commit cdf8a6fb48
"mmc: block: Introduce queue semantics"
deleted the last user of mmc_req_is_special() and it was
a horrible hack to classify requests as "special" or
"not special" to begin with, so delete the helper.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
This switches also the multiple-command ioctl() call to issue
all ioctl()s through the block layer instead of going directly
to the device.
We extend the passed argument with an argument count and loop
over all passed commands in the ioctl() issue function called
from the block layer.
By doing this we are again loosening the grip on the big host
lock, since two calls to mmc_get_card()/mmc_put_card() are
removed.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Avri Altman <Avri.Altman@sandisk.com>
This wraps single ioctl() commands into block requests using
the custom block layer request types REQ_OP_DRV_IN and
REQ_OP_DRV_OUT.
By doing this we are loosening the grip on the big host lock,
since two calls to mmc_get_card()/mmc_put_card() are removed.
We are storing the ioctl() in/out argument as a pointer in
the per-request struct mmc_blk_request container. Since we
now let the block layer allocate this data, blk_get_request()
will allocate it for us and we can immediately dereference
it and use it to pass the argument into the block layer.
We refactor the if/else/if/else ladder in mmc_blk_issue_rq()
as part of the job, keeping some extra attention to the
case when a NULL req is passed into this function and
making that pipeline flush more explicit.
Tested on the ux500 with the userspace:
mmc extcsd read /dev/mmcblk3
resulting in a successful EXTCSD info dump back to the
console.
This commit fixes a starvation issue in the MMC/SD stack
that can be easily provoked in the following way by
issueing the following commands in sequence:
> dd if=/dev/mmcblk3 of=/dev/null bs=1M &
> mmc extcs read /dev/mmcblk3
Before this patch, the extcsd read command would hang
(starve) while waiting for the dd command to finish since
the block layer was holding the card/host lock.
After this patch, the extcsd ioctl() command is nicely
interpersed with the rest of the block commands and we
can issue a bunch of ioctl()s from userspace while there
is some busy block IO going on without any problems.
Conversely userspace ioctl()s can no longer starve
the block layer by holding the card/host lock.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Avri Altman <Avri.Altman@sandisk.com>
The variable is_rpmb is clearly a bool and even assigned true
and false, yet declared as an int.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
The mmc_queue_req is a per-request state container the MMC core uses
to carry bounce buffers, pointers to asynchronous requests and so on.
Currently allocated as a static array of objects, then as a request
comes in, a mmc_queue_req is assigned to it, and used during the
lifetime of the request.
This is backwards compared to how other block layer drivers work:
they usally let the block core provide a per-request struct that get
allocated right beind the struct request, and which can be obtained
using the blk_mq_rq_to_pdu() helper. (The _mq_ infix in this function
name is misleading: it is used by both the old and the MQ block
layer.)
The per-request struct gets allocated to the size stored in the queue
variable .cmd_size initialized using the .init_rq_fn() and
cleaned up using .exit_rq_fn().
The block layer code makes the MMC core rely on this mechanism to
allocate the per-request mmc_queue_req state container.
Doing this make a lot of complicated queue handling go away. We only
need to keep the .qnct that keeps count of how many request are
currently being processed by the MMC layer. The MQ block layer will
replace also this once we transition to it.
Doing this refactoring is necessary to move the ioctl() operations
into custom block layer requests tagged with REQ_OP_DRV_[IN|OUT]
instead of the custom code using the BigMMCHostLock that we have
today: those require that per-request data be obtainable easily from
a request after creating a custom request with e.g.:
struct request *rq = blk_get_request(q, REQ_OP_DRV_IN, __GFP_RECLAIM);
struct mmc_queue_req *mq_rq = req_to_mq_rq(rq);
And this is not possible with the current construction, as the request
is not immediately assigned the per-request state container, but
instead it gets assigned when the request finally enters the MMC
queue, which is way too late for custom requests.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
[Ulf: Folded in the fix to drop a call to blk_cleanup_queue()]
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Heiner Kallweit <hkallweit1@gmail.com>
This option is activated by all multiplatform configs and what
not so we almost always have it turned on, and the memory it
saves is negligible, even more so moving forward. The actual
bounce buffer only gets allocated only when used, the only
thing the ifdefs are saving is a little bit of code.
It is highly improper to have this as a Kconfig option that
get turned on by Kconfig, make this a pure runtime-thing and
let the host decide whether we use bounce buffers. We add a
new property "disable_bounce" to the host struct.
Notice that mmc_queue_calc_bouncesz() already disables the
bounce buffers if host->max_segs != 1, so any arch that has a
maximum number of segments higher than 1 will have bounce
buffers disabled.
The option CONFIG_MMC_BLOCK_BOUNCE is default y so the
majority of platforms in the kernel already have it on, and
it then gets turned off at runtime since most of these have
a host->max_segs > 1. The few exceptions that have
host->max_segs == 1 and still turn off the bounce buffering
are those that disable it in their defconfig.
Those are the following:
arch/arm/configs/colibri_pxa300_defconfig
arch/arm/configs/zeus_defconfig
- Uses MMC_PXA, drivers/mmc/host/pxamci.c
- Sets host->max_segs = NR_SG, which is 1
- This needs its bounce buffer deactivated so we set
host->disable_bounce to true in the host driver
arch/arm/configs/davinci_all_defconfig
- Uses MMC_DAVINCI, drivers/mmc/host/davinci_mmc.c
- This driver sets host->max_segs to MAX_NR_SG, which is 16
- That means this driver anyways disabled bounce buffers
- No special action needed for this platform
arch/arm/configs/lpc32xx_defconfig
arch/arm/configs/nhk8815_defconfig
arch/arm/configs/u300_defconfig
- Uses MMC_ARMMMCI, drivers/mmc/host/mmci.[c|h]
- This driver by default sets host->max_segs to NR_SG,
which is 128, unless a DMA engine is used, and in that case
the number of segments are also > 1
- That means this driver already disables bounce buffers
- No special action needed for these platforms
arch/arm/configs/sama5_defconfig
- Uses MMC_SDHCI, MMC_SDHCI_PLTFM, MMC_SDHCI_OF_AT91, MMC_ATMELMCI
- Uses drivers/mmc/host/sdhci.c
- Normally sets host->max_segs to SDHCI_MAX_SEGS which is 128 and
thus disables bounce buffers
- Sets host->max_segs to 1 if SDHCI_USE_SDMA is set
- SDHCI_USE_SDMA is only set by SDHCI on PCI adapers
- That means that for this platform bounce buffers are already
disabled at runtime
- No special action needed for this platform
arch/blackfin/configs/CM-BF533_defconfig
arch/blackfin/configs/CM-BF537E_defconfig
- Uses MMC_SPI (a simple MMC card connected on SPI pins)
- Uses drivers/mmc/host/mmc_spi.c
- Sets host->max_segs to MMC_SPI_BLOCKSATONCE which is 128
- That means this platform already disables bounce buffers at
runtime
- No special action needed for these platforms
arch/mips/configs/cavium_octeon_defconfig
- Uses MMC_CAVIUM_OCTEON, drivers/mmc/host/cavium.c
- Sets host->max_segs to 16 or 1
- Setting host->disable_bounce to be sure for the 1 case
arch/mips/configs/qi_lb60_defconfig
- Uses MMC_JZ4740, drivers/mmc/host/jz4740_mmc.c
- This sets host->max_segs to 128 so bounce buffers are
already runtime disabled
- No action needed for this platform
It would be interesting to come up with a list of the platforms
that actually end up using bounce buffers. I have not been
able to infer such a list, but it occurs when
host->max_segs == 1 and the bounce buffering is not explicitly
disabled.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Make renesas_sdhi_sys_dmac.c a top-level module file that makes use of
library code supplied by renesas_sdhi_core.c
This is in order to facilitate adding other variants of SDHI;
in particular SDHI using different DMA controllers.
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
[Arnd: Fixed module build error]
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Rename the source file SDHI. A follow-up patch will make it a library
file used by a different top-level module file.
The name "renesas" is chosen as the SDHI driver is applicable to a wider
range of SoCs than SH-Mobile it seems to be a more appropriate name.
However, the SDHI driver source itself, is left as sh_mobile_sdhi to
avoid unnecessary churn.
the name "core" was chosen to reflect the desired role of this file,
to provide core functionality to the sdhi driver. A follow-up patch will
move the file into that role.
Internal symbols have also been renamed to reflect the filename change.
The .name member of struct platform_driver and parameter to
MODULE_ALIAS() have not been changed in order to avoid the complication
of potentially breaking SH SoCs which still use platform drivers.
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Rename the source file for DMA for SDHI as a follow-up to attaching
DMA code to the SDHI driver rather than the tmio_core driver.
The name "renesas" is chosen as the SDHI driver is applicable to a wider
range of SoCs than SH-Mobile it seems to be a more appropriate name.
However, the SDHI driver source itself, is left as sh_mobile_sdhi to
avoid unnecessary churn.
The name sys_dmac was chosen to reflect the type of DMA used.
Internal symbols have also been renamed to reflect the filename change.
A follow-up patch will re-organise the SDHI driver removing
the need for renesas_sdhi_get_dma_ops().
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Refactor DMA support to allow it to be provided by a set of call-backs
that are provided by a host driver. The motivation is to allow multiple
DMA implementations to be provided and instantiated at run-time.
Instantiate the existing DMA implementation from the sh_mobile_sdhi driver
which appears to match the current use-case. This has the side effect
of moving the DMA code from the tmio_core to the sh_mobile_sdhi driver.
A follow-up patch will change the source file for the SDHI DMA
implementation accordingly. Another follow-up patch will re-organise the
SDHI driver removing the need for tmio_mmc_get_dma_ops().
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Reshuffle the comment at the top of the source
dropping filenames and moving up human readable strings.
This seems to be somewhat more useful information to start the
source file with. It is also less fragile, f.e. to file renames.
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
This reverts commit a6db2c8603 ("mmc: dw_mmc: Don't allow Runtime PM for
SDIO cards")'
As dw_mmc now is capable of preventing runtime PM suspend while SDIO IRQs
are enabled, let's drop the less fine-grained method, which is preventing
runtime PM suspend for all SDIO cards - no matter of whether SDIO IRQs are
being enabled or not.
In this way we don't keep the host runtime PM resumed, unless it's really
needed, thus avoiding to waste power.
Especially when SDIO IRQs is supported via a separate out-of-band IRQ line,
which isn't defined by the SDIO standard, typically the SDIO func driver
doesn't enable SDIO IRQs via sdio_claim_irq(). So, for these cases we can
now allow the dwmmc device to be runtime PM suspended in-between requests.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
To be able to handle SDIO IRQs the dw_mmc device needs to be powered and
providing clock to the SDIO card. Therefore, we must not allow the device
to be runtime PM suspended while SDIO IRQs are enabled.
To fix this, let's increase the runtime PM usage count while the mmc core
enables SDIO IRQs. Later when the mmc core tells dw_mmc to disable SDIO
IRQs, we drop the usage count to again allow runtime PM suspend.
This now becomes the default behaviour for dw_mmc. In cases where SDIO IRQs
can be re-routed as GPIO wake-ups during runtime PM suspend, one could
potentially allow runtime PM suspend. However, that will have to be
addressed as a separate change on top of this one.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Convert to use the more lightweight method for processing SDIO IRQs, which
involves the following changes:
- Enable MMC_CAP2_SDIO_IRQ_NOTHREAD when SDIO IRQ is supported and use
sdio_signal_irq() instead of mmc_signal_sdio_irq().
- Mask the SDIO IRQ before signaling a new one to be processed.
- Implement the ->ack_sdio_irq() callback to unmask the SDIO IRQ.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
For hosts not supporting MMC_CAP2_SDIO_IRQ_NOTHREAD but MMC_CAP_SDIO_IRQ,
the SDIO IRQs are processed from a dedicated kernel thread. For these
cases, the host calls mmc_signal_sdio_irq() from its ISR to signal a new
SDIO IRQ.
Signaling an SDIO IRQ makes the host's ->enable_sdio_irq() callback to be
invoked to temporary disable the IRQs, before the kernel thread is woken up
to process it. When processing of the IRQs are completed, they are
re-enabled by the kernel thread, again via invoking the host's
->enable_sdio_irq().
The observation from this, is that the execution path is being unnecessary
complex, as the host driver already knows that it needs to temporary
disable the IRQs before signaling a new one. Moreover, replacing the kernel
thread with a work/workqueue would not only greatly simplify the code, but
also make it more robust.
To address the above problems, let's continue to build upon the support for
MMC_CAP2_SDIO_IRQ_NOTHREAD, as it already implements SDIO IRQs to be
processed without using the clumsy kernel thread and without the ping-pong
calls of the host's ->enable_sdio_irq() callback for each processed IRQ.
Therefore, let's add new API sdio_signal_irq(), which enables hosts to
signal/process SDIO IRQs by using a work/workqueue, rather than using the
kernel thread.
Add also a new host callback ->ack_sdio_irq(), which the work invokes when
the SDIO IRQs have been processed. This informs the host about when it
shall re-enable the SDIO IRQs. Potentially, we could re-use the existing
->enable_sdio_irq() callback instead of adding a new one, however it has
turned out that it's more convenient for hosts to get this information via
a separate callback.
Hosts that wants to use this new method to signal/process SDIO IRQs, must
enable MMC_CAP2_SDIO_IRQ_NOTHREAD and implement the ->ack_sdio_irq()
callback.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
In cases when MMC_CAP2_SDIO_IRQ_NOTHREAD is set, there is a minor window
for when the mmc host could call sdio_run_irqs(), while in fact an SDIO
func driver could have decided to released the SDIO IRQ via a call to
sdio_release_irq(). In this scenario, processing of the SDIO IRQs are done
even if there is none IRQ claimed, which is not what we want.
To prevent this from happen, close the window by validating that at least
one SDIO IRQs is claimed, before deciding to process them.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Douglas Anderson <dianders@chromium.org>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
In case if a pwrseq-emmc has been bound to the host, a call to
mmc_power_up() triggers an eMMC HW reset via the pwrseq_emmc's
->post_power_on() callback. This isn't really what we want, as
mmc_power_up() is called each time when resuming the card.
As a matter of fact, the current approach may also violate the eMMC spec,
as the involved delays managed in pwrseq_emmc assumes both VCC and VCCQ has
been turned on, which isn't the case for VCCQ, unless the regulator is
always on.
Fix this behaviour by aligning to the same procedure used when the mmc host
implements the ->hw_reset() callback and has the MMC_CAP_HW_RESET flag set.
In this way the eMMC HW reset is issued at card detection scan, to cope
with bogus bootloaders and in the error recovery path via the mmc specific
bus_ops->reset() callback.
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
The ->reset() callback is needed to implement a better support for eMMC HW
reset. The following changes will take advantage of the new callback.
Suggested-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Add the missing endianness conversions when printing the USB
device-descriptor idVendor and idProduct fields during probe.
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
At the end of either of the read or write loops len is always zero
and hence the non-zero check on len and return of -EIO is redundant
and can be removed.
Detected by CoverityScan, CID#114293 ("Logically dead code")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
ret is signed however is printed as unsigned fix the same.
If printed as a negative number the result is easier to read.
No functional change.
Signed-off-by: Shubhrajyoti Datta <shubhrajyoti.datta@xilinx.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
First, the logic for translating a register bit to the return code of
exar_get_direction and exar_get_value were wrong. And second, there was
a flip regarding the register bank in exar_get_direction.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Do not allocate resources on behalf of the parent device but on our own.
Otherwise, cleanup does not properly work if gpio-exar is removed but
not the parent device.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
This fixes reloading of the GPIO driver for the same platform device
instance as created by the exar UART driver: First of all, the driver
sets drvdata to its own value during probing and does not restore the
original value on exit. But this won't help anyway as the core clears
drvdata after the driver left.
Set the platform device parent instead.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Samsung pinctrl drivers update for v4.13:
1. Split drivers per ARMv7 and ARMv8 architectures because there
is no need to compile everything on each of them.
2. Fix for possible NULL-pointer dereference after memory allocation
failure.
3. Cleanups (silencing cast warnings, constify, removal of unneeded
casts, removal of modular boiler-plate).
When allocating a zeroed array of objects use devm_kcalloc() instead
of manually calculating the required size and using devm_kzalloc().
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
When the requested number of GPIO lines is 0, return -EINVAL, not
-1 which is -EPERM.
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
We currently shift bits here and there without actually explaining
what we're doing. Add some helper variables with names indicating
their purpose to improve the code readability.
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Currently we ignore the last odd range value, since each chip is
described by two values. Be more strict and require the user to
pass an even number of ranges.
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Invert the logic of the irq_enabled check and only access the private
data after the input is sanitized.
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
We're currently only checking the first character of the input to the
debugfs event files, so a string like '0sdfdsf' is valid and indicates
a falling edge event.
Be more strict and only allow '0', '1', '0\n' & '1\n'.
While we're at it: move the sanitization code before the irq_enabled
check so that we indicate an error on invalid input even if nobody is
waiting for events.
Signed-off-by: Bartosz Golaszewski <brgl@bgdev.pl>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
All IFC version >= 1.0 use 28nm technology for SRAM. Here SRAM has
a requirement to initialize before any read operation performed for
avoiding ECC Error.
So update condition check to initialize SRAM for all IFC version >= 1.0.0
Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Introduce some macros and helpers to avoid magic numbers and
rename macros/functions for clarification.
- We see '| 2' in several places. This means Data Cycle in MAP11 mode.
The Denali User's Guide says bit[1:0] of MAP11 is like follows:
b'00 = Command Cycle
b'01 = Address Cycle
b'10 = Data Cycle
So, this commit added DENALI_MAP11_{CMD,ADDR,DATA} macros.
- We see 'denali->flash_mem + 0x10' in several places, but 0x10 is a
magic number. Actually, this accesses the data port of the Host
Data/Command Interface. So, this commit added DENALI_HOST_DATA.
On the other hand, 'denali->flash_mem' gets access to the address
port, so DENALI_HOST_ADDR was also added.
- We see 'index_addr(denali, cmd, 0x1)' in denali_erase(), but 0x1
is a magic number. 0x1 means the erase operation. Replace 0x1
with DENALI_ERASE.
- Rename index_addr() to denali_host_write() for clarification
- Denali User's Guide says MAP{00,01,10,11} for access mode. Match
the macros with terminology in the IP document.
- Rename struct members as follows:
flash_bank -> active_bank (currently selected bank)
flash_reg -> reg (base address of registers)
flash_mem -> host (base address of host interface)
devnum -> devs_per_cs (devices connected in parallel)
bbtskipbytes -> oob_skip_bytes (number of bytes to skip in OOB)
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Now this driver is ready to remove NAND_SKIP_BBTSCAN.
The BBT descriptors in denali.c are equivalent to the ones in
nand_bbt.c. There is no need to duplicate the equivalent structures.
The with-oob decriptors do not work for this driver anyway.
The bbt_pattern (offs = 8) and the version (veroffs = 12) area
overlaps the ECC area. Set NAND_BBT_NO_OOB flag to use the no_oob
variant of the BBT descriptors.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
For ecc->read_page() and ecc->write_page(), it is possible to call
dma_map_single() against the given buffer. This bypasses the driver
internal bounce buffer and save the memcpy().
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Recent versions of this IP support automatic erased page detection.
If an erased page is detected on reads, the controller does not set
INTR__ECC_UNCOR_ERR, but INTR__ERASED_PAGE.
The detection of erased pages is based on the number of zeros in a
page; if the number of zeros is less than the value in the field
ERASED_THRESHOLD, the page is assumed as erased.
Please note ERASED_THRESHOLD specifies the number of zeros in a _page_
instead of an ECC chunk. Moreover, the controller does not provide a
way to know the actual number of bitflips.
Actually, an erased page (all 0xff) is not an ECC correctable pattern
on the Denali ECC engine. In other words, there may be overlap between
the following two:
[1] a bit pattern reachable from a valid payload + ECC pattern within
ecc.strength bitflips
[2] a bit pattern reachable from an erased state (all 0xff) within
ecc.strength bitflips
So, this feature may intercept ECC correctable patterns, then replace
[1] with [2].
After all, this feature can work safely only when ECC_THRESHOLD == 1,
i.e. detect erased pages without any bitflips. This should be the
case most of the time. If there is a bitflip or more, the driver will
fallback to the software method by using nand_check_erased_ecc_chunk().
Strangely enough, the driver still has to fill the buffer with 0xff
in case of INTR__ERASED_PAGE because the ECC correction engine has
already manipulated the data in the buffer before it judges erased
pages.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
The Denali IP adopts the syndrome page layout; payload and ECC are
interleaved, with BBM area always placed at the beginning of OOB.
The figure below shows the page organization for ecc->steps == 2:
|----------------| |-----------|
| | | |
| | | |
| Payload0 | | |
| | | |
| | | |
| | | |
|----------------| | in-band |
| ECC0 | | area |
|----------------| | |
| | | |
| | | |
| Payload1 | | |
| | | |
| | | |
|----------------| |-----------|
| BBM | | |
|----------------| | |
|Payload1 (cont.)| | |
|----------------| |out-of-band|
| ECC1 | | area |
|----------------| | |
| OOB free | | |
|----------------| |-----------|
The current raw / oob accessors do not take that into consideration,
so in-band and out-of-band data are transferred as stored in the
device. In the case above,
in-band: Payload0 + ECC0 + Payload1(partial)
out-of-band: BBM + Payload1(cont.) + ECC1 + OOB-free
This is wrong. As the comment block of struct nand_ecc_ctrl says,
driver callbacks must hide the specific layout used by the hardware
and always return contiguous in-band and out-of-band data.
The current implementation is completely screwed-up, so read/write
callbacks must be re-worked.
Also, it is reasonable to support PIO transfer in case DMA may not
work for some reasons. Actually, the Data DMA may not be equipped
depending on the configuration of the RTL. This can be checked by
reading the bit 4 of the FEATURES register. Even if the controller
has the DMA support, dma_set_mask() and dma_map_single() could fail.
In either case, the driver can fall back to the PIO transfer. Slower
access would be better than giving up.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
It is not a good idea to re-use macros that represent a specific
register bit field for the transfer direction.
It is true that bit 8 indicates the direction for the MAP10 pipeline
operation and the data DMA operation, but this is not valid across
the IP.
Use a simple flag (write: 1, read: 0) for the direction.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
This driver stores the currently addressed page into denali->page,
which is later read out by helper functions. While I am tackling on
this driver, I often missed to insert "denali->page = page;" where
needed. This makes page_read/write callbacks to get access to a
wrong page, which is a bug hard to figure out.
Instead, I'd rather pass the page via function argument because the
compiler's prototype checks will help to detect bugs.
For the same reason, propagate dma_addr to the DMA helpers instead
of denali->buf.dma_buf .
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
The current bank reset implementation polls the INTR_STATUS register
until interested bits are set. This is not good because:
- polling simply wastes time-slice of the thread
- The while() loop may continue eternally if no bit is set, for
example, due to the controller problem. The denali_wait_for_irq()
uses wait_for_completion_timeout(), which is safer.
We can use interrupt by moving the denali_reset_bank() call below
the interrupt setup.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
The nand_scan_ident() iterates over maxchips, and calls nand_reset()
for each. This driver currently passes the maximum number of banks
(=chip selects) supported by the controller as maxchips. So, maxchips
is typically 4 or 8. Usually, less number of NAND chips are connected
to the controller.
This can be a problem for ONFi devices. Now, this driver implements
->setup_data_interface() hook, so nand_setup_data_interface() issues
Set Features (0xEF) command, which waits until the chip returns R/B#
response. If no chip there, we know it never happens, but the driver
still ends up with waiting for a long time. It will finally bail-out
with timeout error and the driver will work with existing chips, but
unnecessary wait will give a bad user experience.
The denali_nand_reset() polls the INTR__RST_COMP and INTR__TIME_OUT
bits, but they are always set even if not NAND chip is connected to
that bank. To know the chip existence, INTR__INT_ACT bit must be
checked; this flag is set only when R/B# is toggled. Since the Reset
(0xFF) command toggles the R/B# pin, this can be used to know the
actual number of chips, and update denali->max_banks.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
The NAND_CMD_SET_FEATURES support is missing from denali_cmdfunc().
We also see /* TODO: Read OOB data */ comment.
It would be possible to add more commands along with the current
implementation, but having ->cmd_ctrl() seems a better approach from
the discussion with Boris [1].
Rely on the default ->cmdfunc() from the framework and implement the
driver's own ->cmd_ctrl().
This transition also fixes NAND_CMD_STATUS and NAND_CMD_PARAM handling.
NAND_CMD_STATUS was just faked by the register read, so the only valid
bit was the WP bit. NAND_CMD_PARAM was completely broken; not only the
command sent on the bus was NAND_CMD_STATUS instead of NAND_CMD_PARAM,
but also the driver was only reading 8 bytes, while the parameter page
contains several hundreds of bytes.
Also add ->write_byte(), which is needed for write direction commands,
->read/write_buf(16), which will be used some commits later.
->read_word() is not used for now, but the core may call it in the
future.
Now, this driver can drop nand_onfi_get_set_features_notsupp().
[1] https://lkml.org/lkml/2017/3/15/97
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Simplify the interrupt handling and fix issues:
- The register field view of INTR_EN / INTR_STATUS is different
among IP versions. The global macro DENALI_IRQ_ALL is hard-coded
for Intel platforms. The interrupt mask should be determined at
run-time depending on the running platform.
- wait_for_irq() loops do {} while() until interested flags are
asserted. The logic can be simplified.
- The spin_lock() guard seems too complex (and suspicious in a race
condition if wait_for_completion_timeout() bails out by timeout).
- denali->complete is reused again and again, but reinit_completion()
is missing. Add it.
Re-work the code to make it more robust and easier to handle.
While we are here, also rename the jump label "failed_req_irq" to
more appropriate "disable_irq".
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Handling timing parameters in a driver's own way should be avoided
because it duplicates efforts of drivers/mtd/nand/nand_timings.c
Besides, this driver hard-codes Intel specific parameters such as
CLK_X=5, CLK_MULTI=4. Taking a certain device (Samsung K9WAG08U1A)
into account by get_samsung_nand_para() is weird as well.
Now, the core framework provides .setup_data_interface() hook, which
handles timing parameters in a generic manner.
While I am working on this, I found even more issues in the current
code, so fixed the following as well:
- In recent IP versions, WE_2_RE and TWHR2 share the same register.
Likewise for ADDR_2_DATA and TCWAW, CS_SETUP_CNT and TWB. When
updating one, the other must be masked. Otherwise, the other will
be set to 0, then timing settings will be broken.
- The recent IP release expanded the ADDR_2_DATA to 7-bit wide.
This register is related to tADL. As commit 74a332e78e ("mtd:
nand: timings: Fix tADL_min for ONFI 4.0 chips") addressed, the
ONFi 4.0 increased the minimum of tADL to 400 nsec. This may not
fit in the 6-bit ADDR_2_DATA in older versions. Check the IP
revision and handle this correctly, otherwise the register value
would wrap around.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
