Skip to contentSkip to conversation

serial: etraxfs-uart: Fix crash

PATCH In review Guenter Roeck 3 participants 9 messages Nov 2, 2015 – Nov 16, 2015 TTY layer and serial drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 serial: etraxfs-uart: Fix crash

Since commit 7d8c70d8048c ("serial: mctrl-gpio: rename init function"), crisv32 either do not build or crash as follows.

Unable to handle kernel NULL pointer dereference
Linux 4.3.0-rc7-next-20151101 #1 Sun Nov 1 11:41:28 PST 2015
...
Call Trace: [<c0004a0e>] show_stack+0x0/0x9e
[<c004c0c0>] printk+0x0/0x2c
[<c00059d4>] show_registers+0x14a/0x1c2
[<c004c0c0>] printk+0x0/0x2c
[<c0004b52>] die_if_kernel+0x7c/0x9e
[<c0005346>] do_page_fault+0x32e/0x3e6
[<c01dc59c>] of_get_property+0x0/0x2c
[<c01e0558>] of_irq_parse_raw+0x12a/0x376
[<c01dc59c>] of_get_property+0x0/0x2c
[<c0053aca>] get_page_from_freelist+0x73e/0x856
[<c01dc59c>] of_get_property+0x0/0x2c
[<c0008912>] d_mmu_refill+0x10a/0x112
[<c01b488c>] devm_kmalloc+0x40/0x56
[<c01b47d0>] add_dr+0xc/0x1c
[<c01b4800>] devm_add_action+0x2/0x4e
[<c01abdbc>] mctrl_gpio_init_noauto+0x1c/0x76
[<c01abf9e>] mctrl_gpio_init+0x22/0x110

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Fixes 7d8c70d8048c ("serial: mctrl-gpio: rename init function")
Cc Uwe Kleine-König <[email protected]>
Cc Niklas Cassel <[email protected]>

---
It might make sense to merge this patch into the original patch if possible.

 drivers/tty/serial/etraxfs-uart.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/tty/serial/etraxfs-uart.c +1 -1
index 6813e316e9ff..2f80bc7e44fb 100644
@@ -894,7 +894,7 @@ static int etraxfs_uart_probe(struct platform_device *pdev)
 	up->regi_ser = of_iomap(np, 0);
 	up->port.dev = &pdev->dev;
 
-	up->gpios = mctrl_gpio_init(&pdev->dev, 0);
+	up->gpios = mctrl_gpio_init_noauto(&pdev->dev, 0);
 	if (IS_ERR(up->gpios))
 		return PTR_ERR(up->gpios);
 
Signature · 1 line
2.1.4
Acked-by on v1 1/1
#1 lore raw

Hello,

Quoted 15 lines · Guenter Roeck

Since commit 7d8c70d8048c ("serial: mctrl-gpio: rename init function"), crisv32 either do not build or crash as follows.

Unable to handle kernel NULL pointer dereference
Linux 4.3.0-rc7-next-20151101 #1 Sun Nov 1 11:41:28 PST 2015
...
Call Trace: [<c0004a0e>] show_stack+0x0/0x9e
[<c004c0c0>] printk+0x0/0x2c
[<c00059d4>] show_registers+0x14a/0x1c2
[<c004c0c0>] printk+0x0/0x2c
[<c0004b52>] die_if_kernel+0x7c/0x9e
[<c0005346>] do_page_fault+0x32e/0x3e6
[<c01dc59c>] of_get_property+0x0/0x2c
[<c01e0558>] of_irq_parse_raw+0x12a/0x376
[<c01dc59c>] of_get_property+0x0/0x2c
Quoted 8 lines
[<c0053aca>] get_page_from_freelist+0x73e/0x856
[<c01dc59c>] of_get_property+0x0/0x2c
[<c0008912>] d_mmu_refill+0x10a/0x112
[<c01b488c>] devm_kmalloc+0x40/0x56
[<c01b47d0>] add_dr+0xc/0x1c
[<c01b4800>] devm_add_action+0x2/0x4e
[<c01abdbc>] mctrl_gpio_init_noauto+0x1c/0x76
[<c01abf9e>] mctrl_gpio_init+0x22/0x110

There should be a warning at compile time about mismatching parameters.

Quoted 3 lines

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Yes. BTW, 7b9c5162c182 looks broken, too, because the interrupt handling is missing.

Quoted 4 lines
Fixes 7d8c70d8048c ("serial: mctrl-gpio: rename init function")
Cc Uwe Kleine-König <[email protected]>
Cc Niklas Cassel <[email protected]>

Thanks
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#2 lore raw
Quoted 27 lines · Uwe Kleine-König

Hello,

Quoted 15 lines · Guenter Roeck

Since commit 7d8c70d8048c ("serial: mctrl-gpio: rename init function"), crisv32 either do not build or crash as follows.

Unable to handle kernel NULL pointer dereference
Linux 4.3.0-rc7-next-20151101 #1 Sun Nov 1 11:41:28 PST 2015
...
Call Trace: [<c0004a0e>] show_stack+0x0/0x9e
[<c004c0c0>] printk+0x0/0x2c
[<c00059d4>] show_registers+0x14a/0x1c2
[<c004c0c0>] printk+0x0/0x2c
[<c0004b52>] die_if_kernel+0x7c/0x9e
[<c0005346>] do_page_fault+0x32e/0x3e6
[<c01dc59c>] of_get_property+0x0/0x2c
[<c01e0558>] of_irq_parse_raw+0x12a/0x376
[<c01dc59c>] of_get_property+0x0/0x2c
Quoted 8 lines
[<c0053aca>] get_page_from_freelist+0x73e/0x856
[<c01dc59c>] of_get_property+0x0/0x2c
[<c0008912>] d_mmu_refill+0x10a/0x112
[<c01b488c>] devm_kmalloc+0x40/0x56
[<c01b47d0>] add_dr+0xc/0x1c
[<c01b4800>] devm_add_action+0x2/0x4e
[<c01abdbc>] mctrl_gpio_init_noauto+0x1c/0x76
[<c01abf9e>] mctrl_gpio_init+0x22/0x110
Quoted 9 lines

There should be a warning at compile time about mismatching parameters.

Quoted 3 lines

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Yes. BTW, 7b9c5162c182 looks broken, too, because the interrupt handling is missing.

I can not comment on that; all I can say is that it works for me with qemu.

Guenter

#3 lore raw
Quoted 34 lines · Guenter Roeck
Quoted 33 lines · Uwe Kleine-König

Hello,

Quoted 15 lines · Guenter Roeck

Since commit 7d8c70d8048c ("serial: mctrl-gpio: rename init function"), crisv32 either do not build or crash as follows.

Unable to handle kernel NULL pointer dereference
Linux 4.3.0-rc7-next-20151101 #1 Sun Nov 1 11:41:28 PST 2015
...
Call Trace: [<c0004a0e>] show_stack+0x0/0x9e
[<c004c0c0>] printk+0x0/0x2c
[<c00059d4>] show_registers+0x14a/0x1c2
[<c004c0c0>] printk+0x0/0x2c
[<c0004b52>] die_if_kernel+0x7c/0x9e
[<c0005346>] do_page_fault+0x32e/0x3e6
[<c01dc59c>] of_get_property+0x0/0x2c
[<c01e0558>] of_irq_parse_raw+0x12a/0x376
[<c01dc59c>] of_get_property+0x0/0x2c
Quoted 8 lines
[<c0053aca>] get_page_from_freelist+0x73e/0x856
[<c01dc59c>] of_get_property+0x0/0x2c
[<c0008912>] d_mmu_refill+0x10a/0x112
[<c01b488c>] devm_kmalloc+0x40/0x56
[<c01b47d0>] add_dr+0xc/0x1c
[<c01b4800>] devm_add_action+0x2/0x4e
[<c01abdbc>] mctrl_gpio_init_noauto+0x1c/0x76
[<c01abf9e>] mctrl_gpio_init+0x22/0x110

There should be a warning at compile time about mismatching parameters.

Quoted 3 lines

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Quoted 5 lines
Quoted 3 lines

Yes. BTW, 7b9c5162c182 looks broken, too, because the interrupt handling is missing.

I can not comment on that; all I can say is that it works for me with qemu.

What does "works for me" mean? You get output on the qemu console? Fine, but that also works if you don't do any handshaking at all I guess.

Best regards
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#4 lore raw
In reply to Uwe Kleine-König
Quoted 5 lines · Uwe Kleine-König
Quoted 3 lines

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Yes. BTW, 7b9c5162c182 looks broken, too, because the interrupt handling is missing.

The only drivers including serial_mctrl_gpio.h is atmel_serial.c, clps711x.c, mxs-auart.c and etraxfs-uart.c

etraxfs commit
7b9c5162c182 ("serial: etraxfs-uart: use mctrl_gpio helpers for handling modem signals") is heavily influenced by
62b0a1b3e759 ("serial: clps711x: Use mctrl_gpio helpers for handling modem signals")

a driver which also appears to lack mctrl irq setup/handling.

Uwe, are you going to convert all serial drivers that includes serial_mctrl_gpio.h to use the new mctrl_gpio_init? (which also sets up irqs)

Since mctrl gpios are optional in the device tree, I guess mctrl_gpio_init handles if they are missing. So even if etraxfs-uart.c currently lacks mctrl irq setup/handling, if we migrate it to use the new mctrl_gpio_init, we will get this automatically?

#5 lore raw

Helo Niklas,

Quoted 15 lines · Niklas Cassel
Quoted 5 lines · Uwe Kleine-König
Quoted 3 lines

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Yes. BTW, 7b9c5162c182 looks broken, too, because the interrupt handling is missing.

The only drivers including serial_mctrl_gpio.h is atmel_serial.c, clps711x.c, mxs-auart.c and etraxfs-uart.c

etraxfs commit
7b9c5162c182 ("serial: etraxfs-uart: use mctrl_gpio helpers for handling modem signals") is heavily influenced by
62b0a1b3e759 ("serial: clps711x: Use mctrl_gpio helpers for handling modem signals")

Quoted 5 lines

a driver which also appears to lack mctrl irq setup/handling.

Uwe, are you going to convert all serial drivers that includes serial_mctrl_gpio.h to use the new mctrl_gpio_init? (which also sets up irqs)

That is the plan. For drivers like mxs-auart (that do irq handling) this is just simplification. When the irq handling is missing converting to the new mctrl_gpio_init might introduce regressions because it does request_irq which might fail.

Quoted 3 lines

Since mctrl gpios are optional in the device tree, I guess mctrl_gpio_init handles if they are missing. So even if etraxfs-uart.c currently lacks mctrl irq setup/handling, if we migrate it to use the new mctrl_gpio_init, we will get this automatically?

Right, if there are no gpios specified (new and old) mctrl_gpio is a nop.

Best regards
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#6 lore raw

I think that Guenter's patch should be merged asap. Right now the bug it fixes stops us from boot testing in qemu.

For the broken mctrl irq handling part, when Uwe has converted mxs-auart to use the generic irq init, I can do the same for etraxfs + make sure that it gets tested properly on real hardware.

Quoted 36 lines · Uwe Kleine-König

Helo Niklas,

Quoted 14 lines · Niklas Cassel
Quoted 5 lines · Uwe Kleine-König
Quoted 3 lines

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Yes. BTW, 7b9c5162c182 looks broken, too, because the interrupt handling is missing.

The only drivers including serial_mctrl_gpio.h is atmel_serial.c, clps711x.c, mxs-auart.c and etraxfs-uart.c

etraxfs commit
7b9c5162c182 ("serial: etraxfs-uart: use mctrl_gpio helpers for handling modem signals") is heavily influenced by
62b0a1b3e759 ("serial: clps711x: Use mctrl_gpio helpers for handling modem signals")

Quoted 5 lines

a driver which also appears to lack mctrl irq setup/handling.

Uwe, are you going to convert all serial drivers that includes serial_mctrl_gpio.h to use the new mctrl_gpio_init? (which also sets up irqs)

That is the plan. For drivers like mxs-auart (that do irq handling) this is just simplification. When the irq handling is missing converting to the new mctrl_gpio_init might introduce regressions because it does request_irq which might fail.

Quoted 3 lines

Since mctrl gpios are optional in the device tree, I guess mctrl_gpio_init handles if they are missing. So even if etraxfs-uart.c currently lacks mctrl irq setup/handling, if we migrate it to use the new mctrl_gpio_init, we will get this automatically?

Right, if there are no gpios specified (new and old) mctrl_gpio is a nop.

Best regards
Uwe

#7 lore raw
Quoted 3 lines · Niklas Cassel

I think that Guenter's patch should be merged asap. Right now the bug it fixes stops us from boot testing in qemu.

I'll send out a build (failure) report after -rc1 is out. Maybe that will create enough visibility to get the patch accepted (or officially Nacked).

Guenter

Quoted 44 lines

For the broken mctrl irq handling part, when Uwe has converted mxs-auart to use the generic irq init, I can do the same for etraxfs + make sure that it gets tested properly on real hardware.

Quoted 36 lines · Uwe Kleine-König

Helo Niklas,

Quoted 14 lines · Niklas Cassel
Quoted 5 lines · Uwe Kleine-König
Quoted 3 lines

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Yes. BTW, 7b9c5162c182 looks broken, too, because the interrupt handling is missing.

The only drivers including serial_mctrl_gpio.h is atmel_serial.c, clps711x.c, mxs-auart.c and etraxfs-uart.c

etraxfs commit
7b9c5162c182 ("serial: etraxfs-uart: use mctrl_gpio helpers for handling modem signals") is heavily influenced by
62b0a1b3e759 ("serial: clps711x: Use mctrl_gpio helpers for handling modem signals")

Quoted 5 lines

a driver which also appears to lack mctrl irq setup/handling.

Uwe, are you going to convert all serial drivers that includes serial_mctrl_gpio.h to use the new mctrl_gpio_init? (which also sets up irqs)

That is the plan. For drivers like mxs-auart (that do irq handling) this is just simplification. When the irq handling is missing converting to the new mctrl_gpio_init might introduce regressions because it does request_irq which might fail.

Quoted 3 lines

Since mctrl gpios are optional in the device tree, I guess mctrl_gpio_init handles if they are missing. So even if etraxfs-uart.c currently lacks mctrl irq setup/handling, if we migrate it to use the new mctrl_gpio_init, we will get this automatically?

Right, if there are no gpios specified (new and old) mctrl_gpio is a nop.

Best regards
Uwe

Acked-by on v1 1/1
#8 lore raw
In reply to Guenter Roeck
Quoted 51 lines · Guenter Roeck

Since commit 7d8c70d8048c ("serial: mctrl-gpio: rename init function"), crisv32 either do not build or crash as follows.

Unable to handle kernel NULL pointer dereference
Linux 4.3.0-rc7-next-20151101 #1 Sun Nov 1 11:41:28 PST 2015
...
Call Trace: [<c0004a0e>] show_stack+0x0/0x9e
[<c004c0c0>] printk+0x0/0x2c
[<c00059d4>] show_registers+0x14a/0x1c2
[<c004c0c0>] printk+0x0/0x2c
[<c0004b52>] die_if_kernel+0x7c/0x9e
[<c0005346>] do_page_fault+0x32e/0x3e6
[<c01dc59c>] of_get_property+0x0/0x2c
[<c01e0558>] of_irq_parse_raw+0x12a/0x376
[<c01dc59c>] of_get_property+0x0/0x2c
[<c0053aca>] get_page_from_freelist+0x73e/0x856
[<c01dc59c>] of_get_property+0x0/0x2c
[<c0008912>] d_mmu_refill+0x10a/0x112
[<c01b488c>] devm_kmalloc+0x40/0x56
[<c01b47d0>] add_dr+0xc/0x1c
[<c01b4800>] devm_add_action+0x2/0x4e
[<c01abdbc>] mctrl_gpio_init_noauto+0x1c/0x76
[<c01abf9e>] mctrl_gpio_init+0x22/0x110

The function call in the etraxfs-uart driver was not renamed, possibly due to interference with commit 7b9c5162c182 ("serial:
etraxfs-uart: use mctrl_gpio helpers for handling modem signals").

Fixes 7d8c70d8048c ("serial: mctrl-gpio: rename init function")
Cc Uwe Kleine-König <[email protected]>
Cc Niklas Cassel <[email protected]>

---
It might make sense to merge this patch into the original patch if possible.

 drivers/tty/serial/etraxfs-uart.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/tty/serial/etraxfs-uart.c +1 -1
index 6813e316e9ff..2f80bc7e44fb 100644
@@ -894,7 +894,7 @@ static int etraxfs_uart_probe(struct platform_device *pdev)
 	up->regi_ser = of_iomap(np, 0);
 	up->port.dev = &pdev->dev;
 
-	up->gpios = mctrl_gpio_init(&pdev->dev, 0);
+	up->gpios = mctrl_gpio_init_noauto(&pdev->dev, 0);
 	if (IS_ERR(up->gpios))
 		return PTR_ERR(up->gpios);