Skip to contentSkip to conversation

ptp: ocp: fix NULL deref in __handle_s for irig/dcf

PATCH In review Sagi Maimon 3 participants 4 messages Apr 15, 2025

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 ptp: ocp: fix NULL deref in __handle_s for irig/dcf

SMA store/get operations via sysfs can call __handle_signal_outputs or __handle_signal_inputs while irig and dcf pointers remain uninitialized. This leads to a NULL pointer dereference in __handle_s. Add NULL checks for irig and dcf to prevent crashes.

Fixes b286f4e87e32 ("serial: core: Move tty and serdev to be children of serial core port device")

---
Addressed comments from Paolo Abeni:
- https://www.spinics.net/lists/netdev/msg1082406.html
Changes since v1:
- Expanded commit message to clarify the NULL dereference scenario.

---
 drivers/ptp/ptp_ocp.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/ptp/ptp_ocp.c +8 -4
index 7945c6be1f7c..4e4a6f465b01 100644
@@ -2434,15 +2434,19 @@ ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable)
 static void
 __handle_signal_outputs(struct ptp_ocp *bp, u32 val)
 {
-	ptp_ocp_irig_out(bp, val & 0x00100010);
-	ptp_ocp_dcf_out(bp, val & 0x00200020);
+	if (bp->irig_out)
+		ptp_ocp_irig_out(bp, val & 0x00100010);
+	if (bp->dcf_out)
+		ptp_ocp_dcf_out(bp, val & 0x00200020);
 }
 
 static void
 __handle_signal_inputs(struct ptp_ocp *bp, u32 val)
 {
-	ptp_ocp_irig_in(bp, val & 0x00100010);
-	ptp_ocp_dcf_in(bp, val & 0x00200020);
+	if (bp->irig_out)
+		ptp_ocp_irig_in(bp, val & 0x00100010);
+	if (bp->dcf_out)
+		ptp_ocp_dcf_in(bp, val & 0x00200020);
 }
 
 static u32
Signature · 1 line
2.47.0
#1 lore raw
Quoted 29 lines · Sagi Maimon

SMA store/get operations via sysfs can call __handle_signal_outputs or __handle_signal_inputs while irig and dcf pointers remain uninitialized. This leads to a NULL pointer dereference in __handle_s. Add NULL checks for irig and dcf to prevent crashes.

Fixes b286f4e87e32 ("serial: core: Move tty and serdev to be children of serial core port device")

---
Addressed comments from Paolo Abeni:
- https://www.spinics.net/lists/netdev/msg1082406.html
Changes since v1:
- Expanded commit message to clarify the NULL dereference scenario.

---
 drivers/ptp/ptp_ocp.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/ptp/ptp_ocp.c +2 -2
index 7945c6be1f7c..4e4a6f465b01 100644
@@ -2434,15 +2434,19 @@ ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable)
 static void
 __handle_signal_outputs(struct ptp_ocp *bp, u32 val)
 {
-	ptp_ocp_irig_out(bp, val & 0x00100010);
-	ptp_ocp_dcf_out(bp, val & 0x00200020);
+	if (bp->irig_out)
+		ptp_ocp_irig_out(bp, val & 0x00100010);

drivers/ptp/ptp_ocp.c ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable) · line 2446

Quoted 10 lines
drivers/ptp/ptp_ocp.c +3 -2
+	if (bp->dcf_out)
+		ptp_ocp_dcf_out(bp, val & 0x00200020);
 }
 
 static void
 __handle_signal_inputs(struct ptp_ocp *bp, u32 val)
 {
-	ptp_ocp_irig_in(bp, val & 0x00100010);
-	ptp_ocp_dcf_in(bp, val & 0x00200020);
+	if (bp->irig_out)

Why not irig_in? Can we asssume that "in" isn't NULL if "out" isn't?

drivers/ptp/ptp_ocp.c ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable) · line 2448

Quoted 2 lines
drivers/ptp/ptp_ocp.c +2 -0
+		ptp_ocp_irig_in(bp, val & 0x00100010);
+	if (bp->dcf_out)

The same here.

drivers/ptp/ptp_ocp.c ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable) · line 2449

Quoted 1 line
drivers/ptp/ptp_ocp.c +1 -0
+		ptp_ocp_dcf_in(bp, val & 0x00200020);

Just my opinion, I will move these checks into ptp_ocp_...() functions as bp is passed to a function not bp->sth.

Quoted 5 lines
drivers/ptp/ptp_ocp.c +0 -0
 }
 
 static u32
Signature · 1 line
2.47.0
#2 lore raw
Quoted 35 lines · Michal Swiatkowski
Quoted 33 lines · Sagi Maimon

SMA store/get operations via sysfs can call __handle_signal_outputs or __handle_signal_inputs while irig and dcf pointers remain uninitialized. This leads to a NULL pointer dereference in __handle_s. Add NULL checks for irig and dcf to prevent crashes.

Fixes b286f4e87e32 ("serial: core: Move tty and serdev to be children of serial core port device")

---
Addressed comments from Paolo Abeni:
- https://www.spinics.net/lists/netdev/msg1082406.html
Changes since v1:
- Expanded commit message to clarify the NULL dereference scenario.

---
 drivers/ptp/ptp_ocp.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/ptp/ptp_ocp.c +4 -2
index 7945c6be1f7c..4e4a6f465b01 100644
@@ -2434,15 +2434,19 @@ ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable)
 static void
 __handle_signal_outputs(struct ptp_ocp *bp, u32 val)
 {
-     ptp_ocp_irig_out(bp, val & 0x00100010);
-     ptp_ocp_dcf_out(bp, val & 0x00200020);
+     if (bp->irig_out)
+             ptp_ocp_irig_out(bp, val & 0x00100010);
+     if (bp->dcf_out)
+             ptp_ocp_dcf_out(bp, val & 0x00200020);
 }
 
Quoted 8 lines

drivers/ptp/ptp_ocp.c ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable) · line 2446

Quoted 6 lines
drivers/ptp/ptp_ocp.c +1 -2
 static void
 __handle_signal_inputs(struct ptp_ocp *bp, u32 val)
 {
-     ptp_ocp_irig_in(bp, val & 0x00100010);
-     ptp_ocp_dcf_in(bp, val & 0x00200020);
+     if (bp->irig_out)

Why not irig_in? Can we asssume that "in" isn't NULL if "out" isn't?

Ss part of ocp_resource initiation, irig_in and irig_out initiated separately (one can be initiated and the other not ) so we can't assume that

Quoted 14 lines

drivers/ptp/ptp_ocp.c ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable) · line 2448

Quoted 2 lines
drivers/ptp/ptp_ocp.c +2 -0
+             ptp_ocp_irig_in(bp, val & 0x00100010);
+     if (bp->dcf_out)

The same here.

drivers/ptp/ptp_ocp.c ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable) · line 2449

Quoted 1 line
drivers/ptp/ptp_ocp.c +1 -0
+             ptp_ocp_dcf_in(bp, val & 0x00200020);

Just my opinion, I will move these checks into ptp_ocp_...() functions as bp is passed to a function not bp->sth.

Quoted 5 lines
drivers/ptp/ptp_ocp.c +0 -1
 }
 
 static u32
--

2.47.0

maintainer
#3 lore raw
In reply to Sagi Maimon
Quoted 4 lines · Sagi Maimon

SMA store/get operations via sysfs can call __handle_signal_outputs or __handle_signal_inputs while irig and dcf pointers remain uninitialized. This leads to a NULL pointer dereference in __handle_s. Add NULL checks for irig and dcf to prevent crashes.

Ok, looks like I misread the patch previously. IRIG and DCF registers are mapped unconditionally for OCP TimeCard. The functions you are trying to fix are HW-specific functions which you decided to reuse for your hardware. If your hardware doesn't support this function, you have to implement your own HW-specific callbacks used in ocp_adva_sma_op structure. I have to take back my Rb tag, and for this patch it's definitely NAck.

Quoted 41 lines
Fixes b286f4e87e32 ("serial: core: Move tty and serdev to be children of serial core port device")

---
Addressed comments from Paolo Abeni:
- https://www.spinics.net/lists/netdev/msg1082406.html
Changes since v1:
- Expanded commit message to clarify the NULL dereference scenario.

---
  drivers/ptp/ptp_ocp.c | 12 ++++++++----
  1 file changed, 8 insertions(+), 4 deletions(-)
drivers/ptp/ptp_ocp.c +8 -4
index 7945c6be1f7c..4e4a6f465b01 100644
@@ -2434,15 +2434,19 @@ ptp_ocp_dcf_in(struct ptp_ocp *bp, bool enable)
  static void
  __handle_signal_outputs(struct ptp_ocp *bp, u32 val)
  {
-	ptp_ocp_irig_out(bp, val & 0x00100010);
-	ptp_ocp_dcf_out(bp, val & 0x00200020);
+	if (bp->irig_out)
+		ptp_ocp_irig_out(bp, val & 0x00100010);
+	if (bp->dcf_out)
+		ptp_ocp_dcf_out(bp, val & 0x00200020);
  }
 
  static void
  __handle_signal_inputs(struct ptp_ocp *bp, u32 val)
  {
-	ptp_ocp_irig_in(bp, val & 0x00100010);
-	ptp_ocp_dcf_in(bp, val & 0x00200020);
+	if (bp->irig_out)
+		ptp_ocp_irig_in(bp, val & 0x00100010);
+	if (bp->dcf_out)
+		ptp_ocp_dcf_in(bp, val & 0x00200020);
  }
 
  static u32