay Abraham I <kishon@ti.com>
When get a CRC error, start the mmc_retune, it will issue CMD19/CMD21
to do tune, assume there were 10 clock phase need to try, phase 0 to
phase 6 is ok, phase 7 to phase 9 is NG, we try it from 0 to 9, so
the last CMD19/CMD21 will get CRC error, host->need_retune was set and
cause mmc_retune was called, then dead loop of mmc_retune
Signed-off-by: Chaotian Jing <chaotian.jing@mediatek.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Fixes: bd11e8bd03 ("mmc: core: Flag re-tuning is needed on CRC errors")
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The following panic is captured in ker3.14, but the issue still exists
in latest kernel.
---------------------------------------------------------------------
[ 20.738217] c0 3136 (Compiler) Unable to handle kernel NULL pointer dereference
at virtual address 00000578
......
[ 20.738499] c0 3136 (Compiler) PC is at _raw_spin_lock_irqsave+0x24/0x60
[ 20.738527] c0 3136 (Compiler) LR is at _raw_spin_lock_irqsave+0x20/0x60
[ 20.740134] c0 3136 (Compiler) Call trace:
[ 20.740165] c0 3136 (Compiler) [<ffffffc0008ee900>] _raw_spin_lock_irqsave+0x24/0x60
[ 20.740200] c0 3136 (Compiler) [<ffffffc0000dd024>] __wake_up+0x1c/0x54
[ 20.740230] c0 3136 (Compiler) [<ffffffc000639414>] mmc_wait_data_done+0x28/0x34
[ 20.740262] c0 3136 (Compiler) [<ffffffc0006391a0>] mmc_request_done+0xa4/0x220
[ 20.740314] c0 3136 (Compiler) [<ffffffc000656894>] sdhci_tasklet_finish+0xac/0x264
[ 20.740352] c0 3136 (Compiler) [<ffffffc0000a2b58>] tasklet_action+0xa0/0x158
[ 20.740382] c0 3136 (Compiler) [<ffffffc0000a2078>] __do_softirq+0x10c/0x2e4
[ 20.740411] c0 3136 (Compiler) [<ffffffc0000a24bc>] irq_exit+0x8c/0xc0
[ 20.740439] c0 3136 (Compiler) [<ffffffc00008489c>] handle_IRQ+0x48/0xac
[ 20.740469] c0 3136 (Compiler) [<ffffffc000081428>] gic_handle_irq+0x38/0x7c
----------------------------------------------------------------------
Because in SMP, "mrq" has race condition between below two paths:
path1: CPU0: <tasklet context>
static void mmc_wait_data_done(struct mmc_request *mrq)
{
mrq->host->context_info.is_done_rcv = true;
//
// If CPU0 has just finished "is_done_rcv = true" in path1, and at
// this moment, IRQ or ICache line missing happens in CPU0.
// What happens in CPU1 (path2)?
//
// If the mmcqd thread in CPU1(path2) hasn't entered to sleep mode:
// path2 would have chance to break from wait_event_interruptible
// in mmc_wait_for_data_req_done and continue to run for next
// mmc_request (mmc_blk_rw_rq_prep).
//
// Within mmc_blk_rq_prep, mrq is cleared to 0.
// If below line still gets host from "mrq" as the result of
// compiler, the panic happens as we traced.
wake_up_interruptible(&mrq->host->context_info.wait);
}
path2: CPU1: <The mmcqd thread runs mmc_queue_thread>
static int mmc_wait_for_data_req_done(...
{
...
while (1) {
wait_event_interruptible(context_info->wait,
(context_info->is_done_rcv ||
context_info->is_new_req));
static void mmc_blk_rw_rq_prep(...
{
...
memset(brq, 0, sizeof(struct mmc_blk_request));
This issue happens very coincidentally; however adding mdelay(1) in
mmc_wait_data_done as below could duplicate it easily.
static void mmc_wait_data_done(struct mmc_request *mrq)
{
mrq->host->context_info.is_done_rcv = true;
+ mdelay(1);
wake_up_interruptible(&mrq->host->context_info.wait);
}
At runtime, IRQ or ICache line missing may just happen at the same place
of the mdelay(1).
This patch gets the mmc_context_info at the beginning of function, it can
avoid this race condition.
Signed-off-by: Jialing Fu <jlfu@marvell.com>
Tested-by: Shawn Lin <shawn.lin@rock-chips.com>
Fixes: 2220eedfd7 ("mmc: fix async request mechanism ....")
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
For some mass production of kingston eMMCs which adopt Phison's
firmware will meet an unrecoverable data conrruption occasionally
if performing trim due to a firmware bug confirmed by vendor. We
found it on Intel-C3230RK platform. So we add fixup of broken trim
for it.
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
When requesting a trim for several bytes, everything up to the next
erase-group is erased. This causes data corruption.
Signed-off-by: Robin van der Gracht <robin@protonic.nl>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
In the (not so unlikely) case that the mmc controller timeout budget is
enough for exactly one erase-group, the simplification of allowing one
sector has an enormous performance penalty. We optimize this special case
by introducing a flag that prohibits erase-group boundary crossing, so
that we can allow trimming more than one sector at a time.
Signed-off-by: David Jander <david@protonic.nl>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Linus Walleij