Skip to contentSkip to conversation

rt2860sta: Fix argument to linux_pci_unmap_single()

PATCH In review Ben Hutchings 5 participants 8 messages Mar 29, 2010 – Apr 28, 2010 1 message not in the archive Staging

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

PATCH 1/1 rt2860sta: Fix argument to linux_pci_unmap_single()

John Halton wrote in <http://bugs.debian.org/575726>:

Quoted 3 lines

Whenever wpa_supplicant is deactivated (whether by killing the process or during a normal shutdown) I am getting a kerneloops that prevents the computer from completing shutdown. Here is the relevant syslog output:

The backtrace points to an incorrect call from RTMPFreeTxRxRingMemory() into linux_pci_unmap_single(). This appears to have been fixed in Linux 2.6.33 by this change:

commit ca97b8388838ee9ea4b4bad04948f8f7f8a607a3

Author: Bartlomiej Zolnierkiewicz <[email protected]>
Date:   Tue Sep 22 20:44:07 2009 +0200

Staging: rt28x0: updates from vendor's V2.1.0.0 drivers

For stable-2.6.32, just fix this one function call.

---
 drivers/staging/rt2860/common/2860_rtmp_init.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
drivers/staging/rt2860/common/2860_rtmp_init.c +1 -1
index 0bc0fb9..98b0f8e 100644
@@ -716,7 +716,7 @@ VOID RTMPFreeTxRxRingMemory(
 	{
 		if ((pAd->RxRing.Cell[index].DmaBuf.AllocVa) && (pAd->RxRing.Cell[index].pNdisPacket))
 		{
-			PCI_UNMAP_SINGLE(pObj->pci_dev, pAd->RxRing.Cell[index].DmaBuf.AllocPa, pAd->RxRing.Cell[index].DmaBuf.AllocSize, PCI_DMA_FROMDEVICE);
+			PCI_UNMAP_SINGLE(pAd, pAd->RxRing.Cell[index].DmaBuf.AllocPa, pAd->RxRing.Cell[index].DmaBuf.AllocSize, PCI_DMA_FROMDEVICE);
 			RELEASE_NDIS_PACKET(pAd, pAd->RxRing.Cell[index].pNdisPacket, NDIS_STATUS_SUCCESS);
 		}
 	}
Signature · 6 lines
1.7.0.3


-- 
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.
#2 lore raw

PATCH 1/1 rt28xx: Make PCI_{MAP,UNMAP}_SINGLE type-safe

To avoid recurrence of bugs such as <http://bugs.debian.org/575726>,
change the type of the first parameter to linux_pci_{map,unmap}_single() from void * to struct rt_rtmp_adapter *. Also do not define the macros PCI_{MAP,UNMAP}_SINGLE() when building the rt2870sta driver; they are not used and if they were that would be a bug.

---
 drivers/staging/rt2860/rt_linux.h    |   14 +++++---------
 drivers/staging/rt2860/rt_pci_rbus.c |   12 ++++--------
 2 files changed, 9 insertions(+), 17 deletions(-)
drivers/staging/rt2860/rt_linux.h +5 -9
index a7c540f..b370fb2 100644
@@ -455,10 +455,11 @@ void hex_dump(char *str, unsigned char *pSrcBufVA, unsigned int SrcBufLen);
  * Device DMA Access related definitions and data structures.
  **********************************************************************************/
 #ifdef RTMP_MAC_PCI
-dma_addr_t linux_pci_map_single(void *handle, void *ptr, size_t size,
-				int sd_idx, int direction);
-void linux_pci_unmap_single(void *handle, dma_addr_t dma_addr, size_t size,
-			    int direction);
+struct rt_rtmp_adapter;
+dma_addr_t linux_pci_map_single(struct rt_rtmp_adapter *pAd, void *ptr,
+				size_t size, int sd_idx, int direction);
+void linux_pci_unmap_single(struct rt_rtmp_adapter *pAd, dma_addr_t dma_addr,
+			    size_t size, int direction);
 
 #define PCI_MAP_SINGLE(_handle, _ptr, _size, _sd_idx, _dir) \
 	linux_pci_map_single(_handle, _ptr, _size, _sd_idx, _dir)
@@ -475,11 +476,6 @@ void linux_pci_unmap_single(void *handle, dma_addr_t dma_addr, size_t size,
 #define DEV_ALLOC_SKB(_length) \
 	dev_alloc_skb(_length)
 #endif /* RTMP_MAC_PCI // */
-#ifdef RTMP_MAC_USB
-#define PCI_MAP_SINGLE(_handle, _ptr, _size, _dir) (unsigned long)0
-
-#define PCI_UNMAP_SINGLE(_handle, _ptr, _size, _dir)
-#endif /* RTMP_MAC_USB // */
 
 /*
  * unsigned long
drivers/staging/rt2860/rt_pci_rbus.c +4 -8
index e0a0aee..acdf09f 100644
@@ -790,10 +790,9 @@ IRQ_HANDLE_TYPE rt2860_interrupt(int irq, void *dev_instance)
  * invaild or writeback cache
  * and convert virtual address to physical address
  */
-dma_addr_t linux_pci_map_single(void *handle, void *ptr, size_t size,
-				int sd_idx, int direction)
+dma_addr_t linux_pci_map_single(struct rt_rtmp_adapter *pAd, void *ptr,
+				size_t size, int sd_idx, int direction)
 {
-	struct rt_rtmp_adapter *pAd;
 	struct os_cookie *pObj;
 
 	/*
@@ -812,7 +811,6 @@ dma_addr_t linux_pci_map_single(void *handle, void *ptr, size_t size,
 	   sd_idx = -1
 	 */
 
-	pAd = (struct rt_rtmp_adapter *)handle;
 	pObj = (struct os_cookie *)pAd->OS_Cookie;
 
 	if (sd_idx == 1) {
@@ -826,13 +824,11 @@ dma_addr_t linux_pci_map_single(void *handle, void *ptr, size_t size,
 
 }
 
-void linux_pci_unmap_single(void *handle, dma_addr_t dma_addr, size_t size,
-			    int direction)
+void linux_pci_unmap_single(struct rt_rtmp_adapter *pAd, dma_addr_t dma_addr,
+			    size_t size, int direction)
 {
-	struct rt_rtmp_adapter *pAd;
 	struct os_cookie *pObj;
 
-	pAd = (struct rt_rtmp_adapter *)handle;
 	pObj = (struct os_cookie *)pAd->OS_Cookie;
 
 	pci_unmap_single(pObj->pci_dev, dma_addr, size, direction);
Signature · 6 lines
1.7.0.3


-- 
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.
#3 lore raw

Re: [PATCH] rt28xx: Make PCI_{MAP,UNMAP}_SINGLE type-safe

Quoted 11 lines · On Mon, 29 Mar 2010 01:24:45 +0100 Ben Hutchings <[email protected]> wrote:

To avoid recurrence of bugs such as <http://bugs.debian.org/575726>,
change the type of the first parameter to linux_pci_{map,unmap}_single() from void * to struct rt_rtmp_adapter *. Also do not define the macros PCI_{MAP,UNMAP}_SINGLE() when building the rt2870sta driver; they are not used and if they were that would be a bug.

---
 drivers/staging/rt2860/rt_linux.h    |   14 +++++---------
 drivers/staging/rt2860/rt_pci_rbus.c |   12 ++++--------
 2 files changed, 9 insertions(+), 17 deletions(-)

I think using dma_map_single directly instead of inventing a wrapper function make the code more readable.

#4 lore raw

Re: [PATCH] rt28xx: Make PCI_{MAP,UNMAP}_SINGLE type-safe

Quoted 9 lines · FUJITA Tomonori
Quoted 6 lines · On Mon, 29 Mar 2010 01:24:45 +0100 Ben Hutchings <[email protected]> wrote:

To avoid recurrence of bugs such as <http://bugs.debian.org/575726>,
change the type of the first parameter to linux_pci_{map,unmap}_single() from void * to struct rt_rtmp_adapter *. Also do not define the macros PCI_{MAP,UNMAP}_SINGLE() when building the rt2870sta driver; they are not used and if they were that would be a bug.

Quoted 8 lines
Quoted 5 lines
---
 drivers/staging/rt2860/rt_linux.h    |   14 +++++---------
 drivers/staging/rt2860/rt_pci_rbus.c |   12 ++++--------
 2 files changed, 9 insertions(+), 17 deletions(-)

I think using dma_map_single directly instead of inventing a wrapper function make the code more readable.

These functions are not quite simple wrappers, unfortunately. So while I think that is worth doing it is a separate change.

Ben.

Signature · 2 lines
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.
#5 lore raw

Re: [PATCH] rt28xx: Make PCI_{MAP,UNMAP}_SINGLE type-safe

Quoted 16 lines · On Mon, 29 Mar 2010 02:58:48 +0100 Ben Hutchings <[email protected]> wrote:
Quoted 15 lines · FUJITA Tomonori
Quoted 11 lines · On Mon, 29 Mar 2010 01:24:45 +0100 Ben Hutchings <[email protected]> wrote:

To avoid recurrence of bugs such as <http://bugs.debian.org/575726>,
change the type of the first parameter to linux_pci_{map,unmap}_single() from void * to struct rt_rtmp_adapter *. Also do not define the macros PCI_{MAP,UNMAP}_SINGLE() when building the rt2870sta driver; they are not used and if they were that would be a bug.

---
 drivers/staging/rt2860/rt_linux.h    |   14 +++++---------
 drivers/staging/rt2860/rt_pci_rbus.c |   12 ++++--------
 2 files changed, 9 insertions(+), 17 deletions(-)
Quoted 5 lines
Quoted 2 lines

I think using dma_map_single directly instead of inventing a wrapper function make the code more readable.

These functions are not quite simple wrappers, unfortunately. So while I think that is worth doing it is a separate change.

Well, the current wrapper functions looks terrible. Needs to fix them before moving the driver out of the staging anyway, I think.

#6 lore raw

Re: [PATCH] rt28xx: Make PCI_{MAP,UNMAP}_SINGLE type-safe

Quoted 19 lines · FUJITA Tomonori
Quoted 16 lines · On Mon, 29 Mar 2010 02:58:48 +0100 Ben Hutchings <[email protected]> wrote:
Quoted 15 lines · FUJITA Tomonori
Quoted 11 lines · On Mon, 29 Mar 2010 01:24:45 +0100 Ben Hutchings <[email protected]> wrote:

To avoid recurrence of bugs such as <http://bugs.debian.org/575726>,
change the type of the first parameter to linux_pci_{map,unmap}_single() from void * to struct rt_rtmp_adapter *.  Also do not define the macros PCI_{MAP,UNMAP}_SINGLE() when building the rt2870sta driver; they are not used and if they were that would be a bug.

---
 drivers/staging/rt2860/rt_linux.h    |   14 +++++--------- drivers/staging/rt2860/rt_pci_rbus.c |   12 ++++--------

 2 files changed, 9 insertions(+), 17 deletions(-)
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

I think using dma_map_single directly instead of inventing a wrapper function make the code more readable.

These functions are not quite simple wrappers, unfortunately.  So while I think that is worth doing it is a separate change.

Well, the current wrapper functions looks terrible. Needs to fix them before moving the driver out of the staging anyway, I think.

The driver is not expected to make it ever out of staging ... unless someone ports it to mac80211 ... the in kernel rt2x00 drivers are the longer term solutions for supporting this hardware.

#7 lore raw

Re: [PATCH] rt28xx: Make PCI_{MAP,UNMAP}_SINGLE type-safe

In reply to Ben Hutchings
Quoted 7 lines · Ben Hutchings

To avoid recurrence of bugs such as <http://bugs.debian.org/575726>,
change the type of the first parameter to linux_pci_{map,unmap}_single() from void * to struct rt_rtmp_adapter *. Also do not define the macros PCI_{MAP,UNMAP}_SINGLE() when building the rt2870sta driver; they are not used and if they were that would be a bug.

In the future, please do not sign your emails, it just makes it harder to apply them :(

thanks,

greg k-h

#8 lore raw

patch rt2860sta-fix-argument-to-linux_pci_unmap_single.patch added to 2.6.32-stable tree

In reply to Ben Hutchings

This is a note to let you know that we have just queued up the patch titled

Subject: rt2860sta: Fix argument to linux_pci_unmap_single()

to the 2.6.32-stable tree. Its filename is

rt2860sta-fix-argument-to-linux_pci_unmap_single.patch

A git repo of this tree can be found at
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

Quoted 1 line

From [email protected] Mon Mar 29 13:35:12 2010

From: Ben Hutchings <[email protected]>
Date: Mon, 29 Mar 2010 01:09:17 +0100
Subject: rt2860sta: Fix argument to linux_pci_unmap_single()
To: [email protected]
Cc: [email protected], linux-wireless <[email protected]>, John Halton <[email protected]>, Bartlomiej Zolnierkiewicz <[email protected]>
Message-ID: <1269821357.8653.231.camel@localhost>

From: Ben Hutchings <[email protected]>

John Halton wrote in <http://bugs.debian.org/575726>:

Quoted 3 lines

Whenever wpa_supplicant is deactivated (whether by killing the process or during a normal shutdown) I am getting a kerneloops that prevents the computer from completing shutdown. Here is the relevant syslog output:

The backtrace points to an incorrect call from RTMPFreeTxRxRingMemory() into linux_pci_unmap_single(). This appears to have been fixed in Linux 2.6.33 by this change:

commit ca97b8388838ee9ea4b4bad04948f8f7f8a607a3

Author: Bartlomiej Zolnierkiewicz <[email protected]>
Date:   Tue Sep 22 20:44:07 2009 +0200

Staging: rt28x0: updates from vendor's V2.1.0.0 drivers

For stable-2.6.32, just fix this one function call.

---
 drivers/staging/rt2860/common/2860_rtmp_init.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/staging/rt2860/common/2860_rtmp_init.c +1 -1
@@ -716,7 +716,7 @@ VOID RTMPFreeTxRxRingMemory(
 	{
 		if ((pAd->RxRing.Cell[index].DmaBuf.AllocVa) && (pAd->RxRing.Cell[index].pNdisPacket))
 		{
-			PCI_UNMAP_SINGLE(pObj->pci_dev, pAd->RxRing.Cell[index].DmaBuf.AllocPa, pAd->RxRing.Cell[index].DmaBuf.AllocSize, PCI_DMA_FROMDEVICE);
+			PCI_UNMAP_SINGLE(pAd, pAd->RxRing.Cell[index].DmaBuf.AllocPa, pAd->RxRing.Cell[index].DmaBuf.AllocSize, PCI_DMA_FROMDEVICE);
 			RELEASE_NDIS_PACKET(pAd, pAd->RxRing.Cell[index].pNdisPacket, NDIS_STATUS_SUCCESS);
 		}
 	}

Patches currently in stable-queue which might be from [email protected] are