Skip to contentSkip to conversation

dmaengine: fsl-edma: Add lockdep assert for exported function

PATCH Applied Krzysztof Kozlowski 3 participants 9 messages Jun 11, 2020 – Jun 24, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 dmaengine: fsl-edma: Add lockdep assert for exported function

Add lockdep assert for an exported function expected to be called under spin lock. Since this function is called in different modules, the lockdep assert will be self-documenting note about need for locking.

---
 drivers/dma/fsl-edma-common.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/dma/fsl-edma-common.c +2 -0
index 5697c3622699..4550818cca4a 100644
@@ -589,6 +589,8 @@ void fsl_edma_xfer_desc(struct fsl_edma_chan *fsl_chan)
 {
 	struct virt_dma_desc *vdesc;
 
+	lockdep_assert_held(&fsl_chan->vchan.lock);
+
 	vdesc = vchan_next_desc(&fsl_chan->vchan);
 	if (!vdesc)
 		return;
Signature · 1 line
2.7.4
#1 lore raw

PATCH 2/2 dmaengine: fsl-edma: Fix NULL pointer exception in fsl_edma_tx_handler

NULL pointer exception happens occasionally on serial output initiated by login timeout. This was reproduced only if kernel was built with significant debugging options and EDMA driver is used with serial console.

    col-vf50 login: root
    Password:
    Login timed out after 60 seconds.
    Unable to handle kernel NULL pointer dereference at virtual address 00000044
    Internal error: Oops: 5 [#1] ARM
    CPU: 0 PID: 157 Comm: login Not tainted 5.7.0-next-20200610-dirty #4
    Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
      (fsl_edma_tx_handler) from [<8016eb10>] (__handle_irq_event_percpu+0x64/0x304)
      (__handle_irq_event_percpu) from [<8016eddc>] (handle_irq_event_percpu+0x2c/0x7c)
      (handle_irq_event_percpu) from [<8016ee64>] (handle_irq_event+0x38/0x5c)
      (handle_irq_event) from [<801729e4>] (handle_fasteoi_irq+0xa4/0x160)
      (handle_fasteoi_irq) from [<8016ddcc>] (generic_handle_irq+0x34/0x44)
      (generic_handle_irq) from [<8016e40c>] (__handle_domain_irq+0x54/0xa8)
      (__handle_domain_irq) from [<80508bc8>] (gic_handle_irq+0x4c/0x80)
      (gic_handle_irq) from [<80100af0>] (__irq_svc+0x70/0x98)
    Exception stack(0x8459fe80 to 0x8459fec8)
    fe80: 72286b00 e3359f64 00000001 0000412d a0070013 85c98840 85c98840 a0070013
    fea0: 8054e0d4 00000000 00000002 00000000 00000002 8459fed0 8081fbe8 8081fbec
    fec0: 60070013 ffffffff
      (__irq_svc) from [<8081fbec>] (_raw_spin_unlock_irqrestore+0x30/0x58)
      (_raw_spin_unlock_irqrestore) from [<8056cb48>] (uart_flush_buffer+0x88/0xf8)
      (uart_flush_buffer) from [<80554e60>] (tty_ldisc_hangup+0x38/0x1ac)
      (tty_ldisc_hangup) from [<8054c7f4>] (__tty_hangup+0x158/0x2bc)
      (__tty_hangup) from [<80557b90>] (disassociate_ctty.part.1+0x30/0x23c)
      (disassociate_ctty.part.1) from [<8011fc18>] (do_exit+0x580/0xba0)
      (do_exit) from [<801214f8>] (do_group_exit+0x3c/0xb4)
      (do_group_exit) from [<80121580>] (__wake_up_parent+0x0/0x14)

Issue looks like race condition between interrupt handler fsl_edma_tx_handler() (called as result of fsl_edma_xfer_desc()) and terminating the transfer with fsl_edma_terminate_all().

The fsl_edma_tx_handler() handles interrupt for a transfer with already freed edesc and idle==true.

Fixes d6be34fbd39b ("dma: Add Freescale eDMA engine driver support")
---
 drivers/dma/fsl-edma.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/dma/fsl-edma.c +7 -0
index eff7ebd8cf35..90bb72af306c 100644
@@ -45,6 +45,13 @@ static irqreturn_t fsl_edma_tx_handler(int irq, void *dev_id)
 			fsl_chan = &fsl_edma->chans[ch];
 
 			spin_lock(&fsl_chan->vchan.lock);
+
+			if (!fsl_chan->edesc) {
+				/* terminate_all called before */
+				spin_unlock(&fsl_chan->vchan.lock);
+				continue;
+			}
+
 			if (!fsl_chan->edesc->iscyclic) {
 				list_del(&fsl_chan->edesc->vdesc.node);
 				vchan_cookie_complete(&fsl_chan->edesc->vdesc);
Signature · 1 line
2.7.4
#2 lore raw

RE: [PATCH 2/2] dmaengine: fsl-edma: Fix NULL pointer exception in fsl_edma_tx_handler

Quoted 64 lines · On 2020/06/11 20:18 Krzysztof Kozlowski <[email protected]> wrote:

NULL pointer exception happens occasionally on serial output initiated by login timeout. This was reproduced only if kernel was built with significant debugging options and EDMA driver is used with serial console.

col-vf50 login: root
Password:
Login timed out after 60 seconds.
Unable to handle kernel NULL pointer dereference at virtual address
00000044

    Internal error: Oops: 5 [#1] ARM
    CPU: 0 PID: 157 Comm: login Not tainted 5.7.0-next-20200610-dirty #4
    Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
      (fsl_edma_tx_handler) from [<8016eb10>]
(__handle_irq_event_percpu+0x64/0x304)
      (__handle_irq_event_percpu) from [<8016eddc>]
(handle_irq_event_percpu+0x2c/0x7c)
      (handle_irq_event_percpu) from [<8016ee64>]
(handle_irq_event+0x38/0x5c)
      (handle_irq_event) from [<801729e4>]
(handle_fasteoi_irq+0xa4/0x160)
      (handle_fasteoi_irq) from [<8016ddcc>]
(generic_handle_irq+0x34/0x44)
      (generic_handle_irq) from [<8016e40c>]
(__handle_domain_irq+0x54/0xa8)
      (__handle_domain_irq) from [<80508bc8>] (gic_handle_irq+0x4c/0x80)
      (gic_handle_irq) from [<80100af0>] (__irq_svc+0x70/0x98)
    Exception stack(0x8459fe80 to 0x8459fec8)
    fe80: 72286b00 e3359f64 00000001 0000412d a0070013 85c98840
85c98840 a0070013
    fea0: 8054e0d4 00000000 00000002 00000000 00000002 8459fed0
8081fbe8 8081fbec
    fec0: 60070013 ffffffff
      (__irq_svc) from [<8081fbec>]
(_raw_spin_unlock_irqrestore+0x30/0x58)
      (_raw_spin_unlock_irqrestore) from [<8056cb48>]
(uart_flush_buffer+0x88/0xf8)
      (uart_flush_buffer) from [<80554e60>] (tty_ldisc_hangup+0x38/0x1ac)
      (tty_ldisc_hangup) from [<8054c7f4>] (__tty_hangup+0x158/0x2bc)
      (__tty_hangup) from [<80557b90>]
(disassociate_ctty.part.1+0x30/0x23c)
      (disassociate_ctty.part.1) from [<8011fc18>] (do_exit+0x580/0xba0)
      (do_exit) from [<801214f8>] (do_group_exit+0x3c/0xb4)
      (do_group_exit) from [<80121580>] (__wake_up_parent+0x0/0x14)

Issue looks like race condition between interrupt handler fsl_edma_tx_handler() (called as result of fsl_edma_xfer_desc()) and terminating the transfer with fsl_edma_terminate_all().

The fsl_edma_tx_handler() handles interrupt for a transfer with already freed edesc and idle==true.

Fixes d6be34fbd39b ("dma: Add Freescale eDMA engine driver support")
---
 drivers/dma/fsl-edma.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/dma/fsl-edma.c index renamed from drivers/dma/fsl-edma.c +0 -0

eff7ebd8cf35..90bb72af306c 100644

drivers/dma/fsl-edma.c static irqreturn_t fsl_edma_tx_handler(int irq, void · line 50

Quoted 6 lines
drivers/dma/fsl-edma.c +2 -0
@@ -45,6 +45,13 @@ static irqreturn_t fsl_edma_tx_handler(int irq, void
 *dev_id)
 			fsl_chan = &fsl_edma->chans[ch];
 
 			spin_lock(&fsl_chan->vchan.lock);
+
+			if (!fsl_chan->edesc) {

Would you like fix the same potential issue in mcf_edma_tx_handler() of mcf-edma.c?

Quoted 10 lines
drivers/dma/fsl-edma.c +5 -0
+				/* terminate_all called before */
+				spin_unlock(&fsl_chan->vchan.lock);
+				continue;
+			}
+
 			if (!fsl_chan->edesc->iscyclic) {
 				list_del(&fsl_chan->edesc->vdesc.node);

vchan_cookie_complete(&fsl_chan->edesc->vdesc);
--
2.7.4

#3 lore raw

Re: [PATCH 2/2] dmaengine: fsl-edma: Fix NULL pointer exception in fsl_edma_tx_handler

Quoted 68 lines · Robin Gong
Quoted 67 lines · On 2020/06/11 20:18 Krzysztof Kozlowski <[email protected]> wrote:

NULL pointer exception happens occasionally on serial output initiated by login timeout. This was reproduced only if kernel was built with significant debugging options and EDMA driver is used with serial console.

col-vf50 login: root
Password:
Login timed out after 60 seconds.
Unable to handle kernel NULL pointer dereference at virtual address
00000044

    Internal error: Oops: 5 [#1] ARM
    CPU: 0 PID: 157 Comm: login Not tainted 5.7.0-next-20200610-dirty #4
    Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
      (fsl_edma_tx_handler) from [<8016eb10>]
(__handle_irq_event_percpu+0x64/0x304)
      (__handle_irq_event_percpu) from [<8016eddc>]
(handle_irq_event_percpu+0x2c/0x7c)
      (handle_irq_event_percpu) from [<8016ee64>]
(handle_irq_event+0x38/0x5c)
      (handle_irq_event) from [<801729e4>]
(handle_fasteoi_irq+0xa4/0x160)
      (handle_fasteoi_irq) from [<8016ddcc>]
(generic_handle_irq+0x34/0x44)
      (generic_handle_irq) from [<8016e40c>]
(__handle_domain_irq+0x54/0xa8)
      (__handle_domain_irq) from [<80508bc8>] (gic_handle_irq+0x4c/0x80)
      (gic_handle_irq) from [<80100af0>] (__irq_svc+0x70/0x98)
    Exception stack(0x8459fe80 to 0x8459fec8)
    fe80: 72286b00 e3359f64 00000001 0000412d a0070013 85c98840
85c98840 a0070013
    fea0: 8054e0d4 00000000 00000002 00000000 00000002 8459fed0
8081fbe8 8081fbec
    fec0: 60070013 ffffffff
      (__irq_svc) from [<8081fbec>]
(_raw_spin_unlock_irqrestore+0x30/0x58)
      (_raw_spin_unlock_irqrestore) from [<8056cb48>]
(uart_flush_buffer+0x88/0xf8)
      (uart_flush_buffer) from [<80554e60>] (tty_ldisc_hangup+0x38/0x1ac)
      (tty_ldisc_hangup) from [<8054c7f4>] (__tty_hangup+0x158/0x2bc)
      (__tty_hangup) from [<80557b90>]
(disassociate_ctty.part.1+0x30/0x23c)
      (disassociate_ctty.part.1) from [<8011fc18>] (do_exit+0x580/0xba0)
      (do_exit) from [<801214f8>] (do_group_exit+0x3c/0xb4)
      (do_group_exit) from [<80121580>] (__wake_up_parent+0x0/0x14)

Issue looks like race condition between interrupt handler fsl_edma_tx_handler() (called as result of fsl_edma_xfer_desc()) and terminating the transfer with fsl_edma_terminate_all().

The fsl_edma_tx_handler() handles interrupt for a transfer with already freed edesc and idle==true.

Fixes d6be34fbd39b ("dma: Add Freescale eDMA engine driver support")
---
 drivers/dma/fsl-edma.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/dma/fsl-edma.c index renamed from drivers/dma/fsl-edma.c +0 -0

eff7ebd8cf35..90bb72af306c 100644

drivers/dma/fsl-edma.c +0 -0
@@ -45,6 +45,13 @@ static irqreturn_t fsl_edma_tx_handler(int irq, void
 *dev_id)
 			fsl_chan = &fsl_edma->chans[ch];
 
Quoted 5 lines

drivers/dma/fsl-edma.c static irqreturn_t fsl_edma_tx_handler(int irq, void *dev_id) · line 49

Quoted 3 lines
drivers/dma/fsl-edma.c +2 -0
 			spin_lock(&fsl_chan->vchan.lock);
+
+			if (!fsl_chan->edesc) {

Would you like fix the same potential issue in mcf_edma_tx_handler() of mcf-edma.c?

Sure. I'll make another commit as it should be backported to different kernel.

Best regards,
Krzysztof

#4 lore raw

PATCH 1/1 dmaengine: mcf-edma: Fix NULL pointer exception in mcf_edma_tx_handler

In reply to Krzysztof Kozlowski

On Toradex Colibri VF50 (Vybrid VF5xx) with fsl-edma driver NULL pointer exception happens occasionally on serial output initiated by login timeout.

This was reproduced only if kernel was built with significant debugging options and EDMA driver is used with serial console.

Issue looks like a race condition between interrupt handler fsl_edma_tx_handler() (called as a result of fsl_edma_xfer_desc()) and terminating the transfer with fsl_edma_terminate_all().

The fsl_edma_tx_handler() handles interrupt for a transfer with already freed edesc and idle==true.

The mcf-edma driver shares design and lot of code with fsl-edma. It looks like being affected by same problem. Fix this pattern the same way as fix for fsl-edma driver.

Fixes e7a3ff92eaf1 ("dmaengine: fsl-edma: add ColdFire mcf5441x edma support")

---

Not tested on HW.

---
 drivers/dma/mcf-edma.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/dma/mcf-edma.c +7 -0
index e15bd15a9ef6..e12b754e6398 100644
@@ -35,6 +35,13 @@ static irqreturn_t mcf_edma_tx_handler(int irq, void *dev_id)
 			mcf_chan = &mcf_edma->chans[ch];
 
 			spin_lock(&mcf_chan->vchan.lock);
+
+			if (!mcf_chan->edesc) {
+				/* terminate_all called before */
+				spin_unlock(&mcf_chan->vchan.lock);
+				continue;
+			}
+
 			if (!mcf_chan->edesc->iscyclic) {
 				list_del(&mcf_chan->edesc->vdesc.node);
 				vchan_cookie_complete(&mcf_chan->edesc->vdesc);
Signature · 1 line
2.7.4
Reviewed-by on v1 1/1
#5 lore raw

RE: [PATCH] dmaengine: mcf-edma: Fix NULL pointer exception in mcf_edma_tx_handler

Quoted 13 lines · On 2020/06/11 Krzysztof Kozlowski <[email protected]> wrote:

On Toradex Colibri VF50 (Vybrid VF5xx) with fsl-edma driver NULL pointer exception happens occasionally on serial output initiated by login timeout.

This was reproduced only if kernel was built with significant debugging options and EDMA driver is used with serial console.

Issue looks like a race condition between interrupt handler fsl_edma_tx_handler() (called as a result of fsl_edma_xfer_desc()) and terminating the transfer with fsl_edma_terminate_all().

The fsl_edma_tx_handler() handles interrupt for a transfer with already freed edesc and idle==true.

Quoted 8 lines

The mcf-edma driver shares design and lot of code with fsl-edma. It looks like being affected by same problem. Fix this pattern the same way as fix for fsl-edma driver.

Fixes e7a3ff92eaf1 ("dmaengine: fsl-edma: add ColdFire mcf5441x edma

support")

Quoted 29 lines

---

Not tested on HW.

---
 drivers/dma/mcf-edma.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/dma/mcf-edma.c index renamed from drivers/dma/mcf-edma.c +0 -0

e15bd15a9ef6..e12b754e6398 100644

drivers/dma/mcf-edma.c +7 -0
@@ -35,6 +35,13 @@ static irqreturn_t mcf_edma_tx_handler(int irq, void
 *dev_id)
 			mcf_chan = &mcf_edma->chans[ch];
 
 			spin_lock(&mcf_chan->vchan.lock);
+
+			if (!mcf_chan->edesc) {
+				/* terminate_all called before */
+				spin_unlock(&mcf_chan->vchan.lock);
+				continue;
+			}
+
 			if (!mcf_chan->edesc->iscyclic) {
 				list_del(&mcf_chan->edesc->vdesc.node);

vchan_cookie_complete(&mcf_chan->edesc->vdesc);
--
2.7.4

Reviewed-by on v1 2/2
#6 lore raw

RE: [PATCH 2/2] dmaengine: fsl-edma: Fix NULL pointer exception in fsl_edma_tx_handler

In reply to Krzysztof Kozlowski
Quoted 63 lines · On 2020/06/11 20:18 Krzysztof Kozlowski <[email protected]> wrote:

NULL pointer exception happens occasionally on serial output initiated by login timeout. This was reproduced only if kernel was built with significant debugging options and EDMA driver is used with serial console.

col-vf50 login: root
Password:
Login timed out after 60 seconds.
Unable to handle kernel NULL pointer dereference at virtual address
00000044

    Internal error: Oops: 5 [#1] ARM
    CPU: 0 PID: 157 Comm: login Not tainted 5.7.0-next-20200610-dirty #4
    Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
      (fsl_edma_tx_handler) from [<8016eb10>]
(__handle_irq_event_percpu+0x64/0x304)
      (__handle_irq_event_percpu) from [<8016eddc>]
(handle_irq_event_percpu+0x2c/0x7c)
      (handle_irq_event_percpu) from [<8016ee64>]
(handle_irq_event+0x38/0x5c)
      (handle_irq_event) from [<801729e4>]
(handle_fasteoi_irq+0xa4/0x160)
      (handle_fasteoi_irq) from [<8016ddcc>]
(generic_handle_irq+0x34/0x44)
      (generic_handle_irq) from [<8016e40c>]
(__handle_domain_irq+0x54/0xa8)
      (__handle_domain_irq) from [<80508bc8>] (gic_handle_irq+0x4c/0x80)
      (gic_handle_irq) from [<80100af0>] (__irq_svc+0x70/0x98)
    Exception stack(0x8459fe80 to 0x8459fec8)
    fe80: 72286b00 e3359f64 00000001 0000412d a0070013 85c98840
85c98840 a0070013
    fea0: 8054e0d4 00000000 00000002 00000000 00000002 8459fed0
8081fbe8 8081fbec
    fec0: 60070013 ffffffff
      (__irq_svc) from [<8081fbec>]
(_raw_spin_unlock_irqrestore+0x30/0x58)
      (_raw_spin_unlock_irqrestore) from [<8056cb48>]
(uart_flush_buffer+0x88/0xf8)
      (uart_flush_buffer) from [<80554e60>] (tty_ldisc_hangup+0x38/0x1ac)
      (tty_ldisc_hangup) from [<8054c7f4>] (__tty_hangup+0x158/0x2bc)
      (__tty_hangup) from [<80557b90>]
(disassociate_ctty.part.1+0x30/0x23c)
      (disassociate_ctty.part.1) from [<8011fc18>] (do_exit+0x580/0xba0)
      (do_exit) from [<801214f8>] (do_group_exit+0x3c/0xb4)
      (do_group_exit) from [<80121580>] (__wake_up_parent+0x0/0x14)

Issue looks like race condition between interrupt handler fsl_edma_tx_handler() (called as result of fsl_edma_xfer_desc()) and terminating the transfer with fsl_edma_terminate_all().

The fsl_edma_tx_handler() handles interrupt for a transfer with already freed edesc and idle==true.

Fixes d6be34fbd39b ("dma: Add Freescale eDMA engine driver support")
---
 drivers/dma/fsl-edma.c | 7 +++++++
 1 file changed, 7 insertions(+)
drivers/dma/fsl-edma.c index renamed from drivers/dma/fsl-edma.c +0 -0

eff7ebd8cf35..90bb72af306c 100644

drivers/dma/fsl-edma.c static irqreturn_t fsl_edma_tx_handler(int irq, void · line 54

Quoted 10 lines
drivers/dma/fsl-edma.c +6 -0
@@ -45,6 +45,13 @@ static irqreturn_t fsl_edma_tx_handler(int irq, void
 *dev_id)
 			fsl_chan = &fsl_edma->chans[ch];
 
 			spin_lock(&fsl_chan->vchan.lock);
+
+			if (!fsl_chan->edesc) {
+				/* terminate_all called before */
+				spin_unlock(&fsl_chan->vchan.lock);
+				continue;
+			}
Quoted 6 lines
drivers/dma/fsl-edma.c +1 -0
+
 			if (!fsl_chan->edesc->iscyclic) {
 				list_del(&fsl_chan->edesc->vdesc.node);

vchan_cookie_complete(&fsl_chan->edesc->vdesc);
--
2.7.4

Reviewed-by on v1 1/2
#7 lore raw
In reply to Krzysztof Kozlowski
Quoted 5 lines · On 2020/06/11 20:18 Krzysztof Kozlowski <[email protected]> wrote:

Add lockdep assert for an exported function expected to be called under spin lock. Since this function is called in different modules, the lockdep assert will be self-documenting note about need for locking.

Quoted 20 lines
---
 drivers/dma/fsl-edma-common.c | 2 ++
 1 file changed, 2 insertions(+)
(quoted diff) +0 -0

b/drivers/dma/fsl-edma-common.c index 5697c3622699..4550818cca4a 100644

drivers/dma/fsl-edma-common.c +2 -1
@@ -589,6 +589,8 @@ void fsl_edma_xfer_desc(struct fsl_edma_chan
 *fsl_chan)  {
 	struct virt_dma_desc *vdesc;
 
+	lockdep_assert_held(&fsl_chan->vchan.lock);
+
 	vdesc = vchan_next_desc(&fsl_chan->vchan);
 	if (!vdesc)
 		return;
--

2.7.4

maintainer Applied
#8 lore raw
In reply to Krzysztof Kozlowski
Quoted 3 lines · Krzysztof Kozlowski

Add lockdep assert for an exported function expected to be called under spin lock. Since this function is called in different modules, the lockdep assert will be self-documenting note about need for locking.

Applied all to fixes, thanks

Signature · 1 line
~Vinod