Re: Are there still machines with PAGE_SHIFT 16 or larger ?

["Thomas Schmitt" <scdbackup@gmx.net>]

Hi,

Anatoly Pugachev wrote:
> mator@chubaka:~$ getconf PAGESIZE
> 65536
> mator@chubaka:~$ uname -a
> Linux chubaka 5.9.0-1-powerpc64 #1 SMP Debian 5.9.1-1 (2020-10-17)
> ppc64 GNU/Linux
> [...]
> mator@chubaka:~$ md5sum /mnt/1mb
> md5sum: /mnt/1mb: Input/output error

Does it deliver any bytes before erroring out ?

  dd if=/mnt/1mb bs=1 | wc -c

Just to be sure, can you copy the ISO to a Linux machine with <= 32 KiB
page size and verify that the file "1mb" is readable without i/o error
and yields the due checksum ?
(I am not aware how xorriso could have spoiled it, but ...)

The kernel log shows no hint what's wrong. strace cannot, because the
problem happens under mm/readahead in zisofs_fill_pages() deep inside the
kernel.
At least i know where to look for code flaws. I wonder whether this ever
was tested in the last 19 years.

Thanks a lot for doing it now.


Have a nice day :)

Thomas