Skip to contentSkip to conversation

idpf: add missing cpu_to_le32 in idpf_tx_splitq_build_flow_desc

PATCH In review Willem de Bruijn 4 participants 7 messages Aug 3 – Aug 6

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

PATCH v2 1/1 idpf: add missing cpu_to_le32 in idpf_tx_splitq_build_flow_desc

From: Willem de Bruijn <[email protected]>

idpf_tx_splitq_build_flow_desc performs a 32-bit store to &cmd_dtype to set the 8-bit cmd_dtype and zero the adjacent 3-byte timestamp field in a single operation.

Descriptors are in little endian. Add missing cpu_to_le32 and cast to __le32 to ensure the fields are written correctly also on big endian platforms.

Fixes 1a49cf814fe1 ("idpf: add Tx timestamp flows")

---

Changes
v1 -> v2
- add Fixes tag, Tony's Reviewed-by and Cc: [email protected]
v1: https://lore.kernel.org/netdev/[email protected]/

---
 drivers/net/ethernet/intel/idpf/idpf_txrx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/idpf/idpf_txrx.c +1 -1
index c724d429a7aa..91ca75e45463 100644
@@ -2408,7 +2408,7 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
 				    struct idpf_tx_splitq_params *params,
 				    u16 td_cmd, u16 size)
 {
-	*(u32 *)&desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+	*(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));
 	desc->flow.qw1.rxr_bufsize = cpu_to_le16((u16)size);
 	desc->flow.qw1.compl_tag = cpu_to_le16(params->compl_tag);
 }
Signature · 1 line
2.55.0.629.g250fe7f194-goog
Reviewed-by on v2 1/1
#1 lore raw
Quoted 7 lines · Willem de Bruijn

From: Willem de Bruijn <[email protected]>

idpf_tx_splitq_build_flow_desc performs a 32-bit store to &cmd_dtype to set the 8-bit cmd_dtype and zero the adjacent 3-byte timestamp field in a single operation.

Quoted 7 lines

Descriptors are in little endian. Add missing cpu_to_le32 and cast to __le32 to ensure the fields are written correctly also on big endian platforms.

Fixes 1a49cf814fe1 ("idpf: add Tx timestamp flows")

Thanks,
Jason

Reviewed-by on v2 1/1
#2 lore raw
In reply to Willem de Bruijn
Quoted 58 lines

-----Original Message-----
From: Intel-wired-lan <[email protected]> On Behalf Of Willem de Bruijn
Sent: Monday, August 3, 2026 11:06 PM
To: [email protected]
Cc: [email protected]; Nguyen, Anthony L <[email protected]>; Hay, Joshua A <[email protected]>; Kitszel, Przemyslaw <[email protected]>; Willem de Bruijn <[email protected]>
Subject: [Intel-wired-lan] [PATCH iwl-net v2] idpf: add missing cpu_to_le32 in idpf_tx_splitq_build_flow_desc

From: Willem de Bruijn <[email protected]>

idpf_tx_splitq_build_flow_desc performs a 32-bit store to &cmd_dtype to set the 8-bit cmd_dtype and zero the adjacent 3-byte timestamp field in a single operation.

Descriptors are in little endian. Add missing cpu_to_le32 and cast to __le32 to ensure the fields are written correctly also on big endian platforms.

Fixes 1a49cf814fe1 ("idpf: add Tx timestamp flows")

---

Changes
v1 -> v2
- add Fixes tag, Tony's Reviewed-by and Cc: intel-wired-
[email protected]
v1: https://lore.kernel.org/netdev/20260731103137.4000876-1-
[email protected]/

---
 drivers/net/ethernet/intel/idpf/idpf_txrx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/idpf/idpf_txrx.c +1 -1
index c724d429a7aa..91ca75e45463 100644
@@ -2408,7 +2408,7 @@ void idpf_tx_splitq_build_flow_desc(union
 idpf_tx_flex_desc *desc,
 				    struct idpf_tx_splitq_params *params,
 				    u16 td_cmd, u16 size)
 {
-	*(u32 *)&desc->flow.qw1.cmd_dtype = (u8)(params->dtype |
 td_cmd);
+	*(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params-
Quoted 1 line

dtype

+| td_cmd));

drivers/net/ethernet/intel/idpf/idpf_txrx.c +0 -1
 	desc->flow.qw1.rxr_bufsize = cpu_to_le16((u16)size);
 	desc->flow.qw1.compl_tag = cpu_to_le16(params->compl_tag);  }
--
Quoted 1 line

2.55.0.629.g250fe7f194-goog

#3 lore raw
In reply to Willem de Bruijn
Quoted 29 lines · Willem de Bruijn

From: Willem de Bruijn <[email protected]>

idpf_tx_splitq_build_flow_desc performs a 32-bit store to &cmd_dtype to set the 8-bit cmd_dtype and zero the adjacent 3-byte timestamp field in a single operation.

Descriptors are in little endian. Add missing cpu_to_le32 and cast to __le32 to ensure the fields are written correctly also on big endian platforms.

Fixes 1a49cf814fe1 ("idpf: add Tx timestamp flows")

---

Changes
v1 -> v2
- add Fixes tag, Tony's Reviewed-by and Cc: [email protected]
v1: https://lore.kernel.org/netdev/[email protected]/

---
 drivers/net/ethernet/intel/idpf/idpf_txrx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/ethernet/intel/idpf/idpf_txrx.c +0 -0
index c724d429a7aa..91ca75e45463 100644

drivers/net/ethernet/intel/idpf/idpf_txrx.c void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc, · line 2411

Quoted 5 lines
drivers/net/ethernet/intel/idpf/idpf_txrx.c +1 -1
@@ -2408,7 +2408,7 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
 				    struct idpf_tx_splitq_params *params,
 				    u16 td_cmd, u16 size)
 {
-	*(u32 *)&desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+	*(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));

While it technically works, I find it unreadable. Looking at this line without reading commit msg, it looks like a bug where a 4-byte value is written to u8 field. The intent to zero an adjacent, unrelated field is not clear. Consider assigning these 4 fields manually, or adding a comment/function doc.

Thanks,
Marcin

Quoted 3 lines
drivers/net/ethernet/intel/idpf/idpf_txrx.c +0 -0
 	desc->flow.qw1.rxr_bufsize = cpu_to_le16((u16)size);
 	desc->flow.qw1.compl_tag = cpu_to_le16(params->compl_tag);
 }
#4 lore raw
Quoted 27 lines · Marcin Szycik
Quoted 24 lines · Willem de Bruijn

From: Willem de Bruijn <[email protected]>

idpf_tx_splitq_build_flow_desc performs a 32-bit store to &cmd_dtype to set the 8-bit cmd_dtype and zero the adjacent 3-byte timestamp field in a single operation.

Descriptors are in little endian. Add missing cpu_to_le32 and cast to __le32 to ensure the fields are written correctly also on big endian platforms.

Fixes 1a49cf814fe1 ("idpf: add Tx timestamp flows")

---

Changes
v1 -> v2
- add Fixes tag, Tony's Reviewed-by and Cc: [email protected]
v1: https://lore.kernel.org/netdev/[email protected]/

---
 drivers/net/ethernet/intel/idpf/idpf_txrx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 16 lines

drivers/net/ethernet/intel/idpf/idpf_txrx.c void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc, · line 2411

Quoted 10 lines
drivers/net/ethernet/intel/idpf/idpf_txrx.c +1 -1
index c724d429a7aa..91ca75e45463 100644
@@ -2408,7 +2408,7 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
 				    struct idpf_tx_splitq_params *params,
 				    u16 td_cmd, u16 size)
 {
-	*(u32 *)&desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+	*(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));

While it technically works, I find it unreadable. Looking at this line without reading commit msg, it looks like a bug where a 4-byte value is written to u8 field. The intent to zero an adjacent, unrelated field is not clear. Consider assigning these 4 fields manually, or adding a comment/function doc.

This is a bug fix, which generally should be the smallest surgical change only.

That said, I do agree on the style, and a follow-up for net-next will clean that up too:

@@ -2408,7 +2410,12 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
struct idpf_tx_splitq_params *params, u16 td_cmd, u16 size)
{
- *(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));
+ desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+
+ desc->flow.qw1.ts[0] = params->offload.desc_ts[0];
+ desc->flow.qw1.ts[1] = params->offload.desc_ts[1];
+ desc->flow.qw1.ts[2] = params->offload.desc_ts[2];

The fix is queued. If there is consensus, I can respin. But given that we will clean this up in net-next properly, my preference is to keep the minimal fix as is.

#5 lore raw
Quoted 45 lines · Willem de Bruijn
Quoted 27 lines · Marcin Szycik
Quoted 24 lines · Willem de Bruijn

From: Willem de Bruijn <[email protected]>

idpf_tx_splitq_build_flow_desc performs a 32-bit store to &cmd_dtype to set the 8-bit cmd_dtype and zero the adjacent 3-byte timestamp field in a single operation.

Descriptors are in little endian. Add missing cpu_to_le32 and cast to __le32 to ensure the fields are written correctly also on big endian platforms.

Fixes 1a49cf814fe1 ("idpf: add Tx timestamp flows")

---

Changes
v1 -> v2
- add Fixes tag, Tony's Reviewed-by and Cc: [email protected]
v1: https://lore.kernel.org/netdev/[email protected]/

---
 drivers/net/ethernet/intel/idpf/idpf_txrx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 16 lines

drivers/net/ethernet/intel/idpf/idpf_txrx.c void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc, · line 2411

Quoted 10 lines
drivers/net/ethernet/intel/idpf/idpf_txrx.c +1 -1
index c724d429a7aa..91ca75e45463 100644
@@ -2408,7 +2408,7 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
 				    struct idpf_tx_splitq_params *params,
 				    u16 td_cmd, u16 size)
 {
-	*(u32 *)&desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+	*(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));

While it technically works, I find it unreadable. Looking at this line without reading commit msg, it looks like a bug where a 4-byte value is written to u8 field. The intent to zero an adjacent, unrelated field is not clear. Consider assigning these 4 fields manually, or adding a comment/function doc.

Quoted 2 lines

This is a bug fix, which generally should be the smallest surgical change only.

Agreed.

Quoted 14 lines

That said, I do agree on the style, and a follow-up for net-next will clean that up too:

@@ -2408,7 +2410,12 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
struct idpf_tx_splitq_params *params, u16 td_cmd, u16 size)
{
- *(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));
+ desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+
+ desc->flow.qw1.ts[0] = params->offload.desc_ts[0];
+ desc->flow.qw1.ts[1] = params->offload.desc_ts[1];
+ desc->flow.qw1.ts[2] = params->offload.desc_ts[2];

Quoted 3 lines

The fix is queued. If there is consensus, I can respin. But given that we will clean this up in net-next properly, my preference is to keep the minimal fix as is.

I'm fine with this. Was the cleanup patch already posted to next?

Thanks,
Marcin

#6 lore raw
Quoted 66 lines · Marcin Szycik
Quoted 45 lines · Willem de Bruijn
Quoted 27 lines · Marcin Szycik
Quoted 24 lines · Willem de Bruijn

From: Willem de Bruijn <[email protected]>

idpf_tx_splitq_build_flow_desc performs a 32-bit store to &cmd_dtype to set the 8-bit cmd_dtype and zero the adjacent 3-byte timestamp field in a single operation.

Descriptors are in little endian. Add missing cpu_to_le32 and cast to __le32 to ensure the fields are written correctly also on big endian platforms.

Fixes 1a49cf814fe1 ("idpf: add Tx timestamp flows")

---

Changes
v1 -> v2
- add Fixes tag, Tony's Reviewed-by and Cc: [email protected]
v1: https://lore.kernel.org/netdev/[email protected]/

---
 drivers/net/ethernet/intel/idpf/idpf_txrx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 16 lines

drivers/net/ethernet/intel/idpf/idpf_txrx.c void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc, · line 2411

Quoted 10 lines
drivers/net/ethernet/intel/idpf/idpf_txrx.c +1 -1
index c724d429a7aa..91ca75e45463 100644
@@ -2408,7 +2408,7 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
 				    struct idpf_tx_splitq_params *params,
 				    u16 td_cmd, u16 size)
 {
-	*(u32 *)&desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+	*(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));

While it technically works, I find it unreadable. Looking at this line without reading commit msg, it looks like a bug where a 4-byte value is written to u8 field. The intent to zero an adjacent, unrelated field is not clear. Consider assigning these 4 fields manually, or adding a comment/function doc.

Quoted 2 lines

This is a bug fix, which generally should be the smallest surgical change only.

Agreed.

Quoted 14 lines

That said, I do agree on the style, and a follow-up for net-next will clean that up too:

@@ -2408,7 +2410,12 @@ void idpf_tx_splitq_build_flow_desc(union idpf_tx_flex_desc *desc,
struct idpf_tx_splitq_params *params, u16 td_cmd, u16 size)
{
- *(__le32 *)&desc->flow.qw1.cmd_dtype = cpu_to_le32((u8)(params->dtype | td_cmd));
+ desc->flow.qw1.cmd_dtype = (u8)(params->dtype | td_cmd);
+
+ desc->flow.qw1.ts[0] = params->offload.desc_ts[0];
+ desc->flow.qw1.ts[1] = params->offload.desc_ts[1];
+ desc->flow.qw1.ts[2] = params->offload.desc_ts[2];

Quoted 5 lines
Quoted 3 lines

The fix is queued. If there is consensus, I can respin. But given that we will clean this up in net-next properly, my preference is to keep the minimal fix as is.

I'm fine with this.

Great.

Quoted 1 line

Was the cleanup patch already posted to next?

Not yet. I'll try to send the series later today.

It will have a (trivial) conflict with this line. But I will likely need to respin again, so don't want t wait for that (and the merge window).