There is a potential race between fuse_dev_do_write()
and request_wait_answer() contexts as shown below:
TASK 1:
__fuse_request_send():
|--spin_lock(&fiq->waitq.lock);
|--queue_request();
|--spin_unlock(&fiq->waitq.lock);
|--request_wait_answer():
|--if (test_bit(FR_SENT, &req->flags))
<gets pre-empted after it is validated true> TASK 2:
fuse_dev_do_write():
|--clears bit FR_SENT,
|--request_end():
|--sets bit FR_FINISHED
|--spin_lock(&fiq->waitq.lock);
|--list_del_init(&req->intr_entry);
|--spin_unlock(&fiq->waitq.lock);
|--fuse_put_request();
|--queue_interrupt();
<request gets queued to interrupts list>
|--wake_up_locked(&fiq->waitq);
|--wait_event_freezable();
<as FR_FINISHED is set, it returns and then
the caller frees this request>
Now, the next fuse_dev_do_read(), see interrupts list is not empty
and then calls fuse_read_interrupt() which tries to access the request
which is already free'd and gets the below crash:
[11432.401266] Unable to handle kernel paging request at virtual address
6b6b6b6b6b6b6b6b
...
[11432.456168] PC is at __list_del_entry+0x6c/0xc4
[11432.463573] LR is at fuse_dev_do_read+0x1ac/0x474
...
[11432.679999] __list_del_entry+0x6c/0xc4
[11432.687794] fuse_dev_do_read+0x1ac/0x474
[11432.693180] fuse_dev_read+0x6c/0x78
[11432.699082] __vfs_read+0xc0/0xe8
[11432.704459] vfs_read+0x90/0x108
As FR_FINISHED bit is set before deleting the intr_entry with input
queue lock in request completion path, do the testing of this flag and
queueing atomically with the same lock in queue_interrupt().
Change-Id: I84232771cf09a809806568144441fd356b87d46c
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
There is a chance for a CPR controller to use the run-time calibrated
step_quot of lower modes for higher modes. This can sometimes lead
to closed-loop voltage getting stuck at open-loop voltage. To avoid
this, configure APC0/1 and GFX CPR controllers in sdm660 to first use
the default step_quot and then later switch to the run-time calibrated
step_quot.
CRs-Fixed: 2003482
Change-Id: Ic355a3a533ab65d9e2b28c67b7fca7df1dd49d48
Signed-off-by: Tirupathi Reddy <tirupath@codeaurora.org>
There is a possible race condition when debugfs files are concurrently
accessed by multiple threads. Fix this.
CRs-Fixed: 1109420, 1109326
Change-Id: I19e9107079ac8d039b12a37ae612727f824552d4
Signed-off-by: Ankit Sharma <ansharma@codeaurora.org>
Add support for devfreq and cpufreq nodes on SDM630 target
which are needed to scale cpu and DDR frequency.
Change-Id: Ic1051e712a3a6578eb6448538e15be9acb85dc89
Signed-off-by: Santosh Mardi <gsantosh@codeaurora.org>
When an upmigrate ineligible task running on the maximum capacity CPU,
we check if it can be migrated to a lower capacity CPU in tick path.
Add a power cost based check there to prevent the task migration
from a power efficient CPU.
Change-Id: I291c62d7dbf169d5123faba5f5246ad44a7a40dd
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Compute the CPU search mask once by taking task affinity,
cpu_active_mask and cpu_isolated_mask into account and cache
it in cpu_selection_env. This prevents doing the same cpumask
operations multiple times.
Change-Id: I78f35c59e6ee9437b3a522ac7ad856c0251f81ec
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Add CONSTANT_CHARGE_CURRENT_MAX, to set and get fcc.
CRs-Fixed: 2004173
Change-Id: I3dfed0ca7b2a477e1c106955fa112ba5206e3fa4
Signed-off-by: Harry Yang <harryy@codeaurora.org>
Add new SCM ID to register listener when smcinvoke is supported. Then
TZ, depending on how the listener has been registered, will either
behave in the old way (i.e. rsp[1]=QSEE_LISTENER_ID), or the new way
(rsp[1]=app_id) when smcinvoke is supported. And qseecom will first
to register listener with new SCM ID, if failed, then fallback to
use old ID.
Change-Id: Ia27055db6ec8f26705669ee913919b29461121ca
Signed-off-by: Zhen Kong <zkong@codeaurora.org>
The value of device tree property qcom,connector-temp-max-mdegc is
passed to a wrong variable. Fix it.
CRs-Fixed: 1077217
Change-Id: I705467d91dd8e5e9ea4aa0ab231e29c512608f36
Signed-off-by: Harry Yang <harryy@codeaurora.org>
Current driver has two votables for controlling the usb current
and the usb suspend bit. This could lead to a situation where
the icl votable changes current setting while its suspended.
Changing current setting while suspended causes icl changed
interrupt storm.
Fix it by removing usb_suspend_votable and make all the suspend
voters call usb_icl_votable with 0mA to cause suspend.
Update usb_icl_votable to update current settings only when
not suspended. Moreover disable the interrupt while changing
the current settings and enable the interrupt only after the
unsuspend.
Change-Id: Iafc8ee3c934952b8131d374d0529e865424016d5
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
If LINERESET was detected during hibern8 exit, HW recovers
link back to original speed automatically by PA_INIT sequence,
hence remove the full reset.
Change-Id: I181adf4580c0febed817047a40c1d0621d9c9824
Signed-off-by: Venkat Gopalakrishnan <venkatg@codeaurora.org>
Auto hibern8 timer will get reset as part of ufshcd_vops_full_reset(),
so reenable the timer in ufshcd_probe_hba().
Change-Id: I8ff85a9fd02f4a6b527666654728c4146cb61f5e
Signed-off-by: Venkat Gopalakrishnan <venkatg@codeaurora.org>
Set the DMA mask for the SPI device appropriately.
This will alleviate usage of bounce buffers by the dma kernel library and
prevent some of the crashes due to the kernel running out of bounce
buffers.
Change-Id: I7b0b123be6e7bb1e30d0755ca8b1b72ea4e2b5f6
Signed-off-by: Girish Mahadevan <girishm@codeaurora.org>
Parallel charging is enabled only for DCP and HVDCP adapters, add
support to enable parallel charging for SDP/CDP/OCP and FLOAT chargers.
While at it, rename PL_DISABLE_HVDCP_VOTER to PL_DELAY_HVDCP_VOTER.
Change-Id: I79b28ee6fd91739fa9895959d9ca3d399506d388
Signed-off-by: Ashay Jaiswal <ashayj@codeaurora.org>
Read instance type from device tree node to know the storage type
for ICE configuration, which is used to program the corresponding
ICE register during ICE reset.
Change-Id: I0eb423fb84dc89115227abc7c5688df7ae9a4a3c
Signed-off-by: AnilKumar Chimata <anilc@codeaurora.org>
As sdcc and ufs ice nodes are common between sdm660
and sdm630, move sdcc and ufs ice nodes to common
device tree file to avoid duplication of nodes.
Change-Id: I1e018ef44759e8cc3ce201e19a1e99ba6bb601bd
Signed-off-by: Brahmaji K <bkomma@codeaurora.org>
During fast cable connect/disconnect, there is chance that
set_suspend() called as part of USB entering low power mode
and PMIC driver asking to remove pulldowns race each other.
This could cause unclocked access and results in crash. Fix this
issue by using mutex to avoid crash due to the race mentioned.
Change-Id: I1babb545d40b8761774cbdd3a57258ed6bc77398
Signed-off-by: Vijayavardhan Vennapusa <vvreddy@codeaurora.org>
If disconnect happens before connect_work done, there is a chance that
port_usb might be NULL before calling ipa_data_start_rx_tx(). This could
cause crash if it happens. Fix it by taking to local variable under
spin lock protection and check later before calling ipa_data_start_rx_tx()
for queuing RX/TX requests to USB HW.
Change-Id: I82d74e34bb4d29eac225d31dac67bf5d5bc39a79
Signed-off-by: Vijayavardhan Vennapusa <vvreddy@codeaurora.org>
Shared rails such as CX, MX are controlled by the RPM. RPM aggregates
the votes to these rails and sets the rails at a corner voltage to
satisfy the needs of all the masters.
Add support to read and print the CX, MX rail-stats from shared memory.
cat /d/rpm_rail_stats
Number of Rails:2
rail:mx num_corners:4 current_corner:2 last_entered:0
corner:ret time:0
corner:svs time:0
corner:nom time:0
corner:turb time:0
rail:cx num_corners:5 current_corner:1 last_entered:0
corner:ret time:0
corner:svs2 time:0
corner:svs time:0
corner:nom time:0
corner:turb time:0
Change-Id: I730c4accc1bf4c152ef95f9ce9343a58022c2630
Signed-off-by: Karthik Parsha <kparsha@codeaurora.org>
Signed-off-by: Naresh Malladi <namall@codeaurora.org>
Commit b76105e359 ("usb: pd: Read sink capabilities from device
tree") added support in usbpd driver to pass sink capablities from
device tree. But it introduced bug in case of no sink capablities
passed from device tree as in this case, device_property_read_u32_array()
returns -EINVAL. Due to this, usbpd driver probe fails as usbpd_create()
returns errors. Fix this issue by checking if return value > 0 and then
only try to read sink capablities from device tree. If not, use default
sink capabliites defined in usbpd driver.
Change-Id: I51beeba282c6e067881c664d38c2657843a1c716
Signed-off-by: Vijayavardhan Vennapusa <vvreddy@codeaurora.org>
Home key is designed on the PCB level as a compatible solution, but
there is no mechanical hard key installed. Currently, GPIO11 on PM660
has been assigned the interrupt owership to modem, and this will cause
probe failure on gpio-keys driver and fail other gpio keys detection.
So, remove the home key configuration as a temporary solution.
CRs-Fixed: 2006468
Change-Id: Ib42a79e60c0cdf1489a63ba66ce7cde97c0e3f47
Signed-off-by: Yingwei Zhao <cyizhao@codeaurora.org>
