libata: prevent HSM state change race between ISR and PIO

It is possible for ata_sff_flush_pio_task() to set ap->hsm_task_state to
HSM_ST_IDLE in between the time __ata_sff_port_intr() checks for HSM_ST_IDLE
and before it calls ata_sff_hsm_move() causing ata_sff_hsm_move() to BUG().

This problem is hard to reproduce making this patch hard to verify, but this
fix will prevent the race.

I have not been able to reproduce the problem, but here is a crash dump from
a 2.6.32 kernel.

On examining the ata port's state, its hsm_task_state field has a value of HSM_ST_IDLE:

crash> struct ata_port.hsm_task_state ffff881c1121c000
  hsm_task_state = 0

Normally, this should not be possible as ata_sff_hsm_move() was called from ata_sff_host_intr(),
which checks hsm_task_state and won't call ata_sff_hsm_move() if it has a HSM_ST_IDLE value.

PID: 11053  TASK: ffff8816e846cae0  CPU: 0   COMMAND: "sshd"
 #0 [ffff88008ba03960] machine_kexec at ffffffff81038f3b
 #1 [ffff88008ba039c0] crash_kexec at ffffffff810c5d92
 #2 [ffff88008ba03a90] oops_end at ffffffff8152b510
 #3 [ffff88008ba03ac0] die at ffffffff81010e0b
 #4 [ffff88008ba03af0] do_trap at ffffffff8152ad74
 #5 [ffff88008ba03b50] do_invalid_op at ffffffff8100cf95
 #6 [ffff88008ba03bf0] invalid_op at ffffffff8100bf9b
    [exception RIP: ata_sff_hsm_move+317]
    RIP: ffffffff813a77ad  RSP: ffff88008ba03ca0  RFLAGS: 00010097
    RAX: 0000000000000000  RBX: ffff881c1121dc60  RCX: 0000000000000000
    RDX: ffff881c1121dd10  RSI: ffff881c1121dc60  RDI: ffff881c1121c000
    RBP: ffff88008ba03d00   R8: 0000000000000000   R9: 000000000000002e
    R10: 000000000001003f  R11: 000000000000009b  R12: ffff881c1121c000
    R13: 0000000000000000  R14: 0000000000000050  R15: ffff881c1121dd78
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #7 [ffff88008ba03d08] ata_sff_host_intr at ffffffff813a7fbd
 #8 [ffff88008ba03d38] ata_sff_interrupt at ffffffff813a821e
 #9 [ffff88008ba03d78] handle_IRQ_event at ffffffff810e6ec0
--- <IRQ stack> ---
    [exception RIP: pipe_poll+48]
    RIP: ffffffff81192780  RSP: ffff880f26d459b8  RFLAGS: 00000246
    RAX: 0000000000000000  RBX: ffff880f26d459c8  RCX: 0000000000000000
    RDX: 0000000000000001  RSI: 0000000000000000  RDI: ffff881a0539fa80
    RBP: ffffffff8100bb8e   R8: ffff8803b23324a0   R9: 0000000000000000
    R10: ffff880f26d45dd0  R11: 0000000000000008  R12: ffffffff8109b646
    R13: ffff880f26d45948  R14: 0000000000000246  R15: 0000000000000246
    ORIG_RAX: ffffffffffffff10  CS: 0010  SS: 0018
    RIP: 00007f26017435c3  RSP: 00007fffe020c420  RFLAGS: 00000206
    RAX: 0000000000000017  RBX: ffffffff8100b072  RCX: 00007fffe020c45c
    RDX: 00007f2604a3f120  RSI: 00007f2604a3f140  RDI: 000000000000000d
    RBP: 0000000000000000   R8: 00007fffe020e570   R9: 0101010101010101
    R10: 0000000000000000  R11: 0000000000000246  R12: 00007fffe020e5f0
    R13: 00007fffe020e5f4  R14: 00007f26045f373c  R15: 00007fffe020e5e0
    ORIG_RAX: 0000000000000017  CS: 0033  SS: 002b

Somewhere between the ata_sff_hsm_move() check and the ata_sff_host_intr() check, the value changed.
On examining the other cpus to see what else was running, another cpu was running the error handler
routines:

PID: 326    TASK: ffff881c11014aa0  CPU: 1   COMMAND: "scsi_eh_1"
 #0 [ffff88008ba27e90] crash_nmi_callback at ffffffff8102fee6
 #1 [ffff88008ba27ea0] notifier_call_chain at ffffffff8152d515
 #2 [ffff88008ba27ee0] atomic_notifier_call_chain at ffffffff8152d57a
 #3 [ffff88008ba27ef0] notify_die at ffffffff810a154e
 #4 [ffff88008ba27f20] do_nmi at ffffffff8152b1db
 #5 [ffff88008ba27f50] nmi at ffffffff8152aaa0
    [exception RIP: _spin_lock_irqsave+47]
    RIP: ffffffff8152a1ff  RSP: ffff881c11a73aa0  RFLAGS: 00000006
    RAX: 0000000000000001  RBX: ffff881c1121deb8  RCX: 0000000000000000
    RDX: 0000000000000246  RSI: 0000000000000020  RDI: ffff881c122612d8
    RBP: ffff881c11a73aa0   R8: ffff881c17083800   R9: 0000000000000000
    R10: 0000000000000000  R11: 0000000000000000  R12: ffff881c1121c000
    R13: 000000000000001f  R14: ffff881c1121dd50  R15: ffff881c1121dc60
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0000
--- <NMI exception stack> ---
 #6 [ffff881c11a73aa0] _spin_lock_irqsave at ffffffff8152a1ff
 #7 [ffff881c11a73aa8] ata_exec_internal_sg at ffffffff81396fb5
 #8 [ffff881c11a73b58] ata_exec_internal at ffffffff81397109
 #9 [ffff881c11a73bd8] atapi_eh_request_sense at ffffffff813a34eb

Before it tried to acquire a spinlock, ata_exec_internal_sg() called ata_sff_flush_pio_task().
This function will set ap->hsm_task_state to HSM_ST_IDLE, and has no locking around setting this
value. ata_sff_flush_pio_task() can then race with the interrupt handler and potentially set
HSM_ST_IDLE at a fatal moment, which will trigger a kernel BUG.

v2: Fixup comment in ata_sff_flush_pio_task()

tj: Further updated comment.  Use ap->lock instead of shost lock and
    use the [un]lock_irq variant instead of the irqsave/restore one.

Signed-off-by: David Milburn <[email protected]>
Signed-off-by: Tejun Heo <[email protected]>
Cc: [email protected]

drivers/ata/libata-sff.c

@@ -1333,7 +1333,19 @@ void ata_sff_flush_pio_task(struct ata_port *ap)
 	DPRINTK("ENTER\n");
 
 	cancel_delayed_work_sync(&ap->sff_pio_task);
+
+	/*
+	 * We wanna reset the HSM state to IDLE.  If we do so without
+	 * grabbing the port lock, critical sections protected by it which
+	 * expect the HSM state to stay stable may get surprised.  For
+	 * example, we may set IDLE in between the time
+	 * __ata_sff_port_intr() checks for HSM_ST_IDLE and before it calls
+	 * ata_sff_hsm_move() causing ata_sff_hsm_move() to BUG().
+	 */
+	spin_lock_irq(ap->lock);
 	ap->hsm_task_state = HSM_ST_IDLE;
+	spin_unlock_irq(ap->lock);
+
 	ap->sff_pio_task_link = NULL;
 
 	if (ata_msg_ctl(ap))