Skip to contentSkip to conversation

watchdog: dw_wdt: Fix buffer overflow when get timeout

PATCH In review Schspa Shi 2 participants 3 messages Jun 28, 2022 Watchdog device drivers

lore
Share & tools
lore

Schspa Shi

author
#0 lore raw

PATCH 1/1 watchdog: dw_wdt: Fix buffer overflow when get timeout

The top_val can be obtained from device-tree, if it is not configured correctly, there will be buffer overflow.

---
 drivers/watchdog/dw_wdt.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/watchdog/dw_wdt.c +6 -0
index cd578843277e..1f8605c0d712 100644
@@ -155,6 +155,9 @@ static unsigned int dw_wdt_get_min_timeout(struct dw_wdt *dw_wdt)
 			break;
 	}
 
+	if (WARN_ON_ONCE(idx == DW_WDT_NUM_TOPS))
+		idx = DW_WDT_NUM_TOPS - 1;
+
 	return dw_wdt->timeouts[idx].sec;
 }
 
@@ -178,6 +181,9 @@ static unsigned int dw_wdt_get_timeout(struct dw_wdt *dw_wdt)
 			break;
 	}
 
+	if (WARN_ON_ONCE(idx == DW_WDT_NUM_TOPS))
+		idx = DW_WDT_NUM_TOPS - 1;
+
 	/*
 	 * In IRQ mode due to the two stages counter, the actual timeout is
 	 * twice greater than the TOP setting.
Signature · 1 line
2.29.0
maintainer
#1 lore raw
Quoted 13 lines · Schspa Shi

The top_val can be obtained from device-tree, if it is not configured correctly, there will be buffer overflow.

---
  drivers/watchdog/dw_wdt.c | 6 ++++++
  1 file changed, 6 insertions(+)
drivers/watchdog/dw_wdt.c +0 -0
index cd578843277e..1f8605c0d712 100644

drivers/watchdog/dw_wdt.c static unsigned int dw_wdt_get_min_timeout(struct dw_wdt *dw_wdt) · line 160

Quoted 6 lines
drivers/watchdog/dw_wdt.c +3 -0
@@ -155,6 +155,9 @@ static unsigned int dw_wdt_get_min_timeout(struct dw_wdt *dw_wdt)
  			break;
  	}
 
+	if (WARN_ON_ONCE(idx == DW_WDT_NUM_TOPS))
+		idx = DW_WDT_NUM_TOPS - 1;
+

dw_wdt_get_min_timeout() returns the lowest non-0 configurable timeout. The last entry in the timeout array must not be 0, meaning there must be at least one entry in the array where the timeout is not 0. Therefore this situation can not happen.

drivers/watchdog/dw_wdt.c static unsigned int dw_wdt_get_timeout(struct dw_wdt *dw_wdt) · line 186

Quoted 10 lines
drivers/watchdog/dw_wdt.c +3 -0
  	return dw_wdt->timeouts[idx].sec;
  }
 
@@ -178,6 +181,9 @@ static unsigned int dw_wdt_get_timeout(struct dw_wdt *dw_wdt)
  			break;
  	}
 
+	if (WARN_ON_ONCE(idx == DW_WDT_NUM_TOPS))
+		idx = DW_WDT_NUM_TOPS - 1;
+

idx is derived from a top_val value written into WDOG_TIMEOUT_RANGE_REG_OFFSET, and the value written is derived from an entry in the timeouts array. This array contains an entry for each possible top_val. While the array is not sorted by top_val, dw_wdt_handle_tops() still guarantees that an entry exists.

I do not see how bad devicetree data can circumvent that. If it does, please provide an example and explain.

Thanks,
Guenter

author
#2 lore raw
Quoted 13 lines · Guenter Roeck
Quoted 12 lines · Schspa Shi

The top_val can be obtained from device-tree, if it is not configured correctly, there will be buffer overflow.

---
  drivers/watchdog/dw_wdt.c | 6 ++++++
  1 file changed, 6 insertions(+)
drivers/watchdog/dw_wdt.c +0 -0
index cd578843277e..1f8605c0d712 100644
@@ -155,6 +155,9 @@ static unsigned int dw_wdt_get_min_timeout(struct dw_wdt *dw_wdt)
                     break;
Quoted 9 lines

drivers/watchdog/dw_wdt.c static unsigned int dw_wdt_get_min_timeout(struct dw_wdt *dw_wdt) · line 160

Quoted 4 lines
drivers/watchdog/dw_wdt.c +2 -0
     }
  +  if (WARN_ON_ONCE(idx == DW_WDT_NUM_TOPS))
+            idx = DW_WDT_NUM_TOPS - 1;
+

dw_wdt_get_min_timeout() returns the lowest non-0 configurable timeout. The last entry in the timeout array must not be 0, meaning there must be at least one entry in the array where the timeout is not 0. Therefore this situation can not happen.

Yes, I have seen this, you are correct, sorry for the bad patch.

Quoted 10 lines

drivers/watchdog/dw_wdt.c

Quoted 9 lines
drivers/watchdog/dw_wdt.c +0 -0
     return dw_wdt->timeouts[idx].sec;
  }
  @@ -178,6 +181,9 @@ static unsigned int dw_wdt_get_timeout(struct dw_wdt

*dw_wdt)

                     break;
     }

+ if (WARN_ON_ONCE(idx == DW_WDT_NUM_TOPS))

drivers/watchdog/dw_wdt.c +2 -0
+            idx = DW_WDT_NUM_TOPS - 1;
+
Quoted 8 lines

idx is derived from a top_val value written into WDOG_TIMEOUT_RANGE_REG_OFFSET, and the value written is derived from an entry in the timeouts array. This array contains an entry for each possible top_val. While the array is not sorted by top_val, dw_wdt_handle_tops() still guarantees that an entry exists.

I do not see how bad devicetree data can circumvent that. If it does, please provide an example and explain.

My bad, there is no problem.

Quoted 2 lines

Thanks,
Guenter

Signature · 2 lines
BRs
Schspa Shi