22 Feb, 2018
1 commit
-
commit fe0e58048f005fdce315eb4d185e5c160be4ac01 upstream.
This reverts commit 0a44697627d17a66d7dc98f17aeca07ca79c5c20.
This commit was initially intended to fix problems with hs200 and hs400
on some boards, mainly the odroid-c2. The OC2 (Rev 0.2) I have performs
well in this modes, so I could not confirm these issues.We've had several reports about the issues being still present on (some)
OC2, so apparently, this change does not do what it was supposed to do.
Maybe the eMMC signal quality is on the edge on the board. This may
explain the variability we see in term of stability, but this is just a
guess. Lowering the max_frequency to 100Mhz seems to do trick for those
affected by the issueWorse, the commit created new issues (CRC errors and hangs) on other
boards, such as the kvim 1 and 2, the p200 or the libretech-cc.According to amlogic, the Tx phase should not be tuned and left in its
default configuration, so it is best to just revert the commit.Fixes: 0a44697627d1 ("mmc: meson-gx: include tx phase in the tuning process")
Cc: # 4.14+
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson
Signed-off-by: Greg Kroah-Hartman
04 Oct, 2017
3 commits
-
It has been reported that some platforms (odroid-c2) may require
a different tx phase setting to operate at high speed (hs200 and hs400)To improve the situation, this patch includes tx phase in the tuning
process.Fixes: d341ca88eead ("mmc: meson-gx: rework tuning function")
Reported-by: Heiner Kallweit
Signed-off-by: Jerome Brunet
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
Resetting the phase when POWER_ON is set the set_ios() call means that the
phase is reset almost every time the set_ios() is called, while the
expected behavior was to reset the phase on a power cycle.This had gone unnoticed until now because in all mode (except hs400) the
tuning is done after the last to set_ios(). In such case, the tuning
result is used anyway. In HS400, there are a few calls to set_ios() after
the tuning is done, overwriting the tuning result.Resetting the phase on POWER_UP instead of POWER_ON solve the problem.
Fixes: d341ca88eead ("mmc: meson-gx: rework tuning function")
Signed-off-by: Jerome Brunet
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
Using CLK_DIVIDER_ROUND_CLOSEST is unsafe as the mmc clock could be
rounded to a rate higher the specified rate. Removing this flag ensure
that, if the rate needs to be rounded, it will be rounded down.Fixes: 51c5d8447bd7 ("MMC: meson: initial support for GX platforms")
Signed-off-by: Jerome Brunet
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson
31 Aug, 2017
1 commit
-
Using __bf_shf does not compile on arm 32 architecture.
This has gone unnoticed till now cause the driver is only used on arm64.In addition, __bf_shf was already used in the driver without any issue.
It was used on a constant value, so the call was probably optimized
away.Replace __bf_shf by __ffs fixes the problem
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson
30 Aug, 2017
16 commits
-
Rework tuning function of the rx phase. Now that the phase can be
more precisely set using CCF, test more phase setting and find the
largest working window. Then the tuning selected is the one at the
center of the window.This rework allows to use new modes, such as UHS SDR50
Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Initial default tx phase was set to 0 while the datasheet recommends 270.
Some cards fails to initialize with this setting and eMMC mode DDR52 does
not work.Changing this setting to 270 fixes these issues, without any regression
so farReviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Implement voltage switch callback (shamelessly copied from sunxi mmc
driver). This allow, with the appropriate tuning function, to use
SD ultra high speed modes.Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Several phases can be controlled on the meson-gx controller, the core, tx
and rx clock phase. The tx and rx uses delays to allow more fine grained
setting of the phase. To properly compute the phase using delays,
accessing the clock rate is necessary.Instead of ad-hoc functions, use the common clock framework to set the
clock phases (and access the clock rate while doing it).Acked-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Implement the card_busy callback to be able to verify that the
card is done dealing with voltage switch, when the support is
added later on.Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
No functional change, just improve interrupt handler readability
Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
It seems that the mmc clock is also used and required, somehow, by
the controller itself.It is shown during init, when writing to CFG while the divider is set
to 0 will crash the SoC. During a voltage switch, the controller may
crash and the card may then fail to exit busy state if the clock is
stopped.To avoid this, it is best to keep the clock running for the controller,
except during rate change. However, we still need to be able to gate
the clock out of the SoC. Let's use the pinmux for this, and fallback
to gpio mode (pulled-down) when we need to gate the clockReviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
In DDR modes, meson mmc controller requires an input rate twice as fast
as the output rateFixes: 51c5d8447bd7 ("MMC: meson: initial support for GX platforms")
Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Thanks to devm, carrying the clock structure around after init is not
necessary. Rework the function to remove these from the controller host
data.Finally, set initial mmc clock rate before enabling it, simplifying the
exit condition.Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Clean-up clk_set function to prepare the next changes (DDR and clk-stop)
Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Remove conditional write of cfg register. Warn if set_clk fails for some
reason. Consistently use host->dev instead of mixing with mmc_dev(mmc)Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
cfg init function overwrite values set in the clk init function
Remove the cfg pokes from the clk init. Actually, trying to use
the CLK_AUTO, like initially tried in clk_init, would break
the card initializationReviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
On boot, the clock divider value is 0 which is a weird unsupported value.
For example, accessing the cfg register with this value set would crash
the SoC.Previous change removed 0 as possible value for CCF but forgot to properly
initialize the register before registering the clock. This leads to the
CCF finding an illegal value, which it complains about.Initialize the register properly in a standalone patch so the fix can be
picked up if necessary. The change this fixed is: "mmc: meson-gx: remove
CLK_DIVIDER_ALLOW_ZERO clock flag".Reported-by: Neil Armstrong
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Remove unused clock rate defines. These should not be defined but
requested from the clock framework.Also correct typo on the DELAY register
Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
Remove CLK_DIVIDER_ALLOW_ZERO. This flag means that a 1 based divider
with a 0 value will behave as a bypass clockThe mmc divider does not behave like this, a 0 value disables the clock
Remove this flag so CCF never allows a 0 value on this clockFixes: 51c5d8447bd7 ("MMC: meson: initial support for GX platforms")
Reviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson -
CCF generic mux will shift the mask using the value defined in shift
Define the mask accordinglyReviewed-by: Kevin Hilman
Signed-off-by: Jerome Brunet
Signed-off-by: Ulf Hansson
12 Jun, 2017
1 commit
-
There have been reports about SDIO failing with certain WiFi chips in
descriptor chain mode. SD / eMMC are working fine.So let's fall back to bounce buffer mode for command SD_IO_RW_EXTENDED.
This was reported to fix the error.Fixes: 79ed05e329c3 "mmc: meson-gx: add support for descriptor chain mode"
Signed-off-by: Heiner Kallweit
Tested-by: Martin Blumenstingl
Signed-off-by: Ulf Hansson
25 Apr, 2017
18 commits
-
So far a bounce buffer is used to serialize the scatterlist(s).
This overhead can be avoided by switching to descriptor chain mode.
As result the performance is drastically improved. On a Odroid-C2 with
a 128 GB eMMC module raw reads reach 140 MB/s.Prerequisite for descriptor chain mode is that all scatterlist buffers
are 8 byte aligned for 64-bit DMA. That's not always the case, at least
the brcmfmac SDIO WiFi driver is known to cause problems.Therefore, for each request, check upfront whether all scatterlist
buffers are 8 byte aligned and fall back to bounce buffer mode if
that's not the case.Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson -
This patch adds basic tuning which changes the rx clock phase only
until a working setting is found.On a Odroid C2 with 128GB eMMC card and 200 MHz MMC clock only
180° rx clock phase make the system boot w/o CRC errors.With other MMC devices / clock speeds this might be different,
therefore don't change the driver config in general.When retuning skip the currently active parameter set. This avoids
the current problematic config to be chosen again if it causes CRC
errors just occasionally.Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson -
Introduce struct meson_tuning_params for storing the clock phase
configurations. There's no functional change because tx and rx
clock phase were implicitely set to CLK_PHASE_0 before.Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson -
Factor out setting the response configuration bits thus further reducing
complexity of function meson_mmc_start_cmd.Signed-off-by: Heiner Kallweit
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
Factor out setting the block size. This also contributes to reducing
complexity of function meson_mmc_start_cmd.Signed-off-by: Heiner Kallweit
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
CMD23 mode (use "set block count" command before transferring multiple
data blocks) typically is more performant as host / card know upfront
how many data blocks to expect. Therefore add support for this mode to
the driver.Signed-off-by: Heiner Kallweit
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
Currently we use a fixed timeout of 4s for all data transfers. Switch
to dynamic timeout values by making use of data->timeout_ns.Signed-off-by: Heiner Kallweit
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
So far the driver name is used as interrupt description, therefore in
/proc/interrupts it's not possible to tell which interrupt belongs to
which port. Change this by switching to NULL what causes the default
(device name) to be used. In our case that's the DT node name.Signed-off-by: Heiner Kallweit
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
Use GENMASK consistently for all bit masks and switch to using the
bitfield macros FIELD_GET and FIELD_PREP. This hides parts of the
complexity of dealing with bit fields.Signed-off-by: Heiner Kallweit
Reviewed-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
We don't need variable cmd_cfg_timeout, so remove it and simplify the
code a little.Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson
Reviewed-by: Kevin Hilman -
Centralize setting data->bytes_xfered in one place and make sure
that the number of transfererd bytes is set only if we receive
the end-of-chain interrupt and there was no error.Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson
Reviewed-by: Kevin Hilman -
Response reading can be slightly improved by doing it in just one place.
Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson
Reviewed-by: Kevin Hilman -
Simplify the code by using two variables instead of a struct sd_emmc_desc.
Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson
Reviewed-by: Kevin Hilman -
Replace cmd->data with a local variable to simplify code a little.
Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson
Reviewed-by: Kevin Hilman -
The variables holding the configuration bits are zero-initialized.
So we can remove all bit resets.Signed-off-by: Heiner Kallweit
Signed-off-by: Ulf Hansson
Reviewed-by: Kevin Hilman -
This patch fixes the case where meson_mmc_probe function fails before
core_clk is enabled.
Originally, that would result in a try to disable the core clock, and if
it was not already enabled, it would result in a kernel warning.
This issue is similar to the one with div_clk.
Fix it by introducing another error path used only between successfully
enabling the core clock, and successfully enabling the div clock.
That would ensure that core clock is disabled only if it was enabled before.Signed-off-by: Michał Zegan
Signed-off-by: Ulf Hansson -
The mmc host was added in meson_mmc_probe, but never removed in
meson_mmc_remove.
Fix that by removing the host before deallocating other resources.Signed-off-by: Michał Zegan
Tested-by: Michał Zegan
Acked-by: Kevin Hilman
Signed-off-by: Ulf Hansson -
Replace timeout magic numbers with proper constants.
Signed-off-by: Heiner Kallweit
Acked-by: Kevin Hilman
Tested-by: Kevin Hilman
Signed-off-by: Ulf Hansson
