Skip to contentSkip to conversation

avoid unaligned access to data

PATCH In review Olaf Hering 3 participants 9 messages Feb 29, 2004 – Mar 13, 2004

lore
Share & tools
lore

Olaf Hering

author
#0 lore raw

PATCH 1/1 avoid unaligned access to data

the data access in udevinfo is usually unaligned, due to the fact that the struct is at offset strlen(sysfspath). That makes an ia64 kernel rather unhappy. Unless I miss something, this change makes it happy by making an incompatible change to the database.

udevdb.c +17 -0
diff -p -purN udev-018/udevdb.c udev-018.align/udevdb.c
@@ -44,6 +44,8 @@
 static TDB_CONTEXT *udevdb;
 
 
+#define	my_key_align (__alignof__(char *))
+
 int udevdb_add_dev(const char *path, const struct udevice *dev)
 {
 	TDB_DATA key, data;

@@ -57,6 +59,11 @@ int udevdb_add_dev(const char *path, con
 	key.dptr = keystr;
 	key.dsize = strlen(keystr) + 1;
 
+	if (key.dsize & (my_key_align - 1))
+		key.dsize = (key.dsize & ~(my_key_align - 1)) + my_key_align;
+	if (key.dsize > SYSFS_PATH_MAX)
+		key.dsize = SYSFS_PATH_MAX;
+
 	data.dptr = (void *)dev;
 	data.dsize = UDEVICE_LEN;
 

@@ -73,6 +80,11 @@ int udevdb_get_dev(const char *path, str
 	key.dptr = (void *)path;
 	key.dsize = strlen(path) + 1;
 
+	if (key.dsize & (my_key_align - 1))
+		key.dsize = (key.dsize & ~(my_key_align - 1)) + my_key_align;
+	if (key.dsize > SYSFS_PATH_MAX)
+		key.dsize = SYSFS_PATH_MAX;
+
 	data = tdb_fetch(udevdb, key);
 	if (data.dptr = NULL || data.dsize = 0)
 		return -ENODEV;

@@ -96,6 +108,11 @@ int udevdb_delete_dev(const char *path)
 	key.dptr = keystr;
 	key.dsize = strlen(keystr) + 1;
 
+	if (key.dsize & (my_key_align - 1))
+		key.dsize = (key.dsize & ~(my_key_align - 1)) + my_key_align;
+	if (key.dsize > SYSFS_PATH_MAX)
+		key.dsize = SYSFS_PATH_MAX;
+
 	return tdb_delete(udevdb, key);
 }
Signature · 14 lines
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id56&alloc_id438&op=click
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
author
#1 lore raw
Quoted 6 lines · Olaf Hering

the data access in udevinfo is usually unaligned, due to the fact that the struct is at offset strlen(sysfspath). That makes an ia64 kernel rather unhappy. Unless I miss something, this change makes it happy by making an incompatible change to the database.

What is the plan here, Greg?

Signature · 14 lines
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id70&alloc_id638&op�k
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
#2 lore raw
In reply to Olaf Hering
Quoted 10 lines · Olaf Hering
Quoted 6 lines · Olaf Hering

the data access in udevinfo is usually unaligned, due to the fact that the struct is at offset strlen(sysfspath). That makes an ia64 kernel rather unhappy. Unless I miss something, this change makes it happy by making an incompatible change to the database.

What is the plan here, Greg?

Greg didn't see this patch :(

Next time please cc patches to me so I make sure I don't miss them. I tend to get a _lot_ of email...

Anyway, what is the error that happens on a ia64 box?

thanks,

greg k-h

-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id70&alloc_id638&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#3 lore raw
In reply to Olaf Hering
Quoted 11 lines · Olaf Hering

the data access in udevinfo is usually unaligned, due to the fact that the struct is at offset strlen(sysfspath). That makes an ia64 kernel rather unhappy. Unless I miss something, this change makes it happy by making an incompatible change to the database.

diff -p -purN udev-018/udevdb.c udev-018.align/udevdb.c

udevdb.c line 47

Quoted 4 lines
udevdb.c +1 -0
@@ -44,6 +44,8 @@
 static TDB_CONTEXT *udevdb;
 
 
+#define	my_key_align (__alignof__(char *))

Hm, can't we just pad out our data structures to provide the proper alignment? If it's really necessary...

thanks,

greg k-h

-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id70&alloc_id638&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

author
#4 lore raw
In reply to Olaf Hering
Quoted 10 lines · Greg KH
Quoted 9 lines · Olaf Hering
Quoted 6 lines · Olaf Hering

the data access in udevinfo is usually unaligned, due to the fact that the struct is at offset strlen(sysfspath). That makes an ia64 kernel rather unhappy. Unless I miss something, this change makes it happy by making an incompatible change to the database.

Quoted 8 lines
Quoted 1 line

What is the plan here, Greg?

Greg didn't see this patch :(

Next time please cc patches to me so I make sure I don't miss them. I tend to get a _lot_ of email...

Anyway, what is the error that happens on a ia64 box?

it complains about unaligned memory access.

udevinfo(4922): unaligned access to 0x6000000000015747, ip=0x4000000000001dd0

Signature · 14 lines
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id70&alloc_id638&op�k
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
author
#5 lore raw
In reply to Olaf Hering
Quoted 15 lines · Greg KH
Quoted 14 lines · Olaf Hering

the data access in udevinfo is usually unaligned, due to the fact that the struct is at offset strlen(sysfspath). That makes an ia64 kernel rather unhappy. Unless I miss something, this change makes it happy by making an incompatible change to the database.

diff -p -purN udev-018/udevdb.c udev-018.align/udevdb.c

udevdb.c +0 -0
@@ -44,6 +44,8 @@
 static TDB_CONTEXT *udevdb;
Quoted 4 lines

udevdb.c line 47

Quoted 1 line
udevdb.c +1 -0
+#define	my_key_align (__alignof__(char *))

Hm, can't we just pad out our data structures to provide the proper alignment? If it's really necessary...

its not the alignment of struct members, but the alignment of the struct itself. we could copy the struct from the database to an aligned area on the stack or malloc buffer. but thats not faster than the unaligned access.

Signature · 14 lines
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id70&alloc_id638&op�k
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
#6 lore raw
In reply to Olaf Hering
Quoted 16 lines · Olaf Hering
Quoted 14 lines · Greg KH
Quoted 10 lines · Olaf Hering
Quoted 6 lines · Olaf Hering

the data access in udevinfo is usually unaligned, due to the fact that the struct is at offset strlen(sysfspath). That makes an ia64 kernel rather unhappy. Unless I miss something, this change makes it happy by making an incompatible change to the database.

What is the plan here, Greg?

Greg didn't see this patch :(

Quoted 8 lines
Quoted 4 lines

Next time please cc patches to me so I make sure I don't miss them. I tend to get a _lot_ of email...

Anyway, what is the error that happens on a ia64 box?

it complains about unaligned memory access.

udevinfo(4922): unaligned access to 0x6000000000015747, ip=0x4000000000001dd0

So all userspace programs have to be modified to prevent this from happening?

Does it cause a segfault, or just a warning in the syslog?

Seems like a pretty stupid architecture to be complaining of this :)

thanks,

greg k-h

-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id70&alloc_id638&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

author
#7 lore raw
In reply to Olaf Hering
Quoted 1 line · Greg KH

Seems like a pretty stupid architecture to be complaining of this :)

just look at the vendors name..
it results in a warning in dmesg.

Signature · 14 lines
USB is for mice, FireWire is for men!

sUse lINUX ag, nÜRNBERG


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id70&alloc_id638&op�k
_______________________________________________
Linux-hotplug-devel mailing list  http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
#8 lore raw
In reply to Olaf Hering
Quoted 10 lines · Greg KH
Quoted 3 lines · Olaf Hering

it complains about unaligned memory access.

udevinfo(4922): unaligned access to 0x6000000000015747, ip=0x4000000000001dd0

So all userspace programs have to be modified to prevent this from happening?

Does it cause a segfault, or just a warning in the syslog?

Seems like a pretty stupid architecture to be complaining of this :)

Most architectures don't allow unaligned accesses. x86 is one of the few that does allow it.

Linux ia64 will perform unaligned accesses in software for userspace programs when the trap occurs, but will also print a warning.

JE

-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id70&alloc_id638&op=click
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel