[PATCH 1/1] [net/irda]: Fix a misalign access issue.

[PATCH 1/1] [net/irda]: Fix a misalign access issue.

[Bryan Wu]

From: Graf Yang <graf.yang@analog.com>

Signed-off-by: Graf Yang <graf.yang@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
---
 net/irda/discovery.c |    8 +++++---
 net/irda/irlmp.c     |    2 +-
 2 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/net/irda/discovery.c b/net/irda/discovery.c
index bfacef8..a6f99b5 100644
--- a/net/irda/discovery.c
+++ b/net/irda/discovery.c
@@ -40,6 +40,8 @@
 
 #include <net/irda/discovery.h>
 
+#include <asm/unaligned.h>
+
 /*
  * Function irlmp_add_discovery (cachelog, discovery)
  *
@@ -87,7 +89,7 @@ void irlmp_add_discovery(hashbin_t *cachelog, discovery_t *new)
 			 */
 			hashbin_remove_this(cachelog, (irda_queue_t *) node);
 			/* Check if hints bits are unchanged */
-			if(u16ho(node->data.hints) == u16ho(new->data.hints))
+			if (get_unaligned((__u16 *)node->data.hints) == get_unaligned((__u16 *)new->data.hints))
 				/* Set time of first discovery for this node */
 				new->firststamp = node->firststamp;
 			kfree(node);
@@ -281,9 +283,9 @@ struct irda_device_info *irlmp_copy_discoveries(hashbin_t *log, int *pn,
 		/* Mask out the ones we don't want :
 		 * We want to match the discovery mask, and to get only
 		 * the most recent one (unless we want old ones) */
-		if ((u16ho(discovery->data.hints) & mask) &&
+		if ((get_unaligned((__u16 *)discovery->data.hints) & mask) &&
 		    ((old_entries) ||
-		     ((jiffies - discovery->firststamp) < j_timeout)) ) {
+		     ((jiffies - discovery->firststamp) < j_timeout))) {
 			/* Create buffer as needed.
 			 * As this function get called a lot and most time
 			 * we don't have anything to put in the log (we are
diff --git a/net/irda/irlmp.c b/net/irda/irlmp.c
index 1f81f8e..7b95cd6 100644
--- a/net/irda/irlmp.c
+++ b/net/irda/irlmp.c
@@ -1086,7 +1086,7 @@ discovery_t *irlmp_get_discovery_response(void)
 
 	IRDA_ASSERT(irlmp != NULL, return NULL;);
 
-	u16ho(irlmp->discovery_rsp.data.hints) = irlmp->hints.word;
+	put_unaligned(irlmp->hints.word, (__u16 *)irlmp->discovery_rsp.data.hints);
 
 	/*
 	 *  Set character set for device name (we use ASCII), and
-- 
1.5.5

Re: [PATCH 1/1] [net/irda]: Fix a misalign access issue.

[Marcel Holtmann]

Hi Bryan,

> From: Graf Yang <graf.yang@analog.com>
> 
> Signed-off-by: Graf Yang <graf.yang@analog.com>
> Signed-off-by: Bryan Wu <cooloney@kernel.org>
> ---
>  net/irda/discovery.c |    8 +++++---
>  net/irda/irlmp.c     |    2 +-
>  2 files changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/net/irda/discovery.c b/net/irda/discovery.c
> index bfacef8..a6f99b5 100644
> --- a/net/irda/discovery.c
> +++ b/net/irda/discovery.c
> @@ -40,6 +40,8 @@
>  
>  #include <net/irda/discovery.h>
>  
> +#include <asm/unaligned.h>
> +
>  /*
>   * Function irlmp_add_discovery (cachelog, discovery)
>   *
> @@ -87,7 +89,7 @@ void irlmp_add_discovery(hashbin_t *cachelog, discovery_t *new)
>  			 */
>  			hashbin_remove_this(cachelog, (irda_queue_t *) node);
>  			/* Check if hints bits are unchanged */
> -			if(u16ho(node->data.hints) == u16ho(new->data.hints))
> +			if (get_unaligned((__u16 *)node->data.hints) == get_unaligned((__u16 *)new->data.hints))
>  				/* Set time of first discovery for this node */
>  				new->firststamp = node->firststamp;
>  			kfree(node);
> @@ -281,9 +283,9 @@ struct irda_device_info *irlmp_copy_discoveries(hashbin_t *log, int *pn,
>  		/* Mask out the ones we don't want :
>  		 * We want to match the discovery mask, and to get only
>  		 * the most recent one (unless we want old ones) */
> -		if ((u16ho(discovery->data.hints) & mask) &&
> +		if ((get_unaligned((__u16 *)discovery->data.hints) & mask) &&
>  		    ((old_entries) ||
> -		     ((jiffies - discovery->firststamp) < j_timeout)) ) {
> +		     ((jiffies - discovery->firststamp) < j_timeout))) {
>  			/* Create buffer as needed.
>  			 * As this function get called a lot and most time
>  			 * we don't have anything to put in the log (we are
> diff --git a/net/irda/irlmp.c b/net/irda/irlmp.c
> index 1f81f8e..7b95cd6 100644
> --- a/net/irda/irlmp.c
> +++ b/net/irda/irlmp.c
> @@ -1086,7 +1086,7 @@ discovery_t *irlmp_get_discovery_response(void)
>  
>  	IRDA_ASSERT(irlmp != NULL, return NULL;);
>  
> -	u16ho(irlmp->discovery_rsp.data.hints) = irlmp->hints.word;
> +	put_unaligned(irlmp->hints.word, (__u16 *)irlmp->discovery_rsp.data.hints);
>  
>  	/*
>  	 *  Set character set for device name (we use ASCII), and

what about net/irda/irnet/irnet_irda.c and why not also kill the u16ho
macro at the same time.

Regards

Marcel

RE: [PATCH 1/1] [net/irda]: Fix a misalign access issue.

[Yang, Graf]

Hi Marcel,

The root cause is __u8 hints[2] are unaligned because of __u8 charset defined in file include/linux/irda.h:132
struct irda_device_info {
        __u32       saddr;    /* Address of local interface */
        __u32       daddr;    /* Address of remote device */
        char        info[22]; /* Description */
        __u8        charset;  /* Charset used for description */
        __u8        hints[2]; /* Hint bits */
};
And code may access hints[2] as u16. 

As put_unaligned/get_unaligned would set/get value through several instructions which decreases the performance. So I'd like to modify the irda_device_info instead of using put_unaligned/get_unaligned in irlmp.c/discovery.c.

I have tried
struct irda_device_info {
        __u32       saddr;    /* Address of local interface */
        __u32       daddr;    /* Address of remote device */
        char        info[22]; /* Description */
        __u8        charset;  /* Charset used for description */
        __u8        reserve;  /* For align data */ 
        __u8        hints[2]; /* Hint bits */
};

It works with my irda driver. Can I modify it like this?
There are many methods to modify this structure, which method are better?


It seems u16ho is useless, I would remove them from net/irda.

I can't enable irnet, so didn't touch net/irda/irnet/irnet_irda.c.
Would you give some information about how to enable irnet?

BRs,
Graf

-----Original Message-----
From: Marcel Holtmann [mailto:marcel@holtmann.org]
Sent: Mon 5/12/2008 9:15 PM
To: Bryan Wu
Cc: samuel@sortiz.org; irda-users@lists.sourceforge.net; linux-kernel@vger.kernel.org; Graf Yang
Subject: Re: [PATCH 1/1] [net/irda]: Fix a misalign access issue.
 
Hi Bryan,

> From: Graf Yang <graf.yang@analog.com>
> 
> Signed-off-by: Graf Yang <graf.yang@analog.com>
> Signed-off-by: Bryan Wu <cooloney@kernel.org>
> ---
>  net/irda/discovery.c |    8 +++++---
>  net/irda/irlmp.c     |    2 +-
>  2 files changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/net/irda/discovery.c b/net/irda/discovery.c
> index bfacef8..a6f99b5 100644
> --- a/net/irda/discovery.c
> +++ b/net/irda/discovery.c
> @@ -40,6 +40,8 @@
>  
>  #include <net/irda/discovery.h>
>  
> +#include <asm/unaligned.h>
> +
>  /*
>   * Function irlmp_add_discovery (cachelog, discovery)
>   *
> @@ -87,7 +89,7 @@ void irlmp_add_discovery(hashbin_t *cachelog, discovery_t *new)
>  			 */
>  			hashbin_remove_this(cachelog, (irda_queue_t *) node);
>  			/* Check if hints bits are unchanged */
> -			if(u16ho(node->data.hints) == u16ho(new->data.hints))
> +			if (get_unaligned((__u16 *)node->data.hints) == get_unaligned((__u16 *)new->data.hints))
>  				/* Set time of first discovery for this node */
>  				new->firststamp = node->firststamp;
>  			kfree(node);
> @@ -281,9 +283,9 @@ struct irda_device_info *irlmp_copy_discoveries(hashbin_t *log, int *pn,
>  		/* Mask out the ones we don't want :
>  		 * We want to match the discovery mask, and to get only
>  		 * the most recent one (unless we want old ones) */
> -		if ((u16ho(discovery->data.hints) & mask) &&
> +		if ((get_unaligned((__u16 *)discovery->data.hints) & mask) &&
>  		    ((old_entries) ||
> -		     ((jiffies - discovery->firststamp) < j_timeout)) ) {
> +		     ((jiffies - discovery->firststamp) < j_timeout))) {
>  			/* Create buffer as needed.
>  			 * As this function get called a lot and most time
>  			 * we don't have anything to put in the log (we are
> diff --git a/net/irda/irlmp.c b/net/irda/irlmp.c
> index 1f81f8e..7b95cd6 100644
> --- a/net/irda/irlmp.c
> +++ b/net/irda/irlmp.c
> @@ -1086,7 +1086,7 @@ discovery_t *irlmp_get_discovery_response(void)
>  
>  	IRDA_ASSERT(irlmp != NULL, return NULL;);
>  
> -	u16ho(irlmp->discovery_rsp.data.hints) = irlmp->hints.word;
> +	put_unaligned(irlmp->hints.word, (__u16 *)irlmp->discovery_rsp.data.hints);
>  
>  	/*
>  	 *  Set character set for device name (we use ASCII), and

what about net/irda/irnet/irnet_irda.c and why not also kill the u16ho
macro at the same time.

Regards

Marcel

Re: [PATCH 1/1] [net/irda]: Fix a misalign access issue.

[David Miller]

From: "Yang, Graf" <Graf.Yang@analog.com>
Date: Tue, 13 May 2008 12:07:35 +0800

> I have tried
> struct irda_device_info {
>         __u32       saddr;    /* Address of local interface */
>         __u32       daddr;    /* Address of remote device */
>         char        info[22]; /* Description */
>         __u8        charset;  /* Charset used for description */
>         __u8        reserve;  /* For align data */ 
>         __u8        hints[2]; /* Hint bits */
> };
> 
> It works with my irda driver. Can I modify it like this?

This header file and the irda_device_info datastructure are exported
to, and used by, userspace.

Therefore you cannot change this structure's layout without breaking
every single existing IRDA binary out there.

Re: [PATCH 1/1] [net/irda]: Fix a misalign access issue.

[Bryan Wu]

On Tue, May 13, 2008 at 12:14 PM, David Miller <davem@davemloft.net> wrote:
> From: "Yang, Graf" <Graf.Yang@analog.com>
>  Date: Tue, 13 May 2008 12:07:35 +0800
>
>
>  > I have tried
>  > struct irda_device_info {
>  >         __u32       saddr;    /* Address of local interface */
>  >         __u32       daddr;    /* Address of remote device */
>  >         char        info[22]; /* Description */
>  >         __u8        charset;  /* Charset used for description */
>  >         __u8        reserve;  /* For align data */
>  >         __u8        hints[2]; /* Hint bits */
>  > };
>  >
>  > It works with my irda driver. Can I modify it like this?
>
>  This header file and the irda_device_info datastructure are exported
>  to, and used by, userspace.
>
>  Therefore you cannot change this structure's layout without breaking
>  every single existing IRDA binary out there.
>

Exactly, so we plan to add put/get_unaligned method to fix this issue.
After Graf fix it, I will send out the patch

Thanks
-Bryan