Skip to contentSkip to conversation

net: lan743x: Don't sleep in atomic context

PATCH Superseded Moritz Fischer 2 participants 2 messages Jun 25, 2023 – Jun 26, 2023

Subsystems (3)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 net: lan743x: Don't sleep in atomic context

dev_set_rx_mode() grabs a spin_lock, and the lan743x implementation proceeds subsequently to go to sleep using readx_poll_timeout().

Introduce a helper wrapping the readx_poll_timeout_atomic() function and use it to replace the calls to readx_polL_timeout().

---

Changes from v1:
- Added line-breaks
- Changed subject to target net-next
- Removed Tested-by: tag

---
 drivers/net/ethernet/microchip/lan743x_main.c | 22 +++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)
drivers/net/ethernet/microchip/lan743x_main.c +18 -4
index f1bded993edc..4f277ffff1dc 100644
@@ -144,6 +144,19 @@ static int lan743x_csr_light_reset(struct lan743x_adapter *adapter)
 				  !(data & HW_CFG_LRST_), 100000, 10000000);
 }
 
+static int lan743x_csr_wait_for_bit_atomic(struct lan743x_adapter *adapter,
+					   int offset, u32 bit_mask,
+					   int target_value, int udelay_min,
+					   int udelay_max, int count)
+{
+	u32 data;
+
+	return readx_poll_timeout_atomic(LAN743X_CSR_READ_OP, offset, data,
+					 target_value == ((data & bit_mask) ?
+					 1 : 0), udelay_max,
+					 udelay_min * count);
+}
+
 static int lan743x_csr_wait_for_bit(struct lan743x_adapter *adapter,
 				    int offset, u32 bit_mask,
 				    int target_value, int usleep_min,
@@ -736,8 +749,8 @@ static int lan743x_dp_write(struct lan743x_adapter *adapter,
 	u32 dp_sel;
 	int i;
 
-	if (lan743x_csr_wait_for_bit(adapter, DP_SEL, DP_SEL_DPRDY_,
-				     1, 40, 100, 100))
+	if (lan743x_csr_wait_for_bit_atomic(adapter, DP_SEL, DP_SEL_DPRDY_,
+					    1, 40, 100, 100))
 		return -EIO;
 	dp_sel = lan743x_csr_read(adapter, DP_SEL);
 	dp_sel &= ~DP_SEL_MASK_;
@@ -748,8 +761,9 @@ static int lan743x_dp_write(struct lan743x_adapter *adapter,
 		lan743x_csr_write(adapter, DP_ADDR, addr + i);
 		lan743x_csr_write(adapter, DP_DATA_0, buf[i]);
 		lan743x_csr_write(adapter, DP_CMD, DP_CMD_WRITE_);
-		if (lan743x_csr_wait_for_bit(adapter, DP_SEL, DP_SEL_DPRDY_,
-					     1, 40, 100, 100))
+		if (lan743x_csr_wait_for_bit_atomic(adapter, DP_SEL,
+						    DP_SEL_DPRDY_,
+						    1, 40, 100, 100))
 			return -EIO;
 	}
 
Signature · 1 line
2.41.0.178.g377b9f9a00-goog
maintainer
#1 lore raw
Quoted 6 lines · On Sun, 25 Jun 2023 18:23:27 +0000 Moritz Fischer wrote:

dev_set_rx_mode() grabs a spin_lock, and the lan743x implementation proceeds subsequently to go to sleep using readx_poll_timeout().

Introduce a helper wrapping the readx_poll_timeout_atomic() function and use it to replace the calls to readx_polL_timeout().

Quoted 7 lines

---

Changes from v1:
- Added line-breaks
- Changed subject to target net-next
- Removed Tested-by: tag

Sleeping in atomic context is a bug, this is a bug fix. You should add a Fixes tag (probably 23f0703c125be AFAICT) and target the patch at net rather than net-next.
Make sure you CC the authors of the patch under fixes (get_maintainer will point them out if run on a patch file with a Fixes tag) One more nit below...

Quoted 11 lines
 drivers/net/ethernet/microchip/lan743x_main.c | 22 +++++++++++++++----
 1 file changed, 18 insertions(+), 4 deletions(-)
drivers/net/ethernet/microchip/lan743x_main.c +0 -0
index f1bded993edc..4f277ffff1dc 100644
@@ -144,6 +144,19 @@ static int lan743x_csr_light_reset(struct lan743x_adapter *adapter)
 				  !(data & HW_CFG_LRST_), 100000, 10000000);
 }
 

drivers/net/ethernet/microchip/lan743x_main.c static int lan743x_csr_light_reset(struct lan743x_adapter *adapter) · line 156

Quoted 10 lines
drivers/net/ethernet/microchip/lan743x_main.c +10 -0
+static int lan743x_csr_wait_for_bit_atomic(struct lan743x_adapter *adapter,
+					   int offset, u32 bit_mask,
+					   int target_value, int udelay_min,
+					   int udelay_max, int count)
+{
+	u32 data;
+
+	return readx_poll_timeout_atomic(LAN743X_CSR_READ_OP, offset, data,
+					 target_value == ((data & bit_mask) ?
+					 1 : 0), udelay_max,

You can save the awkward wrapping by using a double negation:

					target_value == !!(data & bit_mask)

this is a fairly common form in the kernel.

Signature · 1 line
pw-bot: cr