----- Original Message -----

From: Hillf Danton <redacted>
To: Victor Meyerson <redacted>
Cc: "linux-mips@linux-mips.org" <redacted>; Ralf Baechle <redacted>; LKML <redacted>
Sent: Friday, July 27, 2012 7:47 AM
Subject: Re: Direct I/O bug in kernel

On Wed, Jul 25, 2012 at 1:28 AM, Victor Meyerson
[off-list ref] wrote:

quoted

 Still different checksums and I used the same random-file from my first 

test.

quoted

Then try the fix at
            https://lkml.org/lkml/2012/7/27/54

I tried that patch, although I had to edit a slightly different line as dio_bio_alloc was near line 392 instead of 349 in the version of fs/direct-io.c in my tree.  I still got different checksums between the two files and even different checksums from my earlier attempts.

I am not sure if this helps, but Ralf asked if I can try a different page size to see if this problem occurs.  I originally had CONFIG_PAGE_SIZE_4KB=y and changed it to CONFIG_PAGE_SIZE_16KB=y (via menuconfig).  Having a page size of 16KB (and the above patch not applied) made the checksum on the files match each other and match the file made from the working kernel.

Victor