Patchwork [3/3] ARM: at91/tclib: mask interruptions at shutdown and probe

login
register
mail settings
Submitter Gaël PORTAY
Date Aug. 19, 2014, 10:07 p.m.
Message ID <1408486072-19268-4-git-send-email-gael.portay@gmail.com>
Download mbox | patch
Permalink /patch/381506/
State Not Applicable
Headers show

Comments

Gaël PORTAY - Aug. 19, 2014, 10:07 p.m.
Shutdown properly the timer counter block by masking interruptions. Otherwise,
a segmentation may happen when kexec-ing a new kernel (see backtrace below).
An interruption may happen before the handler is set, leading to a kernel
segmentation fault.

Furthermore, we make sure the interruptions are masked when the driver is
initialized. This will prevent freshly kexec-ed kernel from crashing when
launched from a kernel which does not properly mask interruptions at shutdown.

The backtrace below happened after kexec-ing a new kernel, from a kernel
that did not shut down properly leaving interruptions unmasked.

Unable to handle kernel NULL pointer dereference at virtual address 00000000
pgd = c0004000
[00000000] *pgd=00000000
Internal error: Oops: 80000005 [#1] ARM
Modules linked in:
CPU: 0 PID: 1 Comm: swapper Not tainted 3.16.0+ #144
task: c1828aa0 ti: c182a000 task.ti: c182a000
PC is at 0x0
LR is at ch2_irq+0x28/0x30
pc : [<00000000>]    lr : [<c01db904>]    psr: 000000d3
sp : c182bd38  ip : c182bd48  fp : c182bd44
r10: c0373390  r9 : c1825b00  r8 : 60000053
r7 : 00000000  r6 : 00000000  r5 : 00000013  r4 : c036e800
r3 : 00000000  r2 : 00002004  r1 : c036e760  r0 : c036e760
Flags: nzcv  IRQs off  FIQs off  Mode SVC_32  ISA ARM  Segment kernel
Control: 0005317f  Table: 20004000  DAC: 00000017
Process swapper (pid: 1, stack limit = 0xc182a1c0)
Stack: (0xc182bd38 to 0xc182c000)
bd20:                                                       c182bd7c c182bd48
bd40: c0045430 c01db8ec 00000000 c18c6f40 c182bd74 c1825b00 c035cec4 00000000
bd60: c182be2c 60000053 c1825b34 00000000 c182bd94 c182bd80 c0045570 c0045408
bd80: 00000000 c1825b00 c182bdac c182bd98 c0047f34 c0045550 00000013 c036619c
bda0: c182bdc4 c182bdb0 c0044da4 c0047e98 0000007f 00000013 c182bde4 c182bdc8
bdc0: c0009e34 c0044d8c fefff000 c0046728 60000053 ffffffff c182bdf4 c182bde8
bde0: c00086a8 c0009ddc c182be74 c182bdf8 c000cb80 c0008674 00000000 00000013
be00: 00000000 00014200 c1825b00 c036e800 00000013 c035ed98 60000053 c1825b34
be20: 00000000 c182be74 c182be20 c182be40 c0047994 c0046728 60000053 ffffffff
be40: 00000013 c036e800 c182be64 c1825b00 00000013 c036e800 c035ed98 c03874bc
be60: 00000004 c036e700 c182be94 c182be78 c004689c c0046398 c036e760 c18c6080
be80: 00000000 c035ed10 c182bedc c182be98 c0348b08 c004684c 0000000c c034dac8
bea0: 004c4b3f c028c338 c036e760 00000013 c014ecc8 c18e67e0 c035b9c0 c0348884
bec0: c035b9c0 c182a020 00000000 00000000 c182bf54 c182bee0 c00089fc c0348894
bee0: c00da51c c1ffcc78 c182bf0c c182bef8 c002d100 c002d09c c1ffcc78 00000000
bf00: c182bf54 c182bf10 c002d308 c0336570 c182bf3c c0334e44 00000003 00000003
bf20: 00000030 c0334b44 c0044d74 00000003 00000003 c034dac8 c0350a94 c0373440
bf40: c0373440 00000030 c182bf94 c182bf58 c0336d24 c000890c 00000003 00000003
bf60: c0336560 c182bf64 c182bf64 6e616e0d 00000000 c0272fc8 00000000 00000000
bf80: 00000000 00000000 c182bfac c182bf98 c0272fd8 c0336bd8 c182a000 00000000
bfa0: 00000000 c182bfb0 c00095d0 c0272fd8 00000000 00000000 00000000 00000000
bfc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
bfe0: 00000000 00000000 00000000 00000000 00000013 00000000 374d27cd 33cc33e4
Backtrace:
[<c01db8dc>] (ch2_irq) from [<c0045430>] (handle_irq_event_percpu+0x38/0x148)
[<c00453f8>] (handle_irq_event_percpu) from [<c0045570>] (handle_irq_event+0x30/0x40)
 r10:00000000 r9:c1825b34 r8:60000053 r7:c182be2c r6:00000000 r5:c035cec4
 r4:c1825b00
[<c0045540>] (handle_irq_event) from [<c0047f34>] (handle_fasteoi_irq+0xac/0x11c)
 r4:c1825b00 r3:00000000
[<c0047e88>] (handle_fasteoi_irq) from [<c0044da4>] (generic_handle_irq+0x28/0x38)
 r5:c036619c r4:00000013
[<c0044d7c>] (generic_handle_irq) from [<c0009e34>] (handle_IRQ+0x68/0x88)
 r4:00000013 r3:0000007f
[<c0009dcc>] (handle_IRQ) from [<c00086a8>] (at91_aic_handle_irq+0x44/0x4c)
 r6:ffffffff r5:60000053 r4:c0046728 r3:fefff000
[<c0008664>] (at91_aic_handle_irq) from [<c000cb80>] (__irq_svc+0x40/0x4c)
Exception stack(0xc182bdf8 to 0xc182be40)
bde0:                                                       00000000 00000013
be00: 00000000 00014200 c1825b00 c036e800 00000013 c035ed98 60000053 c1825b34
be20: 00000000 c182be74 c182be20 c182be40 c0047994 c0046728 60000053 ffffffff
[<c0046388>] (__setup_irq) from [<c004689c>] (setup_irq+0x60/0x8c)
 r10:c036e700 r9:00000004 r8:c03874bc r7:c035ed98 r6:c036e800 r5:00000013
 r4:c1825b00
[<c004683c>] (setup_irq) from [<c0348b08>] (tcb_clksrc_init+0x284/0x31c)
 r6:c035ed10 r5:00000000 r4:c18c6080 r3:c036e760
[<c0348884>] (tcb_clksrc_init) from [<c00089fc>] (do_one_initcall+0x100/0x1b4)
 r10:00000000 r9:00000000 r8:c182a020 r7:c035b9c0 r6:c0348884 r5:c035b9c0
 r4:c18e67e0
[<c00088fc>] (do_one_initcall) from [<c0336d24>] (kernel_init_freeable+0x15c/0x224)
 r9:00000030 r8:c0373440 r7:c0373440 r6:c0350a94 r5:c034dac8 r4:00000003
[<c0336bc8>] (kernel_init_freeable) from [<c0272fd8>] (kernel_init+0x10/0xec)
 r9:00000000 r8:00000000 r7:00000000 r6:00000000 r5:c0272fc8 r4:00000000
[<c0272fc8>] (kernel_init) from [<c00095d0>] (ret_from_fork+0x14/0x24)
 r4:00000000 r3:c182a000
Code: bad PC value
---[ end trace 5b30f0017e282e47 ]---
Kernel panic - not syncing: Fatal exception in interrupt

Signed-off-by: Gaël PORTAY <gael.portay@gmail.com>
---
 drivers/misc/atmel_tclib.c | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)
Jean-Christophe PLAGNIOL-VILLARD - Aug. 19, 2014, 10:11 p.m.
Hi,

	This is a bit weird as the clock of the TC should be off and the irq free

	so this should never happened we need to investigate more why this append

Best Regards,
J.
On Aug 20, 2014, at 6:07 AM, Gaël PORTAY <gael.portay@gmail.com> wrote:

> 
> Shutdown properly the timer counter block by masking interruptions. Otherwise,
> a segmentation may happen when kexec-ing a new kernel (see backtrace below).
> An interruption may happen before the handler is set, leading to a kernel
> segmentation fault.
> 
> Furthermore, we make sure the interruptions are masked when the driver is
> initialized. This will prevent freshly kexec-ed kernel from crashing when
> launched from a kernel which does not properly mask interruptions at shutdown.
> 
> The backtrace below happened after kexec-ing a new kernel, from a kernel
> that did not shut down properly leaving interruptions unmasked.
> 
> Unable to handle kernel NULL pointer dereference at virtual address 00000000
> pgd = c0004000
> [00000000] *pgd=00000000
> Internal error: Oops: 80000005 [#1] ARM
> Modules linked in:
> CPU: 0 PID: 1 Comm: swapper Not tainted 3.16.0+ #144
> task: c1828aa0 ti: c182a000 task.ti: c182a000
> PC is at 0x0
> LR is at ch2_irq+0x28/0x30
> pc : [<00000000>]    lr : [<c01db904>]    psr: 000000d3
> sp : c182bd38  ip : c182bd48  fp : c182bd44
> r10: c0373390  r9 : c1825b00  r8 : 60000053
> r7 : 00000000  r6 : 00000000  r5 : 00000013  r4 : c036e800
> r3 : 00000000  r2 : 00002004  r1 : c036e760  r0 : c036e760
> Flags: nzcv  IRQs off  FIQs off  Mode SVC_32  ISA ARM  Segment kernel
> Control: 0005317f  Table: 20004000  DAC: 00000017
> Process swapper (pid: 1, stack limit = 0xc182a1c0)
> Stack: (0xc182bd38 to 0xc182c000)
> bd20:                                                       c182bd7c c182bd48
> bd40: c0045430 c01db8ec 00000000 c18c6f40 c182bd74 c1825b00 c035cec4 00000000
> bd60: c182be2c 60000053 c1825b34 00000000 c182bd94 c182bd80 c0045570 c0045408
> bd80: 00000000 c1825b00 c182bdac c182bd98 c0047f34 c0045550 00000013 c036619c
> bda0: c182bdc4 c182bdb0 c0044da4 c0047e98 0000007f 00000013 c182bde4 c182bdc8
> bdc0: c0009e34 c0044d8c fefff000 c0046728 60000053 ffffffff c182bdf4 c182bde8
> bde0: c00086a8 c0009ddc c182be74 c182bdf8 c000cb80 c0008674 00000000 00000013
> be00: 00000000 00014200 c1825b00 c036e800 00000013 c035ed98 60000053 c1825b34
> be20: 00000000 c182be74 c182be20 c182be40 c0047994 c0046728 60000053 ffffffff
> be40: 00000013 c036e800 c182be64 c1825b00 00000013 c036e800 c035ed98 c03874bc
> be60: 00000004 c036e700 c182be94 c182be78 c004689c c0046398 c036e760 c18c6080
> be80: 00000000 c035ed10 c182bedc c182be98 c0348b08 c004684c 0000000c c034dac8
> bea0: 004c4b3f c028c338 c036e760 00000013 c014ecc8 c18e67e0 c035b9c0 c0348884
> bec0: c035b9c0 c182a020 00000000 00000000 c182bf54 c182bee0 c00089fc c0348894
> bee0: c00da51c c1ffcc78 c182bf0c c182bef8 c002d100 c002d09c c1ffcc78 00000000
> bf00: c182bf54 c182bf10 c002d308 c0336570 c182bf3c c0334e44 00000003 00000003
> bf20: 00000030 c0334b44 c0044d74 00000003 00000003 c034dac8 c0350a94 c0373440
> bf40: c0373440 00000030 c182bf94 c182bf58 c0336d24 c000890c 00000003 00000003
> bf60: c0336560 c182bf64 c182bf64 6e616e0d 00000000 c0272fc8 00000000 00000000
> bf80: 00000000 00000000 c182bfac c182bf98 c0272fd8 c0336bd8 c182a000 00000000
> bfa0: 00000000 c182bfb0 c00095d0 c0272fd8 00000000 00000000 00000000 00000000
> bfc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> bfe0: 00000000 00000000 00000000 00000000 00000013 00000000 374d27cd 33cc33e4
> Backtrace:
> [<c01db8dc>] (ch2_irq) from [<c0045430>] (handle_irq_event_percpu+0x38/0x148)
> [<c00453f8>] (handle_irq_event_percpu) from [<c0045570>] (handle_irq_event+0x30/0x40)
> r10:00000000 r9:c1825b34 r8:60000053 r7:c182be2c r6:00000000 r5:c035cec4
> r4:c1825b00
> [<c0045540>] (handle_irq_event) from [<c0047f34>] (handle_fasteoi_irq+0xac/0x11c)
> r4:c1825b00 r3:00000000
> [<c0047e88>] (handle_fasteoi_irq) from [<c0044da4>] (generic_handle_irq+0x28/0x38)
> r5:c036619c r4:00000013
> [<c0044d7c>] (generic_handle_irq) from [<c0009e34>] (handle_IRQ+0x68/0x88)
> r4:00000013 r3:0000007f
> [<c0009dcc>] (handle_IRQ) from [<c00086a8>] (at91_aic_handle_irq+0x44/0x4c)
> r6:ffffffff r5:60000053 r4:c0046728 r3:fefff000
> [<c0008664>] (at91_aic_handle_irq) from [<c000cb80>] (__irq_svc+0x40/0x4c)
> Exception stack(0xc182bdf8 to 0xc182be40)
> bde0:                                                       00000000 00000013
> be00: 00000000 00014200 c1825b00 c036e800 00000013 c035ed98 60000053 c1825b34
> be20: 00000000 c182be74 c182be20 c182be40 c0047994 c0046728 60000053 ffffffff
> [<c0046388>] (__setup_irq) from [<c004689c>] (setup_irq+0x60/0x8c)
> r10:c036e700 r9:00000004 r8:c03874bc r7:c035ed98 r6:c036e800 r5:00000013
> r4:c1825b00
> [<c004683c>] (setup_irq) from [<c0348b08>] (tcb_clksrc_init+0x284/0x31c)
> r6:c035ed10 r5:00000000 r4:c18c6080 r3:c036e760
> [<c0348884>] (tcb_clksrc_init) from [<c00089fc>] (do_one_initcall+0x100/0x1b4)
> r10:00000000 r9:00000000 r8:c182a020 r7:c035b9c0 r6:c0348884 r5:c035b9c0
> r4:c18e67e0
> [<c00088fc>] (do_one_initcall) from [<c0336d24>] (kernel_init_freeable+0x15c/0x224)
> r9:00000030 r8:c0373440 r7:c0373440 r6:c0350a94 r5:c034dac8 r4:00000003
> [<c0336bc8>] (kernel_init_freeable) from [<c0272fd8>] (kernel_init+0x10/0xec)
> r9:00000000 r8:00000000 r7:00000000 r6:00000000 r5:c0272fc8 r4:00000000
> [<c0272fc8>] (kernel_init) from [<c00095d0>] (ret_from_fork+0x14/0x24)
> r4:00000000 r3:c182a000
> Code: bad PC value
> ---[ end trace 5b30f0017e282e47 ]---
> Kernel panic - not syncing: Fatal exception in interrupt
> 
> Signed-off-by: Gaël PORTAY <gael.portay@gmail.com>
> ---
> drivers/misc/atmel_tclib.c | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
> 
> diff --git a/drivers/misc/atmel_tclib.c b/drivers/misc/atmel_tclib.c
> index d505d1e..a680151 100644
> --- a/drivers/misc/atmel_tclib.c
> +++ b/drivers/misc/atmel_tclib.c
> @@ -109,6 +109,7 @@ static int __init tc_probe(struct platform_device *pdev)
> 	struct clk	*clk;
> 	int		irq;
> 	struct resource	*r;
> +	unsigned int	i;
> 
> 	irq = platform_get_irq(pdev, 0);
> 	if (irq < 0)
> @@ -157,18 +158,33 @@ static int __init tc_probe(struct platform_device *pdev)
> 	if (tc->irq[2] < 0)
> 		tc->irq[2] = irq;
> 
> +	for (i = 0; i < 3; i++)
> +		__raw_writel(0xff, tc->regs + ATMEL_TC_REG(i, IDR));
> +
> 	spin_lock(&tc_list_lock);
> 	list_add_tail(&tc->node, &tc_list);
> 	spin_unlock(&tc_list_lock);
> 
> +	platform_set_drvdata(pdev, tc);
> +
> 	return 0;
> }
> 
> +static void tc_shutdown (struct platform_device *pdev)
> +{
> +	int i;
> +	struct atmel_tc *tc = platform_get_drvdata(pdev);
> +
> +	for (i = 0; i < 3; i++)
> +		__raw_writel(0xff, tc->regs + ATMEL_TC_REG(i, IDR));
> +}
> +
> static struct platform_driver tc_driver = {
> 	.driver = {
> 		.name	= "atmel_tcb",
> 		.of_match_table	= of_match_ptr(atmel_tcb_dt_ids),
> 	},
> +	.shutdown = tc_shutdown,
> };
> 
> static int __init tc_init(void)
> -- 
> 1.9.3
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-pwm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Boris BREZILLON - Aug. 19, 2014, 11:01 p.m.
Hi Jean-Christophe,

On Wed, 20 Aug 2014 06:11:17 +0800
Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com> wrote:

> Hi,
> 
> 	This is a bit weird as the clock of the TC should be off and the irq free
> 
> 	so this should never happened we need to investigate more why this append

I may have found the source of this bug.

As Gael stated, when you're kexec-ing a new kernel your previous kernel
could be using the tbc_clksrc driver (and especially the clkevent
device). Thus the kernel might have planned a timer event and then been
asked to shutdown the machine (requested by the kexec code).
In this case the AIC interrupt connected to the TC Block is disabled
but not the interrupts within the TCB IP (IDR registers), possibly
leaving a pending interrupt before booting the new kernel.

When the tcb_clksrc driver is loaded by the new kernel it enables the
interrupt line by calling setup_irq [1] while the clockevent device is
not registered yet [2]. Thus the event_handler is still NULL when the
AIC line connected to the TCB is unmasked. Remember that an interrupt
is still pending on this HW block, which will lead to an immediate call
to the ch2_irq handler, which tries to call the event_handler, which in
turns is NULL because clkevent device registration has not taken place
at this moment => Kernel panic.
ITOH, we can't register the clkevent device before the irq handler is
set up, because we should be ready to handle clkevent request at the
time clockevents_config_and_register is called.

This leaves two solution:
 1) disable the TCB irqs (using TCB IDR registers) before calling
 setup_irq in the tcb_clksrc driver
 2) disable the TCB irqs at the tclib level (as proposed by Gael)

I prefer solution #2 because it fixes the bug for all TCB users (not
just the tcb_clksrc driver).

Best Regards,

Boris

[1]http://lxr.free-electrons.com/source/drivers/clocksource/tcb_clksrc.c#L207
[2]http://lxr.free-electrons.com/source/drivers/clocksource/tcb_clksrc.c#L211

> 
> Best Regards,
> J.
> On Aug 20, 2014, at 6:07 AM, Gaël PORTAY <gael.portay@gmail.com> wrote:
> 
> > 
> > Shutdown properly the timer counter block by masking interruptions. Otherwise,
> > a segmentation may happen when kexec-ing a new kernel (see backtrace below).
> > An interruption may happen before the handler is set, leading to a kernel
> > segmentation fault.
> > 
> > Furthermore, we make sure the interruptions are masked when the driver is
> > initialized. This will prevent freshly kexec-ed kernel from crashing when
> > launched from a kernel which does not properly mask interruptions at shutdown.
> > 
> > The backtrace below happened after kexec-ing a new kernel, from a kernel
> > that did not shut down properly leaving interruptions unmasked.
> > 
> > Unable to handle kernel NULL pointer dereference at virtual address 00000000
> > pgd = c0004000
> > [00000000] *pgd=00000000
> > Internal error: Oops: 80000005 [#1] ARM
> > Modules linked in:
> > CPU: 0 PID: 1 Comm: swapper Not tainted 3.16.0+ #144
> > task: c1828aa0 ti: c182a000 task.ti: c182a000
> > PC is at 0x0
> > LR is at ch2_irq+0x28/0x30
> > pc : [<00000000>]    lr : [<c01db904>]    psr: 000000d3
> > sp : c182bd38  ip : c182bd48  fp : c182bd44
> > r10: c0373390  r9 : c1825b00  r8 : 60000053
> > r7 : 00000000  r6 : 00000000  r5 : 00000013  r4 : c036e800
> > r3 : 00000000  r2 : 00002004  r1 : c036e760  r0 : c036e760
> > Flags: nzcv  IRQs off  FIQs off  Mode SVC_32  ISA ARM  Segment kernel
> > Control: 0005317f  Table: 20004000  DAC: 00000017
> > Process swapper (pid: 1, stack limit = 0xc182a1c0)
> > Stack: (0xc182bd38 to 0xc182c000)
> > bd20:                                                       c182bd7c c182bd48
> > bd40: c0045430 c01db8ec 00000000 c18c6f40 c182bd74 c1825b00 c035cec4 00000000
> > bd60: c182be2c 60000053 c1825b34 00000000 c182bd94 c182bd80 c0045570 c0045408
> > bd80: 00000000 c1825b00 c182bdac c182bd98 c0047f34 c0045550 00000013 c036619c
> > bda0: c182bdc4 c182bdb0 c0044da4 c0047e98 0000007f 00000013 c182bde4 c182bdc8
> > bdc0: c0009e34 c0044d8c fefff000 c0046728 60000053 ffffffff c182bdf4 c182bde8
> > bde0: c00086a8 c0009ddc c182be74 c182bdf8 c000cb80 c0008674 00000000 00000013
> > be00: 00000000 00014200 c1825b00 c036e800 00000013 c035ed98 60000053 c1825b34
> > be20: 00000000 c182be74 c182be20 c182be40 c0047994 c0046728 60000053 ffffffff
> > be40: 00000013 c036e800 c182be64 c1825b00 00000013 c036e800 c035ed98 c03874bc
> > be60: 00000004 c036e700 c182be94 c182be78 c004689c c0046398 c036e760 c18c6080
> > be80: 00000000 c035ed10 c182bedc c182be98 c0348b08 c004684c 0000000c c034dac8
> > bea0: 004c4b3f c028c338 c036e760 00000013 c014ecc8 c18e67e0 c035b9c0 c0348884
> > bec0: c035b9c0 c182a020 00000000 00000000 c182bf54 c182bee0 c00089fc c0348894
> > bee0: c00da51c c1ffcc78 c182bf0c c182bef8 c002d100 c002d09c c1ffcc78 00000000
> > bf00: c182bf54 c182bf10 c002d308 c0336570 c182bf3c c0334e44 00000003 00000003
> > bf20: 00000030 c0334b44 c0044d74 00000003 00000003 c034dac8 c0350a94 c0373440
> > bf40: c0373440 00000030 c182bf94 c182bf58 c0336d24 c000890c 00000003 00000003
> > bf60: c0336560 c182bf64 c182bf64 6e616e0d 00000000 c0272fc8 00000000 00000000
> > bf80: 00000000 00000000 c182bfac c182bf98 c0272fd8 c0336bd8 c182a000 00000000
> > bfa0: 00000000 c182bfb0 c00095d0 c0272fd8 00000000 00000000 00000000 00000000
> > bfc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
> > bfe0: 00000000 00000000 00000000 00000000 00000013 00000000 374d27cd 33cc33e4
> > Backtrace:
> > [<c01db8dc>] (ch2_irq) from [<c0045430>] (handle_irq_event_percpu+0x38/0x148)
> > [<c00453f8>] (handle_irq_event_percpu) from [<c0045570>] (handle_irq_event+0x30/0x40)
> > r10:00000000 r9:c1825b34 r8:60000053 r7:c182be2c r6:00000000 r5:c035cec4
> > r4:c1825b00
> > [<c0045540>] (handle_irq_event) from [<c0047f34>] (handle_fasteoi_irq+0xac/0x11c)
> > r4:c1825b00 r3:00000000
> > [<c0047e88>] (handle_fasteoi_irq) from [<c0044da4>] (generic_handle_irq+0x28/0x38)
> > r5:c036619c r4:00000013
> > [<c0044d7c>] (generic_handle_irq) from [<c0009e34>] (handle_IRQ+0x68/0x88)
> > r4:00000013 r3:0000007f
> > [<c0009dcc>] (handle_IRQ) from [<c00086a8>] (at91_aic_handle_irq+0x44/0x4c)
> > r6:ffffffff r5:60000053 r4:c0046728 r3:fefff000
> > [<c0008664>] (at91_aic_handle_irq) from [<c000cb80>] (__irq_svc+0x40/0x4c)
> > Exception stack(0xc182bdf8 to 0xc182be40)
> > bde0:                                                       00000000 00000013
> > be00: 00000000 00014200 c1825b00 c036e800 00000013 c035ed98 60000053 c1825b34
> > be20: 00000000 c182be74 c182be20 c182be40 c0047994 c0046728 60000053 ffffffff
> > [<c0046388>] (__setup_irq) from [<c004689c>] (setup_irq+0x60/0x8c)
> > r10:c036e700 r9:00000004 r8:c03874bc r7:c035ed98 r6:c036e800 r5:00000013
> > r4:c1825b00
> > [<c004683c>] (setup_irq) from [<c0348b08>] (tcb_clksrc_init+0x284/0x31c)
> > r6:c035ed10 r5:00000000 r4:c18c6080 r3:c036e760
> > [<c0348884>] (tcb_clksrc_init) from [<c00089fc>] (do_one_initcall+0x100/0x1b4)
> > r10:00000000 r9:00000000 r8:c182a020 r7:c035b9c0 r6:c0348884 r5:c035b9c0
> > r4:c18e67e0
> > [<c00088fc>] (do_one_initcall) from [<c0336d24>] (kernel_init_freeable+0x15c/0x224)
> > r9:00000030 r8:c0373440 r7:c0373440 r6:c0350a94 r5:c034dac8 r4:00000003
> > [<c0336bc8>] (kernel_init_freeable) from [<c0272fd8>] (kernel_init+0x10/0xec)
> > r9:00000000 r8:00000000 r7:00000000 r6:00000000 r5:c0272fc8 r4:00000000
> > [<c0272fc8>] (kernel_init) from [<c00095d0>] (ret_from_fork+0x14/0x24)
> > r4:00000000 r3:c182a000
> > Code: bad PC value
> > ---[ end trace 5b30f0017e282e47 ]---
> > Kernel panic - not syncing: Fatal exception in interrupt
> > 
> > Signed-off-by: Gaël PORTAY <gael.portay@gmail.com>
> > ---
> > drivers/misc/atmel_tclib.c | 16 ++++++++++++++++
> > 1 file changed, 16 insertions(+)
> > 
> > diff --git a/drivers/misc/atmel_tclib.c b/drivers/misc/atmel_tclib.c
> > index d505d1e..a680151 100644
> > --- a/drivers/misc/atmel_tclib.c
> > +++ b/drivers/misc/atmel_tclib.c
> > @@ -109,6 +109,7 @@ static int __init tc_probe(struct platform_device *pdev)
> > 	struct clk	*clk;
> > 	int		irq;
> > 	struct resource	*r;
> > +	unsigned int	i;
> > 
> > 	irq = platform_get_irq(pdev, 0);
> > 	if (irq < 0)
> > @@ -157,18 +158,33 @@ static int __init tc_probe(struct platform_device *pdev)
> > 	if (tc->irq[2] < 0)
> > 		tc->irq[2] = irq;
> > 
> > +	for (i = 0; i < 3; i++)
> > +		__raw_writel(0xff, tc->regs + ATMEL_TC_REG(i, IDR));
> > +
> > 	spin_lock(&tc_list_lock);
> > 	list_add_tail(&tc->node, &tc_list);
> > 	spin_unlock(&tc_list_lock);
> > 
> > +	platform_set_drvdata(pdev, tc);
> > +
> > 	return 0;
> > }
> > 
> > +static void tc_shutdown (struct platform_device *pdev)
> > +{
> > +	int i;
> > +	struct atmel_tc *tc = platform_get_drvdata(pdev);
> > +
> > +	for (i = 0; i < 3; i++)
> > +		__raw_writel(0xff, tc->regs + ATMEL_TC_REG(i, IDR));
> > +}
> > +
> > static struct platform_driver tc_driver = {
> > 	.driver = {
> > 		.name	= "atmel_tcb",
> > 		.of_match_table	= of_match_ptr(atmel_tcb_dt_ids),
> > 	},
> > +	.shutdown = tc_shutdown,
> > };
> > 
> > static int __init tc_init(void)
> > -- 
> > 1.9.3
> > 
>
Thierry Reding - Aug. 20, 2014, 7:31 a.m.
On Wed, Aug 20, 2014 at 01:01:30AM +0200, Boris BREZILLON wrote:
> Hi Jean-Christophe,
> 
> On Wed, 20 Aug 2014 06:11:17 +0800
> Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com> wrote:
> 
> > Hi,
> > 
> > 	This is a bit weird as the clock of the TC should be off and the irq free
> > 
> > 	so this should never happened we need to investigate more why this append
> 
> I may have found the source of this bug.
> 
> As Gael stated, when you're kexec-ing a new kernel your previous kernel
> could be using the tbc_clksrc driver (and especially the clkevent
> device). Thus the kernel might have planned a timer event and then been
> asked to shutdown the machine (requested by the kexec code).
> In this case the AIC interrupt connected to the TC Block is disabled
> but not the interrupts within the TCB IP (IDR registers), possibly
> leaving a pending interrupt before booting the new kernel.
> 
> When the tcb_clksrc driver is loaded by the new kernel it enables the
> interrupt line by calling setup_irq [1] while the clockevent device is
> not registered yet [2]. Thus the event_handler is still NULL when the
> AIC line connected to the TCB is unmasked. Remember that an interrupt
> is still pending on this HW block, which will lead to an immediate call
> to the ch2_irq handler, which tries to call the event_handler, which in
> turns is NULL because clkevent device registration has not taken place
> at this moment => Kernel panic.
> ITOH, we can't register the clkevent device before the irq handler is
> set up, because we should be ready to handle clkevent request at the
> time clockevents_config_and_register is called.
> 
> This leaves two solution:
>  1) disable the TCB irqs (using TCB IDR registers) before calling
>  setup_irq in the tcb_clksrc driver
>  2) disable the TCB irqs at the tclib level (as proposed by Gael)
> 
> I prefer solution #2 because it fixes the bug for all TCB users (not
> just the tcb_clksrc driver).

Wouldn't a more proper fix be to only enable the IRQ (setup_irq()) once
everything has properly been set up? That's certainly how all other
drivers are doing this. Generally I think it's best to assume that an
interrupt can fire at any point after it's been enabled, so everything
should be set up prior to enabling it.

Also, does anyone know why this driver uses setup_irq() rather than the
more idiomatic request_irq()?

Thierry
Boris BREZILLON - Aug. 20, 2014, 8:14 a.m.
Hi Thierry,

On Wed, 20 Aug 2014 09:31:13 +0200
Thierry Reding <thierry.reding@gmail.com> wrote:

> On Wed, Aug 20, 2014 at 01:01:30AM +0200, Boris BREZILLON wrote:
> > Hi Jean-Christophe,
> > 
> > On Wed, 20 Aug 2014 06:11:17 +0800
> > Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com> wrote:
> > 
> > > Hi,
> > > 
> > > 	This is a bit weird as the clock of the TC should be off and the irq free
> > > 
> > > 	so this should never happened we need to investigate more why this append
> > 
> > I may have found the source of this bug.
> > 
> > As Gael stated, when you're kexec-ing a new kernel your previous kernel
> > could be using the tbc_clksrc driver (and especially the clkevent
> > device). Thus the kernel might have planned a timer event and then been
> > asked to shutdown the machine (requested by the kexec code).
> > In this case the AIC interrupt connected to the TC Block is disabled
> > but not the interrupts within the TCB IP (IDR registers), possibly
> > leaving a pending interrupt before booting the new kernel.
> > 
> > When the tcb_clksrc driver is loaded by the new kernel it enables the
> > interrupt line by calling setup_irq [1] while the clockevent device is
> > not registered yet [2]. Thus the event_handler is still NULL when the
> > AIC line connected to the TCB is unmasked. Remember that an interrupt
> > is still pending on this HW block, which will lead to an immediate call
> > to the ch2_irq handler, which tries to call the event_handler, which in
> > turns is NULL because clkevent device registration has not taken place
> > at this moment => Kernel panic.
> > ITOH, we can't register the clkevent device before the irq handler is
> > set up, because we should be ready to handle clkevent request at the
> > time clockevents_config_and_register is called.
> > 
> > This leaves two solution:
> >  1) disable the TCB irqs (using TCB IDR registers) before calling
> >  setup_irq in the tcb_clksrc driver
> >  2) disable the TCB irqs at the tclib level (as proposed by Gael)
> > 
> > I prefer solution #2 because it fixes the bug for all TCB users (not
> > just the tcb_clksrc driver).
> 
> Wouldn't a more proper fix be to only enable the IRQ (setup_irq()) once
> everything has properly been set up? That's certainly how all other
> drivers are doing this. Generally I think it's best to assume that an
> interrupt can fire at any point after it's been enabled, so everything
> should be set up prior to enabling it.

Sure. And, AFAIK, another common practice is to disable all interrupts
and acknowledge all pending interrupts before registering a new irq
handler to avoid inheriting peripheral dirty state from previous usage
(either the bootloader, or the previous kernel when using kexec).

This being said, I really think we should leave the HW in a clean state
when shutdown is called. And disabling interrupts at the tclib level
(in a shutdown callback) ensure that.

> 
> Also, does anyone know why this driver uses setup_irq() rather than the
> more idiomatic request_irq()?

Because nobody has sanitized this driver yet ;-).

Best Regards,

Boris
Thierry Reding - Aug. 20, 2014, 8:28 a.m.
On Wed, Aug 20, 2014 at 10:14:22AM +0200, Boris BREZILLON wrote:
> Hi Thierry,
> 
> On Wed, 20 Aug 2014 09:31:13 +0200
> Thierry Reding <thierry.reding@gmail.com> wrote:
> 
> > On Wed, Aug 20, 2014 at 01:01:30AM +0200, Boris BREZILLON wrote:
> > > Hi Jean-Christophe,
> > > 
> > > On Wed, 20 Aug 2014 06:11:17 +0800
> > > Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com> wrote:
> > > 
> > > > Hi,
> > > > 
> > > > 	This is a bit weird as the clock of the TC should be off and the irq free
> > > > 
> > > > 	so this should never happened we need to investigate more why this append
> > > 
> > > I may have found the source of this bug.
> > > 
> > > As Gael stated, when you're kexec-ing a new kernel your previous kernel
> > > could be using the tbc_clksrc driver (and especially the clkevent
> > > device). Thus the kernel might have planned a timer event and then been
> > > asked to shutdown the machine (requested by the kexec code).
> > > In this case the AIC interrupt connected to the TC Block is disabled
> > > but not the interrupts within the TCB IP (IDR registers), possibly
> > > leaving a pending interrupt before booting the new kernel.
> > > 
> > > When the tcb_clksrc driver is loaded by the new kernel it enables the
> > > interrupt line by calling setup_irq [1] while the clockevent device is
> > > not registered yet [2]. Thus the event_handler is still NULL when the
> > > AIC line connected to the TCB is unmasked. Remember that an interrupt
> > > is still pending on this HW block, which will lead to an immediate call
> > > to the ch2_irq handler, which tries to call the event_handler, which in
> > > turns is NULL because clkevent device registration has not taken place
> > > at this moment => Kernel panic.
> > > ITOH, we can't register the clkevent device before the irq handler is
> > > set up, because we should be ready to handle clkevent request at the
> > > time clockevents_config_and_register is called.
> > > 
> > > This leaves two solution:
> > >  1) disable the TCB irqs (using TCB IDR registers) before calling
> > >  setup_irq in the tcb_clksrc driver
> > >  2) disable the TCB irqs at the tclib level (as proposed by Gael)
> > > 
> > > I prefer solution #2 because it fixes the bug for all TCB users (not
> > > just the tcb_clksrc driver).
> > 
> > Wouldn't a more proper fix be to only enable the IRQ (setup_irq()) once
> > everything has properly been set up? That's certainly how all other
> > drivers are doing this. Generally I think it's best to assume that an
> > interrupt can fire at any point after it's been enabled, so everything
> > should be set up prior to enabling it.
> 
> Sure. And, AFAIK, another common practice is to disable all interrupts
> and acknowledge all pending interrupts before registering a new irq
> handler to avoid inheriting peripheral dirty state from previous usage
> (either the bootloader, or the previous kernel when using kexec).

Discarding all pending interrupts may not always be what we want. And
masking interrupts prior to registering the handler isn't always going
to work either (shared interrupts), so device drivers should always set
things up in the correct order.

> This being said, I really think we should leave the HW in a clean state
> when shutdown is called. And disabling interrupts at the tclib level
> (in a shutdown callback) ensure that.

Yes, cleaning up the hardware and disabling interrupts upon shutdown
seems like a good idea to me in addition to the above.

Thierry
Boris BREZILLON - Aug. 20, 2014, 9:06 a.m.
On Wed, 20 Aug 2014 10:28:20 +0200
Thierry Reding <thierry.reding@gmail.com> wrote:

> On Wed, Aug 20, 2014 at 10:14:22AM +0200, Boris BREZILLON wrote:
> > Hi Thierry,
> > 
> > On Wed, 20 Aug 2014 09:31:13 +0200
> > Thierry Reding <thierry.reding@gmail.com> wrote:
> > 
> > > On Wed, Aug 20, 2014 at 01:01:30AM +0200, Boris BREZILLON wrote:
> > > > Hi Jean-Christophe,
> > > > 
> > > > On Wed, 20 Aug 2014 06:11:17 +0800
> > > > Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com> wrote:
> > > > 
> > > > > Hi,
> > > > > 
> > > > > 	This is a bit weird as the clock of the TC should be off and the irq free
> > > > > 
> > > > > 	so this should never happened we need to investigate more why this append
> > > > 
> > > > I may have found the source of this bug.
> > > > 
> > > > As Gael stated, when you're kexec-ing a new kernel your previous kernel
> > > > could be using the tbc_clksrc driver (and especially the clkevent
> > > > device). Thus the kernel might have planned a timer event and then been
> > > > asked to shutdown the machine (requested by the kexec code).
> > > > In this case the AIC interrupt connected to the TC Block is disabled
> > > > but not the interrupts within the TCB IP (IDR registers), possibly
> > > > leaving a pending interrupt before booting the new kernel.
> > > > 
> > > > When the tcb_clksrc driver is loaded by the new kernel it enables the
> > > > interrupt line by calling setup_irq [1] while the clockevent device is
> > > > not registered yet [2]. Thus the event_handler is still NULL when the
> > > > AIC line connected to the TCB is unmasked. Remember that an interrupt
> > > > is still pending on this HW block, which will lead to an immediate call
> > > > to the ch2_irq handler, which tries to call the event_handler, which in
> > > > turns is NULL because clkevent device registration has not taken place
> > > > at this moment => Kernel panic.
> > > > ITOH, we can't register the clkevent device before the irq handler is
> > > > set up, because we should be ready to handle clkevent request at the
> > > > time clockevents_config_and_register is called.
> > > > 
> > > > This leaves two solution:
> > > >  1) disable the TCB irqs (using TCB IDR registers) before calling
> > > >  setup_irq in the tcb_clksrc driver
> > > >  2) disable the TCB irqs at the tclib level (as proposed by Gael)
> > > > 
> > > > I prefer solution #2 because it fixes the bug for all TCB users (not
> > > > just the tcb_clksrc driver).
> > > 
> > > Wouldn't a more proper fix be to only enable the IRQ (setup_irq()) once
> > > everything has properly been set up? That's certainly how all other
> > > drivers are doing this. Generally I think it's best to assume that an
> > > interrupt can fire at any point after it's been enabled, so everything
> > > should be set up prior to enabling it.
> > 
> > Sure. And, AFAIK, another common practice is to disable all interrupts
> > and acknowledge all pending interrupts before registering a new irq
> > handler to avoid inheriting peripheral dirty state from previous usage
> > (either the bootloader, or the previous kernel when using kexec).
> 
> Discarding all pending interrupts may not always be what we want. And
> masking interrupts prior to registering the handler isn't always going
> to work either (shared interrupts), so device drivers should always set
> things up in the correct order.
> 

I meant disabling/acknowledging interrupts within the HW block not
the interrupt line connected to the interrupt controller (which indeed
can be shared among several peripherals).
The TCB IP provides SR (Status Register) to acknowledge interrupts at
the TCB level and IER/IDR/ISR (Interrupt Enable/Disable/Status
Register) to manipulate TCB interrupts.

> > This being said, I really think we should leave the HW in a clean state
> > when shutdown is called. And disabling interrupts at the tclib level
> > (in a shutdown callback) ensure that.
> 
> Yes, cleaning up the hardware and disabling interrupts upon shutdown
> seems like a good idea to me in addition to the above.
> 
> Thierry
Thierry Reding - Aug. 20, 2014, 9:48 a.m.
On Wed, Aug 20, 2014 at 11:06:25AM +0200, Boris BREZILLON wrote:
> On Wed, 20 Aug 2014 10:28:20 +0200
> Thierry Reding <thierry.reding@gmail.com> wrote:
> 
> > On Wed, Aug 20, 2014 at 10:14:22AM +0200, Boris BREZILLON wrote:
> > > Hi Thierry,
> > > 
> > > On Wed, 20 Aug 2014 09:31:13 +0200
> > > Thierry Reding <thierry.reding@gmail.com> wrote:
> > > 
> > > > On Wed, Aug 20, 2014 at 01:01:30AM +0200, Boris BREZILLON wrote:
> > > > > Hi Jean-Christophe,
> > > > > 
> > > > > On Wed, 20 Aug 2014 06:11:17 +0800
> > > > > Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com> wrote:
> > > > > 
> > > > > > Hi,
> > > > > > 
> > > > > > 	This is a bit weird as the clock of the TC should be off and the irq free
> > > > > > 
> > > > > > 	so this should never happened we need to investigate more why this append
> > > > > 
> > > > > I may have found the source of this bug.
> > > > > 
> > > > > As Gael stated, when you're kexec-ing a new kernel your previous kernel
> > > > > could be using the tbc_clksrc driver (and especially the clkevent
> > > > > device). Thus the kernel might have planned a timer event and then been
> > > > > asked to shutdown the machine (requested by the kexec code).
> > > > > In this case the AIC interrupt connected to the TC Block is disabled
> > > > > but not the interrupts within the TCB IP (IDR registers), possibly
> > > > > leaving a pending interrupt before booting the new kernel.
> > > > > 
> > > > > When the tcb_clksrc driver is loaded by the new kernel it enables the
> > > > > interrupt line by calling setup_irq [1] while the clockevent device is
> > > > > not registered yet [2]. Thus the event_handler is still NULL when the
> > > > > AIC line connected to the TCB is unmasked. Remember that an interrupt
> > > > > is still pending on this HW block, which will lead to an immediate call
> > > > > to the ch2_irq handler, which tries to call the event_handler, which in
> > > > > turns is NULL because clkevent device registration has not taken place
> > > > > at this moment => Kernel panic.
> > > > > ITOH, we can't register the clkevent device before the irq handler is
> > > > > set up, because we should be ready to handle clkevent request at the
> > > > > time clockevents_config_and_register is called.
> > > > > 
> > > > > This leaves two solution:
> > > > >  1) disable the TCB irqs (using TCB IDR registers) before calling
> > > > >  setup_irq in the tcb_clksrc driver
> > > > >  2) disable the TCB irqs at the tclib level (as proposed by Gael)
> > > > > 
> > > > > I prefer solution #2 because it fixes the bug for all TCB users (not
> > > > > just the tcb_clksrc driver).
> > > > 
> > > > Wouldn't a more proper fix be to only enable the IRQ (setup_irq()) once
> > > > everything has properly been set up? That's certainly how all other
> > > > drivers are doing this. Generally I think it's best to assume that an
> > > > interrupt can fire at any point after it's been enabled, so everything
> > > > should be set up prior to enabling it.
> > > 
> > > Sure. And, AFAIK, another common practice is to disable all interrupts
> > > and acknowledge all pending interrupts before registering a new irq
> > > handler to avoid inheriting peripheral dirty state from previous usage
> > > (either the bootloader, or the previous kernel when using kexec).
> > 
> > Discarding all pending interrupts may not always be what we want. And
> > masking interrupts prior to registering the handler isn't always going
> > to work either (shared interrupts), so device drivers should always set
> > things up in the correct order.
> > 
> 
> I meant disabling/acknowledging interrupts within the HW block not
> the interrupt line connected to the interrupt controller (which indeed
> can be shared among several peripherals).
> The TCB IP provides SR (Status Register) to acknowledge interrupts at
> the TCB level and IER/IDR/ISR (Interrupt Enable/Disable/Status
> Register) to manipulate TCB interrupts.

But when you share interrupts, then when an incoming interrupt will
cause all handlers to be called, so you still need to set it up
properly.

Thierry
Boris BREZILLON - Aug. 20, 2014, 10:07 a.m.
On Wed, 20 Aug 2014 11:48:08 +0200
Thierry Reding <thierry.reding@gmail.com> wrote:

> On Wed, Aug 20, 2014 at 11:06:25AM +0200, Boris BREZILLON wrote:
> > On Wed, 20 Aug 2014 10:28:20 +0200
> > Thierry Reding <thierry.reding@gmail.com> wrote:
> > 
> > > On Wed, Aug 20, 2014 at 10:14:22AM +0200, Boris BREZILLON wrote:
> > > > Hi Thierry,
> > > > 
> > > > On Wed, 20 Aug 2014 09:31:13 +0200
> > > > Thierry Reding <thierry.reding@gmail.com> wrote:
> > > > 
> > > > > On Wed, Aug 20, 2014 at 01:01:30AM +0200, Boris BREZILLON wrote:
> > > > > > Hi Jean-Christophe,
> > > > > > 
> > > > > > On Wed, 20 Aug 2014 06:11:17 +0800
> > > > > > Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com> wrote:
> > > > > > 
> > > > > > > Hi,
> > > > > > > 
> > > > > > > 	This is a bit weird as the clock of the TC should be off and the irq free
> > > > > > > 
> > > > > > > 	so this should never happened we need to investigate more why this append
> > > > > > 
> > > > > > I may have found the source of this bug.
> > > > > > 
> > > > > > As Gael stated, when you're kexec-ing a new kernel your previous kernel
> > > > > > could be using the tbc_clksrc driver (and especially the clkevent
> > > > > > device). Thus the kernel might have planned a timer event and then been
> > > > > > asked to shutdown the machine (requested by the kexec code).
> > > > > > In this case the AIC interrupt connected to the TC Block is disabled
> > > > > > but not the interrupts within the TCB IP (IDR registers), possibly
> > > > > > leaving a pending interrupt before booting the new kernel.
> > > > > > 
> > > > > > When the tcb_clksrc driver is loaded by the new kernel it enables the
> > > > > > interrupt line by calling setup_irq [1] while the clockevent device is
> > > > > > not registered yet [2]. Thus the event_handler is still NULL when the
> > > > > > AIC line connected to the TCB is unmasked. Remember that an interrupt
> > > > > > is still pending on this HW block, which will lead to an immediate call
> > > > > > to the ch2_irq handler, which tries to call the event_handler, which in
> > > > > > turns is NULL because clkevent device registration has not taken place
> > > > > > at this moment => Kernel panic.
> > > > > > ITOH, we can't register the clkevent device before the irq handler is
> > > > > > set up, because we should be ready to handle clkevent request at the
> > > > > > time clockevents_config_and_register is called.
> > > > > > 
> > > > > > This leaves two solution:
> > > > > >  1) disable the TCB irqs (using TCB IDR registers) before calling
> > > > > >  setup_irq in the tcb_clksrc driver
> > > > > >  2) disable the TCB irqs at the tclib level (as proposed by Gael)
> > > > > > 
> > > > > > I prefer solution #2 because it fixes the bug for all TCB users (not
> > > > > > just the tcb_clksrc driver).
> > > > > 
> > > > > Wouldn't a more proper fix be to only enable the IRQ (setup_irq()) once
> > > > > everything has properly been set up? That's certainly how all other
> > > > > drivers are doing this. Generally I think it's best to assume that an
> > > > > interrupt can fire at any point after it's been enabled, so everything
> > > > > should be set up prior to enabling it.
> > > > 
> > > > Sure. And, AFAIK, another common practice is to disable all interrupts
> > > > and acknowledge all pending interrupts before registering a new irq
> > > > handler to avoid inheriting peripheral dirty state from previous usage
> > > > (either the bootloader, or the previous kernel when using kexec).
> > > 
> > > Discarding all pending interrupts may not always be what we want. And
> > > masking interrupts prior to registering the handler isn't always going
> > > to work either (shared interrupts), so device drivers should always set
> > > things up in the correct order.
> > > 
> > 
> > I meant disabling/acknowledging interrupts within the HW block not
> > the interrupt line connected to the interrupt controller (which indeed
> > can be shared among several peripherals).
> > The TCB IP provides SR (Status Register) to acknowledge interrupts at
> > the TCB level and IER/IDR/ISR (Interrupt Enable/Disable/Status
> > Register) to manipulate TCB interrupts.
> 
> But when you share interrupts, then when an incoming interrupt will
> cause all handlers to be called, so you still need to set it up
> properly.

Right, I forgot about that one (even if we could mask the status
register with the interrupt status register to avoid calling the
event_handler when the interrupt is not enabled).

Anyway I agree with you on this point: everything should be ready when
calling request_irq.
Arnd Bergmann - Aug. 21, 2014, 3:32 a.m.
On Wednesday 20 August 2014, Gaël PORTAY wrote:
> +static void tc_shutdown (struct platform_device *pdev)
> +{
> +       int i;
> +       struct atmel_tc *tc = platform_get_drvdata(pdev);
> +
> +       for (i = 0; i < 3; i++)
> +               __raw_writel(0xff, tc->regs + ATMEL_TC_REG(i, IDR));
> +}

In general, __raw_readl/__raw_writel are not meant to be called by device drivers.

Just use readl/writel by default, or readl_relaxed/writel_relaxed if the code is
performance critical and you are sure it is safe to use them.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-pwm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/misc/atmel_tclib.c b/drivers/misc/atmel_tclib.c
index d505d1e..a680151 100644
--- a/drivers/misc/atmel_tclib.c
+++ b/drivers/misc/atmel_tclib.c
@@ -109,6 +109,7 @@  static int __init tc_probe(struct platform_device *pdev)
 	struct clk	*clk;
 	int		irq;
 	struct resource	*r;
+	unsigned int	i;
 
 	irq = platform_get_irq(pdev, 0);
 	if (irq < 0)
@@ -157,18 +158,33 @@  static int __init tc_probe(struct platform_device *pdev)
 	if (tc->irq[2] < 0)
 		tc->irq[2] = irq;
 
+	for (i = 0; i < 3; i++)
+		__raw_writel(0xff, tc->regs + ATMEL_TC_REG(i, IDR));
+
 	spin_lock(&tc_list_lock);
 	list_add_tail(&tc->node, &tc_list);
 	spin_unlock(&tc_list_lock);
 
+	platform_set_drvdata(pdev, tc);
+
 	return 0;
 }
 
+static void tc_shutdown (struct platform_device *pdev)
+{
+	int i;
+	struct atmel_tc *tc = platform_get_drvdata(pdev);
+
+	for (i = 0; i < 3; i++)
+		__raw_writel(0xff, tc->regs + ATMEL_TC_REG(i, IDR));
+}
+
 static struct platform_driver tc_driver = {
 	.driver = {
 		.name	= "atmel_tcb",
 		.of_match_table	= of_match_ptr(atmel_tcb_dt_ids),
 	},
+	.shutdown = tc_shutdown,
 };
 
 static int __init tc_init(void)