Skip to contentSkip to conversation

drivers: staging: wilc1000: Replace message queue with standard Linux lists

PATCH In review Chandra S Gorentla 3 participants 10 messages Sep 28, 2015 – Sep 30, 2015 Staging

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 drivers: staging: wilc1000: Replace message queue with standard Linux lists

- The message queue is replaced with standard Linux linked list
- kmem_cache is used for list members
- A check for return value of receive method is added
- GFP_ATOMIC is changed to GFP_KERNEL
- A few other related minor changes

---
- Comments of Dan Carpenter are taken care
- If receive method return value not checked, case statement for previous message may get executed if not done
- Indication of usage of 'kmem_cache' is added to the description
- Memory allocation donw outside spin_lock; GFP_ATOMIC replaced with GFP_KERNEL
- spin_lock, spin_unlock are matched
- goto statements removed
- Unrelated changes moved to a different patch
- PRINT_ERR can be taken up in a seperate patch

 drivers/staging/wilc1000/host_interface.c |  7 ++-
 drivers/staging/wilc1000/wilc_msgqueue.c  | 73 +++++++++++++------------------
 drivers/staging/wilc1000/wilc_platform.h  |  5 ++-
 3 files changed, 40 insertions(+), 45 deletions(-)
drivers/staging/wilc1000/host_interface.c +6 -1
index 62f4a8a..954656d 100644
@@ -4304,11 +4304,16 @@ static int hostIFthread(void *pvArg)
 	u32 u32Ret;
 	struct host_if_msg msg;
 	tstrWILC_WFIDrv *pstrWFIDrv;
+	int ret;
 
 	memset(&msg, 0, sizeof(struct host_if_msg));
 
 	while (1) {
-		wilc_mq_recv(&gMsgQHostIF, &msg, sizeof(struct host_if_msg), &u32Ret);
+		ret = wilc_mq_recv(&gMsgQHostIF, &msg,
+					sizeof(struct host_if_msg), &u32Ret);
+		if (ret)
+			continue;
+
 		pstrWFIDrv = (tstrWILC_WFIDrv *)msg.drvHandler;
 		if (msg.id == HOST_IF_MSG_EXIT) {
 			PRINT_D(GENERIC_DBG, "THREAD: Exiting HostIfThread\n");
drivers/staging/wilc1000/wilc_msgqueue.c +31 -42

142 lines not on this page · raw message

drivers/staging/wilc1000/wilc_platform.h +3 -2
index b763616..3871658 100644
@@ -20,7 +20,7 @@
 typedef struct __Message_struct {
 	void *pvBuffer;
 	u32 u32Length;
-	struct __Message_struct *pstrNext;
+	struct list_head link;
 } Message;
 
 typedef struct __MessageQueue_struct {
@@ -28,7 +28,8 @@ typedef struct __MessageQueue_struct {
 	spinlock_t strCriticalSection;
 	bool bExiting;
 	u32 u32ReceiversCount;
-	Message *pstrMessageList;
+	struct list_head msg_list_head;
+	struct kmem_cache *msg_cache;
 } WILC_MsgQueueHandle;
 
 
Signature · 1 line
2.1.4
#1 lore raw

PATCH 2/2 drivers: staging: wilc1000: wilc_msgqueue.c: Remove code that no effect

---
 drivers/staging/wilc1000/wilc_msgqueue.c | 5 -----
 1 file changed, 5 deletions(-)
drivers/staging/wilc1000/wilc_msgqueue.c +0 -5
index a01ada4..1a411d3 100644
@@ -133,11 +133,6 @@ int wilc_mq_recv(WILC_MsgQueueHandle *pHandle,
 	down(&pHandle->hSem);
 
 	/* other non-timeout scenarios */
-	if (result) {
-		PRINT_ER("Non-timeout\n");
-		return result;
-	}
-
 	if (pHandle->bExiting) {
 		PRINT_ER("pHandle fail\n");
 		return -EFAULT;
Signature · 1 line
2.1.4
#2 lore raw

PATCH 2/2 drivers: staging: wilc1000: wilc_msgqueue.c: Remove ineffective code

---
 drivers/staging/wilc1000/wilc_msgqueue.c | 5 -----
 1 file changed, 5 deletions(-)
drivers/staging/wilc1000/wilc_msgqueue.c +0 -5
index a01ada4..1a411d3 100644
@@ -133,11 +133,6 @@ int wilc_mq_recv(WILC_MsgQueueHandle *pHandle,
 	down(&pHandle->hSem);
 
 	/* other non-timeout scenarios */
-	if (result) {
-		PRINT_ER("Non-timeout\n");
-		return result;
-	}
-
 	if (pHandle->bExiting) {
 		PRINT_ER("pHandle fail\n");
 		return -EFAULT;
Signature · 1 line
2.1.4
#3 lore raw

Re: [PATCH 2/2] drivers: staging: wilc1000: wilc_msgqueue.c: Remove code that no effect

In reply to Chandra S Gorentla

Please do not review this. This is duplicate to -
[PATCH 2/2] drivers: staging: wilc1000: wilc_msgqueue.c: Remove ineffective code

Quoted 24 lines · Chandra S Gorentla
---
 drivers/staging/wilc1000/wilc_msgqueue.c | 5 -----
 1 file changed, 5 deletions(-)
drivers/staging/wilc1000/wilc_msgqueue.c +0 -5
index a01ada4..1a411d3 100644
@@ -133,11 +133,6 @@ int wilc_mq_recv(WILC_MsgQueueHandle *pHandle,
 	down(&pHandle->hSem);
 
 	/* other non-timeout scenarios */
-	if (result) {
-		PRINT_ER("Non-timeout\n");
-		return result;
-	}
-
 	if (pHandle->bExiting) {
 		PRINT_ER("pHandle fail\n");
 		return -EFAULT;
Signature · 1 line
2.1.4
maintainer
#4 lore raw

Re: [PATCH 2/2] drivers: staging: wilc1000: wilc_msgqueue.c: Remove code that no effect

Quoted 2 lines · Chandra Gorentla

Please do not review this. This is duplicate to -
[PATCH 2/2] drivers: staging: wilc1000: wilc_msgqueue.c: Remove ineffective code

I don't understand, you sent 2 2/2 patches, which one do I look at?

Please just resend the whole series.

thanks,

greg k-h

maintainer
#5 lore raw
In reply to Chandra S Gorentla
Quoted 2 lines · Chandra S Gorentla

- The message queue is replaced with standard Linux linked list
- kmem_cache is used for list members

Why?

Quoted 2 lines

- A check for return value of receive method is added
- GFP_ATOMIC is changed to GFP_KERNEL

Why? Are you sure that is safe?

Quoted 1 line

- A few other related minor changes

Be specific please.

thanks,

greg k-h

#6 lore raw
Quoted 5 lines · Greg KH
Quoted 2 lines · Chandra S Gorentla

- The message queue is replaced with standard Linux linked list
- kmem_cache is used for list members

Why?

The list entires are of fixed type and repeatedly allocated and deallocated for messages.

Quoted 5 lines
Quoted 3 lines

- A check for return value of receive method is added
- GFP_ATOMIC is changed to GFP_KERNEL

Why? Are you sure that is safe?

The allocations are moved out of 'spin_lock_irqsave' and 'spin_unlock_irqrestore'. Hence they are safe.

Quoted 4 lines
Quoted 2 lines

- A few other related minor changes

Be specific please.

OK. I will try to send seperate patches.

Quoted 4 lines

thanks,

greg k-h

Thank you,
chandra

#7 lore raw
In reply to Chandra S Gorentla
Quoted 5 lines · Chandra S Gorentla

- The message queue is replaced with standard Linux linked list
- kmem_cache is used for list members
- A check for return value of receive method is added
- GFP_ATOMIC is changed to GFP_KERNEL
- A few other related minor changes

These should be listed and explained.

drivers/staging/wilc1000/host_interface.c static int hostIFthread(void *pvArg) · line 4316

Quoted 8 lines
drivers/staging/wilc1000/host_interface.c +5 -1
 	while (1) {
-		wilc_mq_recv(&gMsgQHostIF, &msg, sizeof(struct host_if_msg), &u32Ret);
+		ret = wilc_mq_recv(&gMsgQHostIF, &msg,
+					sizeof(struct host_if_msg), &u32Ret);
+		if (ret)
+			continue;
+

I asked before if this was a forever loop and never got a response. Also what does this have to do with list macros?

regards,
dan carpenter

#8 lore raw
Quoted 8 lines · Dan Carpenter
Quoted 5 lines · Chandra S Gorentla

- The message queue is replaced with standard Linux linked list
- kmem_cache is used for list members
- A check for return value of receive method is added
- GFP_ATOMIC is changed to GFP_KERNEL
- A few other related minor changes

These should be listed and explained.

OK. I will try to send seperate patches.

Quoted 12 lines

drivers/staging/wilc1000/host_interface.c static int hostIFthread(void *pvArg) · line 4316

Quoted 9 lines
drivers/staging/wilc1000/host_interface.c +5 -1
 	while (1) {
-		wilc_mq_recv(&gMsgQHostIF, &msg, sizeof(struct host_if_msg), &u32Ret);
+		ret = wilc_mq_recv(&gMsgQHostIF, &msg,
+					sizeof(struct host_if_msg), &u32Ret);
+		if (ret)
+			continue;
+

I asked before if this was a forever loop and never got a response. Also what does this have to do with list macros?

The only exit condition of this loop is to receive a message 'HOST_IF_MSG_EXIT'. If this check is not there and 'wilc_mq_recv' returns an error, the switch case below it will be executed for the previously received message.

I will send this change in a different patch.

Quoted 3 lines

regards,
dan carpenter

#9 lore raw
Quoted 10 lines · Chandra Gorentla
Quoted 10 lines

drivers/staging/wilc1000/host_interface.c static int hostIFthread(void *pvArg) · line 4316

Quoted 7 lines
drivers/staging/wilc1000/host_interface.c +5 -1
 	while (1) {
-		wilc_mq_recv(&gMsgQHostIF, &msg, sizeof(struct host_if_msg), &u32Ret);
+		ret = wilc_mq_recv(&gMsgQHostIF, &msg,
+					sizeof(struct host_if_msg), &u32Ret);
+		if (ret)
+			continue;
+

I asked before if this was a forever loop and never got a response. Also what does this have to do with list macros?

Quoted 4 lines

The only exit condition of this loop is to receive a message 'HOST_IF_MSG_EXIT'. If this check is not there and 'wilc_mq_recv' returns an error, the switch case below it will be executed for the previously received message.

Oh, hm... It looks like wilc_mq_recv() can return -EFAULT, -EOVERFLOW or success here. If it returns -EFUALT is calling wilc_mq_recv() again really the right thing? I suspect we should break in that case. We should probably at least sleep for a bit intsead of looping so tightly if it returns -EOVERFLOW?

I'm not sure.

regards,
dan carpenter