On 28 August 2013 04:49, Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> No need, I took them as-is, thanks for doing this work.
Thanks Greg..
But I think the current commit log is poor and doesn't communicate the problem
well.. I have prepared another patchset with the same diff and updated
log (below),
I can resend them if you are willing to drop current version and take next one..
----------x--------------x---------------------
The current driver triggers a lockdep warning for RT kernel if
tty_flip_buffer_push() is called with uart_port->lock locked. This
never shows
up on UP kernels and comes up only on SMP kernels on RT kernels.
This happened due to following commit in rt kernels:
commit 714cc6dc551136b8f88b53361b504a5f624308b9
Author: Ingo Molnar <mingo@xxxxxxx>
Date: Fri Jul 3 08:30:01 2009 -0500
serial: 8250: Call flush_to_ldisc when the irq is threaded
Signed-off-by: Ingo Molnar <mingo@xxxxxxx>
---
drivers/tty/tty_buffer.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/tty/tty_buffer.c b/drivers/tty/tty_buffer.c
index 9121c1f..94d3215 100644
--- a/drivers/tty/tty_buffer.c
+++ b/drivers/tty/tty_buffer.c
@@ -517,10 +517,15 @@ void tty_flip_buffer_push(struct tty_port *port)
buf->tail->commit = buf->tail->used;
spin_unlock_irqrestore(&buf->lock, flags);
+#ifndef CONFIG_PREEMPT_RT_FULL
if (port->low_latency)
flush_to_ldisc(&buf->work);
else
schedule_work(&buf->work);
+#else
+ flush_to_ldisc(&buf->work);
+#endif
+
}
EXPORT_SYMBOL(tty_flip_buffer_push);
And the crash looks like this (produced with samsung.c driver):
-----
kernel BUG at
/home/arm/work/kernel/linaro/lng/lng.git/kernel/rtmutex.c:738!
Internal error: Oops - BUG: 0 [#1] PREEMPT SMP THUMB2
Modules linked in:
CPU: 0 PID: 1895 Comm: irq/85-12c20000 Not tainted 3.10.6-rt3+ #7
task: ef2d9cc0 ti: ee2b0000 task.ti: ee2b0000
PC is at rt_spin_lock_slowlock+0x18c/0x1a0
LR is at rt_spin_lock_slowlock+0x3f/0x1a0
pc : [<80406b9c>] lr : [<80406a4f>] psr: 60070033
sp : ee2b1d68 ip : 00000000 fp : 00000000
r10: ee6e641e r9 : ef2d9cc0 r8 : 00000000
r7 : ee6e7800 r6 : 00000000 r5 : ee2b0000 r4 : 806a9ccc
r3 : ef2d9cc0 r2 : 00000000 r1 : ef2d9cc0 r0 : 00000000
Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA Thumb Segment kernel
Control: 50c5387d Table: aeac406a DAC: 00000015
Process irq/85-12c20000 (pid: 1895, stack limit = 0xee2b0238)
Stack: (0xee2b1d68 to 0xee2b2000)
[snip]
[<80406b9c>] (rt_spin_lock_slowlock+0x18c/0x1a0) from
[<80288639>] (uart_start+0x1d/0x30)
[<80288639>] (uart_start+0x1d/0x30) from [<80278af5>]
(n_tty_receive_buf+0x9d/0xbf0)
[<80278af5>] (n_tty_receive_buf+0x9d/0xbf0) from [<8027bbe5>]
(flush_to_ldisc+0xb5/0xe8)
[<8027bbe5>] (flush_to_ldisc+0xb5/0xe8) from [<8028d4db>]
(s3c24xx_serial_rx_chars+0xbf/0x19c)
[<8028d4db>] (s3c24xx_serial_rx_chars+0xbf/0x19c) from
[<8028dbdd>] (s3c64xx_serial_handle_irq+0x31/0x38)
[<8028dbdd>] (s3c64xx_serial_handle_irq+0x31/0x38) from
[<80067db9>] (irq_forced_thread_fn+0x19/0x38)
[<80067db9>] (irq_forced_thread_fn+0x19/0x38) from
[<80067e91>] (irq_thread+0xb9/0x118)
[<80067e91>] (irq_thread+0xb9/0x118) from [<80033561>]
(kthread+0x6d/0x74)
[<80033561>] (kthread+0x6d/0x74) from [<8000cf3d>]
(ret_from_fork+0x11/0x20)
Code: f7ff fae4 e7ac de02 (de02) 1d23
---[ end trace 0000000000000002 ]---
note: irq/85-12c20000[1895] exited with preempt_count 1
Unable to handle kernel paging request at virtual address ffffffec
pgd = 80004000
[ffffffec] *pgd=af7fd821, *pte=00000000, *ppte=00000000
Internal error: Oops: 17 [#2] PREEMPT SMP THUMB2
Modules linked in:
CPU: 0 PID: 1895 Comm: irq/85-12c20000 Tainted: G D 3.10.6-rt3+ #7
task: ef2d9cc0 ti: ee2b0000 task.ti: ee2b0000
PC is at kthread_data+0xa/0x10
LR is at irq_thread_dtor+0x21/0x80
[snip]
[<80033886>] (kthread_data+0xa/0x10) from [<80067d1d>]
(irq_thread_dtor+0x21/0x80)
[<80067d1d>] (irq_thread_dtor+0x21/0x80) from [<8003136f>]
(task_work_run+0x73/0xac)
[<8003136f>] (task_work_run+0x73/0xac) from [<8001f619>]
(do_exit+0x215/0x6fc)
[<8001f619>] (do_exit+0x215/0x6fc) from [<8000fcf1>] (die+0x155/0x1e8)
[<8000fcf1>] (die+0x155/0x1e8) from [<80008235>]
(do_undefinstr+0x69/0x14c)
[<80008235>] (do_undefinstr+0x69/0x14c) from [<8000cac1>]
(__und_svc_finish+0x1/0x40)
Exception stack(0xee2b1d20 to 0xee2b1d68)
1d20: 00000000 ef2d9cc0 00000000 ef2d9cc0 806a9ccc ee2b0000
00000000 ee6e7800
1d40: 00000000 ef2d9cc0 ee6e641e 00000000 00000000 ee2b1d68
80406a4f 80406b9c
1d60: 60070033 ffffffff
[<8000cac1>] (__und_svc_finish+0x1/0x40) from [<80406b9c>]
(rt_spin_lock_slowlock+0x18c/0x1a0)
[<80406b9c>] (rt_spin_lock_slowlock+0x18c/0x1a0) from
[<80288639>] (uart_start+0x1d/0x30)
[<80288639>] (uart_start+0x1d/0x30) from [<80278af5>]
(n_tty_receive_buf+0x9d/0xbf0)
[<80278af5>] (n_tty_receive_buf+0x9d/0xbf0) from [<8027bbe5>]
(flush_to_ldisc+0xb5/0xe8)
[<8027bbe5>] (flush_to_ldisc+0xb5/0xe8) from [<8028d4db>]
(s3c24xx_serial_rx_chars+0xbf/0x19c)
[<8028d4db>] (s3c24xx_serial_rx_chars+0xbf/0x19c) from
[<8028dbdd>] (s3c64xx_serial_handle_irq+0x31/0x38)
[<8028dbdd>] (s3c64xx_serial_handle_irq+0x31/0x38) from
[<80067db9>] (irq_forced_thread_fn+0x19/0x38)
[<80067db9>] (irq_forced_thread_fn+0x19/0x38) from
[<80067e91>] (irq_thread+0xb9/0x118)
[<80067e91>] (irq_thread+0xb9/0x118) from [<80033561>]
(kthread+0x6d/0x74)
[<80033561>] (kthread+0x6d/0x74) from [<8000cf3d>]
(ret_from_fork+0x11/0x20)
Moreoever port->lock isn't really required before calling
tty_flip_buffer_push()
and hence its better to make changes in mainline kernel instead of
RT kernel.
Release the port lock before calling tty_flip_buffer_push() and reacquire it
after the call.
Similar stuff was already done for few other drivers in the past, like:
commit 2389b272168ceec056ca1d8a870a97fa9c26e11a
Author: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Date: Tue May 29 21:53:50 2007 +0100
[ARM] 4417/1: Serial: Fix AMBA drivers locking
Cc: Tobias Klauser <tklauser@xxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe linux-serial" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
