From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pali =?utf-8?B?Um9ow6Fy?= Subject: Re: Potential data race in psmouse_interrupt Date: Thu, 23 Jul 2015 15:08:44 +0200 Message-ID: <20150723130844.GA29125@pali> References: Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Andrey Konovalov Cc: Dmitry Torokhov , Hans de Goede , Mathias Gottschlag , Shailendra Verma , Rusty Russell , "Luis R. Rodriguez" , Thomas Hellstrom , linux-input@vger.kernel.org, linux-kernel@vger.kernel.org, Dmitry Vyukov , Kostya Serebryany , Alexander Potapenko , ktsan@googlegroups.com List-Id: linux-input@vger.kernel.org On Wednesday 22 July 2015 15:14:42 Andrey Konovalov wrote: > Hi! >=20 > We are working on a dynamic data race detector for Linux kernel calle= d > KernelThreadSanitizer (ktsan) (https://github.com/google/ktsan/wiki). >=20 > Here is a report we got while running ktsan (upstream revision > 21bdb584af8cca7c6df3c44cba268be050a234eb). >=20 > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D > ThreadSanitizer: data-race in psmouse_interrupt >=20 > Write of size 1 by thread T1 (K1): > [] psmouse_interrupt+0xd2/0x690 > drivers/input/mouse/psmouse-base.c:337 > [] serio_interrupt+0x66/0xb0 drivers/input/serio/s= erio.c:1008 > [] i8042_interrupt+0x240/0x4d0 > drivers/input/serio/i8042.c:541 > [] handle_irq_event_percpu+0x62/0x2a0 kernel/irq/h= andle.c:143 > [] handle_irq_event+0xa2/0xf0 kernel/irq/handle.c:= 192 > [] handle_edge_irq+0x104/0x380 kernel/irq/chip.c:6= 08 > [< inlined >] handle_irq+0xf6/0x220 generic_handle_irq_desc > include/linux/irqdesc.h:146 > [] handle_irq+0xf6/0x220 arch/x86/kernel/irq_64.c:= 81 > [] do_IRQ+0x5e/0x110 arch/x86/kernel/irq.c:223 > [] ret_from_intr+0x0/0x1f arch/x86/entry/entry_64.= S:601 > [< inlined >] irq_exit+0x98/0xa0 invoke_softirq kernel/softir= q.c:350 > [] irq_exit+0x98/0xa0 kernel/softirq.c:391 > [< inlined >] smp_apic_timer_interrupt+0x63/0x80 exiting_irq > ./arch/x86/include/asm/apic.h:655 > [] smp_apic_timer_interrupt+0x63/0x80 > arch/x86/kernel/apic/apic.c:915 > [] apic_timer_interrupt+0x6b/0x70 > arch/x86/entry/entry_64.S:782 > [] kvasprintf+0x5d/0x90 lib/kasprintf.c:24 > [] kasprintf+0x4e/0x70 lib/kasprintf.c:40 > [< inlined >] module_add_driver+0x8f/0x180 make_driver_name > drivers/base/module.c:18 > [] module_add_driver+0x8f/0x180 drivers/base/modul= e.c:63 > [] bus_add_driver+0x204/0x340 drivers/base/bus.c:7= 13 > [] driver_register+0xdb/0x190 drivers/base/driver.= c:168 > [] __hid_register_driver+0xb6/0x100 > drivers/hid/hid-core.c:2677 > [] apple_driver_init+0x24/0x31 drivers/hid/hid-app= le.c:574 > [] do_one_initcall+0xb9/0x220 init/main.c:798 > [< inlined >] kernel_init_freeable+0x2ca/0x397 > do_initcall_level init/main.c:863 > [< inlined >] kernel_init_freeable+0x2ca/0x397 do_initcalls > init/main.c:871 > [< inlined >] kernel_init_freeable+0x2ca/0x397 do_basic_setup > init/main.c:890 > [] kernel_init_freeable+0x2ca/0x397 init/main.c:10= 13 > [] kernel_init+0x16/0x140 init/main.c:941 > [] ret_from_fork+0x3f/0x70 arch/x86/entry/entry_64= =2ES:526 > DBG: cpu =3D ffff88053fd9fe68 > DBG: cpu id =3D 3 >=20 > Previous read of size 1 by thread T509 (K506): > [< inlined >] cypress_send_ext_cmd+0x193/0x500 > cypress_ps2_read_cmd_status drivers/input/mouse/cypress_ps2.c:123 > [] cypress_send_ext_cmd+0x193/0x500 > drivers/input/mouse/cypress_ps2.c:192 > [] cypress_detect+0x2a/0xb0 > drivers/input/mouse/cypress_ps2.c:208 > [] psmouse_extensions+0x1f6/0x4e0 > drivers/input/mouse/psmouse-base.c:849 (discriminator 1) > [] psmouse_switch_protocol+0x2ad/0x330 > drivers/input/mouse/psmouse-base.c:1467 > [] psmouse_connect+0x2e6/0x450 > drivers/input/mouse/psmouse-base.c:1554 (discriminator 4) > [< inlined >] serio_driver_probe+0x57/0x80 > serio_connect_driver drivers/input/serio/serio.c:65 > [] serio_driver_probe+0x57/0x80 > drivers/input/serio/serio.c:796 > [< inlined >] driver_probe_device+0x1da/0x460 really_probe > drivers/base/dd.c:308 > [] driver_probe_device+0x1da/0x460 drivers/base/dd= =2Ec:414 > [] __driver_attach+0xc4/0xd0 drivers/base/dd.c:612 > [] bus_for_each_dev+0x93/0xe0 drivers/base/bus.c:3= 14 > [] driver_attach+0x33/0x50 drivers/base/dd.c:631 > [< inlined >] serio_handle_event+0x2d3/0x3b0 > serio_attach_driver drivers/input/serio/serio.c:826 > [] serio_handle_event+0x2d3/0x3b0 > drivers/input/serio/serio.c:245 > [] process_one_work+0x2a9/0x6d0 kernel/workqueue.c= :2036 > [] worker_thread+0xbf/0x710 kernel/workqueue.c:217= 0 > [] kthread+0x15c/0x180 kernel/kthread.c:207 > [] ret_from_fork+0x3f/0x70 arch/x86/entry/entry_64= =2ES:526 > DBG: cpu =3D 0 >=20 > DBG: addr: ffff88019d703909 > DBG: first offset: 1, second offset: 1 > DBG: T1 clock: {T1: 17075020, T509: 142163} > DBG: T509 clock: {T509: 142167} > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >=20 > It seems that accessing 'psmouse->pktcnt' in > 'psmouse->packet[psmouse->pktcnt++] =3D data' in 'psmouse_interrupt()= ' > is racing with 'wait_event_timeout(ps2dev->wait, (psmouse->pktcnt >=3D > pktsize), msecs_to_jiffies(CYTP_CMD_TIMEOUT))' in > 'cypress_ps2_read_cmd_status()'. > If that's the case, it's possible that 'psmouse->pktcnt' is > incremented, but no data is written yet, so '(psmouse->pktcnt >=3D > pktsize)' will evaluate to true and the following 'memcpy' will acces= s > invalid data. >=20 > Could you confirm if this is a real data race? >=20 > Thanks in advance! >=20 Hi Andrey, I would like to ask you, did not you noticed some problem from Sanitize= r with psmouse-base.c and alps.c code? There is unknown problem when alps.c receive invalid data from psmouse_interrupt/serio_interrupt/i8042_interrupt. And when I read your email with race condition & accesing invalid data in cypress driver. I'm thinking if there cannot be something like that. --=20 Pali Roh=C3=A1r pali.rohar@gmail.com