Re: (sashiko review) [PATCH] mm/damon/sysfs: dealloc repeat_call_control if damon_call() fails

[SeongJae Park]

Forwarding Sashiko review for doing discussions via mails.

# review url: https://sashiko.dev/#/patchset/20260327003224.55752-1-sj@xxxxxxxxxx
# start of sashiko.dev inline review
commit aba7f94d0accfda5ab9b573157f691de9e9acf00
Author: SeongJae Park <sj@xxxxxxxxxx>

mm/damon/sysfs: dealloc repeat_call_control if damon_call() fails

This commit attempts to fix a memory leak where repeat_call_control is
not deallocated if damon_call() fails because the kdamond stopped before
the invocation. It adds a check to free repeat_call_control when
damon_call() returns an error.

Link: https://lore.kernel.org/20260320020630.962-1-sj@xxxxxxxxxx

> diff --git a/mm/damon/sysfs.c b/mm/damon/sysfs.c
> index 6a44a2f3d8fc..eefa959aa30a 100644
> --- a/mm/damon/sysfs.c
> +++ b/mm/damon/sysfs.c
> @@ -1670,7 +1670,8 @@ static int damon_sysfs_turn_damon_on(struct damon_sysfs_kdamond *kdamond)
>  	repeat_call_control->data = kdamond;
>  	repeat_call_control->repeat = true;
>  	repeat_call_control->dealloc_on_cancel = true;
> -	damon_call(ctx, repeat_call_control);
> +	if (damon_call(ctx, repeat_call_control))
> +		kfree(repeat_call_control);

Does this fully resolve the leak of repeat_call_control?

If a sysfs writer thread calls damon_call() during kdamond shutdown,
after the worker thread in kdamond_fn() flushes pending controls but
before it sets ctx->kdamond = NULL:

kdamond_fn()
   kdamond_call(ctx, true);
   ...
   // Before ctx->kdamond = NULL

Would damon_call() add the new request to ctx->call_controls and return 0
since ctx->kdamond is not yet NULL and repeat is true?

If so, does this leave repeat_call_control in the queue and leaked since
the worker is already exiting and this check is bypassed?

Could this same window also cause a deadlock for synchronous commands like
DAMON_SYSFS_CMD_UPDATE_SCHEMES_STATS?

If a synchronous command with repeat = false is added to the queue during
this same window, would the caller hang indefinitely on
wait_for_completion() since the worker thread will never process the queue
again?

>  	return err;
>  }


# end of sashiko.dev inline review
# review url: https://sashiko.dev/#/patchset/20260327003224.55752-1-sj@xxxxxxxxxx
#
# hkml [1] generated a draft of this mail.  It can be regenerated
# using below command:
#
#     hkml patch sashiko_dev --for_forwarding \
#             20260327003224.55752-1-sj@xxxxxxxxxx
#
# [1] https://github.com/sjp38/hackermail

Sent using hkml (https://github.com/sjp38/hackermail)