In CQ mode, the legacy commands will be sent after CQ is put in
halt state. But when deferred resume is enabled, the resume
happens in mmc_wait_for_req(), which will enable CQ again
overiding the caller's state. This causes legacy commands to fail.
Fix this by moving the deferred resume from all the places to
mmc_rpm_hold() which is always the first step done in both
CQ/legacy commands and thus avoids the above issue scneario.
Fix: fbb7cbf8ea (mmc: core: Add deferred resume for command queue mode)
Change-Id: I97fb9d654968b294dc1fb6986b8b9431e22233c0
Signed-off-by: Veerabhadrarao Badiganti <vbadigan@codeaurora.org>
MMC_PM_WAKE_SDIO_IRQ flag is used by platform driver(sdhci-msm
here) to determine if SDIO wakeup IRQ needs to be enabled
or not. If this flag is not removed in every mmc_resume_host,
then while suspend, platform driver will configure the sdio
wakeup IRQ even though wlan is disabled from UI.
This flag can be enabled by wlan driver before entring into suspend.
In this case since Wlan is not enable and thus not entering WoW mode,
pull up present on this SDIO GPIO line will pull the line high (which
otherwise is pulled low by wlan firmware in WoW mode and triggers
interrupt by pulling it high in WoW mode) which eventually will
trigger a false SDIO wakeup interrupt thus breaking the suspend
everytime.
Hence disable the flag everytime in mmc_resume_host and let
wlan driver enables it (-by sdio_set_host_pm_flags)
before entering into suspend if needed
for sdio wakeup part.
Change-Id: Iecf9eef4d91a741116bc8b8bb59c49ae7f2891fd
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
Power supply to card QCA9377 & QCA6574 is used for wlan and
bluetooth as well.
Now if SDIO card is switched to 1.8V signaling voltage, it will not
switch back to 3.3V without a power cycle. CMD0 will not make SD (SDIO)
card enter 3.3V signalling voltage. Since the power supply
for mentioned card is also used to power bluetooth along with
wlan, so card cannot be power cycled for wlan alone from such mmc
host
This patch adds a capability for such host which cannot power cycle
the card. Thus during resume of sdio (mmc_sdio_resume), it checks
if host has MMC_CAP2_BROKEN_PWR_CYCLE capability set. In case if this
capability is set we do mmc_sdio_power_restore.
This patch also fixes the resume path that if there is an oldcard
present and it's an UHS card then CMD11 may give no response. In that
case we need to switch only host signal voltage to 1.8V as card is
already in 1.8V.
Without this patch software assumes that card is switched to 3.3V on
power cycle and tries voltage switch sequence which fails
(because host driver will not control this power supply and thus
card will never enter into 3.3V supply). Thus UHS mode is not entered.
Change-Id: I1fae116fd3c5b3854761ac09da1ffec7f6ddfec9
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
commit 32ecd320db39bcb007679ed42f283740641b81ea upstream.
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>
Signed-off-by: Willy Tarreau <w@1wt.eu>
Do not resume the eMMC device as part of the system resume.
Only resume it when there's a request to be processed.
CRs-fixed: 972118
Change-Id: Ia4b3fcafb3fb5eabaad9d32bd776a03f45d4ab40
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Signed-off-by: Pavan Anamula <pavana@codeaurora.org>
To remove bus vote we expect clk-gating to kick in. But when SDcard is not
inserted, we are never turning the clocks on and hence gating will never
kick in. Thus bus voting will be always put to true.
Fix this by notifying the sdhci_msm about the pm_status only when card
is present(if host->bus_ops is NOT NULL) and remove bus vote in case of
error in mmc_suspend_host/mmc_resume_host.
Change-Id: I1f244e312f0716f99cdc53c991d37b102d40b8bc
Signed-off-by: Sayali Lokhande <sayalil@codeaurora.org>
This patch does following-
This adds an API(mmc_cmdq_hw_reset), for RESET_ALL
of SDHCI, power cycle mmc card and
reintialize it, which enables cmdq as well(if supported).
This acquires claim_host before calling mmc_power_restore.
mmc_power_restore should be called with claim_host acquired,
since this function is required from non-claim-host context in
cmdq error handling.
Change-Id: I31c4449dead1d4ad4e10a4822cce2298ec5fb4b6
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
Call mmc_/cmdq_post_req with tag number instead of
mrq. This is needed in error handling as part of multiple
request error handling.
Change-Id: I175432639d28378ec74669e31deb4d1667c49bb8
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
This is to ensure that clock scaling work is either cancelled
or completely done if it is already running before the MMC/SD
suspend is started. Otherwise, the system suspend may fail
unnecessarily if it cannot claim the host due to pending
clock scaling work.
Change-Id: Icc465a607451b29d51e27b733e506060ca9f9dc0
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
commit 71f8a4b81d040b3d094424197ca2f1bf811b1245 upstream.
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>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Earlier we disabled the cache during suspend, which meant a flush was
internally at the eMMC performed as well.
To simplify code we can make use of the mmc_flush_cache(), during mmc
suspend, which makes the mmc_cache_ctrl() redundant so then we can
remove it.
Change-Id: I80c6cbf9ad3e81943aee4b844c267c8e4f360257
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Acked-by: Seungwon Jeon <tgih.jun@samsung.com>
Signed-off-by: Chris Ball <chris@printf.net>
Git-commit: 1a6fd1533dfa409e48f377db1d92ef1cc6b0c4ce
Git-repo: https://android.googlesource.com/kernel/common.git
[kaushalk@codeaurora.org: Resolved context conflicts]
Signed-off-by: Kaushal Kumar <kaushalk@codeaurora.org>
Timeout handler executed on softirq context. cmdq dumpstate() api resumes
platform device and it can't be executed on softirq context. Request
completion callback schedules error handler work in case of timeout error.
This change moves CQE registers dump to the error handler callback.
Change-Id: Iea26ca5240f6031218dcf374cafcf2708df1f125
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
Currently getting status/ext_csd using debugfs or IOCTL
calls in cmdq mode is not working.
Fix it by halting the cmdq engine and making sure that
card queue is empty before issuing these cmds.
Change-Id: Idb89def9ff5c2fee6866759b9a8c652049552933
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
When CONFIG_MMC_BLOCK_DEFERRED_RESUME was enabled,
if sleep and wake up the phone, then remove SD card,
the removal event will not be detected. This change
will ensure removal detection of SD card even if it
has not been resumed if the card is removed after
waking up the phone.
Change-Id: Ib99d1915b6fc150979c77dc3fbf6af79e5558db6
Signed-off-by: xiaonian <xiaonian@codeaurora.org>
During system suspend CQE halted by queue suspend even before mmc host
suspend. When mmc host suspend calls halt helper, CQE is already halted.
This change ignores error value of halt helper.
Change-Id: I2d1456333f1d04f0e7a77485443b80a0ccaa3b4e
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
Need to cancel clock scaling work only when scaling was enabled and
the card is in the legacy (non-cq mode).
Change-Id: Ibfe00e54c495e249d0937345ffc6e6335940d21e
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
This change is needed bacause mmc_rpm_release() api is used in softirq
context, when usage counter goes to zero, it will trigger pm core runtime
suspend callback.
Change-Id: I18b1d193952a83336d6539c495509091c24e7d41
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
Add a check to mmc_reset_clk_scale_stats to check if clk_scaling
is enabled before resetting the statistics.
Change-Id: I025a150f68ab19733d3184b1573428cf788f6605
Signed-off-by: Maya Erez <merez@codeaurora.org>
Enable clock gating for cmdq by adding respective
API to clk hold/release. CMDQ also uses the same
legacy clock gating API.
Change-Id: I3d4f28cedb24ff2292ab08bdd7470358cf134dd5
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
Add clock scaling support for command queuing mode.
In CQ mode we count busy time from sending the first data request
until the completion of all the pending data requests.
This is done by starting the busy time on the first data request,
stopping it at the end of each data request and restarting it as
long as there are pending data requests.
The need to scale up/down the clocks is checked before sending each
request. Scaling the clocks up and down is done immediately in the
cmdqd context.
Before scaling the clocks the queue has to be drained and halted,
to allow sending the switch and tuning commands.
Change-Id: Ie856284261afc9a7f02489d2c9ab6a438e609f57
Signed-off-by: Maya Erez <merez@codeaurora.org>
Discard is supported in CMDQ mode only when device queue is empty.
Hence, discard commands should be sent using DCMD slot with
QBR (Queue Barrier) flag set.
Change-Id: I630091cbd94ffcdcec71626257f912c15fd2e21e
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
On error, the CMDQ engine stops processing requests. It is then
halted and error handled.
The error have been categorized as below:
1. Command error
a. time-out
- invalidate all pending tags & requeue
- reset both card & controller
b. crc
- end the error mrq
- tune
- unhalt
2. Data error
a. time-out
- invalidate all pending tags & requeue
- reset both card and controller
b. crc
- end the error mrq
- tune
- unhalt
3. RED error
This is device specific error and is not recoverable.
The card and controller are reset in this case and all
pending tags are invalidated and requeued.
Change-Id: I791d05f6b31d8f9b35a56fe85007b320c14e8b46
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Halt is a controller feature that can change the controller mode
from command-queue to legacy. This feature is very helpful in
error cases.
Change-Id: I7f1465b609afed68886256bd605d4019716964f4
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Adds flush request support to command-queue. This uses DCMD
feature of the controller for sending commands in
command-queue mode. DCMD is a direct command feature that uses
a pre-configured slot for sending commands other than Class 11.
Change-Id: Iebf6b74173dc91b0dc7230d1e87c65983d15148e
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Command queueing is defined in eMMC-5.1. It is designed for
higher performance by ensuring upto 32 requests to be serviced
at a time.
Adds read/write support for CMDQ enabled devices.
Change-Id: I136ddea8e5ca57eb4f85ca6e72c60001a7e24f78
Signed-off-by: Sujit Reddy Thumma <sthumma@codeaurora.org>
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
No need to check for valid state when we want to
scale down the clocks. As this state validity
check is to only address that clk scaling
doesnt scale up (which will send tuning
cmds) during definite cmds sequences (like
erase operation).
CRs-Fixed: 800478
Change-Id: Ia7a578434d7db8983493788ab2d7f9383d703169
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
In case card detect IRQ is triggered before mmc_add_host(), then
there could be a potential race between mmc_power_off() which gets
called from mmc_rescan() of card detect IRQ handler and
mmc_start_host() which gets called from mmc_add_host(). This may
turn off the clocks while mmc_start_host() is still running and thus
may result in an un-clocked register access.
Change-Id: I522df75ba0ad23f065127e015ad825075be94596
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
The eMMC card can go into "reset state" due to several
reasons and without host involvement,
e.g. low voltage detection.
This kind of eMMC card reset requires eMMC host to
re-initialize the eMMC card.
In case of eMMC card reset, eMMC card firmware sets its
state to "reset state" which allows CMD0 only.
In case the eMMC host is in the middle of FLUSH request it will
get timeout which will cause the driver to requeue the flush
request and try it later on.
The problem with this is that no one is going to initialize
the eMMC card to move out of "reset state", hence, the flush
request will keep on failing over and over again.
This commit adds a device reset for the case where flush request
fails due to timeout.
Change-Id: Ic268e13cb19d0cdce0070174bf7847e6253e4513
Signed-off-by: Talel Shenhar <tatias@codeaurora.org>
This change adds a read-modify-write logic to BKOPS feature enable.
It is required in order to avoid overriding other fields defined
in BKOPS_EN register.
Change-Id: I689f5cd14d9ec1bb881f503a0418026a59e6c197
Signed-off-by: Talel Shenhar <tatias@codeaurora.org>
Include one more state of MMC_LOAD_INIT when resume (which
calls for mmc clk init algorithm) is called
to notify sdhci with SDHCI_PERMORMANCE_MODE_INIT.
Change-Id: I80661d7c9a927ed16925c0b9c00ac7abec608cee
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
The unclocked access could happen in the following scenario -
1. mmcqd thread is waiting for previously submitted io to be done
2. before this io is done, it is woken up due to is_urgent event
3. mmcqd thread is running and is processing the is_urgent event
4. Just about the same time the previously submitted io is done and
it sets is_done_rcv flag, tries to wake up mmcqd thread and release
host clock so as to gate the clocks.
If step 3 and 4 happen at the same time, then clocks will be gated off
before is_urgent is completely handled. Hence, make sure to hold the
clock before invoking get_xfer_remain and stop_request host->op.
Change-Id: If03177bf5e28cc217cc8e86b10d81d4adbab78b9
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
This reverts commit 3ec1ed2a8d.
The original 10ms delay is enough to reach power on from power
up. Increase delay is workaround to resovle resume Re-init
retries time out error. The root cause is additional QDSD block
would impact SD fucntionality as they are multiplex with SD
interface. If cd-gpio is HW floating and has a pulse, QDSD 100ms
search process would be triggerred is prior to SD block and SD
interface work as QDSD instead of SD.
The fix is merged, so revert the workaround.
Change-Id: I647be754de4a6a13678cbb1456a9db88c2714f94
Signed-off-by: Guoping Yu <guopingy@codeaurora.org>
When eMMC cache is disabled, the card has to flush all the data in its
cache.
This change increases the timeout for cache flush operation and also uses
the cache flush timeout in cache disable operation.
Change-Id: I373050a99816c95e784a957de4271955a511b9f9
CRs-Fixed: 723701
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
Add a capability to selectively increase the delay between
power_up and power_on. This delay was arrived at after performing
a series of experiments to find the minimum value. It is required
in some scenarios to avoid command timeouts during resume.
Change-Id: I857e740aa20a33e08515d054e5c9bf128db229c6
Signed-off-by: Guoping Yu <guopingy@codeaurora.org>
If the card is removed, the wakelock has to be held for
an extended period of time to enable the user-space to
respond. The code to extend the wakelock has been
duplicated. This patch removes this duplicate code.
CRs-fixed: 684724
Change-Id: I96e4e9dec46ada7853724c36f0c2c6329f696e5d
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
This changes ensures that pointer to card struct is valid.
Change-Id: I11405cd7dd3a1a43fdb3010152a07c1c7944fc9d
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
The only error returned by stop_request(), is returned when there is
no current request to stop. This case is handled by mmc_stop_request()
caller.
Change-Id: I09bb30593d6accefef5d550881d87b6bf603296c
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
Fix a race condition that can lead to null pointer dereference
while the MMC transfers are going on.
1) mmc_request_done() -> mmc_wait_for_data_done ->
step1: update is_done_rcv
step2: wake_up sleeping thread (mmcqd) waiting for is_done_rcv
2) mmcqd -> mmc_wait_for_data_req_done
step4: wait for is_done_rcv or is_new_req
step5: is_new_req set from block layer context and mmcqd
is woken up
step6: let's say step1 is done, so complete the current request
step7: fetch new request and issue to host layer
step8: fetch one more request and wait for previous request to
complete
In the above execution contexts, if step4-step8 happens between step1 and
step2 a null pointer dereference is observed -
[ 29.483302] Unable to handle kernel NULL pointer dereference
at virtual address 00000488
[ 29.490366] pgd = c0004000
[ 29.493054] [00000488] *pgd=00000000
[ 29.518937] PC is at do_raw_spin_lock+0x8/0x13c
[ 29.523445] LR is at _raw_spin_lock_irqsave+0x20/0x28
[ 30.108789] [<c0339cd4>] (do_raw_spin_lock+0x8/0x13c) from
[ 30.118418] [<c095d178>] (_raw_spin_lock_irqsave+0x20/0x28) from
[ 30.127445] [<c0142ef4>] (__wake_up+0x20/0x50) from
[ 30.136124] [<c0663f70>] (mmc_request_done+0x30c/0x368) from
[ 30.145932] [<c067bd98>] (sdhci_tasklet_finish+0x130/0x13c) from
Change-Id: I9a21431b5fd9bb9bbcb5c18a9895096fe845e64b
Signed-off-by: Sujit Reddy Thumma <sthumma@codeaurora.org>
There could be a race between suspend and mmc_rescan as follows:
At resume:
-> mmc_pm_notify
-> mmc_rescan is scheduled
Suspend is triggered:
-> mmc_pm_notify
-> wait for mmc_rescan/cancel it
-> rescan acquires wakelock if card is detected
so as to give user-space a chance to respond
-> PM framework cancels suspend since wakelock is active
Hence, release the wake-lock if its active in suspend prepare
and check for rescan_disable before acquiring wake-lock.
This prefers suspend over notifying the user-space of the status
of the card, which is fine since with suspend user-space is frozen
too.
CRs-fixed: 655281
Change-Id: I83783589f9b541d3af73931e9a7b0b144a25538e
Signed-off-by: Asutosh Das <asutoshd@codeaurora.org>
Luca Stefani
LuK1337
Veerabhadrarao Badiganti
Ritesh Harjani
Matt Gumbel
Asutosh Das
Sayali Lokhande
Linux Build Service Account
Sahitya Tummala
Jialing Fu
Ulf Hansson
Konstantin Dorfman
xiaonian
Maya Erez
Talel Shenhar
Chuanxiao Dong
Guoping Yu
Sujit Reddy Thumma