slot ranges from 0 to 31, left shifting 1 by slot is undefined when slot
is 31
Change-Id: I0c2e6d278ff593cee07736627cd87692f45e2da9
Signed-off-by: Anand Saminathan <anans@google.com>
Remove blanket assumption of empty data segment area to avoid
issues while writing descriptors.
Signed-off-by: Rohit Ner <rohitner@google.com>
Change-Id: If6ff0426c44c949af1065082ed8a23ed936b5f3e
Host must only set UICCMD if HCS.UCRDY is set to 1.
At present, SW polls for UCRDY only before sending DME_GET.
Generalise this behaviour for DME_SET, DME_LINKSTARTUP,
DME_HIBERNATE_EXIT by moving polling logic inside ufshc_send_uic_cmd.
Signed-off-by: Rohit Ner <rohitner@google.com>
Change-Id: Iece777f803a660fdd144a073834c221e889371a6
Adds a common function to poll for interrupt status which reports errors
and returns error codes
Signed-off-by: Anand Saminathan <anans@google.com>
Change-Id: Ie5df036dc979e984871de4ae7e974b994296ca4c
The PRDT length field contains the count of the entries in the PRDT. See
JEDEC Standard No. 223E, section 6.1.1, "UTP Transfer Request
Descriptor," page 66. Previously we were setting the PRDT length field
to the number of bytes in the PRDT divided by four (the size in units of
32 bits). This was incorrect according to the spec.
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: I960771e6ce57002872392993042fae9ec505447e
Previously, if the image being read exceeded 12,800 KB (or 50 PRDT
entries of size 256 KB), the UFS driver would not flush the entire
Physical Region Descriptor Table (PRDT). This would cause the UFS host
controller to read empty PRDT entries, which eventually would make the
system crash. This change updates the UFS driver to flush the entire
PRDT, irrespective of the size of the image being read.
This change also throws an error if the memory allocated for UFS
descriptors is not sufficient to hold the entire Physical Region
Descriptor Table (PRDT).
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: I291dc62748992481be3cc156ce1474a6e3990ea9
Currently the UFS driver places the Command UPIU, Response UPIU, and
PRDT immediately after the UTP Transfer Request Descriptor. This space
would normally be reserved for other slots in the UTP Transfer Request
List, but because we always use slot zero, the other slots in the UTP
Transfer Request List are never used and this is okay.
Because the Command UPIU, Response UPIU, and PRDT are placed inside the
UTP Transfer Request List, the UFS driver would break if two or more
slots were used at the same time. Therefore, in a sense the
get_empty_slot() function is misleading. It gives developers the
illusion that they can use two or more slots simultaneously but in
reality they cannot.
This change deletes the get_empty_slot() function and replaces it with
is_slot_available() so that only one slot can be used.
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: I57f316640a1cdd56493505ede61f3012ceb2f305
The Device Present (DP) field is set to '1' after host controller
receive 'SUCCESS' return code on the response of the DME_LINKSTARTUP UIC
CMD during host controller initialization.
JEDEC Standard No. 223E
Page 28
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: I9db0374c1df3530d64187b9e449cde3b27d63072
Unit Attention Condition (UAC) gets set on a warm reset. Sending any
command (other than INQUIRY and REPORT LUNs) clears UAC, so its good to
add some retries when UAC is encountered
Signed-off-by: Anand Saminathan <anans@google.com>
Change-Id: Ia03b916d68565d0f3d25086b7f6d8c51d557b64f
Initialising these registers with header address will not work when
get_empty_slot returns anything other than 0 because these registers
should point to the starting of transfer request list instead of the
current descriptor
Change-Id: I60b3b59e2be6e2635a59b14dd1f11d93b9d95a1f
Signed-off-by: anans <anans@google.com>
The entire packet including UPIUs and PRDT is 0x400 but the controller
just looks for the header (32-bytes) from the UTRL
Change-Id: Ibd5d22b4a841c107fdf6447d598c5c600998e0f8
Signed-off-by: Anand Saminathan <anans@google.com>
utrlbau should point to header and not upiu
this is the case everywhere except for ufs_prepare_cmd
Signed-off-by: anans <anans@google.com>
Change-Id: I02695824c1409124a60e63c3a7ff3278a4dc5fa8
These polling loops are not required according to the spec
Signed-off-by: anans <anans@google.com>
Change-Id: I50d832ba495f30cc7a0553c84e58b747d51e0a4e
This change replaces the polling loop with fixed number of retries,
returns error values and handles them in ufs_enum.
Signed-off-by: Rohit Ner <rohitner@google.com>
Change-Id: Ia769ef26703c7525091e55ff46aaae4637db933c
according to the spec, the response to read attr comes in the
ts.attr.value field and not in the data segment.
Signed-off-by: anans <anans@google.com>
Change-Id: Iaf21883bb7e364fd7c7e4bccb33359367a0cf99d
ufs controller needs to be disabled if already enabled, without
this we noticed a crash at linkstartup during reinit
Signed-off-by: anans <anans@google.com>
Change-Id: I523c5d57c1d34f6404a6368ee3f364fbffd2e542
time taken for device init varies based on different devices,
instead of waiting for 200ms - we can poll on fdevice init
until it gets cleared, similar to what linux does
Change-Id: I571649231732fde0cd6d5be89b6f14fe905fcaff
Signed-off-by: anans <anans@google.com>
The result variable is not being used so it's better to delete it.
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: Icae614076ce1ba7cdc86267473d59a8bec682f6c
nutrs is set in ufs_enum (used by get_empty_slot), this will not
be assigned if UFS_FLAGS_SKIPINIT is set in flags during init and
might end up crashing read/write commands
Change-Id: I1517b69c56741fd5bf4ef0ebc1fc8738746233d7
Signed-off-by: anans <anans@google.com>
Fix software cache maintenance issues that can happen when
cpu prefetches data before DMA operations are complete.
This change fixes two cases one for ufs_read_blocks and
other for ufs_check_resp, in both cases invalidation of
buffer was done before the DMA operation completed.
This caused cpu prefetcher to bring data into cache
before DMA completed and caused UFS read failures.
The changes also removes unwanted cache operations to
local variable utrd which is not consumed by UFS host
controller and zeroing out buffer in ufs_read_capacity.
Change-Id: I9a288eb19d6705f6fa8bdb0b817a6411235fd8b6
Signed-off-by: Channagoud kadabi <kadabi@google.com>
new function for sending commands and reuses that function in the
driver, this can also be used to have retries for specific
commands in the future
Signed-off-by: anans <anans@google.com>
Change-Id: Ie01f36ff8e2df072db4d97929d293b80ed24f04b
ufs_read_blocks always zeros out the buffer before passing
to UFS for DMA. We don't need to zero out buf before reading
from UFS storage, this change remove the memset in ufs_read_blocks.
Signed-off-by: Channa Kadabi <kadabi@google.com>
Change-Id: I8029a7ea07fbd8cce29b383c80a3cfc782c5b7ec
Previously ufs_write_blocks was memsetting the write buffer before
calling ufs_prepare_cmd, causing zeros to be written to UFS. This change
deletes the memset call so the original buffer contents get written to
UFS.
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: I3299f11b30e6d7d409408ce11a6759c88607ee18
The utrd struct is allocated on the stack by ufs_check_resp's caller.
Invalidating the utrd struct is unnecessary since it's only read from,
and can cause other values stored on the stack (e.g. link register) to
be inadvertently invalidated.
Change-Id: Icd455b52beb2677fafc083d68d0bfa0645b7194b
Signed-off-by: Wing Li <wingers@google.com>
This change aims to make the UFS code more robust by performing a
controller reset if linkstartup fails. This idea was borrowed from
Linux's ufshcd_link_startup function.
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: I6b52148d1bf155b11198dc82a39b1120057adaaf
This change aims to make the UFS code more robust by adding retry logic
and timeouts to ufshc_reset. We also define a new function
ufshc_hce_enable for Host Controller Enable (HCE). The inner and outer
retry pattern is based on Linux's ufshcd_hba_execute_hce function.
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: I9403a5a25d3ca50af5f2f9a65b774f6a2d7a9626
This change aims to make the UFS code more robust by removing asserts
and adding retry logic. We also reduce repetition by reusing
ufshc_send_uic_cmd for DME_GET and DME_SET commands.
Signed-off-by: Jorge Troncoso <jatron@google.com>
Change-Id: Id70aa1687d5ca78dc7d47234372255ac5a04a612
We've seen issues with some THG based UFS chips, where
after reset the LUNs don't always enumerate properly.
After some debugging, we found that extending the mdelay
here seems to resolve the issue by giving the chips enough
time to complete reset.
Change-Id: I848f810b2438ed6ad3d33db614c61d2cef9ac400
Signed-off-by: John Stultz <john.stultz@linaro.org>
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: e0ea0928d5 ("Fix gpio includes of mt8173 platform
to avoid collision."). More recently, this patch has had similar
problems: 46f9b2c3a2 ("drivers: add tzc380 support").
This problem was introduced in commit 4ecca33988 ("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>
Hynix ufs has deviations on hi36xx platform which will result
in ufs bursts transfer failures at a very low probability.
To fix the problem, the Hynix device must set the register
VS_DebugSaveConfigTime to 0x10, which will set time reference
for SaveConfigTime is 250 ns. The time reference for SaveConfigTime
is 40 ns by default.
Signed-off-by: fengbaopeng <fengbaopeng@hisilicon.com>
If UFS device is initialized, we could just make it out of
hibernation by UFS_FLAGS_SKIPINIT. And vendor's dirver is always
focus on PHY setting. We could use UFS driver directly if it
exits from hibernation.
There're eight LUNs in UFS device. The UFS driver only provides
the read/write API with LUN. User could define his own read/write
since user may want to access different LUNs.
Signed-off-by: Haojian Zhuang <haojian.zhuang@linaro.org>
