GitLab

Enforce full include path for includes. Deprecate old paths.

The following folders inside include/lib have been left unchanged:

- include/lib/cpus/${ARCH}
- include/lib/el3_runtime/${ARCH}

The reason for this change is that having a global namespace for
includes isn't a good idea. It defeats one of the advantages of having
folders and it introduces problems that are sometimes subtle (because
you may not know the header you are actually including if there are two
of them).

For example, this patch had to be created because two headers were
called the same way: e0ea0928 ("Fix gpio includes of mt8173 platform
to avoid collision."). More recently, this patch has had similar
problems: 46f9b2c3 ("drivers: add tzc380 support").

This problem was introduced in commit 4ecca339

 ("Move include and
source files to logical locations"). At that time, there weren't too
many headers so it wasn't a real issue. However, time has shown that
this creates problems.

Platforms that want to preserve the way they include headers may add the
removed paths to PLAT_INCLUDES, but this is discouraged.

Change-Id: I39dc53ed98f9e297a5966e723d1936d6ccf2fc8f
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>

Running TF-A 2.0 and later seems to cause a regression on HiKey 620.

    NOTICE:  BL2: v2.0(release):v2.0
    NOTICE:  BL2: Built : 17:41:23, Dec 17 2018
    NOTICE:  acpu_dvfs_set_freq: set acpu freq success!ERROR:   CMD1 failed after 100 retries
    ERROR:   BL2: Failed to load image (-5)

The reason seems to be that during emmc enumeration when BL2 sends the command
    OCR_SECTOR_MODE | OCR_VDD_MIN_2V7 | OCR_VDD_MIN_1V7

it for some reason takes some more time to get a reply. So a delay with
mdelay(1), seems to not be enough any longer and therefore we increase it to
mdelay(10) instead which makes the device boot as expected again.
Signed-off-by: Joakim Bech <joakim.bech@linaro.org>

Signed-off-by: Yann Gautier <yann.gautier@st.com>

EXT_CSD command needs to access data from eMMC device. Add the
operation of polling eMMC device status. Make sure the command is
finished.
Signed-off-by: Haojian Zhuang <haojian.zhuang@linaro.org>

Respect official response type and update response to follow
official specification.
All the MMC_RESPONSE_R(_x) are replaced with each corresponding define.

Partly revert 2a82a9c9

 for dw_mmc.c:
Responses R1, R1B and R5 have CRC.
Signed-off-by: Lionel Debieve <lionel.debieve@st.com>
Signed-off-by: Yann Gautier <yann.gautier@st.com>

The Poplar is broken on eMMC initialization because of commit
2a82a9c9 ("drivers: emmc: dw_mmc: Add response flag into response ID
definition").  It changes the driver behavior on response type handling
in dw_send_cmd(), because MMC_RESPONSE_R(2) and MMC_RESPONSE_R2 are
different things.  MMC core is still sending the former while we already
changed to check the latter in dw_mmc driver.

This patch fixes R2 response type in MMC core code.  It's the same
thing as what commit 94522ff7

 ("drivers: mmc: Fix R3 response type
definition") does for R3 response.

With this fix, Poplar is back to work.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>

Add missing response type for SWITCH command and STOP_TRANSMISSION
so that controller can be configured accordingly.

[bod: ported this change from Jun's eMMC patches to the MMC driver]
Signed-off-by: Jun Nie <jun.nie@linaro.org>
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

The R3 response type definition should be (1 << 0). Make sure we define the
expected response code in the appropriate fashion.
Signed-off-by: Bryan O'Donoghue <bryan.odonoghue@linaro.org>

Because of -Werror, this causes a build error.

Change-Id: I37a8c4bbfe3f2ced5e17981a2814985919ad483b
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>

It should set buswidth and speed of mmc controller before accessing
mmc.
Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Haojian Zhuang <haojian.zhuang@linaro.org>

DMA is always used in mmc driver. So the buffer address should
always follow the DMA limitation.

There're same requirement in mmc_read_blocks()/mmc_write_blocks()
on parameter buf. Since parameter buf comes from io_block driver,
it's already handled in io_block driver.

At here, just make the minimum address alignment on 16 chars.
Signed-off-by: Haojian Zhuang <haojian.zhuang@linaro.org>

mmc_read_blocks()/mmc_write_blocks() derived from io_block_ops_t
type. It means that lba param should be integer type, not
unsigned integer type.
Signed-off-by: Haojian Zhuang <haojian.zhuang@linaro.org>

Sending CMD8 before CMD1 just causes to fetch data failure in eMMC.
Check whether it's eMMC first. If it's eMMC, send CMD1 command instead.
Signed-off-by: Haojian Zhuang <haojian.zhuang@linaro.org>

A new function mmc_reset_to_idle is also created.
Signed-off-by: Yann Gautier <yann.gautier@st.com>

This change is largely based on existing eMMC framework by Haojian Zhuang
(@hzhuang1).

The MMC framework supports both eMMC and SD card devices. It was
written as a new framework since breaking few eMMC framework APIs.

At card probe and after the reset to idle command (CMD0), a Send
Interface Condition Command is sent (CMD8) to distinguish between
eMMC and SD card devices. eMMC devices go through the same
sequence as in the former eMMC framework. Else the framework
uses commands dedicated to SD-cards for init or frequency switch.

A structure is created to share info with the driver. It stores:
- the MMC type (eMMC, SD or SD HC)
- the device size
- the max frequency supported by the device
- the block size: 512 for eMMC and SD-HC and read from CSD
 structure for older SD-cards

Restriction to align buffers on block size has been removed.
Cache maintenance was removed and is expected to be done in the platform
or device driver.

The MMC framework includes some MISRA compliance coding style
maybe not yet ported in the existing eMMC framework.

Fixes ARM-software/tf-issues#597
Signed-off-by: Yann Gautier <yann.gautier@st.com>