From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-pl1-f180.google.com (mail-pl1-f180.google.com [209.85.214.180]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id D7872388E72 for ; Thu, 10 Sep 2026 16:55:47 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.214.180 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1789059351; cv=none; b=IXEJebIEMkwK99BSB8tifFYk6bhP7/uBdhPmK4ZkH4raQkCn1uqU97udu4hGz4Za1S2Y26hFxAqH4rm8pGbdBuBLmUwp9tWY7dflfxih+49gKXry3jUS2V9XtaXWhOeCULtSuk/TNCd3P4WoI3lDlK7ozljJRwuJfQ243rtAW5o= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1789059351; c=relaxed/simple; bh=rz3yn+C2eMXZq6DDOcOzTXjjZhcgJFRoGT8TBDOfLM8=; h=From:To:Cc:Subject:Date:Message-ID:In-Reply-To:References: MIME-Version; b=t9cURgzXd4cDkYbYqgZ/v+35ocKJ5lzhtVLQmto4VFrClykXzfP9mLWuCQIhQG7+RKgH8kShByeZli+ENqR2v7HWiRTWET5f4MRzfpro51AtKbIgpo5URVEx7rRvz269wwKOumMNmHypcXWBW0ca0zdsHm3JKhpv730OVC/w1fY= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com; spf=pass smtp.mailfrom=gmail.com; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b=VAHyT+iS; arc=none smtp.client-ip=209.85.214.180 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=gmail.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="VAHyT+iS" Received: by mail-pl1-f180.google.com with SMTP id d9443c01a7336-2caced6038eso178505ad.0 for ; Thu, 10 Sep 2026 09:55:47 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20251104; t=1789059347; x=1789664147; darn=vger.kernel.org; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:from:to:cc:subject:date :message-id:reply-to:content-type; bh=XhI6MBvRN0gP2RgUEHdvGpUA0bye96IQsEzTBS/VwdM=; b=VAHyT+iSi6/EwZsmASElUloGOoXv6X54zDBElHMQxmydoDshiN1jN3bPeaYtsSTvnJ 6YZi3ZFoPFBkoKy7+r7Olpa6zAxKdhZdPjG2moPL/CRR3ccL3Njc3vIfAb4ABeC00o7N J+iVGmwknW6CXluVtYuWNj4/nJGI5vO/n95Mb7AtXQ4BUKWYsqbjJyIojBuYW3/Yr7qH gjZt9EgjwUmARucjVV68GRf1zp12cL/HPGWcvLWdDC11KVqPPuCLfAyDRqn/i5yMNxdV jVvpWKAHHfbJCnzsZKJRx3lRQY0ThK2/yvYGcTiOAugwjZtQsjtDQ6n0Di9/d2Txofy4 Bsug== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1789059347; x=1789664147; h=content-transfer-encoding:mime-version:references:in-reply-to :message-id:date:subject:cc:to:from:x-gm-gg:x-gm-message-state:from :to:cc:subject:date:message-id:reply-to:content-type; bh=XhI6MBvRN0gP2RgUEHdvGpUA0bye96IQsEzTBS/VwdM=; b=cUi3MKkLQWbPz3gBR/Dr6F2YQRh7ECxMW1wQ/i+aMh1NlMAuqz074kwtDJ2CdEtXgp FiGzRqDBG8KUVCcyUZCutyl6BP8EgNDrq4FPrVBLwZ37Fk5iPsvL6bPsvVofUl5tumfe KJ0lR9Ixxwr9aFFpet6xI0r3CuGoUu6zIpmQSfyq3ZP9yO/L52ve7GRkkE+UaTjyn6QB aOmRGxy+Y6x8Uk9JPUe7wK+YA3+uZRyePKdIMTUU4ypS3v0qOmypfIwqM2V2BW9RMuWZ iJCAHtt1IdBWVeaKFfh6MlMUGRLzvhkhEMELhZvfbhoDowrNamKDYZ5oPI1JQOPJJ7Vx 3viw== X-Gm-Message-State: AFuF++kCln2nZcG7sR1DFK+EX8ExoBzURhtJA2GYsa1TMAj3s7DSGJJ7 Eltf8u1N6VJEUR4YCFVOYydTmXtmlBncbzj76ZAFJvVr7yUrY1UYlBj2 X-Gm-Gg: AYBFou2lKj1trYztceW139K1DEO90mgl/+I2ChcsMTA5E//90jDoXpsms3ln7VASpta YnQGmoMIFZWhxR3yN49KVV2/ADyFO0aNts3zyMI13FjOsb+dw4axPksy3v3driEreyjqpi2jOKB wyfqWcyH4mDTgnevy21xY5xq/zaJCh+PUpjsBryWHwDp7loyE0e5YtU59eNPtsbvhnCylRS9njs fjZSrAuyNI3QSlIielQ9gRCpqT1CaKJ4ny5YG97ASLJKDyWqSPPWv3wJ0Cwj1jgsYLBw0RX0vow FXov/cF0cXiHfgYlguY7tgm9QUstiXhBvCvIk+Y0HKOBakCO5UFZREOjrkxwSTdFcTu9on3SaQi jFFXE2t6rPw7MADMmmHO+KXAQWj0e/iwjoq6glzdYC0ZXlnnZ/YqGt5u41X+3ydwJ6hw1nhg5/B FHqGj5Hpx47HfmhvrpV5Au4mCWy1gR+/PP/d+Mv1k44MvLmeXFlZB5YsIBR8c9QlTT7+rASvurC ShmfJ/UWs/EpQWVwLA= X-Received: by 2002:a17:903:2308:b0:2d8:d4d2:d13a with SMTP id d9443c01a7336-2dd0eec8eddmr58642945ad.22.1789059347086; Thu, 10 Sep 2026 09:55:47 -0700 (PDT) Received: from thangnn-ASUS.. ([2405:4802:1d38:5c70:ced9:ef96:e152:2608]) by smtp.gmail.com with ESMTPSA id 5a478bee46e88-3369856a9f7sm32600640eec.5.2026.09.10.09.55.45 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 10 Sep 2026 09:55:46 -0700 (PDT) From: Nguyen Ngoc Thang To: Jiri Kosina , Benjamin Tissoires Cc: linux-input@vger.kernel.org, linux-kernel@vger.kernel.org, syzbot+093e05755c2d99caab91@syzkaller.appspotmail.com, Dmitry Torokhov , sashiko-bot@kernel.org, Nguyen Ngoc Thang Subject: Re: [PATCH v2] HID: multitouch: use GFP_ATOMIC for report/application/usage allocations Date: Thu, 10 Sep 2026 23:55:41 +0700 Message-ID: <20260910165541.37601-1-ngocthang2710.1999@gmail.com> X-Mailer: git-send-email 2.43.0 In-Reply-To: <20260910164229.CC7E51F0089E@smtp.kernel.org> References: <20260910164229.CC7E51F0089E@smtp.kernel.org> Precedence: bulk X-Mailing-List: linux-input@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit mt_allocate_report_data() and mt_allocate_application() devm_kzalloc() with GFP_KERNEL, but they are reachable from mt_event()/mt_report() via hid_report_raw_event(), which can run in (soft)irq context off a USB URB completion (e.g. dummy_hcd's hrtimer softirq). GFP_KERNEL there may call into fs_reclaim, which lockdep flags as an inconsistent {SOFTIRQ-ON-W} -> {IN-SOFTIRQ-W} use of fs_reclaim, alongside a "sleeping function called from invalid context" splat. Switch both to GFP_ATOMIC since neither is on a path that needs to sleep. mt_allocate_usage() is left at GFP_KERNEL: it is only reached from mt_touch_input_mapping(), which runs at probe time (hid_hw_start()), never from mt_event(), so it does not need the downgrade. While auditing that same softirq path, td->applications turned out to be walked and mutated without any lock: mt_allocate_application() can list_add_tail() to it from mt_event()/mt_report() in (soft)irq context, while mt_release_contacts() (timer callback) and mt_set_quirks() (sysfs write) each list_for_each_entry() over it from their own, independent contexts. On SMP those can run concurrently and corrupt the list. Add application_list_lock (spin_lock_irqsave, since one user is a timer/irq context) around every access to td->applications, and hold it across mt_find_application()'s whole search-then-allocate sequence so two callers can never both decide to create the same application. Reported-by: syzbot+093e05755c2d99caab91@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=093e05755c2d99caab91 Signed-off-by: Nguyen Ngoc Thang --- v3: - Add application_list_lock to serialize td->applications against the concurrent softirq/timer/sysfs accessors flagged by sashiko-bot (pre-existing race, not introduced by this patch, but fixed here since it's on the same softirq-safety path). v2: - Keep mt_allocate_usage() at GFP_KERNEL. It is only reached from mt_touch_input_mapping(), at probe time, never from mt_event(), so the GFP_ATOMIC downgrade there was unnecessary. (feedback from sashiko-bot) --- drivers/hid/hid-multitouch.c | 32 ++++++++++++++++++++++++++++++-- 1 file changed, 30 insertions(+), 2 deletions(-) diff --git a/drivers/hid/hid-multitouch.c b/drivers/hid/hid-multitouch.c index 2c41bacab1ca..dd17bf4abfae 100644 --- a/drivers/hid/hid-multitouch.c +++ b/drivers/hid/hid-multitouch.c @@ -37,6 +37,7 @@ #include #include #include +#include #include #include #include @@ -186,6 +187,11 @@ struct mt_device { bool serial_maybe; /* need to check for serial protocol */ struct list_head applications; + spinlock_t application_list_lock; /* protects applications, incl. + * concurrent add from softirq + * vs. list_for_each_entry from + * timer/sysfs context + */ struct list_head reports; }; @@ -477,6 +483,7 @@ static ssize_t mt_set_quirks(struct device *dev, struct hid_device *hdev = to_hid_device(dev); struct mt_device *td = hid_get_drvdata(hdev); struct mt_application *application; + unsigned long flags; unsigned long val; @@ -485,11 +492,13 @@ static ssize_t mt_set_quirks(struct device *dev, td->mtclass.quirks = val; + spin_lock_irqsave(&td->application_list_lock, flags); list_for_each_entry(application, &td->applications, list) { application->quirks = val; if (!application->have_contact_count) application->quirks &= ~MT_QUIRK_CONTACT_CNT_ACCURATE; } + spin_unlock_irqrestore(&td->application_list_lock, flags); return count; } @@ -604,6 +613,7 @@ static struct mt_usages *mt_allocate_usage(struct hid_device *hdev, { struct mt_usages *usage; + /* only called from mt_touch_input_mapping(), at probe time */ usage = devm_kzalloc(&hdev->dev, sizeof(*usage), GFP_KERNEL); if (!usage) return NULL; @@ -627,14 +637,16 @@ static struct mt_usages *mt_allocate_usage(struct hid_device *hdev, return usage; } +/* called with td->application_list_lock held */ static struct mt_application *mt_allocate_application(struct mt_device *td, struct hid_report *report) { unsigned int application = report->application; struct mt_application *mt_application; + /* may run from hid_report_raw_event() in (soft)irq context */ mt_application = devm_kzalloc(&td->hdev->dev, sizeof(*mt_application), - GFP_KERNEL); + GFP_ATOMIC); if (!mt_application) return NULL; @@ -667,6 +679,15 @@ static struct mt_application *mt_find_application(struct mt_device *td, { unsigned int application = report->application; struct mt_application *tmp, *mt_application = NULL; + unsigned long flags; + + /* + * Hold the lock across the search and the fallback allocation so + * a concurrent mt_release_contacts()/mt_set_quirks() never walks + * the list mid-insert, and two callers can never both decide to + * allocate the same application. + */ + spin_lock_irqsave(&td->application_list_lock, flags); list_for_each_entry(tmp, &td->applications, list) { if (application == tmp->application) { @@ -681,6 +702,8 @@ static struct mt_application *mt_find_application(struct mt_device *td, if (!mt_application) mt_application = mt_allocate_application(td, report); + spin_unlock_irqrestore(&td->application_list_lock, flags); + return mt_application; } @@ -692,7 +715,8 @@ static struct mt_report_data *mt_allocate_report_data(struct mt_device *td, struct hid_field *field; int r, n; - rdata = devm_kzalloc(&td->hdev->dev, sizeof(*rdata), GFP_KERNEL); + /* may run from hid_report_raw_event() in (soft)irq context */ + rdata = devm_kzalloc(&td->hdev->dev, sizeof(*rdata), GFP_ATOMIC); if (!rdata) return NULL; @@ -2060,6 +2084,7 @@ static void mt_release_contacts(struct hid_device *hid) struct hid_input *hidinput; struct mt_application *application; struct mt_device *td = hid_get_drvdata(hid); + unsigned long flags; list_for_each_entry(hidinput, &hid->inputs, list) { struct input_dev *input_dev = hidinput->input; @@ -2077,9 +2102,11 @@ static void mt_release_contacts(struct hid_device *hid) } } + spin_lock_irqsave(&td->application_list_lock, flags); list_for_each_entry(application, &td->applications, list) { application->num_received = 0; } + spin_unlock_irqrestore(&td->application_list_lock, flags); } static void mt_expired_timeout(struct timer_list *t) @@ -2127,6 +2154,7 @@ static int mt_probe(struct hid_device *hdev, const struct hid_device_id *id) hid_set_drvdata(hdev, td); INIT_LIST_HEAD(&td->applications); + spin_lock_init(&td->application_list_lock); INIT_LIST_HEAD(&td->reports); if (id->vendor == HID_ANY_ID && id->product == HID_ANY_ID) -- 2.43.0