Skip to contentSkip to conversation

e100: fix shift-out-of-bounds in e100_eeprom_load()

PATCH In review Malathi 3 participants 3 messages Aug 11 – Aug 12

Subsystems (3)
lore
Share & tools
lore

Malathi

author
#0 lore raw

PATCH v2 1/1 e100: fix shift-out-of-bounds in e100_eeprom_load()

e100_eeprom_load() and e100_eeprom_save() start with an address length of 8 and call e100_eeprom_read() to auto-detect the real EEPROM address length. e100_eeprom_read() adjusts the length with

	*addr_len -= (i - 16);

based on when the EEPROM drives a dummy zero onto EEDO. A malfunctioning or emulated device that drives EEDO low too early makes (i - 16) exceed the current length, underflowing the u16 addr_len to a large value such as 65529.

That value is then used as a shift count:

	nic->eeprom_wc = 1 << addr_len;

which is undefined behaviour and additionally overflows the fixed-size nic->eeprom[256] cache.

  UBSAN: shift-out-of-bounds in drivers/net/ethernet/intel/e100.c:768:21
  shift exponent 65529 is too large for 32-bit type 'int'

The same corrupted addr_len is also fed back into e100_eeprom_read() for every subsequent word, where it is used as a shift count again:

	cmd_addr_data = ((op_read << *addr_len) | addr) << 16;

so validating the length only once at the caller is not enough.

Clamp the length in e100_eeprom_read() so the subtraction can never underflow the u16, and reject a zero or out-of-range length in e100_eeprom_load() and e100_eeprom_save() before using it. The EEPROM cache holds at most 256 words, so a valid address length is in [1, 8].

---
v2:
- Drop the Fixes: tag and retarget to net-next; the underflow is only reachable with malfunctioning or emulated hardware, so this is a hardening change and not stable material.
- Clamp addr_len inside e100_eeprom_read() so the auto-detect subtraction can never underflow the u16. The corrupted length was otherwise reused as a shift count for every subsequent word, so validating it only once at the callers (as in v1) was not enough.
- Also reject a zero address length in e100_eeprom_load() and e100_eeprom_save().

 drivers/net/ethernet/intel/e100.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)
drivers/net/ethernet/intel/e100.c +16 -1
index 29960762e64a..26a7c0aaa6e2 100644
@@ -744,7 +744,12 @@ static __le16 e100_eeprom_read(struct nic *nic, u16 *addr_len, u16 addr)
 		 * complete address.  Use this to adjust addr_len. */
 		ctrl = ioread8(&nic->csr->eeprom_ctrl_lo);
 		if (!(ctrl & eedo) && i > 16) {
-			*addr_len -= (i - 16);
+			u16 len = i - 16;
+
+			if (len > *addr_len)
+				*addr_len = 0;
+			else
+				*addr_len -= len;
 			i = 17;
 		}
 
@@ -765,6 +770,11 @@ static int e100_eeprom_load(struct nic *nic)
 
 	/* Try reading with an 8-bit addr len to discover actual addr len */
 	e100_eeprom_read(nic, &addr_len, 0);
+	if (!addr_len || addr_len > 8) {
+		netif_err(nic, probe, nic->netdev,
+			  "invalid EEPROM address length %u\n", addr_len);
+		return -EINVAL;
+	}
 	nic->eeprom_wc = 1 << addr_len;
 
 	for (addr = 0; addr < nic->eeprom_wc; addr++) {
@@ -791,6 +801,11 @@ static int e100_eeprom_save(struct nic *nic, u16 start, u16 count)
 
 	/* Try reading with an 8-bit addr len to discover actual addr len */
 	e100_eeprom_read(nic, &addr_len, 0);
+	if (!addr_len || addr_len > 8) {
+		netif_err(nic, probe, nic->netdev,
+			  "invalid EEPROM address length %u\n", addr_len);
+		return -EINVAL;
+	}
 	nic->eeprom_wc = 1 << addr_len;
 
 	if (start + count >= nic->eeprom_wc)
Signature · 1 line
2.43.0
Reviewed-by on v2 1/1
#1 lore raw
Quoted 118 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Malathi
Sent: Tuesday, August 11, 2026 7:08 AM
To: Nguyen, Anthony L <[email protected]>; Kitszel, Przemyslaw <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]; Malathi <[email protected]>; [email protected]
Subject: [Intel-wired-lan] [PATCH net-next v2] e100: fix shift-out-of- bounds in e100_eeprom_load()

e100_eeprom_load() and e100_eeprom_save() start with an address length of 8 and call e100_eeprom_read() to auto-detect the real EEPROM address length. e100_eeprom_read() adjusts the length with

	*addr_len -= (i - 16);

based on when the EEPROM drives a dummy zero onto EEDO. A malfunctioning or emulated device that drives EEDO low too early makes (i - 16) exceed the current length, underflowing the u16 addr_len to a large value such as 65529.

That value is then used as a shift count:

	nic->eeprom_wc = 1 << addr_len;

which is undefined behaviour and additionally overflows the fixed-size nic->eeprom[256] cache.

  UBSAN: shift-out-of-bounds in

drivers/net/ethernet/intel/e100.c:768:21
shift exponent 65529 is too large for 32-bit type 'int'

The same corrupted addr_len is also fed back into e100_eeprom_read() for every subsequent word, where it is used as a shift count again:

	cmd_addr_data = ((op_read << *addr_len) | addr) << 16;

so validating the length only once at the caller is not enough.

Clamp the length in e100_eeprom_read() so the subtraction can never underflow the u16, and reject a zero or out-of-range length in e100_eeprom_load() and e100_eeprom_save() before using it. The EEPROM cache holds at most 256 words, so a valid address length is in [1, 8].

---
v2:
- Drop the Fixes: tag and retarget to net-next; the underflow is only reachable with malfunctioning or emulated hardware, so this is a hardening change and not stable material.
- Clamp addr_len inside e100_eeprom_read() so the auto-detect subtraction can never underflow the u16. The corrupted length was otherwise reused as a shift count for every subsequent word, so validating it only once at the callers (as in v1) was not enough.
- Also reject a zero address length in e100_eeprom_load() and e100_eeprom_save().

 drivers/net/ethernet/intel/e100.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)
drivers/net/ethernet/intel/e100.c +6 -1
index 29960762e64a..26a7c0aaa6e2 100644
@@ -744,7 +744,12 @@ static __le16 e100_eeprom_read(struct nic *nic,
 u16 *addr_len, u16 addr)
 		 * complete address.  Use this to adjust addr_len. */
 		ctrl = ioread8(&nic->csr->eeprom_ctrl_lo);
 		if (!(ctrl & eedo) && i > 16) {
-			*addr_len -= (i - 16);
+			u16 len = i - 16;
+
+			if (len > *addr_len)
+				*addr_len = 0;
+			else
+				*addr_len -= len;
 			i = 17;
 		}
@@ -765,6 +770,11 @@ static int e100_eeprom_load(struct nic *nic)
 
 	/* Try reading with an 8-bit addr len to discover actual addr

len */

drivers/net/ethernet/intel/e100.c +3 -0
 	e100_eeprom_read(nic, &addr_len, 0);
+	if (!addr_len || addr_len > 8) {
+		netif_err(nic, probe, nic->netdev,
+			  "invalid EEPROM address length %u\n",

addr_len);

drivers/net/ethernet/intel/e100.c +3 -0
+		return -EINVAL;
+	}
 	nic->eeprom_wc = 1 << addr_len;
 
 	for (addr = 0; addr < nic->eeprom_wc; addr++) { @@ -791,6
+801,11 @@ static int e100_eeprom_save(struct nic *nic, u16 start, u16

count)

/* Try reading with an 8-bit addr len to discover actual addr
len */

drivers/net/ethernet/intel/e100.c +3 -0
 	e100_eeprom_read(nic, &addr_len, 0);
+	if (!addr_len || addr_len > 8) {
+		netif_err(nic, probe, nic->netdev,
+			  "invalid EEPROM address length %u\n",

addr_len);

drivers/net/ethernet/intel/e100.c +2 -1
+		return -EINVAL;
+	}
 	nic->eeprom_wc = 1 << addr_len;
 
 	if (start + count >= nic->eeprom_wc)
--
Quoted 1 line

2.43.0

#2 lore raw
Quoted 116 lines · Aleksandr
Quoted 116 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Malathi
Sent: Tuesday, August 11, 2026 7:08 AM
To: Nguyen, Anthony L <[email protected]>; Kitszel, Przemyslaw <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]; Malathi <[email protected]>; [email protected]
Subject: [Intel-wired-lan] [PATCH net-next v2] e100: fix shift-out-of- bounds in e100_eeprom_load()

e100_eeprom_load() and e100_eeprom_save() start with an address length of 8 and call e100_eeprom_read() to auto-detect the real EEPROM address length. e100_eeprom_read() adjusts the length with

	*addr_len -= (i - 16);

based on when the EEPROM drives a dummy zero onto EEDO. A malfunctioning or emulated device that drives EEDO low too early makes (i - 16) exceed the current length, underflowing the u16 addr_len to a large value such as 65529.

That value is then used as a shift count:

	nic->eeprom_wc = 1 << addr_len;

which is undefined behaviour and additionally overflows the fixed-size nic->eeprom[256] cache.

  UBSAN: shift-out-of-bounds in

drivers/net/ethernet/intel/e100.c:768:21
shift exponent 65529 is too large for 32-bit type 'int'

The same corrupted addr_len is also fed back into e100_eeprom_read() for every subsequent word, where it is used as a shift count again:

	cmd_addr_data = ((op_read << *addr_len) | addr) << 16;

so validating the length only once at the caller is not enough.

Clamp the length in e100_eeprom_read() so the subtraction can never underflow the u16, and reject a zero or out-of-range length in e100_eeprom_load() and e100_eeprom_save() before using it. The EEPROM cache holds at most 256 words, so a valid address length is in [1, 8].

---
v2:
- Drop the Fixes: tag and retarget to net-next; the underflow is only reachable with malfunctioning or emulated hardware, so this is a hardening change and not stable material.
- Clamp addr_len inside e100_eeprom_read() so the auto-detect subtraction can never underflow the u16. The corrupted length was otherwise reused as a shift count for every subsequent word, so validating it only once at the callers (as in v1) was not enough.
- Also reject a zero address length in e100_eeprom_load() and e100_eeprom_save().

 drivers/net/ethernet/intel/e100.c | 17 ++++++++++++++++-
 1 file changed, 16 insertions(+), 1 deletion(-)
drivers/net/ethernet/intel/e100.c +6 -1
index 29960762e64a..26a7c0aaa6e2 100644
@@ -744,7 +744,12 @@ static __le16 e100_eeprom_read(struct nic *nic,
 u16 *addr_len, u16 addr)
 		 * complete address.  Use this to adjust addr_len. */
 		ctrl = ioread8(&nic->csr->eeprom_ctrl_lo);
 		if (!(ctrl & eedo) && i > 16) {
-			*addr_len -= (i - 16);
+			u16 len = i - 16;
+
+			if (len > *addr_len)
+				*addr_len = 0;
+			else
+				*addr_len -= len;
 			i = 17;
 		}
@@ -765,6 +770,11 @@ static int e100_eeprom_load(struct nic *nic)
 
 	/* Try reading with an 8-bit addr len to discover actual addr

len */

drivers/net/ethernet/intel/e100.c +3 -0
 	e100_eeprom_read(nic, &addr_len, 0);
+	if (!addr_len || addr_len > 8) {
+		netif_err(nic, probe, nic->netdev,
+			  "invalid EEPROM address length %u\n",

addr_len);

drivers/net/ethernet/intel/e100.c +3 -0
+		return -EINVAL;
+	}
 	nic->eeprom_wc = 1 << addr_len;
 
 	for (addr = 0; addr < nic->eeprom_wc; addr++) { @@ -791,6
+801,11 @@ static int e100_eeprom_save(struct nic *nic, u16 start, u16

count)

/* Try reading with an 8-bit addr len to discover actual addr
len */

drivers/net/ethernet/intel/e100.c +3 -0
 	e100_eeprom_read(nic, &addr_len, 0);
+	if (!addr_len || addr_len > 8) {
+		netif_err(nic, probe, nic->netdev,
+			  "invalid EEPROM address length %u\n",

addr_len);

drivers/net/ethernet/intel/e100.c +2 -0
+		return -EINVAL;
+	}
 	nic->eeprom_wc = 1 << addr_len;
Quoted 5 lines
Quoted 3 lines

if (start + count >= nic->eeprom_wc)
--
2.43.0

Hi Aleksandr Loktionov,

could you please clarify something regarding this issue?

I submitted this code as part of the bug report on July 7 v1[1] and July 10 v2 [2] and we were already discussing it.

Later on July 10, the same code lines were submitted again by her.

I noticed that her submission was reviewed, even though i had already shared the same code earlier.

Could you please clarify why her submission was reviewed instead of the original one i submitted?

I also sent you reply on v2 [2]. Could you please review that as well when you get a chance?

I just want to understand how the submissions are being considered in this case and make sure the contribution timeline is clear.

[1] https://lore.kernel.org/all/[email protected]/T/
[2] https://lore.kernel.org/all/[email protected]/T/

Thank you,
Pavan