Skip to content

IPTS sleep improvements - #188

Open
jaydenmilne wants to merge 6 commits into
linux-surface:v7.2-surface-develfrom
jaydens-tasty-slop:ipts-fix-eintr-livelock
Open

jaydenmilne wants to merge 6 commits into
linux-surface:v7.2-surface-develfrom
jaydens-tasty-slop:ipts-fix-eintr-livelock

Conversation

@jaydenmilne

Copy link
Copy Markdown

I've had issues with IPTS blocking sleep and hibernations, resulting in a hot, dead laptop in my bag. I sent Claude after the issues then ran some soak tests, and it seems to have resolved the problem.

I had it run 339 suspend then hibernate cycles over 10 hours (most with the hibernate step simulated). None failed to freeze, there were no kernel warnings, no cases of the driver giving up, and 0 leaked HID devices. All of these issues were observed in kernel logs beforehand.

I also let it hibernate several times the past week. Every time the driver came back and it slept correctly.

Jayden Milne and others added 6 commits November 1, 2026 21:20
ipts_mei_send() calls mei_cldev_send() again for as long as it returns
-EINTR. mei_cldev_send() queues the message and then waits interruptibly
for the transfer to complete, returning -EINTR if a signal is pending.
Since the signal stays pending until the task returns to userspace, every
retry fails immediately, and the loop never terminates. The task spins in
the kernel, cannot be killed, and blocks the freezer. Every retry also
queues the message again.

On EDS v1 devices, iptsd switches the device to multitouch mode on
startup, which restarts the device from iptsd's process context. If iptsd
is stopped during that restart (e.g. by a system-sleep hook at the
intermediate wakeup of suspend-then-hibernate), it gets stuck, and the
system fails to hibernate:

  Freezing user space processes failed after 20.006 seconds (1 tasks refusing to freeze, wq_busy=0):
  task:iptsd           state:R  running task ...
   __mei_cl_send_timeout+0x197/0x360 [mei]
   ipts_mei_send+0x46/0x90 [ipts]
   ipts_cmd_send+0x8c/0xf0 [ipts]
   ipts_control_get_device_info+0x49/0xec [ipts]
   ipts_control_start.cold+0x24/0x271 [ipts]
   ipts_control_restart+0x2f/0x40 [ipts]

Return the error instead of retrying.

Signed-off-by: Jayden Milne <linux-surface@jayd.ml>
Patchset: ipts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ipts_thread_stop() waits for the completion that the thread signals right
before returning, and only then calls kthread_stop(). By that time the
thread may have fully exited, and since nothing holds a reference to its
task_struct, kthread_stop() operates on freed memory:

  refcount_t: addition on 0; use-after-free.
  WARNING: lib/refcount.c:25 at refcount_warn_saturate+0x6a/0x90
  Call Trace:
   kthread_stop+0x16c/0x180
   ipts_thread_stop+0x32/0x60 [ipts]
   ipts_receiver_stop+0x22/0x40 [ipts]
   _ipts_control_stop.part.0+0x29/0x50 [ipts]
   ipts_control_restart+0x21/0x40 [ipts]

Take a reference when starting the thread, and drop it with
kthread_stop_put().

Signed-off-by: Jayden Milne <linux-surface@jayd.ml>
Patchset: ipts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…iled

The receiver thread returns an error if the sensor does not shut down
cleanly (e.g. "READY_FOR_DATA: cmd failed: 5"). The thread has exited at
that point, but _ipts_control_stop() returns early and leaves the buffers
allocated, and ipts_control_stop() skips freeing the HID device.

MEI removes and probes the device again on every suspend and hibernation.
Since the old HID device is never destroyed, it keeps its hidraw node, the
new one gets a different node, and iptsd is left talking to the wrong
device, breaking touch until the next reboot:

  READY_FOR_DATA: cmd failed: 5
  Failed to wait for data: -53
  Failed to stop receiver loop: -53
  Failed to stop receiver: -53
  Failed to stop IPTS: -53
  hid-generic 0000:1B96:005E.0163: input,hidraw3: ...

Log the error and finish tearing down the device.

Signed-off-by: Jayden Milne <linux-surface@jayd.ml>
Patchset: ipts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
When sending a command is interrupted, or waiting for its response times
out, the ME may still answer later. That response stays in the list of
received messages until the device is removed, where it can be mistaken
for the response to a later command with the same code.

This is not just a leak. After a restart, a stale READY_FOR_DATA response
makes the event mode receiver process a buffer that has no data, and if
its status is an error, the next shutdown of the receiver fails:

  IPTS running in event mode
  Failed to process buffer: -19
  FEEDBACK: recv failed: -11
  Failed to send feedback: -11
  Stopping IPTS
  READY_FOR_DATA: cmd failed: 5
  Failed to wait for data: -53

Drop all unconsumed responses when starting the device, before sending
the first command.

Signed-off-by: Jayden Milne <linux-surface@jayd.ml>
Patchset: ipts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On EDS v1 devices, switching modes restarts the device. If the restart
fails (e.g. because the calling task has a signal pending), the device
stays stopped and the HID device inactive. Since ipts_hid_raw_request()
rejects requests while the HID device is inactive, userspace cannot
retry, and touch input stays broken until the module is reloaded or the
MEI client device is probed again.

When the restart fails, restore the previous mode and restart the device
from a work item instead, retrying a few times. Restoring the previous
mode means that the device comes back in the mode that works without
iptsd, and that iptsd can switch modes again once it is restarted.

Serialize restarts with a mutex, and make sure that no restart is queued
or running when the device is removed.

Signed-off-by: Jayden Milne <linux-surface@jayd.ml>
Patchset: ipts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…itch

After a command times out, the ME client can get stuck one response
behind: each response is delivered only when the next command is sent.
Every later command then times out as well. The restart attempts cycle
between failing on GET_DEVICE_INFO and failing on SET_MODE until they give
up, and the device stays stopped.

Waiting does not clear this, but reloading the module does. Disconnect
from the ME client and connect again before each restart attempt, like a
re-probe does. The cause of the stuck response is not known.

Signed-off-by: Jayden Milne <linux-surface@jayd.ml>
Patchset: ipts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@jaydenmilne

Copy link
Copy Markdown
Author

Sorry - realized there are some competing PRs here. I'll dig into it

@loafecho

Copy link
Copy Markdown

@jaydenmilne your three commits are in #196, rebased onto #187 with your authorship and sign-off kept; each has a [loafecho: ...] note on what changed in the rebase. Details on #187.

@jaydenmilne

Copy link
Copy Markdown
Author

Sweet - I'll build your PR and then run the soak test tonight

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants