Re: fdatasync() error in shred from coreutils-5.2.1 on AIX 5.2

[Albert Chin-A-Young]

On Thu, May 20, 2004 at 11:24:38PM -0700, Paul Eggert wrote:
> Albert Chin <address@hidden> writes:
> 
> >   245772: kwrite(3, "XXXXXXXXXXXXXXXXXXXXXXXX".., 12288)  = 12288
> 
> It appears that kwrite is not failing, even though we must be writing
> past the end of the disk at this point.  This sounds like it could be
> a bug in the /dev/lv00 drivers.
> 
> shred should try "lseek (fd, 0, SEEK_END)": did that fail too?
> It wasn't in your truss extract.

The only lseek I see in the truss output is the following. It appears
very early.
  245772: open("/dev/lv00", 0400004001)                   = 3
  245772: fstatx(3, 0x2FF21F98, 128, 010)                 = 0
  245772: klseek(3, 0, 0, 0x00000002)                     = 0
  245772: klseek(3, 0, 0, 0x00000000)                     = 0

> Do you get the same failure when you write to a raw device?

Yes:
  gshred: /dev/hdisk1: pass 1/3 (random)...1004GiB
  gshred: /dev/hdisk1: pass 1/3 (random)...1005GiB
  gshred: /dev/hdisk1: pass 1/3 (random)...1006GiB
  ...
  gshred: /dev/hdisk1: pass 1/3 (random)...1022GiB
  gshred: /dev/hdisk1: pass 1/3 (random)...1023GiB
  gshred: /dev/hdisk1: pass 1/3 (random)...1.0TiB
  error writing at offset 1099511635968: A system call received a
  parameter that is not valid.

> Is there some approved-by-AIX method to find out the size of a device
> like /dev/lv00?

Probably. I'll try to figure it out.

-- 
albert chin (address@hidden)