UDF filesystem has some bugs on truncating
Hi all,
I found that UDF has bugs on truncating.
When you do this:
dd if=/dev/zero of=aaa bs=1024k count=2 seek=3000
, Linux will hang and die.
The platform is Linux 2.6.16 on MIPS malta board.
Regards,
Colin
Ar Gwe, 2006-06-16 am 13:47 +0800, ysgrifennodd colin:
Quoted 5 lines
Hi all,
I found that UDF has bugs on truncating.
When you do this:dd if=/dev/zero of=aaa bs=1024k count=2 seek=3000, Linux will hang and die.
I'm unable to duplicate this on x86-64, do you have more information, eg backtraces ?
Ar Gwe, 2006-06-16 am 13:47 +0800, ysgrifennodd colin:
Quoted 6 lines
Hi all,
I found that UDF has bugs on truncating.
When you do this:dd if=/dev/zero of=aaa bs=1024k count=2 seek=3000, Linux will hang and die.
The platform is Linux 2.6.16 on MIPS malta board.
Ok I eventually sort of reproduced this on x86-64. It took a while because in my environment I see a crash 2 or 3 hours after the test is run, and that crash is on hardware that doesn't otherwise crash and seems to be repeatable.
Alan
Hi Alan,
Do you use the same command with me?
The parameter "seek" of "dd if=/dev/zero of=aaa bs=1024k count=2 seek=3000" will inform "dd" to use truncate system call to expand file in no time. In normal situations, the file will grow up to 3G in about 2 seconds. This dd command on ext3 partition can run very well on my platform, but it will fail on UDF partition.
Regards,
Colin
----- Original Message -----
From: "Alan Cox" <[email protected]>
To: "colin" <[email protected]>
Cc: <[email protected]>
Sent: Monday, June 19, 2006 5:32 PM
Subject: Re: UDF filesystem has some bugs on truncating
Quoted 15 lines
Ar Gwe, 2006-06-16 am 13:47 +0800, ysgrifennodd colin:
Quoted 6 lines
Hi all,
I found that UDF has bugs on truncating.
When you do this:dd if=/dev/zero of=aaa bs=1024k count=2 seek=3000, Linux will hang and die.
The platform is Linux 2.6.16 on MIPS malta board.Ok I eventually sort of reproduced this on x86-64. It took a while because in my environment I see a crash 2 or 3 hours after the test is run, and that crash is on hardware that doesn't otherwise crash and seems to be repeatable.
Alan